xref: /openbmc/linux/include/linux/netdevice.h (revision 98a52692)
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, <waltje@uWalt.NL.Mugnet.ORG>
13  *		Corey Minyard <wf-rch!minyard@relay.EU.net>
14  *		Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15  *		Alan Cox, <alan@lxorguk.ukuu.org.uk>
16  *		Bjorn Ekwall. <bj0rn@blox.se>
17  *              Pekka Riikonen <priikone@poseidon.pspt.fi>
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 
83 typedef u32 xdp_features_t;
84 
85 void synchronize_net(void);
86 void netdev_set_default_ethtool_ops(struct net_device *dev,
87 				    const struct ethtool_ops *ops);
88 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
89 
90 /* Backlog congestion levels */
91 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
92 #define NET_RX_DROP		1	/* packet dropped */
93 
94 #define MAX_NEST_DEV 8
95 
96 /*
97  * Transmit return codes: transmit return codes originate from three different
98  * namespaces:
99  *
100  * - qdisc return codes
101  * - driver transmit return codes
102  * - errno values
103  *
104  * Drivers are allowed to return any one of those in their hard_start_xmit()
105  * function. Real network devices commonly used with qdiscs should only return
106  * the driver transmit return codes though - when qdiscs are used, the actual
107  * transmission happens asynchronously, so the value is not propagated to
108  * higher layers. Virtual network devices transmit synchronously; in this case
109  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
110  * others are propagated to higher layers.
111  */
112 
113 /* qdisc ->enqueue() return codes. */
114 #define NET_XMIT_SUCCESS	0x00
115 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
116 #define NET_XMIT_CN		0x02	/* congestion notification	*/
117 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
118 
119 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
120  * indicates that the device will soon be dropping packets, or already drops
121  * some packets of the same priority; prompting us to send less aggressively. */
122 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
123 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
124 
125 /* Driver transmit return codes */
126 #define NETDEV_TX_MASK		0xf0
127 
128 enum netdev_tx {
129 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
130 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
131 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
132 };
133 typedef enum netdev_tx netdev_tx_t;
134 
135 /*
136  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
137  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
138  */
139 static inline bool dev_xmit_complete(int rc)
140 {
141 	/*
142 	 * Positive cases with an skb consumed by a driver:
143 	 * - successful transmission (rc == NETDEV_TX_OK)
144 	 * - error while transmitting (rc < 0)
145 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
146 	 */
147 	if (likely(rc < NET_XMIT_MASK))
148 		return true;
149 
150 	return false;
151 }
152 
153 /*
154  *	Compute the worst-case header length according to the protocols
155  *	used.
156  */
157 
158 #if defined(CONFIG_HYPERV_NET)
159 # define LL_MAX_HEADER 128
160 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
161 # if defined(CONFIG_MAC80211_MESH)
162 #  define LL_MAX_HEADER 128
163 # else
164 #  define LL_MAX_HEADER 96
165 # endif
166 #else
167 # define LL_MAX_HEADER 32
168 #endif
169 
170 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
171     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
172 #define MAX_HEADER LL_MAX_HEADER
173 #else
174 #define MAX_HEADER (LL_MAX_HEADER + 48)
175 #endif
176 
177 /*
178  *	Old network device statistics. Fields are native words
179  *	(unsigned long) so they can be read and written atomically.
180  */
181 
182 #define NET_DEV_STAT(FIELD)			\
183 	union {					\
184 		unsigned long FIELD;		\
185 		atomic_long_t __##FIELD;	\
186 	}
187 
188 struct net_device_stats {
189 	NET_DEV_STAT(rx_packets);
190 	NET_DEV_STAT(tx_packets);
191 	NET_DEV_STAT(rx_bytes);
192 	NET_DEV_STAT(tx_bytes);
193 	NET_DEV_STAT(rx_errors);
194 	NET_DEV_STAT(tx_errors);
195 	NET_DEV_STAT(rx_dropped);
196 	NET_DEV_STAT(tx_dropped);
197 	NET_DEV_STAT(multicast);
198 	NET_DEV_STAT(collisions);
199 	NET_DEV_STAT(rx_length_errors);
200 	NET_DEV_STAT(rx_over_errors);
201 	NET_DEV_STAT(rx_crc_errors);
202 	NET_DEV_STAT(rx_frame_errors);
203 	NET_DEV_STAT(rx_fifo_errors);
204 	NET_DEV_STAT(rx_missed_errors);
205 	NET_DEV_STAT(tx_aborted_errors);
206 	NET_DEV_STAT(tx_carrier_errors);
207 	NET_DEV_STAT(tx_fifo_errors);
208 	NET_DEV_STAT(tx_heartbeat_errors);
209 	NET_DEV_STAT(tx_window_errors);
210 	NET_DEV_STAT(rx_compressed);
211 	NET_DEV_STAT(tx_compressed);
212 };
213 #undef NET_DEV_STAT
214 
215 /* per-cpu stats, allocated on demand.
216  * Try to fit them in a single cache line, for dev_get_stats() sake.
217  */
218 struct net_device_core_stats {
219 	unsigned long	rx_dropped;
220 	unsigned long	tx_dropped;
221 	unsigned long	rx_nohandler;
222 	unsigned long	rx_otherhost_dropped;
223 } __aligned(4 * sizeof(unsigned long));
224 
225 #include <linux/cache.h>
226 #include <linux/skbuff.h>
227 
228 #ifdef CONFIG_RPS
229 #include <linux/static_key.h>
230 extern struct static_key_false rps_needed;
231 extern struct static_key_false rfs_needed;
232 #endif
233 
234 struct neighbour;
235 struct neigh_parms;
236 struct sk_buff;
237 
238 struct netdev_hw_addr {
239 	struct list_head	list;
240 	struct rb_node		node;
241 	unsigned char		addr[MAX_ADDR_LEN];
242 	unsigned char		type;
243 #define NETDEV_HW_ADDR_T_LAN		1
244 #define NETDEV_HW_ADDR_T_SAN		2
245 #define NETDEV_HW_ADDR_T_UNICAST	3
246 #define NETDEV_HW_ADDR_T_MULTICAST	4
247 	bool			global_use;
248 	int			sync_cnt;
249 	int			refcount;
250 	int			synced;
251 	struct rcu_head		rcu_head;
252 };
253 
254 struct netdev_hw_addr_list {
255 	struct list_head	list;
256 	int			count;
257 
258 	/* Auxiliary tree for faster lookup on addition and deletion */
259 	struct rb_root		tree;
260 };
261 
262 #define netdev_hw_addr_list_count(l) ((l)->count)
263 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
264 #define netdev_hw_addr_list_for_each(ha, l) \
265 	list_for_each_entry(ha, &(l)->list, list)
266 
267 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
268 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
269 #define netdev_for_each_uc_addr(ha, dev) \
270 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
271 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
272 	netdev_for_each_uc_addr((_ha), (_dev)) \
273 		if ((_ha)->sync_cnt)
274 
275 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
276 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
277 #define netdev_for_each_mc_addr(ha, dev) \
278 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
279 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
280 	netdev_for_each_mc_addr((_ha), (_dev)) \
281 		if ((_ha)->sync_cnt)
282 
283 struct hh_cache {
284 	unsigned int	hh_len;
285 	seqlock_t	hh_lock;
286 
287 	/* cached hardware header; allow for machine alignment needs.        */
288 #define HH_DATA_MOD	16
289 #define HH_DATA_OFF(__len) \
290 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
291 #define HH_DATA_ALIGN(__len) \
292 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
293 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
294 };
295 
296 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
297  * Alternative is:
298  *   dev->hard_header_len ? (dev->hard_header_len +
299  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
300  *
301  * We could use other alignment values, but we must maintain the
302  * relationship HH alignment <= LL alignment.
303  */
304 #define LL_RESERVED_SPACE(dev) \
305 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
306 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
307 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
308 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
309 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
310 
311 struct header_ops {
312 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
313 			   unsigned short type, const void *daddr,
314 			   const void *saddr, unsigned int len);
315 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
316 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
317 	void	(*cache_update)(struct hh_cache *hh,
318 				const struct net_device *dev,
319 				const unsigned char *haddr);
320 	bool	(*validate)(const char *ll_header, unsigned int len);
321 	__be16	(*parse_protocol)(const struct sk_buff *skb);
322 };
323 
324 /* These flag bits are private to the generic network queueing
325  * layer; they may not be explicitly referenced by any other
326  * code.
327  */
328 
329 enum netdev_state_t {
330 	__LINK_STATE_START,
331 	__LINK_STATE_PRESENT,
332 	__LINK_STATE_NOCARRIER,
333 	__LINK_STATE_LINKWATCH_PENDING,
334 	__LINK_STATE_DORMANT,
335 	__LINK_STATE_TESTING,
336 };
337 
338 struct gro_list {
339 	struct list_head	list;
340 	int			count;
341 };
342 
343 /*
344  * size of gro hash buckets, must less than bit number of
345  * napi_struct::gro_bitmask
346  */
347 #define GRO_HASH_BUCKETS	8
348 
349 /*
350  * Structure for NAPI scheduling similar to tasklet but with weighting
351  */
352 struct napi_struct {
353 	/* The poll_list must only be managed by the entity which
354 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
355 	 * whoever atomically sets that bit can add this napi_struct
356 	 * to the per-CPU poll_list, and whoever clears that bit
357 	 * can remove from the list right before clearing the bit.
358 	 */
359 	struct list_head	poll_list;
360 
361 	unsigned long		state;
362 	int			weight;
363 	int			defer_hard_irqs_count;
364 	unsigned long		gro_bitmask;
365 	int			(*poll)(struct napi_struct *, int);
366 #ifdef CONFIG_NETPOLL
367 	/* CPU actively polling if netpoll is configured */
368 	int			poll_owner;
369 #endif
370 	/* CPU on which NAPI has been scheduled for processing */
371 	int			list_owner;
372 	struct net_device	*dev;
373 	struct gro_list		gro_hash[GRO_HASH_BUCKETS];
374 	struct sk_buff		*skb;
375 	struct list_head	rx_list; /* Pending GRO_NORMAL skbs */
376 	int			rx_count; /* length of rx_list */
377 	unsigned int		napi_id;
378 	struct hrtimer		timer;
379 	struct task_struct	*thread;
380 	/* control-path-only fields follow */
381 	struct list_head	dev_list;
382 	struct hlist_node	napi_hash_node;
383 };
384 
385 enum {
386 	NAPI_STATE_SCHED,		/* Poll is scheduled */
387 	NAPI_STATE_MISSED,		/* reschedule a napi */
388 	NAPI_STATE_DISABLE,		/* Disable pending */
389 	NAPI_STATE_NPSVC,		/* Netpoll - don't dequeue from poll_list */
390 	NAPI_STATE_LISTED,		/* NAPI added to system lists */
391 	NAPI_STATE_NO_BUSY_POLL,	/* Do not add in napi_hash, no busy polling */
392 	NAPI_STATE_IN_BUSY_POLL,	/* sk_busy_loop() owns this NAPI */
393 	NAPI_STATE_PREFER_BUSY_POLL,	/* prefer busy-polling over softirq processing*/
394 	NAPI_STATE_THREADED,		/* The poll is performed inside its own thread*/
395 	NAPI_STATE_SCHED_THREADED,	/* Napi is currently scheduled in threaded mode */
396 };
397 
398 enum {
399 	NAPIF_STATE_SCHED		= BIT(NAPI_STATE_SCHED),
400 	NAPIF_STATE_MISSED		= BIT(NAPI_STATE_MISSED),
401 	NAPIF_STATE_DISABLE		= BIT(NAPI_STATE_DISABLE),
402 	NAPIF_STATE_NPSVC		= BIT(NAPI_STATE_NPSVC),
403 	NAPIF_STATE_LISTED		= BIT(NAPI_STATE_LISTED),
404 	NAPIF_STATE_NO_BUSY_POLL	= BIT(NAPI_STATE_NO_BUSY_POLL),
405 	NAPIF_STATE_IN_BUSY_POLL	= BIT(NAPI_STATE_IN_BUSY_POLL),
406 	NAPIF_STATE_PREFER_BUSY_POLL	= BIT(NAPI_STATE_PREFER_BUSY_POLL),
407 	NAPIF_STATE_THREADED		= BIT(NAPI_STATE_THREADED),
408 	NAPIF_STATE_SCHED_THREADED	= BIT(NAPI_STATE_SCHED_THREADED),
409 };
410 
411 enum gro_result {
412 	GRO_MERGED,
413 	GRO_MERGED_FREE,
414 	GRO_HELD,
415 	GRO_NORMAL,
416 	GRO_CONSUMED,
417 };
418 typedef enum gro_result gro_result_t;
419 
420 /*
421  * enum rx_handler_result - Possible return values for rx_handlers.
422  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
423  * further.
424  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
425  * case skb->dev was changed by rx_handler.
426  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
427  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
428  *
429  * rx_handlers are functions called from inside __netif_receive_skb(), to do
430  * special processing of the skb, prior to delivery to protocol handlers.
431  *
432  * Currently, a net_device can only have a single rx_handler registered. Trying
433  * to register a second rx_handler will return -EBUSY.
434  *
435  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
436  * To unregister a rx_handler on a net_device, use
437  * netdev_rx_handler_unregister().
438  *
439  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
440  * do with the skb.
441  *
442  * If the rx_handler consumed the skb in some way, it should return
443  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
444  * the skb to be delivered in some other way.
445  *
446  * If the rx_handler changed skb->dev, to divert the skb to another
447  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
448  * new device will be called if it exists.
449  *
450  * If the rx_handler decides the skb should be ignored, it should return
451  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
452  * are registered on exact device (ptype->dev == skb->dev).
453  *
454  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
455  * delivered, it should return RX_HANDLER_PASS.
456  *
457  * A device without a registered rx_handler will behave as if rx_handler
458  * returned RX_HANDLER_PASS.
459  */
460 
461 enum rx_handler_result {
462 	RX_HANDLER_CONSUMED,
463 	RX_HANDLER_ANOTHER,
464 	RX_HANDLER_EXACT,
465 	RX_HANDLER_PASS,
466 };
467 typedef enum rx_handler_result rx_handler_result_t;
468 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
469 
470 void __napi_schedule(struct napi_struct *n);
471 void __napi_schedule_irqoff(struct napi_struct *n);
472 
473 static inline bool napi_disable_pending(struct napi_struct *n)
474 {
475 	return test_bit(NAPI_STATE_DISABLE, &n->state);
476 }
477 
478 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
479 {
480 	return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
481 }
482 
483 bool napi_schedule_prep(struct napi_struct *n);
484 
485 /**
486  *	napi_schedule - schedule NAPI poll
487  *	@n: NAPI context
488  *
489  * Schedule NAPI poll routine to be called if it is not already
490  * running.
491  */
492 static inline void napi_schedule(struct napi_struct *n)
493 {
494 	if (napi_schedule_prep(n))
495 		__napi_schedule(n);
496 }
497 
498 /**
499  *	napi_schedule_irqoff - schedule NAPI poll
500  *	@n: NAPI context
501  *
502  * Variant of napi_schedule(), assuming hard irqs are masked.
503  */
504 static inline void napi_schedule_irqoff(struct napi_struct *n)
505 {
506 	if (napi_schedule_prep(n))
507 		__napi_schedule_irqoff(n);
508 }
509 
510 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
511 static inline bool napi_reschedule(struct napi_struct *napi)
512 {
513 	if (napi_schedule_prep(napi)) {
514 		__napi_schedule(napi);
515 		return true;
516 	}
517 	return false;
518 }
519 
520 /**
521  * napi_complete_done - NAPI processing complete
522  * @n: NAPI context
523  * @work_done: number of packets processed
524  *
525  * Mark NAPI processing as complete. Should only be called if poll budget
526  * has not been completely consumed.
527  * Prefer over napi_complete().
528  * Return false if device should avoid rearming interrupts.
529  */
530 bool napi_complete_done(struct napi_struct *n, int work_done);
531 
532 static inline bool napi_complete(struct napi_struct *n)
533 {
534 	return napi_complete_done(n, 0);
535 }
536 
537 int dev_set_threaded(struct net_device *dev, bool threaded);
538 
539 /**
540  *	napi_disable - prevent NAPI from scheduling
541  *	@n: NAPI context
542  *
543  * Stop NAPI from being scheduled on this context.
544  * Waits till any outstanding processing completes.
545  */
546 void napi_disable(struct napi_struct *n);
547 
548 void napi_enable(struct napi_struct *n);
549 
550 /**
551  *	napi_synchronize - wait until NAPI is not running
552  *	@n: NAPI context
553  *
554  * Wait until NAPI is done being scheduled on this context.
555  * Waits till any outstanding processing completes but
556  * does not disable future activations.
557  */
558 static inline void napi_synchronize(const struct napi_struct *n)
559 {
560 	if (IS_ENABLED(CONFIG_SMP))
561 		while (test_bit(NAPI_STATE_SCHED, &n->state))
562 			msleep(1);
563 	else
564 		barrier();
565 }
566 
567 /**
568  *	napi_if_scheduled_mark_missed - if napi is running, set the
569  *	NAPIF_STATE_MISSED
570  *	@n: NAPI context
571  *
572  * If napi is running, set the NAPIF_STATE_MISSED, and return true if
573  * NAPI is scheduled.
574  **/
575 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
576 {
577 	unsigned long val, new;
578 
579 	val = READ_ONCE(n->state);
580 	do {
581 		if (val & NAPIF_STATE_DISABLE)
582 			return true;
583 
584 		if (!(val & NAPIF_STATE_SCHED))
585 			return false;
586 
587 		new = val | NAPIF_STATE_MISSED;
588 	} while (!try_cmpxchg(&n->state, &val, new));
589 
590 	return true;
591 }
592 
593 enum netdev_queue_state_t {
594 	__QUEUE_STATE_DRV_XOFF,
595 	__QUEUE_STATE_STACK_XOFF,
596 	__QUEUE_STATE_FROZEN,
597 };
598 
599 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
600 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
601 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
602 
603 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
604 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
605 					QUEUE_STATE_FROZEN)
606 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
607 					QUEUE_STATE_FROZEN)
608 
609 /*
610  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
611  * netif_tx_* functions below are used to manipulate this flag.  The
612  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
613  * queue independently.  The netif_xmit_*stopped functions below are called
614  * to check if the queue has been stopped by the driver or stack (either
615  * of the XOFF bits are set in the state).  Drivers should not need to call
616  * netif_xmit*stopped functions, they should only be using netif_tx_*.
617  */
618 
619 struct netdev_queue {
620 /*
621  * read-mostly part
622  */
623 	struct net_device	*dev;
624 	netdevice_tracker	dev_tracker;
625 
626 	struct Qdisc __rcu	*qdisc;
627 	struct Qdisc __rcu	*qdisc_sleeping;
628 #ifdef CONFIG_SYSFS
629 	struct kobject		kobj;
630 #endif
631 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
632 	int			numa_node;
633 #endif
634 	unsigned long		tx_maxrate;
635 	/*
636 	 * Number of TX timeouts for this queue
637 	 * (/sys/class/net/DEV/Q/trans_timeout)
638 	 */
639 	atomic_long_t		trans_timeout;
640 
641 	/* Subordinate device that the queue has been assigned to */
642 	struct net_device	*sb_dev;
643 #ifdef CONFIG_XDP_SOCKETS
644 	struct xsk_buff_pool    *pool;
645 #endif
646 /*
647  * write-mostly part
648  */
649 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
650 	int			xmit_lock_owner;
651 	/*
652 	 * Time (in jiffies) of last Tx
653 	 */
654 	unsigned long		trans_start;
655 
656 	unsigned long		state;
657 
658 #ifdef CONFIG_BQL
659 	struct dql		dql;
660 #endif
661 } ____cacheline_aligned_in_smp;
662 
663 extern int sysctl_fb_tunnels_only_for_init_net;
664 extern int sysctl_devconf_inherit_init_net;
665 
666 /*
667  * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
668  *                                     == 1 : For initns only
669  *                                     == 2 : For none.
670  */
671 static inline bool net_has_fallback_tunnels(const struct net *net)
672 {
673 #if IS_ENABLED(CONFIG_SYSCTL)
674 	int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
675 
676 	return !fb_tunnels_only_for_init_net ||
677 		(net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
678 #else
679 	return true;
680 #endif
681 }
682 
683 static inline int net_inherit_devconf(void)
684 {
685 #if IS_ENABLED(CONFIG_SYSCTL)
686 	return READ_ONCE(sysctl_devconf_inherit_init_net);
687 #else
688 	return 0;
689 #endif
690 }
691 
692 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
693 {
694 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
695 	return q->numa_node;
696 #else
697 	return NUMA_NO_NODE;
698 #endif
699 }
700 
701 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
702 {
703 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
704 	q->numa_node = node;
705 #endif
706 }
707 
708 #ifdef CONFIG_RPS
709 /*
710  * This structure holds an RPS map which can be of variable length.  The
711  * map is an array of CPUs.
712  */
713 struct rps_map {
714 	unsigned int len;
715 	struct rcu_head rcu;
716 	u16 cpus[];
717 };
718 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
719 
720 /*
721  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
722  * tail pointer for that CPU's input queue at the time of last enqueue, and
723  * a hardware filter index.
724  */
725 struct rps_dev_flow {
726 	u16 cpu;
727 	u16 filter;
728 	unsigned int last_qtail;
729 };
730 #define RPS_NO_FILTER 0xffff
731 
732 /*
733  * The rps_dev_flow_table structure contains a table of flow mappings.
734  */
735 struct rps_dev_flow_table {
736 	unsigned int mask;
737 	struct rcu_head rcu;
738 	struct rps_dev_flow flows[];
739 };
740 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
741     ((_num) * sizeof(struct rps_dev_flow)))
742 
743 /*
744  * The rps_sock_flow_table contains mappings of flows to the last CPU
745  * on which they were processed by the application (set in recvmsg).
746  * Each entry is a 32bit value. Upper part is the high-order bits
747  * of flow hash, lower part is CPU number.
748  * rps_cpu_mask is used to partition the space, depending on number of
749  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
750  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
751  * meaning we use 32-6=26 bits for the hash.
752  */
753 struct rps_sock_flow_table {
754 	u32	mask;
755 
756 	u32	ents[] ____cacheline_aligned_in_smp;
757 };
758 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
759 
760 #define RPS_NO_CPU 0xffff
761 
762 extern u32 rps_cpu_mask;
763 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
764 
765 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
766 					u32 hash)
767 {
768 	if (table && hash) {
769 		unsigned int index = hash & table->mask;
770 		u32 val = hash & ~rps_cpu_mask;
771 
772 		/* We only give a hint, preemption can change CPU under us */
773 		val |= raw_smp_processor_id();
774 
775 		/* The following WRITE_ONCE() is paired with the READ_ONCE()
776 		 * here, and another one in get_rps_cpu().
777 		 */
778 		if (READ_ONCE(table->ents[index]) != val)
779 			WRITE_ONCE(table->ents[index], val);
780 	}
781 }
782 
783 #ifdef CONFIG_RFS_ACCEL
784 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
785 			 u16 filter_id);
786 #endif
787 #endif /* CONFIG_RPS */
788 
789 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
790 enum xps_map_type {
791 	XPS_CPUS = 0,
792 	XPS_RXQS,
793 	XPS_MAPS_MAX,
794 };
795 
796 #ifdef CONFIG_XPS
797 /*
798  * This structure holds an XPS map which can be of variable length.  The
799  * map is an array of queues.
800  */
801 struct xps_map {
802 	unsigned int len;
803 	unsigned int alloc_len;
804 	struct rcu_head rcu;
805 	u16 queues[];
806 };
807 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
808 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
809        - sizeof(struct xps_map)) / sizeof(u16))
810 
811 /*
812  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
813  *
814  * We keep track of the number of cpus/rxqs used when the struct is allocated,
815  * in nr_ids. This will help not accessing out-of-bound memory.
816  *
817  * We keep track of the number of traffic classes used when the struct is
818  * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
819  * not crossing its upper bound, as the original dev->num_tc can be updated in
820  * the meantime.
821  */
822 struct xps_dev_maps {
823 	struct rcu_head rcu;
824 	unsigned int nr_ids;
825 	s16 num_tc;
826 	struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
827 };
828 
829 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +	\
830 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
831 
832 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
833 	(_rxqs * (_tcs) * sizeof(struct xps_map *)))
834 
835 #endif /* CONFIG_XPS */
836 
837 #define TC_MAX_QUEUE	16
838 #define TC_BITMASK	15
839 /* HW offloaded queuing disciplines txq count and offset maps */
840 struct netdev_tc_txq {
841 	u16 count;
842 	u16 offset;
843 };
844 
845 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
846 /*
847  * This structure is to hold information about the device
848  * configured to run FCoE protocol stack.
849  */
850 struct netdev_fcoe_hbainfo {
851 	char	manufacturer[64];
852 	char	serial_number[64];
853 	char	hardware_version[64];
854 	char	driver_version[64];
855 	char	optionrom_version[64];
856 	char	firmware_version[64];
857 	char	model[256];
858 	char	model_description[256];
859 };
860 #endif
861 
862 #define MAX_PHYS_ITEM_ID_LEN 32
863 
864 /* This structure holds a unique identifier to identify some
865  * physical item (port for example) used by a netdevice.
866  */
867 struct netdev_phys_item_id {
868 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
869 	unsigned char id_len;
870 };
871 
872 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
873 					    struct netdev_phys_item_id *b)
874 {
875 	return a->id_len == b->id_len &&
876 	       memcmp(a->id, b->id, a->id_len) == 0;
877 }
878 
879 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
880 				       struct sk_buff *skb,
881 				       struct net_device *sb_dev);
882 
883 enum net_device_path_type {
884 	DEV_PATH_ETHERNET = 0,
885 	DEV_PATH_VLAN,
886 	DEV_PATH_BRIDGE,
887 	DEV_PATH_PPPOE,
888 	DEV_PATH_DSA,
889 	DEV_PATH_MTK_WDMA,
890 };
891 
892 struct net_device_path {
893 	enum net_device_path_type	type;
894 	const struct net_device		*dev;
895 	union {
896 		struct {
897 			u16		id;
898 			__be16		proto;
899 			u8		h_dest[ETH_ALEN];
900 		} encap;
901 		struct {
902 			enum {
903 				DEV_PATH_BR_VLAN_KEEP,
904 				DEV_PATH_BR_VLAN_TAG,
905 				DEV_PATH_BR_VLAN_UNTAG,
906 				DEV_PATH_BR_VLAN_UNTAG_HW,
907 			}		vlan_mode;
908 			u16		vlan_id;
909 			__be16		vlan_proto;
910 		} bridge;
911 		struct {
912 			int port;
913 			u16 proto;
914 		} dsa;
915 		struct {
916 			u8 wdma_idx;
917 			u8 queue;
918 			u16 wcid;
919 			u8 bss;
920 		} mtk_wdma;
921 	};
922 };
923 
924 #define NET_DEVICE_PATH_STACK_MAX	5
925 #define NET_DEVICE_PATH_VLAN_MAX	2
926 
927 struct net_device_path_stack {
928 	int			num_paths;
929 	struct net_device_path	path[NET_DEVICE_PATH_STACK_MAX];
930 };
931 
932 struct net_device_path_ctx {
933 	const struct net_device *dev;
934 	u8			daddr[ETH_ALEN];
935 
936 	int			num_vlans;
937 	struct {
938 		u16		id;
939 		__be16		proto;
940 	} vlan[NET_DEVICE_PATH_VLAN_MAX];
941 };
942 
943 enum tc_setup_type {
944 	TC_QUERY_CAPS,
945 	TC_SETUP_QDISC_MQPRIO,
946 	TC_SETUP_CLSU32,
947 	TC_SETUP_CLSFLOWER,
948 	TC_SETUP_CLSMATCHALL,
949 	TC_SETUP_CLSBPF,
950 	TC_SETUP_BLOCK,
951 	TC_SETUP_QDISC_CBS,
952 	TC_SETUP_QDISC_RED,
953 	TC_SETUP_QDISC_PRIO,
954 	TC_SETUP_QDISC_MQ,
955 	TC_SETUP_QDISC_ETF,
956 	TC_SETUP_ROOT_QDISC,
957 	TC_SETUP_QDISC_GRED,
958 	TC_SETUP_QDISC_TAPRIO,
959 	TC_SETUP_FT,
960 	TC_SETUP_QDISC_ETS,
961 	TC_SETUP_QDISC_TBF,
962 	TC_SETUP_QDISC_FIFO,
963 	TC_SETUP_QDISC_HTB,
964 	TC_SETUP_ACT,
965 };
966 
967 /* These structures hold the attributes of bpf state that are being passed
968  * to the netdevice through the bpf op.
969  */
970 enum bpf_netdev_command {
971 	/* Set or clear a bpf program used in the earliest stages of packet
972 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
973 	 * is responsible for calling bpf_prog_put on any old progs that are
974 	 * stored. In case of error, the callee need not release the new prog
975 	 * reference, but on success it takes ownership and must bpf_prog_put
976 	 * when it is no longer used.
977 	 */
978 	XDP_SETUP_PROG,
979 	XDP_SETUP_PROG_HW,
980 	/* BPF program for offload callbacks, invoked at program load time. */
981 	BPF_OFFLOAD_MAP_ALLOC,
982 	BPF_OFFLOAD_MAP_FREE,
983 	XDP_SETUP_XSK_POOL,
984 };
985 
986 struct bpf_prog_offload_ops;
987 struct netlink_ext_ack;
988 struct xdp_umem;
989 struct xdp_dev_bulk_queue;
990 struct bpf_xdp_link;
991 
992 enum bpf_xdp_mode {
993 	XDP_MODE_SKB = 0,
994 	XDP_MODE_DRV = 1,
995 	XDP_MODE_HW = 2,
996 	__MAX_XDP_MODE
997 };
998 
999 struct bpf_xdp_entity {
1000 	struct bpf_prog *prog;
1001 	struct bpf_xdp_link *link;
1002 };
1003 
1004 struct netdev_bpf {
1005 	enum bpf_netdev_command command;
1006 	union {
1007 		/* XDP_SETUP_PROG */
1008 		struct {
1009 			u32 flags;
1010 			struct bpf_prog *prog;
1011 			struct netlink_ext_ack *extack;
1012 		};
1013 		/* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1014 		struct {
1015 			struct bpf_offloaded_map *offmap;
1016 		};
1017 		/* XDP_SETUP_XSK_POOL */
1018 		struct {
1019 			struct xsk_buff_pool *pool;
1020 			u16 queue_id;
1021 		} xsk;
1022 	};
1023 };
1024 
1025 /* Flags for ndo_xsk_wakeup. */
1026 #define XDP_WAKEUP_RX (1 << 0)
1027 #define XDP_WAKEUP_TX (1 << 1)
1028 
1029 #ifdef CONFIG_XFRM_OFFLOAD
1030 struct xfrmdev_ops {
1031 	int	(*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1032 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
1033 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
1034 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
1035 				       struct xfrm_state *x);
1036 	void	(*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1037 	void	(*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1038 	int	(*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1039 	void	(*xdo_dev_policy_delete) (struct xfrm_policy *x);
1040 	void	(*xdo_dev_policy_free) (struct xfrm_policy *x);
1041 };
1042 #endif
1043 
1044 struct dev_ifalias {
1045 	struct rcu_head rcuhead;
1046 	char ifalias[];
1047 };
1048 
1049 struct devlink;
1050 struct tlsdev_ops;
1051 
1052 struct netdev_net_notifier {
1053 	struct list_head list;
1054 	struct notifier_block *nb;
1055 };
1056 
1057 /*
1058  * This structure defines the management hooks for network devices.
1059  * The following hooks can be defined; unless noted otherwise, they are
1060  * optional and can be filled with a null pointer.
1061  *
1062  * int (*ndo_init)(struct net_device *dev);
1063  *     This function is called once when a network device is registered.
1064  *     The network device can use this for any late stage initialization
1065  *     or semantic validation. It can fail with an error code which will
1066  *     be propagated back to register_netdev.
1067  *
1068  * void (*ndo_uninit)(struct net_device *dev);
1069  *     This function is called when device is unregistered or when registration
1070  *     fails. It is not called if init fails.
1071  *
1072  * int (*ndo_open)(struct net_device *dev);
1073  *     This function is called when a network device transitions to the up
1074  *     state.
1075  *
1076  * int (*ndo_stop)(struct net_device *dev);
1077  *     This function is called when a network device transitions to the down
1078  *     state.
1079  *
1080  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1081  *                               struct net_device *dev);
1082  *	Called when a packet needs to be transmitted.
1083  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
1084  *	the queue before that can happen; it's for obsolete devices and weird
1085  *	corner cases, but the stack really does a non-trivial amount
1086  *	of useless work if you return NETDEV_TX_BUSY.
1087  *	Required; cannot be NULL.
1088  *
1089  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1090  *					   struct net_device *dev
1091  *					   netdev_features_t features);
1092  *	Called by core transmit path to determine if device is capable of
1093  *	performing offload operations on a given packet. This is to give
1094  *	the device an opportunity to implement any restrictions that cannot
1095  *	be otherwise expressed by feature flags. The check is called with
1096  *	the set of features that the stack has calculated and it returns
1097  *	those the driver believes to be appropriate.
1098  *
1099  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1100  *                         struct net_device *sb_dev);
1101  *	Called to decide which queue to use when device supports multiple
1102  *	transmit queues.
1103  *
1104  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1105  *	This function is called to allow device receiver to make
1106  *	changes to configuration when multicast or promiscuous is enabled.
1107  *
1108  * void (*ndo_set_rx_mode)(struct net_device *dev);
1109  *	This function is called device changes address list filtering.
1110  *	If driver handles unicast address filtering, it should set
1111  *	IFF_UNICAST_FLT in its priv_flags.
1112  *
1113  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1114  *	This function  is called when the Media Access Control address
1115  *	needs to be changed. If this interface is not defined, the
1116  *	MAC address can not be changed.
1117  *
1118  * int (*ndo_validate_addr)(struct net_device *dev);
1119  *	Test if Media Access Control address is valid for the device.
1120  *
1121  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1122  *	Old-style ioctl entry point. This is used internally by the
1123  *	appletalk and ieee802154 subsystems but is no longer called by
1124  *	the device ioctl handler.
1125  *
1126  * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1127  *	Used by the bonding driver for its device specific ioctls:
1128  *	SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1129  *	SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1130  *
1131  * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1132  *	Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1133  *	SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1134  *
1135  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1136  *	Used to set network devices bus interface parameters. This interface
1137  *	is retained for legacy reasons; new devices should use the bus
1138  *	interface (PCI) for low level management.
1139  *
1140  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1141  *	Called when a user wants to change the Maximum Transfer Unit
1142  *	of a device.
1143  *
1144  * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1145  *	Callback used when the transmitter has not made any progress
1146  *	for dev->watchdog ticks.
1147  *
1148  * void (*ndo_get_stats64)(struct net_device *dev,
1149  *                         struct rtnl_link_stats64 *storage);
1150  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1151  *	Called when a user wants to get the network device usage
1152  *	statistics. Drivers must do one of the following:
1153  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
1154  *	   rtnl_link_stats64 structure passed by the caller.
1155  *	2. Define @ndo_get_stats to update a net_device_stats structure
1156  *	   (which should normally be dev->stats) and return a pointer to
1157  *	   it. The structure may be changed asynchronously only if each
1158  *	   field is written atomically.
1159  *	3. Update dev->stats asynchronously and atomically, and define
1160  *	   neither operation.
1161  *
1162  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1163  *	Return true if this device supports offload stats of this attr_id.
1164  *
1165  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1166  *	void *attr_data)
1167  *	Get statistics for offload operations by attr_id. Write it into the
1168  *	attr_data pointer.
1169  *
1170  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1171  *	If device supports VLAN filtering this function is called when a
1172  *	VLAN id is registered.
1173  *
1174  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1175  *	If device supports VLAN filtering this function is called when a
1176  *	VLAN id is unregistered.
1177  *
1178  * void (*ndo_poll_controller)(struct net_device *dev);
1179  *
1180  *	SR-IOV management functions.
1181  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1182  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1183  *			  u8 qos, __be16 proto);
1184  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1185  *			  int max_tx_rate);
1186  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1187  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1188  * int (*ndo_get_vf_config)(struct net_device *dev,
1189  *			    int vf, struct ifla_vf_info *ivf);
1190  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1191  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1192  *			  struct nlattr *port[]);
1193  *
1194  *      Enable or disable the VF ability to query its RSS Redirection Table and
1195  *      Hash Key. This is needed since on some devices VF share this information
1196  *      with PF and querying it may introduce a theoretical security risk.
1197  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1198  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1199  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1200  *		       void *type_data);
1201  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
1202  *	This is always called from the stack with the rtnl lock held and netif
1203  *	tx queues stopped. This allows the netdevice to perform queue
1204  *	management safely.
1205  *
1206  *	Fiber Channel over Ethernet (FCoE) offload functions.
1207  * int (*ndo_fcoe_enable)(struct net_device *dev);
1208  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
1209  *	so the underlying device can perform whatever needed configuration or
1210  *	initialization to support acceleration of FCoE traffic.
1211  *
1212  * int (*ndo_fcoe_disable)(struct net_device *dev);
1213  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
1214  *	so the underlying device can perform whatever needed clean-ups to
1215  *	stop supporting acceleration of FCoE traffic.
1216  *
1217  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1218  *			     struct scatterlist *sgl, unsigned int sgc);
1219  *	Called when the FCoE Initiator wants to initialize an I/O that
1220  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1221  *	perform necessary setup and returns 1 to indicate the device is set up
1222  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1223  *
1224  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1225  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1226  *	indicated by the FC exchange id 'xid', so the underlying device can
1227  *	clean up and reuse resources for later DDP requests.
1228  *
1229  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1230  *			      struct scatterlist *sgl, unsigned int sgc);
1231  *	Called when the FCoE Target wants to initialize an I/O that
1232  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1233  *	perform necessary setup and returns 1 to indicate the device is set up
1234  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1235  *
1236  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1237  *			       struct netdev_fcoe_hbainfo *hbainfo);
1238  *	Called when the FCoE Protocol stack wants information on the underlying
1239  *	device. This information is utilized by the FCoE protocol stack to
1240  *	register attributes with Fiber Channel management service as per the
1241  *	FC-GS Fabric Device Management Information(FDMI) specification.
1242  *
1243  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1244  *	Called when the underlying device wants to override default World Wide
1245  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1246  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1247  *	protocol stack to use.
1248  *
1249  *	RFS acceleration.
1250  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1251  *			    u16 rxq_index, u32 flow_id);
1252  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1253  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1254  *	Return the filter ID on success, or a negative error code.
1255  *
1256  *	Slave management functions (for bridge, bonding, etc).
1257  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1258  *	Called to make another netdev an underling.
1259  *
1260  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1261  *	Called to release previously enslaved netdev.
1262  *
1263  * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1264  *					    struct sk_buff *skb,
1265  *					    bool all_slaves);
1266  *	Get the xmit slave of master device. If all_slaves is true, function
1267  *	assume all the slaves can transmit.
1268  *
1269  *      Feature/offload setting functions.
1270  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1271  *		netdev_features_t features);
1272  *	Adjusts the requested feature flags according to device-specific
1273  *	constraints, and returns the resulting flags. Must not modify
1274  *	the device state.
1275  *
1276  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1277  *	Called to update device configuration to new features. Passed
1278  *	feature set might be less than what was returned by ndo_fix_features()).
1279  *	Must return >0 or -errno if it changed dev->features itself.
1280  *
1281  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1282  *		      struct net_device *dev,
1283  *		      const unsigned char *addr, u16 vid, u16 flags,
1284  *		      struct netlink_ext_ack *extack);
1285  *	Adds an FDB entry to dev for addr.
1286  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1287  *		      struct net_device *dev,
1288  *		      const unsigned char *addr, u16 vid)
1289  *	Deletes the FDB entry from dev coresponding to addr.
1290  * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1291  *			   struct net_device *dev,
1292  *			   u16 vid,
1293  *			   struct netlink_ext_ack *extack);
1294  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1295  *		       struct net_device *dev, struct net_device *filter_dev,
1296  *		       int *idx)
1297  *	Used to add FDB entries to dump requests. Implementers should add
1298  *	entries to skb and update idx with the number of entries.
1299  *
1300  * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1301  *		      u16 nlmsg_flags, struct netlink_ext_ack *extack);
1302  *	Adds an MDB entry to dev.
1303  * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1304  *		      struct netlink_ext_ack *extack);
1305  *	Deletes the MDB entry from dev.
1306  * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1307  *		       struct netlink_callback *cb);
1308  *	Dumps MDB entries from dev. The first argument (marker) in the netlink
1309  *	callback is used by core rtnetlink code.
1310  *
1311  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1312  *			     u16 flags, struct netlink_ext_ack *extack)
1313  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1314  *			     struct net_device *dev, u32 filter_mask,
1315  *			     int nlflags)
1316  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1317  *			     u16 flags);
1318  *
1319  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1320  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1321  *	which do not represent real hardware may define this to allow their
1322  *	userspace components to manage their virtual carrier state. Devices
1323  *	that determine carrier state from physical hardware properties (eg
1324  *	network cables) or protocol-dependent mechanisms (eg
1325  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1326  *
1327  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1328  *			       struct netdev_phys_item_id *ppid);
1329  *	Called to get ID of physical port of this device. If driver does
1330  *	not implement this, it is assumed that the hw is not able to have
1331  *	multiple net devices on single physical port.
1332  *
1333  * int (*ndo_get_port_parent_id)(struct net_device *dev,
1334  *				 struct netdev_phys_item_id *ppid)
1335  *	Called to get the parent ID of the physical port of this device.
1336  *
1337  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1338  *				 struct net_device *dev)
1339  *	Called by upper layer devices to accelerate switching or other
1340  *	station functionality into hardware. 'pdev is the lowerdev
1341  *	to use for the offload and 'dev' is the net device that will
1342  *	back the offload. Returns a pointer to the private structure
1343  *	the upper layer will maintain.
1344  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1345  *	Called by upper layer device to delete the station created
1346  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1347  *	the station and priv is the structure returned by the add
1348  *	operation.
1349  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1350  *			     int queue_index, u32 maxrate);
1351  *	Called when a user wants to set a max-rate limitation of specific
1352  *	TX queue.
1353  * int (*ndo_get_iflink)(const struct net_device *dev);
1354  *	Called to get the iflink value of this device.
1355  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1356  *	This function is used to get egress tunnel information for given skb.
1357  *	This is useful for retrieving outer tunnel header parameters while
1358  *	sampling packet.
1359  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1360  *	This function is used to specify the headroom that the skb must
1361  *	consider when allocation skb during packet reception. Setting
1362  *	appropriate rx headroom value allows avoiding skb head copy on
1363  *	forward. Setting a negative value resets the rx headroom to the
1364  *	default value.
1365  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1366  *	This function is used to set or query state related to XDP on the
1367  *	netdevice and manage BPF offload. See definition of
1368  *	enum bpf_netdev_command for details.
1369  * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1370  *			u32 flags);
1371  *	This function is used to submit @n XDP packets for transmit on a
1372  *	netdevice. Returns number of frames successfully transmitted, frames
1373  *	that got dropped are freed/returned via xdp_return_frame().
1374  *	Returns negative number, means general error invoking ndo, meaning
1375  *	no frames were xmit'ed and core-caller will free all frames.
1376  * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1377  *					        struct xdp_buff *xdp);
1378  *      Get the xmit slave of master device based on the xdp_buff.
1379  * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1380  *      This function is used to wake up the softirq, ksoftirqd or kthread
1381  *	responsible for sending and/or receiving packets on a specific
1382  *	queue id bound to an AF_XDP socket. The flags field specifies if
1383  *	only RX, only Tx, or both should be woken up using the flags
1384  *	XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1385  * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1386  *			 int cmd);
1387  *	Add, change, delete or get information on an IPv4 tunnel.
1388  * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1389  *	If a device is paired with a peer device, return the peer instance.
1390  *	The caller must be under RCU read context.
1391  * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1392  *     Get the forwarding path to reach the real device from the HW destination address
1393  * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1394  *			     const struct skb_shared_hwtstamps *hwtstamps,
1395  *			     bool cycles);
1396  *	Get hardware timestamp based on normal/adjustable time or free running
1397  *	cycle counter. This function is required if physical clock supports a
1398  *	free running cycle counter.
1399  *
1400  * int (*ndo_hwtstamp_get)(struct net_device *dev,
1401  *			   struct kernel_hwtstamp_config *kernel_config);
1402  *	Get the currently configured hardware timestamping parameters for the
1403  *	NIC device.
1404  *
1405  * int (*ndo_hwtstamp_set)(struct net_device *dev,
1406  *			   struct kernel_hwtstamp_config *kernel_config,
1407  *			   struct netlink_ext_ack *extack);
1408  *	Change the hardware timestamping parameters for NIC device.
1409  */
1410 struct net_device_ops {
1411 	int			(*ndo_init)(struct net_device *dev);
1412 	void			(*ndo_uninit)(struct net_device *dev);
1413 	int			(*ndo_open)(struct net_device *dev);
1414 	int			(*ndo_stop)(struct net_device *dev);
1415 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1416 						  struct net_device *dev);
1417 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1418 						      struct net_device *dev,
1419 						      netdev_features_t features);
1420 	u16			(*ndo_select_queue)(struct net_device *dev,
1421 						    struct sk_buff *skb,
1422 						    struct net_device *sb_dev);
1423 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1424 						       int flags);
1425 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1426 	int			(*ndo_set_mac_address)(struct net_device *dev,
1427 						       void *addr);
1428 	int			(*ndo_validate_addr)(struct net_device *dev);
1429 	int			(*ndo_do_ioctl)(struct net_device *dev,
1430 					        struct ifreq *ifr, int cmd);
1431 	int			(*ndo_eth_ioctl)(struct net_device *dev,
1432 						 struct ifreq *ifr, int cmd);
1433 	int			(*ndo_siocbond)(struct net_device *dev,
1434 						struct ifreq *ifr, int cmd);
1435 	int			(*ndo_siocwandev)(struct net_device *dev,
1436 						  struct if_settings *ifs);
1437 	int			(*ndo_siocdevprivate)(struct net_device *dev,
1438 						      struct ifreq *ifr,
1439 						      void __user *data, int cmd);
1440 	int			(*ndo_set_config)(struct net_device *dev,
1441 					          struct ifmap *map);
1442 	int			(*ndo_change_mtu)(struct net_device *dev,
1443 						  int new_mtu);
1444 	int			(*ndo_neigh_setup)(struct net_device *dev,
1445 						   struct neigh_parms *);
1446 	void			(*ndo_tx_timeout) (struct net_device *dev,
1447 						   unsigned int txqueue);
1448 
1449 	void			(*ndo_get_stats64)(struct net_device *dev,
1450 						   struct rtnl_link_stats64 *storage);
1451 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1452 	int			(*ndo_get_offload_stats)(int attr_id,
1453 							 const struct net_device *dev,
1454 							 void *attr_data);
1455 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1456 
1457 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1458 						       __be16 proto, u16 vid);
1459 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1460 						        __be16 proto, u16 vid);
1461 #ifdef CONFIG_NET_POLL_CONTROLLER
1462 	void                    (*ndo_poll_controller)(struct net_device *dev);
1463 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1464 						     struct netpoll_info *info);
1465 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1466 #endif
1467 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1468 						  int queue, u8 *mac);
1469 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1470 						   int queue, u16 vlan,
1471 						   u8 qos, __be16 proto);
1472 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1473 						   int vf, int min_tx_rate,
1474 						   int max_tx_rate);
1475 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1476 						       int vf, bool setting);
1477 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1478 						    int vf, bool setting);
1479 	int			(*ndo_get_vf_config)(struct net_device *dev,
1480 						     int vf,
1481 						     struct ifla_vf_info *ivf);
1482 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1483 							 int vf, int link_state);
1484 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1485 						    int vf,
1486 						    struct ifla_vf_stats
1487 						    *vf_stats);
1488 	int			(*ndo_set_vf_port)(struct net_device *dev,
1489 						   int vf,
1490 						   struct nlattr *port[]);
1491 	int			(*ndo_get_vf_port)(struct net_device *dev,
1492 						   int vf, struct sk_buff *skb);
1493 	int			(*ndo_get_vf_guid)(struct net_device *dev,
1494 						   int vf,
1495 						   struct ifla_vf_guid *node_guid,
1496 						   struct ifla_vf_guid *port_guid);
1497 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1498 						   int vf, u64 guid,
1499 						   int guid_type);
1500 	int			(*ndo_set_vf_rss_query_en)(
1501 						   struct net_device *dev,
1502 						   int vf, bool setting);
1503 	int			(*ndo_setup_tc)(struct net_device *dev,
1504 						enum tc_setup_type type,
1505 						void *type_data);
1506 #if IS_ENABLED(CONFIG_FCOE)
1507 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1508 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1509 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1510 						      u16 xid,
1511 						      struct scatterlist *sgl,
1512 						      unsigned int sgc);
1513 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1514 						     u16 xid);
1515 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1516 						       u16 xid,
1517 						       struct scatterlist *sgl,
1518 						       unsigned int sgc);
1519 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1520 							struct netdev_fcoe_hbainfo *hbainfo);
1521 #endif
1522 
1523 #if IS_ENABLED(CONFIG_LIBFCOE)
1524 #define NETDEV_FCOE_WWNN 0
1525 #define NETDEV_FCOE_WWPN 1
1526 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1527 						    u64 *wwn, int type);
1528 #endif
1529 
1530 #ifdef CONFIG_RFS_ACCEL
1531 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1532 						     const struct sk_buff *skb,
1533 						     u16 rxq_index,
1534 						     u32 flow_id);
1535 #endif
1536 	int			(*ndo_add_slave)(struct net_device *dev,
1537 						 struct net_device *slave_dev,
1538 						 struct netlink_ext_ack *extack);
1539 	int			(*ndo_del_slave)(struct net_device *dev,
1540 						 struct net_device *slave_dev);
1541 	struct net_device*	(*ndo_get_xmit_slave)(struct net_device *dev,
1542 						      struct sk_buff *skb,
1543 						      bool all_slaves);
1544 	struct net_device*	(*ndo_sk_get_lower_dev)(struct net_device *dev,
1545 							struct sock *sk);
1546 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1547 						    netdev_features_t features);
1548 	int			(*ndo_set_features)(struct net_device *dev,
1549 						    netdev_features_t features);
1550 	int			(*ndo_neigh_construct)(struct net_device *dev,
1551 						       struct neighbour *n);
1552 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1553 						     struct neighbour *n);
1554 
1555 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1556 					       struct nlattr *tb[],
1557 					       struct net_device *dev,
1558 					       const unsigned char *addr,
1559 					       u16 vid,
1560 					       u16 flags,
1561 					       struct netlink_ext_ack *extack);
1562 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1563 					       struct nlattr *tb[],
1564 					       struct net_device *dev,
1565 					       const unsigned char *addr,
1566 					       u16 vid, struct netlink_ext_ack *extack);
1567 	int			(*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1568 						    struct nlattr *tb[],
1569 						    struct net_device *dev,
1570 						    u16 vid,
1571 						    struct netlink_ext_ack *extack);
1572 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1573 						struct netlink_callback *cb,
1574 						struct net_device *dev,
1575 						struct net_device *filter_dev,
1576 						int *idx);
1577 	int			(*ndo_fdb_get)(struct sk_buff *skb,
1578 					       struct nlattr *tb[],
1579 					       struct net_device *dev,
1580 					       const unsigned char *addr,
1581 					       u16 vid, u32 portid, u32 seq,
1582 					       struct netlink_ext_ack *extack);
1583 	int			(*ndo_mdb_add)(struct net_device *dev,
1584 					       struct nlattr *tb[],
1585 					       u16 nlmsg_flags,
1586 					       struct netlink_ext_ack *extack);
1587 	int			(*ndo_mdb_del)(struct net_device *dev,
1588 					       struct nlattr *tb[],
1589 					       struct netlink_ext_ack *extack);
1590 	int			(*ndo_mdb_dump)(struct net_device *dev,
1591 						struct sk_buff *skb,
1592 						struct netlink_callback *cb);
1593 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1594 						      struct nlmsghdr *nlh,
1595 						      u16 flags,
1596 						      struct netlink_ext_ack *extack);
1597 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1598 						      u32 pid, u32 seq,
1599 						      struct net_device *dev,
1600 						      u32 filter_mask,
1601 						      int nlflags);
1602 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1603 						      struct nlmsghdr *nlh,
1604 						      u16 flags);
1605 	int			(*ndo_change_carrier)(struct net_device *dev,
1606 						      bool new_carrier);
1607 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1608 							struct netdev_phys_item_id *ppid);
1609 	int			(*ndo_get_port_parent_id)(struct net_device *dev,
1610 							  struct netdev_phys_item_id *ppid);
1611 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1612 							  char *name, size_t len);
1613 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1614 							struct net_device *dev);
1615 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1616 							void *priv);
1617 
1618 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1619 						      int queue_index,
1620 						      u32 maxrate);
1621 	int			(*ndo_get_iflink)(const struct net_device *dev);
1622 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1623 						       struct sk_buff *skb);
1624 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1625 						       int needed_headroom);
1626 	int			(*ndo_bpf)(struct net_device *dev,
1627 					   struct netdev_bpf *bpf);
1628 	int			(*ndo_xdp_xmit)(struct net_device *dev, int n,
1629 						struct xdp_frame **xdp,
1630 						u32 flags);
1631 	struct net_device *	(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1632 							  struct xdp_buff *xdp);
1633 	int			(*ndo_xsk_wakeup)(struct net_device *dev,
1634 						  u32 queue_id, u32 flags);
1635 	int			(*ndo_tunnel_ctl)(struct net_device *dev,
1636 						  struct ip_tunnel_parm *p, int cmd);
1637 	struct net_device *	(*ndo_get_peer_dev)(struct net_device *dev);
1638 	int                     (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1639                                                          struct net_device_path *path);
1640 	ktime_t			(*ndo_get_tstamp)(struct net_device *dev,
1641 						  const struct skb_shared_hwtstamps *hwtstamps,
1642 						  bool cycles);
1643 	int			(*ndo_hwtstamp_get)(struct net_device *dev,
1644 						    struct kernel_hwtstamp_config *kernel_config);
1645 	int			(*ndo_hwtstamp_set)(struct net_device *dev,
1646 						    struct kernel_hwtstamp_config *kernel_config,
1647 						    struct netlink_ext_ack *extack);
1648 };
1649 
1650 /**
1651  * enum netdev_priv_flags - &struct net_device priv_flags
1652  *
1653  * These are the &struct net_device, they are only set internally
1654  * by drivers and used in the kernel. These flags are invisible to
1655  * userspace; this means that the order of these flags can change
1656  * during any kernel release.
1657  *
1658  * You should have a pretty good reason to be extending these flags.
1659  *
1660  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1661  * @IFF_EBRIDGE: Ethernet bridging device
1662  * @IFF_BONDING: bonding master or slave
1663  * @IFF_ISATAP: ISATAP interface (RFC4214)
1664  * @IFF_WAN_HDLC: WAN HDLC device
1665  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1666  *	release skb->dst
1667  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1668  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1669  * @IFF_MACVLAN_PORT: device used as macvlan port
1670  * @IFF_BRIDGE_PORT: device used as bridge port
1671  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1672  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1673  * @IFF_UNICAST_FLT: Supports unicast filtering
1674  * @IFF_TEAM_PORT: device used as team port
1675  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1676  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1677  *	change when it's running
1678  * @IFF_MACVLAN: Macvlan device
1679  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1680  *	underlying stacked devices
1681  * @IFF_L3MDEV_MASTER: device is an L3 master device
1682  * @IFF_NO_QUEUE: device can run without qdisc attached
1683  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1684  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1685  * @IFF_TEAM: device is a team device
1686  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1687  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1688  *	entity (i.e. the master device for bridged veth)
1689  * @IFF_MACSEC: device is a MACsec device
1690  * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1691  * @IFF_FAILOVER: device is a failover master device
1692  * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1693  * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1694  * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1695  * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1696  *	skb_headlen(skb) == 0 (data starts from frag0)
1697  * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1698  * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1699  *	ndo_hwtstamp_set() for all timestamp requests regardless of source,
1700  *	even if those aren't HWTSTAMP_SOURCE_NETDEV.
1701  */
1702 enum netdev_priv_flags {
1703 	IFF_802_1Q_VLAN			= 1<<0,
1704 	IFF_EBRIDGE			= 1<<1,
1705 	IFF_BONDING			= 1<<2,
1706 	IFF_ISATAP			= 1<<3,
1707 	IFF_WAN_HDLC			= 1<<4,
1708 	IFF_XMIT_DST_RELEASE		= 1<<5,
1709 	IFF_DONT_BRIDGE			= 1<<6,
1710 	IFF_DISABLE_NETPOLL		= 1<<7,
1711 	IFF_MACVLAN_PORT		= 1<<8,
1712 	IFF_BRIDGE_PORT			= 1<<9,
1713 	IFF_OVS_DATAPATH		= 1<<10,
1714 	IFF_TX_SKB_SHARING		= 1<<11,
1715 	IFF_UNICAST_FLT			= 1<<12,
1716 	IFF_TEAM_PORT			= 1<<13,
1717 	IFF_SUPP_NOFCS			= 1<<14,
1718 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1719 	IFF_MACVLAN			= 1<<16,
1720 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1721 	IFF_L3MDEV_MASTER		= 1<<18,
1722 	IFF_NO_QUEUE			= 1<<19,
1723 	IFF_OPENVSWITCH			= 1<<20,
1724 	IFF_L3MDEV_SLAVE		= 1<<21,
1725 	IFF_TEAM			= 1<<22,
1726 	IFF_RXFH_CONFIGURED		= 1<<23,
1727 	IFF_PHONY_HEADROOM		= 1<<24,
1728 	IFF_MACSEC			= 1<<25,
1729 	IFF_NO_RX_HANDLER		= 1<<26,
1730 	IFF_FAILOVER			= 1<<27,
1731 	IFF_FAILOVER_SLAVE		= 1<<28,
1732 	IFF_L3MDEV_RX_HANDLER		= 1<<29,
1733 	IFF_NO_ADDRCONF			= BIT_ULL(30),
1734 	IFF_TX_SKB_NO_LINEAR		= BIT_ULL(31),
1735 	IFF_CHANGE_PROTO_DOWN		= BIT_ULL(32),
1736 	IFF_SEE_ALL_HWTSTAMP_REQUESTS	= BIT_ULL(33),
1737 };
1738 
1739 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1740 #define IFF_EBRIDGE			IFF_EBRIDGE
1741 #define IFF_BONDING			IFF_BONDING
1742 #define IFF_ISATAP			IFF_ISATAP
1743 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1744 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1745 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1746 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1747 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1748 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1749 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1750 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1751 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1752 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1753 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1754 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1755 #define IFF_MACVLAN			IFF_MACVLAN
1756 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1757 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1758 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1759 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1760 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1761 #define IFF_TEAM			IFF_TEAM
1762 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1763 #define IFF_PHONY_HEADROOM		IFF_PHONY_HEADROOM
1764 #define IFF_MACSEC			IFF_MACSEC
1765 #define IFF_NO_RX_HANDLER		IFF_NO_RX_HANDLER
1766 #define IFF_FAILOVER			IFF_FAILOVER
1767 #define IFF_FAILOVER_SLAVE		IFF_FAILOVER_SLAVE
1768 #define IFF_L3MDEV_RX_HANDLER		IFF_L3MDEV_RX_HANDLER
1769 #define IFF_TX_SKB_NO_LINEAR		IFF_TX_SKB_NO_LINEAR
1770 
1771 /* Specifies the type of the struct net_device::ml_priv pointer */
1772 enum netdev_ml_priv_type {
1773 	ML_PRIV_NONE,
1774 	ML_PRIV_CAN,
1775 };
1776 
1777 /**
1778  *	struct net_device - The DEVICE structure.
1779  *
1780  *	Actually, this whole structure is a big mistake.  It mixes I/O
1781  *	data with strictly "high-level" data, and it has to know about
1782  *	almost every data structure used in the INET module.
1783  *
1784  *	@name:	This is the first field of the "visible" part of this structure
1785  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1786  *		of the interface.
1787  *
1788  *	@name_node:	Name hashlist node
1789  *	@ifalias:	SNMP alias
1790  *	@mem_end:	Shared memory end
1791  *	@mem_start:	Shared memory start
1792  *	@base_addr:	Device I/O address
1793  *	@irq:		Device IRQ number
1794  *
1795  *	@state:		Generic network queuing layer state, see netdev_state_t
1796  *	@dev_list:	The global list of network devices
1797  *	@napi_list:	List entry used for polling NAPI devices
1798  *	@unreg_list:	List entry  when we are unregistering the
1799  *			device; see the function unregister_netdev
1800  *	@close_list:	List entry used when we are closing the device
1801  *	@ptype_all:     Device-specific packet handlers for all protocols
1802  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1803  *
1804  *	@adj_list:	Directly linked devices, like slaves for bonding
1805  *	@features:	Currently active device features
1806  *	@hw_features:	User-changeable features
1807  *
1808  *	@wanted_features:	User-requested features
1809  *	@vlan_features:		Mask of features inheritable by VLAN devices
1810  *
1811  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1812  *				This field indicates what encapsulation
1813  *				offloads the hardware is capable of doing,
1814  *				and drivers will need to set them appropriately.
1815  *
1816  *	@mpls_features:	Mask of features inheritable by MPLS
1817  *	@gso_partial_features: value(s) from NETIF_F_GSO\*
1818  *
1819  *	@ifindex:	interface index
1820  *	@group:		The group the device belongs to
1821  *
1822  *	@stats:		Statistics struct, which was left as a legacy, use
1823  *			rtnl_link_stats64 instead
1824  *
1825  *	@core_stats:	core networking counters,
1826  *			do not use this in drivers
1827  *	@carrier_up_count:	Number of times the carrier has been up
1828  *	@carrier_down_count:	Number of times the carrier has been down
1829  *
1830  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1831  *				instead of ioctl,
1832  *				see <net/iw_handler.h> for details.
1833  *	@wireless_data:	Instance data managed by the core of wireless extensions
1834  *
1835  *	@netdev_ops:	Includes several pointers to callbacks,
1836  *			if one wants to override the ndo_*() functions
1837  *	@xdp_metadata_ops:	Includes pointers to XDP metadata callbacks.
1838  *	@ethtool_ops:	Management operations
1839  *	@l3mdev_ops:	Layer 3 master device operations
1840  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1841  *			discovery handling. Necessary for e.g. 6LoWPAN.
1842  *	@xfrmdev_ops:	Transformation offload operations
1843  *	@tlsdev_ops:	Transport Layer Security offload operations
1844  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1845  *			of Layer 2 headers.
1846  *
1847  *	@flags:		Interface flags (a la BSD)
1848  *	@xdp_features:	XDP capability supported by the device
1849  *	@priv_flags:	Like 'flags' but invisible to userspace,
1850  *			see if.h for the definitions
1851  *	@gflags:	Global flags ( kept as legacy )
1852  *	@padded:	How much padding added by alloc_netdev()
1853  *	@operstate:	RFC2863 operstate
1854  *	@link_mode:	Mapping policy to operstate
1855  *	@if_port:	Selectable AUI, TP, ...
1856  *	@dma:		DMA channel
1857  *	@mtu:		Interface MTU value
1858  *	@min_mtu:	Interface Minimum MTU value
1859  *	@max_mtu:	Interface Maximum MTU value
1860  *	@type:		Interface hardware type
1861  *	@hard_header_len: Maximum hardware header length.
1862  *	@min_header_len:  Minimum hardware header length
1863  *
1864  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1865  *			  cases can this be guaranteed
1866  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1867  *			  cases can this be guaranteed. Some cases also use
1868  *			  LL_MAX_HEADER instead to allocate the skb
1869  *
1870  *	interface address info:
1871  *
1872  * 	@perm_addr:		Permanent hw address
1873  * 	@addr_assign_type:	Hw address assignment type
1874  * 	@addr_len:		Hardware address length
1875  *	@upper_level:		Maximum depth level of upper devices.
1876  *	@lower_level:		Maximum depth level of lower devices.
1877  *	@neigh_priv_len:	Used in neigh_alloc()
1878  * 	@dev_id:		Used to differentiate devices that share
1879  * 				the same link layer address
1880  * 	@dev_port:		Used to differentiate devices that share
1881  * 				the same function
1882  *	@addr_list_lock:	XXX: need comments on this one
1883  *	@name_assign_type:	network interface name assignment type
1884  *	@uc_promisc:		Counter that indicates promiscuous mode
1885  *				has been enabled due to the need to listen to
1886  *				additional unicast addresses in a device that
1887  *				does not implement ndo_set_rx_mode()
1888  *	@uc:			unicast mac addresses
1889  *	@mc:			multicast mac addresses
1890  *	@dev_addrs:		list of device hw addresses
1891  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1892  *	@promiscuity:		Number of times the NIC is told to work in
1893  *				promiscuous mode; if it becomes 0 the NIC will
1894  *				exit promiscuous mode
1895  *	@allmulti:		Counter, enables or disables allmulticast mode
1896  *
1897  *	@vlan_info:	VLAN info
1898  *	@dsa_ptr:	dsa specific data
1899  *	@tipc_ptr:	TIPC specific data
1900  *	@atalk_ptr:	AppleTalk link
1901  *	@ip_ptr:	IPv4 specific data
1902  *	@ip6_ptr:	IPv6 specific data
1903  *	@ax25_ptr:	AX.25 specific data
1904  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1905  *	@ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1906  *			 device struct
1907  *	@mpls_ptr:	mpls_dev struct pointer
1908  *	@mctp_ptr:	MCTP specific data
1909  *
1910  *	@dev_addr:	Hw address (before bcast,
1911  *			because most packets are unicast)
1912  *
1913  *	@_rx:			Array of RX queues
1914  *	@num_rx_queues:		Number of RX queues
1915  *				allocated at register_netdev() time
1916  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1917  *	@xdp_prog:		XDP sockets filter program pointer
1918  *	@gro_flush_timeout:	timeout for GRO layer in NAPI
1919  *	@napi_defer_hard_irqs:	If not zero, provides a counter that would
1920  *				allow to avoid NIC hard IRQ, on busy queues.
1921  *
1922  *	@rx_handler:		handler for received packets
1923  *	@rx_handler_data: 	XXX: need comments on this one
1924  *	@tcx_ingress:		BPF & clsact qdisc specific data for ingress processing
1925  *	@ingress_queue:		XXX: need comments on this one
1926  *	@nf_hooks_ingress:	netfilter hooks executed for ingress packets
1927  *	@broadcast:		hw bcast address
1928  *
1929  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1930  *			indexed by RX queue number. Assigned by driver.
1931  *			This must only be set if the ndo_rx_flow_steer
1932  *			operation is defined
1933  *	@index_hlist:		Device index hash chain
1934  *
1935  *	@_tx:			Array of TX queues
1936  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1937  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1938  *	@qdisc:			Root qdisc from userspace point of view
1939  *	@tx_queue_len:		Max frames per queue allowed
1940  *	@tx_global_lock: 	XXX: need comments on this one
1941  *	@xdp_bulkq:		XDP device bulk queue
1942  *	@xps_maps:		all CPUs/RXQs maps for XPS device
1943  *
1944  *	@xps_maps:	XXX: need comments on this one
1945  *	@tcx_egress:		BPF & clsact qdisc specific data for egress processing
1946  *	@nf_hooks_egress:	netfilter hooks executed for egress packets
1947  *	@qdisc_hash:		qdisc hash table
1948  *	@watchdog_timeo:	Represents the timeout that is used by
1949  *				the watchdog (see dev_watchdog())
1950  *	@watchdog_timer:	List of timers
1951  *
1952  *	@proto_down_reason:	reason a netdev interface is held down
1953  *	@pcpu_refcnt:		Number of references to this device
1954  *	@dev_refcnt:		Number of references to this device
1955  *	@refcnt_tracker:	Tracker directory for tracked references to this device
1956  *	@todo_list:		Delayed register/unregister
1957  *	@link_watch_list:	XXX: need comments on this one
1958  *
1959  *	@reg_state:		Register/unregister state machine
1960  *	@dismantle:		Device is going to be freed
1961  *	@rtnl_link_state:	This enum represents the phases of creating
1962  *				a new link
1963  *
1964  *	@needs_free_netdev:	Should unregister perform free_netdev?
1965  *	@priv_destructor:	Called from unregister
1966  *	@npinfo:		XXX: need comments on this one
1967  * 	@nd_net:		Network namespace this network device is inside
1968  *
1969  * 	@ml_priv:	Mid-layer private
1970  *	@ml_priv_type:  Mid-layer private type
1971  * 	@lstats:	Loopback statistics
1972  * 	@tstats:	Tunnel statistics
1973  * 	@dstats:	Dummy statistics
1974  * 	@vstats:	Virtual ethernet statistics
1975  *
1976  *	@garp_port:	GARP
1977  *	@mrp_port:	MRP
1978  *
1979  *	@dm_private:	Drop monitor private
1980  *
1981  *	@dev:		Class/net/name entry
1982  *	@sysfs_groups:	Space for optional device, statistics and wireless
1983  *			sysfs groups
1984  *
1985  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1986  *	@rtnl_link_ops:	Rtnl_link_ops
1987  *
1988  *	@gso_max_size:	Maximum size of generic segmentation offload
1989  *	@tso_max_size:	Device (as in HW) limit on the max TSO request size
1990  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1991  *			NIC for GSO
1992  *	@tso_max_segs:	Device (as in HW) limit on the max TSO segment count
1993  * 	@gso_ipv4_max_size:	Maximum size of generic segmentation offload,
1994  * 				for IPv4.
1995  *
1996  *	@dcbnl_ops:	Data Center Bridging netlink ops
1997  *	@num_tc:	Number of traffic classes in the net device
1998  *	@tc_to_txq:	XXX: need comments on this one
1999  *	@prio_tc_map:	XXX: need comments on this one
2000  *
2001  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
2002  *
2003  *	@priomap:	XXX: need comments on this one
2004  *	@phydev:	Physical device may attach itself
2005  *			for hardware timestamping
2006  *	@sfp_bus:	attached &struct sfp_bus structure.
2007  *
2008  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2009  *
2010  *	@proto_down:	protocol port state information can be sent to the
2011  *			switch driver and used to set the phys state of the
2012  *			switch port.
2013  *
2014  *	@wol_enabled:	Wake-on-LAN is enabled
2015  *
2016  *	@threaded:	napi threaded mode is enabled
2017  *
2018  *	@net_notifier_list:	List of per-net netdev notifier block
2019  *				that follow this device when it is moved
2020  *				to another network namespace.
2021  *
2022  *	@macsec_ops:    MACsec offloading ops
2023  *
2024  *	@udp_tunnel_nic_info:	static structure describing the UDP tunnel
2025  *				offload capabilities of the device
2026  *	@udp_tunnel_nic:	UDP tunnel offload state
2027  *	@xdp_state:		stores info on attached XDP BPF programs
2028  *
2029  *	@nested_level:	Used as a parameter of spin_lock_nested() of
2030  *			dev->addr_list_lock.
2031  *	@unlink_list:	As netif_addr_lock() can be called recursively,
2032  *			keep a list of interfaces to be deleted.
2033  *	@gro_max_size:	Maximum size of aggregated packet in generic
2034  *			receive offload (GRO)
2035  * 	@gro_ipv4_max_size:	Maximum size of aggregated packet in generic
2036  * 				receive offload (GRO), for IPv4.
2037  *	@xdp_zc_max_segs:	Maximum number of segments supported by AF_XDP
2038  *				zero copy driver
2039  *
2040  *	@dev_addr_shadow:	Copy of @dev_addr to catch direct writes.
2041  *	@linkwatch_dev_tracker:	refcount tracker used by linkwatch.
2042  *	@watchdog_dev_tracker:	refcount tracker used by watchdog.
2043  *	@dev_registered_tracker:	tracker for reference held while
2044  *					registered
2045  *	@offload_xstats_l3:	L3 HW stats for this netdevice.
2046  *
2047  *	@devlink_port:	Pointer to related devlink port structure.
2048  *			Assigned by a driver before netdev registration using
2049  *			SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2050  *			during the time netdevice is registered.
2051  *
2052  *	FIXME: cleanup struct net_device such that network protocol info
2053  *	moves out.
2054  */
2055 
2056 struct net_device {
2057 	char			name[IFNAMSIZ];
2058 	struct netdev_name_node	*name_node;
2059 	struct dev_ifalias	__rcu *ifalias;
2060 	/*
2061 	 *	I/O specific fields
2062 	 *	FIXME: Merge these and struct ifmap into one
2063 	 */
2064 	unsigned long		mem_end;
2065 	unsigned long		mem_start;
2066 	unsigned long		base_addr;
2067 
2068 	/*
2069 	 *	Some hardware also needs these fields (state,dev_list,
2070 	 *	napi_list,unreg_list,close_list) but they are not
2071 	 *	part of the usual set specified in Space.c.
2072 	 */
2073 
2074 	unsigned long		state;
2075 
2076 	struct list_head	dev_list;
2077 	struct list_head	napi_list;
2078 	struct list_head	unreg_list;
2079 	struct list_head	close_list;
2080 	struct list_head	ptype_all;
2081 	struct list_head	ptype_specific;
2082 
2083 	struct {
2084 		struct list_head upper;
2085 		struct list_head lower;
2086 	} adj_list;
2087 
2088 	/* Read-mostly cache-line for fast-path access */
2089 	unsigned int		flags;
2090 	xdp_features_t		xdp_features;
2091 	unsigned long long	priv_flags;
2092 	const struct net_device_ops *netdev_ops;
2093 	const struct xdp_metadata_ops *xdp_metadata_ops;
2094 	int			ifindex;
2095 	unsigned short		gflags;
2096 	unsigned short		hard_header_len;
2097 
2098 	/* Note : dev->mtu is often read without holding a lock.
2099 	 * Writers usually hold RTNL.
2100 	 * It is recommended to use READ_ONCE() to annotate the reads,
2101 	 * and to use WRITE_ONCE() to annotate the writes.
2102 	 */
2103 	unsigned int		mtu;
2104 	unsigned short		needed_headroom;
2105 	unsigned short		needed_tailroom;
2106 
2107 	netdev_features_t	features;
2108 	netdev_features_t	hw_features;
2109 	netdev_features_t	wanted_features;
2110 	netdev_features_t	vlan_features;
2111 	netdev_features_t	hw_enc_features;
2112 	netdev_features_t	mpls_features;
2113 	netdev_features_t	gso_partial_features;
2114 
2115 	unsigned int		min_mtu;
2116 	unsigned int		max_mtu;
2117 	unsigned short		type;
2118 	unsigned char		min_header_len;
2119 	unsigned char		name_assign_type;
2120 
2121 	int			group;
2122 
2123 	struct net_device_stats	stats; /* not used by modern drivers */
2124 
2125 	struct net_device_core_stats __percpu *core_stats;
2126 
2127 	/* Stats to monitor link on/off, flapping */
2128 	atomic_t		carrier_up_count;
2129 	atomic_t		carrier_down_count;
2130 
2131 #ifdef CONFIG_WIRELESS_EXT
2132 	const struct iw_handler_def *wireless_handlers;
2133 	struct iw_public_data	*wireless_data;
2134 #endif
2135 	const struct ethtool_ops *ethtool_ops;
2136 #ifdef CONFIG_NET_L3_MASTER_DEV
2137 	const struct l3mdev_ops	*l3mdev_ops;
2138 #endif
2139 #if IS_ENABLED(CONFIG_IPV6)
2140 	const struct ndisc_ops *ndisc_ops;
2141 #endif
2142 
2143 #ifdef CONFIG_XFRM_OFFLOAD
2144 	const struct xfrmdev_ops *xfrmdev_ops;
2145 #endif
2146 
2147 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2148 	const struct tlsdev_ops *tlsdev_ops;
2149 #endif
2150 
2151 	const struct header_ops *header_ops;
2152 
2153 	unsigned char		operstate;
2154 	unsigned char		link_mode;
2155 
2156 	unsigned char		if_port;
2157 	unsigned char		dma;
2158 
2159 	/* Interface address info. */
2160 	unsigned char		perm_addr[MAX_ADDR_LEN];
2161 	unsigned char		addr_assign_type;
2162 	unsigned char		addr_len;
2163 	unsigned char		upper_level;
2164 	unsigned char		lower_level;
2165 
2166 	unsigned short		neigh_priv_len;
2167 	unsigned short          dev_id;
2168 	unsigned short          dev_port;
2169 	unsigned short		padded;
2170 
2171 	spinlock_t		addr_list_lock;
2172 	int			irq;
2173 
2174 	struct netdev_hw_addr_list	uc;
2175 	struct netdev_hw_addr_list	mc;
2176 	struct netdev_hw_addr_list	dev_addrs;
2177 
2178 #ifdef CONFIG_SYSFS
2179 	struct kset		*queues_kset;
2180 #endif
2181 #ifdef CONFIG_LOCKDEP
2182 	struct list_head	unlink_list;
2183 #endif
2184 	unsigned int		promiscuity;
2185 	unsigned int		allmulti;
2186 	bool			uc_promisc;
2187 #ifdef CONFIG_LOCKDEP
2188 	unsigned char		nested_level;
2189 #endif
2190 
2191 
2192 	/* Protocol-specific pointers */
2193 
2194 	struct in_device __rcu	*ip_ptr;
2195 	struct inet6_dev __rcu	*ip6_ptr;
2196 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2197 	struct vlan_info __rcu	*vlan_info;
2198 #endif
2199 #if IS_ENABLED(CONFIG_NET_DSA)
2200 	struct dsa_port		*dsa_ptr;
2201 #endif
2202 #if IS_ENABLED(CONFIG_TIPC)
2203 	struct tipc_bearer __rcu *tipc_ptr;
2204 #endif
2205 #if IS_ENABLED(CONFIG_ATALK)
2206 	void 			*atalk_ptr;
2207 #endif
2208 #if IS_ENABLED(CONFIG_AX25)
2209 	void			*ax25_ptr;
2210 #endif
2211 #if IS_ENABLED(CONFIG_CFG80211)
2212 	struct wireless_dev	*ieee80211_ptr;
2213 #endif
2214 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2215 	struct wpan_dev		*ieee802154_ptr;
2216 #endif
2217 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2218 	struct mpls_dev __rcu	*mpls_ptr;
2219 #endif
2220 #if IS_ENABLED(CONFIG_MCTP)
2221 	struct mctp_dev __rcu	*mctp_ptr;
2222 #endif
2223 
2224 /*
2225  * Cache lines mostly used on receive path (including eth_type_trans())
2226  */
2227 	/* Interface address info used in eth_type_trans() */
2228 	const unsigned char	*dev_addr;
2229 
2230 	struct netdev_rx_queue	*_rx;
2231 	unsigned int		num_rx_queues;
2232 	unsigned int		real_num_rx_queues;
2233 
2234 	struct bpf_prog __rcu	*xdp_prog;
2235 	unsigned long		gro_flush_timeout;
2236 	int			napi_defer_hard_irqs;
2237 #define GRO_LEGACY_MAX_SIZE	65536u
2238 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2239  * and shinfo->gso_segs is a 16bit field.
2240  */
2241 #define GRO_MAX_SIZE		(8 * 65535u)
2242 	unsigned int		gro_max_size;
2243 	unsigned int		gro_ipv4_max_size;
2244 	unsigned int		xdp_zc_max_segs;
2245 	rx_handler_func_t __rcu	*rx_handler;
2246 	void __rcu		*rx_handler_data;
2247 #ifdef CONFIG_NET_XGRESS
2248 	struct bpf_mprog_entry __rcu *tcx_ingress;
2249 #endif
2250 	struct netdev_queue __rcu *ingress_queue;
2251 #ifdef CONFIG_NETFILTER_INGRESS
2252 	struct nf_hook_entries __rcu *nf_hooks_ingress;
2253 #endif
2254 
2255 	unsigned char		broadcast[MAX_ADDR_LEN];
2256 #ifdef CONFIG_RFS_ACCEL
2257 	struct cpu_rmap		*rx_cpu_rmap;
2258 #endif
2259 	struct hlist_node	index_hlist;
2260 
2261 /*
2262  * Cache lines mostly used on transmit path
2263  */
2264 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
2265 	unsigned int		num_tx_queues;
2266 	unsigned int		real_num_tx_queues;
2267 	struct Qdisc __rcu	*qdisc;
2268 	unsigned int		tx_queue_len;
2269 	spinlock_t		tx_global_lock;
2270 
2271 	struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2272 
2273 #ifdef CONFIG_XPS
2274 	struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2275 #endif
2276 #ifdef CONFIG_NET_XGRESS
2277 	struct bpf_mprog_entry __rcu *tcx_egress;
2278 #endif
2279 #ifdef CONFIG_NETFILTER_EGRESS
2280 	struct nf_hook_entries __rcu *nf_hooks_egress;
2281 #endif
2282 
2283 #ifdef CONFIG_NET_SCHED
2284 	DECLARE_HASHTABLE	(qdisc_hash, 4);
2285 #endif
2286 	/* These may be needed for future network-power-down code. */
2287 	struct timer_list	watchdog_timer;
2288 	int			watchdog_timeo;
2289 
2290 	u32                     proto_down_reason;
2291 
2292 	struct list_head	todo_list;
2293 
2294 #ifdef CONFIG_PCPU_DEV_REFCNT
2295 	int __percpu		*pcpu_refcnt;
2296 #else
2297 	refcount_t		dev_refcnt;
2298 #endif
2299 	struct ref_tracker_dir	refcnt_tracker;
2300 
2301 	struct list_head	link_watch_list;
2302 
2303 	enum { NETREG_UNINITIALIZED=0,
2304 	       NETREG_REGISTERED,	/* completed register_netdevice */
2305 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
2306 	       NETREG_UNREGISTERED,	/* completed unregister todo */
2307 	       NETREG_RELEASED,		/* called free_netdev */
2308 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
2309 	} reg_state:8;
2310 
2311 	bool dismantle;
2312 
2313 	enum {
2314 		RTNL_LINK_INITIALIZED,
2315 		RTNL_LINK_INITIALIZING,
2316 	} rtnl_link_state:16;
2317 
2318 	bool needs_free_netdev;
2319 	void (*priv_destructor)(struct net_device *dev);
2320 
2321 #ifdef CONFIG_NETPOLL
2322 	struct netpoll_info __rcu	*npinfo;
2323 #endif
2324 
2325 	possible_net_t			nd_net;
2326 
2327 	/* mid-layer private */
2328 	void				*ml_priv;
2329 	enum netdev_ml_priv_type	ml_priv_type;
2330 
2331 	union {
2332 		struct pcpu_lstats __percpu		*lstats;
2333 		struct pcpu_sw_netstats __percpu	*tstats;
2334 		struct pcpu_dstats __percpu		*dstats;
2335 	};
2336 
2337 #if IS_ENABLED(CONFIG_GARP)
2338 	struct garp_port __rcu	*garp_port;
2339 #endif
2340 #if IS_ENABLED(CONFIG_MRP)
2341 	struct mrp_port __rcu	*mrp_port;
2342 #endif
2343 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2344 	struct dm_hw_stat_delta __rcu *dm_private;
2345 #endif
2346 	struct device		dev;
2347 	const struct attribute_group *sysfs_groups[4];
2348 	const struct attribute_group *sysfs_rx_queue_group;
2349 
2350 	const struct rtnl_link_ops *rtnl_link_ops;
2351 
2352 	/* for setting kernel sock attribute on TCP connection setup */
2353 #define GSO_MAX_SEGS		65535u
2354 #define GSO_LEGACY_MAX_SIZE	65536u
2355 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2356  * and shinfo->gso_segs is a 16bit field.
2357  */
2358 #define GSO_MAX_SIZE		(8 * GSO_MAX_SEGS)
2359 
2360 	unsigned int		gso_max_size;
2361 #define TSO_LEGACY_MAX_SIZE	65536
2362 #define TSO_MAX_SIZE		UINT_MAX
2363 	unsigned int		tso_max_size;
2364 	u16			gso_max_segs;
2365 #define TSO_MAX_SEGS		U16_MAX
2366 	u16			tso_max_segs;
2367 	unsigned int		gso_ipv4_max_size;
2368 
2369 #ifdef CONFIG_DCB
2370 	const struct dcbnl_rtnl_ops *dcbnl_ops;
2371 #endif
2372 	s16			num_tc;
2373 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
2374 	u8			prio_tc_map[TC_BITMASK + 1];
2375 
2376 #if IS_ENABLED(CONFIG_FCOE)
2377 	unsigned int		fcoe_ddp_xid;
2378 #endif
2379 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2380 	struct netprio_map __rcu *priomap;
2381 #endif
2382 	struct phy_device	*phydev;
2383 	struct sfp_bus		*sfp_bus;
2384 	struct lock_class_key	*qdisc_tx_busylock;
2385 	bool			proto_down;
2386 	unsigned		wol_enabled:1;
2387 	unsigned		threaded:1;
2388 
2389 	struct list_head	net_notifier_list;
2390 
2391 #if IS_ENABLED(CONFIG_MACSEC)
2392 	/* MACsec management functions */
2393 	const struct macsec_ops *macsec_ops;
2394 #endif
2395 	const struct udp_tunnel_nic_info	*udp_tunnel_nic_info;
2396 	struct udp_tunnel_nic	*udp_tunnel_nic;
2397 
2398 	/* protected by rtnl_lock */
2399 	struct bpf_xdp_entity	xdp_state[__MAX_XDP_MODE];
2400 
2401 	u8 dev_addr_shadow[MAX_ADDR_LEN];
2402 	netdevice_tracker	linkwatch_dev_tracker;
2403 	netdevice_tracker	watchdog_dev_tracker;
2404 	netdevice_tracker	dev_registered_tracker;
2405 	struct rtnl_hw_stats64	*offload_xstats_l3;
2406 
2407 	struct devlink_port	*devlink_port;
2408 };
2409 #define to_net_dev(d) container_of(d, struct net_device, dev)
2410 
2411 /*
2412  * Driver should use this to assign devlink port instance to a netdevice
2413  * before it registers the netdevice. Therefore devlink_port is static
2414  * during the netdev lifetime after it is registered.
2415  */
2416 #define SET_NETDEV_DEVLINK_PORT(dev, port)			\
2417 ({								\
2418 	WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED);	\
2419 	((dev)->devlink_port = (port));				\
2420 })
2421 
2422 static inline bool netif_elide_gro(const struct net_device *dev)
2423 {
2424 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2425 		return true;
2426 	return false;
2427 }
2428 
2429 #define	NETDEV_ALIGN		32
2430 
2431 static inline
2432 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2433 {
2434 	return dev->prio_tc_map[prio & TC_BITMASK];
2435 }
2436 
2437 static inline
2438 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2439 {
2440 	if (tc >= dev->num_tc)
2441 		return -EINVAL;
2442 
2443 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2444 	return 0;
2445 }
2446 
2447 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2448 void netdev_reset_tc(struct net_device *dev);
2449 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2450 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2451 
2452 static inline
2453 int netdev_get_num_tc(struct net_device *dev)
2454 {
2455 	return dev->num_tc;
2456 }
2457 
2458 static inline void net_prefetch(void *p)
2459 {
2460 	prefetch(p);
2461 #if L1_CACHE_BYTES < 128
2462 	prefetch((u8 *)p + L1_CACHE_BYTES);
2463 #endif
2464 }
2465 
2466 static inline void net_prefetchw(void *p)
2467 {
2468 	prefetchw(p);
2469 #if L1_CACHE_BYTES < 128
2470 	prefetchw((u8 *)p + L1_CACHE_BYTES);
2471 #endif
2472 }
2473 
2474 void netdev_unbind_sb_channel(struct net_device *dev,
2475 			      struct net_device *sb_dev);
2476 int netdev_bind_sb_channel_queue(struct net_device *dev,
2477 				 struct net_device *sb_dev,
2478 				 u8 tc, u16 count, u16 offset);
2479 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2480 static inline int netdev_get_sb_channel(struct net_device *dev)
2481 {
2482 	return max_t(int, -dev->num_tc, 0);
2483 }
2484 
2485 static inline
2486 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2487 					 unsigned int index)
2488 {
2489 	DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2490 	return &dev->_tx[index];
2491 }
2492 
2493 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2494 						    const struct sk_buff *skb)
2495 {
2496 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2497 }
2498 
2499 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2500 					    void (*f)(struct net_device *,
2501 						      struct netdev_queue *,
2502 						      void *),
2503 					    void *arg)
2504 {
2505 	unsigned int i;
2506 
2507 	for (i = 0; i < dev->num_tx_queues; i++)
2508 		f(dev, &dev->_tx[i], arg);
2509 }
2510 
2511 #define netdev_lockdep_set_classes(dev)				\
2512 {								\
2513 	static struct lock_class_key qdisc_tx_busylock_key;	\
2514 	static struct lock_class_key qdisc_xmit_lock_key;	\
2515 	static struct lock_class_key dev_addr_list_lock_key;	\
2516 	unsigned int i;						\
2517 								\
2518 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
2519 	lockdep_set_class(&(dev)->addr_list_lock,		\
2520 			  &dev_addr_list_lock_key);		\
2521 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
2522 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
2523 				  &qdisc_xmit_lock_key);	\
2524 }
2525 
2526 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2527 		     struct net_device *sb_dev);
2528 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2529 					 struct sk_buff *skb,
2530 					 struct net_device *sb_dev);
2531 
2532 /* returns the headroom that the master device needs to take in account
2533  * when forwarding to this dev
2534  */
2535 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2536 {
2537 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2538 }
2539 
2540 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2541 {
2542 	if (dev->netdev_ops->ndo_set_rx_headroom)
2543 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2544 }
2545 
2546 /* set the device rx headroom to the dev's default */
2547 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2548 {
2549 	netdev_set_rx_headroom(dev, -1);
2550 }
2551 
2552 static inline void *netdev_get_ml_priv(struct net_device *dev,
2553 				       enum netdev_ml_priv_type type)
2554 {
2555 	if (dev->ml_priv_type != type)
2556 		return NULL;
2557 
2558 	return dev->ml_priv;
2559 }
2560 
2561 static inline void netdev_set_ml_priv(struct net_device *dev,
2562 				      void *ml_priv,
2563 				      enum netdev_ml_priv_type type)
2564 {
2565 	WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2566 	     "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2567 	     dev->ml_priv_type, type);
2568 	WARN(!dev->ml_priv_type && dev->ml_priv,
2569 	     "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2570 
2571 	dev->ml_priv = ml_priv;
2572 	dev->ml_priv_type = type;
2573 }
2574 
2575 /*
2576  * Net namespace inlines
2577  */
2578 static inline
2579 struct net *dev_net(const struct net_device *dev)
2580 {
2581 	return read_pnet(&dev->nd_net);
2582 }
2583 
2584 static inline
2585 void dev_net_set(struct net_device *dev, struct net *net)
2586 {
2587 	write_pnet(&dev->nd_net, net);
2588 }
2589 
2590 /**
2591  *	netdev_priv - access network device private data
2592  *	@dev: network device
2593  *
2594  * Get network device private data
2595  */
2596 static inline void *netdev_priv(const struct net_device *dev)
2597 {
2598 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2599 }
2600 
2601 /* Set the sysfs physical device reference for the network logical device
2602  * if set prior to registration will cause a symlink during initialization.
2603  */
2604 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2605 
2606 /* Set the sysfs device type for the network logical device to allow
2607  * fine-grained identification of different network device types. For
2608  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2609  */
2610 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2611 
2612 /* Default NAPI poll() weight
2613  * Device drivers are strongly advised to not use bigger value
2614  */
2615 #define NAPI_POLL_WEIGHT 64
2616 
2617 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2618 			   int (*poll)(struct napi_struct *, int), int weight);
2619 
2620 /**
2621  * netif_napi_add() - initialize a NAPI context
2622  * @dev:  network device
2623  * @napi: NAPI context
2624  * @poll: polling function
2625  *
2626  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2627  * *any* of the other NAPI-related functions.
2628  */
2629 static inline void
2630 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2631 	       int (*poll)(struct napi_struct *, int))
2632 {
2633 	netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2634 }
2635 
2636 static inline void
2637 netif_napi_add_tx_weight(struct net_device *dev,
2638 			 struct napi_struct *napi,
2639 			 int (*poll)(struct napi_struct *, int),
2640 			 int weight)
2641 {
2642 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2643 	netif_napi_add_weight(dev, napi, poll, weight);
2644 }
2645 
2646 /**
2647  * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2648  * @dev:  network device
2649  * @napi: NAPI context
2650  * @poll: polling function
2651  *
2652  * This variant of netif_napi_add() should be used from drivers using NAPI
2653  * to exclusively poll a TX queue.
2654  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2655  */
2656 static inline void netif_napi_add_tx(struct net_device *dev,
2657 				     struct napi_struct *napi,
2658 				     int (*poll)(struct napi_struct *, int))
2659 {
2660 	netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2661 }
2662 
2663 /**
2664  *  __netif_napi_del - remove a NAPI context
2665  *  @napi: NAPI context
2666  *
2667  * Warning: caller must observe RCU grace period before freeing memory
2668  * containing @napi. Drivers might want to call this helper to combine
2669  * all the needed RCU grace periods into a single one.
2670  */
2671 void __netif_napi_del(struct napi_struct *napi);
2672 
2673 /**
2674  *  netif_napi_del - remove a NAPI context
2675  *  @napi: NAPI context
2676  *
2677  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2678  */
2679 static inline void netif_napi_del(struct napi_struct *napi)
2680 {
2681 	__netif_napi_del(napi);
2682 	synchronize_net();
2683 }
2684 
2685 struct packet_type {
2686 	__be16			type;	/* This is really htons(ether_type). */
2687 	bool			ignore_outgoing;
2688 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2689 	netdevice_tracker	dev_tracker;
2690 	int			(*func) (struct sk_buff *,
2691 					 struct net_device *,
2692 					 struct packet_type *,
2693 					 struct net_device *);
2694 	void			(*list_func) (struct list_head *,
2695 					      struct packet_type *,
2696 					      struct net_device *);
2697 	bool			(*id_match)(struct packet_type *ptype,
2698 					    struct sock *sk);
2699 	struct net		*af_packet_net;
2700 	void			*af_packet_priv;
2701 	struct list_head	list;
2702 };
2703 
2704 struct offload_callbacks {
2705 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2706 						netdev_features_t features);
2707 	struct sk_buff		*(*gro_receive)(struct list_head *head,
2708 						struct sk_buff *skb);
2709 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2710 };
2711 
2712 struct packet_offload {
2713 	__be16			 type;	/* This is really htons(ether_type). */
2714 	u16			 priority;
2715 	struct offload_callbacks callbacks;
2716 	struct list_head	 list;
2717 };
2718 
2719 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2720 struct pcpu_sw_netstats {
2721 	u64_stats_t		rx_packets;
2722 	u64_stats_t		rx_bytes;
2723 	u64_stats_t		tx_packets;
2724 	u64_stats_t		tx_bytes;
2725 	struct u64_stats_sync   syncp;
2726 } __aligned(4 * sizeof(u64));
2727 
2728 struct pcpu_lstats {
2729 	u64_stats_t packets;
2730 	u64_stats_t bytes;
2731 	struct u64_stats_sync syncp;
2732 } __aligned(2 * sizeof(u64));
2733 
2734 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2735 
2736 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2737 {
2738 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2739 
2740 	u64_stats_update_begin(&tstats->syncp);
2741 	u64_stats_add(&tstats->rx_bytes, len);
2742 	u64_stats_inc(&tstats->rx_packets);
2743 	u64_stats_update_end(&tstats->syncp);
2744 }
2745 
2746 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2747 					  unsigned int packets,
2748 					  unsigned int len)
2749 {
2750 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2751 
2752 	u64_stats_update_begin(&tstats->syncp);
2753 	u64_stats_add(&tstats->tx_bytes, len);
2754 	u64_stats_add(&tstats->tx_packets, packets);
2755 	u64_stats_update_end(&tstats->syncp);
2756 }
2757 
2758 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2759 {
2760 	struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2761 
2762 	u64_stats_update_begin(&lstats->syncp);
2763 	u64_stats_add(&lstats->bytes, len);
2764 	u64_stats_inc(&lstats->packets);
2765 	u64_stats_update_end(&lstats->syncp);
2766 }
2767 
2768 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2769 ({									\
2770 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2771 	if (pcpu_stats)	{						\
2772 		int __cpu;						\
2773 		for_each_possible_cpu(__cpu) {				\
2774 			typeof(type) *stat;				\
2775 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2776 			u64_stats_init(&stat->syncp);			\
2777 		}							\
2778 	}								\
2779 	pcpu_stats;							\
2780 })
2781 
2782 #define netdev_alloc_pcpu_stats(type)					\
2783 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2784 
2785 #define devm_netdev_alloc_pcpu_stats(dev, type)				\
2786 ({									\
2787 	typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2788 	if (pcpu_stats) {						\
2789 		int __cpu;						\
2790 		for_each_possible_cpu(__cpu) {				\
2791 			typeof(type) *stat;				\
2792 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2793 			u64_stats_init(&stat->syncp);			\
2794 		}							\
2795 	}								\
2796 	pcpu_stats;							\
2797 })
2798 
2799 enum netdev_lag_tx_type {
2800 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2801 	NETDEV_LAG_TX_TYPE_RANDOM,
2802 	NETDEV_LAG_TX_TYPE_BROADCAST,
2803 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2804 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2805 	NETDEV_LAG_TX_TYPE_HASH,
2806 };
2807 
2808 enum netdev_lag_hash {
2809 	NETDEV_LAG_HASH_NONE,
2810 	NETDEV_LAG_HASH_L2,
2811 	NETDEV_LAG_HASH_L34,
2812 	NETDEV_LAG_HASH_L23,
2813 	NETDEV_LAG_HASH_E23,
2814 	NETDEV_LAG_HASH_E34,
2815 	NETDEV_LAG_HASH_VLAN_SRCMAC,
2816 	NETDEV_LAG_HASH_UNKNOWN,
2817 };
2818 
2819 struct netdev_lag_upper_info {
2820 	enum netdev_lag_tx_type tx_type;
2821 	enum netdev_lag_hash hash_type;
2822 };
2823 
2824 struct netdev_lag_lower_state_info {
2825 	u8 link_up : 1,
2826 	   tx_enabled : 1;
2827 };
2828 
2829 #include <linux/notifier.h>
2830 
2831 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2832  * and the rtnetlink notification exclusion list in rtnetlink_event() when
2833  * adding new types.
2834  */
2835 enum netdev_cmd {
2836 	NETDEV_UP	= 1,	/* For now you can't veto a device up/down */
2837 	NETDEV_DOWN,
2838 	NETDEV_REBOOT,		/* Tell a protocol stack a network interface
2839 				   detected a hardware crash and restarted
2840 				   - we can use this eg to kick tcp sessions
2841 				   once done */
2842 	NETDEV_CHANGE,		/* Notify device state change */
2843 	NETDEV_REGISTER,
2844 	NETDEV_UNREGISTER,
2845 	NETDEV_CHANGEMTU,	/* notify after mtu change happened */
2846 	NETDEV_CHANGEADDR,	/* notify after the address change */
2847 	NETDEV_PRE_CHANGEADDR,	/* notify before the address change */
2848 	NETDEV_GOING_DOWN,
2849 	NETDEV_CHANGENAME,
2850 	NETDEV_FEAT_CHANGE,
2851 	NETDEV_BONDING_FAILOVER,
2852 	NETDEV_PRE_UP,
2853 	NETDEV_PRE_TYPE_CHANGE,
2854 	NETDEV_POST_TYPE_CHANGE,
2855 	NETDEV_POST_INIT,
2856 	NETDEV_PRE_UNINIT,
2857 	NETDEV_RELEASE,
2858 	NETDEV_NOTIFY_PEERS,
2859 	NETDEV_JOIN,
2860 	NETDEV_CHANGEUPPER,
2861 	NETDEV_RESEND_IGMP,
2862 	NETDEV_PRECHANGEMTU,	/* notify before mtu change happened */
2863 	NETDEV_CHANGEINFODATA,
2864 	NETDEV_BONDING_INFO,
2865 	NETDEV_PRECHANGEUPPER,
2866 	NETDEV_CHANGELOWERSTATE,
2867 	NETDEV_UDP_TUNNEL_PUSH_INFO,
2868 	NETDEV_UDP_TUNNEL_DROP_INFO,
2869 	NETDEV_CHANGE_TX_QUEUE_LEN,
2870 	NETDEV_CVLAN_FILTER_PUSH_INFO,
2871 	NETDEV_CVLAN_FILTER_DROP_INFO,
2872 	NETDEV_SVLAN_FILTER_PUSH_INFO,
2873 	NETDEV_SVLAN_FILTER_DROP_INFO,
2874 	NETDEV_OFFLOAD_XSTATS_ENABLE,
2875 	NETDEV_OFFLOAD_XSTATS_DISABLE,
2876 	NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2877 	NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2878 	NETDEV_XDP_FEAT_CHANGE,
2879 };
2880 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2881 
2882 int register_netdevice_notifier(struct notifier_block *nb);
2883 int unregister_netdevice_notifier(struct notifier_block *nb);
2884 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2885 int unregister_netdevice_notifier_net(struct net *net,
2886 				      struct notifier_block *nb);
2887 int register_netdevice_notifier_dev_net(struct net_device *dev,
2888 					struct notifier_block *nb,
2889 					struct netdev_net_notifier *nn);
2890 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2891 					  struct notifier_block *nb,
2892 					  struct netdev_net_notifier *nn);
2893 
2894 struct netdev_notifier_info {
2895 	struct net_device	*dev;
2896 	struct netlink_ext_ack	*extack;
2897 };
2898 
2899 struct netdev_notifier_info_ext {
2900 	struct netdev_notifier_info info; /* must be first */
2901 	union {
2902 		u32 mtu;
2903 	} ext;
2904 };
2905 
2906 struct netdev_notifier_change_info {
2907 	struct netdev_notifier_info info; /* must be first */
2908 	unsigned int flags_changed;
2909 };
2910 
2911 struct netdev_notifier_changeupper_info {
2912 	struct netdev_notifier_info info; /* must be first */
2913 	struct net_device *upper_dev; /* new upper dev */
2914 	bool master; /* is upper dev master */
2915 	bool linking; /* is the notification for link or unlink */
2916 	void *upper_info; /* upper dev info */
2917 };
2918 
2919 struct netdev_notifier_changelowerstate_info {
2920 	struct netdev_notifier_info info; /* must be first */
2921 	void *lower_state_info; /* is lower dev state */
2922 };
2923 
2924 struct netdev_notifier_pre_changeaddr_info {
2925 	struct netdev_notifier_info info; /* must be first */
2926 	const unsigned char *dev_addr;
2927 };
2928 
2929 enum netdev_offload_xstats_type {
2930 	NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2931 };
2932 
2933 struct netdev_notifier_offload_xstats_info {
2934 	struct netdev_notifier_info info; /* must be first */
2935 	enum netdev_offload_xstats_type type;
2936 
2937 	union {
2938 		/* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2939 		struct netdev_notifier_offload_xstats_rd *report_delta;
2940 		/* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2941 		struct netdev_notifier_offload_xstats_ru *report_used;
2942 	};
2943 };
2944 
2945 int netdev_offload_xstats_enable(struct net_device *dev,
2946 				 enum netdev_offload_xstats_type type,
2947 				 struct netlink_ext_ack *extack);
2948 int netdev_offload_xstats_disable(struct net_device *dev,
2949 				  enum netdev_offload_xstats_type type);
2950 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2951 				   enum netdev_offload_xstats_type type);
2952 int netdev_offload_xstats_get(struct net_device *dev,
2953 			      enum netdev_offload_xstats_type type,
2954 			      struct rtnl_hw_stats64 *stats, bool *used,
2955 			      struct netlink_ext_ack *extack);
2956 void
2957 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2958 				   const struct rtnl_hw_stats64 *stats);
2959 void
2960 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2961 void netdev_offload_xstats_push_delta(struct net_device *dev,
2962 				      enum netdev_offload_xstats_type type,
2963 				      const struct rtnl_hw_stats64 *stats);
2964 
2965 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2966 					     struct net_device *dev)
2967 {
2968 	info->dev = dev;
2969 	info->extack = NULL;
2970 }
2971 
2972 static inline struct net_device *
2973 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2974 {
2975 	return info->dev;
2976 }
2977 
2978 static inline struct netlink_ext_ack *
2979 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2980 {
2981 	return info->extack;
2982 }
2983 
2984 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2985 int call_netdevice_notifiers_info(unsigned long val,
2986 				  struct netdev_notifier_info *info);
2987 
2988 extern rwlock_t				dev_base_lock;		/* Device list lock */
2989 
2990 #define for_each_netdev(net, d)		\
2991 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2992 #define for_each_netdev_reverse(net, d)	\
2993 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2994 #define for_each_netdev_rcu(net, d)		\
2995 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2996 #define for_each_netdev_safe(net, d, n)	\
2997 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2998 #define for_each_netdev_continue(net, d)		\
2999 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3000 #define for_each_netdev_continue_reverse(net, d)		\
3001 		list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3002 						     dev_list)
3003 #define for_each_netdev_continue_rcu(net, d)		\
3004 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3005 #define for_each_netdev_in_bond_rcu(bond, slave)	\
3006 		for_each_netdev_rcu(&init_net, slave)	\
3007 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3008 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
3009 
3010 #define for_each_netdev_dump(net, d, ifindex)				\
3011 	xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3012 
3013 static inline struct net_device *next_net_device(struct net_device *dev)
3014 {
3015 	struct list_head *lh;
3016 	struct net *net;
3017 
3018 	net = dev_net(dev);
3019 	lh = dev->dev_list.next;
3020 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3021 }
3022 
3023 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3024 {
3025 	struct list_head *lh;
3026 	struct net *net;
3027 
3028 	net = dev_net(dev);
3029 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3030 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3031 }
3032 
3033 static inline struct net_device *first_net_device(struct net *net)
3034 {
3035 	return list_empty(&net->dev_base_head) ? NULL :
3036 		net_device_entry(net->dev_base_head.next);
3037 }
3038 
3039 static inline struct net_device *first_net_device_rcu(struct net *net)
3040 {
3041 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3042 
3043 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3044 }
3045 
3046 int netdev_boot_setup_check(struct net_device *dev);
3047 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3048 				       const char *hwaddr);
3049 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3050 void dev_add_pack(struct packet_type *pt);
3051 void dev_remove_pack(struct packet_type *pt);
3052 void __dev_remove_pack(struct packet_type *pt);
3053 void dev_add_offload(struct packet_offload *po);
3054 void dev_remove_offload(struct packet_offload *po);
3055 
3056 int dev_get_iflink(const struct net_device *dev);
3057 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3058 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3059 			  struct net_device_path_stack *stack);
3060 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3061 				      unsigned short mask);
3062 struct net_device *dev_get_by_name(struct net *net, const char *name);
3063 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3064 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3065 bool netdev_name_in_use(struct net *net, const char *name);
3066 int dev_alloc_name(struct net_device *dev, const char *name);
3067 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3068 void dev_close(struct net_device *dev);
3069 void dev_close_many(struct list_head *head, bool unlink);
3070 void dev_disable_lro(struct net_device *dev);
3071 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3072 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3073 		     struct net_device *sb_dev);
3074 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3075 		       struct net_device *sb_dev);
3076 
3077 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3078 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3079 
3080 static inline int dev_queue_xmit(struct sk_buff *skb)
3081 {
3082 	return __dev_queue_xmit(skb, NULL);
3083 }
3084 
3085 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3086 				       struct net_device *sb_dev)
3087 {
3088 	return __dev_queue_xmit(skb, sb_dev);
3089 }
3090 
3091 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3092 {
3093 	int ret;
3094 
3095 	ret = __dev_direct_xmit(skb, queue_id);
3096 	if (!dev_xmit_complete(ret))
3097 		kfree_skb(skb);
3098 	return ret;
3099 }
3100 
3101 int register_netdevice(struct net_device *dev);
3102 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3103 void unregister_netdevice_many(struct list_head *head);
3104 static inline void unregister_netdevice(struct net_device *dev)
3105 {
3106 	unregister_netdevice_queue(dev, NULL);
3107 }
3108 
3109 int netdev_refcnt_read(const struct net_device *dev);
3110 void free_netdev(struct net_device *dev);
3111 void netdev_freemem(struct net_device *dev);
3112 int init_dummy_netdev(struct net_device *dev);
3113 
3114 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3115 					 struct sk_buff *skb,
3116 					 bool all_slaves);
3117 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3118 					    struct sock *sk);
3119 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3120 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3121 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3122 				       netdevice_tracker *tracker, gfp_t gfp);
3123 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3124 				      netdevice_tracker *tracker, gfp_t gfp);
3125 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3126 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3127 
3128 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3129 				  unsigned short type,
3130 				  const void *daddr, const void *saddr,
3131 				  unsigned int len)
3132 {
3133 	if (!dev->header_ops || !dev->header_ops->create)
3134 		return 0;
3135 
3136 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3137 }
3138 
3139 static inline int dev_parse_header(const struct sk_buff *skb,
3140 				   unsigned char *haddr)
3141 {
3142 	const struct net_device *dev = skb->dev;
3143 
3144 	if (!dev->header_ops || !dev->header_ops->parse)
3145 		return 0;
3146 	return dev->header_ops->parse(skb, haddr);
3147 }
3148 
3149 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3150 {
3151 	const struct net_device *dev = skb->dev;
3152 
3153 	if (!dev->header_ops || !dev->header_ops->parse_protocol)
3154 		return 0;
3155 	return dev->header_ops->parse_protocol(skb);
3156 }
3157 
3158 /* ll_header must have at least hard_header_len allocated */
3159 static inline bool dev_validate_header(const struct net_device *dev,
3160 				       char *ll_header, int len)
3161 {
3162 	if (likely(len >= dev->hard_header_len))
3163 		return true;
3164 	if (len < dev->min_header_len)
3165 		return false;
3166 
3167 	if (capable(CAP_SYS_RAWIO)) {
3168 		memset(ll_header + len, 0, dev->hard_header_len - len);
3169 		return true;
3170 	}
3171 
3172 	if (dev->header_ops && dev->header_ops->validate)
3173 		return dev->header_ops->validate(ll_header, len);
3174 
3175 	return false;
3176 }
3177 
3178 static inline bool dev_has_header(const struct net_device *dev)
3179 {
3180 	return dev->header_ops && dev->header_ops->create;
3181 }
3182 
3183 /*
3184  * Incoming packets are placed on per-CPU queues
3185  */
3186 struct softnet_data {
3187 	struct list_head	poll_list;
3188 	struct sk_buff_head	process_queue;
3189 
3190 	/* stats */
3191 	unsigned int		processed;
3192 	unsigned int		time_squeeze;
3193 #ifdef CONFIG_RPS
3194 	struct softnet_data	*rps_ipi_list;
3195 #endif
3196 
3197 	bool			in_net_rx_action;
3198 	bool			in_napi_threaded_poll;
3199 
3200 #ifdef CONFIG_NET_FLOW_LIMIT
3201 	struct sd_flow_limit __rcu *flow_limit;
3202 #endif
3203 	struct Qdisc		*output_queue;
3204 	struct Qdisc		**output_queue_tailp;
3205 	struct sk_buff		*completion_queue;
3206 #ifdef CONFIG_XFRM_OFFLOAD
3207 	struct sk_buff_head	xfrm_backlog;
3208 #endif
3209 	/* written and read only by owning cpu: */
3210 	struct {
3211 		u16 recursion;
3212 		u8  more;
3213 #ifdef CONFIG_NET_EGRESS
3214 		u8  skip_txqueue;
3215 #endif
3216 	} xmit;
3217 #ifdef CONFIG_RPS
3218 	/* input_queue_head should be written by cpu owning this struct,
3219 	 * and only read by other cpus. Worth using a cache line.
3220 	 */
3221 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
3222 
3223 	/* Elements below can be accessed between CPUs for RPS/RFS */
3224 	call_single_data_t	csd ____cacheline_aligned_in_smp;
3225 	struct softnet_data	*rps_ipi_next;
3226 	unsigned int		cpu;
3227 	unsigned int		input_queue_tail;
3228 #endif
3229 	unsigned int		received_rps;
3230 	unsigned int		dropped;
3231 	struct sk_buff_head	input_pkt_queue;
3232 	struct napi_struct	backlog;
3233 
3234 	/* Another possibly contended cache line */
3235 	spinlock_t		defer_lock ____cacheline_aligned_in_smp;
3236 	int			defer_count;
3237 	int			defer_ipi_scheduled;
3238 	struct sk_buff		*defer_list;
3239 	call_single_data_t	defer_csd;
3240 };
3241 
3242 static inline void input_queue_head_incr(struct softnet_data *sd)
3243 {
3244 #ifdef CONFIG_RPS
3245 	sd->input_queue_head++;
3246 #endif
3247 }
3248 
3249 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3250 					      unsigned int *qtail)
3251 {
3252 #ifdef CONFIG_RPS
3253 	*qtail = ++sd->input_queue_tail;
3254 #endif
3255 }
3256 
3257 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3258 
3259 static inline int dev_recursion_level(void)
3260 {
3261 	return this_cpu_read(softnet_data.xmit.recursion);
3262 }
3263 
3264 #define XMIT_RECURSION_LIMIT	8
3265 static inline bool dev_xmit_recursion(void)
3266 {
3267 	return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3268 			XMIT_RECURSION_LIMIT);
3269 }
3270 
3271 static inline void dev_xmit_recursion_inc(void)
3272 {
3273 	__this_cpu_inc(softnet_data.xmit.recursion);
3274 }
3275 
3276 static inline void dev_xmit_recursion_dec(void)
3277 {
3278 	__this_cpu_dec(softnet_data.xmit.recursion);
3279 }
3280 
3281 void __netif_schedule(struct Qdisc *q);
3282 void netif_schedule_queue(struct netdev_queue *txq);
3283 
3284 static inline void netif_tx_schedule_all(struct net_device *dev)
3285 {
3286 	unsigned int i;
3287 
3288 	for (i = 0; i < dev->num_tx_queues; i++)
3289 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
3290 }
3291 
3292 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3293 {
3294 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3295 }
3296 
3297 /**
3298  *	netif_start_queue - allow transmit
3299  *	@dev: network device
3300  *
3301  *	Allow upper layers to call the device hard_start_xmit routine.
3302  */
3303 static inline void netif_start_queue(struct net_device *dev)
3304 {
3305 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3306 }
3307 
3308 static inline void netif_tx_start_all_queues(struct net_device *dev)
3309 {
3310 	unsigned int i;
3311 
3312 	for (i = 0; i < dev->num_tx_queues; i++) {
3313 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3314 		netif_tx_start_queue(txq);
3315 	}
3316 }
3317 
3318 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3319 
3320 /**
3321  *	netif_wake_queue - restart transmit
3322  *	@dev: network device
3323  *
3324  *	Allow upper layers to call the device hard_start_xmit routine.
3325  *	Used for flow control when transmit resources are available.
3326  */
3327 static inline void netif_wake_queue(struct net_device *dev)
3328 {
3329 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3330 }
3331 
3332 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3333 {
3334 	unsigned int i;
3335 
3336 	for (i = 0; i < dev->num_tx_queues; i++) {
3337 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3338 		netif_tx_wake_queue(txq);
3339 	}
3340 }
3341 
3342 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3343 {
3344 	/* Must be an atomic op see netif_txq_try_stop() */
3345 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3346 }
3347 
3348 /**
3349  *	netif_stop_queue - stop transmitted packets
3350  *	@dev: network device
3351  *
3352  *	Stop upper layers calling the device hard_start_xmit routine.
3353  *	Used for flow control when transmit resources are unavailable.
3354  */
3355 static inline void netif_stop_queue(struct net_device *dev)
3356 {
3357 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3358 }
3359 
3360 void netif_tx_stop_all_queues(struct net_device *dev);
3361 
3362 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3363 {
3364 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3365 }
3366 
3367 /**
3368  *	netif_queue_stopped - test if transmit queue is flowblocked
3369  *	@dev: network device
3370  *
3371  *	Test if transmit queue on device is currently unable to send.
3372  */
3373 static inline bool netif_queue_stopped(const struct net_device *dev)
3374 {
3375 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3376 }
3377 
3378 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3379 {
3380 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3381 }
3382 
3383 static inline bool
3384 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3385 {
3386 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3387 }
3388 
3389 static inline bool
3390 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3391 {
3392 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3393 }
3394 
3395 /**
3396  *	netdev_queue_set_dql_min_limit - set dql minimum limit
3397  *	@dev_queue: pointer to transmit queue
3398  *	@min_limit: dql minimum limit
3399  *
3400  * Forces xmit_more() to return true until the minimum threshold
3401  * defined by @min_limit is reached (or until the tx queue is
3402  * empty). Warning: to be use with care, misuse will impact the
3403  * latency.
3404  */
3405 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3406 						  unsigned int min_limit)
3407 {
3408 #ifdef CONFIG_BQL
3409 	dev_queue->dql.min_limit = min_limit;
3410 #endif
3411 }
3412 
3413 /**
3414  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3415  *	@dev_queue: pointer to transmit queue
3416  *
3417  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3418  * to give appropriate hint to the CPU.
3419  */
3420 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3421 {
3422 #ifdef CONFIG_BQL
3423 	prefetchw(&dev_queue->dql.num_queued);
3424 #endif
3425 }
3426 
3427 /**
3428  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3429  *	@dev_queue: pointer to transmit queue
3430  *
3431  * BQL enabled drivers might use this helper in their TX completion path,
3432  * to give appropriate hint to the CPU.
3433  */
3434 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3435 {
3436 #ifdef CONFIG_BQL
3437 	prefetchw(&dev_queue->dql.limit);
3438 #endif
3439 }
3440 
3441 /**
3442  *	netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3443  *	@dev_queue: network device queue
3444  *	@bytes: number of bytes queued to the device queue
3445  *
3446  *	Report the number of bytes queued for sending/completion to the network
3447  *	device hardware queue. @bytes should be a good approximation and should
3448  *	exactly match netdev_completed_queue() @bytes.
3449  *	This is typically called once per packet, from ndo_start_xmit().
3450  */
3451 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3452 					unsigned int bytes)
3453 {
3454 #ifdef CONFIG_BQL
3455 	dql_queued(&dev_queue->dql, bytes);
3456 
3457 	if (likely(dql_avail(&dev_queue->dql) >= 0))
3458 		return;
3459 
3460 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3461 
3462 	/*
3463 	 * The XOFF flag must be set before checking the dql_avail below,
3464 	 * because in netdev_tx_completed_queue we update the dql_completed
3465 	 * before checking the XOFF flag.
3466 	 */
3467 	smp_mb();
3468 
3469 	/* check again in case another CPU has just made room avail */
3470 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3471 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3472 #endif
3473 }
3474 
3475 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3476  * that they should not test BQL status themselves.
3477  * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3478  * skb of a batch.
3479  * Returns true if the doorbell must be used to kick the NIC.
3480  */
3481 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3482 					  unsigned int bytes,
3483 					  bool xmit_more)
3484 {
3485 	if (xmit_more) {
3486 #ifdef CONFIG_BQL
3487 		dql_queued(&dev_queue->dql, bytes);
3488 #endif
3489 		return netif_tx_queue_stopped(dev_queue);
3490 	}
3491 	netdev_tx_sent_queue(dev_queue, bytes);
3492 	return true;
3493 }
3494 
3495 /**
3496  *	netdev_sent_queue - report the number of bytes queued to hardware
3497  *	@dev: network device
3498  *	@bytes: number of bytes queued to the hardware device queue
3499  *
3500  *	Report the number of bytes queued for sending/completion to the network
3501  *	device hardware queue#0. @bytes should be a good approximation and should
3502  *	exactly match netdev_completed_queue() @bytes.
3503  *	This is typically called once per packet, from ndo_start_xmit().
3504  */
3505 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3506 {
3507 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3508 }
3509 
3510 static inline bool __netdev_sent_queue(struct net_device *dev,
3511 				       unsigned int bytes,
3512 				       bool xmit_more)
3513 {
3514 	return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3515 				      xmit_more);
3516 }
3517 
3518 /**
3519  *	netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3520  *	@dev_queue: network device queue
3521  *	@pkts: number of packets (currently ignored)
3522  *	@bytes: number of bytes dequeued from the device queue
3523  *
3524  *	Must be called at most once per TX completion round (and not per
3525  *	individual packet), so that BQL can adjust its limits appropriately.
3526  */
3527 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3528 					     unsigned int pkts, unsigned int bytes)
3529 {
3530 #ifdef CONFIG_BQL
3531 	if (unlikely(!bytes))
3532 		return;
3533 
3534 	dql_completed(&dev_queue->dql, bytes);
3535 
3536 	/*
3537 	 * Without the memory barrier there is a small possiblity that
3538 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3539 	 * be stopped forever
3540 	 */
3541 	smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3542 
3543 	if (unlikely(dql_avail(&dev_queue->dql) < 0))
3544 		return;
3545 
3546 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3547 		netif_schedule_queue(dev_queue);
3548 #endif
3549 }
3550 
3551 /**
3552  * 	netdev_completed_queue - report bytes and packets completed by device
3553  * 	@dev: network device
3554  * 	@pkts: actual number of packets sent over the medium
3555  * 	@bytes: actual number of bytes sent over the medium
3556  *
3557  * 	Report the number of bytes and packets transmitted by the network device
3558  * 	hardware queue over the physical medium, @bytes must exactly match the
3559  * 	@bytes amount passed to netdev_sent_queue()
3560  */
3561 static inline void netdev_completed_queue(struct net_device *dev,
3562 					  unsigned int pkts, unsigned int bytes)
3563 {
3564 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3565 }
3566 
3567 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3568 {
3569 #ifdef CONFIG_BQL
3570 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3571 	dql_reset(&q->dql);
3572 #endif
3573 }
3574 
3575 /**
3576  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3577  * 	@dev_queue: network device
3578  *
3579  * 	Reset the bytes and packet count of a network device and clear the
3580  * 	software flow control OFF bit for this network device
3581  */
3582 static inline void netdev_reset_queue(struct net_device *dev_queue)
3583 {
3584 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3585 }
3586 
3587 /**
3588  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3589  * 	@dev: network device
3590  * 	@queue_index: given tx queue index
3591  *
3592  * 	Returns 0 if given tx queue index >= number of device tx queues,
3593  * 	otherwise returns the originally passed tx queue index.
3594  */
3595 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3596 {
3597 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3598 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3599 				     dev->name, queue_index,
3600 				     dev->real_num_tx_queues);
3601 		return 0;
3602 	}
3603 
3604 	return queue_index;
3605 }
3606 
3607 /**
3608  *	netif_running - test if up
3609  *	@dev: network device
3610  *
3611  *	Test if the device has been brought up.
3612  */
3613 static inline bool netif_running(const struct net_device *dev)
3614 {
3615 	return test_bit(__LINK_STATE_START, &dev->state);
3616 }
3617 
3618 /*
3619  * Routines to manage the subqueues on a device.  We only need start,
3620  * stop, and a check if it's stopped.  All other device management is
3621  * done at the overall netdevice level.
3622  * Also test the device if we're multiqueue.
3623  */
3624 
3625 /**
3626  *	netif_start_subqueue - allow sending packets on subqueue
3627  *	@dev: network device
3628  *	@queue_index: sub queue index
3629  *
3630  * Start individual transmit queue of a device with multiple transmit queues.
3631  */
3632 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3633 {
3634 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3635 
3636 	netif_tx_start_queue(txq);
3637 }
3638 
3639 /**
3640  *	netif_stop_subqueue - stop sending packets on subqueue
3641  *	@dev: network device
3642  *	@queue_index: sub queue index
3643  *
3644  * Stop individual transmit queue of a device with multiple transmit queues.
3645  */
3646 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3647 {
3648 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3649 	netif_tx_stop_queue(txq);
3650 }
3651 
3652 /**
3653  *	__netif_subqueue_stopped - test status of subqueue
3654  *	@dev: network device
3655  *	@queue_index: sub queue index
3656  *
3657  * Check individual transmit queue of a device with multiple transmit queues.
3658  */
3659 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3660 					    u16 queue_index)
3661 {
3662 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3663 
3664 	return netif_tx_queue_stopped(txq);
3665 }
3666 
3667 /**
3668  *	netif_subqueue_stopped - test status of subqueue
3669  *	@dev: network device
3670  *	@skb: sub queue buffer pointer
3671  *
3672  * Check individual transmit queue of a device with multiple transmit queues.
3673  */
3674 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3675 					  struct sk_buff *skb)
3676 {
3677 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3678 }
3679 
3680 /**
3681  *	netif_wake_subqueue - allow sending packets on subqueue
3682  *	@dev: network device
3683  *	@queue_index: sub queue index
3684  *
3685  * Resume individual transmit queue of a device with multiple transmit queues.
3686  */
3687 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3688 {
3689 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3690 
3691 	netif_tx_wake_queue(txq);
3692 }
3693 
3694 #ifdef CONFIG_XPS
3695 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3696 			u16 index);
3697 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3698 			  u16 index, enum xps_map_type type);
3699 
3700 /**
3701  *	netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3702  *	@j: CPU/Rx queue index
3703  *	@mask: bitmask of all cpus/rx queues
3704  *	@nr_bits: number of bits in the bitmask
3705  *
3706  * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3707  */
3708 static inline bool netif_attr_test_mask(unsigned long j,
3709 					const unsigned long *mask,
3710 					unsigned int nr_bits)
3711 {
3712 	cpu_max_bits_warn(j, nr_bits);
3713 	return test_bit(j, mask);
3714 }
3715 
3716 /**
3717  *	netif_attr_test_online - Test for online CPU/Rx queue
3718  *	@j: CPU/Rx queue index
3719  *	@online_mask: bitmask for CPUs/Rx queues that are online
3720  *	@nr_bits: number of bits in the bitmask
3721  *
3722  * Returns true if a CPU/Rx queue is online.
3723  */
3724 static inline bool netif_attr_test_online(unsigned long j,
3725 					  const unsigned long *online_mask,
3726 					  unsigned int nr_bits)
3727 {
3728 	cpu_max_bits_warn(j, nr_bits);
3729 
3730 	if (online_mask)
3731 		return test_bit(j, online_mask);
3732 
3733 	return (j < nr_bits);
3734 }
3735 
3736 /**
3737  *	netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3738  *	@n: CPU/Rx queue index
3739  *	@srcp: the cpumask/Rx queue mask pointer
3740  *	@nr_bits: number of bits in the bitmask
3741  *
3742  * Returns >= nr_bits if no further CPUs/Rx queues set.
3743  */
3744 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3745 					       unsigned int nr_bits)
3746 {
3747 	/* -1 is a legal arg here. */
3748 	if (n != -1)
3749 		cpu_max_bits_warn(n, nr_bits);
3750 
3751 	if (srcp)
3752 		return find_next_bit(srcp, nr_bits, n + 1);
3753 
3754 	return n + 1;
3755 }
3756 
3757 /**
3758  *	netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3759  *	@n: CPU/Rx queue index
3760  *	@src1p: the first CPUs/Rx queues mask pointer
3761  *	@src2p: the second CPUs/Rx queues mask pointer
3762  *	@nr_bits: number of bits in the bitmask
3763  *
3764  * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3765  */
3766 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3767 					  const unsigned long *src2p,
3768 					  unsigned int nr_bits)
3769 {
3770 	/* -1 is a legal arg here. */
3771 	if (n != -1)
3772 		cpu_max_bits_warn(n, nr_bits);
3773 
3774 	if (src1p && src2p)
3775 		return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3776 	else if (src1p)
3777 		return find_next_bit(src1p, nr_bits, n + 1);
3778 	else if (src2p)
3779 		return find_next_bit(src2p, nr_bits, n + 1);
3780 
3781 	return n + 1;
3782 }
3783 #else
3784 static inline int netif_set_xps_queue(struct net_device *dev,
3785 				      const struct cpumask *mask,
3786 				      u16 index)
3787 {
3788 	return 0;
3789 }
3790 
3791 static inline int __netif_set_xps_queue(struct net_device *dev,
3792 					const unsigned long *mask,
3793 					u16 index, enum xps_map_type type)
3794 {
3795 	return 0;
3796 }
3797 #endif
3798 
3799 /**
3800  *	netif_is_multiqueue - test if device has multiple transmit queues
3801  *	@dev: network device
3802  *
3803  * Check if device has multiple transmit queues
3804  */
3805 static inline bool netif_is_multiqueue(const struct net_device *dev)
3806 {
3807 	return dev->num_tx_queues > 1;
3808 }
3809 
3810 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3811 
3812 #ifdef CONFIG_SYSFS
3813 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3814 #else
3815 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3816 						unsigned int rxqs)
3817 {
3818 	dev->real_num_rx_queues = rxqs;
3819 	return 0;
3820 }
3821 #endif
3822 int netif_set_real_num_queues(struct net_device *dev,
3823 			      unsigned int txq, unsigned int rxq);
3824 
3825 int netif_get_num_default_rss_queues(void);
3826 
3827 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3828 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3829 
3830 /*
3831  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3832  * interrupt context or with hardware interrupts being disabled.
3833  * (in_hardirq() || irqs_disabled())
3834  *
3835  * We provide four helpers that can be used in following contexts :
3836  *
3837  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3838  *  replacing kfree_skb(skb)
3839  *
3840  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3841  *  Typically used in place of consume_skb(skb) in TX completion path
3842  *
3843  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3844  *  replacing kfree_skb(skb)
3845  *
3846  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3847  *  and consumed a packet. Used in place of consume_skb(skb)
3848  */
3849 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3850 {
3851 	dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3852 }
3853 
3854 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3855 {
3856 	dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3857 }
3858 
3859 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3860 {
3861 	dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3862 }
3863 
3864 static inline void dev_consume_skb_any(struct sk_buff *skb)
3865 {
3866 	dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3867 }
3868 
3869 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3870 			     struct bpf_prog *xdp_prog);
3871 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3872 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3873 int netif_rx(struct sk_buff *skb);
3874 int __netif_rx(struct sk_buff *skb);
3875 
3876 int netif_receive_skb(struct sk_buff *skb);
3877 int netif_receive_skb_core(struct sk_buff *skb);
3878 void netif_receive_skb_list_internal(struct list_head *head);
3879 void netif_receive_skb_list(struct list_head *head);
3880 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3881 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3882 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3883 void napi_get_frags_check(struct napi_struct *napi);
3884 gro_result_t napi_gro_frags(struct napi_struct *napi);
3885 struct packet_offload *gro_find_receive_by_type(__be16 type);
3886 struct packet_offload *gro_find_complete_by_type(__be16 type);
3887 
3888 static inline void napi_free_frags(struct napi_struct *napi)
3889 {
3890 	kfree_skb(napi->skb);
3891 	napi->skb = NULL;
3892 }
3893 
3894 bool netdev_is_rx_handler_busy(struct net_device *dev);
3895 int netdev_rx_handler_register(struct net_device *dev,
3896 			       rx_handler_func_t *rx_handler,
3897 			       void *rx_handler_data);
3898 void netdev_rx_handler_unregister(struct net_device *dev);
3899 
3900 bool dev_valid_name(const char *name);
3901 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3902 {
3903 	return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3904 }
3905 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3906 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3907 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3908 		void __user *data, bool *need_copyout);
3909 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3910 int generic_hwtstamp_get_lower(struct net_device *dev,
3911 			       struct kernel_hwtstamp_config *kernel_cfg);
3912 int generic_hwtstamp_set_lower(struct net_device *dev,
3913 			       struct kernel_hwtstamp_config *kernel_cfg,
3914 			       struct netlink_ext_ack *extack);
3915 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3916 unsigned int dev_get_flags(const struct net_device *);
3917 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3918 		       struct netlink_ext_ack *extack);
3919 int dev_change_flags(struct net_device *dev, unsigned int flags,
3920 		     struct netlink_ext_ack *extack);
3921 int dev_set_alias(struct net_device *, const char *, size_t);
3922 int dev_get_alias(const struct net_device *, char *, size_t);
3923 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3924 			       const char *pat, int new_ifindex);
3925 static inline
3926 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3927 			     const char *pat)
3928 {
3929 	return __dev_change_net_namespace(dev, net, pat, 0);
3930 }
3931 int __dev_set_mtu(struct net_device *, int);
3932 int dev_set_mtu(struct net_device *, int);
3933 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3934 			      struct netlink_ext_ack *extack);
3935 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3936 			struct netlink_ext_ack *extack);
3937 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3938 			     struct netlink_ext_ack *extack);
3939 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3940 int dev_get_port_parent_id(struct net_device *dev,
3941 			   struct netdev_phys_item_id *ppid, bool recurse);
3942 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3943 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3944 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3945 				    struct netdev_queue *txq, int *ret);
3946 
3947 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3948 u8 dev_xdp_prog_count(struct net_device *dev);
3949 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3950 
3951 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3952 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3953 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3954 bool is_skb_forwardable(const struct net_device *dev,
3955 			const struct sk_buff *skb);
3956 
3957 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3958 						 const struct sk_buff *skb,
3959 						 const bool check_mtu)
3960 {
3961 	const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3962 	unsigned int len;
3963 
3964 	if (!(dev->flags & IFF_UP))
3965 		return false;
3966 
3967 	if (!check_mtu)
3968 		return true;
3969 
3970 	len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3971 	if (skb->len <= len)
3972 		return true;
3973 
3974 	/* if TSO is enabled, we don't care about the length as the packet
3975 	 * could be forwarded without being segmented before
3976 	 */
3977 	if (skb_is_gso(skb))
3978 		return true;
3979 
3980 	return false;
3981 }
3982 
3983 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
3984 
3985 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
3986 {
3987 	/* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
3988 	struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
3989 
3990 	if (likely(p))
3991 		return p;
3992 
3993 	return netdev_core_stats_alloc(dev);
3994 }
3995 
3996 #define DEV_CORE_STATS_INC(FIELD)						\
3997 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev)		\
3998 {										\
3999 	struct net_device_core_stats __percpu *p;				\
4000 										\
4001 	p = dev_core_stats(dev);						\
4002 	if (p)									\
4003 		this_cpu_inc(p->FIELD);						\
4004 }
4005 DEV_CORE_STATS_INC(rx_dropped)
4006 DEV_CORE_STATS_INC(tx_dropped)
4007 DEV_CORE_STATS_INC(rx_nohandler)
4008 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4009 
4010 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4011 					       struct sk_buff *skb,
4012 					       const bool check_mtu)
4013 {
4014 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4015 	    unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4016 		dev_core_stats_rx_dropped_inc(dev);
4017 		kfree_skb(skb);
4018 		return NET_RX_DROP;
4019 	}
4020 
4021 	skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4022 	skb->priority = 0;
4023 	return 0;
4024 }
4025 
4026 bool dev_nit_active(struct net_device *dev);
4027 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4028 
4029 static inline void __dev_put(struct net_device *dev)
4030 {
4031 	if (dev) {
4032 #ifdef CONFIG_PCPU_DEV_REFCNT
4033 		this_cpu_dec(*dev->pcpu_refcnt);
4034 #else
4035 		refcount_dec(&dev->dev_refcnt);
4036 #endif
4037 	}
4038 }
4039 
4040 static inline void __dev_hold(struct net_device *dev)
4041 {
4042 	if (dev) {
4043 #ifdef CONFIG_PCPU_DEV_REFCNT
4044 		this_cpu_inc(*dev->pcpu_refcnt);
4045 #else
4046 		refcount_inc(&dev->dev_refcnt);
4047 #endif
4048 	}
4049 }
4050 
4051 static inline void __netdev_tracker_alloc(struct net_device *dev,
4052 					  netdevice_tracker *tracker,
4053 					  gfp_t gfp)
4054 {
4055 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4056 	ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4057 #endif
4058 }
4059 
4060 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4061  * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4062  */
4063 static inline void netdev_tracker_alloc(struct net_device *dev,
4064 					netdevice_tracker *tracker, gfp_t gfp)
4065 {
4066 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4067 	refcount_dec(&dev->refcnt_tracker.no_tracker);
4068 	__netdev_tracker_alloc(dev, tracker, gfp);
4069 #endif
4070 }
4071 
4072 static inline void netdev_tracker_free(struct net_device *dev,
4073 				       netdevice_tracker *tracker)
4074 {
4075 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4076 	ref_tracker_free(&dev->refcnt_tracker, tracker);
4077 #endif
4078 }
4079 
4080 static inline void netdev_hold(struct net_device *dev,
4081 			       netdevice_tracker *tracker, gfp_t gfp)
4082 {
4083 	if (dev) {
4084 		__dev_hold(dev);
4085 		__netdev_tracker_alloc(dev, tracker, gfp);
4086 	}
4087 }
4088 
4089 static inline void netdev_put(struct net_device *dev,
4090 			      netdevice_tracker *tracker)
4091 {
4092 	if (dev) {
4093 		netdev_tracker_free(dev, tracker);
4094 		__dev_put(dev);
4095 	}
4096 }
4097 
4098 /**
4099  *	dev_hold - get reference to device
4100  *	@dev: network device
4101  *
4102  * Hold reference to device to keep it from being freed.
4103  * Try using netdev_hold() instead.
4104  */
4105 static inline void dev_hold(struct net_device *dev)
4106 {
4107 	netdev_hold(dev, NULL, GFP_ATOMIC);
4108 }
4109 
4110 /**
4111  *	dev_put - release reference to device
4112  *	@dev: network device
4113  *
4114  * Release reference to device to allow it to be freed.
4115  * Try using netdev_put() instead.
4116  */
4117 static inline void dev_put(struct net_device *dev)
4118 {
4119 	netdev_put(dev, NULL);
4120 }
4121 
4122 static inline void netdev_ref_replace(struct net_device *odev,
4123 				      struct net_device *ndev,
4124 				      netdevice_tracker *tracker,
4125 				      gfp_t gfp)
4126 {
4127 	if (odev)
4128 		netdev_tracker_free(odev, tracker);
4129 
4130 	__dev_hold(ndev);
4131 	__dev_put(odev);
4132 
4133 	if (ndev)
4134 		__netdev_tracker_alloc(ndev, tracker, gfp);
4135 }
4136 
4137 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4138  * and _off may be called from IRQ context, but it is caller
4139  * who is responsible for serialization of these calls.
4140  *
4141  * The name carrier is inappropriate, these functions should really be
4142  * called netif_lowerlayer_*() because they represent the state of any
4143  * kind of lower layer not just hardware media.
4144  */
4145 void linkwatch_fire_event(struct net_device *dev);
4146 
4147 /**
4148  *	netif_carrier_ok - test if carrier present
4149  *	@dev: network device
4150  *
4151  * Check if carrier is present on device
4152  */
4153 static inline bool netif_carrier_ok(const struct net_device *dev)
4154 {
4155 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4156 }
4157 
4158 unsigned long dev_trans_start(struct net_device *dev);
4159 
4160 void __netdev_watchdog_up(struct net_device *dev);
4161 
4162 void netif_carrier_on(struct net_device *dev);
4163 void netif_carrier_off(struct net_device *dev);
4164 void netif_carrier_event(struct net_device *dev);
4165 
4166 /**
4167  *	netif_dormant_on - mark device as dormant.
4168  *	@dev: network device
4169  *
4170  * Mark device as dormant (as per RFC2863).
4171  *
4172  * The dormant state indicates that the relevant interface is not
4173  * actually in a condition to pass packets (i.e., it is not 'up') but is
4174  * in a "pending" state, waiting for some external event.  For "on-
4175  * demand" interfaces, this new state identifies the situation where the
4176  * interface is waiting for events to place it in the up state.
4177  */
4178 static inline void netif_dormant_on(struct net_device *dev)
4179 {
4180 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4181 		linkwatch_fire_event(dev);
4182 }
4183 
4184 /**
4185  *	netif_dormant_off - set device as not dormant.
4186  *	@dev: network device
4187  *
4188  * Device is not in dormant state.
4189  */
4190 static inline void netif_dormant_off(struct net_device *dev)
4191 {
4192 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4193 		linkwatch_fire_event(dev);
4194 }
4195 
4196 /**
4197  *	netif_dormant - test if device is dormant
4198  *	@dev: network device
4199  *
4200  * Check if device is dormant.
4201  */
4202 static inline bool netif_dormant(const struct net_device *dev)
4203 {
4204 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
4205 }
4206 
4207 
4208 /**
4209  *	netif_testing_on - mark device as under test.
4210  *	@dev: network device
4211  *
4212  * Mark device as under test (as per RFC2863).
4213  *
4214  * The testing state indicates that some test(s) must be performed on
4215  * the interface. After completion, of the test, the interface state
4216  * will change to up, dormant, or down, as appropriate.
4217  */
4218 static inline void netif_testing_on(struct net_device *dev)
4219 {
4220 	if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4221 		linkwatch_fire_event(dev);
4222 }
4223 
4224 /**
4225  *	netif_testing_off - set device as not under test.
4226  *	@dev: network device
4227  *
4228  * Device is not in testing state.
4229  */
4230 static inline void netif_testing_off(struct net_device *dev)
4231 {
4232 	if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4233 		linkwatch_fire_event(dev);
4234 }
4235 
4236 /**
4237  *	netif_testing - test if device is under test
4238  *	@dev: network device
4239  *
4240  * Check if device is under test
4241  */
4242 static inline bool netif_testing(const struct net_device *dev)
4243 {
4244 	return test_bit(__LINK_STATE_TESTING, &dev->state);
4245 }
4246 
4247 
4248 /**
4249  *	netif_oper_up - test if device is operational
4250  *	@dev: network device
4251  *
4252  * Check if carrier is operational
4253  */
4254 static inline bool netif_oper_up(const struct net_device *dev)
4255 {
4256 	return (dev->operstate == IF_OPER_UP ||
4257 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4258 }
4259 
4260 /**
4261  *	netif_device_present - is device available or removed
4262  *	@dev: network device
4263  *
4264  * Check if device has not been removed from system.
4265  */
4266 static inline bool netif_device_present(const struct net_device *dev)
4267 {
4268 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
4269 }
4270 
4271 void netif_device_detach(struct net_device *dev);
4272 
4273 void netif_device_attach(struct net_device *dev);
4274 
4275 /*
4276  * Network interface message level settings
4277  */
4278 
4279 enum {
4280 	NETIF_MSG_DRV_BIT,
4281 	NETIF_MSG_PROBE_BIT,
4282 	NETIF_MSG_LINK_BIT,
4283 	NETIF_MSG_TIMER_BIT,
4284 	NETIF_MSG_IFDOWN_BIT,
4285 	NETIF_MSG_IFUP_BIT,
4286 	NETIF_MSG_RX_ERR_BIT,
4287 	NETIF_MSG_TX_ERR_BIT,
4288 	NETIF_MSG_TX_QUEUED_BIT,
4289 	NETIF_MSG_INTR_BIT,
4290 	NETIF_MSG_TX_DONE_BIT,
4291 	NETIF_MSG_RX_STATUS_BIT,
4292 	NETIF_MSG_PKTDATA_BIT,
4293 	NETIF_MSG_HW_BIT,
4294 	NETIF_MSG_WOL_BIT,
4295 
4296 	/* When you add a new bit above, update netif_msg_class_names array
4297 	 * in net/ethtool/common.c
4298 	 */
4299 	NETIF_MSG_CLASS_COUNT,
4300 };
4301 /* Both ethtool_ops interface and internal driver implementation use u32 */
4302 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4303 
4304 #define __NETIF_MSG_BIT(bit)	((u32)1 << (bit))
4305 #define __NETIF_MSG(name)	__NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4306 
4307 #define NETIF_MSG_DRV		__NETIF_MSG(DRV)
4308 #define NETIF_MSG_PROBE		__NETIF_MSG(PROBE)
4309 #define NETIF_MSG_LINK		__NETIF_MSG(LINK)
4310 #define NETIF_MSG_TIMER		__NETIF_MSG(TIMER)
4311 #define NETIF_MSG_IFDOWN	__NETIF_MSG(IFDOWN)
4312 #define NETIF_MSG_IFUP		__NETIF_MSG(IFUP)
4313 #define NETIF_MSG_RX_ERR	__NETIF_MSG(RX_ERR)
4314 #define NETIF_MSG_TX_ERR	__NETIF_MSG(TX_ERR)
4315 #define NETIF_MSG_TX_QUEUED	__NETIF_MSG(TX_QUEUED)
4316 #define NETIF_MSG_INTR		__NETIF_MSG(INTR)
4317 #define NETIF_MSG_TX_DONE	__NETIF_MSG(TX_DONE)
4318 #define NETIF_MSG_RX_STATUS	__NETIF_MSG(RX_STATUS)
4319 #define NETIF_MSG_PKTDATA	__NETIF_MSG(PKTDATA)
4320 #define NETIF_MSG_HW		__NETIF_MSG(HW)
4321 #define NETIF_MSG_WOL		__NETIF_MSG(WOL)
4322 
4323 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
4324 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
4325 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
4326 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
4327 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
4328 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
4329 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
4330 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
4331 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4332 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
4333 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
4334 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
4335 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
4336 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
4337 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
4338 
4339 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4340 {
4341 	/* use default */
4342 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4343 		return default_msg_enable_bits;
4344 	if (debug_value == 0)	/* no output */
4345 		return 0;
4346 	/* set low N bits */
4347 	return (1U << debug_value) - 1;
4348 }
4349 
4350 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4351 {
4352 	spin_lock(&txq->_xmit_lock);
4353 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4354 	WRITE_ONCE(txq->xmit_lock_owner, cpu);
4355 }
4356 
4357 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4358 {
4359 	__acquire(&txq->_xmit_lock);
4360 	return true;
4361 }
4362 
4363 static inline void __netif_tx_release(struct netdev_queue *txq)
4364 {
4365 	__release(&txq->_xmit_lock);
4366 }
4367 
4368 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4369 {
4370 	spin_lock_bh(&txq->_xmit_lock);
4371 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4372 	WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4373 }
4374 
4375 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4376 {
4377 	bool ok = spin_trylock(&txq->_xmit_lock);
4378 
4379 	if (likely(ok)) {
4380 		/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4381 		WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4382 	}
4383 	return ok;
4384 }
4385 
4386 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4387 {
4388 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4389 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4390 	spin_unlock(&txq->_xmit_lock);
4391 }
4392 
4393 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4394 {
4395 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4396 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4397 	spin_unlock_bh(&txq->_xmit_lock);
4398 }
4399 
4400 /*
4401  * txq->trans_start can be read locklessly from dev_watchdog()
4402  */
4403 static inline void txq_trans_update(struct netdev_queue *txq)
4404 {
4405 	if (txq->xmit_lock_owner != -1)
4406 		WRITE_ONCE(txq->trans_start, jiffies);
4407 }
4408 
4409 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4410 {
4411 	unsigned long now = jiffies;
4412 
4413 	if (READ_ONCE(txq->trans_start) != now)
4414 		WRITE_ONCE(txq->trans_start, now);
4415 }
4416 
4417 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4418 static inline void netif_trans_update(struct net_device *dev)
4419 {
4420 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4421 
4422 	txq_trans_cond_update(txq);
4423 }
4424 
4425 /**
4426  *	netif_tx_lock - grab network device transmit lock
4427  *	@dev: network device
4428  *
4429  * Get network device transmit lock
4430  */
4431 void netif_tx_lock(struct net_device *dev);
4432 
4433 static inline void netif_tx_lock_bh(struct net_device *dev)
4434 {
4435 	local_bh_disable();
4436 	netif_tx_lock(dev);
4437 }
4438 
4439 void netif_tx_unlock(struct net_device *dev);
4440 
4441 static inline void netif_tx_unlock_bh(struct net_device *dev)
4442 {
4443 	netif_tx_unlock(dev);
4444 	local_bh_enable();
4445 }
4446 
4447 #define HARD_TX_LOCK(dev, txq, cpu) {			\
4448 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4449 		__netif_tx_lock(txq, cpu);		\
4450 	} else {					\
4451 		__netif_tx_acquire(txq);		\
4452 	}						\
4453 }
4454 
4455 #define HARD_TX_TRYLOCK(dev, txq)			\
4456 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
4457 		__netif_tx_trylock(txq) :		\
4458 		__netif_tx_acquire(txq))
4459 
4460 #define HARD_TX_UNLOCK(dev, txq) {			\
4461 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4462 		__netif_tx_unlock(txq);			\
4463 	} else {					\
4464 		__netif_tx_release(txq);		\
4465 	}						\
4466 }
4467 
4468 static inline void netif_tx_disable(struct net_device *dev)
4469 {
4470 	unsigned int i;
4471 	int cpu;
4472 
4473 	local_bh_disable();
4474 	cpu = smp_processor_id();
4475 	spin_lock(&dev->tx_global_lock);
4476 	for (i = 0; i < dev->num_tx_queues; i++) {
4477 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4478 
4479 		__netif_tx_lock(txq, cpu);
4480 		netif_tx_stop_queue(txq);
4481 		__netif_tx_unlock(txq);
4482 	}
4483 	spin_unlock(&dev->tx_global_lock);
4484 	local_bh_enable();
4485 }
4486 
4487 static inline void netif_addr_lock(struct net_device *dev)
4488 {
4489 	unsigned char nest_level = 0;
4490 
4491 #ifdef CONFIG_LOCKDEP
4492 	nest_level = dev->nested_level;
4493 #endif
4494 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4495 }
4496 
4497 static inline void netif_addr_lock_bh(struct net_device *dev)
4498 {
4499 	unsigned char nest_level = 0;
4500 
4501 #ifdef CONFIG_LOCKDEP
4502 	nest_level = dev->nested_level;
4503 #endif
4504 	local_bh_disable();
4505 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4506 }
4507 
4508 static inline void netif_addr_unlock(struct net_device *dev)
4509 {
4510 	spin_unlock(&dev->addr_list_lock);
4511 }
4512 
4513 static inline void netif_addr_unlock_bh(struct net_device *dev)
4514 {
4515 	spin_unlock_bh(&dev->addr_list_lock);
4516 }
4517 
4518 /*
4519  * dev_addrs walker. Should be used only for read access. Call with
4520  * rcu_read_lock held.
4521  */
4522 #define for_each_dev_addr(dev, ha) \
4523 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4524 
4525 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4526 
4527 void ether_setup(struct net_device *dev);
4528 
4529 /* Support for loadable net-drivers */
4530 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4531 				    unsigned char name_assign_type,
4532 				    void (*setup)(struct net_device *),
4533 				    unsigned int txqs, unsigned int rxqs);
4534 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4535 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4536 
4537 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4538 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4539 			 count)
4540 
4541 int register_netdev(struct net_device *dev);
4542 void unregister_netdev(struct net_device *dev);
4543 
4544 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4545 
4546 /* General hardware address lists handling functions */
4547 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4548 		   struct netdev_hw_addr_list *from_list, int addr_len);
4549 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4550 		      struct netdev_hw_addr_list *from_list, int addr_len);
4551 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4552 		       struct net_device *dev,
4553 		       int (*sync)(struct net_device *, const unsigned char *),
4554 		       int (*unsync)(struct net_device *,
4555 				     const unsigned char *));
4556 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4557 			   struct net_device *dev,
4558 			   int (*sync)(struct net_device *,
4559 				       const unsigned char *, int),
4560 			   int (*unsync)(struct net_device *,
4561 					 const unsigned char *, int));
4562 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4563 			      struct net_device *dev,
4564 			      int (*unsync)(struct net_device *,
4565 					    const unsigned char *, int));
4566 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4567 			  struct net_device *dev,
4568 			  int (*unsync)(struct net_device *,
4569 					const unsigned char *));
4570 void __hw_addr_init(struct netdev_hw_addr_list *list);
4571 
4572 /* Functions used for device addresses handling */
4573 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4574 		  const void *addr, size_t len);
4575 
4576 static inline void
4577 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4578 {
4579 	dev_addr_mod(dev, 0, addr, len);
4580 }
4581 
4582 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4583 {
4584 	__dev_addr_set(dev, addr, dev->addr_len);
4585 }
4586 
4587 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4588 		 unsigned char addr_type);
4589 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4590 		 unsigned char addr_type);
4591 
4592 /* Functions used for unicast addresses handling */
4593 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4594 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4595 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4596 int dev_uc_sync(struct net_device *to, struct net_device *from);
4597 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4598 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4599 void dev_uc_flush(struct net_device *dev);
4600 void dev_uc_init(struct net_device *dev);
4601 
4602 /**
4603  *  __dev_uc_sync - Synchonize device's unicast list
4604  *  @dev:  device to sync
4605  *  @sync: function to call if address should be added
4606  *  @unsync: function to call if address should be removed
4607  *
4608  *  Add newly added addresses to the interface, and release
4609  *  addresses that have been deleted.
4610  */
4611 static inline int __dev_uc_sync(struct net_device *dev,
4612 				int (*sync)(struct net_device *,
4613 					    const unsigned char *),
4614 				int (*unsync)(struct net_device *,
4615 					      const unsigned char *))
4616 {
4617 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4618 }
4619 
4620 /**
4621  *  __dev_uc_unsync - Remove synchronized addresses from device
4622  *  @dev:  device to sync
4623  *  @unsync: function to call if address should be removed
4624  *
4625  *  Remove all addresses that were added to the device by dev_uc_sync().
4626  */
4627 static inline void __dev_uc_unsync(struct net_device *dev,
4628 				   int (*unsync)(struct net_device *,
4629 						 const unsigned char *))
4630 {
4631 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
4632 }
4633 
4634 /* Functions used for multicast addresses handling */
4635 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4636 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4637 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4638 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4639 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4640 int dev_mc_sync(struct net_device *to, struct net_device *from);
4641 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4642 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4643 void dev_mc_flush(struct net_device *dev);
4644 void dev_mc_init(struct net_device *dev);
4645 
4646 /**
4647  *  __dev_mc_sync - Synchonize device's multicast list
4648  *  @dev:  device to sync
4649  *  @sync: function to call if address should be added
4650  *  @unsync: function to call if address should be removed
4651  *
4652  *  Add newly added addresses to the interface, and release
4653  *  addresses that have been deleted.
4654  */
4655 static inline int __dev_mc_sync(struct net_device *dev,
4656 				int (*sync)(struct net_device *,
4657 					    const unsigned char *),
4658 				int (*unsync)(struct net_device *,
4659 					      const unsigned char *))
4660 {
4661 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4662 }
4663 
4664 /**
4665  *  __dev_mc_unsync - Remove synchronized addresses from device
4666  *  @dev:  device to sync
4667  *  @unsync: function to call if address should be removed
4668  *
4669  *  Remove all addresses that were added to the device by dev_mc_sync().
4670  */
4671 static inline void __dev_mc_unsync(struct net_device *dev,
4672 				   int (*unsync)(struct net_device *,
4673 						 const unsigned char *))
4674 {
4675 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
4676 }
4677 
4678 /* Functions used for secondary unicast and multicast support */
4679 void dev_set_rx_mode(struct net_device *dev);
4680 int dev_set_promiscuity(struct net_device *dev, int inc);
4681 int dev_set_allmulti(struct net_device *dev, int inc);
4682 void netdev_state_change(struct net_device *dev);
4683 void __netdev_notify_peers(struct net_device *dev);
4684 void netdev_notify_peers(struct net_device *dev);
4685 void netdev_features_change(struct net_device *dev);
4686 /* Load a device via the kmod */
4687 void dev_load(struct net *net, const char *name);
4688 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4689 					struct rtnl_link_stats64 *storage);
4690 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4691 			     const struct net_device_stats *netdev_stats);
4692 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4693 			   const struct pcpu_sw_netstats __percpu *netstats);
4694 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4695 
4696 extern int		netdev_max_backlog;
4697 extern int		dev_rx_weight;
4698 extern int		dev_tx_weight;
4699 extern int		gro_normal_batch;
4700 
4701 enum {
4702 	NESTED_SYNC_IMM_BIT,
4703 	NESTED_SYNC_TODO_BIT,
4704 };
4705 
4706 #define __NESTED_SYNC_BIT(bit)	((u32)1 << (bit))
4707 #define __NESTED_SYNC(name)	__NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4708 
4709 #define NESTED_SYNC_IMM		__NESTED_SYNC(IMM)
4710 #define NESTED_SYNC_TODO	__NESTED_SYNC(TODO)
4711 
4712 struct netdev_nested_priv {
4713 	unsigned char flags;
4714 	void *data;
4715 };
4716 
4717 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4718 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4719 						     struct list_head **iter);
4720 
4721 /* iterate through upper list, must be called under RCU read lock */
4722 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4723 	for (iter = &(dev)->adj_list.upper, \
4724 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4725 	     updev; \
4726 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4727 
4728 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4729 				  int (*fn)(struct net_device *upper_dev,
4730 					    struct netdev_nested_priv *priv),
4731 				  struct netdev_nested_priv *priv);
4732 
4733 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4734 				  struct net_device *upper_dev);
4735 
4736 bool netdev_has_any_upper_dev(struct net_device *dev);
4737 
4738 void *netdev_lower_get_next_private(struct net_device *dev,
4739 				    struct list_head **iter);
4740 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4741 					struct list_head **iter);
4742 
4743 #define netdev_for_each_lower_private(dev, priv, iter) \
4744 	for (iter = (dev)->adj_list.lower.next, \
4745 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
4746 	     priv; \
4747 	     priv = netdev_lower_get_next_private(dev, &(iter)))
4748 
4749 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4750 	for (iter = &(dev)->adj_list.lower, \
4751 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4752 	     priv; \
4753 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4754 
4755 void *netdev_lower_get_next(struct net_device *dev,
4756 				struct list_head **iter);
4757 
4758 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4759 	for (iter = (dev)->adj_list.lower.next, \
4760 	     ldev = netdev_lower_get_next(dev, &(iter)); \
4761 	     ldev; \
4762 	     ldev = netdev_lower_get_next(dev, &(iter)))
4763 
4764 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4765 					     struct list_head **iter);
4766 int netdev_walk_all_lower_dev(struct net_device *dev,
4767 			      int (*fn)(struct net_device *lower_dev,
4768 					struct netdev_nested_priv *priv),
4769 			      struct netdev_nested_priv *priv);
4770 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4771 				  int (*fn)(struct net_device *lower_dev,
4772 					    struct netdev_nested_priv *priv),
4773 				  struct netdev_nested_priv *priv);
4774 
4775 void *netdev_adjacent_get_private(struct list_head *adj_list);
4776 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4777 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4778 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4779 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4780 			  struct netlink_ext_ack *extack);
4781 int netdev_master_upper_dev_link(struct net_device *dev,
4782 				 struct net_device *upper_dev,
4783 				 void *upper_priv, void *upper_info,
4784 				 struct netlink_ext_ack *extack);
4785 void netdev_upper_dev_unlink(struct net_device *dev,
4786 			     struct net_device *upper_dev);
4787 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4788 				   struct net_device *new_dev,
4789 				   struct net_device *dev,
4790 				   struct netlink_ext_ack *extack);
4791 void netdev_adjacent_change_commit(struct net_device *old_dev,
4792 				   struct net_device *new_dev,
4793 				   struct net_device *dev);
4794 void netdev_adjacent_change_abort(struct net_device *old_dev,
4795 				  struct net_device *new_dev,
4796 				  struct net_device *dev);
4797 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4798 void *netdev_lower_dev_get_private(struct net_device *dev,
4799 				   struct net_device *lower_dev);
4800 void netdev_lower_state_changed(struct net_device *lower_dev,
4801 				void *lower_state_info);
4802 
4803 /* RSS keys are 40 or 52 bytes long */
4804 #define NETDEV_RSS_KEY_LEN 52
4805 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4806 void netdev_rss_key_fill(void *buffer, size_t len);
4807 
4808 int skb_checksum_help(struct sk_buff *skb);
4809 int skb_crc32c_csum_help(struct sk_buff *skb);
4810 int skb_csum_hwoffload_help(struct sk_buff *skb,
4811 			    const netdev_features_t features);
4812 
4813 struct netdev_bonding_info {
4814 	ifslave	slave;
4815 	ifbond	master;
4816 };
4817 
4818 struct netdev_notifier_bonding_info {
4819 	struct netdev_notifier_info info; /* must be first */
4820 	struct netdev_bonding_info  bonding_info;
4821 };
4822 
4823 void netdev_bonding_info_change(struct net_device *dev,
4824 				struct netdev_bonding_info *bonding_info);
4825 
4826 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4827 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4828 #else
4829 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4830 				  const void *data)
4831 {
4832 }
4833 #endif
4834 
4835 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4836 
4837 static inline bool can_checksum_protocol(netdev_features_t features,
4838 					 __be16 protocol)
4839 {
4840 	if (protocol == htons(ETH_P_FCOE))
4841 		return !!(features & NETIF_F_FCOE_CRC);
4842 
4843 	/* Assume this is an IP checksum (not SCTP CRC) */
4844 
4845 	if (features & NETIF_F_HW_CSUM) {
4846 		/* Can checksum everything */
4847 		return true;
4848 	}
4849 
4850 	switch (protocol) {
4851 	case htons(ETH_P_IP):
4852 		return !!(features & NETIF_F_IP_CSUM);
4853 	case htons(ETH_P_IPV6):
4854 		return !!(features & NETIF_F_IPV6_CSUM);
4855 	default:
4856 		return false;
4857 	}
4858 }
4859 
4860 #ifdef CONFIG_BUG
4861 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4862 #else
4863 static inline void netdev_rx_csum_fault(struct net_device *dev,
4864 					struct sk_buff *skb)
4865 {
4866 }
4867 #endif
4868 /* rx skb timestamps */
4869 void net_enable_timestamp(void);
4870 void net_disable_timestamp(void);
4871 
4872 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4873 					const struct skb_shared_hwtstamps *hwtstamps,
4874 					bool cycles)
4875 {
4876 	const struct net_device_ops *ops = dev->netdev_ops;
4877 
4878 	if (ops->ndo_get_tstamp)
4879 		return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4880 
4881 	return hwtstamps->hwtstamp;
4882 }
4883 
4884 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4885 					      struct sk_buff *skb, struct net_device *dev,
4886 					      bool more)
4887 {
4888 	__this_cpu_write(softnet_data.xmit.more, more);
4889 	return ops->ndo_start_xmit(skb, dev);
4890 }
4891 
4892 static inline bool netdev_xmit_more(void)
4893 {
4894 	return __this_cpu_read(softnet_data.xmit.more);
4895 }
4896 
4897 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4898 					    struct netdev_queue *txq, bool more)
4899 {
4900 	const struct net_device_ops *ops = dev->netdev_ops;
4901 	netdev_tx_t rc;
4902 
4903 	rc = __netdev_start_xmit(ops, skb, dev, more);
4904 	if (rc == NETDEV_TX_OK)
4905 		txq_trans_update(txq);
4906 
4907 	return rc;
4908 }
4909 
4910 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4911 				const void *ns);
4912 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4913 				 const void *ns);
4914 
4915 extern const struct kobj_ns_type_operations net_ns_type_operations;
4916 
4917 const char *netdev_drivername(const struct net_device *dev);
4918 
4919 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4920 							  netdev_features_t f2)
4921 {
4922 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4923 		if (f1 & NETIF_F_HW_CSUM)
4924 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4925 		else
4926 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4927 	}
4928 
4929 	return f1 & f2;
4930 }
4931 
4932 static inline netdev_features_t netdev_get_wanted_features(
4933 	struct net_device *dev)
4934 {
4935 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4936 }
4937 netdev_features_t netdev_increment_features(netdev_features_t all,
4938 	netdev_features_t one, netdev_features_t mask);
4939 
4940 /* Allow TSO being used on stacked device :
4941  * Performing the GSO segmentation before last device
4942  * is a performance improvement.
4943  */
4944 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4945 							netdev_features_t mask)
4946 {
4947 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4948 }
4949 
4950 int __netdev_update_features(struct net_device *dev);
4951 void netdev_update_features(struct net_device *dev);
4952 void netdev_change_features(struct net_device *dev);
4953 
4954 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4955 					struct net_device *dev);
4956 
4957 netdev_features_t passthru_features_check(struct sk_buff *skb,
4958 					  struct net_device *dev,
4959 					  netdev_features_t features);
4960 netdev_features_t netif_skb_features(struct sk_buff *skb);
4961 void skb_warn_bad_offload(const struct sk_buff *skb);
4962 
4963 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4964 {
4965 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4966 
4967 	/* check flags correspondence */
4968 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4969 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4970 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4971 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4972 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4973 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4974 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4975 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4976 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4977 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4978 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4979 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4980 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4981 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4982 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4983 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4984 	BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4985 	BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4986 	BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4987 
4988 	return (features & feature) == feature;
4989 }
4990 
4991 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4992 {
4993 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4994 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4995 }
4996 
4997 static inline bool netif_needs_gso(struct sk_buff *skb,
4998 				   netdev_features_t features)
4999 {
5000 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5001 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5002 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5003 }
5004 
5005 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5006 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5007 void netif_inherit_tso_max(struct net_device *to,
5008 			   const struct net_device *from);
5009 
5010 static inline bool netif_is_macsec(const struct net_device *dev)
5011 {
5012 	return dev->priv_flags & IFF_MACSEC;
5013 }
5014 
5015 static inline bool netif_is_macvlan(const struct net_device *dev)
5016 {
5017 	return dev->priv_flags & IFF_MACVLAN;
5018 }
5019 
5020 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5021 {
5022 	return dev->priv_flags & IFF_MACVLAN_PORT;
5023 }
5024 
5025 static inline bool netif_is_bond_master(const struct net_device *dev)
5026 {
5027 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5028 }
5029 
5030 static inline bool netif_is_bond_slave(const struct net_device *dev)
5031 {
5032 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5033 }
5034 
5035 static inline bool netif_supports_nofcs(struct net_device *dev)
5036 {
5037 	return dev->priv_flags & IFF_SUPP_NOFCS;
5038 }
5039 
5040 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5041 {
5042 	return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5043 }
5044 
5045 static inline bool netif_is_l3_master(const struct net_device *dev)
5046 {
5047 	return dev->priv_flags & IFF_L3MDEV_MASTER;
5048 }
5049 
5050 static inline bool netif_is_l3_slave(const struct net_device *dev)
5051 {
5052 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
5053 }
5054 
5055 static inline int dev_sdif(const struct net_device *dev)
5056 {
5057 #ifdef CONFIG_NET_L3_MASTER_DEV
5058 	if (netif_is_l3_slave(dev))
5059 		return dev->ifindex;
5060 #endif
5061 	return 0;
5062 }
5063 
5064 static inline bool netif_is_bridge_master(const struct net_device *dev)
5065 {
5066 	return dev->priv_flags & IFF_EBRIDGE;
5067 }
5068 
5069 static inline bool netif_is_bridge_port(const struct net_device *dev)
5070 {
5071 	return dev->priv_flags & IFF_BRIDGE_PORT;
5072 }
5073 
5074 static inline bool netif_is_ovs_master(const struct net_device *dev)
5075 {
5076 	return dev->priv_flags & IFF_OPENVSWITCH;
5077 }
5078 
5079 static inline bool netif_is_ovs_port(const struct net_device *dev)
5080 {
5081 	return dev->priv_flags & IFF_OVS_DATAPATH;
5082 }
5083 
5084 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5085 {
5086 	return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5087 }
5088 
5089 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5090 {
5091 	return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5092 }
5093 
5094 static inline bool netif_is_team_master(const struct net_device *dev)
5095 {
5096 	return dev->priv_flags & IFF_TEAM;
5097 }
5098 
5099 static inline bool netif_is_team_port(const struct net_device *dev)
5100 {
5101 	return dev->priv_flags & IFF_TEAM_PORT;
5102 }
5103 
5104 static inline bool netif_is_lag_master(const struct net_device *dev)
5105 {
5106 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
5107 }
5108 
5109 static inline bool netif_is_lag_port(const struct net_device *dev)
5110 {
5111 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5112 }
5113 
5114 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5115 {
5116 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
5117 }
5118 
5119 static inline bool netif_is_failover(const struct net_device *dev)
5120 {
5121 	return dev->priv_flags & IFF_FAILOVER;
5122 }
5123 
5124 static inline bool netif_is_failover_slave(const struct net_device *dev)
5125 {
5126 	return dev->priv_flags & IFF_FAILOVER_SLAVE;
5127 }
5128 
5129 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5130 static inline void netif_keep_dst(struct net_device *dev)
5131 {
5132 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5133 }
5134 
5135 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5136 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5137 {
5138 	/* TODO: reserve and use an additional IFF bit, if we get more users */
5139 	return netif_is_macsec(dev);
5140 }
5141 
5142 extern struct pernet_operations __net_initdata loopback_net_ops;
5143 
5144 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5145 
5146 /* netdev_printk helpers, similar to dev_printk */
5147 
5148 static inline const char *netdev_name(const struct net_device *dev)
5149 {
5150 	if (!dev->name[0] || strchr(dev->name, '%'))
5151 		return "(unnamed net_device)";
5152 	return dev->name;
5153 }
5154 
5155 static inline const char *netdev_reg_state(const struct net_device *dev)
5156 {
5157 	switch (dev->reg_state) {
5158 	case NETREG_UNINITIALIZED: return " (uninitialized)";
5159 	case NETREG_REGISTERED: return "";
5160 	case NETREG_UNREGISTERING: return " (unregistering)";
5161 	case NETREG_UNREGISTERED: return " (unregistered)";
5162 	case NETREG_RELEASED: return " (released)";
5163 	case NETREG_DUMMY: return " (dummy)";
5164 	}
5165 
5166 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5167 	return " (unknown)";
5168 }
5169 
5170 #define MODULE_ALIAS_NETDEV(device) \
5171 	MODULE_ALIAS("netdev-" device)
5172 
5173 /*
5174  * netdev_WARN() acts like dev_printk(), but with the key difference
5175  * of using a WARN/WARN_ON to get the message out, including the
5176  * file/line information and a backtrace.
5177  */
5178 #define netdev_WARN(dev, format, args...)			\
5179 	WARN(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5180 	     netdev_reg_state(dev), ##args)
5181 
5182 #define netdev_WARN_ONCE(dev, format, args...)				\
5183 	WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5184 		  netdev_reg_state(dev), ##args)
5185 
5186 /*
5187  *	The list of packet types we will receive (as opposed to discard)
5188  *	and the routines to invoke.
5189  *
5190  *	Why 16. Because with 16 the only overlap we get on a hash of the
5191  *	low nibble of the protocol value is RARP/SNAP/X.25.
5192  *
5193  *		0800	IP
5194  *		0001	802.3
5195  *		0002	AX.25
5196  *		0004	802.2
5197  *		8035	RARP
5198  *		0005	SNAP
5199  *		0805	X.25
5200  *		0806	ARP
5201  *		8137	IPX
5202  *		0009	Localtalk
5203  *		86DD	IPv6
5204  */
5205 #define PTYPE_HASH_SIZE	(16)
5206 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
5207 
5208 extern struct list_head ptype_all __read_mostly;
5209 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5210 
5211 extern struct net_device *blackhole_netdev;
5212 
5213 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5214 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5215 #define DEV_STATS_ADD(DEV, FIELD, VAL) 	\
5216 		atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5217 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5218 
5219 #endif	/* _LINUX_NETDEVICE_H */
5220