xref: /linux-6.15/include/linux/netdevice.h (revision 6151ff9c)
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		Definitions for the Interfaces handler.
8  *
9  * Version:	@(#)dev.h	1.0.10	08/12/93
10  *
11  * Authors:	Ross Biro
12  *		Fred N. van Kempen, <[email protected]>
13  *		Corey Minyard <[email protected]>
14  *		Donald J. Becker, <[email protected]>
15  *		Alan Cox, <[email protected]>
16  *		Bjorn Ekwall. <[email protected]>
17  *              Pekka Riikonen <[email protected]>
18  *
19  *		Moved to /usr/include/linux for NET3
20  */
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
23 
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
31 #include <asm/local.h>
32 
33 #include <linux/percpu.h>
34 #include <linux/rculist.h>
35 #include <linux/workqueue.h>
36 #include <linux/dynamic_queue_limits.h>
37 
38 #include <net/net_namespace.h>
39 #ifdef CONFIG_DCB
40 #include <net/dcbnl.h>
41 #endif
42 #include <net/netprio_cgroup.h>
43 
44 #include <linux/netdev_features.h>
45 #include <linux/neighbour.h>
46 #include <uapi/linux/netdevice.h>
47 #include <uapi/linux/if_bonding.h>
48 #include <uapi/linux/pkt_cls.h>
49 #include <uapi/linux/netdev.h>
50 #include <linux/hashtable.h>
51 #include <linux/rbtree.h>
52 #include <net/net_trackers.h>
53 #include <net/net_debug.h>
54 #include <net/dropreason-core.h>
55 
56 struct netpoll_info;
57 struct device;
58 struct ethtool_ops;
59 struct kernel_hwtstamp_config;
60 struct phy_device;
61 struct dsa_port;
62 struct ip_tunnel_parm;
63 struct macsec_context;
64 struct macsec_ops;
65 struct netdev_name_node;
66 struct sd_flow_limit;
67 struct sfp_bus;
68 /* 802.11 specific */
69 struct wireless_dev;
70 /* 802.15.4 specific */
71 struct wpan_dev;
72 struct mpls_dev;
73 /* UDP Tunnel offloads */
74 struct udp_tunnel_info;
75 struct udp_tunnel_nic_info;
76 struct udp_tunnel_nic;
77 struct bpf_prog;
78 struct xdp_buff;
79 struct xdp_frame;
80 struct xdp_metadata_ops;
81 struct xdp_md;
82 /* DPLL specific */
83 struct dpll_pin;
84 
85 typedef u32 xdp_features_t;
86 
87 void synchronize_net(void);
88 void netdev_set_default_ethtool_ops(struct net_device *dev,
89 				    const struct ethtool_ops *ops);
90 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
91 
92 /* Backlog congestion levels */
93 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
94 #define NET_RX_DROP		1	/* packet dropped */
95 
96 #define MAX_NEST_DEV 8
97 
98 /*
99  * Transmit return codes: transmit return codes originate from three different
100  * namespaces:
101  *
102  * - qdisc return codes
103  * - driver transmit return codes
104  * - errno values
105  *
106  * Drivers are allowed to return any one of those in their hard_start_xmit()
107  * function. Real network devices commonly used with qdiscs should only return
108  * the driver transmit return codes though - when qdiscs are used, the actual
109  * transmission happens asynchronously, so the value is not propagated to
110  * higher layers. Virtual network devices transmit synchronously; in this case
111  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
112  * others are propagated to higher layers.
113  */
114 
115 /* qdisc ->enqueue() return codes. */
116 #define NET_XMIT_SUCCESS	0x00
117 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
118 #define NET_XMIT_CN		0x02	/* congestion notification	*/
119 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
120 
121 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
122  * indicates that the device will soon be dropping packets, or already drops
123  * some packets of the same priority; prompting us to send less aggressively. */
124 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
125 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
126 
127 /* Driver transmit return codes */
128 #define NETDEV_TX_MASK		0xf0
129 
130 enum netdev_tx {
131 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
132 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
133 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
134 };
135 typedef enum netdev_tx netdev_tx_t;
136 
137 /*
138  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
139  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
140  */
141 static inline bool dev_xmit_complete(int rc)
142 {
143 	/*
144 	 * Positive cases with an skb consumed by a driver:
145 	 * - successful transmission (rc == NETDEV_TX_OK)
146 	 * - error while transmitting (rc < 0)
147 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
148 	 */
149 	if (likely(rc < NET_XMIT_MASK))
150 		return true;
151 
152 	return false;
153 }
154 
155 /*
156  *	Compute the worst-case header length according to the protocols
157  *	used.
158  */
159 
160 #if defined(CONFIG_HYPERV_NET)
161 # define LL_MAX_HEADER 128
162 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
163 # if defined(CONFIG_MAC80211_MESH)
164 #  define LL_MAX_HEADER 128
165 # else
166 #  define LL_MAX_HEADER 96
167 # endif
168 #else
169 # define LL_MAX_HEADER 32
170 #endif
171 
172 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
173     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
174 #define MAX_HEADER LL_MAX_HEADER
175 #else
176 #define MAX_HEADER (LL_MAX_HEADER + 48)
177 #endif
178 
179 /*
180  *	Old network device statistics. Fields are native words
181  *	(unsigned long) so they can be read and written atomically.
182  */
183 
184 #define NET_DEV_STAT(FIELD)			\
185 	union {					\
186 		unsigned long FIELD;		\
187 		atomic_long_t __##FIELD;	\
188 	}
189 
190 struct net_device_stats {
191 	NET_DEV_STAT(rx_packets);
192 	NET_DEV_STAT(tx_packets);
193 	NET_DEV_STAT(rx_bytes);
194 	NET_DEV_STAT(tx_bytes);
195 	NET_DEV_STAT(rx_errors);
196 	NET_DEV_STAT(tx_errors);
197 	NET_DEV_STAT(rx_dropped);
198 	NET_DEV_STAT(tx_dropped);
199 	NET_DEV_STAT(multicast);
200 	NET_DEV_STAT(collisions);
201 	NET_DEV_STAT(rx_length_errors);
202 	NET_DEV_STAT(rx_over_errors);
203 	NET_DEV_STAT(rx_crc_errors);
204 	NET_DEV_STAT(rx_frame_errors);
205 	NET_DEV_STAT(rx_fifo_errors);
206 	NET_DEV_STAT(rx_missed_errors);
207 	NET_DEV_STAT(tx_aborted_errors);
208 	NET_DEV_STAT(tx_carrier_errors);
209 	NET_DEV_STAT(tx_fifo_errors);
210 	NET_DEV_STAT(tx_heartbeat_errors);
211 	NET_DEV_STAT(tx_window_errors);
212 	NET_DEV_STAT(rx_compressed);
213 	NET_DEV_STAT(tx_compressed);
214 };
215 #undef NET_DEV_STAT
216 
217 /* per-cpu stats, allocated on demand.
218  * Try to fit them in a single cache line, for dev_get_stats() sake.
219  */
220 struct net_device_core_stats {
221 	unsigned long	rx_dropped;
222 	unsigned long	tx_dropped;
223 	unsigned long	rx_nohandler;
224 	unsigned long	rx_otherhost_dropped;
225 } __aligned(4 * sizeof(unsigned long));
226 
227 #include <linux/cache.h>
228 #include <linux/skbuff.h>
229 
230 #ifdef CONFIG_RPS
231 #include <linux/static_key.h>
232 extern struct static_key_false rps_needed;
233 extern struct static_key_false rfs_needed;
234 #endif
235 
236 struct neighbour;
237 struct neigh_parms;
238 struct sk_buff;
239 
240 struct netdev_hw_addr {
241 	struct list_head	list;
242 	struct rb_node		node;
243 	unsigned char		addr[MAX_ADDR_LEN];
244 	unsigned char		type;
245 #define NETDEV_HW_ADDR_T_LAN		1
246 #define NETDEV_HW_ADDR_T_SAN		2
247 #define NETDEV_HW_ADDR_T_UNICAST	3
248 #define NETDEV_HW_ADDR_T_MULTICAST	4
249 	bool			global_use;
250 	int			sync_cnt;
251 	int			refcount;
252 	int			synced;
253 	struct rcu_head		rcu_head;
254 };
255 
256 struct netdev_hw_addr_list {
257 	struct list_head	list;
258 	int			count;
259 
260 	/* Auxiliary tree for faster lookup on addition and deletion */
261 	struct rb_root		tree;
262 };
263 
264 #define netdev_hw_addr_list_count(l) ((l)->count)
265 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
266 #define netdev_hw_addr_list_for_each(ha, l) \
267 	list_for_each_entry(ha, &(l)->list, list)
268 
269 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
270 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
271 #define netdev_for_each_uc_addr(ha, dev) \
272 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
273 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
274 	netdev_for_each_uc_addr((_ha), (_dev)) \
275 		if ((_ha)->sync_cnt)
276 
277 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
278 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
279 #define netdev_for_each_mc_addr(ha, dev) \
280 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
281 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
282 	netdev_for_each_mc_addr((_ha), (_dev)) \
283 		if ((_ha)->sync_cnt)
284 
285 struct hh_cache {
286 	unsigned int	hh_len;
287 	seqlock_t	hh_lock;
288 
289 	/* cached hardware header; allow for machine alignment needs.        */
290 #define HH_DATA_MOD	16
291 #define HH_DATA_OFF(__len) \
292 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
293 #define HH_DATA_ALIGN(__len) \
294 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
295 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
296 };
297 
298 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
299  * Alternative is:
300  *   dev->hard_header_len ? (dev->hard_header_len +
301  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
302  *
303  * We could use other alignment values, but we must maintain the
304  * relationship HH alignment <= LL alignment.
305  */
306 #define LL_RESERVED_SPACE(dev) \
307 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
308 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
309 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
310 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
311 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
312 
313 struct header_ops {
314 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
315 			   unsigned short type, const void *daddr,
316 			   const void *saddr, unsigned int len);
317 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
318 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
319 	void	(*cache_update)(struct hh_cache *hh,
320 				const struct net_device *dev,
321 				const unsigned char *haddr);
322 	bool	(*validate)(const char *ll_header, unsigned int len);
323 	__be16	(*parse_protocol)(const struct sk_buff *skb);
324 };
325 
326 /* These flag bits are private to the generic network queueing
327  * layer; they may not be explicitly referenced by any other
328  * code.
329  */
330 
331 enum netdev_state_t {
332 	__LINK_STATE_START,
333 	__LINK_STATE_PRESENT,
334 	__LINK_STATE_NOCARRIER,
335 	__LINK_STATE_LINKWATCH_PENDING,
336 	__LINK_STATE_DORMANT,
337 	__LINK_STATE_TESTING,
338 };
339 
340 struct gro_list {
341 	struct list_head	list;
342 	int			count;
343 };
344 
345 /*
346  * size of gro hash buckets, must less than bit number of
347  * napi_struct::gro_bitmask
348  */
349 #define GRO_HASH_BUCKETS	8
350 
351 /*
352  * Structure for NAPI scheduling similar to tasklet but with weighting
353  */
354 struct napi_struct {
355 	/* The poll_list must only be managed by the entity which
356 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
357 	 * whoever atomically sets that bit can add this napi_struct
358 	 * to the per-CPU poll_list, and whoever clears that bit
359 	 * can remove from the list right before clearing the bit.
360 	 */
361 	struct list_head	poll_list;
362 
363 	unsigned long		state;
364 	int			weight;
365 	int			defer_hard_irqs_count;
366 	unsigned long		gro_bitmask;
367 	int			(*poll)(struct napi_struct *, int);
368 #ifdef CONFIG_NETPOLL
369 	/* CPU actively polling if netpoll is configured */
370 	int			poll_owner;
371 #endif
372 	/* CPU on which NAPI has been scheduled for processing */
373 	int			list_owner;
374 	struct net_device	*dev;
375 	struct gro_list		gro_hash[GRO_HASH_BUCKETS];
376 	struct sk_buff		*skb;
377 	struct list_head	rx_list; /* Pending GRO_NORMAL skbs */
378 	int			rx_count; /* length of rx_list */
379 	unsigned int		napi_id;
380 	struct hrtimer		timer;
381 	struct task_struct	*thread;
382 	/* control-path-only fields follow */
383 	struct list_head	dev_list;
384 	struct hlist_node	napi_hash_node;
385 };
386 
387 enum {
388 	NAPI_STATE_SCHED,		/* Poll is scheduled */
389 	NAPI_STATE_MISSED,		/* reschedule a napi */
390 	NAPI_STATE_DISABLE,		/* Disable pending */
391 	NAPI_STATE_NPSVC,		/* Netpoll - don't dequeue from poll_list */
392 	NAPI_STATE_LISTED,		/* NAPI added to system lists */
393 	NAPI_STATE_NO_BUSY_POLL,	/* Do not add in napi_hash, no busy polling */
394 	NAPI_STATE_IN_BUSY_POLL,	/* sk_busy_loop() owns this NAPI */
395 	NAPI_STATE_PREFER_BUSY_POLL,	/* prefer busy-polling over softirq processing*/
396 	NAPI_STATE_THREADED,		/* The poll is performed inside its own thread*/
397 	NAPI_STATE_SCHED_THREADED,	/* Napi is currently scheduled in threaded mode */
398 };
399 
400 enum {
401 	NAPIF_STATE_SCHED		= BIT(NAPI_STATE_SCHED),
402 	NAPIF_STATE_MISSED		= BIT(NAPI_STATE_MISSED),
403 	NAPIF_STATE_DISABLE		= BIT(NAPI_STATE_DISABLE),
404 	NAPIF_STATE_NPSVC		= BIT(NAPI_STATE_NPSVC),
405 	NAPIF_STATE_LISTED		= BIT(NAPI_STATE_LISTED),
406 	NAPIF_STATE_NO_BUSY_POLL	= BIT(NAPI_STATE_NO_BUSY_POLL),
407 	NAPIF_STATE_IN_BUSY_POLL	= BIT(NAPI_STATE_IN_BUSY_POLL),
408 	NAPIF_STATE_PREFER_BUSY_POLL	= BIT(NAPI_STATE_PREFER_BUSY_POLL),
409 	NAPIF_STATE_THREADED		= BIT(NAPI_STATE_THREADED),
410 	NAPIF_STATE_SCHED_THREADED	= BIT(NAPI_STATE_SCHED_THREADED),
411 };
412 
413 enum gro_result {
414 	GRO_MERGED,
415 	GRO_MERGED_FREE,
416 	GRO_HELD,
417 	GRO_NORMAL,
418 	GRO_CONSUMED,
419 };
420 typedef enum gro_result gro_result_t;
421 
422 /*
423  * enum rx_handler_result - Possible return values for rx_handlers.
424  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
425  * further.
426  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
427  * case skb->dev was changed by rx_handler.
428  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
429  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
430  *
431  * rx_handlers are functions called from inside __netif_receive_skb(), to do
432  * special processing of the skb, prior to delivery to protocol handlers.
433  *
434  * Currently, a net_device can only have a single rx_handler registered. Trying
435  * to register a second rx_handler will return -EBUSY.
436  *
437  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
438  * To unregister a rx_handler on a net_device, use
439  * netdev_rx_handler_unregister().
440  *
441  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
442  * do with the skb.
443  *
444  * If the rx_handler consumed the skb in some way, it should return
445  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
446  * the skb to be delivered in some other way.
447  *
448  * If the rx_handler changed skb->dev, to divert the skb to another
449  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
450  * new device will be called if it exists.
451  *
452  * If the rx_handler decides the skb should be ignored, it should return
453  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
454  * are registered on exact device (ptype->dev == skb->dev).
455  *
456  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
457  * delivered, it should return RX_HANDLER_PASS.
458  *
459  * A device without a registered rx_handler will behave as if rx_handler
460  * returned RX_HANDLER_PASS.
461  */
462 
463 enum rx_handler_result {
464 	RX_HANDLER_CONSUMED,
465 	RX_HANDLER_ANOTHER,
466 	RX_HANDLER_EXACT,
467 	RX_HANDLER_PASS,
468 };
469 typedef enum rx_handler_result rx_handler_result_t;
470 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
471 
472 void __napi_schedule(struct napi_struct *n);
473 void __napi_schedule_irqoff(struct napi_struct *n);
474 
475 static inline bool napi_disable_pending(struct napi_struct *n)
476 {
477 	return test_bit(NAPI_STATE_DISABLE, &n->state);
478 }
479 
480 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
481 {
482 	return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
483 }
484 
485 bool napi_schedule_prep(struct napi_struct *n);
486 
487 /**
488  *	napi_schedule - schedule NAPI poll
489  *	@n: NAPI context
490  *
491  * Schedule NAPI poll routine to be called if it is not already
492  * running.
493  * Return true if we schedule a NAPI or false if not.
494  * Refer to napi_schedule_prep() for additional reason on why
495  * a NAPI might not be scheduled.
496  */
497 static inline bool napi_schedule(struct napi_struct *n)
498 {
499 	if (napi_schedule_prep(n)) {
500 		__napi_schedule(n);
501 		return true;
502 	}
503 
504 	return false;
505 }
506 
507 /**
508  *	napi_schedule_irqoff - schedule NAPI poll
509  *	@n: NAPI context
510  *
511  * Variant of napi_schedule(), assuming hard irqs are masked.
512  */
513 static inline void napi_schedule_irqoff(struct napi_struct *n)
514 {
515 	if (napi_schedule_prep(n))
516 		__napi_schedule_irqoff(n);
517 }
518 
519 /**
520  * napi_complete_done - NAPI processing complete
521  * @n: NAPI context
522  * @work_done: number of packets processed
523  *
524  * Mark NAPI processing as complete. Should only be called if poll budget
525  * has not been completely consumed.
526  * Prefer over napi_complete().
527  * Return false if device should avoid rearming interrupts.
528  */
529 bool napi_complete_done(struct napi_struct *n, int work_done);
530 
531 static inline bool napi_complete(struct napi_struct *n)
532 {
533 	return napi_complete_done(n, 0);
534 }
535 
536 int dev_set_threaded(struct net_device *dev, bool threaded);
537 
538 /**
539  *	napi_disable - prevent NAPI from scheduling
540  *	@n: NAPI context
541  *
542  * Stop NAPI from being scheduled on this context.
543  * Waits till any outstanding processing completes.
544  */
545 void napi_disable(struct napi_struct *n);
546 
547 void napi_enable(struct napi_struct *n);
548 
549 /**
550  *	napi_synchronize - wait until NAPI is not running
551  *	@n: NAPI context
552  *
553  * Wait until NAPI is done being scheduled on this context.
554  * Waits till any outstanding processing completes but
555  * does not disable future activations.
556  */
557 static inline void napi_synchronize(const struct napi_struct *n)
558 {
559 	if (IS_ENABLED(CONFIG_SMP))
560 		while (test_bit(NAPI_STATE_SCHED, &n->state))
561 			msleep(1);
562 	else
563 		barrier();
564 }
565 
566 /**
567  *	napi_if_scheduled_mark_missed - if napi is running, set the
568  *	NAPIF_STATE_MISSED
569  *	@n: NAPI context
570  *
571  * If napi is running, set the NAPIF_STATE_MISSED, and return true if
572  * NAPI is scheduled.
573  **/
574 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
575 {
576 	unsigned long val, new;
577 
578 	val = READ_ONCE(n->state);
579 	do {
580 		if (val & NAPIF_STATE_DISABLE)
581 			return true;
582 
583 		if (!(val & NAPIF_STATE_SCHED))
584 			return false;
585 
586 		new = val | NAPIF_STATE_MISSED;
587 	} while (!try_cmpxchg(&n->state, &val, new));
588 
589 	return true;
590 }
591 
592 enum netdev_queue_state_t {
593 	__QUEUE_STATE_DRV_XOFF,
594 	__QUEUE_STATE_STACK_XOFF,
595 	__QUEUE_STATE_FROZEN,
596 };
597 
598 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
599 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
600 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
601 
602 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
603 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
604 					QUEUE_STATE_FROZEN)
605 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
606 					QUEUE_STATE_FROZEN)
607 
608 /*
609  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
610  * netif_tx_* functions below are used to manipulate this flag.  The
611  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
612  * queue independently.  The netif_xmit_*stopped functions below are called
613  * to check if the queue has been stopped by the driver or stack (either
614  * of the XOFF bits are set in the state).  Drivers should not need to call
615  * netif_xmit*stopped functions, they should only be using netif_tx_*.
616  */
617 
618 struct netdev_queue {
619 /*
620  * read-mostly part
621  */
622 	struct net_device	*dev;
623 	netdevice_tracker	dev_tracker;
624 
625 	struct Qdisc __rcu	*qdisc;
626 	struct Qdisc __rcu	*qdisc_sleeping;
627 #ifdef CONFIG_SYSFS
628 	struct kobject		kobj;
629 #endif
630 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
631 	int			numa_node;
632 #endif
633 	unsigned long		tx_maxrate;
634 	/*
635 	 * Number of TX timeouts for this queue
636 	 * (/sys/class/net/DEV/Q/trans_timeout)
637 	 */
638 	atomic_long_t		trans_timeout;
639 
640 	/* Subordinate device that the queue has been assigned to */
641 	struct net_device	*sb_dev;
642 #ifdef CONFIG_XDP_SOCKETS
643 	struct xsk_buff_pool    *pool;
644 #endif
645 /*
646  * write-mostly part
647  */
648 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
649 	int			xmit_lock_owner;
650 	/*
651 	 * Time (in jiffies) of last Tx
652 	 */
653 	unsigned long		trans_start;
654 
655 	unsigned long		state;
656 
657 #ifdef CONFIG_BQL
658 	struct dql		dql;
659 #endif
660 } ____cacheline_aligned_in_smp;
661 
662 extern int sysctl_fb_tunnels_only_for_init_net;
663 extern int sysctl_devconf_inherit_init_net;
664 
665 /*
666  * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
667  *                                     == 1 : For initns only
668  *                                     == 2 : For none.
669  */
670 static inline bool net_has_fallback_tunnels(const struct net *net)
671 {
672 #if IS_ENABLED(CONFIG_SYSCTL)
673 	int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
674 
675 	return !fb_tunnels_only_for_init_net ||
676 		(net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
677 #else
678 	return true;
679 #endif
680 }
681 
682 static inline int net_inherit_devconf(void)
683 {
684 #if IS_ENABLED(CONFIG_SYSCTL)
685 	return READ_ONCE(sysctl_devconf_inherit_init_net);
686 #else
687 	return 0;
688 #endif
689 }
690 
691 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
692 {
693 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
694 	return q->numa_node;
695 #else
696 	return NUMA_NO_NODE;
697 #endif
698 }
699 
700 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
701 {
702 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
703 	q->numa_node = node;
704 #endif
705 }
706 
707 #ifdef CONFIG_RPS
708 /*
709  * This structure holds an RPS map which can be of variable length.  The
710  * map is an array of CPUs.
711  */
712 struct rps_map {
713 	unsigned int len;
714 	struct rcu_head rcu;
715 	u16 cpus[];
716 };
717 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
718 
719 /*
720  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
721  * tail pointer for that CPU's input queue at the time of last enqueue, and
722  * a hardware filter index.
723  */
724 struct rps_dev_flow {
725 	u16 cpu;
726 	u16 filter;
727 	unsigned int last_qtail;
728 };
729 #define RPS_NO_FILTER 0xffff
730 
731 /*
732  * The rps_dev_flow_table structure contains a table of flow mappings.
733  */
734 struct rps_dev_flow_table {
735 	unsigned int mask;
736 	struct rcu_head rcu;
737 	struct rps_dev_flow flows[];
738 };
739 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
740     ((_num) * sizeof(struct rps_dev_flow)))
741 
742 /*
743  * The rps_sock_flow_table contains mappings of flows to the last CPU
744  * on which they were processed by the application (set in recvmsg).
745  * Each entry is a 32bit value. Upper part is the high-order bits
746  * of flow hash, lower part is CPU number.
747  * rps_cpu_mask is used to partition the space, depending on number of
748  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
749  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
750  * meaning we use 32-6=26 bits for the hash.
751  */
752 struct rps_sock_flow_table {
753 	u32	mask;
754 
755 	u32	ents[] ____cacheline_aligned_in_smp;
756 };
757 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
758 
759 #define RPS_NO_CPU 0xffff
760 
761 extern u32 rps_cpu_mask;
762 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
763 
764 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
765 					u32 hash)
766 {
767 	if (table && hash) {
768 		unsigned int index = hash & table->mask;
769 		u32 val = hash & ~rps_cpu_mask;
770 
771 		/* We only give a hint, preemption can change CPU under us */
772 		val |= raw_smp_processor_id();
773 
774 		/* The following WRITE_ONCE() is paired with the READ_ONCE()
775 		 * here, and another one in get_rps_cpu().
776 		 */
777 		if (READ_ONCE(table->ents[index]) != val)
778 			WRITE_ONCE(table->ents[index], val);
779 	}
780 }
781 
782 #ifdef CONFIG_RFS_ACCEL
783 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
784 			 u16 filter_id);
785 #endif
786 #endif /* CONFIG_RPS */
787 
788 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
789 enum xps_map_type {
790 	XPS_CPUS = 0,
791 	XPS_RXQS,
792 	XPS_MAPS_MAX,
793 };
794 
795 #ifdef CONFIG_XPS
796 /*
797  * This structure holds an XPS map which can be of variable length.  The
798  * map is an array of queues.
799  */
800 struct xps_map {
801 	unsigned int len;
802 	unsigned int alloc_len;
803 	struct rcu_head rcu;
804 	u16 queues[];
805 };
806 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
807 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
808        - sizeof(struct xps_map)) / sizeof(u16))
809 
810 /*
811  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
812  *
813  * We keep track of the number of cpus/rxqs used when the struct is allocated,
814  * in nr_ids. This will help not accessing out-of-bound memory.
815  *
816  * We keep track of the number of traffic classes used when the struct is
817  * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
818  * not crossing its upper bound, as the original dev->num_tc can be updated in
819  * the meantime.
820  */
821 struct xps_dev_maps {
822 	struct rcu_head rcu;
823 	unsigned int nr_ids;
824 	s16 num_tc;
825 	struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
826 };
827 
828 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +	\
829 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
830 
831 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
832 	(_rxqs * (_tcs) * sizeof(struct xps_map *)))
833 
834 #endif /* CONFIG_XPS */
835 
836 #define TC_MAX_QUEUE	16
837 #define TC_BITMASK	15
838 /* HW offloaded queuing disciplines txq count and offset maps */
839 struct netdev_tc_txq {
840 	u16 count;
841 	u16 offset;
842 };
843 
844 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
845 /*
846  * This structure is to hold information about the device
847  * configured to run FCoE protocol stack.
848  */
849 struct netdev_fcoe_hbainfo {
850 	char	manufacturer[64];
851 	char	serial_number[64];
852 	char	hardware_version[64];
853 	char	driver_version[64];
854 	char	optionrom_version[64];
855 	char	firmware_version[64];
856 	char	model[256];
857 	char	model_description[256];
858 };
859 #endif
860 
861 #define MAX_PHYS_ITEM_ID_LEN 32
862 
863 /* This structure holds a unique identifier to identify some
864  * physical item (port for example) used by a netdevice.
865  */
866 struct netdev_phys_item_id {
867 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
868 	unsigned char id_len;
869 };
870 
871 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
872 					    struct netdev_phys_item_id *b)
873 {
874 	return a->id_len == b->id_len &&
875 	       memcmp(a->id, b->id, a->id_len) == 0;
876 }
877 
878 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
879 				       struct sk_buff *skb,
880 				       struct net_device *sb_dev);
881 
882 enum net_device_path_type {
883 	DEV_PATH_ETHERNET = 0,
884 	DEV_PATH_VLAN,
885 	DEV_PATH_BRIDGE,
886 	DEV_PATH_PPPOE,
887 	DEV_PATH_DSA,
888 	DEV_PATH_MTK_WDMA,
889 };
890 
891 struct net_device_path {
892 	enum net_device_path_type	type;
893 	const struct net_device		*dev;
894 	union {
895 		struct {
896 			u16		id;
897 			__be16		proto;
898 			u8		h_dest[ETH_ALEN];
899 		} encap;
900 		struct {
901 			enum {
902 				DEV_PATH_BR_VLAN_KEEP,
903 				DEV_PATH_BR_VLAN_TAG,
904 				DEV_PATH_BR_VLAN_UNTAG,
905 				DEV_PATH_BR_VLAN_UNTAG_HW,
906 			}		vlan_mode;
907 			u16		vlan_id;
908 			__be16		vlan_proto;
909 		} bridge;
910 		struct {
911 			int port;
912 			u16 proto;
913 		} dsa;
914 		struct {
915 			u8 wdma_idx;
916 			u8 queue;
917 			u16 wcid;
918 			u8 bss;
919 			u8 amsdu;
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 nlmsghdr *nlh, struct net_device *dev,
1291  *			   struct netlink_ext_ack *extack);
1292  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1293  *		       struct net_device *dev, struct net_device *filter_dev,
1294  *		       int *idx)
1295  *	Used to add FDB entries to dump requests. Implementers should add
1296  *	entries to skb and update idx with the number of entries.
1297  *
1298  * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1299  *		      u16 nlmsg_flags, struct netlink_ext_ack *extack);
1300  *	Adds an MDB entry to dev.
1301  * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1302  *		      struct netlink_ext_ack *extack);
1303  *	Deletes the MDB entry from dev.
1304  * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1305  *		       struct netlink_callback *cb);
1306  *	Dumps MDB entries from dev. The first argument (marker) in the netlink
1307  *	callback is used by core rtnetlink code.
1308  *
1309  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1310  *			     u16 flags, struct netlink_ext_ack *extack)
1311  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1312  *			     struct net_device *dev, u32 filter_mask,
1313  *			     int nlflags)
1314  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1315  *			     u16 flags);
1316  *
1317  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1318  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1319  *	which do not represent real hardware may define this to allow their
1320  *	userspace components to manage their virtual carrier state. Devices
1321  *	that determine carrier state from physical hardware properties (eg
1322  *	network cables) or protocol-dependent mechanisms (eg
1323  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1324  *
1325  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1326  *			       struct netdev_phys_item_id *ppid);
1327  *	Called to get ID of physical port of this device. If driver does
1328  *	not implement this, it is assumed that the hw is not able to have
1329  *	multiple net devices on single physical port.
1330  *
1331  * int (*ndo_get_port_parent_id)(struct net_device *dev,
1332  *				 struct netdev_phys_item_id *ppid)
1333  *	Called to get the parent ID of the physical port of this device.
1334  *
1335  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1336  *				 struct net_device *dev)
1337  *	Called by upper layer devices to accelerate switching or other
1338  *	station functionality into hardware. 'pdev is the lowerdev
1339  *	to use for the offload and 'dev' is the net device that will
1340  *	back the offload. Returns a pointer to the private structure
1341  *	the upper layer will maintain.
1342  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1343  *	Called by upper layer device to delete the station created
1344  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1345  *	the station and priv is the structure returned by the add
1346  *	operation.
1347  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1348  *			     int queue_index, u32 maxrate);
1349  *	Called when a user wants to set a max-rate limitation of specific
1350  *	TX queue.
1351  * int (*ndo_get_iflink)(const struct net_device *dev);
1352  *	Called to get the iflink value of this device.
1353  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1354  *	This function is used to get egress tunnel information for given skb.
1355  *	This is useful for retrieving outer tunnel header parameters while
1356  *	sampling packet.
1357  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1358  *	This function is used to specify the headroom that the skb must
1359  *	consider when allocation skb during packet reception. Setting
1360  *	appropriate rx headroom value allows avoiding skb head copy on
1361  *	forward. Setting a negative value resets the rx headroom to the
1362  *	default value.
1363  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1364  *	This function is used to set or query state related to XDP on the
1365  *	netdevice and manage BPF offload. See definition of
1366  *	enum bpf_netdev_command for details.
1367  * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1368  *			u32 flags);
1369  *	This function is used to submit @n XDP packets for transmit on a
1370  *	netdevice. Returns number of frames successfully transmitted, frames
1371  *	that got dropped are freed/returned via xdp_return_frame().
1372  *	Returns negative number, means general error invoking ndo, meaning
1373  *	no frames were xmit'ed and core-caller will free all frames.
1374  * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1375  *					        struct xdp_buff *xdp);
1376  *      Get the xmit slave of master device based on the xdp_buff.
1377  * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1378  *      This function is used to wake up the softirq, ksoftirqd or kthread
1379  *	responsible for sending and/or receiving packets on a specific
1380  *	queue id bound to an AF_XDP socket. The flags field specifies if
1381  *	only RX, only Tx, or both should be woken up using the flags
1382  *	XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1383  * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1384  *			 int cmd);
1385  *	Add, change, delete or get information on an IPv4 tunnel.
1386  * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1387  *	If a device is paired with a peer device, return the peer instance.
1388  *	The caller must be under RCU read context.
1389  * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1390  *     Get the forwarding path to reach the real device from the HW destination address
1391  * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1392  *			     const struct skb_shared_hwtstamps *hwtstamps,
1393  *			     bool cycles);
1394  *	Get hardware timestamp based on normal/adjustable time or free running
1395  *	cycle counter. This function is required if physical clock supports a
1396  *	free running cycle counter.
1397  *
1398  * int (*ndo_hwtstamp_get)(struct net_device *dev,
1399  *			   struct kernel_hwtstamp_config *kernel_config);
1400  *	Get the currently configured hardware timestamping parameters for the
1401  *	NIC device.
1402  *
1403  * int (*ndo_hwtstamp_set)(struct net_device *dev,
1404  *			   struct kernel_hwtstamp_config *kernel_config,
1405  *			   struct netlink_ext_ack *extack);
1406  *	Change the hardware timestamping parameters for NIC device.
1407  */
1408 struct net_device_ops {
1409 	int			(*ndo_init)(struct net_device *dev);
1410 	void			(*ndo_uninit)(struct net_device *dev);
1411 	int			(*ndo_open)(struct net_device *dev);
1412 	int			(*ndo_stop)(struct net_device *dev);
1413 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1414 						  struct net_device *dev);
1415 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1416 						      struct net_device *dev,
1417 						      netdev_features_t features);
1418 	u16			(*ndo_select_queue)(struct net_device *dev,
1419 						    struct sk_buff *skb,
1420 						    struct net_device *sb_dev);
1421 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1422 						       int flags);
1423 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1424 	int			(*ndo_set_mac_address)(struct net_device *dev,
1425 						       void *addr);
1426 	int			(*ndo_validate_addr)(struct net_device *dev);
1427 	int			(*ndo_do_ioctl)(struct net_device *dev,
1428 					        struct ifreq *ifr, int cmd);
1429 	int			(*ndo_eth_ioctl)(struct net_device *dev,
1430 						 struct ifreq *ifr, int cmd);
1431 	int			(*ndo_siocbond)(struct net_device *dev,
1432 						struct ifreq *ifr, int cmd);
1433 	int			(*ndo_siocwandev)(struct net_device *dev,
1434 						  struct if_settings *ifs);
1435 	int			(*ndo_siocdevprivate)(struct net_device *dev,
1436 						      struct ifreq *ifr,
1437 						      void __user *data, int cmd);
1438 	int			(*ndo_set_config)(struct net_device *dev,
1439 					          struct ifmap *map);
1440 	int			(*ndo_change_mtu)(struct net_device *dev,
1441 						  int new_mtu);
1442 	int			(*ndo_neigh_setup)(struct net_device *dev,
1443 						   struct neigh_parms *);
1444 	void			(*ndo_tx_timeout) (struct net_device *dev,
1445 						   unsigned int txqueue);
1446 
1447 	void			(*ndo_get_stats64)(struct net_device *dev,
1448 						   struct rtnl_link_stats64 *storage);
1449 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1450 	int			(*ndo_get_offload_stats)(int attr_id,
1451 							 const struct net_device *dev,
1452 							 void *attr_data);
1453 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1454 
1455 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1456 						       __be16 proto, u16 vid);
1457 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1458 						        __be16 proto, u16 vid);
1459 #ifdef CONFIG_NET_POLL_CONTROLLER
1460 	void                    (*ndo_poll_controller)(struct net_device *dev);
1461 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1462 						     struct netpoll_info *info);
1463 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1464 #endif
1465 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1466 						  int queue, u8 *mac);
1467 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1468 						   int queue, u16 vlan,
1469 						   u8 qos, __be16 proto);
1470 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1471 						   int vf, int min_tx_rate,
1472 						   int max_tx_rate);
1473 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1474 						       int vf, bool setting);
1475 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1476 						    int vf, bool setting);
1477 	int			(*ndo_get_vf_config)(struct net_device *dev,
1478 						     int vf,
1479 						     struct ifla_vf_info *ivf);
1480 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1481 							 int vf, int link_state);
1482 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1483 						    int vf,
1484 						    struct ifla_vf_stats
1485 						    *vf_stats);
1486 	int			(*ndo_set_vf_port)(struct net_device *dev,
1487 						   int vf,
1488 						   struct nlattr *port[]);
1489 	int			(*ndo_get_vf_port)(struct net_device *dev,
1490 						   int vf, struct sk_buff *skb);
1491 	int			(*ndo_get_vf_guid)(struct net_device *dev,
1492 						   int vf,
1493 						   struct ifla_vf_guid *node_guid,
1494 						   struct ifla_vf_guid *port_guid);
1495 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1496 						   int vf, u64 guid,
1497 						   int guid_type);
1498 	int			(*ndo_set_vf_rss_query_en)(
1499 						   struct net_device *dev,
1500 						   int vf, bool setting);
1501 	int			(*ndo_setup_tc)(struct net_device *dev,
1502 						enum tc_setup_type type,
1503 						void *type_data);
1504 #if IS_ENABLED(CONFIG_FCOE)
1505 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1506 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1507 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1508 						      u16 xid,
1509 						      struct scatterlist *sgl,
1510 						      unsigned int sgc);
1511 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1512 						     u16 xid);
1513 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1514 						       u16 xid,
1515 						       struct scatterlist *sgl,
1516 						       unsigned int sgc);
1517 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1518 							struct netdev_fcoe_hbainfo *hbainfo);
1519 #endif
1520 
1521 #if IS_ENABLED(CONFIG_LIBFCOE)
1522 #define NETDEV_FCOE_WWNN 0
1523 #define NETDEV_FCOE_WWPN 1
1524 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1525 						    u64 *wwn, int type);
1526 #endif
1527 
1528 #ifdef CONFIG_RFS_ACCEL
1529 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1530 						     const struct sk_buff *skb,
1531 						     u16 rxq_index,
1532 						     u32 flow_id);
1533 #endif
1534 	int			(*ndo_add_slave)(struct net_device *dev,
1535 						 struct net_device *slave_dev,
1536 						 struct netlink_ext_ack *extack);
1537 	int			(*ndo_del_slave)(struct net_device *dev,
1538 						 struct net_device *slave_dev);
1539 	struct net_device*	(*ndo_get_xmit_slave)(struct net_device *dev,
1540 						      struct sk_buff *skb,
1541 						      bool all_slaves);
1542 	struct net_device*	(*ndo_sk_get_lower_dev)(struct net_device *dev,
1543 							struct sock *sk);
1544 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1545 						    netdev_features_t features);
1546 	int			(*ndo_set_features)(struct net_device *dev,
1547 						    netdev_features_t features);
1548 	int			(*ndo_neigh_construct)(struct net_device *dev,
1549 						       struct neighbour *n);
1550 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1551 						     struct neighbour *n);
1552 
1553 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1554 					       struct nlattr *tb[],
1555 					       struct net_device *dev,
1556 					       const unsigned char *addr,
1557 					       u16 vid,
1558 					       u16 flags,
1559 					       struct netlink_ext_ack *extack);
1560 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1561 					       struct nlattr *tb[],
1562 					       struct net_device *dev,
1563 					       const unsigned char *addr,
1564 					       u16 vid, struct netlink_ext_ack *extack);
1565 	int			(*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1566 						    struct net_device *dev,
1567 						    struct netlink_ext_ack *extack);
1568 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1569 						struct netlink_callback *cb,
1570 						struct net_device *dev,
1571 						struct net_device *filter_dev,
1572 						int *idx);
1573 	int			(*ndo_fdb_get)(struct sk_buff *skb,
1574 					       struct nlattr *tb[],
1575 					       struct net_device *dev,
1576 					       const unsigned char *addr,
1577 					       u16 vid, u32 portid, u32 seq,
1578 					       struct netlink_ext_ack *extack);
1579 	int			(*ndo_mdb_add)(struct net_device *dev,
1580 					       struct nlattr *tb[],
1581 					       u16 nlmsg_flags,
1582 					       struct netlink_ext_ack *extack);
1583 	int			(*ndo_mdb_del)(struct net_device *dev,
1584 					       struct nlattr *tb[],
1585 					       struct netlink_ext_ack *extack);
1586 	int			(*ndo_mdb_dump)(struct net_device *dev,
1587 						struct sk_buff *skb,
1588 						struct netlink_callback *cb);
1589 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1590 						      struct nlmsghdr *nlh,
1591 						      u16 flags,
1592 						      struct netlink_ext_ack *extack);
1593 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1594 						      u32 pid, u32 seq,
1595 						      struct net_device *dev,
1596 						      u32 filter_mask,
1597 						      int nlflags);
1598 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1599 						      struct nlmsghdr *nlh,
1600 						      u16 flags);
1601 	int			(*ndo_change_carrier)(struct net_device *dev,
1602 						      bool new_carrier);
1603 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1604 							struct netdev_phys_item_id *ppid);
1605 	int			(*ndo_get_port_parent_id)(struct net_device *dev,
1606 							  struct netdev_phys_item_id *ppid);
1607 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1608 							  char *name, size_t len);
1609 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1610 							struct net_device *dev);
1611 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1612 							void *priv);
1613 
1614 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1615 						      int queue_index,
1616 						      u32 maxrate);
1617 	int			(*ndo_get_iflink)(const struct net_device *dev);
1618 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1619 						       struct sk_buff *skb);
1620 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1621 						       int needed_headroom);
1622 	int			(*ndo_bpf)(struct net_device *dev,
1623 					   struct netdev_bpf *bpf);
1624 	int			(*ndo_xdp_xmit)(struct net_device *dev, int n,
1625 						struct xdp_frame **xdp,
1626 						u32 flags);
1627 	struct net_device *	(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1628 							  struct xdp_buff *xdp);
1629 	int			(*ndo_xsk_wakeup)(struct net_device *dev,
1630 						  u32 queue_id, u32 flags);
1631 	int			(*ndo_tunnel_ctl)(struct net_device *dev,
1632 						  struct ip_tunnel_parm *p, int cmd);
1633 	struct net_device *	(*ndo_get_peer_dev)(struct net_device *dev);
1634 	int                     (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1635                                                          struct net_device_path *path);
1636 	ktime_t			(*ndo_get_tstamp)(struct net_device *dev,
1637 						  const struct skb_shared_hwtstamps *hwtstamps,
1638 						  bool cycles);
1639 	int			(*ndo_hwtstamp_get)(struct net_device *dev,
1640 						    struct kernel_hwtstamp_config *kernel_config);
1641 	int			(*ndo_hwtstamp_set)(struct net_device *dev,
1642 						    struct kernel_hwtstamp_config *kernel_config,
1643 						    struct netlink_ext_ack *extack);
1644 };
1645 
1646 /**
1647  * enum netdev_priv_flags - &struct net_device priv_flags
1648  *
1649  * These are the &struct net_device, they are only set internally
1650  * by drivers and used in the kernel. These flags are invisible to
1651  * userspace; this means that the order of these flags can change
1652  * during any kernel release.
1653  *
1654  * You should have a pretty good reason to be extending these flags.
1655  *
1656  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1657  * @IFF_EBRIDGE: Ethernet bridging device
1658  * @IFF_BONDING: bonding master or slave
1659  * @IFF_ISATAP: ISATAP interface (RFC4214)
1660  * @IFF_WAN_HDLC: WAN HDLC device
1661  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1662  *	release skb->dst
1663  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1664  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1665  * @IFF_MACVLAN_PORT: device used as macvlan port
1666  * @IFF_BRIDGE_PORT: device used as bridge port
1667  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1668  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1669  * @IFF_UNICAST_FLT: Supports unicast filtering
1670  * @IFF_TEAM_PORT: device used as team port
1671  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1672  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1673  *	change when it's running
1674  * @IFF_MACVLAN: Macvlan device
1675  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1676  *	underlying stacked devices
1677  * @IFF_L3MDEV_MASTER: device is an L3 master device
1678  * @IFF_NO_QUEUE: device can run without qdisc attached
1679  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1680  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1681  * @IFF_TEAM: device is a team device
1682  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1683  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1684  *	entity (i.e. the master device for bridged veth)
1685  * @IFF_MACSEC: device is a MACsec device
1686  * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1687  * @IFF_FAILOVER: device is a failover master device
1688  * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1689  * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1690  * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1691  * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1692  *	skb_headlen(skb) == 0 (data starts from frag0)
1693  * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1694  * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1695  *	ndo_hwtstamp_set() for all timestamp requests regardless of source,
1696  *	even if those aren't HWTSTAMP_SOURCE_NETDEV.
1697  */
1698 enum netdev_priv_flags {
1699 	IFF_802_1Q_VLAN			= 1<<0,
1700 	IFF_EBRIDGE			= 1<<1,
1701 	IFF_BONDING			= 1<<2,
1702 	IFF_ISATAP			= 1<<3,
1703 	IFF_WAN_HDLC			= 1<<4,
1704 	IFF_XMIT_DST_RELEASE		= 1<<5,
1705 	IFF_DONT_BRIDGE			= 1<<6,
1706 	IFF_DISABLE_NETPOLL		= 1<<7,
1707 	IFF_MACVLAN_PORT		= 1<<8,
1708 	IFF_BRIDGE_PORT			= 1<<9,
1709 	IFF_OVS_DATAPATH		= 1<<10,
1710 	IFF_TX_SKB_SHARING		= 1<<11,
1711 	IFF_UNICAST_FLT			= 1<<12,
1712 	IFF_TEAM_PORT			= 1<<13,
1713 	IFF_SUPP_NOFCS			= 1<<14,
1714 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1715 	IFF_MACVLAN			= 1<<16,
1716 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1717 	IFF_L3MDEV_MASTER		= 1<<18,
1718 	IFF_NO_QUEUE			= 1<<19,
1719 	IFF_OPENVSWITCH			= 1<<20,
1720 	IFF_L3MDEV_SLAVE		= 1<<21,
1721 	IFF_TEAM			= 1<<22,
1722 	IFF_RXFH_CONFIGURED		= 1<<23,
1723 	IFF_PHONY_HEADROOM		= 1<<24,
1724 	IFF_MACSEC			= 1<<25,
1725 	IFF_NO_RX_HANDLER		= 1<<26,
1726 	IFF_FAILOVER			= 1<<27,
1727 	IFF_FAILOVER_SLAVE		= 1<<28,
1728 	IFF_L3MDEV_RX_HANDLER		= 1<<29,
1729 	IFF_NO_ADDRCONF			= BIT_ULL(30),
1730 	IFF_TX_SKB_NO_LINEAR		= BIT_ULL(31),
1731 	IFF_CHANGE_PROTO_DOWN		= BIT_ULL(32),
1732 	IFF_SEE_ALL_HWTSTAMP_REQUESTS	= BIT_ULL(33),
1733 };
1734 
1735 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1736 #define IFF_EBRIDGE			IFF_EBRIDGE
1737 #define IFF_BONDING			IFF_BONDING
1738 #define IFF_ISATAP			IFF_ISATAP
1739 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1740 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1741 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1742 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1743 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1744 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1745 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1746 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1747 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1748 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1749 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1750 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1751 #define IFF_MACVLAN			IFF_MACVLAN
1752 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1753 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1754 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1755 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1756 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1757 #define IFF_TEAM			IFF_TEAM
1758 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1759 #define IFF_PHONY_HEADROOM		IFF_PHONY_HEADROOM
1760 #define IFF_MACSEC			IFF_MACSEC
1761 #define IFF_NO_RX_HANDLER		IFF_NO_RX_HANDLER
1762 #define IFF_FAILOVER			IFF_FAILOVER
1763 #define IFF_FAILOVER_SLAVE		IFF_FAILOVER_SLAVE
1764 #define IFF_L3MDEV_RX_HANDLER		IFF_L3MDEV_RX_HANDLER
1765 #define IFF_TX_SKB_NO_LINEAR		IFF_TX_SKB_NO_LINEAR
1766 
1767 /* Specifies the type of the struct net_device::ml_priv pointer */
1768 enum netdev_ml_priv_type {
1769 	ML_PRIV_NONE,
1770 	ML_PRIV_CAN,
1771 };
1772 
1773 /**
1774  *	struct net_device - The DEVICE structure.
1775  *
1776  *	Actually, this whole structure is a big mistake.  It mixes I/O
1777  *	data with strictly "high-level" data, and it has to know about
1778  *	almost every data structure used in the INET module.
1779  *
1780  *	@name:	This is the first field of the "visible" part of this structure
1781  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1782  *		of the interface.
1783  *
1784  *	@name_node:	Name hashlist node
1785  *	@ifalias:	SNMP alias
1786  *	@mem_end:	Shared memory end
1787  *	@mem_start:	Shared memory start
1788  *	@base_addr:	Device I/O address
1789  *	@irq:		Device IRQ number
1790  *
1791  *	@state:		Generic network queuing layer state, see netdev_state_t
1792  *	@dev_list:	The global list of network devices
1793  *	@napi_list:	List entry used for polling NAPI devices
1794  *	@unreg_list:	List entry  when we are unregistering the
1795  *			device; see the function unregister_netdev
1796  *	@close_list:	List entry used when we are closing the device
1797  *	@ptype_all:     Device-specific packet handlers for all protocols
1798  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1799  *
1800  *	@adj_list:	Directly linked devices, like slaves for bonding
1801  *	@features:	Currently active device features
1802  *	@hw_features:	User-changeable features
1803  *
1804  *	@wanted_features:	User-requested features
1805  *	@vlan_features:		Mask of features inheritable by VLAN devices
1806  *
1807  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1808  *				This field indicates what encapsulation
1809  *				offloads the hardware is capable of doing,
1810  *				and drivers will need to set them appropriately.
1811  *
1812  *	@mpls_features:	Mask of features inheritable by MPLS
1813  *	@gso_partial_features: value(s) from NETIF_F_GSO\*
1814  *
1815  *	@ifindex:	interface index
1816  *	@group:		The group the device belongs to
1817  *
1818  *	@stats:		Statistics struct, which was left as a legacy, use
1819  *			rtnl_link_stats64 instead
1820  *
1821  *	@core_stats:	core networking counters,
1822  *			do not use this in drivers
1823  *	@carrier_up_count:	Number of times the carrier has been up
1824  *	@carrier_down_count:	Number of times the carrier has been down
1825  *
1826  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1827  *				instead of ioctl,
1828  *				see <net/iw_handler.h> for details.
1829  *	@wireless_data:	Instance data managed by the core of wireless extensions
1830  *
1831  *	@netdev_ops:	Includes several pointers to callbacks,
1832  *			if one wants to override the ndo_*() functions
1833  *	@xdp_metadata_ops:	Includes pointers to XDP metadata callbacks.
1834  *	@ethtool_ops:	Management operations
1835  *	@l3mdev_ops:	Layer 3 master device operations
1836  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1837  *			discovery handling. Necessary for e.g. 6LoWPAN.
1838  *	@xfrmdev_ops:	Transformation offload operations
1839  *	@tlsdev_ops:	Transport Layer Security offload operations
1840  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1841  *			of Layer 2 headers.
1842  *
1843  *	@flags:		Interface flags (a la BSD)
1844  *	@xdp_features:	XDP capability supported by the device
1845  *	@priv_flags:	Like 'flags' but invisible to userspace,
1846  *			see if.h for the definitions
1847  *	@gflags:	Global flags ( kept as legacy )
1848  *	@padded:	How much padding added by alloc_netdev()
1849  *	@operstate:	RFC2863 operstate
1850  *	@link_mode:	Mapping policy to operstate
1851  *	@if_port:	Selectable AUI, TP, ...
1852  *	@dma:		DMA channel
1853  *	@mtu:		Interface MTU value
1854  *	@min_mtu:	Interface Minimum MTU value
1855  *	@max_mtu:	Interface Maximum MTU value
1856  *	@type:		Interface hardware type
1857  *	@hard_header_len: Maximum hardware header length.
1858  *	@min_header_len:  Minimum hardware header length
1859  *
1860  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1861  *			  cases can this be guaranteed
1862  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1863  *			  cases can this be guaranteed. Some cases also use
1864  *			  LL_MAX_HEADER instead to allocate the skb
1865  *
1866  *	interface address info:
1867  *
1868  * 	@perm_addr:		Permanent hw address
1869  * 	@addr_assign_type:	Hw address assignment type
1870  * 	@addr_len:		Hardware address length
1871  *	@upper_level:		Maximum depth level of upper devices.
1872  *	@lower_level:		Maximum depth level of lower devices.
1873  *	@neigh_priv_len:	Used in neigh_alloc()
1874  * 	@dev_id:		Used to differentiate devices that share
1875  * 				the same link layer address
1876  * 	@dev_port:		Used to differentiate devices that share
1877  * 				the same function
1878  *	@addr_list_lock:	XXX: need comments on this one
1879  *	@name_assign_type:	network interface name assignment type
1880  *	@uc_promisc:		Counter that indicates promiscuous mode
1881  *				has been enabled due to the need to listen to
1882  *				additional unicast addresses in a device that
1883  *				does not implement ndo_set_rx_mode()
1884  *	@uc:			unicast mac addresses
1885  *	@mc:			multicast mac addresses
1886  *	@dev_addrs:		list of device hw addresses
1887  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1888  *	@promiscuity:		Number of times the NIC is told to work in
1889  *				promiscuous mode; if it becomes 0 the NIC will
1890  *				exit promiscuous mode
1891  *	@allmulti:		Counter, enables or disables allmulticast mode
1892  *
1893  *	@vlan_info:	VLAN info
1894  *	@dsa_ptr:	dsa specific data
1895  *	@tipc_ptr:	TIPC specific data
1896  *	@atalk_ptr:	AppleTalk link
1897  *	@ip_ptr:	IPv4 specific data
1898  *	@ip6_ptr:	IPv6 specific data
1899  *	@ax25_ptr:	AX.25 specific data
1900  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1901  *	@ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1902  *			 device struct
1903  *	@mpls_ptr:	mpls_dev struct pointer
1904  *	@mctp_ptr:	MCTP specific data
1905  *
1906  *	@dev_addr:	Hw address (before bcast,
1907  *			because most packets are unicast)
1908  *
1909  *	@_rx:			Array of RX queues
1910  *	@num_rx_queues:		Number of RX queues
1911  *				allocated at register_netdev() time
1912  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1913  *	@xdp_prog:		XDP sockets filter program pointer
1914  *	@gro_flush_timeout:	timeout for GRO layer in NAPI
1915  *	@napi_defer_hard_irqs:	If not zero, provides a counter that would
1916  *				allow to avoid NIC hard IRQ, on busy queues.
1917  *
1918  *	@rx_handler:		handler for received packets
1919  *	@rx_handler_data: 	XXX: need comments on this one
1920  *	@tcx_ingress:		BPF & clsact qdisc specific data for ingress processing
1921  *	@ingress_queue:		XXX: need comments on this one
1922  *	@nf_hooks_ingress:	netfilter hooks executed for ingress packets
1923  *	@broadcast:		hw bcast address
1924  *
1925  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1926  *			indexed by RX queue number. Assigned by driver.
1927  *			This must only be set if the ndo_rx_flow_steer
1928  *			operation is defined
1929  *	@index_hlist:		Device index hash chain
1930  *
1931  *	@_tx:			Array of TX queues
1932  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1933  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1934  *	@qdisc:			Root qdisc from userspace point of view
1935  *	@tx_queue_len:		Max frames per queue allowed
1936  *	@tx_global_lock: 	XXX: need comments on this one
1937  *	@xdp_bulkq:		XDP device bulk queue
1938  *	@xps_maps:		all CPUs/RXQs maps for XPS device
1939  *
1940  *	@xps_maps:	XXX: need comments on this one
1941  *	@tcx_egress:		BPF & clsact qdisc specific data for egress processing
1942  *	@nf_hooks_egress:	netfilter hooks executed for egress packets
1943  *	@qdisc_hash:		qdisc hash table
1944  *	@watchdog_timeo:	Represents the timeout that is used by
1945  *				the watchdog (see dev_watchdog())
1946  *	@watchdog_timer:	List of timers
1947  *
1948  *	@proto_down_reason:	reason a netdev interface is held down
1949  *	@pcpu_refcnt:		Number of references to this device
1950  *	@dev_refcnt:		Number of references to this device
1951  *	@refcnt_tracker:	Tracker directory for tracked references to this device
1952  *	@todo_list:		Delayed register/unregister
1953  *	@link_watch_list:	XXX: need comments on this one
1954  *
1955  *	@reg_state:		Register/unregister state machine
1956  *	@dismantle:		Device is going to be freed
1957  *	@rtnl_link_state:	This enum represents the phases of creating
1958  *				a new link
1959  *
1960  *	@needs_free_netdev:	Should unregister perform free_netdev?
1961  *	@priv_destructor:	Called from unregister
1962  *	@npinfo:		XXX: need comments on this one
1963  * 	@nd_net:		Network namespace this network device is inside
1964  *
1965  * 	@ml_priv:	Mid-layer private
1966  *	@ml_priv_type:  Mid-layer private type
1967  * 	@lstats:	Loopback statistics
1968  * 	@tstats:	Tunnel statistics
1969  * 	@dstats:	Dummy statistics
1970  * 	@vstats:	Virtual ethernet statistics
1971  *
1972  *	@garp_port:	GARP
1973  *	@mrp_port:	MRP
1974  *
1975  *	@dm_private:	Drop monitor private
1976  *
1977  *	@dev:		Class/net/name entry
1978  *	@sysfs_groups:	Space for optional device, statistics and wireless
1979  *			sysfs groups
1980  *
1981  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1982  *	@rtnl_link_ops:	Rtnl_link_ops
1983  *
1984  *	@gso_max_size:	Maximum size of generic segmentation offload
1985  *	@tso_max_size:	Device (as in HW) limit on the max TSO request size
1986  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1987  *			NIC for GSO
1988  *	@tso_max_segs:	Device (as in HW) limit on the max TSO segment count
1989  * 	@gso_ipv4_max_size:	Maximum size of generic segmentation offload,
1990  * 				for IPv4.
1991  *
1992  *	@dcbnl_ops:	Data Center Bridging netlink ops
1993  *	@num_tc:	Number of traffic classes in the net device
1994  *	@tc_to_txq:	XXX: need comments on this one
1995  *	@prio_tc_map:	XXX: need comments on this one
1996  *
1997  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1998  *
1999  *	@priomap:	XXX: need comments on this one
2000  *	@phydev:	Physical device may attach itself
2001  *			for hardware timestamping
2002  *	@sfp_bus:	attached &struct sfp_bus structure.
2003  *
2004  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2005  *
2006  *	@proto_down:	protocol port state information can be sent to the
2007  *			switch driver and used to set the phys state of the
2008  *			switch port.
2009  *
2010  *	@wol_enabled:	Wake-on-LAN is enabled
2011  *
2012  *	@threaded:	napi threaded mode is enabled
2013  *
2014  *	@net_notifier_list:	List of per-net netdev notifier block
2015  *				that follow this device when it is moved
2016  *				to another network namespace.
2017  *
2018  *	@macsec_ops:    MACsec offloading ops
2019  *
2020  *	@udp_tunnel_nic_info:	static structure describing the UDP tunnel
2021  *				offload capabilities of the device
2022  *	@udp_tunnel_nic:	UDP tunnel offload state
2023  *	@xdp_state:		stores info on attached XDP BPF programs
2024  *
2025  *	@nested_level:	Used as a parameter of spin_lock_nested() of
2026  *			dev->addr_list_lock.
2027  *	@unlink_list:	As netif_addr_lock() can be called recursively,
2028  *			keep a list of interfaces to be deleted.
2029  *	@gro_max_size:	Maximum size of aggregated packet in generic
2030  *			receive offload (GRO)
2031  * 	@gro_ipv4_max_size:	Maximum size of aggregated packet in generic
2032  * 				receive offload (GRO), for IPv4.
2033  *	@xdp_zc_max_segs:	Maximum number of segments supported by AF_XDP
2034  *				zero copy driver
2035  *
2036  *	@dev_addr_shadow:	Copy of @dev_addr to catch direct writes.
2037  *	@linkwatch_dev_tracker:	refcount tracker used by linkwatch.
2038  *	@watchdog_dev_tracker:	refcount tracker used by watchdog.
2039  *	@dev_registered_tracker:	tracker for reference held while
2040  *					registered
2041  *	@offload_xstats_l3:	L3 HW stats for this netdevice.
2042  *
2043  *	@devlink_port:	Pointer to related devlink port structure.
2044  *			Assigned by a driver before netdev registration using
2045  *			SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2046  *			during the time netdevice is registered.
2047  *
2048  *	@dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2049  *		   where the clock is recovered.
2050  *
2051  *	FIXME: cleanup struct net_device such that network protocol info
2052  *	moves out.
2053  */
2054 
2055 struct net_device {
2056 	char			name[IFNAMSIZ];
2057 	struct netdev_name_node	*name_node;
2058 	struct dev_ifalias	__rcu *ifalias;
2059 	/*
2060 	 *	I/O specific fields
2061 	 *	FIXME: Merge these and struct ifmap into one
2062 	 */
2063 	unsigned long		mem_end;
2064 	unsigned long		mem_start;
2065 	unsigned long		base_addr;
2066 
2067 	/*
2068 	 *	Some hardware also needs these fields (state,dev_list,
2069 	 *	napi_list,unreg_list,close_list) but they are not
2070 	 *	part of the usual set specified in Space.c.
2071 	 */
2072 
2073 	unsigned long		state;
2074 
2075 	struct list_head	dev_list;
2076 	struct list_head	napi_list;
2077 	struct list_head	unreg_list;
2078 	struct list_head	close_list;
2079 	struct list_head	ptype_all;
2080 	struct list_head	ptype_specific;
2081 
2082 	struct {
2083 		struct list_head upper;
2084 		struct list_head lower;
2085 	} adj_list;
2086 
2087 	/* Read-mostly cache-line for fast-path access */
2088 	unsigned int		flags;
2089 	xdp_features_t		xdp_features;
2090 	unsigned long long	priv_flags;
2091 	const struct net_device_ops *netdev_ops;
2092 	const struct xdp_metadata_ops *xdp_metadata_ops;
2093 	int			ifindex;
2094 	unsigned short		gflags;
2095 	unsigned short		hard_header_len;
2096 
2097 	/* Note : dev->mtu is often read without holding a lock.
2098 	 * Writers usually hold RTNL.
2099 	 * It is recommended to use READ_ONCE() to annotate the reads,
2100 	 * and to use WRITE_ONCE() to annotate the writes.
2101 	 */
2102 	unsigned int		mtu;
2103 	unsigned short		needed_headroom;
2104 	unsigned short		needed_tailroom;
2105 
2106 	netdev_features_t	features;
2107 	netdev_features_t	hw_features;
2108 	netdev_features_t	wanted_features;
2109 	netdev_features_t	vlan_features;
2110 	netdev_features_t	hw_enc_features;
2111 	netdev_features_t	mpls_features;
2112 	netdev_features_t	gso_partial_features;
2113 
2114 	unsigned int		min_mtu;
2115 	unsigned int		max_mtu;
2116 	unsigned short		type;
2117 	unsigned char		min_header_len;
2118 	unsigned char		name_assign_type;
2119 
2120 	int			group;
2121 
2122 	struct net_device_stats	stats; /* not used by modern drivers */
2123 
2124 	struct net_device_core_stats __percpu *core_stats;
2125 
2126 	/* Stats to monitor link on/off, flapping */
2127 	atomic_t		carrier_up_count;
2128 	atomic_t		carrier_down_count;
2129 
2130 #ifdef CONFIG_WIRELESS_EXT
2131 	const struct iw_handler_def *wireless_handlers;
2132 	struct iw_public_data	*wireless_data;
2133 #endif
2134 	const struct ethtool_ops *ethtool_ops;
2135 #ifdef CONFIG_NET_L3_MASTER_DEV
2136 	const struct l3mdev_ops	*l3mdev_ops;
2137 #endif
2138 #if IS_ENABLED(CONFIG_IPV6)
2139 	const struct ndisc_ops *ndisc_ops;
2140 #endif
2141 
2142 #ifdef CONFIG_XFRM_OFFLOAD
2143 	const struct xfrmdev_ops *xfrmdev_ops;
2144 #endif
2145 
2146 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2147 	const struct tlsdev_ops *tlsdev_ops;
2148 #endif
2149 
2150 	const struct header_ops *header_ops;
2151 
2152 	unsigned char		operstate;
2153 	unsigned char		link_mode;
2154 
2155 	unsigned char		if_port;
2156 	unsigned char		dma;
2157 
2158 	/* Interface address info. */
2159 	unsigned char		perm_addr[MAX_ADDR_LEN];
2160 	unsigned char		addr_assign_type;
2161 	unsigned char		addr_len;
2162 	unsigned char		upper_level;
2163 	unsigned char		lower_level;
2164 
2165 	unsigned short		neigh_priv_len;
2166 	unsigned short          dev_id;
2167 	unsigned short          dev_port;
2168 	unsigned short		padded;
2169 
2170 	spinlock_t		addr_list_lock;
2171 	int			irq;
2172 
2173 	struct netdev_hw_addr_list	uc;
2174 	struct netdev_hw_addr_list	mc;
2175 	struct netdev_hw_addr_list	dev_addrs;
2176 
2177 #ifdef CONFIG_SYSFS
2178 	struct kset		*queues_kset;
2179 #endif
2180 #ifdef CONFIG_LOCKDEP
2181 	struct list_head	unlink_list;
2182 #endif
2183 	unsigned int		promiscuity;
2184 	unsigned int		allmulti;
2185 	bool			uc_promisc;
2186 #ifdef CONFIG_LOCKDEP
2187 	unsigned char		nested_level;
2188 #endif
2189 
2190 
2191 	/* Protocol-specific pointers */
2192 
2193 	struct in_device __rcu	*ip_ptr;
2194 	struct inet6_dev __rcu	*ip6_ptr;
2195 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2196 	struct vlan_info __rcu	*vlan_info;
2197 #endif
2198 #if IS_ENABLED(CONFIG_NET_DSA)
2199 	struct dsa_port		*dsa_ptr;
2200 #endif
2201 #if IS_ENABLED(CONFIG_TIPC)
2202 	struct tipc_bearer __rcu *tipc_ptr;
2203 #endif
2204 #if IS_ENABLED(CONFIG_ATALK)
2205 	void 			*atalk_ptr;
2206 #endif
2207 #if IS_ENABLED(CONFIG_AX25)
2208 	void			*ax25_ptr;
2209 #endif
2210 #if IS_ENABLED(CONFIG_CFG80211)
2211 	struct wireless_dev	*ieee80211_ptr;
2212 #endif
2213 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2214 	struct wpan_dev		*ieee802154_ptr;
2215 #endif
2216 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2217 	struct mpls_dev __rcu	*mpls_ptr;
2218 #endif
2219 #if IS_ENABLED(CONFIG_MCTP)
2220 	struct mctp_dev __rcu	*mctp_ptr;
2221 #endif
2222 
2223 /*
2224  * Cache lines mostly used on receive path (including eth_type_trans())
2225  */
2226 	/* Interface address info used in eth_type_trans() */
2227 	const unsigned char	*dev_addr;
2228 
2229 	struct netdev_rx_queue	*_rx;
2230 	unsigned int		num_rx_queues;
2231 	unsigned int		real_num_rx_queues;
2232 
2233 	struct bpf_prog __rcu	*xdp_prog;
2234 	unsigned long		gro_flush_timeout;
2235 	int			napi_defer_hard_irqs;
2236 #define GRO_LEGACY_MAX_SIZE	65536u
2237 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2238  * and shinfo->gso_segs is a 16bit field.
2239  */
2240 #define GRO_MAX_SIZE		(8 * 65535u)
2241 	unsigned int		gro_max_size;
2242 	unsigned int		gro_ipv4_max_size;
2243 	unsigned int		xdp_zc_max_segs;
2244 	rx_handler_func_t __rcu	*rx_handler;
2245 	void __rcu		*rx_handler_data;
2246 #ifdef CONFIG_NET_XGRESS
2247 	struct bpf_mprog_entry __rcu *tcx_ingress;
2248 #endif
2249 	struct netdev_queue __rcu *ingress_queue;
2250 #ifdef CONFIG_NETFILTER_INGRESS
2251 	struct nf_hook_entries __rcu *nf_hooks_ingress;
2252 #endif
2253 
2254 	unsigned char		broadcast[MAX_ADDR_LEN];
2255 #ifdef CONFIG_RFS_ACCEL
2256 	struct cpu_rmap		*rx_cpu_rmap;
2257 #endif
2258 	struct hlist_node	index_hlist;
2259 
2260 /*
2261  * Cache lines mostly used on transmit path
2262  */
2263 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
2264 	unsigned int		num_tx_queues;
2265 	unsigned int		real_num_tx_queues;
2266 	struct Qdisc __rcu	*qdisc;
2267 	unsigned int		tx_queue_len;
2268 	spinlock_t		tx_global_lock;
2269 
2270 	struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2271 
2272 #ifdef CONFIG_XPS
2273 	struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2274 #endif
2275 #ifdef CONFIG_NET_XGRESS
2276 	struct bpf_mprog_entry __rcu *tcx_egress;
2277 #endif
2278 #ifdef CONFIG_NETFILTER_EGRESS
2279 	struct nf_hook_entries __rcu *nf_hooks_egress;
2280 #endif
2281 
2282 #ifdef CONFIG_NET_SCHED
2283 	DECLARE_HASHTABLE	(qdisc_hash, 4);
2284 #endif
2285 	/* These may be needed for future network-power-down code. */
2286 	struct timer_list	watchdog_timer;
2287 	int			watchdog_timeo;
2288 
2289 	u32                     proto_down_reason;
2290 
2291 	struct list_head	todo_list;
2292 
2293 #ifdef CONFIG_PCPU_DEV_REFCNT
2294 	int __percpu		*pcpu_refcnt;
2295 #else
2296 	refcount_t		dev_refcnt;
2297 #endif
2298 	struct ref_tracker_dir	refcnt_tracker;
2299 
2300 	struct list_head	link_watch_list;
2301 
2302 	enum { NETREG_UNINITIALIZED=0,
2303 	       NETREG_REGISTERED,	/* completed register_netdevice */
2304 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
2305 	       NETREG_UNREGISTERED,	/* completed unregister todo */
2306 	       NETREG_RELEASED,		/* called free_netdev */
2307 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
2308 	} reg_state:8;
2309 
2310 	bool dismantle;
2311 
2312 	enum {
2313 		RTNL_LINK_INITIALIZED,
2314 		RTNL_LINK_INITIALIZING,
2315 	} rtnl_link_state:16;
2316 
2317 	bool needs_free_netdev;
2318 	void (*priv_destructor)(struct net_device *dev);
2319 
2320 #ifdef CONFIG_NETPOLL
2321 	struct netpoll_info __rcu	*npinfo;
2322 #endif
2323 
2324 	possible_net_t			nd_net;
2325 
2326 	/* mid-layer private */
2327 	void				*ml_priv;
2328 	enum netdev_ml_priv_type	ml_priv_type;
2329 
2330 	union {
2331 		struct pcpu_lstats __percpu		*lstats;
2332 		struct pcpu_sw_netstats __percpu	*tstats;
2333 		struct pcpu_dstats __percpu		*dstats;
2334 	};
2335 
2336 #if IS_ENABLED(CONFIG_GARP)
2337 	struct garp_port __rcu	*garp_port;
2338 #endif
2339 #if IS_ENABLED(CONFIG_MRP)
2340 	struct mrp_port __rcu	*mrp_port;
2341 #endif
2342 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2343 	struct dm_hw_stat_delta __rcu *dm_private;
2344 #endif
2345 	struct device		dev;
2346 	const struct attribute_group *sysfs_groups[4];
2347 	const struct attribute_group *sysfs_rx_queue_group;
2348 
2349 	const struct rtnl_link_ops *rtnl_link_ops;
2350 
2351 	/* for setting kernel sock attribute on TCP connection setup */
2352 #define GSO_MAX_SEGS		65535u
2353 #define GSO_LEGACY_MAX_SIZE	65536u
2354 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2355  * and shinfo->gso_segs is a 16bit field.
2356  */
2357 #define GSO_MAX_SIZE		(8 * GSO_MAX_SEGS)
2358 
2359 	unsigned int		gso_max_size;
2360 #define TSO_LEGACY_MAX_SIZE	65536
2361 #define TSO_MAX_SIZE		UINT_MAX
2362 	unsigned int		tso_max_size;
2363 	u16			gso_max_segs;
2364 #define TSO_MAX_SEGS		U16_MAX
2365 	u16			tso_max_segs;
2366 	unsigned int		gso_ipv4_max_size;
2367 
2368 #ifdef CONFIG_DCB
2369 	const struct dcbnl_rtnl_ops *dcbnl_ops;
2370 #endif
2371 	s16			num_tc;
2372 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
2373 	u8			prio_tc_map[TC_BITMASK + 1];
2374 
2375 #if IS_ENABLED(CONFIG_FCOE)
2376 	unsigned int		fcoe_ddp_xid;
2377 #endif
2378 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2379 	struct netprio_map __rcu *priomap;
2380 #endif
2381 	struct phy_device	*phydev;
2382 	struct sfp_bus		*sfp_bus;
2383 	struct lock_class_key	*qdisc_tx_busylock;
2384 	bool			proto_down;
2385 	unsigned		wol_enabled:1;
2386 	unsigned		threaded:1;
2387 
2388 	struct list_head	net_notifier_list;
2389 
2390 #if IS_ENABLED(CONFIG_MACSEC)
2391 	/* MACsec management functions */
2392 	const struct macsec_ops *macsec_ops;
2393 #endif
2394 	const struct udp_tunnel_nic_info	*udp_tunnel_nic_info;
2395 	struct udp_tunnel_nic	*udp_tunnel_nic;
2396 
2397 	/* protected by rtnl_lock */
2398 	struct bpf_xdp_entity	xdp_state[__MAX_XDP_MODE];
2399 
2400 	u8 dev_addr_shadow[MAX_ADDR_LEN];
2401 	netdevice_tracker	linkwatch_dev_tracker;
2402 	netdevice_tracker	watchdog_dev_tracker;
2403 	netdevice_tracker	dev_registered_tracker;
2404 	struct rtnl_hw_stats64	*offload_xstats_l3;
2405 
2406 	struct devlink_port	*devlink_port;
2407 
2408 #if IS_ENABLED(CONFIG_DPLL)
2409 	struct dpll_pin		*dpll_pin;
2410 #endif
2411 };
2412 #define to_net_dev(d) container_of(d, struct net_device, dev)
2413 
2414 /*
2415  * Driver should use this to assign devlink port instance to a netdevice
2416  * before it registers the netdevice. Therefore devlink_port is static
2417  * during the netdev lifetime after it is registered.
2418  */
2419 #define SET_NETDEV_DEVLINK_PORT(dev, port)			\
2420 ({								\
2421 	WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED);	\
2422 	((dev)->devlink_port = (port));				\
2423 })
2424 
2425 static inline bool netif_elide_gro(const struct net_device *dev)
2426 {
2427 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2428 		return true;
2429 	return false;
2430 }
2431 
2432 #define	NETDEV_ALIGN		32
2433 
2434 static inline
2435 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2436 {
2437 	return dev->prio_tc_map[prio & TC_BITMASK];
2438 }
2439 
2440 static inline
2441 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2442 {
2443 	if (tc >= dev->num_tc)
2444 		return -EINVAL;
2445 
2446 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2447 	return 0;
2448 }
2449 
2450 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2451 void netdev_reset_tc(struct net_device *dev);
2452 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2453 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2454 
2455 static inline
2456 int netdev_get_num_tc(struct net_device *dev)
2457 {
2458 	return dev->num_tc;
2459 }
2460 
2461 static inline void net_prefetch(void *p)
2462 {
2463 	prefetch(p);
2464 #if L1_CACHE_BYTES < 128
2465 	prefetch((u8 *)p + L1_CACHE_BYTES);
2466 #endif
2467 }
2468 
2469 static inline void net_prefetchw(void *p)
2470 {
2471 	prefetchw(p);
2472 #if L1_CACHE_BYTES < 128
2473 	prefetchw((u8 *)p + L1_CACHE_BYTES);
2474 #endif
2475 }
2476 
2477 void netdev_unbind_sb_channel(struct net_device *dev,
2478 			      struct net_device *sb_dev);
2479 int netdev_bind_sb_channel_queue(struct net_device *dev,
2480 				 struct net_device *sb_dev,
2481 				 u8 tc, u16 count, u16 offset);
2482 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2483 static inline int netdev_get_sb_channel(struct net_device *dev)
2484 {
2485 	return max_t(int, -dev->num_tc, 0);
2486 }
2487 
2488 static inline
2489 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2490 					 unsigned int index)
2491 {
2492 	DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2493 	return &dev->_tx[index];
2494 }
2495 
2496 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2497 						    const struct sk_buff *skb)
2498 {
2499 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2500 }
2501 
2502 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2503 					    void (*f)(struct net_device *,
2504 						      struct netdev_queue *,
2505 						      void *),
2506 					    void *arg)
2507 {
2508 	unsigned int i;
2509 
2510 	for (i = 0; i < dev->num_tx_queues; i++)
2511 		f(dev, &dev->_tx[i], arg);
2512 }
2513 
2514 #define netdev_lockdep_set_classes(dev)				\
2515 {								\
2516 	static struct lock_class_key qdisc_tx_busylock_key;	\
2517 	static struct lock_class_key qdisc_xmit_lock_key;	\
2518 	static struct lock_class_key dev_addr_list_lock_key;	\
2519 	unsigned int i;						\
2520 								\
2521 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
2522 	lockdep_set_class(&(dev)->addr_list_lock,		\
2523 			  &dev_addr_list_lock_key);		\
2524 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
2525 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
2526 				  &qdisc_xmit_lock_key);	\
2527 }
2528 
2529 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2530 		     struct net_device *sb_dev);
2531 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2532 					 struct sk_buff *skb,
2533 					 struct net_device *sb_dev);
2534 
2535 /* returns the headroom that the master device needs to take in account
2536  * when forwarding to this dev
2537  */
2538 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2539 {
2540 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2541 }
2542 
2543 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2544 {
2545 	if (dev->netdev_ops->ndo_set_rx_headroom)
2546 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2547 }
2548 
2549 /* set the device rx headroom to the dev's default */
2550 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2551 {
2552 	netdev_set_rx_headroom(dev, -1);
2553 }
2554 
2555 static inline void *netdev_get_ml_priv(struct net_device *dev,
2556 				       enum netdev_ml_priv_type type)
2557 {
2558 	if (dev->ml_priv_type != type)
2559 		return NULL;
2560 
2561 	return dev->ml_priv;
2562 }
2563 
2564 static inline void netdev_set_ml_priv(struct net_device *dev,
2565 				      void *ml_priv,
2566 				      enum netdev_ml_priv_type type)
2567 {
2568 	WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2569 	     "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2570 	     dev->ml_priv_type, type);
2571 	WARN(!dev->ml_priv_type && dev->ml_priv,
2572 	     "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2573 
2574 	dev->ml_priv = ml_priv;
2575 	dev->ml_priv_type = type;
2576 }
2577 
2578 /*
2579  * Net namespace inlines
2580  */
2581 static inline
2582 struct net *dev_net(const struct net_device *dev)
2583 {
2584 	return read_pnet(&dev->nd_net);
2585 }
2586 
2587 static inline
2588 void dev_net_set(struct net_device *dev, struct net *net)
2589 {
2590 	write_pnet(&dev->nd_net, net);
2591 }
2592 
2593 /**
2594  *	netdev_priv - access network device private data
2595  *	@dev: network device
2596  *
2597  * Get network device private data
2598  */
2599 static inline void *netdev_priv(const struct net_device *dev)
2600 {
2601 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2602 }
2603 
2604 /* Set the sysfs physical device reference for the network logical device
2605  * if set prior to registration will cause a symlink during initialization.
2606  */
2607 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2608 
2609 /* Set the sysfs device type for the network logical device to allow
2610  * fine-grained identification of different network device types. For
2611  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2612  */
2613 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2614 
2615 /* Default NAPI poll() weight
2616  * Device drivers are strongly advised to not use bigger value
2617  */
2618 #define NAPI_POLL_WEIGHT 64
2619 
2620 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2621 			   int (*poll)(struct napi_struct *, int), int weight);
2622 
2623 /**
2624  * netif_napi_add() - initialize a NAPI context
2625  * @dev:  network device
2626  * @napi: NAPI context
2627  * @poll: polling function
2628  *
2629  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2630  * *any* of the other NAPI-related functions.
2631  */
2632 static inline void
2633 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2634 	       int (*poll)(struct napi_struct *, int))
2635 {
2636 	netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2637 }
2638 
2639 static inline void
2640 netif_napi_add_tx_weight(struct net_device *dev,
2641 			 struct napi_struct *napi,
2642 			 int (*poll)(struct napi_struct *, int),
2643 			 int weight)
2644 {
2645 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2646 	netif_napi_add_weight(dev, napi, poll, weight);
2647 }
2648 
2649 /**
2650  * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2651  * @dev:  network device
2652  * @napi: NAPI context
2653  * @poll: polling function
2654  *
2655  * This variant of netif_napi_add() should be used from drivers using NAPI
2656  * to exclusively poll a TX queue.
2657  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2658  */
2659 static inline void netif_napi_add_tx(struct net_device *dev,
2660 				     struct napi_struct *napi,
2661 				     int (*poll)(struct napi_struct *, int))
2662 {
2663 	netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2664 }
2665 
2666 /**
2667  *  __netif_napi_del - remove a NAPI context
2668  *  @napi: NAPI context
2669  *
2670  * Warning: caller must observe RCU grace period before freeing memory
2671  * containing @napi. Drivers might want to call this helper to combine
2672  * all the needed RCU grace periods into a single one.
2673  */
2674 void __netif_napi_del(struct napi_struct *napi);
2675 
2676 /**
2677  *  netif_napi_del - remove a NAPI context
2678  *  @napi: NAPI context
2679  *
2680  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2681  */
2682 static inline void netif_napi_del(struct napi_struct *napi)
2683 {
2684 	__netif_napi_del(napi);
2685 	synchronize_net();
2686 }
2687 
2688 struct packet_type {
2689 	__be16			type;	/* This is really htons(ether_type). */
2690 	bool			ignore_outgoing;
2691 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2692 	netdevice_tracker	dev_tracker;
2693 	int			(*func) (struct sk_buff *,
2694 					 struct net_device *,
2695 					 struct packet_type *,
2696 					 struct net_device *);
2697 	void			(*list_func) (struct list_head *,
2698 					      struct packet_type *,
2699 					      struct net_device *);
2700 	bool			(*id_match)(struct packet_type *ptype,
2701 					    struct sock *sk);
2702 	struct net		*af_packet_net;
2703 	void			*af_packet_priv;
2704 	struct list_head	list;
2705 };
2706 
2707 struct offload_callbacks {
2708 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2709 						netdev_features_t features);
2710 	struct sk_buff		*(*gro_receive)(struct list_head *head,
2711 						struct sk_buff *skb);
2712 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2713 };
2714 
2715 struct packet_offload {
2716 	__be16			 type;	/* This is really htons(ether_type). */
2717 	u16			 priority;
2718 	struct offload_callbacks callbacks;
2719 	struct list_head	 list;
2720 };
2721 
2722 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2723 struct pcpu_sw_netstats {
2724 	u64_stats_t		rx_packets;
2725 	u64_stats_t		rx_bytes;
2726 	u64_stats_t		tx_packets;
2727 	u64_stats_t		tx_bytes;
2728 	struct u64_stats_sync   syncp;
2729 } __aligned(4 * sizeof(u64));
2730 
2731 struct pcpu_lstats {
2732 	u64_stats_t packets;
2733 	u64_stats_t bytes;
2734 	struct u64_stats_sync syncp;
2735 } __aligned(2 * sizeof(u64));
2736 
2737 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2738 
2739 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2740 {
2741 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2742 
2743 	u64_stats_update_begin(&tstats->syncp);
2744 	u64_stats_add(&tstats->rx_bytes, len);
2745 	u64_stats_inc(&tstats->rx_packets);
2746 	u64_stats_update_end(&tstats->syncp);
2747 }
2748 
2749 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2750 					  unsigned int packets,
2751 					  unsigned int len)
2752 {
2753 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2754 
2755 	u64_stats_update_begin(&tstats->syncp);
2756 	u64_stats_add(&tstats->tx_bytes, len);
2757 	u64_stats_add(&tstats->tx_packets, packets);
2758 	u64_stats_update_end(&tstats->syncp);
2759 }
2760 
2761 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2762 {
2763 	struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2764 
2765 	u64_stats_update_begin(&lstats->syncp);
2766 	u64_stats_add(&lstats->bytes, len);
2767 	u64_stats_inc(&lstats->packets);
2768 	u64_stats_update_end(&lstats->syncp);
2769 }
2770 
2771 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2772 ({									\
2773 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2774 	if (pcpu_stats)	{						\
2775 		int __cpu;						\
2776 		for_each_possible_cpu(__cpu) {				\
2777 			typeof(type) *stat;				\
2778 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2779 			u64_stats_init(&stat->syncp);			\
2780 		}							\
2781 	}								\
2782 	pcpu_stats;							\
2783 })
2784 
2785 #define netdev_alloc_pcpu_stats(type)					\
2786 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2787 
2788 #define devm_netdev_alloc_pcpu_stats(dev, type)				\
2789 ({									\
2790 	typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2791 	if (pcpu_stats) {						\
2792 		int __cpu;						\
2793 		for_each_possible_cpu(__cpu) {				\
2794 			typeof(type) *stat;				\
2795 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2796 			u64_stats_init(&stat->syncp);			\
2797 		}							\
2798 	}								\
2799 	pcpu_stats;							\
2800 })
2801 
2802 enum netdev_lag_tx_type {
2803 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2804 	NETDEV_LAG_TX_TYPE_RANDOM,
2805 	NETDEV_LAG_TX_TYPE_BROADCAST,
2806 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2807 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2808 	NETDEV_LAG_TX_TYPE_HASH,
2809 };
2810 
2811 enum netdev_lag_hash {
2812 	NETDEV_LAG_HASH_NONE,
2813 	NETDEV_LAG_HASH_L2,
2814 	NETDEV_LAG_HASH_L34,
2815 	NETDEV_LAG_HASH_L23,
2816 	NETDEV_LAG_HASH_E23,
2817 	NETDEV_LAG_HASH_E34,
2818 	NETDEV_LAG_HASH_VLAN_SRCMAC,
2819 	NETDEV_LAG_HASH_UNKNOWN,
2820 };
2821 
2822 struct netdev_lag_upper_info {
2823 	enum netdev_lag_tx_type tx_type;
2824 	enum netdev_lag_hash hash_type;
2825 };
2826 
2827 struct netdev_lag_lower_state_info {
2828 	u8 link_up : 1,
2829 	   tx_enabled : 1;
2830 };
2831 
2832 #include <linux/notifier.h>
2833 
2834 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2835  * and the rtnetlink notification exclusion list in rtnetlink_event() when
2836  * adding new types.
2837  */
2838 enum netdev_cmd {
2839 	NETDEV_UP	= 1,	/* For now you can't veto a device up/down */
2840 	NETDEV_DOWN,
2841 	NETDEV_REBOOT,		/* Tell a protocol stack a network interface
2842 				   detected a hardware crash and restarted
2843 				   - we can use this eg to kick tcp sessions
2844 				   once done */
2845 	NETDEV_CHANGE,		/* Notify device state change */
2846 	NETDEV_REGISTER,
2847 	NETDEV_UNREGISTER,
2848 	NETDEV_CHANGEMTU,	/* notify after mtu change happened */
2849 	NETDEV_CHANGEADDR,	/* notify after the address change */
2850 	NETDEV_PRE_CHANGEADDR,	/* notify before the address change */
2851 	NETDEV_GOING_DOWN,
2852 	NETDEV_CHANGENAME,
2853 	NETDEV_FEAT_CHANGE,
2854 	NETDEV_BONDING_FAILOVER,
2855 	NETDEV_PRE_UP,
2856 	NETDEV_PRE_TYPE_CHANGE,
2857 	NETDEV_POST_TYPE_CHANGE,
2858 	NETDEV_POST_INIT,
2859 	NETDEV_PRE_UNINIT,
2860 	NETDEV_RELEASE,
2861 	NETDEV_NOTIFY_PEERS,
2862 	NETDEV_JOIN,
2863 	NETDEV_CHANGEUPPER,
2864 	NETDEV_RESEND_IGMP,
2865 	NETDEV_PRECHANGEMTU,	/* notify before mtu change happened */
2866 	NETDEV_CHANGEINFODATA,
2867 	NETDEV_BONDING_INFO,
2868 	NETDEV_PRECHANGEUPPER,
2869 	NETDEV_CHANGELOWERSTATE,
2870 	NETDEV_UDP_TUNNEL_PUSH_INFO,
2871 	NETDEV_UDP_TUNNEL_DROP_INFO,
2872 	NETDEV_CHANGE_TX_QUEUE_LEN,
2873 	NETDEV_CVLAN_FILTER_PUSH_INFO,
2874 	NETDEV_CVLAN_FILTER_DROP_INFO,
2875 	NETDEV_SVLAN_FILTER_PUSH_INFO,
2876 	NETDEV_SVLAN_FILTER_DROP_INFO,
2877 	NETDEV_OFFLOAD_XSTATS_ENABLE,
2878 	NETDEV_OFFLOAD_XSTATS_DISABLE,
2879 	NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2880 	NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2881 	NETDEV_XDP_FEAT_CHANGE,
2882 };
2883 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2884 
2885 int register_netdevice_notifier(struct notifier_block *nb);
2886 int unregister_netdevice_notifier(struct notifier_block *nb);
2887 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2888 int unregister_netdevice_notifier_net(struct net *net,
2889 				      struct notifier_block *nb);
2890 int register_netdevice_notifier_dev_net(struct net_device *dev,
2891 					struct notifier_block *nb,
2892 					struct netdev_net_notifier *nn);
2893 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2894 					  struct notifier_block *nb,
2895 					  struct netdev_net_notifier *nn);
2896 
2897 struct netdev_notifier_info {
2898 	struct net_device	*dev;
2899 	struct netlink_ext_ack	*extack;
2900 };
2901 
2902 struct netdev_notifier_info_ext {
2903 	struct netdev_notifier_info info; /* must be first */
2904 	union {
2905 		u32 mtu;
2906 	} ext;
2907 };
2908 
2909 struct netdev_notifier_change_info {
2910 	struct netdev_notifier_info info; /* must be first */
2911 	unsigned int flags_changed;
2912 };
2913 
2914 struct netdev_notifier_changeupper_info {
2915 	struct netdev_notifier_info info; /* must be first */
2916 	struct net_device *upper_dev; /* new upper dev */
2917 	bool master; /* is upper dev master */
2918 	bool linking; /* is the notification for link or unlink */
2919 	void *upper_info; /* upper dev info */
2920 };
2921 
2922 struct netdev_notifier_changelowerstate_info {
2923 	struct netdev_notifier_info info; /* must be first */
2924 	void *lower_state_info; /* is lower dev state */
2925 };
2926 
2927 struct netdev_notifier_pre_changeaddr_info {
2928 	struct netdev_notifier_info info; /* must be first */
2929 	const unsigned char *dev_addr;
2930 };
2931 
2932 enum netdev_offload_xstats_type {
2933 	NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2934 };
2935 
2936 struct netdev_notifier_offload_xstats_info {
2937 	struct netdev_notifier_info info; /* must be first */
2938 	enum netdev_offload_xstats_type type;
2939 
2940 	union {
2941 		/* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2942 		struct netdev_notifier_offload_xstats_rd *report_delta;
2943 		/* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2944 		struct netdev_notifier_offload_xstats_ru *report_used;
2945 	};
2946 };
2947 
2948 int netdev_offload_xstats_enable(struct net_device *dev,
2949 				 enum netdev_offload_xstats_type type,
2950 				 struct netlink_ext_ack *extack);
2951 int netdev_offload_xstats_disable(struct net_device *dev,
2952 				  enum netdev_offload_xstats_type type);
2953 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2954 				   enum netdev_offload_xstats_type type);
2955 int netdev_offload_xstats_get(struct net_device *dev,
2956 			      enum netdev_offload_xstats_type type,
2957 			      struct rtnl_hw_stats64 *stats, bool *used,
2958 			      struct netlink_ext_ack *extack);
2959 void
2960 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2961 				   const struct rtnl_hw_stats64 *stats);
2962 void
2963 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2964 void netdev_offload_xstats_push_delta(struct net_device *dev,
2965 				      enum netdev_offload_xstats_type type,
2966 				      const struct rtnl_hw_stats64 *stats);
2967 
2968 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2969 					     struct net_device *dev)
2970 {
2971 	info->dev = dev;
2972 	info->extack = NULL;
2973 }
2974 
2975 static inline struct net_device *
2976 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2977 {
2978 	return info->dev;
2979 }
2980 
2981 static inline struct netlink_ext_ack *
2982 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2983 {
2984 	return info->extack;
2985 }
2986 
2987 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2988 int call_netdevice_notifiers_info(unsigned long val,
2989 				  struct netdev_notifier_info *info);
2990 
2991 extern rwlock_t				dev_base_lock;		/* Device list lock */
2992 
2993 #define for_each_netdev(net, d)		\
2994 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2995 #define for_each_netdev_reverse(net, d)	\
2996 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2997 #define for_each_netdev_rcu(net, d)		\
2998 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2999 #define for_each_netdev_safe(net, d, n)	\
3000 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3001 #define for_each_netdev_continue(net, d)		\
3002 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3003 #define for_each_netdev_continue_reverse(net, d)		\
3004 		list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3005 						     dev_list)
3006 #define for_each_netdev_continue_rcu(net, d)		\
3007 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3008 #define for_each_netdev_in_bond_rcu(bond, slave)	\
3009 		for_each_netdev_rcu(&init_net, slave)	\
3010 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3011 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
3012 
3013 #define for_each_netdev_dump(net, d, ifindex)				\
3014 	xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3015 
3016 static inline struct net_device *next_net_device(struct net_device *dev)
3017 {
3018 	struct list_head *lh;
3019 	struct net *net;
3020 
3021 	net = dev_net(dev);
3022 	lh = dev->dev_list.next;
3023 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3024 }
3025 
3026 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3027 {
3028 	struct list_head *lh;
3029 	struct net *net;
3030 
3031 	net = dev_net(dev);
3032 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3033 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3034 }
3035 
3036 static inline struct net_device *first_net_device(struct net *net)
3037 {
3038 	return list_empty(&net->dev_base_head) ? NULL :
3039 		net_device_entry(net->dev_base_head.next);
3040 }
3041 
3042 static inline struct net_device *first_net_device_rcu(struct net *net)
3043 {
3044 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3045 
3046 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3047 }
3048 
3049 int netdev_boot_setup_check(struct net_device *dev);
3050 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3051 				       const char *hwaddr);
3052 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3053 void dev_add_pack(struct packet_type *pt);
3054 void dev_remove_pack(struct packet_type *pt);
3055 void __dev_remove_pack(struct packet_type *pt);
3056 void dev_add_offload(struct packet_offload *po);
3057 void dev_remove_offload(struct packet_offload *po);
3058 
3059 int dev_get_iflink(const struct net_device *dev);
3060 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3061 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3062 			  struct net_device_path_stack *stack);
3063 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3064 				      unsigned short mask);
3065 struct net_device *dev_get_by_name(struct net *net, const char *name);
3066 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3067 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3068 bool netdev_name_in_use(struct net *net, const char *name);
3069 int dev_alloc_name(struct net_device *dev, const char *name);
3070 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3071 void dev_close(struct net_device *dev);
3072 void dev_close_many(struct list_head *head, bool unlink);
3073 void dev_disable_lro(struct net_device *dev);
3074 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3075 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3076 		     struct net_device *sb_dev);
3077 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3078 		       struct net_device *sb_dev);
3079 
3080 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3081 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3082 
3083 static inline int dev_queue_xmit(struct sk_buff *skb)
3084 {
3085 	return __dev_queue_xmit(skb, NULL);
3086 }
3087 
3088 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3089 				       struct net_device *sb_dev)
3090 {
3091 	return __dev_queue_xmit(skb, sb_dev);
3092 }
3093 
3094 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3095 {
3096 	int ret;
3097 
3098 	ret = __dev_direct_xmit(skb, queue_id);
3099 	if (!dev_xmit_complete(ret))
3100 		kfree_skb(skb);
3101 	return ret;
3102 }
3103 
3104 int register_netdevice(struct net_device *dev);
3105 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3106 void unregister_netdevice_many(struct list_head *head);
3107 static inline void unregister_netdevice(struct net_device *dev)
3108 {
3109 	unregister_netdevice_queue(dev, NULL);
3110 }
3111 
3112 int netdev_refcnt_read(const struct net_device *dev);
3113 void free_netdev(struct net_device *dev);
3114 void netdev_freemem(struct net_device *dev);
3115 int init_dummy_netdev(struct net_device *dev);
3116 
3117 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3118 					 struct sk_buff *skb,
3119 					 bool all_slaves);
3120 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3121 					    struct sock *sk);
3122 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3123 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3124 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3125 				       netdevice_tracker *tracker, gfp_t gfp);
3126 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3127 				      netdevice_tracker *tracker, gfp_t gfp);
3128 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3129 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3130 
3131 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3132 				  unsigned short type,
3133 				  const void *daddr, const void *saddr,
3134 				  unsigned int len)
3135 {
3136 	if (!dev->header_ops || !dev->header_ops->create)
3137 		return 0;
3138 
3139 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3140 }
3141 
3142 static inline int dev_parse_header(const struct sk_buff *skb,
3143 				   unsigned char *haddr)
3144 {
3145 	const struct net_device *dev = skb->dev;
3146 
3147 	if (!dev->header_ops || !dev->header_ops->parse)
3148 		return 0;
3149 	return dev->header_ops->parse(skb, haddr);
3150 }
3151 
3152 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3153 {
3154 	const struct net_device *dev = skb->dev;
3155 
3156 	if (!dev->header_ops || !dev->header_ops->parse_protocol)
3157 		return 0;
3158 	return dev->header_ops->parse_protocol(skb);
3159 }
3160 
3161 /* ll_header must have at least hard_header_len allocated */
3162 static inline bool dev_validate_header(const struct net_device *dev,
3163 				       char *ll_header, int len)
3164 {
3165 	if (likely(len >= dev->hard_header_len))
3166 		return true;
3167 	if (len < dev->min_header_len)
3168 		return false;
3169 
3170 	if (capable(CAP_SYS_RAWIO)) {
3171 		memset(ll_header + len, 0, dev->hard_header_len - len);
3172 		return true;
3173 	}
3174 
3175 	if (dev->header_ops && dev->header_ops->validate)
3176 		return dev->header_ops->validate(ll_header, len);
3177 
3178 	return false;
3179 }
3180 
3181 static inline bool dev_has_header(const struct net_device *dev)
3182 {
3183 	return dev->header_ops && dev->header_ops->create;
3184 }
3185 
3186 /*
3187  * Incoming packets are placed on per-CPU queues
3188  */
3189 struct softnet_data {
3190 	struct list_head	poll_list;
3191 	struct sk_buff_head	process_queue;
3192 
3193 	/* stats */
3194 	unsigned int		processed;
3195 	unsigned int		time_squeeze;
3196 #ifdef CONFIG_RPS
3197 	struct softnet_data	*rps_ipi_list;
3198 #endif
3199 
3200 	bool			in_net_rx_action;
3201 	bool			in_napi_threaded_poll;
3202 
3203 #ifdef CONFIG_NET_FLOW_LIMIT
3204 	struct sd_flow_limit __rcu *flow_limit;
3205 #endif
3206 	struct Qdisc		*output_queue;
3207 	struct Qdisc		**output_queue_tailp;
3208 	struct sk_buff		*completion_queue;
3209 #ifdef CONFIG_XFRM_OFFLOAD
3210 	struct sk_buff_head	xfrm_backlog;
3211 #endif
3212 	/* written and read only by owning cpu: */
3213 	struct {
3214 		u16 recursion;
3215 		u8  more;
3216 #ifdef CONFIG_NET_EGRESS
3217 		u8  skip_txqueue;
3218 #endif
3219 	} xmit;
3220 #ifdef CONFIG_RPS
3221 	/* input_queue_head should be written by cpu owning this struct,
3222 	 * and only read by other cpus. Worth using a cache line.
3223 	 */
3224 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
3225 
3226 	/* Elements below can be accessed between CPUs for RPS/RFS */
3227 	call_single_data_t	csd ____cacheline_aligned_in_smp;
3228 	struct softnet_data	*rps_ipi_next;
3229 	unsigned int		cpu;
3230 	unsigned int		input_queue_tail;
3231 #endif
3232 	unsigned int		received_rps;
3233 	unsigned int		dropped;
3234 	struct sk_buff_head	input_pkt_queue;
3235 	struct napi_struct	backlog;
3236 
3237 	/* Another possibly contended cache line */
3238 	spinlock_t		defer_lock ____cacheline_aligned_in_smp;
3239 	int			defer_count;
3240 	int			defer_ipi_scheduled;
3241 	struct sk_buff		*defer_list;
3242 	call_single_data_t	defer_csd;
3243 };
3244 
3245 static inline void input_queue_head_incr(struct softnet_data *sd)
3246 {
3247 #ifdef CONFIG_RPS
3248 	sd->input_queue_head++;
3249 #endif
3250 }
3251 
3252 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3253 					      unsigned int *qtail)
3254 {
3255 #ifdef CONFIG_RPS
3256 	*qtail = ++sd->input_queue_tail;
3257 #endif
3258 }
3259 
3260 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3261 
3262 static inline int dev_recursion_level(void)
3263 {
3264 	return this_cpu_read(softnet_data.xmit.recursion);
3265 }
3266 
3267 #define XMIT_RECURSION_LIMIT	8
3268 static inline bool dev_xmit_recursion(void)
3269 {
3270 	return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3271 			XMIT_RECURSION_LIMIT);
3272 }
3273 
3274 static inline void dev_xmit_recursion_inc(void)
3275 {
3276 	__this_cpu_inc(softnet_data.xmit.recursion);
3277 }
3278 
3279 static inline void dev_xmit_recursion_dec(void)
3280 {
3281 	__this_cpu_dec(softnet_data.xmit.recursion);
3282 }
3283 
3284 void __netif_schedule(struct Qdisc *q);
3285 void netif_schedule_queue(struct netdev_queue *txq);
3286 
3287 static inline void netif_tx_schedule_all(struct net_device *dev)
3288 {
3289 	unsigned int i;
3290 
3291 	for (i = 0; i < dev->num_tx_queues; i++)
3292 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
3293 }
3294 
3295 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3296 {
3297 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3298 }
3299 
3300 /**
3301  *	netif_start_queue - allow transmit
3302  *	@dev: network device
3303  *
3304  *	Allow upper layers to call the device hard_start_xmit routine.
3305  */
3306 static inline void netif_start_queue(struct net_device *dev)
3307 {
3308 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3309 }
3310 
3311 static inline void netif_tx_start_all_queues(struct net_device *dev)
3312 {
3313 	unsigned int i;
3314 
3315 	for (i = 0; i < dev->num_tx_queues; i++) {
3316 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3317 		netif_tx_start_queue(txq);
3318 	}
3319 }
3320 
3321 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3322 
3323 /**
3324  *	netif_wake_queue - restart transmit
3325  *	@dev: network device
3326  *
3327  *	Allow upper layers to call the device hard_start_xmit routine.
3328  *	Used for flow control when transmit resources are available.
3329  */
3330 static inline void netif_wake_queue(struct net_device *dev)
3331 {
3332 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3333 }
3334 
3335 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3336 {
3337 	unsigned int i;
3338 
3339 	for (i = 0; i < dev->num_tx_queues; i++) {
3340 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3341 		netif_tx_wake_queue(txq);
3342 	}
3343 }
3344 
3345 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3346 {
3347 	/* Must be an atomic op see netif_txq_try_stop() */
3348 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3349 }
3350 
3351 /**
3352  *	netif_stop_queue - stop transmitted packets
3353  *	@dev: network device
3354  *
3355  *	Stop upper layers calling the device hard_start_xmit routine.
3356  *	Used for flow control when transmit resources are unavailable.
3357  */
3358 static inline void netif_stop_queue(struct net_device *dev)
3359 {
3360 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3361 }
3362 
3363 void netif_tx_stop_all_queues(struct net_device *dev);
3364 
3365 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3366 {
3367 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3368 }
3369 
3370 /**
3371  *	netif_queue_stopped - test if transmit queue is flowblocked
3372  *	@dev: network device
3373  *
3374  *	Test if transmit queue on device is currently unable to send.
3375  */
3376 static inline bool netif_queue_stopped(const struct net_device *dev)
3377 {
3378 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3379 }
3380 
3381 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3382 {
3383 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3384 }
3385 
3386 static inline bool
3387 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3388 {
3389 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3390 }
3391 
3392 static inline bool
3393 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3394 {
3395 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3396 }
3397 
3398 /**
3399  *	netdev_queue_set_dql_min_limit - set dql minimum limit
3400  *	@dev_queue: pointer to transmit queue
3401  *	@min_limit: dql minimum limit
3402  *
3403  * Forces xmit_more() to return true until the minimum threshold
3404  * defined by @min_limit is reached (or until the tx queue is
3405  * empty). Warning: to be use with care, misuse will impact the
3406  * latency.
3407  */
3408 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3409 						  unsigned int min_limit)
3410 {
3411 #ifdef CONFIG_BQL
3412 	dev_queue->dql.min_limit = min_limit;
3413 #endif
3414 }
3415 
3416 /**
3417  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3418  *	@dev_queue: pointer to transmit queue
3419  *
3420  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3421  * to give appropriate hint to the CPU.
3422  */
3423 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3424 {
3425 #ifdef CONFIG_BQL
3426 	prefetchw(&dev_queue->dql.num_queued);
3427 #endif
3428 }
3429 
3430 /**
3431  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3432  *	@dev_queue: pointer to transmit queue
3433  *
3434  * BQL enabled drivers might use this helper in their TX completion path,
3435  * to give appropriate hint to the CPU.
3436  */
3437 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3438 {
3439 #ifdef CONFIG_BQL
3440 	prefetchw(&dev_queue->dql.limit);
3441 #endif
3442 }
3443 
3444 /**
3445  *	netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3446  *	@dev_queue: network device queue
3447  *	@bytes: number of bytes queued to the device queue
3448  *
3449  *	Report the number of bytes queued for sending/completion to the network
3450  *	device hardware queue. @bytes should be a good approximation and should
3451  *	exactly match netdev_completed_queue() @bytes.
3452  *	This is typically called once per packet, from ndo_start_xmit().
3453  */
3454 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3455 					unsigned int bytes)
3456 {
3457 #ifdef CONFIG_BQL
3458 	dql_queued(&dev_queue->dql, bytes);
3459 
3460 	if (likely(dql_avail(&dev_queue->dql) >= 0))
3461 		return;
3462 
3463 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3464 
3465 	/*
3466 	 * The XOFF flag must be set before checking the dql_avail below,
3467 	 * because in netdev_tx_completed_queue we update the dql_completed
3468 	 * before checking the XOFF flag.
3469 	 */
3470 	smp_mb();
3471 
3472 	/* check again in case another CPU has just made room avail */
3473 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3474 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3475 #endif
3476 }
3477 
3478 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3479  * that they should not test BQL status themselves.
3480  * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3481  * skb of a batch.
3482  * Returns true if the doorbell must be used to kick the NIC.
3483  */
3484 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3485 					  unsigned int bytes,
3486 					  bool xmit_more)
3487 {
3488 	if (xmit_more) {
3489 #ifdef CONFIG_BQL
3490 		dql_queued(&dev_queue->dql, bytes);
3491 #endif
3492 		return netif_tx_queue_stopped(dev_queue);
3493 	}
3494 	netdev_tx_sent_queue(dev_queue, bytes);
3495 	return true;
3496 }
3497 
3498 /**
3499  *	netdev_sent_queue - report the number of bytes queued to hardware
3500  *	@dev: network device
3501  *	@bytes: number of bytes queued to the hardware device queue
3502  *
3503  *	Report the number of bytes queued for sending/completion to the network
3504  *	device hardware queue#0. @bytes should be a good approximation and should
3505  *	exactly match netdev_completed_queue() @bytes.
3506  *	This is typically called once per packet, from ndo_start_xmit().
3507  */
3508 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3509 {
3510 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3511 }
3512 
3513 static inline bool __netdev_sent_queue(struct net_device *dev,
3514 				       unsigned int bytes,
3515 				       bool xmit_more)
3516 {
3517 	return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3518 				      xmit_more);
3519 }
3520 
3521 /**
3522  *	netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3523  *	@dev_queue: network device queue
3524  *	@pkts: number of packets (currently ignored)
3525  *	@bytes: number of bytes dequeued from the device queue
3526  *
3527  *	Must be called at most once per TX completion round (and not per
3528  *	individual packet), so that BQL can adjust its limits appropriately.
3529  */
3530 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3531 					     unsigned int pkts, unsigned int bytes)
3532 {
3533 #ifdef CONFIG_BQL
3534 	if (unlikely(!bytes))
3535 		return;
3536 
3537 	dql_completed(&dev_queue->dql, bytes);
3538 
3539 	/*
3540 	 * Without the memory barrier there is a small possiblity that
3541 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3542 	 * be stopped forever
3543 	 */
3544 	smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3545 
3546 	if (unlikely(dql_avail(&dev_queue->dql) < 0))
3547 		return;
3548 
3549 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3550 		netif_schedule_queue(dev_queue);
3551 #endif
3552 }
3553 
3554 /**
3555  * 	netdev_completed_queue - report bytes and packets completed by device
3556  * 	@dev: network device
3557  * 	@pkts: actual number of packets sent over the medium
3558  * 	@bytes: actual number of bytes sent over the medium
3559  *
3560  * 	Report the number of bytes and packets transmitted by the network device
3561  * 	hardware queue over the physical medium, @bytes must exactly match the
3562  * 	@bytes amount passed to netdev_sent_queue()
3563  */
3564 static inline void netdev_completed_queue(struct net_device *dev,
3565 					  unsigned int pkts, unsigned int bytes)
3566 {
3567 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3568 }
3569 
3570 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3571 {
3572 #ifdef CONFIG_BQL
3573 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3574 	dql_reset(&q->dql);
3575 #endif
3576 }
3577 
3578 /**
3579  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3580  * 	@dev_queue: network device
3581  *
3582  * 	Reset the bytes and packet count of a network device and clear the
3583  * 	software flow control OFF bit for this network device
3584  */
3585 static inline void netdev_reset_queue(struct net_device *dev_queue)
3586 {
3587 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3588 }
3589 
3590 /**
3591  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3592  * 	@dev: network device
3593  * 	@queue_index: given tx queue index
3594  *
3595  * 	Returns 0 if given tx queue index >= number of device tx queues,
3596  * 	otherwise returns the originally passed tx queue index.
3597  */
3598 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3599 {
3600 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3601 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3602 				     dev->name, queue_index,
3603 				     dev->real_num_tx_queues);
3604 		return 0;
3605 	}
3606 
3607 	return queue_index;
3608 }
3609 
3610 /**
3611  *	netif_running - test if up
3612  *	@dev: network device
3613  *
3614  *	Test if the device has been brought up.
3615  */
3616 static inline bool netif_running(const struct net_device *dev)
3617 {
3618 	return test_bit(__LINK_STATE_START, &dev->state);
3619 }
3620 
3621 /*
3622  * Routines to manage the subqueues on a device.  We only need start,
3623  * stop, and a check if it's stopped.  All other device management is
3624  * done at the overall netdevice level.
3625  * Also test the device if we're multiqueue.
3626  */
3627 
3628 /**
3629  *	netif_start_subqueue - allow sending packets on subqueue
3630  *	@dev: network device
3631  *	@queue_index: sub queue index
3632  *
3633  * Start individual transmit queue of a device with multiple transmit queues.
3634  */
3635 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3636 {
3637 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3638 
3639 	netif_tx_start_queue(txq);
3640 }
3641 
3642 /**
3643  *	netif_stop_subqueue - stop sending packets on subqueue
3644  *	@dev: network device
3645  *	@queue_index: sub queue index
3646  *
3647  * Stop individual transmit queue of a device with multiple transmit queues.
3648  */
3649 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3650 {
3651 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3652 	netif_tx_stop_queue(txq);
3653 }
3654 
3655 /**
3656  *	__netif_subqueue_stopped - test status of subqueue
3657  *	@dev: network device
3658  *	@queue_index: sub queue index
3659  *
3660  * Check individual transmit queue of a device with multiple transmit queues.
3661  */
3662 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3663 					    u16 queue_index)
3664 {
3665 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3666 
3667 	return netif_tx_queue_stopped(txq);
3668 }
3669 
3670 /**
3671  *	netif_subqueue_stopped - test status of subqueue
3672  *	@dev: network device
3673  *	@skb: sub queue buffer pointer
3674  *
3675  * Check individual transmit queue of a device with multiple transmit queues.
3676  */
3677 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3678 					  struct sk_buff *skb)
3679 {
3680 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3681 }
3682 
3683 /**
3684  *	netif_wake_subqueue - allow sending packets on subqueue
3685  *	@dev: network device
3686  *	@queue_index: sub queue index
3687  *
3688  * Resume individual transmit queue of a device with multiple transmit queues.
3689  */
3690 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3691 {
3692 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3693 
3694 	netif_tx_wake_queue(txq);
3695 }
3696 
3697 #ifdef CONFIG_XPS
3698 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3699 			u16 index);
3700 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3701 			  u16 index, enum xps_map_type type);
3702 
3703 /**
3704  *	netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3705  *	@j: CPU/Rx queue index
3706  *	@mask: bitmask of all cpus/rx queues
3707  *	@nr_bits: number of bits in the bitmask
3708  *
3709  * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3710  */
3711 static inline bool netif_attr_test_mask(unsigned long j,
3712 					const unsigned long *mask,
3713 					unsigned int nr_bits)
3714 {
3715 	cpu_max_bits_warn(j, nr_bits);
3716 	return test_bit(j, mask);
3717 }
3718 
3719 /**
3720  *	netif_attr_test_online - Test for online CPU/Rx queue
3721  *	@j: CPU/Rx queue index
3722  *	@online_mask: bitmask for CPUs/Rx queues that are online
3723  *	@nr_bits: number of bits in the bitmask
3724  *
3725  * Returns true if a CPU/Rx queue is online.
3726  */
3727 static inline bool netif_attr_test_online(unsigned long j,
3728 					  const unsigned long *online_mask,
3729 					  unsigned int nr_bits)
3730 {
3731 	cpu_max_bits_warn(j, nr_bits);
3732 
3733 	if (online_mask)
3734 		return test_bit(j, online_mask);
3735 
3736 	return (j < nr_bits);
3737 }
3738 
3739 /**
3740  *	netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3741  *	@n: CPU/Rx queue index
3742  *	@srcp: the cpumask/Rx queue mask pointer
3743  *	@nr_bits: number of bits in the bitmask
3744  *
3745  * Returns >= nr_bits if no further CPUs/Rx queues set.
3746  */
3747 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3748 					       unsigned int nr_bits)
3749 {
3750 	/* -1 is a legal arg here. */
3751 	if (n != -1)
3752 		cpu_max_bits_warn(n, nr_bits);
3753 
3754 	if (srcp)
3755 		return find_next_bit(srcp, nr_bits, n + 1);
3756 
3757 	return n + 1;
3758 }
3759 
3760 /**
3761  *	netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3762  *	@n: CPU/Rx queue index
3763  *	@src1p: the first CPUs/Rx queues mask pointer
3764  *	@src2p: the second CPUs/Rx queues mask pointer
3765  *	@nr_bits: number of bits in the bitmask
3766  *
3767  * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3768  */
3769 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3770 					  const unsigned long *src2p,
3771 					  unsigned int nr_bits)
3772 {
3773 	/* -1 is a legal arg here. */
3774 	if (n != -1)
3775 		cpu_max_bits_warn(n, nr_bits);
3776 
3777 	if (src1p && src2p)
3778 		return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3779 	else if (src1p)
3780 		return find_next_bit(src1p, nr_bits, n + 1);
3781 	else if (src2p)
3782 		return find_next_bit(src2p, nr_bits, n + 1);
3783 
3784 	return n + 1;
3785 }
3786 #else
3787 static inline int netif_set_xps_queue(struct net_device *dev,
3788 				      const struct cpumask *mask,
3789 				      u16 index)
3790 {
3791 	return 0;
3792 }
3793 
3794 static inline int __netif_set_xps_queue(struct net_device *dev,
3795 					const unsigned long *mask,
3796 					u16 index, enum xps_map_type type)
3797 {
3798 	return 0;
3799 }
3800 #endif
3801 
3802 /**
3803  *	netif_is_multiqueue - test if device has multiple transmit queues
3804  *	@dev: network device
3805  *
3806  * Check if device has multiple transmit queues
3807  */
3808 static inline bool netif_is_multiqueue(const struct net_device *dev)
3809 {
3810 	return dev->num_tx_queues > 1;
3811 }
3812 
3813 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3814 
3815 #ifdef CONFIG_SYSFS
3816 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3817 #else
3818 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3819 						unsigned int rxqs)
3820 {
3821 	dev->real_num_rx_queues = rxqs;
3822 	return 0;
3823 }
3824 #endif
3825 int netif_set_real_num_queues(struct net_device *dev,
3826 			      unsigned int txq, unsigned int rxq);
3827 
3828 int netif_get_num_default_rss_queues(void);
3829 
3830 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3831 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3832 
3833 /*
3834  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3835  * interrupt context or with hardware interrupts being disabled.
3836  * (in_hardirq() || irqs_disabled())
3837  *
3838  * We provide four helpers that can be used in following contexts :
3839  *
3840  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3841  *  replacing kfree_skb(skb)
3842  *
3843  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3844  *  Typically used in place of consume_skb(skb) in TX completion path
3845  *
3846  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3847  *  replacing kfree_skb(skb)
3848  *
3849  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3850  *  and consumed a packet. Used in place of consume_skb(skb)
3851  */
3852 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3853 {
3854 	dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3855 }
3856 
3857 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3858 {
3859 	dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3860 }
3861 
3862 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3863 {
3864 	dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3865 }
3866 
3867 static inline void dev_consume_skb_any(struct sk_buff *skb)
3868 {
3869 	dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3870 }
3871 
3872 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3873 			     struct bpf_prog *xdp_prog);
3874 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3875 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3876 int netif_rx(struct sk_buff *skb);
3877 int __netif_rx(struct sk_buff *skb);
3878 
3879 int netif_receive_skb(struct sk_buff *skb);
3880 int netif_receive_skb_core(struct sk_buff *skb);
3881 void netif_receive_skb_list_internal(struct list_head *head);
3882 void netif_receive_skb_list(struct list_head *head);
3883 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3884 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3885 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3886 void napi_get_frags_check(struct napi_struct *napi);
3887 gro_result_t napi_gro_frags(struct napi_struct *napi);
3888 struct packet_offload *gro_find_receive_by_type(__be16 type);
3889 struct packet_offload *gro_find_complete_by_type(__be16 type);
3890 
3891 static inline void napi_free_frags(struct napi_struct *napi)
3892 {
3893 	kfree_skb(napi->skb);
3894 	napi->skb = NULL;
3895 }
3896 
3897 bool netdev_is_rx_handler_busy(struct net_device *dev);
3898 int netdev_rx_handler_register(struct net_device *dev,
3899 			       rx_handler_func_t *rx_handler,
3900 			       void *rx_handler_data);
3901 void netdev_rx_handler_unregister(struct net_device *dev);
3902 
3903 bool dev_valid_name(const char *name);
3904 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3905 {
3906 	return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3907 }
3908 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3909 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3910 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3911 		void __user *data, bool *need_copyout);
3912 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3913 int generic_hwtstamp_get_lower(struct net_device *dev,
3914 			       struct kernel_hwtstamp_config *kernel_cfg);
3915 int generic_hwtstamp_set_lower(struct net_device *dev,
3916 			       struct kernel_hwtstamp_config *kernel_cfg,
3917 			       struct netlink_ext_ack *extack);
3918 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3919 unsigned int dev_get_flags(const struct net_device *);
3920 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3921 		       struct netlink_ext_ack *extack);
3922 int dev_change_flags(struct net_device *dev, unsigned int flags,
3923 		     struct netlink_ext_ack *extack);
3924 int dev_set_alias(struct net_device *, const char *, size_t);
3925 int dev_get_alias(const struct net_device *, char *, size_t);
3926 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3927 			       const char *pat, int new_ifindex);
3928 static inline
3929 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3930 			     const char *pat)
3931 {
3932 	return __dev_change_net_namespace(dev, net, pat, 0);
3933 }
3934 int __dev_set_mtu(struct net_device *, int);
3935 int dev_set_mtu(struct net_device *, int);
3936 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3937 			      struct netlink_ext_ack *extack);
3938 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3939 			struct netlink_ext_ack *extack);
3940 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3941 			     struct netlink_ext_ack *extack);
3942 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3943 int dev_get_port_parent_id(struct net_device *dev,
3944 			   struct netdev_phys_item_id *ppid, bool recurse);
3945 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3946 void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin);
3947 void netdev_dpll_pin_clear(struct net_device *dev);
3948 
3949 static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev)
3950 {
3951 #if IS_ENABLED(CONFIG_DPLL)
3952 	return dev->dpll_pin;
3953 #else
3954 	return NULL;
3955 #endif
3956 }
3957 
3958 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3959 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3960 				    struct netdev_queue *txq, int *ret);
3961 
3962 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3963 u8 dev_xdp_prog_count(struct net_device *dev);
3964 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3965 
3966 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3967 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3968 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3969 bool is_skb_forwardable(const struct net_device *dev,
3970 			const struct sk_buff *skb);
3971 
3972 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3973 						 const struct sk_buff *skb,
3974 						 const bool check_mtu)
3975 {
3976 	const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3977 	unsigned int len;
3978 
3979 	if (!(dev->flags & IFF_UP))
3980 		return false;
3981 
3982 	if (!check_mtu)
3983 		return true;
3984 
3985 	len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3986 	if (skb->len <= len)
3987 		return true;
3988 
3989 	/* if TSO is enabled, we don't care about the length as the packet
3990 	 * could be forwarded without being segmented before
3991 	 */
3992 	if (skb_is_gso(skb))
3993 		return true;
3994 
3995 	return false;
3996 }
3997 
3998 void netdev_core_stats_inc(struct net_device *dev, u32 offset);
3999 
4000 #define DEV_CORE_STATS_INC(FIELD)						\
4001 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev)		\
4002 {										\
4003 	netdev_core_stats_inc(dev,						\
4004 			offsetof(struct net_device_core_stats, FIELD));		\
4005 }
4006 DEV_CORE_STATS_INC(rx_dropped)
4007 DEV_CORE_STATS_INC(tx_dropped)
4008 DEV_CORE_STATS_INC(rx_nohandler)
4009 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4010 #undef DEV_CORE_STATS_INC
4011 
4012 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4013 					       struct sk_buff *skb,
4014 					       const bool check_mtu)
4015 {
4016 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4017 	    unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4018 		dev_core_stats_rx_dropped_inc(dev);
4019 		kfree_skb(skb);
4020 		return NET_RX_DROP;
4021 	}
4022 
4023 	skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4024 	skb->priority = 0;
4025 	return 0;
4026 }
4027 
4028 bool dev_nit_active(struct net_device *dev);
4029 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4030 
4031 static inline void __dev_put(struct net_device *dev)
4032 {
4033 	if (dev) {
4034 #ifdef CONFIG_PCPU_DEV_REFCNT
4035 		this_cpu_dec(*dev->pcpu_refcnt);
4036 #else
4037 		refcount_dec(&dev->dev_refcnt);
4038 #endif
4039 	}
4040 }
4041 
4042 static inline void __dev_hold(struct net_device *dev)
4043 {
4044 	if (dev) {
4045 #ifdef CONFIG_PCPU_DEV_REFCNT
4046 		this_cpu_inc(*dev->pcpu_refcnt);
4047 #else
4048 		refcount_inc(&dev->dev_refcnt);
4049 #endif
4050 	}
4051 }
4052 
4053 static inline void __netdev_tracker_alloc(struct net_device *dev,
4054 					  netdevice_tracker *tracker,
4055 					  gfp_t gfp)
4056 {
4057 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4058 	ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4059 #endif
4060 }
4061 
4062 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4063  * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4064  */
4065 static inline void netdev_tracker_alloc(struct net_device *dev,
4066 					netdevice_tracker *tracker, gfp_t gfp)
4067 {
4068 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4069 	refcount_dec(&dev->refcnt_tracker.no_tracker);
4070 	__netdev_tracker_alloc(dev, tracker, gfp);
4071 #endif
4072 }
4073 
4074 static inline void netdev_tracker_free(struct net_device *dev,
4075 				       netdevice_tracker *tracker)
4076 {
4077 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4078 	ref_tracker_free(&dev->refcnt_tracker, tracker);
4079 #endif
4080 }
4081 
4082 static inline void netdev_hold(struct net_device *dev,
4083 			       netdevice_tracker *tracker, gfp_t gfp)
4084 {
4085 	if (dev) {
4086 		__dev_hold(dev);
4087 		__netdev_tracker_alloc(dev, tracker, gfp);
4088 	}
4089 }
4090 
4091 static inline void netdev_put(struct net_device *dev,
4092 			      netdevice_tracker *tracker)
4093 {
4094 	if (dev) {
4095 		netdev_tracker_free(dev, tracker);
4096 		__dev_put(dev);
4097 	}
4098 }
4099 
4100 /**
4101  *	dev_hold - get reference to device
4102  *	@dev: network device
4103  *
4104  * Hold reference to device to keep it from being freed.
4105  * Try using netdev_hold() instead.
4106  */
4107 static inline void dev_hold(struct net_device *dev)
4108 {
4109 	netdev_hold(dev, NULL, GFP_ATOMIC);
4110 }
4111 
4112 /**
4113  *	dev_put - release reference to device
4114  *	@dev: network device
4115  *
4116  * Release reference to device to allow it to be freed.
4117  * Try using netdev_put() instead.
4118  */
4119 static inline void dev_put(struct net_device *dev)
4120 {
4121 	netdev_put(dev, NULL);
4122 }
4123 
4124 static inline void netdev_ref_replace(struct net_device *odev,
4125 				      struct net_device *ndev,
4126 				      netdevice_tracker *tracker,
4127 				      gfp_t gfp)
4128 {
4129 	if (odev)
4130 		netdev_tracker_free(odev, tracker);
4131 
4132 	__dev_hold(ndev);
4133 	__dev_put(odev);
4134 
4135 	if (ndev)
4136 		__netdev_tracker_alloc(ndev, tracker, gfp);
4137 }
4138 
4139 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4140  * and _off may be called from IRQ context, but it is caller
4141  * who is responsible for serialization of these calls.
4142  *
4143  * The name carrier is inappropriate, these functions should really be
4144  * called netif_lowerlayer_*() because they represent the state of any
4145  * kind of lower layer not just hardware media.
4146  */
4147 void linkwatch_fire_event(struct net_device *dev);
4148 
4149 /**
4150  *	netif_carrier_ok - test if carrier present
4151  *	@dev: network device
4152  *
4153  * Check if carrier is present on device
4154  */
4155 static inline bool netif_carrier_ok(const struct net_device *dev)
4156 {
4157 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4158 }
4159 
4160 unsigned long dev_trans_start(struct net_device *dev);
4161 
4162 void __netdev_watchdog_up(struct net_device *dev);
4163 
4164 void netif_carrier_on(struct net_device *dev);
4165 void netif_carrier_off(struct net_device *dev);
4166 void netif_carrier_event(struct net_device *dev);
4167 
4168 /**
4169  *	netif_dormant_on - mark device as dormant.
4170  *	@dev: network device
4171  *
4172  * Mark device as dormant (as per RFC2863).
4173  *
4174  * The dormant state indicates that the relevant interface is not
4175  * actually in a condition to pass packets (i.e., it is not 'up') but is
4176  * in a "pending" state, waiting for some external event.  For "on-
4177  * demand" interfaces, this new state identifies the situation where the
4178  * interface is waiting for events to place it in the up state.
4179  */
4180 static inline void netif_dormant_on(struct net_device *dev)
4181 {
4182 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4183 		linkwatch_fire_event(dev);
4184 }
4185 
4186 /**
4187  *	netif_dormant_off - set device as not dormant.
4188  *	@dev: network device
4189  *
4190  * Device is not in dormant state.
4191  */
4192 static inline void netif_dormant_off(struct net_device *dev)
4193 {
4194 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4195 		linkwatch_fire_event(dev);
4196 }
4197 
4198 /**
4199  *	netif_dormant - test if device is dormant
4200  *	@dev: network device
4201  *
4202  * Check if device is dormant.
4203  */
4204 static inline bool netif_dormant(const struct net_device *dev)
4205 {
4206 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
4207 }
4208 
4209 
4210 /**
4211  *	netif_testing_on - mark device as under test.
4212  *	@dev: network device
4213  *
4214  * Mark device as under test (as per RFC2863).
4215  *
4216  * The testing state indicates that some test(s) must be performed on
4217  * the interface. After completion, of the test, the interface state
4218  * will change to up, dormant, or down, as appropriate.
4219  */
4220 static inline void netif_testing_on(struct net_device *dev)
4221 {
4222 	if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4223 		linkwatch_fire_event(dev);
4224 }
4225 
4226 /**
4227  *	netif_testing_off - set device as not under test.
4228  *	@dev: network device
4229  *
4230  * Device is not in testing state.
4231  */
4232 static inline void netif_testing_off(struct net_device *dev)
4233 {
4234 	if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4235 		linkwatch_fire_event(dev);
4236 }
4237 
4238 /**
4239  *	netif_testing - test if device is under test
4240  *	@dev: network device
4241  *
4242  * Check if device is under test
4243  */
4244 static inline bool netif_testing(const struct net_device *dev)
4245 {
4246 	return test_bit(__LINK_STATE_TESTING, &dev->state);
4247 }
4248 
4249 
4250 /**
4251  *	netif_oper_up - test if device is operational
4252  *	@dev: network device
4253  *
4254  * Check if carrier is operational
4255  */
4256 static inline bool netif_oper_up(const struct net_device *dev)
4257 {
4258 	return (dev->operstate == IF_OPER_UP ||
4259 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4260 }
4261 
4262 /**
4263  *	netif_device_present - is device available or removed
4264  *	@dev: network device
4265  *
4266  * Check if device has not been removed from system.
4267  */
4268 static inline bool netif_device_present(const struct net_device *dev)
4269 {
4270 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
4271 }
4272 
4273 void netif_device_detach(struct net_device *dev);
4274 
4275 void netif_device_attach(struct net_device *dev);
4276 
4277 /*
4278  * Network interface message level settings
4279  */
4280 
4281 enum {
4282 	NETIF_MSG_DRV_BIT,
4283 	NETIF_MSG_PROBE_BIT,
4284 	NETIF_MSG_LINK_BIT,
4285 	NETIF_MSG_TIMER_BIT,
4286 	NETIF_MSG_IFDOWN_BIT,
4287 	NETIF_MSG_IFUP_BIT,
4288 	NETIF_MSG_RX_ERR_BIT,
4289 	NETIF_MSG_TX_ERR_BIT,
4290 	NETIF_MSG_TX_QUEUED_BIT,
4291 	NETIF_MSG_INTR_BIT,
4292 	NETIF_MSG_TX_DONE_BIT,
4293 	NETIF_MSG_RX_STATUS_BIT,
4294 	NETIF_MSG_PKTDATA_BIT,
4295 	NETIF_MSG_HW_BIT,
4296 	NETIF_MSG_WOL_BIT,
4297 
4298 	/* When you add a new bit above, update netif_msg_class_names array
4299 	 * in net/ethtool/common.c
4300 	 */
4301 	NETIF_MSG_CLASS_COUNT,
4302 };
4303 /* Both ethtool_ops interface and internal driver implementation use u32 */
4304 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4305 
4306 #define __NETIF_MSG_BIT(bit)	((u32)1 << (bit))
4307 #define __NETIF_MSG(name)	__NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4308 
4309 #define NETIF_MSG_DRV		__NETIF_MSG(DRV)
4310 #define NETIF_MSG_PROBE		__NETIF_MSG(PROBE)
4311 #define NETIF_MSG_LINK		__NETIF_MSG(LINK)
4312 #define NETIF_MSG_TIMER		__NETIF_MSG(TIMER)
4313 #define NETIF_MSG_IFDOWN	__NETIF_MSG(IFDOWN)
4314 #define NETIF_MSG_IFUP		__NETIF_MSG(IFUP)
4315 #define NETIF_MSG_RX_ERR	__NETIF_MSG(RX_ERR)
4316 #define NETIF_MSG_TX_ERR	__NETIF_MSG(TX_ERR)
4317 #define NETIF_MSG_TX_QUEUED	__NETIF_MSG(TX_QUEUED)
4318 #define NETIF_MSG_INTR		__NETIF_MSG(INTR)
4319 #define NETIF_MSG_TX_DONE	__NETIF_MSG(TX_DONE)
4320 #define NETIF_MSG_RX_STATUS	__NETIF_MSG(RX_STATUS)
4321 #define NETIF_MSG_PKTDATA	__NETIF_MSG(PKTDATA)
4322 #define NETIF_MSG_HW		__NETIF_MSG(HW)
4323 #define NETIF_MSG_WOL		__NETIF_MSG(WOL)
4324 
4325 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
4326 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
4327 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
4328 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
4329 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
4330 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
4331 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
4332 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
4333 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4334 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
4335 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
4336 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
4337 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
4338 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
4339 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
4340 
4341 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4342 {
4343 	/* use default */
4344 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4345 		return default_msg_enable_bits;
4346 	if (debug_value == 0)	/* no output */
4347 		return 0;
4348 	/* set low N bits */
4349 	return (1U << debug_value) - 1;
4350 }
4351 
4352 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4353 {
4354 	spin_lock(&txq->_xmit_lock);
4355 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4356 	WRITE_ONCE(txq->xmit_lock_owner, cpu);
4357 }
4358 
4359 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4360 {
4361 	__acquire(&txq->_xmit_lock);
4362 	return true;
4363 }
4364 
4365 static inline void __netif_tx_release(struct netdev_queue *txq)
4366 {
4367 	__release(&txq->_xmit_lock);
4368 }
4369 
4370 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4371 {
4372 	spin_lock_bh(&txq->_xmit_lock);
4373 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4374 	WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4375 }
4376 
4377 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4378 {
4379 	bool ok = spin_trylock(&txq->_xmit_lock);
4380 
4381 	if (likely(ok)) {
4382 		/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4383 		WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4384 	}
4385 	return ok;
4386 }
4387 
4388 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4389 {
4390 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4391 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4392 	spin_unlock(&txq->_xmit_lock);
4393 }
4394 
4395 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4396 {
4397 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4398 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4399 	spin_unlock_bh(&txq->_xmit_lock);
4400 }
4401 
4402 /*
4403  * txq->trans_start can be read locklessly from dev_watchdog()
4404  */
4405 static inline void txq_trans_update(struct netdev_queue *txq)
4406 {
4407 	if (txq->xmit_lock_owner != -1)
4408 		WRITE_ONCE(txq->trans_start, jiffies);
4409 }
4410 
4411 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4412 {
4413 	unsigned long now = jiffies;
4414 
4415 	if (READ_ONCE(txq->trans_start) != now)
4416 		WRITE_ONCE(txq->trans_start, now);
4417 }
4418 
4419 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4420 static inline void netif_trans_update(struct net_device *dev)
4421 {
4422 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4423 
4424 	txq_trans_cond_update(txq);
4425 }
4426 
4427 /**
4428  *	netif_tx_lock - grab network device transmit lock
4429  *	@dev: network device
4430  *
4431  * Get network device transmit lock
4432  */
4433 void netif_tx_lock(struct net_device *dev);
4434 
4435 static inline void netif_tx_lock_bh(struct net_device *dev)
4436 {
4437 	local_bh_disable();
4438 	netif_tx_lock(dev);
4439 }
4440 
4441 void netif_tx_unlock(struct net_device *dev);
4442 
4443 static inline void netif_tx_unlock_bh(struct net_device *dev)
4444 {
4445 	netif_tx_unlock(dev);
4446 	local_bh_enable();
4447 }
4448 
4449 #define HARD_TX_LOCK(dev, txq, cpu) {			\
4450 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4451 		__netif_tx_lock(txq, cpu);		\
4452 	} else {					\
4453 		__netif_tx_acquire(txq);		\
4454 	}						\
4455 }
4456 
4457 #define HARD_TX_TRYLOCK(dev, txq)			\
4458 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
4459 		__netif_tx_trylock(txq) :		\
4460 		__netif_tx_acquire(txq))
4461 
4462 #define HARD_TX_UNLOCK(dev, txq) {			\
4463 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4464 		__netif_tx_unlock(txq);			\
4465 	} else {					\
4466 		__netif_tx_release(txq);		\
4467 	}						\
4468 }
4469 
4470 static inline void netif_tx_disable(struct net_device *dev)
4471 {
4472 	unsigned int i;
4473 	int cpu;
4474 
4475 	local_bh_disable();
4476 	cpu = smp_processor_id();
4477 	spin_lock(&dev->tx_global_lock);
4478 	for (i = 0; i < dev->num_tx_queues; i++) {
4479 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4480 
4481 		__netif_tx_lock(txq, cpu);
4482 		netif_tx_stop_queue(txq);
4483 		__netif_tx_unlock(txq);
4484 	}
4485 	spin_unlock(&dev->tx_global_lock);
4486 	local_bh_enable();
4487 }
4488 
4489 static inline void netif_addr_lock(struct net_device *dev)
4490 {
4491 	unsigned char nest_level = 0;
4492 
4493 #ifdef CONFIG_LOCKDEP
4494 	nest_level = dev->nested_level;
4495 #endif
4496 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4497 }
4498 
4499 static inline void netif_addr_lock_bh(struct net_device *dev)
4500 {
4501 	unsigned char nest_level = 0;
4502 
4503 #ifdef CONFIG_LOCKDEP
4504 	nest_level = dev->nested_level;
4505 #endif
4506 	local_bh_disable();
4507 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4508 }
4509 
4510 static inline void netif_addr_unlock(struct net_device *dev)
4511 {
4512 	spin_unlock(&dev->addr_list_lock);
4513 }
4514 
4515 static inline void netif_addr_unlock_bh(struct net_device *dev)
4516 {
4517 	spin_unlock_bh(&dev->addr_list_lock);
4518 }
4519 
4520 /*
4521  * dev_addrs walker. Should be used only for read access. Call with
4522  * rcu_read_lock held.
4523  */
4524 #define for_each_dev_addr(dev, ha) \
4525 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4526 
4527 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4528 
4529 void ether_setup(struct net_device *dev);
4530 
4531 /* Support for loadable net-drivers */
4532 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4533 				    unsigned char name_assign_type,
4534 				    void (*setup)(struct net_device *),
4535 				    unsigned int txqs, unsigned int rxqs);
4536 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4537 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4538 
4539 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4540 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4541 			 count)
4542 
4543 int register_netdev(struct net_device *dev);
4544 void unregister_netdev(struct net_device *dev);
4545 
4546 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4547 
4548 /* General hardware address lists handling functions */
4549 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4550 		   struct netdev_hw_addr_list *from_list, int addr_len);
4551 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4552 		      struct netdev_hw_addr_list *from_list, int addr_len);
4553 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4554 		       struct net_device *dev,
4555 		       int (*sync)(struct net_device *, const unsigned char *),
4556 		       int (*unsync)(struct net_device *,
4557 				     const unsigned char *));
4558 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4559 			   struct net_device *dev,
4560 			   int (*sync)(struct net_device *,
4561 				       const unsigned char *, int),
4562 			   int (*unsync)(struct net_device *,
4563 					 const unsigned char *, int));
4564 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4565 			      struct net_device *dev,
4566 			      int (*unsync)(struct net_device *,
4567 					    const unsigned char *, int));
4568 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4569 			  struct net_device *dev,
4570 			  int (*unsync)(struct net_device *,
4571 					const unsigned char *));
4572 void __hw_addr_init(struct netdev_hw_addr_list *list);
4573 
4574 /* Functions used for device addresses handling */
4575 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4576 		  const void *addr, size_t len);
4577 
4578 static inline void
4579 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4580 {
4581 	dev_addr_mod(dev, 0, addr, len);
4582 }
4583 
4584 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4585 {
4586 	__dev_addr_set(dev, addr, dev->addr_len);
4587 }
4588 
4589 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4590 		 unsigned char addr_type);
4591 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4592 		 unsigned char addr_type);
4593 
4594 /* Functions used for unicast addresses handling */
4595 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4596 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4597 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4598 int dev_uc_sync(struct net_device *to, struct net_device *from);
4599 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4600 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4601 void dev_uc_flush(struct net_device *dev);
4602 void dev_uc_init(struct net_device *dev);
4603 
4604 /**
4605  *  __dev_uc_sync - Synchonize device's unicast list
4606  *  @dev:  device to sync
4607  *  @sync: function to call if address should be added
4608  *  @unsync: function to call if address should be removed
4609  *
4610  *  Add newly added addresses to the interface, and release
4611  *  addresses that have been deleted.
4612  */
4613 static inline int __dev_uc_sync(struct net_device *dev,
4614 				int (*sync)(struct net_device *,
4615 					    const unsigned char *),
4616 				int (*unsync)(struct net_device *,
4617 					      const unsigned char *))
4618 {
4619 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4620 }
4621 
4622 /**
4623  *  __dev_uc_unsync - Remove synchronized addresses from device
4624  *  @dev:  device to sync
4625  *  @unsync: function to call if address should be removed
4626  *
4627  *  Remove all addresses that were added to the device by dev_uc_sync().
4628  */
4629 static inline void __dev_uc_unsync(struct net_device *dev,
4630 				   int (*unsync)(struct net_device *,
4631 						 const unsigned char *))
4632 {
4633 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
4634 }
4635 
4636 /* Functions used for multicast addresses handling */
4637 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4638 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4639 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4640 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4641 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4642 int dev_mc_sync(struct net_device *to, struct net_device *from);
4643 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4644 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4645 void dev_mc_flush(struct net_device *dev);
4646 void dev_mc_init(struct net_device *dev);
4647 
4648 /**
4649  *  __dev_mc_sync - Synchonize device's multicast list
4650  *  @dev:  device to sync
4651  *  @sync: function to call if address should be added
4652  *  @unsync: function to call if address should be removed
4653  *
4654  *  Add newly added addresses to the interface, and release
4655  *  addresses that have been deleted.
4656  */
4657 static inline int __dev_mc_sync(struct net_device *dev,
4658 				int (*sync)(struct net_device *,
4659 					    const unsigned char *),
4660 				int (*unsync)(struct net_device *,
4661 					      const unsigned char *))
4662 {
4663 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4664 }
4665 
4666 /**
4667  *  __dev_mc_unsync - Remove synchronized addresses from device
4668  *  @dev:  device to sync
4669  *  @unsync: function to call if address should be removed
4670  *
4671  *  Remove all addresses that were added to the device by dev_mc_sync().
4672  */
4673 static inline void __dev_mc_unsync(struct net_device *dev,
4674 				   int (*unsync)(struct net_device *,
4675 						 const unsigned char *))
4676 {
4677 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
4678 }
4679 
4680 /* Functions used for secondary unicast and multicast support */
4681 void dev_set_rx_mode(struct net_device *dev);
4682 int dev_set_promiscuity(struct net_device *dev, int inc);
4683 int dev_set_allmulti(struct net_device *dev, int inc);
4684 void netdev_state_change(struct net_device *dev);
4685 void __netdev_notify_peers(struct net_device *dev);
4686 void netdev_notify_peers(struct net_device *dev);
4687 void netdev_features_change(struct net_device *dev);
4688 /* Load a device via the kmod */
4689 void dev_load(struct net *net, const char *name);
4690 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4691 					struct rtnl_link_stats64 *storage);
4692 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4693 			     const struct net_device_stats *netdev_stats);
4694 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4695 			   const struct pcpu_sw_netstats __percpu *netstats);
4696 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4697 
4698 extern int		netdev_max_backlog;
4699 extern int		dev_rx_weight;
4700 extern int		dev_tx_weight;
4701 extern int		gro_normal_batch;
4702 
4703 enum {
4704 	NESTED_SYNC_IMM_BIT,
4705 	NESTED_SYNC_TODO_BIT,
4706 };
4707 
4708 #define __NESTED_SYNC_BIT(bit)	((u32)1 << (bit))
4709 #define __NESTED_SYNC(name)	__NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4710 
4711 #define NESTED_SYNC_IMM		__NESTED_SYNC(IMM)
4712 #define NESTED_SYNC_TODO	__NESTED_SYNC(TODO)
4713 
4714 struct netdev_nested_priv {
4715 	unsigned char flags;
4716 	void *data;
4717 };
4718 
4719 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4720 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4721 						     struct list_head **iter);
4722 
4723 /* iterate through upper list, must be called under RCU read lock */
4724 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4725 	for (iter = &(dev)->adj_list.upper, \
4726 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4727 	     updev; \
4728 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4729 
4730 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4731 				  int (*fn)(struct net_device *upper_dev,
4732 					    struct netdev_nested_priv *priv),
4733 				  struct netdev_nested_priv *priv);
4734 
4735 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4736 				  struct net_device *upper_dev);
4737 
4738 bool netdev_has_any_upper_dev(struct net_device *dev);
4739 
4740 void *netdev_lower_get_next_private(struct net_device *dev,
4741 				    struct list_head **iter);
4742 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4743 					struct list_head **iter);
4744 
4745 #define netdev_for_each_lower_private(dev, priv, iter) \
4746 	for (iter = (dev)->adj_list.lower.next, \
4747 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
4748 	     priv; \
4749 	     priv = netdev_lower_get_next_private(dev, &(iter)))
4750 
4751 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4752 	for (iter = &(dev)->adj_list.lower, \
4753 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4754 	     priv; \
4755 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4756 
4757 void *netdev_lower_get_next(struct net_device *dev,
4758 				struct list_head **iter);
4759 
4760 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4761 	for (iter = (dev)->adj_list.lower.next, \
4762 	     ldev = netdev_lower_get_next(dev, &(iter)); \
4763 	     ldev; \
4764 	     ldev = netdev_lower_get_next(dev, &(iter)))
4765 
4766 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4767 					     struct list_head **iter);
4768 int netdev_walk_all_lower_dev(struct net_device *dev,
4769 			      int (*fn)(struct net_device *lower_dev,
4770 					struct netdev_nested_priv *priv),
4771 			      struct netdev_nested_priv *priv);
4772 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4773 				  int (*fn)(struct net_device *lower_dev,
4774 					    struct netdev_nested_priv *priv),
4775 				  struct netdev_nested_priv *priv);
4776 
4777 void *netdev_adjacent_get_private(struct list_head *adj_list);
4778 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4779 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4780 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4781 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4782 			  struct netlink_ext_ack *extack);
4783 int netdev_master_upper_dev_link(struct net_device *dev,
4784 				 struct net_device *upper_dev,
4785 				 void *upper_priv, void *upper_info,
4786 				 struct netlink_ext_ack *extack);
4787 void netdev_upper_dev_unlink(struct net_device *dev,
4788 			     struct net_device *upper_dev);
4789 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4790 				   struct net_device *new_dev,
4791 				   struct net_device *dev,
4792 				   struct netlink_ext_ack *extack);
4793 void netdev_adjacent_change_commit(struct net_device *old_dev,
4794 				   struct net_device *new_dev,
4795 				   struct net_device *dev);
4796 void netdev_adjacent_change_abort(struct net_device *old_dev,
4797 				  struct net_device *new_dev,
4798 				  struct net_device *dev);
4799 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4800 void *netdev_lower_dev_get_private(struct net_device *dev,
4801 				   struct net_device *lower_dev);
4802 void netdev_lower_state_changed(struct net_device *lower_dev,
4803 				void *lower_state_info);
4804 
4805 /* RSS keys are 40 or 52 bytes long */
4806 #define NETDEV_RSS_KEY_LEN 52
4807 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4808 void netdev_rss_key_fill(void *buffer, size_t len);
4809 
4810 int skb_checksum_help(struct sk_buff *skb);
4811 int skb_crc32c_csum_help(struct sk_buff *skb);
4812 int skb_csum_hwoffload_help(struct sk_buff *skb,
4813 			    const netdev_features_t features);
4814 
4815 struct netdev_bonding_info {
4816 	ifslave	slave;
4817 	ifbond	master;
4818 };
4819 
4820 struct netdev_notifier_bonding_info {
4821 	struct netdev_notifier_info info; /* must be first */
4822 	struct netdev_bonding_info  bonding_info;
4823 };
4824 
4825 void netdev_bonding_info_change(struct net_device *dev,
4826 				struct netdev_bonding_info *bonding_info);
4827 
4828 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4829 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4830 #else
4831 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4832 				  const void *data)
4833 {
4834 }
4835 #endif
4836 
4837 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4838 
4839 static inline bool can_checksum_protocol(netdev_features_t features,
4840 					 __be16 protocol)
4841 {
4842 	if (protocol == htons(ETH_P_FCOE))
4843 		return !!(features & NETIF_F_FCOE_CRC);
4844 
4845 	/* Assume this is an IP checksum (not SCTP CRC) */
4846 
4847 	if (features & NETIF_F_HW_CSUM) {
4848 		/* Can checksum everything */
4849 		return true;
4850 	}
4851 
4852 	switch (protocol) {
4853 	case htons(ETH_P_IP):
4854 		return !!(features & NETIF_F_IP_CSUM);
4855 	case htons(ETH_P_IPV6):
4856 		return !!(features & NETIF_F_IPV6_CSUM);
4857 	default:
4858 		return false;
4859 	}
4860 }
4861 
4862 #ifdef CONFIG_BUG
4863 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4864 #else
4865 static inline void netdev_rx_csum_fault(struct net_device *dev,
4866 					struct sk_buff *skb)
4867 {
4868 }
4869 #endif
4870 /* rx skb timestamps */
4871 void net_enable_timestamp(void);
4872 void net_disable_timestamp(void);
4873 
4874 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4875 					const struct skb_shared_hwtstamps *hwtstamps,
4876 					bool cycles)
4877 {
4878 	const struct net_device_ops *ops = dev->netdev_ops;
4879 
4880 	if (ops->ndo_get_tstamp)
4881 		return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4882 
4883 	return hwtstamps->hwtstamp;
4884 }
4885 
4886 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4887 					      struct sk_buff *skb, struct net_device *dev,
4888 					      bool more)
4889 {
4890 	__this_cpu_write(softnet_data.xmit.more, more);
4891 	return ops->ndo_start_xmit(skb, dev);
4892 }
4893 
4894 static inline bool netdev_xmit_more(void)
4895 {
4896 	return __this_cpu_read(softnet_data.xmit.more);
4897 }
4898 
4899 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4900 					    struct netdev_queue *txq, bool more)
4901 {
4902 	const struct net_device_ops *ops = dev->netdev_ops;
4903 	netdev_tx_t rc;
4904 
4905 	rc = __netdev_start_xmit(ops, skb, dev, more);
4906 	if (rc == NETDEV_TX_OK)
4907 		txq_trans_update(txq);
4908 
4909 	return rc;
4910 }
4911 
4912 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4913 				const void *ns);
4914 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4915 				 const void *ns);
4916 
4917 extern const struct kobj_ns_type_operations net_ns_type_operations;
4918 
4919 const char *netdev_drivername(const struct net_device *dev);
4920 
4921 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4922 							  netdev_features_t f2)
4923 {
4924 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4925 		if (f1 & NETIF_F_HW_CSUM)
4926 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4927 		else
4928 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4929 	}
4930 
4931 	return f1 & f2;
4932 }
4933 
4934 static inline netdev_features_t netdev_get_wanted_features(
4935 	struct net_device *dev)
4936 {
4937 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4938 }
4939 netdev_features_t netdev_increment_features(netdev_features_t all,
4940 	netdev_features_t one, netdev_features_t mask);
4941 
4942 /* Allow TSO being used on stacked device :
4943  * Performing the GSO segmentation before last device
4944  * is a performance improvement.
4945  */
4946 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4947 							netdev_features_t mask)
4948 {
4949 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4950 }
4951 
4952 int __netdev_update_features(struct net_device *dev);
4953 void netdev_update_features(struct net_device *dev);
4954 void netdev_change_features(struct net_device *dev);
4955 
4956 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4957 					struct net_device *dev);
4958 
4959 netdev_features_t passthru_features_check(struct sk_buff *skb,
4960 					  struct net_device *dev,
4961 					  netdev_features_t features);
4962 netdev_features_t netif_skb_features(struct sk_buff *skb);
4963 void skb_warn_bad_offload(const struct sk_buff *skb);
4964 
4965 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4966 {
4967 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4968 
4969 	/* check flags correspondence */
4970 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4971 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4972 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4973 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4974 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4975 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4976 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4977 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4978 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4979 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4980 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4981 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4982 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4983 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4984 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4985 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4986 	BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4987 	BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4988 	BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4989 
4990 	return (features & feature) == feature;
4991 }
4992 
4993 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4994 {
4995 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4996 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4997 }
4998 
4999 static inline bool netif_needs_gso(struct sk_buff *skb,
5000 				   netdev_features_t features)
5001 {
5002 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5003 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5004 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5005 }
5006 
5007 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5008 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5009 void netif_inherit_tso_max(struct net_device *to,
5010 			   const struct net_device *from);
5011 
5012 static inline bool netif_is_macsec(const struct net_device *dev)
5013 {
5014 	return dev->priv_flags & IFF_MACSEC;
5015 }
5016 
5017 static inline bool netif_is_macvlan(const struct net_device *dev)
5018 {
5019 	return dev->priv_flags & IFF_MACVLAN;
5020 }
5021 
5022 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5023 {
5024 	return dev->priv_flags & IFF_MACVLAN_PORT;
5025 }
5026 
5027 static inline bool netif_is_bond_master(const struct net_device *dev)
5028 {
5029 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5030 }
5031 
5032 static inline bool netif_is_bond_slave(const struct net_device *dev)
5033 {
5034 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5035 }
5036 
5037 static inline bool netif_supports_nofcs(struct net_device *dev)
5038 {
5039 	return dev->priv_flags & IFF_SUPP_NOFCS;
5040 }
5041 
5042 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5043 {
5044 	return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5045 }
5046 
5047 static inline bool netif_is_l3_master(const struct net_device *dev)
5048 {
5049 	return dev->priv_flags & IFF_L3MDEV_MASTER;
5050 }
5051 
5052 static inline bool netif_is_l3_slave(const struct net_device *dev)
5053 {
5054 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
5055 }
5056 
5057 static inline int dev_sdif(const struct net_device *dev)
5058 {
5059 #ifdef CONFIG_NET_L3_MASTER_DEV
5060 	if (netif_is_l3_slave(dev))
5061 		return dev->ifindex;
5062 #endif
5063 	return 0;
5064 }
5065 
5066 static inline bool netif_is_bridge_master(const struct net_device *dev)
5067 {
5068 	return dev->priv_flags & IFF_EBRIDGE;
5069 }
5070 
5071 static inline bool netif_is_bridge_port(const struct net_device *dev)
5072 {
5073 	return dev->priv_flags & IFF_BRIDGE_PORT;
5074 }
5075 
5076 static inline bool netif_is_ovs_master(const struct net_device *dev)
5077 {
5078 	return dev->priv_flags & IFF_OPENVSWITCH;
5079 }
5080 
5081 static inline bool netif_is_ovs_port(const struct net_device *dev)
5082 {
5083 	return dev->priv_flags & IFF_OVS_DATAPATH;
5084 }
5085 
5086 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5087 {
5088 	return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5089 }
5090 
5091 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5092 {
5093 	return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5094 }
5095 
5096 static inline bool netif_is_team_master(const struct net_device *dev)
5097 {
5098 	return dev->priv_flags & IFF_TEAM;
5099 }
5100 
5101 static inline bool netif_is_team_port(const struct net_device *dev)
5102 {
5103 	return dev->priv_flags & IFF_TEAM_PORT;
5104 }
5105 
5106 static inline bool netif_is_lag_master(const struct net_device *dev)
5107 {
5108 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
5109 }
5110 
5111 static inline bool netif_is_lag_port(const struct net_device *dev)
5112 {
5113 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5114 }
5115 
5116 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5117 {
5118 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
5119 }
5120 
5121 static inline bool netif_is_failover(const struct net_device *dev)
5122 {
5123 	return dev->priv_flags & IFF_FAILOVER;
5124 }
5125 
5126 static inline bool netif_is_failover_slave(const struct net_device *dev)
5127 {
5128 	return dev->priv_flags & IFF_FAILOVER_SLAVE;
5129 }
5130 
5131 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5132 static inline void netif_keep_dst(struct net_device *dev)
5133 {
5134 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5135 }
5136 
5137 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5138 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5139 {
5140 	/* TODO: reserve and use an additional IFF bit, if we get more users */
5141 	return netif_is_macsec(dev);
5142 }
5143 
5144 extern struct pernet_operations __net_initdata loopback_net_ops;
5145 
5146 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5147 
5148 /* netdev_printk helpers, similar to dev_printk */
5149 
5150 static inline const char *netdev_name(const struct net_device *dev)
5151 {
5152 	if (!dev->name[0] || strchr(dev->name, '%'))
5153 		return "(unnamed net_device)";
5154 	return dev->name;
5155 }
5156 
5157 static inline const char *netdev_reg_state(const struct net_device *dev)
5158 {
5159 	switch (dev->reg_state) {
5160 	case NETREG_UNINITIALIZED: return " (uninitialized)";
5161 	case NETREG_REGISTERED: return "";
5162 	case NETREG_UNREGISTERING: return " (unregistering)";
5163 	case NETREG_UNREGISTERED: return " (unregistered)";
5164 	case NETREG_RELEASED: return " (released)";
5165 	case NETREG_DUMMY: return " (dummy)";
5166 	}
5167 
5168 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5169 	return " (unknown)";
5170 }
5171 
5172 #define MODULE_ALIAS_NETDEV(device) \
5173 	MODULE_ALIAS("netdev-" device)
5174 
5175 /*
5176  * netdev_WARN() acts like dev_printk(), but with the key difference
5177  * of using a WARN/WARN_ON to get the message out, including the
5178  * file/line information and a backtrace.
5179  */
5180 #define netdev_WARN(dev, format, args...)			\
5181 	WARN(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5182 	     netdev_reg_state(dev), ##args)
5183 
5184 #define netdev_WARN_ONCE(dev, format, args...)				\
5185 	WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5186 		  netdev_reg_state(dev), ##args)
5187 
5188 /*
5189  *	The list of packet types we will receive (as opposed to discard)
5190  *	and the routines to invoke.
5191  *
5192  *	Why 16. Because with 16 the only overlap we get on a hash of the
5193  *	low nibble of the protocol value is RARP/SNAP/X.25.
5194  *
5195  *		0800	IP
5196  *		0001	802.3
5197  *		0002	AX.25
5198  *		0004	802.2
5199  *		8035	RARP
5200  *		0005	SNAP
5201  *		0805	X.25
5202  *		0806	ARP
5203  *		8137	IPX
5204  *		0009	Localtalk
5205  *		86DD	IPv6
5206  */
5207 #define PTYPE_HASH_SIZE	(16)
5208 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
5209 
5210 extern struct list_head ptype_all __read_mostly;
5211 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5212 
5213 extern struct net_device *blackhole_netdev;
5214 
5215 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5216 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5217 #define DEV_STATS_ADD(DEV, FIELD, VAL) 	\
5218 		atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5219 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5220 
5221 #endif	/* _LINUX_NETDEVICE_H */
5222