xref: /linux-6.15/include/linux/netdevice.h (revision d677aebd)
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 #include <linux/netdev_features.h>
44 #include <linux/neighbour.h>
45 #include <linux/netdevice_xmit.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_kern;
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 struct ethtool_netdev_state;
83 struct phy_link_topology;
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 struct neighbour;
231 struct neigh_parms;
232 struct sk_buff;
233 
234 struct netdev_hw_addr {
235 	struct list_head	list;
236 	struct rb_node		node;
237 	unsigned char		addr[MAX_ADDR_LEN];
238 	unsigned char		type;
239 #define NETDEV_HW_ADDR_T_LAN		1
240 #define NETDEV_HW_ADDR_T_SAN		2
241 #define NETDEV_HW_ADDR_T_UNICAST	3
242 #define NETDEV_HW_ADDR_T_MULTICAST	4
243 	bool			global_use;
244 	int			sync_cnt;
245 	int			refcount;
246 	int			synced;
247 	struct rcu_head		rcu_head;
248 };
249 
250 struct netdev_hw_addr_list {
251 	struct list_head	list;
252 	int			count;
253 
254 	/* Auxiliary tree for faster lookup on addition and deletion */
255 	struct rb_root		tree;
256 };
257 
258 #define netdev_hw_addr_list_count(l) ((l)->count)
259 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
260 #define netdev_hw_addr_list_for_each(ha, l) \
261 	list_for_each_entry(ha, &(l)->list, list)
262 
263 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
264 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
265 #define netdev_for_each_uc_addr(ha, dev) \
266 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
267 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
268 	netdev_for_each_uc_addr((_ha), (_dev)) \
269 		if ((_ha)->sync_cnt)
270 
271 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
272 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
273 #define netdev_for_each_mc_addr(ha, dev) \
274 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
275 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
276 	netdev_for_each_mc_addr((_ha), (_dev)) \
277 		if ((_ha)->sync_cnt)
278 
279 struct hh_cache {
280 	unsigned int	hh_len;
281 	seqlock_t	hh_lock;
282 
283 	/* cached hardware header; allow for machine alignment needs.        */
284 #define HH_DATA_MOD	16
285 #define HH_DATA_OFF(__len) \
286 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
287 #define HH_DATA_ALIGN(__len) \
288 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
289 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
290 };
291 
292 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
293  * Alternative is:
294  *   dev->hard_header_len ? (dev->hard_header_len +
295  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
296  *
297  * We could use other alignment values, but we must maintain the
298  * relationship HH alignment <= LL alignment.
299  */
300 #define LL_RESERVED_SPACE(dev) \
301 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
302 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
303 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
304 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
305 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
306 
307 struct header_ops {
308 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
309 			   unsigned short type, const void *daddr,
310 			   const void *saddr, unsigned int len);
311 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
312 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
313 	void	(*cache_update)(struct hh_cache *hh,
314 				const struct net_device *dev,
315 				const unsigned char *haddr);
316 	bool	(*validate)(const char *ll_header, unsigned int len);
317 	__be16	(*parse_protocol)(const struct sk_buff *skb);
318 };
319 
320 /* These flag bits are private to the generic network queueing
321  * layer; they may not be explicitly referenced by any other
322  * code.
323  */
324 
325 enum netdev_state_t {
326 	__LINK_STATE_START,
327 	__LINK_STATE_PRESENT,
328 	__LINK_STATE_NOCARRIER,
329 	__LINK_STATE_LINKWATCH_PENDING,
330 	__LINK_STATE_DORMANT,
331 	__LINK_STATE_TESTING,
332 };
333 
334 struct gro_list {
335 	struct list_head	list;
336 	int			count;
337 };
338 
339 /*
340  * size of gro hash buckets, must less than bit number of
341  * napi_struct::gro_bitmask
342  */
343 #define GRO_HASH_BUCKETS	8
344 
345 /*
346  * Structure for NAPI scheduling similar to tasklet but with weighting
347  */
348 struct napi_struct {
349 	/* The poll_list must only be managed by the entity which
350 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
351 	 * whoever atomically sets that bit can add this napi_struct
352 	 * to the per-CPU poll_list, and whoever clears that bit
353 	 * can remove from the list right before clearing the bit.
354 	 */
355 	struct list_head	poll_list;
356 
357 	unsigned long		state;
358 	int			weight;
359 	u32			defer_hard_irqs_count;
360 	unsigned long		gro_bitmask;
361 	int			(*poll)(struct napi_struct *, int);
362 #ifdef CONFIG_NETPOLL
363 	/* CPU actively polling if netpoll is configured */
364 	int			poll_owner;
365 #endif
366 	/* CPU on which NAPI has been scheduled for processing */
367 	int			list_owner;
368 	struct net_device	*dev;
369 	struct gro_list		gro_hash[GRO_HASH_BUCKETS];
370 	struct sk_buff		*skb;
371 	struct list_head	rx_list; /* Pending GRO_NORMAL skbs */
372 	int			rx_count; /* length of rx_list */
373 	unsigned int		napi_id;
374 	struct hrtimer		timer;
375 	struct task_struct	*thread;
376 	/* control-path-only fields follow */
377 	struct list_head	dev_list;
378 	struct hlist_node	napi_hash_node;
379 	int			irq;
380 };
381 
382 enum {
383 	NAPI_STATE_SCHED,		/* Poll is scheduled */
384 	NAPI_STATE_MISSED,		/* reschedule a napi */
385 	NAPI_STATE_DISABLE,		/* Disable pending */
386 	NAPI_STATE_NPSVC,		/* Netpoll - don't dequeue from poll_list */
387 	NAPI_STATE_LISTED,		/* NAPI added to system lists */
388 	NAPI_STATE_NO_BUSY_POLL,	/* Do not add in napi_hash, no busy polling */
389 	NAPI_STATE_IN_BUSY_POLL,	/* sk_busy_loop() owns this NAPI */
390 	NAPI_STATE_PREFER_BUSY_POLL,	/* prefer busy-polling over softirq processing*/
391 	NAPI_STATE_THREADED,		/* The poll is performed inside its own thread*/
392 	NAPI_STATE_SCHED_THREADED,	/* Napi is currently scheduled in threaded mode */
393 };
394 
395 enum {
396 	NAPIF_STATE_SCHED		= BIT(NAPI_STATE_SCHED),
397 	NAPIF_STATE_MISSED		= BIT(NAPI_STATE_MISSED),
398 	NAPIF_STATE_DISABLE		= BIT(NAPI_STATE_DISABLE),
399 	NAPIF_STATE_NPSVC		= BIT(NAPI_STATE_NPSVC),
400 	NAPIF_STATE_LISTED		= BIT(NAPI_STATE_LISTED),
401 	NAPIF_STATE_NO_BUSY_POLL	= BIT(NAPI_STATE_NO_BUSY_POLL),
402 	NAPIF_STATE_IN_BUSY_POLL	= BIT(NAPI_STATE_IN_BUSY_POLL),
403 	NAPIF_STATE_PREFER_BUSY_POLL	= BIT(NAPI_STATE_PREFER_BUSY_POLL),
404 	NAPIF_STATE_THREADED		= BIT(NAPI_STATE_THREADED),
405 	NAPIF_STATE_SCHED_THREADED	= BIT(NAPI_STATE_SCHED_THREADED),
406 };
407 
408 enum gro_result {
409 	GRO_MERGED,
410 	GRO_MERGED_FREE,
411 	GRO_HELD,
412 	GRO_NORMAL,
413 	GRO_CONSUMED,
414 };
415 typedef enum gro_result gro_result_t;
416 
417 /*
418  * enum rx_handler_result - Possible return values for rx_handlers.
419  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
420  * further.
421  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
422  * case skb->dev was changed by rx_handler.
423  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
424  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
425  *
426  * rx_handlers are functions called from inside __netif_receive_skb(), to do
427  * special processing of the skb, prior to delivery to protocol handlers.
428  *
429  * Currently, a net_device can only have a single rx_handler registered. Trying
430  * to register a second rx_handler will return -EBUSY.
431  *
432  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
433  * To unregister a rx_handler on a net_device, use
434  * netdev_rx_handler_unregister().
435  *
436  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
437  * do with the skb.
438  *
439  * If the rx_handler consumed the skb in some way, it should return
440  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
441  * the skb to be delivered in some other way.
442  *
443  * If the rx_handler changed skb->dev, to divert the skb to another
444  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
445  * new device will be called if it exists.
446  *
447  * If the rx_handler decides the skb should be ignored, it should return
448  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
449  * are registered on exact device (ptype->dev == skb->dev).
450  *
451  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
452  * delivered, it should return RX_HANDLER_PASS.
453  *
454  * A device without a registered rx_handler will behave as if rx_handler
455  * returned RX_HANDLER_PASS.
456  */
457 
458 enum rx_handler_result {
459 	RX_HANDLER_CONSUMED,
460 	RX_HANDLER_ANOTHER,
461 	RX_HANDLER_EXACT,
462 	RX_HANDLER_PASS,
463 };
464 typedef enum rx_handler_result rx_handler_result_t;
465 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
466 
467 void __napi_schedule(struct napi_struct *n);
468 void __napi_schedule_irqoff(struct napi_struct *n);
469 
470 static inline bool napi_disable_pending(struct napi_struct *n)
471 {
472 	return test_bit(NAPI_STATE_DISABLE, &n->state);
473 }
474 
475 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
476 {
477 	return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
478 }
479 
480 /**
481  * napi_is_scheduled - test if NAPI is scheduled
482  * @n: NAPI context
483  *
484  * This check is "best-effort". With no locking implemented,
485  * a NAPI can be scheduled or terminate right after this check
486  * and produce not precise results.
487  *
488  * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
489  * should not be used normally and napi_schedule should be
490  * used instead.
491  *
492  * Use only if the driver really needs to check if a NAPI
493  * is scheduled for example in the context of delayed timer
494  * that can be skipped if a NAPI is already scheduled.
495  *
496  * Return True if NAPI is scheduled, False otherwise.
497  */
498 static inline bool napi_is_scheduled(struct napi_struct *n)
499 {
500 	return test_bit(NAPI_STATE_SCHED, &n->state);
501 }
502 
503 bool napi_schedule_prep(struct napi_struct *n);
504 
505 /**
506  *	napi_schedule - schedule NAPI poll
507  *	@n: NAPI context
508  *
509  * Schedule NAPI poll routine to be called if it is not already
510  * running.
511  * Return true if we schedule a NAPI or false if not.
512  * Refer to napi_schedule_prep() for additional reason on why
513  * a NAPI might not be scheduled.
514  */
515 static inline bool napi_schedule(struct napi_struct *n)
516 {
517 	if (napi_schedule_prep(n)) {
518 		__napi_schedule(n);
519 		return true;
520 	}
521 
522 	return false;
523 }
524 
525 /**
526  *	napi_schedule_irqoff - schedule NAPI poll
527  *	@n: NAPI context
528  *
529  * Variant of napi_schedule(), assuming hard irqs are masked.
530  */
531 static inline void napi_schedule_irqoff(struct napi_struct *n)
532 {
533 	if (napi_schedule_prep(n))
534 		__napi_schedule_irqoff(n);
535 }
536 
537 /**
538  * napi_complete_done - NAPI processing complete
539  * @n: NAPI context
540  * @work_done: number of packets processed
541  *
542  * Mark NAPI processing as complete. Should only be called if poll budget
543  * has not been completely consumed.
544  * Prefer over napi_complete().
545  * Return false if device should avoid rearming interrupts.
546  */
547 bool napi_complete_done(struct napi_struct *n, int work_done);
548 
549 static inline bool napi_complete(struct napi_struct *n)
550 {
551 	return napi_complete_done(n, 0);
552 }
553 
554 int dev_set_threaded(struct net_device *dev, bool threaded);
555 
556 /**
557  *	napi_disable - prevent NAPI from scheduling
558  *	@n: NAPI context
559  *
560  * Stop NAPI from being scheduled on this context.
561  * Waits till any outstanding processing completes.
562  */
563 void napi_disable(struct napi_struct *n);
564 
565 void napi_enable(struct napi_struct *n);
566 
567 /**
568  *	napi_synchronize - wait until NAPI is not running
569  *	@n: NAPI context
570  *
571  * Wait until NAPI is done being scheduled on this context.
572  * Waits till any outstanding processing completes but
573  * does not disable future activations.
574  */
575 static inline void napi_synchronize(const struct napi_struct *n)
576 {
577 	if (IS_ENABLED(CONFIG_SMP))
578 		while (test_bit(NAPI_STATE_SCHED, &n->state))
579 			msleep(1);
580 	else
581 		barrier();
582 }
583 
584 /**
585  *	napi_if_scheduled_mark_missed - if napi is running, set the
586  *	NAPIF_STATE_MISSED
587  *	@n: NAPI context
588  *
589  * If napi is running, set the NAPIF_STATE_MISSED, and return true if
590  * NAPI is scheduled.
591  **/
592 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
593 {
594 	unsigned long val, new;
595 
596 	val = READ_ONCE(n->state);
597 	do {
598 		if (val & NAPIF_STATE_DISABLE)
599 			return true;
600 
601 		if (!(val & NAPIF_STATE_SCHED))
602 			return false;
603 
604 		new = val | NAPIF_STATE_MISSED;
605 	} while (!try_cmpxchg(&n->state, &val, new));
606 
607 	return true;
608 }
609 
610 enum netdev_queue_state_t {
611 	__QUEUE_STATE_DRV_XOFF,
612 	__QUEUE_STATE_STACK_XOFF,
613 	__QUEUE_STATE_FROZEN,
614 };
615 
616 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
617 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
618 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
619 
620 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
621 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
622 					QUEUE_STATE_FROZEN)
623 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
624 					QUEUE_STATE_FROZEN)
625 
626 /*
627  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
628  * netif_tx_* functions below are used to manipulate this flag.  The
629  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
630  * queue independently.  The netif_xmit_*stopped functions below are called
631  * to check if the queue has been stopped by the driver or stack (either
632  * of the XOFF bits are set in the state).  Drivers should not need to call
633  * netif_xmit*stopped functions, they should only be using netif_tx_*.
634  */
635 
636 struct netdev_queue {
637 /*
638  * read-mostly part
639  */
640 	struct net_device	*dev;
641 	netdevice_tracker	dev_tracker;
642 
643 	struct Qdisc __rcu	*qdisc;
644 	struct Qdisc __rcu	*qdisc_sleeping;
645 #ifdef CONFIG_SYSFS
646 	struct kobject		kobj;
647 #endif
648 	unsigned long		tx_maxrate;
649 	/*
650 	 * Number of TX timeouts for this queue
651 	 * (/sys/class/net/DEV/Q/trans_timeout)
652 	 */
653 	atomic_long_t		trans_timeout;
654 
655 	/* Subordinate device that the queue has been assigned to */
656 	struct net_device	*sb_dev;
657 #ifdef CONFIG_XDP_SOCKETS
658 	struct xsk_buff_pool    *pool;
659 #endif
660 
661 /*
662  * write-mostly part
663  */
664 #ifdef CONFIG_BQL
665 	struct dql		dql;
666 #endif
667 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
668 	int			xmit_lock_owner;
669 	/*
670 	 * Time (in jiffies) of last Tx
671 	 */
672 	unsigned long		trans_start;
673 
674 	unsigned long		state;
675 
676 /*
677  * slow- / control-path part
678  */
679 	/* NAPI instance for the queue
680 	 * Readers and writers must hold RTNL
681 	 */
682 	struct napi_struct	*napi;
683 
684 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
685 	int			numa_node;
686 #endif
687 } ____cacheline_aligned_in_smp;
688 
689 extern int sysctl_fb_tunnels_only_for_init_net;
690 extern int sysctl_devconf_inherit_init_net;
691 
692 /*
693  * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
694  *                                     == 1 : For initns only
695  *                                     == 2 : For none.
696  */
697 static inline bool net_has_fallback_tunnels(const struct net *net)
698 {
699 #if IS_ENABLED(CONFIG_SYSCTL)
700 	int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
701 
702 	return !fb_tunnels_only_for_init_net ||
703 		(net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
704 #else
705 	return true;
706 #endif
707 }
708 
709 static inline int net_inherit_devconf(void)
710 {
711 #if IS_ENABLED(CONFIG_SYSCTL)
712 	return READ_ONCE(sysctl_devconf_inherit_init_net);
713 #else
714 	return 0;
715 #endif
716 }
717 
718 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
719 {
720 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
721 	return q->numa_node;
722 #else
723 	return NUMA_NO_NODE;
724 #endif
725 }
726 
727 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
728 {
729 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
730 	q->numa_node = node;
731 #endif
732 }
733 
734 #ifdef CONFIG_RFS_ACCEL
735 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
736 			 u16 filter_id);
737 #endif
738 
739 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
740 enum xps_map_type {
741 	XPS_CPUS = 0,
742 	XPS_RXQS,
743 	XPS_MAPS_MAX,
744 };
745 
746 #ifdef CONFIG_XPS
747 /*
748  * This structure holds an XPS map which can be of variable length.  The
749  * map is an array of queues.
750  */
751 struct xps_map {
752 	unsigned int len;
753 	unsigned int alloc_len;
754 	struct rcu_head rcu;
755 	u16 queues[];
756 };
757 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
758 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
759        - sizeof(struct xps_map)) / sizeof(u16))
760 
761 /*
762  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
763  *
764  * We keep track of the number of cpus/rxqs used when the struct is allocated,
765  * in nr_ids. This will help not accessing out-of-bound memory.
766  *
767  * We keep track of the number of traffic classes used when the struct is
768  * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
769  * not crossing its upper bound, as the original dev->num_tc can be updated in
770  * the meantime.
771  */
772 struct xps_dev_maps {
773 	struct rcu_head rcu;
774 	unsigned int nr_ids;
775 	s16 num_tc;
776 	struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
777 };
778 
779 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +	\
780 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
781 
782 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
783 	(_rxqs * (_tcs) * sizeof(struct xps_map *)))
784 
785 #endif /* CONFIG_XPS */
786 
787 #define TC_MAX_QUEUE	16
788 #define TC_BITMASK	15
789 /* HW offloaded queuing disciplines txq count and offset maps */
790 struct netdev_tc_txq {
791 	u16 count;
792 	u16 offset;
793 };
794 
795 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
796 /*
797  * This structure is to hold information about the device
798  * configured to run FCoE protocol stack.
799  */
800 struct netdev_fcoe_hbainfo {
801 	char	manufacturer[64];
802 	char	serial_number[64];
803 	char	hardware_version[64];
804 	char	driver_version[64];
805 	char	optionrom_version[64];
806 	char	firmware_version[64];
807 	char	model[256];
808 	char	model_description[256];
809 };
810 #endif
811 
812 #define MAX_PHYS_ITEM_ID_LEN 32
813 
814 /* This structure holds a unique identifier to identify some
815  * physical item (port for example) used by a netdevice.
816  */
817 struct netdev_phys_item_id {
818 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
819 	unsigned char id_len;
820 };
821 
822 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
823 					    struct netdev_phys_item_id *b)
824 {
825 	return a->id_len == b->id_len &&
826 	       memcmp(a->id, b->id, a->id_len) == 0;
827 }
828 
829 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
830 				       struct sk_buff *skb,
831 				       struct net_device *sb_dev);
832 
833 enum net_device_path_type {
834 	DEV_PATH_ETHERNET = 0,
835 	DEV_PATH_VLAN,
836 	DEV_PATH_BRIDGE,
837 	DEV_PATH_PPPOE,
838 	DEV_PATH_DSA,
839 	DEV_PATH_MTK_WDMA,
840 };
841 
842 struct net_device_path {
843 	enum net_device_path_type	type;
844 	const struct net_device		*dev;
845 	union {
846 		struct {
847 			u16		id;
848 			__be16		proto;
849 			u8		h_dest[ETH_ALEN];
850 		} encap;
851 		struct {
852 			enum {
853 				DEV_PATH_BR_VLAN_KEEP,
854 				DEV_PATH_BR_VLAN_TAG,
855 				DEV_PATH_BR_VLAN_UNTAG,
856 				DEV_PATH_BR_VLAN_UNTAG_HW,
857 			}		vlan_mode;
858 			u16		vlan_id;
859 			__be16		vlan_proto;
860 		} bridge;
861 		struct {
862 			int port;
863 			u16 proto;
864 		} dsa;
865 		struct {
866 			u8 wdma_idx;
867 			u8 queue;
868 			u16 wcid;
869 			u8 bss;
870 			u8 amsdu;
871 		} mtk_wdma;
872 	};
873 };
874 
875 #define NET_DEVICE_PATH_STACK_MAX	5
876 #define NET_DEVICE_PATH_VLAN_MAX	2
877 
878 struct net_device_path_stack {
879 	int			num_paths;
880 	struct net_device_path	path[NET_DEVICE_PATH_STACK_MAX];
881 };
882 
883 struct net_device_path_ctx {
884 	const struct net_device *dev;
885 	u8			daddr[ETH_ALEN];
886 
887 	int			num_vlans;
888 	struct {
889 		u16		id;
890 		__be16		proto;
891 	} vlan[NET_DEVICE_PATH_VLAN_MAX];
892 };
893 
894 enum tc_setup_type {
895 	TC_QUERY_CAPS,
896 	TC_SETUP_QDISC_MQPRIO,
897 	TC_SETUP_CLSU32,
898 	TC_SETUP_CLSFLOWER,
899 	TC_SETUP_CLSMATCHALL,
900 	TC_SETUP_CLSBPF,
901 	TC_SETUP_BLOCK,
902 	TC_SETUP_QDISC_CBS,
903 	TC_SETUP_QDISC_RED,
904 	TC_SETUP_QDISC_PRIO,
905 	TC_SETUP_QDISC_MQ,
906 	TC_SETUP_QDISC_ETF,
907 	TC_SETUP_ROOT_QDISC,
908 	TC_SETUP_QDISC_GRED,
909 	TC_SETUP_QDISC_TAPRIO,
910 	TC_SETUP_FT,
911 	TC_SETUP_QDISC_ETS,
912 	TC_SETUP_QDISC_TBF,
913 	TC_SETUP_QDISC_FIFO,
914 	TC_SETUP_QDISC_HTB,
915 	TC_SETUP_ACT,
916 };
917 
918 /* These structures hold the attributes of bpf state that are being passed
919  * to the netdevice through the bpf op.
920  */
921 enum bpf_netdev_command {
922 	/* Set or clear a bpf program used in the earliest stages of packet
923 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
924 	 * is responsible for calling bpf_prog_put on any old progs that are
925 	 * stored. In case of error, the callee need not release the new prog
926 	 * reference, but on success it takes ownership and must bpf_prog_put
927 	 * when it is no longer used.
928 	 */
929 	XDP_SETUP_PROG,
930 	XDP_SETUP_PROG_HW,
931 	/* BPF program for offload callbacks, invoked at program load time. */
932 	BPF_OFFLOAD_MAP_ALLOC,
933 	BPF_OFFLOAD_MAP_FREE,
934 	XDP_SETUP_XSK_POOL,
935 };
936 
937 struct bpf_prog_offload_ops;
938 struct netlink_ext_ack;
939 struct xdp_umem;
940 struct xdp_dev_bulk_queue;
941 struct bpf_xdp_link;
942 
943 enum bpf_xdp_mode {
944 	XDP_MODE_SKB = 0,
945 	XDP_MODE_DRV = 1,
946 	XDP_MODE_HW = 2,
947 	__MAX_XDP_MODE
948 };
949 
950 struct bpf_xdp_entity {
951 	struct bpf_prog *prog;
952 	struct bpf_xdp_link *link;
953 };
954 
955 struct netdev_bpf {
956 	enum bpf_netdev_command command;
957 	union {
958 		/* XDP_SETUP_PROG */
959 		struct {
960 			u32 flags;
961 			struct bpf_prog *prog;
962 			struct netlink_ext_ack *extack;
963 		};
964 		/* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
965 		struct {
966 			struct bpf_offloaded_map *offmap;
967 		};
968 		/* XDP_SETUP_XSK_POOL */
969 		struct {
970 			struct xsk_buff_pool *pool;
971 			u16 queue_id;
972 		} xsk;
973 	};
974 };
975 
976 /* Flags for ndo_xsk_wakeup. */
977 #define XDP_WAKEUP_RX (1 << 0)
978 #define XDP_WAKEUP_TX (1 << 1)
979 
980 #ifdef CONFIG_XFRM_OFFLOAD
981 struct xfrmdev_ops {
982 	int	(*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
983 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
984 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
985 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
986 				       struct xfrm_state *x);
987 	void	(*xdo_dev_state_advance_esn) (struct xfrm_state *x);
988 	void	(*xdo_dev_state_update_stats) (struct xfrm_state *x);
989 	int	(*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
990 	void	(*xdo_dev_policy_delete) (struct xfrm_policy *x);
991 	void	(*xdo_dev_policy_free) (struct xfrm_policy *x);
992 };
993 #endif
994 
995 struct dev_ifalias {
996 	struct rcu_head rcuhead;
997 	char ifalias[];
998 };
999 
1000 struct devlink;
1001 struct tlsdev_ops;
1002 
1003 struct netdev_net_notifier {
1004 	struct list_head list;
1005 	struct notifier_block *nb;
1006 };
1007 
1008 /*
1009  * This structure defines the management hooks for network devices.
1010  * The following hooks can be defined; unless noted otherwise, they are
1011  * optional and can be filled with a null pointer.
1012  *
1013  * int (*ndo_init)(struct net_device *dev);
1014  *     This function is called once when a network device is registered.
1015  *     The network device can use this for any late stage initialization
1016  *     or semantic validation. It can fail with an error code which will
1017  *     be propagated back to register_netdev.
1018  *
1019  * void (*ndo_uninit)(struct net_device *dev);
1020  *     This function is called when device is unregistered or when registration
1021  *     fails. It is not called if init fails.
1022  *
1023  * int (*ndo_open)(struct net_device *dev);
1024  *     This function is called when a network device transitions to the up
1025  *     state.
1026  *
1027  * int (*ndo_stop)(struct net_device *dev);
1028  *     This function is called when a network device transitions to the down
1029  *     state.
1030  *
1031  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1032  *                               struct net_device *dev);
1033  *	Called when a packet needs to be transmitted.
1034  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
1035  *	the queue before that can happen; it's for obsolete devices and weird
1036  *	corner cases, but the stack really does a non-trivial amount
1037  *	of useless work if you return NETDEV_TX_BUSY.
1038  *	Required; cannot be NULL.
1039  *
1040  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1041  *					   struct net_device *dev
1042  *					   netdev_features_t features);
1043  *	Called by core transmit path to determine if device is capable of
1044  *	performing offload operations on a given packet. This is to give
1045  *	the device an opportunity to implement any restrictions that cannot
1046  *	be otherwise expressed by feature flags. The check is called with
1047  *	the set of features that the stack has calculated and it returns
1048  *	those the driver believes to be appropriate.
1049  *
1050  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1051  *                         struct net_device *sb_dev);
1052  *	Called to decide which queue to use when device supports multiple
1053  *	transmit queues.
1054  *
1055  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1056  *	This function is called to allow device receiver to make
1057  *	changes to configuration when multicast or promiscuous is enabled.
1058  *
1059  * void (*ndo_set_rx_mode)(struct net_device *dev);
1060  *	This function is called device changes address list filtering.
1061  *	If driver handles unicast address filtering, it should set
1062  *	IFF_UNICAST_FLT in its priv_flags.
1063  *
1064  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1065  *	This function  is called when the Media Access Control address
1066  *	needs to be changed. If this interface is not defined, the
1067  *	MAC address can not be changed.
1068  *
1069  * int (*ndo_validate_addr)(struct net_device *dev);
1070  *	Test if Media Access Control address is valid for the device.
1071  *
1072  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1073  *	Old-style ioctl entry point. This is used internally by the
1074  *	appletalk and ieee802154 subsystems but is no longer called by
1075  *	the device ioctl handler.
1076  *
1077  * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1078  *	Used by the bonding driver for its device specific ioctls:
1079  *	SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1080  *	SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1081  *
1082  * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1083  *	Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1084  *	SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1085  *
1086  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1087  *	Used to set network devices bus interface parameters. This interface
1088  *	is retained for legacy reasons; new devices should use the bus
1089  *	interface (PCI) for low level management.
1090  *
1091  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1092  *	Called when a user wants to change the Maximum Transfer Unit
1093  *	of a device.
1094  *
1095  * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1096  *	Callback used when the transmitter has not made any progress
1097  *	for dev->watchdog ticks.
1098  *
1099  * void (*ndo_get_stats64)(struct net_device *dev,
1100  *                         struct rtnl_link_stats64 *storage);
1101  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1102  *	Called when a user wants to get the network device usage
1103  *	statistics. Drivers must do one of the following:
1104  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
1105  *	   rtnl_link_stats64 structure passed by the caller.
1106  *	2. Define @ndo_get_stats to update a net_device_stats structure
1107  *	   (which should normally be dev->stats) and return a pointer to
1108  *	   it. The structure may be changed asynchronously only if each
1109  *	   field is written atomically.
1110  *	3. Update dev->stats asynchronously and atomically, and define
1111  *	   neither operation.
1112  *
1113  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1114  *	Return true if this device supports offload stats of this attr_id.
1115  *
1116  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1117  *	void *attr_data)
1118  *	Get statistics for offload operations by attr_id. Write it into the
1119  *	attr_data pointer.
1120  *
1121  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1122  *	If device supports VLAN filtering this function is called when a
1123  *	VLAN id is registered.
1124  *
1125  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1126  *	If device supports VLAN filtering this function is called when a
1127  *	VLAN id is unregistered.
1128  *
1129  * void (*ndo_poll_controller)(struct net_device *dev);
1130  *
1131  *	SR-IOV management functions.
1132  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1133  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1134  *			  u8 qos, __be16 proto);
1135  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1136  *			  int max_tx_rate);
1137  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1138  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1139  * int (*ndo_get_vf_config)(struct net_device *dev,
1140  *			    int vf, struct ifla_vf_info *ivf);
1141  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1142  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1143  *			  struct nlattr *port[]);
1144  *
1145  *      Enable or disable the VF ability to query its RSS Redirection Table and
1146  *      Hash Key. This is needed since on some devices VF share this information
1147  *      with PF and querying it may introduce a theoretical security risk.
1148  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1149  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1150  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1151  *		       void *type_data);
1152  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
1153  *	This is always called from the stack with the rtnl lock held and netif
1154  *	tx queues stopped. This allows the netdevice to perform queue
1155  *	management safely.
1156  *
1157  *	Fiber Channel over Ethernet (FCoE) offload functions.
1158  * int (*ndo_fcoe_enable)(struct net_device *dev);
1159  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
1160  *	so the underlying device can perform whatever needed configuration or
1161  *	initialization to support acceleration of FCoE traffic.
1162  *
1163  * int (*ndo_fcoe_disable)(struct net_device *dev);
1164  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
1165  *	so the underlying device can perform whatever needed clean-ups to
1166  *	stop supporting acceleration of FCoE traffic.
1167  *
1168  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1169  *			     struct scatterlist *sgl, unsigned int sgc);
1170  *	Called when the FCoE Initiator wants to initialize an I/O that
1171  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1172  *	perform necessary setup and returns 1 to indicate the device is set up
1173  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1174  *
1175  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1176  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1177  *	indicated by the FC exchange id 'xid', so the underlying device can
1178  *	clean up and reuse resources for later DDP requests.
1179  *
1180  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1181  *			      struct scatterlist *sgl, unsigned int sgc);
1182  *	Called when the FCoE Target wants to initialize an I/O that
1183  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1184  *	perform necessary setup and returns 1 to indicate the device is set up
1185  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1186  *
1187  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1188  *			       struct netdev_fcoe_hbainfo *hbainfo);
1189  *	Called when the FCoE Protocol stack wants information on the underlying
1190  *	device. This information is utilized by the FCoE protocol stack to
1191  *	register attributes with Fiber Channel management service as per the
1192  *	FC-GS Fabric Device Management Information(FDMI) specification.
1193  *
1194  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1195  *	Called when the underlying device wants to override default World Wide
1196  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1197  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1198  *	protocol stack to use.
1199  *
1200  *	RFS acceleration.
1201  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1202  *			    u16 rxq_index, u32 flow_id);
1203  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1204  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1205  *	Return the filter ID on success, or a negative error code.
1206  *
1207  *	Slave management functions (for bridge, bonding, etc).
1208  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1209  *	Called to make another netdev an underling.
1210  *
1211  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1212  *	Called to release previously enslaved netdev.
1213  *
1214  * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1215  *					    struct sk_buff *skb,
1216  *					    bool all_slaves);
1217  *	Get the xmit slave of master device. If all_slaves is true, function
1218  *	assume all the slaves can transmit.
1219  *
1220  *      Feature/offload setting functions.
1221  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1222  *		netdev_features_t features);
1223  *	Adjusts the requested feature flags according to device-specific
1224  *	constraints, and returns the resulting flags. Must not modify
1225  *	the device state.
1226  *
1227  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1228  *	Called to update device configuration to new features. Passed
1229  *	feature set might be less than what was returned by ndo_fix_features()).
1230  *	Must return >0 or -errno if it changed dev->features itself.
1231  *
1232  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1233  *		      struct net_device *dev,
1234  *		      const unsigned char *addr, u16 vid, u16 flags,
1235  *		      struct netlink_ext_ack *extack);
1236  *	Adds an FDB entry to dev for addr.
1237  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1238  *		      struct net_device *dev,
1239  *		      const unsigned char *addr, u16 vid)
1240  *	Deletes the FDB entry from dev corresponding to addr.
1241  * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1242  *			   struct netlink_ext_ack *extack);
1243  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1244  *		       struct net_device *dev, struct net_device *filter_dev,
1245  *		       int *idx)
1246  *	Used to add FDB entries to dump requests. Implementers should add
1247  *	entries to skb and update idx with the number of entries.
1248  *
1249  * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1250  *		      u16 nlmsg_flags, struct netlink_ext_ack *extack);
1251  *	Adds an MDB entry to dev.
1252  * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1253  *		      struct netlink_ext_ack *extack);
1254  *	Deletes the MDB entry from dev.
1255  * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1256  *			   struct netlink_ext_ack *extack);
1257  *	Bulk deletes MDB entries from dev.
1258  * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1259  *		       struct netlink_callback *cb);
1260  *	Dumps MDB entries from dev. The first argument (marker) in the netlink
1261  *	callback is used by core rtnetlink code.
1262  *
1263  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1264  *			     u16 flags, struct netlink_ext_ack *extack)
1265  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1266  *			     struct net_device *dev, u32 filter_mask,
1267  *			     int nlflags)
1268  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1269  *			     u16 flags);
1270  *
1271  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1272  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1273  *	which do not represent real hardware may define this to allow their
1274  *	userspace components to manage their virtual carrier state. Devices
1275  *	that determine carrier state from physical hardware properties (eg
1276  *	network cables) or protocol-dependent mechanisms (eg
1277  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1278  *
1279  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1280  *			       struct netdev_phys_item_id *ppid);
1281  *	Called to get ID of physical port of this device. If driver does
1282  *	not implement this, it is assumed that the hw is not able to have
1283  *	multiple net devices on single physical port.
1284  *
1285  * int (*ndo_get_port_parent_id)(struct net_device *dev,
1286  *				 struct netdev_phys_item_id *ppid)
1287  *	Called to get the parent ID of the physical port of this device.
1288  *
1289  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1290  *				 struct net_device *dev)
1291  *	Called by upper layer devices to accelerate switching or other
1292  *	station functionality into hardware. 'pdev is the lowerdev
1293  *	to use for the offload and 'dev' is the net device that will
1294  *	back the offload. Returns a pointer to the private structure
1295  *	the upper layer will maintain.
1296  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1297  *	Called by upper layer device to delete the station created
1298  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1299  *	the station and priv is the structure returned by the add
1300  *	operation.
1301  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1302  *			     int queue_index, u32 maxrate);
1303  *	Called when a user wants to set a max-rate limitation of specific
1304  *	TX queue.
1305  * int (*ndo_get_iflink)(const struct net_device *dev);
1306  *	Called to get the iflink value of this device.
1307  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1308  *	This function is used to get egress tunnel information for given skb.
1309  *	This is useful for retrieving outer tunnel header parameters while
1310  *	sampling packet.
1311  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1312  *	This function is used to specify the headroom that the skb must
1313  *	consider when allocation skb during packet reception. Setting
1314  *	appropriate rx headroom value allows avoiding skb head copy on
1315  *	forward. Setting a negative value resets the rx headroom to the
1316  *	default value.
1317  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1318  *	This function is used to set or query state related to XDP on the
1319  *	netdevice and manage BPF offload. See definition of
1320  *	enum bpf_netdev_command for details.
1321  * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1322  *			u32 flags);
1323  *	This function is used to submit @n XDP packets for transmit on a
1324  *	netdevice. Returns number of frames successfully transmitted, frames
1325  *	that got dropped are freed/returned via xdp_return_frame().
1326  *	Returns negative number, means general error invoking ndo, meaning
1327  *	no frames were xmit'ed and core-caller will free all frames.
1328  * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1329  *					        struct xdp_buff *xdp);
1330  *      Get the xmit slave of master device based on the xdp_buff.
1331  * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1332  *      This function is used to wake up the softirq, ksoftirqd or kthread
1333  *	responsible for sending and/or receiving packets on a specific
1334  *	queue id bound to an AF_XDP socket. The flags field specifies if
1335  *	only RX, only Tx, or both should be woken up using the flags
1336  *	XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1337  * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1338  *			 int cmd);
1339  *	Add, change, delete or get information on an IPv4 tunnel.
1340  * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1341  *	If a device is paired with a peer device, return the peer instance.
1342  *	The caller must be under RCU read context.
1343  * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1344  *     Get the forwarding path to reach the real device from the HW destination address
1345  * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1346  *			     const struct skb_shared_hwtstamps *hwtstamps,
1347  *			     bool cycles);
1348  *	Get hardware timestamp based on normal/adjustable time or free running
1349  *	cycle counter. This function is required if physical clock supports a
1350  *	free running cycle counter.
1351  *
1352  * int (*ndo_hwtstamp_get)(struct net_device *dev,
1353  *			   struct kernel_hwtstamp_config *kernel_config);
1354  *	Get the currently configured hardware timestamping parameters for the
1355  *	NIC device.
1356  *
1357  * int (*ndo_hwtstamp_set)(struct net_device *dev,
1358  *			   struct kernel_hwtstamp_config *kernel_config,
1359  *			   struct netlink_ext_ack *extack);
1360  *	Change the hardware timestamping parameters for NIC device.
1361  */
1362 struct net_device_ops {
1363 	int			(*ndo_init)(struct net_device *dev);
1364 	void			(*ndo_uninit)(struct net_device *dev);
1365 	int			(*ndo_open)(struct net_device *dev);
1366 	int			(*ndo_stop)(struct net_device *dev);
1367 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1368 						  struct net_device *dev);
1369 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1370 						      struct net_device *dev,
1371 						      netdev_features_t features);
1372 	u16			(*ndo_select_queue)(struct net_device *dev,
1373 						    struct sk_buff *skb,
1374 						    struct net_device *sb_dev);
1375 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1376 						       int flags);
1377 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1378 	int			(*ndo_set_mac_address)(struct net_device *dev,
1379 						       void *addr);
1380 	int			(*ndo_validate_addr)(struct net_device *dev);
1381 	int			(*ndo_do_ioctl)(struct net_device *dev,
1382 					        struct ifreq *ifr, int cmd);
1383 	int			(*ndo_eth_ioctl)(struct net_device *dev,
1384 						 struct ifreq *ifr, int cmd);
1385 	int			(*ndo_siocbond)(struct net_device *dev,
1386 						struct ifreq *ifr, int cmd);
1387 	int			(*ndo_siocwandev)(struct net_device *dev,
1388 						  struct if_settings *ifs);
1389 	int			(*ndo_siocdevprivate)(struct net_device *dev,
1390 						      struct ifreq *ifr,
1391 						      void __user *data, int cmd);
1392 	int			(*ndo_set_config)(struct net_device *dev,
1393 					          struct ifmap *map);
1394 	int			(*ndo_change_mtu)(struct net_device *dev,
1395 						  int new_mtu);
1396 	int			(*ndo_neigh_setup)(struct net_device *dev,
1397 						   struct neigh_parms *);
1398 	void			(*ndo_tx_timeout) (struct net_device *dev,
1399 						   unsigned int txqueue);
1400 
1401 	void			(*ndo_get_stats64)(struct net_device *dev,
1402 						   struct rtnl_link_stats64 *storage);
1403 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1404 	int			(*ndo_get_offload_stats)(int attr_id,
1405 							 const struct net_device *dev,
1406 							 void *attr_data);
1407 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1408 
1409 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1410 						       __be16 proto, u16 vid);
1411 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1412 						        __be16 proto, u16 vid);
1413 #ifdef CONFIG_NET_POLL_CONTROLLER
1414 	void                    (*ndo_poll_controller)(struct net_device *dev);
1415 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1416 						     struct netpoll_info *info);
1417 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1418 #endif
1419 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1420 						  int queue, u8 *mac);
1421 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1422 						   int queue, u16 vlan,
1423 						   u8 qos, __be16 proto);
1424 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1425 						   int vf, int min_tx_rate,
1426 						   int max_tx_rate);
1427 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1428 						       int vf, bool setting);
1429 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1430 						    int vf, bool setting);
1431 	int			(*ndo_get_vf_config)(struct net_device *dev,
1432 						     int vf,
1433 						     struct ifla_vf_info *ivf);
1434 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1435 							 int vf, int link_state);
1436 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1437 						    int vf,
1438 						    struct ifla_vf_stats
1439 						    *vf_stats);
1440 	int			(*ndo_set_vf_port)(struct net_device *dev,
1441 						   int vf,
1442 						   struct nlattr *port[]);
1443 	int			(*ndo_get_vf_port)(struct net_device *dev,
1444 						   int vf, struct sk_buff *skb);
1445 	int			(*ndo_get_vf_guid)(struct net_device *dev,
1446 						   int vf,
1447 						   struct ifla_vf_guid *node_guid,
1448 						   struct ifla_vf_guid *port_guid);
1449 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1450 						   int vf, u64 guid,
1451 						   int guid_type);
1452 	int			(*ndo_set_vf_rss_query_en)(
1453 						   struct net_device *dev,
1454 						   int vf, bool setting);
1455 	int			(*ndo_setup_tc)(struct net_device *dev,
1456 						enum tc_setup_type type,
1457 						void *type_data);
1458 #if IS_ENABLED(CONFIG_FCOE)
1459 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1460 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1461 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1462 						      u16 xid,
1463 						      struct scatterlist *sgl,
1464 						      unsigned int sgc);
1465 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1466 						     u16 xid);
1467 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1468 						       u16 xid,
1469 						       struct scatterlist *sgl,
1470 						       unsigned int sgc);
1471 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1472 							struct netdev_fcoe_hbainfo *hbainfo);
1473 #endif
1474 
1475 #if IS_ENABLED(CONFIG_LIBFCOE)
1476 #define NETDEV_FCOE_WWNN 0
1477 #define NETDEV_FCOE_WWPN 1
1478 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1479 						    u64 *wwn, int type);
1480 #endif
1481 
1482 #ifdef CONFIG_RFS_ACCEL
1483 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1484 						     const struct sk_buff *skb,
1485 						     u16 rxq_index,
1486 						     u32 flow_id);
1487 #endif
1488 	int			(*ndo_add_slave)(struct net_device *dev,
1489 						 struct net_device *slave_dev,
1490 						 struct netlink_ext_ack *extack);
1491 	int			(*ndo_del_slave)(struct net_device *dev,
1492 						 struct net_device *slave_dev);
1493 	struct net_device*	(*ndo_get_xmit_slave)(struct net_device *dev,
1494 						      struct sk_buff *skb,
1495 						      bool all_slaves);
1496 	struct net_device*	(*ndo_sk_get_lower_dev)(struct net_device *dev,
1497 							struct sock *sk);
1498 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1499 						    netdev_features_t features);
1500 	int			(*ndo_set_features)(struct net_device *dev,
1501 						    netdev_features_t features);
1502 	int			(*ndo_neigh_construct)(struct net_device *dev,
1503 						       struct neighbour *n);
1504 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1505 						     struct neighbour *n);
1506 
1507 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1508 					       struct nlattr *tb[],
1509 					       struct net_device *dev,
1510 					       const unsigned char *addr,
1511 					       u16 vid,
1512 					       u16 flags,
1513 					       struct netlink_ext_ack *extack);
1514 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1515 					       struct nlattr *tb[],
1516 					       struct net_device *dev,
1517 					       const unsigned char *addr,
1518 					       u16 vid, struct netlink_ext_ack *extack);
1519 	int			(*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1520 						    struct net_device *dev,
1521 						    struct netlink_ext_ack *extack);
1522 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1523 						struct netlink_callback *cb,
1524 						struct net_device *dev,
1525 						struct net_device *filter_dev,
1526 						int *idx);
1527 	int			(*ndo_fdb_get)(struct sk_buff *skb,
1528 					       struct nlattr *tb[],
1529 					       struct net_device *dev,
1530 					       const unsigned char *addr,
1531 					       u16 vid, u32 portid, u32 seq,
1532 					       struct netlink_ext_ack *extack);
1533 	int			(*ndo_mdb_add)(struct net_device *dev,
1534 					       struct nlattr *tb[],
1535 					       u16 nlmsg_flags,
1536 					       struct netlink_ext_ack *extack);
1537 	int			(*ndo_mdb_del)(struct net_device *dev,
1538 					       struct nlattr *tb[],
1539 					       struct netlink_ext_ack *extack);
1540 	int			(*ndo_mdb_del_bulk)(struct net_device *dev,
1541 						    struct nlattr *tb[],
1542 						    struct netlink_ext_ack *extack);
1543 	int			(*ndo_mdb_dump)(struct net_device *dev,
1544 						struct sk_buff *skb,
1545 						struct netlink_callback *cb);
1546 	int			(*ndo_mdb_get)(struct net_device *dev,
1547 					       struct nlattr *tb[], u32 portid,
1548 					       u32 seq,
1549 					       struct netlink_ext_ack *extack);
1550 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1551 						      struct nlmsghdr *nlh,
1552 						      u16 flags,
1553 						      struct netlink_ext_ack *extack);
1554 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1555 						      u32 pid, u32 seq,
1556 						      struct net_device *dev,
1557 						      u32 filter_mask,
1558 						      int nlflags);
1559 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1560 						      struct nlmsghdr *nlh,
1561 						      u16 flags);
1562 	int			(*ndo_change_carrier)(struct net_device *dev,
1563 						      bool new_carrier);
1564 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1565 							struct netdev_phys_item_id *ppid);
1566 	int			(*ndo_get_port_parent_id)(struct net_device *dev,
1567 							  struct netdev_phys_item_id *ppid);
1568 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1569 							  char *name, size_t len);
1570 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1571 							struct net_device *dev);
1572 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1573 							void *priv);
1574 
1575 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1576 						      int queue_index,
1577 						      u32 maxrate);
1578 	int			(*ndo_get_iflink)(const struct net_device *dev);
1579 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1580 						       struct sk_buff *skb);
1581 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1582 						       int needed_headroom);
1583 	int			(*ndo_bpf)(struct net_device *dev,
1584 					   struct netdev_bpf *bpf);
1585 	int			(*ndo_xdp_xmit)(struct net_device *dev, int n,
1586 						struct xdp_frame **xdp,
1587 						u32 flags);
1588 	struct net_device *	(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1589 							  struct xdp_buff *xdp);
1590 	int			(*ndo_xsk_wakeup)(struct net_device *dev,
1591 						  u32 queue_id, u32 flags);
1592 	int			(*ndo_tunnel_ctl)(struct net_device *dev,
1593 						  struct ip_tunnel_parm_kern *p,
1594 						  int cmd);
1595 	struct net_device *	(*ndo_get_peer_dev)(struct net_device *dev);
1596 	int                     (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1597                                                          struct net_device_path *path);
1598 	ktime_t			(*ndo_get_tstamp)(struct net_device *dev,
1599 						  const struct skb_shared_hwtstamps *hwtstamps,
1600 						  bool cycles);
1601 	int			(*ndo_hwtstamp_get)(struct net_device *dev,
1602 						    struct kernel_hwtstamp_config *kernel_config);
1603 	int			(*ndo_hwtstamp_set)(struct net_device *dev,
1604 						    struct kernel_hwtstamp_config *kernel_config,
1605 						    struct netlink_ext_ack *extack);
1606 
1607 #if IS_ENABLED(CONFIG_NET_SHAPER)
1608 	/**
1609 	 * @net_shaper_ops: Device shaping offload operations
1610 	 * see include/net/net_shapers.h
1611 	 */
1612 	const struct net_shaper_ops *net_shaper_ops;
1613 #endif
1614 };
1615 
1616 /**
1617  * enum netdev_priv_flags - &struct net_device priv_flags
1618  *
1619  * These are the &struct net_device, they are only set internally
1620  * by drivers and used in the kernel. These flags are invisible to
1621  * userspace; this means that the order of these flags can change
1622  * during any kernel release.
1623  *
1624  * You should add bitfield booleans after either net_device::priv_flags
1625  * (hotpath) or ::threaded (slowpath) instead of extending these flags.
1626  *
1627  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1628  * @IFF_EBRIDGE: Ethernet bridging device
1629  * @IFF_BONDING: bonding master or slave
1630  * @IFF_ISATAP: ISATAP interface (RFC4214)
1631  * @IFF_WAN_HDLC: WAN HDLC device
1632  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1633  *	release skb->dst
1634  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1635  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1636  * @IFF_MACVLAN_PORT: device used as macvlan port
1637  * @IFF_BRIDGE_PORT: device used as bridge port
1638  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1639  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1640  * @IFF_UNICAST_FLT: Supports unicast filtering
1641  * @IFF_TEAM_PORT: device used as team port
1642  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1643  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1644  *	change when it's running
1645  * @IFF_MACVLAN: Macvlan device
1646  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1647  *	underlying stacked devices
1648  * @IFF_L3MDEV_MASTER: device is an L3 master device
1649  * @IFF_NO_QUEUE: device can run without qdisc attached
1650  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1651  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1652  * @IFF_TEAM: device is a team device
1653  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1654  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1655  *	entity (i.e. the master device for bridged veth)
1656  * @IFF_MACSEC: device is a MACsec device
1657  * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1658  * @IFF_FAILOVER: device is a failover master device
1659  * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1660  * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1661  * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1662  * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1663  *	skb_headlen(skb) == 0 (data starts from frag0)
1664  */
1665 enum netdev_priv_flags {
1666 	IFF_802_1Q_VLAN			= 1<<0,
1667 	IFF_EBRIDGE			= 1<<1,
1668 	IFF_BONDING			= 1<<2,
1669 	IFF_ISATAP			= 1<<3,
1670 	IFF_WAN_HDLC			= 1<<4,
1671 	IFF_XMIT_DST_RELEASE		= 1<<5,
1672 	IFF_DONT_BRIDGE			= 1<<6,
1673 	IFF_DISABLE_NETPOLL		= 1<<7,
1674 	IFF_MACVLAN_PORT		= 1<<8,
1675 	IFF_BRIDGE_PORT			= 1<<9,
1676 	IFF_OVS_DATAPATH		= 1<<10,
1677 	IFF_TX_SKB_SHARING		= 1<<11,
1678 	IFF_UNICAST_FLT			= 1<<12,
1679 	IFF_TEAM_PORT			= 1<<13,
1680 	IFF_SUPP_NOFCS			= 1<<14,
1681 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1682 	IFF_MACVLAN			= 1<<16,
1683 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1684 	IFF_L3MDEV_MASTER		= 1<<18,
1685 	IFF_NO_QUEUE			= 1<<19,
1686 	IFF_OPENVSWITCH			= 1<<20,
1687 	IFF_L3MDEV_SLAVE		= 1<<21,
1688 	IFF_TEAM			= 1<<22,
1689 	IFF_RXFH_CONFIGURED		= 1<<23,
1690 	IFF_PHONY_HEADROOM		= 1<<24,
1691 	IFF_MACSEC			= 1<<25,
1692 	IFF_NO_RX_HANDLER		= 1<<26,
1693 	IFF_FAILOVER			= 1<<27,
1694 	IFF_FAILOVER_SLAVE		= 1<<28,
1695 	IFF_L3MDEV_RX_HANDLER		= 1<<29,
1696 	IFF_NO_ADDRCONF			= BIT_ULL(30),
1697 	IFF_TX_SKB_NO_LINEAR		= BIT_ULL(31),
1698 };
1699 
1700 /* Specifies the type of the struct net_device::ml_priv pointer */
1701 enum netdev_ml_priv_type {
1702 	ML_PRIV_NONE,
1703 	ML_PRIV_CAN,
1704 };
1705 
1706 enum netdev_stat_type {
1707 	NETDEV_PCPU_STAT_NONE,
1708 	NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1709 	NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1710 	NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1711 };
1712 
1713 enum netdev_reg_state {
1714 	NETREG_UNINITIALIZED = 0,
1715 	NETREG_REGISTERED,	/* completed register_netdevice */
1716 	NETREG_UNREGISTERING,	/* called unregister_netdevice */
1717 	NETREG_UNREGISTERED,	/* completed unregister todo */
1718 	NETREG_RELEASED,	/* called free_netdev */
1719 	NETREG_DUMMY,		/* dummy device for NAPI poll */
1720 };
1721 
1722 /**
1723  *	struct net_device - The DEVICE structure.
1724  *
1725  *	Actually, this whole structure is a big mistake.  It mixes I/O
1726  *	data with strictly "high-level" data, and it has to know about
1727  *	almost every data structure used in the INET module.
1728  *
1729  *	@priv_flags:	flags invisible to userspace defined as bits, see
1730  *			enum netdev_priv_flags for the definitions
1731  *	@lltx:		device supports lockless Tx. Deprecated for real HW
1732  *			drivers. Mainly used by logical interfaces, such as
1733  *			bonding and tunnels
1734  *
1735  *	@name:	This is the first field of the "visible" part of this structure
1736  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1737  *		of the interface.
1738  *
1739  *	@name_node:	Name hashlist node
1740  *	@ifalias:	SNMP alias
1741  *	@mem_end:	Shared memory end
1742  *	@mem_start:	Shared memory start
1743  *	@base_addr:	Device I/O address
1744  *	@irq:		Device IRQ number
1745  *
1746  *	@state:		Generic network queuing layer state, see netdev_state_t
1747  *	@dev_list:	The global list of network devices
1748  *	@napi_list:	List entry used for polling NAPI devices
1749  *	@unreg_list:	List entry  when we are unregistering the
1750  *			device; see the function unregister_netdev
1751  *	@close_list:	List entry used when we are closing the device
1752  *	@ptype_all:     Device-specific packet handlers for all protocols
1753  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1754  *
1755  *	@adj_list:	Directly linked devices, like slaves for bonding
1756  *	@features:	Currently active device features
1757  *	@hw_features:	User-changeable features
1758  *
1759  *	@wanted_features:	User-requested features
1760  *	@vlan_features:		Mask of features inheritable by VLAN devices
1761  *
1762  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1763  *				This field indicates what encapsulation
1764  *				offloads the hardware is capable of doing,
1765  *				and drivers will need to set them appropriately.
1766  *
1767  *	@mpls_features:	Mask of features inheritable by MPLS
1768  *	@gso_partial_features: value(s) from NETIF_F_GSO\*
1769  *
1770  *	@ifindex:	interface index
1771  *	@group:		The group the device belongs to
1772  *
1773  *	@stats:		Statistics struct, which was left as a legacy, use
1774  *			rtnl_link_stats64 instead
1775  *
1776  *	@core_stats:	core networking counters,
1777  *			do not use this in drivers
1778  *	@carrier_up_count:	Number of times the carrier has been up
1779  *	@carrier_down_count:	Number of times the carrier has been down
1780  *
1781  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1782  *				instead of ioctl,
1783  *				see <net/iw_handler.h> for details.
1784  *	@wireless_data:	Instance data managed by the core of wireless extensions
1785  *
1786  *	@netdev_ops:	Includes several pointers to callbacks,
1787  *			if one wants to override the ndo_*() functions
1788  *	@xdp_metadata_ops:	Includes pointers to XDP metadata callbacks.
1789  *	@xsk_tx_metadata_ops:	Includes pointers to AF_XDP TX metadata callbacks.
1790  *	@ethtool_ops:	Management operations
1791  *	@l3mdev_ops:	Layer 3 master device operations
1792  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1793  *			discovery handling. Necessary for e.g. 6LoWPAN.
1794  *	@xfrmdev_ops:	Transformation offload operations
1795  *	@tlsdev_ops:	Transport Layer Security offload operations
1796  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1797  *			of Layer 2 headers.
1798  *
1799  *	@flags:		Interface flags (a la BSD)
1800  *	@xdp_features:	XDP capability supported by the device
1801  *	@gflags:	Global flags ( kept as legacy )
1802  *	@priv_len:	Size of the ->priv flexible array
1803  *	@priv:		Flexible array containing private data
1804  *	@operstate:	RFC2863 operstate
1805  *	@link_mode:	Mapping policy to operstate
1806  *	@if_port:	Selectable AUI, TP, ...
1807  *	@dma:		DMA channel
1808  *	@mtu:		Interface MTU value
1809  *	@min_mtu:	Interface Minimum MTU value
1810  *	@max_mtu:	Interface Maximum MTU value
1811  *	@type:		Interface hardware type
1812  *	@hard_header_len: Maximum hardware header length.
1813  *	@min_header_len:  Minimum hardware header length
1814  *
1815  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1816  *			  cases can this be guaranteed
1817  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1818  *			  cases can this be guaranteed. Some cases also use
1819  *			  LL_MAX_HEADER instead to allocate the skb
1820  *
1821  *	interface address info:
1822  *
1823  * 	@perm_addr:		Permanent hw address
1824  * 	@addr_assign_type:	Hw address assignment type
1825  * 	@addr_len:		Hardware address length
1826  *	@upper_level:		Maximum depth level of upper devices.
1827  *	@lower_level:		Maximum depth level of lower devices.
1828  *	@neigh_priv_len:	Used in neigh_alloc()
1829  * 	@dev_id:		Used to differentiate devices that share
1830  * 				the same link layer address
1831  * 	@dev_port:		Used to differentiate devices that share
1832  * 				the same function
1833  *	@addr_list_lock:	XXX: need comments on this one
1834  *	@name_assign_type:	network interface name assignment type
1835  *	@uc_promisc:		Counter that indicates promiscuous mode
1836  *				has been enabled due to the need to listen to
1837  *				additional unicast addresses in a device that
1838  *				does not implement ndo_set_rx_mode()
1839  *	@uc:			unicast mac addresses
1840  *	@mc:			multicast mac addresses
1841  *	@dev_addrs:		list of device hw addresses
1842  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1843  *	@promiscuity:		Number of times the NIC is told to work in
1844  *				promiscuous mode; if it becomes 0 the NIC will
1845  *				exit promiscuous mode
1846  *	@allmulti:		Counter, enables or disables allmulticast mode
1847  *
1848  *	@vlan_info:	VLAN info
1849  *	@dsa_ptr:	dsa specific data
1850  *	@tipc_ptr:	TIPC specific data
1851  *	@atalk_ptr:	AppleTalk link
1852  *	@ip_ptr:	IPv4 specific data
1853  *	@ip6_ptr:	IPv6 specific data
1854  *	@ax25_ptr:	AX.25 specific data
1855  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1856  *	@ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1857  *			 device struct
1858  *	@mpls_ptr:	mpls_dev struct pointer
1859  *	@mctp_ptr:	MCTP specific data
1860  *
1861  *	@dev_addr:	Hw address (before bcast,
1862  *			because most packets are unicast)
1863  *
1864  *	@_rx:			Array of RX queues
1865  *	@num_rx_queues:		Number of RX queues
1866  *				allocated at register_netdev() time
1867  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1868  *	@xdp_prog:		XDP sockets filter program pointer
1869  *	@gro_flush_timeout:	timeout for GRO layer in NAPI
1870  *	@napi_defer_hard_irqs:	If not zero, provides a counter that would
1871  *				allow to avoid NIC hard IRQ, on busy queues.
1872  *
1873  *	@rx_handler:		handler for received packets
1874  *	@rx_handler_data: 	XXX: need comments on this one
1875  *	@tcx_ingress:		BPF & clsact qdisc specific data for ingress processing
1876  *	@ingress_queue:		XXX: need comments on this one
1877  *	@nf_hooks_ingress:	netfilter hooks executed for ingress packets
1878  *	@broadcast:		hw bcast address
1879  *
1880  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1881  *			indexed by RX queue number. Assigned by driver.
1882  *			This must only be set if the ndo_rx_flow_steer
1883  *			operation is defined
1884  *	@index_hlist:		Device index hash chain
1885  *
1886  *	@_tx:			Array of TX queues
1887  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1888  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1889  *	@qdisc:			Root qdisc from userspace point of view
1890  *	@tx_queue_len:		Max frames per queue allowed
1891  *	@tx_global_lock: 	XXX: need comments on this one
1892  *	@xdp_bulkq:		XDP device bulk queue
1893  *	@xps_maps:		all CPUs/RXQs maps for XPS device
1894  *
1895  *	@xps_maps:	XXX: need comments on this one
1896  *	@tcx_egress:		BPF & clsact qdisc specific data for egress processing
1897  *	@nf_hooks_egress:	netfilter hooks executed for egress packets
1898  *	@qdisc_hash:		qdisc hash table
1899  *	@watchdog_timeo:	Represents the timeout that is used by
1900  *				the watchdog (see dev_watchdog())
1901  *	@watchdog_timer:	List of timers
1902  *
1903  *	@proto_down_reason:	reason a netdev interface is held down
1904  *	@pcpu_refcnt:		Number of references to this device
1905  *	@dev_refcnt:		Number of references to this device
1906  *	@refcnt_tracker:	Tracker directory for tracked references to this device
1907  *	@todo_list:		Delayed register/unregister
1908  *	@link_watch_list:	XXX: need comments on this one
1909  *
1910  *	@reg_state:		Register/unregister state machine
1911  *	@dismantle:		Device is going to be freed
1912  *	@rtnl_link_state:	This enum represents the phases of creating
1913  *				a new link
1914  *
1915  *	@needs_free_netdev:	Should unregister perform free_netdev?
1916  *	@priv_destructor:	Called from unregister
1917  *	@npinfo:		XXX: need comments on this one
1918  * 	@nd_net:		Network namespace this network device is inside
1919  *
1920  * 	@ml_priv:	Mid-layer private
1921  *	@ml_priv_type:  Mid-layer private type
1922  *
1923  *	@pcpu_stat_type:	Type of device statistics which the core should
1924  *				allocate/free: none, lstats, tstats, dstats. none
1925  *				means the driver is handling statistics allocation/
1926  *				freeing internally.
1927  *	@lstats:		Loopback statistics: packets, bytes
1928  *	@tstats:		Tunnel statistics: RX/TX packets, RX/TX bytes
1929  *	@dstats:		Dummy statistics: RX/TX/drop packets, RX/TX bytes
1930  *
1931  *	@garp_port:	GARP
1932  *	@mrp_port:	MRP
1933  *
1934  *	@dm_private:	Drop monitor private
1935  *
1936  *	@dev:		Class/net/name entry
1937  *	@sysfs_groups:	Space for optional device, statistics and wireless
1938  *			sysfs groups
1939  *
1940  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1941  *	@rtnl_link_ops:	Rtnl_link_ops
1942  *	@stat_ops:	Optional ops for queue-aware statistics
1943  *	@queue_mgmt_ops:	Optional ops for queue management
1944  *
1945  *	@gso_max_size:	Maximum size of generic segmentation offload
1946  *	@tso_max_size:	Device (as in HW) limit on the max TSO request size
1947  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1948  *			NIC for GSO
1949  *	@tso_max_segs:	Device (as in HW) limit on the max TSO segment count
1950  * 	@gso_ipv4_max_size:	Maximum size of generic segmentation offload,
1951  * 				for IPv4.
1952  *
1953  *	@dcbnl_ops:	Data Center Bridging netlink ops
1954  *	@num_tc:	Number of traffic classes in the net device
1955  *	@tc_to_txq:	XXX: need comments on this one
1956  *	@prio_tc_map:	XXX: need comments on this one
1957  *
1958  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1959  *
1960  *	@priomap:	XXX: need comments on this one
1961  *	@link_topo:	Physical link topology tracking attached PHYs
1962  *	@phydev:	Physical device may attach itself
1963  *			for hardware timestamping
1964  *	@sfp_bus:	attached &struct sfp_bus structure.
1965  *
1966  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1967  *
1968  *	@proto_down:	protocol port state information can be sent to the
1969  *			switch driver and used to set the phys state of the
1970  *			switch port.
1971  *
1972  *	@threaded:	napi threaded mode is enabled
1973  *
1974  *	@see_all_hwtstamp_requests: device wants to see calls to
1975  *			ndo_hwtstamp_set() for all timestamp requests
1976  *			regardless of source, even if those aren't
1977  *			HWTSTAMP_SOURCE_NETDEV
1978  *	@change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN
1979  *	@netns_local: interface can't change network namespaces
1980  *	@fcoe_mtu:	device supports maximum FCoE MTU, 2158 bytes
1981  *
1982  *	@net_notifier_list:	List of per-net netdev notifier block
1983  *				that follow this device when it is moved
1984  *				to another network namespace.
1985  *
1986  *	@macsec_ops:    MACsec offloading ops
1987  *
1988  *	@udp_tunnel_nic_info:	static structure describing the UDP tunnel
1989  *				offload capabilities of the device
1990  *	@udp_tunnel_nic:	UDP tunnel offload state
1991  *	@ethtool:	ethtool related state
1992  *	@xdp_state:		stores info on attached XDP BPF programs
1993  *
1994  *	@nested_level:	Used as a parameter of spin_lock_nested() of
1995  *			dev->addr_list_lock.
1996  *	@unlink_list:	As netif_addr_lock() can be called recursively,
1997  *			keep a list of interfaces to be deleted.
1998  *	@gro_max_size:	Maximum size of aggregated packet in generic
1999  *			receive offload (GRO)
2000  * 	@gro_ipv4_max_size:	Maximum size of aggregated packet in generic
2001  * 				receive offload (GRO), for IPv4.
2002  *	@xdp_zc_max_segs:	Maximum number of segments supported by AF_XDP
2003  *				zero copy driver
2004  *
2005  *	@dev_addr_shadow:	Copy of @dev_addr to catch direct writes.
2006  *	@linkwatch_dev_tracker:	refcount tracker used by linkwatch.
2007  *	@watchdog_dev_tracker:	refcount tracker used by watchdog.
2008  *	@dev_registered_tracker:	tracker for reference held while
2009  *					registered
2010  *	@offload_xstats_l3:	L3 HW stats for this netdevice.
2011  *
2012  *	@devlink_port:	Pointer to related devlink port structure.
2013  *			Assigned by a driver before netdev registration using
2014  *			SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2015  *			during the time netdevice is registered.
2016  *
2017  *	@dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2018  *		   where the clock is recovered.
2019  *
2020  *	@max_pacing_offload_horizon: max EDT offload horizon in nsec.
2021  *
2022  *	FIXME: cleanup struct net_device such that network protocol info
2023  *	moves out.
2024  */
2025 
2026 struct net_device {
2027 	/* Cacheline organization can be found documented in
2028 	 * Documentation/networking/net_cachelines/net_device.rst.
2029 	 * Please update the document when adding new fields.
2030 	 */
2031 
2032 	/* TX read-mostly hotpath */
2033 	__cacheline_group_begin(net_device_read_tx);
2034 	struct_group(priv_flags_fast,
2035 		unsigned long		priv_flags:32;
2036 		unsigned long		lltx:1;
2037 	);
2038 	const struct net_device_ops *netdev_ops;
2039 	const struct header_ops *header_ops;
2040 	struct netdev_queue	*_tx;
2041 	netdev_features_t	gso_partial_features;
2042 	unsigned int		real_num_tx_queues;
2043 	unsigned int		gso_max_size;
2044 	unsigned int		gso_ipv4_max_size;
2045 	u16			gso_max_segs;
2046 	s16			num_tc;
2047 	/* Note : dev->mtu is often read without holding a lock.
2048 	 * Writers usually hold RTNL.
2049 	 * It is recommended to use READ_ONCE() to annotate the reads,
2050 	 * and to use WRITE_ONCE() to annotate the writes.
2051 	 */
2052 	unsigned int		mtu;
2053 	unsigned short		needed_headroom;
2054 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
2055 #ifdef CONFIG_XPS
2056 	struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2057 #endif
2058 #ifdef CONFIG_NETFILTER_EGRESS
2059 	struct nf_hook_entries __rcu *nf_hooks_egress;
2060 #endif
2061 #ifdef CONFIG_NET_XGRESS
2062 	struct bpf_mprog_entry __rcu *tcx_egress;
2063 #endif
2064 	__cacheline_group_end(net_device_read_tx);
2065 
2066 	/* TXRX read-mostly hotpath */
2067 	__cacheline_group_begin(net_device_read_txrx);
2068 	union {
2069 		struct pcpu_lstats __percpu		*lstats;
2070 		struct pcpu_sw_netstats __percpu	*tstats;
2071 		struct pcpu_dstats __percpu		*dstats;
2072 	};
2073 	unsigned long		state;
2074 	unsigned int		flags;
2075 	unsigned short		hard_header_len;
2076 	netdev_features_t	features;
2077 	struct inet6_dev __rcu	*ip6_ptr;
2078 	__cacheline_group_end(net_device_read_txrx);
2079 
2080 	/* RX read-mostly hotpath */
2081 	__cacheline_group_begin(net_device_read_rx);
2082 	struct bpf_prog __rcu	*xdp_prog;
2083 	struct list_head	ptype_specific;
2084 	int			ifindex;
2085 	unsigned int		real_num_rx_queues;
2086 	struct netdev_rx_queue	*_rx;
2087 	unsigned long		gro_flush_timeout;
2088 	u32			napi_defer_hard_irqs;
2089 	unsigned int		gro_max_size;
2090 	unsigned int		gro_ipv4_max_size;
2091 	rx_handler_func_t __rcu	*rx_handler;
2092 	void __rcu		*rx_handler_data;
2093 	possible_net_t			nd_net;
2094 #ifdef CONFIG_NETPOLL
2095 	struct netpoll_info __rcu	*npinfo;
2096 #endif
2097 #ifdef CONFIG_NET_XGRESS
2098 	struct bpf_mprog_entry __rcu *tcx_ingress;
2099 #endif
2100 	__cacheline_group_end(net_device_read_rx);
2101 
2102 	char			name[IFNAMSIZ];
2103 	struct netdev_name_node	*name_node;
2104 	struct dev_ifalias	__rcu *ifalias;
2105 	/*
2106 	 *	I/O specific fields
2107 	 *	FIXME: Merge these and struct ifmap into one
2108 	 */
2109 	unsigned long		mem_end;
2110 	unsigned long		mem_start;
2111 	unsigned long		base_addr;
2112 
2113 	/*
2114 	 *	Some hardware also needs these fields (state,dev_list,
2115 	 *	napi_list,unreg_list,close_list) but they are not
2116 	 *	part of the usual set specified in Space.c.
2117 	 */
2118 
2119 
2120 	struct list_head	dev_list;
2121 	struct list_head	napi_list;
2122 	struct list_head	unreg_list;
2123 	struct list_head	close_list;
2124 	struct list_head	ptype_all;
2125 
2126 	struct {
2127 		struct list_head upper;
2128 		struct list_head lower;
2129 	} adj_list;
2130 
2131 	/* Read-mostly cache-line for fast-path access */
2132 	xdp_features_t		xdp_features;
2133 	const struct xdp_metadata_ops *xdp_metadata_ops;
2134 	const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2135 	unsigned short		gflags;
2136 
2137 	unsigned short		needed_tailroom;
2138 
2139 	netdev_features_t	hw_features;
2140 	netdev_features_t	wanted_features;
2141 	netdev_features_t	vlan_features;
2142 	netdev_features_t	hw_enc_features;
2143 	netdev_features_t	mpls_features;
2144 
2145 	unsigned int		min_mtu;
2146 	unsigned int		max_mtu;
2147 	unsigned short		type;
2148 	unsigned char		min_header_len;
2149 	unsigned char		name_assign_type;
2150 
2151 	int			group;
2152 
2153 	struct net_device_stats	stats; /* not used by modern drivers */
2154 
2155 	struct net_device_core_stats __percpu *core_stats;
2156 
2157 	/* Stats to monitor link on/off, flapping */
2158 	atomic_t		carrier_up_count;
2159 	atomic_t		carrier_down_count;
2160 
2161 #ifdef CONFIG_WIRELESS_EXT
2162 	const struct iw_handler_def *wireless_handlers;
2163 	struct iw_public_data	*wireless_data;
2164 #endif
2165 	const struct ethtool_ops *ethtool_ops;
2166 #ifdef CONFIG_NET_L3_MASTER_DEV
2167 	const struct l3mdev_ops	*l3mdev_ops;
2168 #endif
2169 #if IS_ENABLED(CONFIG_IPV6)
2170 	const struct ndisc_ops *ndisc_ops;
2171 #endif
2172 
2173 #ifdef CONFIG_XFRM_OFFLOAD
2174 	const struct xfrmdev_ops *xfrmdev_ops;
2175 #endif
2176 
2177 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2178 	const struct tlsdev_ops *tlsdev_ops;
2179 #endif
2180 
2181 	unsigned int		operstate;
2182 	unsigned char		link_mode;
2183 
2184 	unsigned char		if_port;
2185 	unsigned char		dma;
2186 
2187 	/* Interface address info. */
2188 	unsigned char		perm_addr[MAX_ADDR_LEN];
2189 	unsigned char		addr_assign_type;
2190 	unsigned char		addr_len;
2191 	unsigned char		upper_level;
2192 	unsigned char		lower_level;
2193 
2194 	unsigned short		neigh_priv_len;
2195 	unsigned short          dev_id;
2196 	unsigned short          dev_port;
2197 	int			irq;
2198 	u32			priv_len;
2199 
2200 	spinlock_t		addr_list_lock;
2201 
2202 	struct netdev_hw_addr_list	uc;
2203 	struct netdev_hw_addr_list	mc;
2204 	struct netdev_hw_addr_list	dev_addrs;
2205 
2206 #ifdef CONFIG_SYSFS
2207 	struct kset		*queues_kset;
2208 #endif
2209 #ifdef CONFIG_LOCKDEP
2210 	struct list_head	unlink_list;
2211 #endif
2212 	unsigned int		promiscuity;
2213 	unsigned int		allmulti;
2214 	bool			uc_promisc;
2215 #ifdef CONFIG_LOCKDEP
2216 	unsigned char		nested_level;
2217 #endif
2218 
2219 
2220 	/* Protocol-specific pointers */
2221 	struct in_device __rcu	*ip_ptr;
2222 	/** @fib_nh_head: nexthops associated with this netdev */
2223 	struct hlist_head	fib_nh_head;
2224 
2225 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2226 	struct vlan_info __rcu	*vlan_info;
2227 #endif
2228 #if IS_ENABLED(CONFIG_NET_DSA)
2229 	struct dsa_port		*dsa_ptr;
2230 #endif
2231 #if IS_ENABLED(CONFIG_TIPC)
2232 	struct tipc_bearer __rcu *tipc_ptr;
2233 #endif
2234 #if IS_ENABLED(CONFIG_ATALK)
2235 	void 			*atalk_ptr;
2236 #endif
2237 #if IS_ENABLED(CONFIG_AX25)
2238 	void			*ax25_ptr;
2239 #endif
2240 #if IS_ENABLED(CONFIG_CFG80211)
2241 	struct wireless_dev	*ieee80211_ptr;
2242 #endif
2243 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2244 	struct wpan_dev		*ieee802154_ptr;
2245 #endif
2246 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2247 	struct mpls_dev __rcu	*mpls_ptr;
2248 #endif
2249 #if IS_ENABLED(CONFIG_MCTP)
2250 	struct mctp_dev __rcu	*mctp_ptr;
2251 #endif
2252 
2253 /*
2254  * Cache lines mostly used on receive path (including eth_type_trans())
2255  */
2256 	/* Interface address info used in eth_type_trans() */
2257 	const unsigned char	*dev_addr;
2258 
2259 	unsigned int		num_rx_queues;
2260 #define GRO_LEGACY_MAX_SIZE	65536u
2261 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2262  * and shinfo->gso_segs is a 16bit field.
2263  */
2264 #define GRO_MAX_SIZE		(8 * 65535u)
2265 	unsigned int		xdp_zc_max_segs;
2266 	struct netdev_queue __rcu *ingress_queue;
2267 #ifdef CONFIG_NETFILTER_INGRESS
2268 	struct nf_hook_entries __rcu *nf_hooks_ingress;
2269 #endif
2270 
2271 	unsigned char		broadcast[MAX_ADDR_LEN];
2272 #ifdef CONFIG_RFS_ACCEL
2273 	struct cpu_rmap		*rx_cpu_rmap;
2274 #endif
2275 	struct hlist_node	index_hlist;
2276 
2277 /*
2278  * Cache lines mostly used on transmit path
2279  */
2280 	unsigned int		num_tx_queues;
2281 	struct Qdisc __rcu	*qdisc;
2282 	unsigned int		tx_queue_len;
2283 	spinlock_t		tx_global_lock;
2284 
2285 	struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2286 
2287 #ifdef CONFIG_NET_SCHED
2288 	DECLARE_HASHTABLE	(qdisc_hash, 4);
2289 #endif
2290 	/* These may be needed for future network-power-down code. */
2291 	struct timer_list	watchdog_timer;
2292 	int			watchdog_timeo;
2293 
2294 	u32                     proto_down_reason;
2295 
2296 	struct list_head	todo_list;
2297 
2298 #ifdef CONFIG_PCPU_DEV_REFCNT
2299 	int __percpu		*pcpu_refcnt;
2300 #else
2301 	refcount_t		dev_refcnt;
2302 #endif
2303 	struct ref_tracker_dir	refcnt_tracker;
2304 
2305 	struct list_head	link_watch_list;
2306 
2307 	u8 reg_state;
2308 
2309 	bool dismantle;
2310 
2311 	enum {
2312 		RTNL_LINK_INITIALIZED,
2313 		RTNL_LINK_INITIALIZING,
2314 	} rtnl_link_state:16;
2315 
2316 	bool needs_free_netdev;
2317 	void (*priv_destructor)(struct net_device *dev);
2318 
2319 	/* mid-layer private */
2320 	void				*ml_priv;
2321 	enum netdev_ml_priv_type	ml_priv_type;
2322 
2323 	enum netdev_stat_type		pcpu_stat_type:8;
2324 
2325 #if IS_ENABLED(CONFIG_GARP)
2326 	struct garp_port __rcu	*garp_port;
2327 #endif
2328 #if IS_ENABLED(CONFIG_MRP)
2329 	struct mrp_port __rcu	*mrp_port;
2330 #endif
2331 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2332 	struct dm_hw_stat_delta __rcu *dm_private;
2333 #endif
2334 	struct device		dev;
2335 	const struct attribute_group *sysfs_groups[4];
2336 	const struct attribute_group *sysfs_rx_queue_group;
2337 
2338 	const struct rtnl_link_ops *rtnl_link_ops;
2339 
2340 	const struct netdev_stat_ops *stat_ops;
2341 
2342 	const struct netdev_queue_mgmt_ops *queue_mgmt_ops;
2343 
2344 	/* for setting kernel sock attribute on TCP connection setup */
2345 #define GSO_MAX_SEGS		65535u
2346 #define GSO_LEGACY_MAX_SIZE	65536u
2347 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2348  * and shinfo->gso_segs is a 16bit field.
2349  */
2350 #define GSO_MAX_SIZE		(8 * GSO_MAX_SEGS)
2351 
2352 #define TSO_LEGACY_MAX_SIZE	65536
2353 #define TSO_MAX_SIZE		UINT_MAX
2354 	unsigned int		tso_max_size;
2355 #define TSO_MAX_SEGS		U16_MAX
2356 	u16			tso_max_segs;
2357 
2358 #ifdef CONFIG_DCB
2359 	const struct dcbnl_rtnl_ops *dcbnl_ops;
2360 #endif
2361 	u8			prio_tc_map[TC_BITMASK + 1];
2362 
2363 #if IS_ENABLED(CONFIG_FCOE)
2364 	unsigned int		fcoe_ddp_xid;
2365 #endif
2366 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2367 	struct netprio_map __rcu *priomap;
2368 #endif
2369 	struct phy_link_topology	*link_topo;
2370 	struct phy_device	*phydev;
2371 	struct sfp_bus		*sfp_bus;
2372 	struct lock_class_key	*qdisc_tx_busylock;
2373 	bool			proto_down;
2374 	bool			threaded;
2375 
2376 	/* priv_flags_slow, ungrouped to save space */
2377 	unsigned long		see_all_hwtstamp_requests:1;
2378 	unsigned long		change_proto_down:1;
2379 	unsigned long		netns_local:1;
2380 	unsigned long		fcoe_mtu:1;
2381 
2382 	struct list_head	net_notifier_list;
2383 
2384 #if IS_ENABLED(CONFIG_MACSEC)
2385 	/* MACsec management functions */
2386 	const struct macsec_ops *macsec_ops;
2387 #endif
2388 	const struct udp_tunnel_nic_info	*udp_tunnel_nic_info;
2389 	struct udp_tunnel_nic	*udp_tunnel_nic;
2390 
2391 	struct ethtool_netdev_state *ethtool;
2392 
2393 	/* protected by rtnl_lock */
2394 	struct bpf_xdp_entity	xdp_state[__MAX_XDP_MODE];
2395 
2396 	u8 dev_addr_shadow[MAX_ADDR_LEN];
2397 	netdevice_tracker	linkwatch_dev_tracker;
2398 	netdevice_tracker	watchdog_dev_tracker;
2399 	netdevice_tracker	dev_registered_tracker;
2400 	struct rtnl_hw_stats64	*offload_xstats_l3;
2401 
2402 	struct devlink_port	*devlink_port;
2403 
2404 #if IS_ENABLED(CONFIG_DPLL)
2405 	struct dpll_pin	__rcu	*dpll_pin;
2406 #endif
2407 #if IS_ENABLED(CONFIG_PAGE_POOL)
2408 	/** @page_pools: page pools created for this netdevice */
2409 	struct hlist_head	page_pools;
2410 #endif
2411 
2412 	/** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */
2413 	struct dim_irq_moder	*irq_moder;
2414 
2415 	u64			max_pacing_offload_horizon;
2416 
2417 	/**
2418 	 * @lock: protects @net_shaper_hierarchy, feel free to use for other
2419 	 * netdev-scope protection. Ordering: take after rtnl_lock.
2420 	 */
2421 	struct mutex		lock;
2422 
2423 #if IS_ENABLED(CONFIG_NET_SHAPER)
2424 	/**
2425 	 * @net_shaper_hierarchy: data tracking the current shaper status
2426 	 *  see include/net/net_shapers.h
2427 	 */
2428 	struct net_shaper_hierarchy *net_shaper_hierarchy;
2429 #endif
2430 	u8			priv[] ____cacheline_aligned
2431 				       __counted_by(priv_len);
2432 } ____cacheline_aligned;
2433 #define to_net_dev(d) container_of(d, struct net_device, dev)
2434 
2435 /*
2436  * Driver should use this to assign devlink port instance to a netdevice
2437  * before it registers the netdevice. Therefore devlink_port is static
2438  * during the netdev lifetime after it is registered.
2439  */
2440 #define SET_NETDEV_DEVLINK_PORT(dev, port)			\
2441 ({								\
2442 	WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED);	\
2443 	((dev)->devlink_port = (port));				\
2444 })
2445 
2446 static inline bool netif_elide_gro(const struct net_device *dev)
2447 {
2448 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2449 		return true;
2450 	return false;
2451 }
2452 
2453 #define	NETDEV_ALIGN		32
2454 
2455 static inline
2456 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2457 {
2458 	return dev->prio_tc_map[prio & TC_BITMASK];
2459 }
2460 
2461 static inline
2462 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2463 {
2464 	if (tc >= dev->num_tc)
2465 		return -EINVAL;
2466 
2467 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2468 	return 0;
2469 }
2470 
2471 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2472 void netdev_reset_tc(struct net_device *dev);
2473 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2474 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2475 
2476 static inline
2477 int netdev_get_num_tc(struct net_device *dev)
2478 {
2479 	return dev->num_tc;
2480 }
2481 
2482 static inline void net_prefetch(void *p)
2483 {
2484 	prefetch(p);
2485 #if L1_CACHE_BYTES < 128
2486 	prefetch((u8 *)p + L1_CACHE_BYTES);
2487 #endif
2488 }
2489 
2490 static inline void net_prefetchw(void *p)
2491 {
2492 	prefetchw(p);
2493 #if L1_CACHE_BYTES < 128
2494 	prefetchw((u8 *)p + L1_CACHE_BYTES);
2495 #endif
2496 }
2497 
2498 void netdev_unbind_sb_channel(struct net_device *dev,
2499 			      struct net_device *sb_dev);
2500 int netdev_bind_sb_channel_queue(struct net_device *dev,
2501 				 struct net_device *sb_dev,
2502 				 u8 tc, u16 count, u16 offset);
2503 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2504 static inline int netdev_get_sb_channel(struct net_device *dev)
2505 {
2506 	return max_t(int, -dev->num_tc, 0);
2507 }
2508 
2509 static inline
2510 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2511 					 unsigned int index)
2512 {
2513 	DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2514 	return &dev->_tx[index];
2515 }
2516 
2517 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2518 						    const struct sk_buff *skb)
2519 {
2520 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2521 }
2522 
2523 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2524 					    void (*f)(struct net_device *,
2525 						      struct netdev_queue *,
2526 						      void *),
2527 					    void *arg)
2528 {
2529 	unsigned int i;
2530 
2531 	for (i = 0; i < dev->num_tx_queues; i++)
2532 		f(dev, &dev->_tx[i], arg);
2533 }
2534 
2535 #define netdev_lockdep_set_classes(dev)				\
2536 {								\
2537 	static struct lock_class_key qdisc_tx_busylock_key;	\
2538 	static struct lock_class_key qdisc_xmit_lock_key;	\
2539 	static struct lock_class_key dev_addr_list_lock_key;	\
2540 	unsigned int i;						\
2541 								\
2542 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
2543 	lockdep_set_class(&(dev)->addr_list_lock,		\
2544 			  &dev_addr_list_lock_key);		\
2545 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
2546 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
2547 				  &qdisc_xmit_lock_key);	\
2548 }
2549 
2550 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2551 		     struct net_device *sb_dev);
2552 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2553 					 struct sk_buff *skb,
2554 					 struct net_device *sb_dev);
2555 
2556 /* returns the headroom that the master device needs to take in account
2557  * when forwarding to this dev
2558  */
2559 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2560 {
2561 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2562 }
2563 
2564 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2565 {
2566 	if (dev->netdev_ops->ndo_set_rx_headroom)
2567 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2568 }
2569 
2570 /* set the device rx headroom to the dev's default */
2571 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2572 {
2573 	netdev_set_rx_headroom(dev, -1);
2574 }
2575 
2576 static inline void *netdev_get_ml_priv(struct net_device *dev,
2577 				       enum netdev_ml_priv_type type)
2578 {
2579 	if (dev->ml_priv_type != type)
2580 		return NULL;
2581 
2582 	return dev->ml_priv;
2583 }
2584 
2585 static inline void netdev_set_ml_priv(struct net_device *dev,
2586 				      void *ml_priv,
2587 				      enum netdev_ml_priv_type type)
2588 {
2589 	WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2590 	     "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2591 	     dev->ml_priv_type, type);
2592 	WARN(!dev->ml_priv_type && dev->ml_priv,
2593 	     "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2594 
2595 	dev->ml_priv = ml_priv;
2596 	dev->ml_priv_type = type;
2597 }
2598 
2599 /*
2600  * Net namespace inlines
2601  */
2602 static inline
2603 struct net *dev_net(const struct net_device *dev)
2604 {
2605 	return read_pnet(&dev->nd_net);
2606 }
2607 
2608 static inline
2609 void dev_net_set(struct net_device *dev, struct net *net)
2610 {
2611 	write_pnet(&dev->nd_net, net);
2612 }
2613 
2614 /**
2615  *	netdev_priv - access network device private data
2616  *	@dev: network device
2617  *
2618  * Get network device private data
2619  */
2620 static inline void *netdev_priv(const struct net_device *dev)
2621 {
2622 	return (void *)dev->priv;
2623 }
2624 
2625 /* Set the sysfs physical device reference for the network logical device
2626  * if set prior to registration will cause a symlink during initialization.
2627  */
2628 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2629 
2630 /* Set the sysfs device type for the network logical device to allow
2631  * fine-grained identification of different network device types. For
2632  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2633  */
2634 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2635 
2636 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2637 			  enum netdev_queue_type type,
2638 			  struct napi_struct *napi);
2639 
2640 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2641 {
2642 	napi->irq = irq;
2643 }
2644 
2645 /* Default NAPI poll() weight
2646  * Device drivers are strongly advised to not use bigger value
2647  */
2648 #define NAPI_POLL_WEIGHT 64
2649 
2650 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2651 			   int (*poll)(struct napi_struct *, int), int weight);
2652 
2653 /**
2654  * netif_napi_add() - initialize a NAPI context
2655  * @dev:  network device
2656  * @napi: NAPI context
2657  * @poll: polling function
2658  *
2659  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2660  * *any* of the other NAPI-related functions.
2661  */
2662 static inline void
2663 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2664 	       int (*poll)(struct napi_struct *, int))
2665 {
2666 	netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2667 }
2668 
2669 static inline void
2670 netif_napi_add_tx_weight(struct net_device *dev,
2671 			 struct napi_struct *napi,
2672 			 int (*poll)(struct napi_struct *, int),
2673 			 int weight)
2674 {
2675 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2676 	netif_napi_add_weight(dev, napi, poll, weight);
2677 }
2678 
2679 /**
2680  * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2681  * @dev:  network device
2682  * @napi: NAPI context
2683  * @poll: polling function
2684  *
2685  * This variant of netif_napi_add() should be used from drivers using NAPI
2686  * to exclusively poll a TX queue.
2687  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2688  */
2689 static inline void netif_napi_add_tx(struct net_device *dev,
2690 				     struct napi_struct *napi,
2691 				     int (*poll)(struct napi_struct *, int))
2692 {
2693 	netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2694 }
2695 
2696 /**
2697  *  __netif_napi_del - remove a NAPI context
2698  *  @napi: NAPI context
2699  *
2700  * Warning: caller must observe RCU grace period before freeing memory
2701  * containing @napi. Drivers might want to call this helper to combine
2702  * all the needed RCU grace periods into a single one.
2703  */
2704 void __netif_napi_del(struct napi_struct *napi);
2705 
2706 /**
2707  *  netif_napi_del - remove a NAPI context
2708  *  @napi: NAPI context
2709  *
2710  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2711  */
2712 static inline void netif_napi_del(struct napi_struct *napi)
2713 {
2714 	__netif_napi_del(napi);
2715 	synchronize_net();
2716 }
2717 
2718 struct packet_type {
2719 	__be16			type;	/* This is really htons(ether_type). */
2720 	bool			ignore_outgoing;
2721 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2722 	netdevice_tracker	dev_tracker;
2723 	int			(*func) (struct sk_buff *,
2724 					 struct net_device *,
2725 					 struct packet_type *,
2726 					 struct net_device *);
2727 	void			(*list_func) (struct list_head *,
2728 					      struct packet_type *,
2729 					      struct net_device *);
2730 	bool			(*id_match)(struct packet_type *ptype,
2731 					    struct sock *sk);
2732 	struct net		*af_packet_net;
2733 	void			*af_packet_priv;
2734 	struct list_head	list;
2735 };
2736 
2737 struct offload_callbacks {
2738 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2739 						netdev_features_t features);
2740 	struct sk_buff		*(*gro_receive)(struct list_head *head,
2741 						struct sk_buff *skb);
2742 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2743 };
2744 
2745 struct packet_offload {
2746 	__be16			 type;	/* This is really htons(ether_type). */
2747 	u16			 priority;
2748 	struct offload_callbacks callbacks;
2749 	struct list_head	 list;
2750 };
2751 
2752 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2753 struct pcpu_sw_netstats {
2754 	u64_stats_t		rx_packets;
2755 	u64_stats_t		rx_bytes;
2756 	u64_stats_t		tx_packets;
2757 	u64_stats_t		tx_bytes;
2758 	struct u64_stats_sync   syncp;
2759 } __aligned(4 * sizeof(u64));
2760 
2761 struct pcpu_dstats {
2762 	u64_stats_t		rx_packets;
2763 	u64_stats_t		rx_bytes;
2764 	u64_stats_t		rx_drops;
2765 	u64_stats_t		tx_packets;
2766 	u64_stats_t		tx_bytes;
2767 	u64_stats_t		tx_drops;
2768 	struct u64_stats_sync	syncp;
2769 } __aligned(8 * sizeof(u64));
2770 
2771 struct pcpu_lstats {
2772 	u64_stats_t packets;
2773 	u64_stats_t bytes;
2774 	struct u64_stats_sync syncp;
2775 } __aligned(2 * sizeof(u64));
2776 
2777 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2778 
2779 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2780 {
2781 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2782 
2783 	u64_stats_update_begin(&tstats->syncp);
2784 	u64_stats_add(&tstats->rx_bytes, len);
2785 	u64_stats_inc(&tstats->rx_packets);
2786 	u64_stats_update_end(&tstats->syncp);
2787 }
2788 
2789 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2790 					  unsigned int packets,
2791 					  unsigned int len)
2792 {
2793 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2794 
2795 	u64_stats_update_begin(&tstats->syncp);
2796 	u64_stats_add(&tstats->tx_bytes, len);
2797 	u64_stats_add(&tstats->tx_packets, packets);
2798 	u64_stats_update_end(&tstats->syncp);
2799 }
2800 
2801 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2802 {
2803 	struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2804 
2805 	u64_stats_update_begin(&lstats->syncp);
2806 	u64_stats_add(&lstats->bytes, len);
2807 	u64_stats_inc(&lstats->packets);
2808 	u64_stats_update_end(&lstats->syncp);
2809 }
2810 
2811 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2812 ({									\
2813 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2814 	if (pcpu_stats)	{						\
2815 		int __cpu;						\
2816 		for_each_possible_cpu(__cpu) {				\
2817 			typeof(type) *stat;				\
2818 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2819 			u64_stats_init(&stat->syncp);			\
2820 		}							\
2821 	}								\
2822 	pcpu_stats;							\
2823 })
2824 
2825 #define netdev_alloc_pcpu_stats(type)					\
2826 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2827 
2828 #define devm_netdev_alloc_pcpu_stats(dev, type)				\
2829 ({									\
2830 	typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2831 	if (pcpu_stats) {						\
2832 		int __cpu;						\
2833 		for_each_possible_cpu(__cpu) {				\
2834 			typeof(type) *stat;				\
2835 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2836 			u64_stats_init(&stat->syncp);			\
2837 		}							\
2838 	}								\
2839 	pcpu_stats;							\
2840 })
2841 
2842 enum netdev_lag_tx_type {
2843 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2844 	NETDEV_LAG_TX_TYPE_RANDOM,
2845 	NETDEV_LAG_TX_TYPE_BROADCAST,
2846 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2847 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2848 	NETDEV_LAG_TX_TYPE_HASH,
2849 };
2850 
2851 enum netdev_lag_hash {
2852 	NETDEV_LAG_HASH_NONE,
2853 	NETDEV_LAG_HASH_L2,
2854 	NETDEV_LAG_HASH_L34,
2855 	NETDEV_LAG_HASH_L23,
2856 	NETDEV_LAG_HASH_E23,
2857 	NETDEV_LAG_HASH_E34,
2858 	NETDEV_LAG_HASH_VLAN_SRCMAC,
2859 	NETDEV_LAG_HASH_UNKNOWN,
2860 };
2861 
2862 struct netdev_lag_upper_info {
2863 	enum netdev_lag_tx_type tx_type;
2864 	enum netdev_lag_hash hash_type;
2865 };
2866 
2867 struct netdev_lag_lower_state_info {
2868 	u8 link_up : 1,
2869 	   tx_enabled : 1;
2870 };
2871 
2872 #include <linux/notifier.h>
2873 
2874 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2875  * and the rtnetlink notification exclusion list in rtnetlink_event() when
2876  * adding new types.
2877  */
2878 enum netdev_cmd {
2879 	NETDEV_UP	= 1,	/* For now you can't veto a device up/down */
2880 	NETDEV_DOWN,
2881 	NETDEV_REBOOT,		/* Tell a protocol stack a network interface
2882 				   detected a hardware crash and restarted
2883 				   - we can use this eg to kick tcp sessions
2884 				   once done */
2885 	NETDEV_CHANGE,		/* Notify device state change */
2886 	NETDEV_REGISTER,
2887 	NETDEV_UNREGISTER,
2888 	NETDEV_CHANGEMTU,	/* notify after mtu change happened */
2889 	NETDEV_CHANGEADDR,	/* notify after the address change */
2890 	NETDEV_PRE_CHANGEADDR,	/* notify before the address change */
2891 	NETDEV_GOING_DOWN,
2892 	NETDEV_CHANGENAME,
2893 	NETDEV_FEAT_CHANGE,
2894 	NETDEV_BONDING_FAILOVER,
2895 	NETDEV_PRE_UP,
2896 	NETDEV_PRE_TYPE_CHANGE,
2897 	NETDEV_POST_TYPE_CHANGE,
2898 	NETDEV_POST_INIT,
2899 	NETDEV_PRE_UNINIT,
2900 	NETDEV_RELEASE,
2901 	NETDEV_NOTIFY_PEERS,
2902 	NETDEV_JOIN,
2903 	NETDEV_CHANGEUPPER,
2904 	NETDEV_RESEND_IGMP,
2905 	NETDEV_PRECHANGEMTU,	/* notify before mtu change happened */
2906 	NETDEV_CHANGEINFODATA,
2907 	NETDEV_BONDING_INFO,
2908 	NETDEV_PRECHANGEUPPER,
2909 	NETDEV_CHANGELOWERSTATE,
2910 	NETDEV_UDP_TUNNEL_PUSH_INFO,
2911 	NETDEV_UDP_TUNNEL_DROP_INFO,
2912 	NETDEV_CHANGE_TX_QUEUE_LEN,
2913 	NETDEV_CVLAN_FILTER_PUSH_INFO,
2914 	NETDEV_CVLAN_FILTER_DROP_INFO,
2915 	NETDEV_SVLAN_FILTER_PUSH_INFO,
2916 	NETDEV_SVLAN_FILTER_DROP_INFO,
2917 	NETDEV_OFFLOAD_XSTATS_ENABLE,
2918 	NETDEV_OFFLOAD_XSTATS_DISABLE,
2919 	NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2920 	NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2921 	NETDEV_XDP_FEAT_CHANGE,
2922 };
2923 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2924 
2925 int register_netdevice_notifier(struct notifier_block *nb);
2926 int unregister_netdevice_notifier(struct notifier_block *nb);
2927 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2928 int unregister_netdevice_notifier_net(struct net *net,
2929 				      struct notifier_block *nb);
2930 int register_netdevice_notifier_dev_net(struct net_device *dev,
2931 					struct notifier_block *nb,
2932 					struct netdev_net_notifier *nn);
2933 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2934 					  struct notifier_block *nb,
2935 					  struct netdev_net_notifier *nn);
2936 
2937 struct netdev_notifier_info {
2938 	struct net_device	*dev;
2939 	struct netlink_ext_ack	*extack;
2940 };
2941 
2942 struct netdev_notifier_info_ext {
2943 	struct netdev_notifier_info info; /* must be first */
2944 	union {
2945 		u32 mtu;
2946 	} ext;
2947 };
2948 
2949 struct netdev_notifier_change_info {
2950 	struct netdev_notifier_info info; /* must be first */
2951 	unsigned int flags_changed;
2952 };
2953 
2954 struct netdev_notifier_changeupper_info {
2955 	struct netdev_notifier_info info; /* must be first */
2956 	struct net_device *upper_dev; /* new upper dev */
2957 	bool master; /* is upper dev master */
2958 	bool linking; /* is the notification for link or unlink */
2959 	void *upper_info; /* upper dev info */
2960 };
2961 
2962 struct netdev_notifier_changelowerstate_info {
2963 	struct netdev_notifier_info info; /* must be first */
2964 	void *lower_state_info; /* is lower dev state */
2965 };
2966 
2967 struct netdev_notifier_pre_changeaddr_info {
2968 	struct netdev_notifier_info info; /* must be first */
2969 	const unsigned char *dev_addr;
2970 };
2971 
2972 enum netdev_offload_xstats_type {
2973 	NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2974 };
2975 
2976 struct netdev_notifier_offload_xstats_info {
2977 	struct netdev_notifier_info info; /* must be first */
2978 	enum netdev_offload_xstats_type type;
2979 
2980 	union {
2981 		/* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2982 		struct netdev_notifier_offload_xstats_rd *report_delta;
2983 		/* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2984 		struct netdev_notifier_offload_xstats_ru *report_used;
2985 	};
2986 };
2987 
2988 int netdev_offload_xstats_enable(struct net_device *dev,
2989 				 enum netdev_offload_xstats_type type,
2990 				 struct netlink_ext_ack *extack);
2991 int netdev_offload_xstats_disable(struct net_device *dev,
2992 				  enum netdev_offload_xstats_type type);
2993 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2994 				   enum netdev_offload_xstats_type type);
2995 int netdev_offload_xstats_get(struct net_device *dev,
2996 			      enum netdev_offload_xstats_type type,
2997 			      struct rtnl_hw_stats64 *stats, bool *used,
2998 			      struct netlink_ext_ack *extack);
2999 void
3000 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3001 				   const struct rtnl_hw_stats64 *stats);
3002 void
3003 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3004 void netdev_offload_xstats_push_delta(struct net_device *dev,
3005 				      enum netdev_offload_xstats_type type,
3006 				      const struct rtnl_hw_stats64 *stats);
3007 
3008 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3009 					     struct net_device *dev)
3010 {
3011 	info->dev = dev;
3012 	info->extack = NULL;
3013 }
3014 
3015 static inline struct net_device *
3016 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3017 {
3018 	return info->dev;
3019 }
3020 
3021 static inline struct netlink_ext_ack *
3022 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3023 {
3024 	return info->extack;
3025 }
3026 
3027 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3028 int call_netdevice_notifiers_info(unsigned long val,
3029 				  struct netdev_notifier_info *info);
3030 
3031 #define for_each_netdev(net, d)		\
3032 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3033 #define for_each_netdev_reverse(net, d)	\
3034 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3035 #define for_each_netdev_rcu(net, d)		\
3036 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3037 #define for_each_netdev_safe(net, d, n)	\
3038 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3039 #define for_each_netdev_continue(net, d)		\
3040 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3041 #define for_each_netdev_continue_reverse(net, d)		\
3042 		list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3043 						     dev_list)
3044 #define for_each_netdev_continue_rcu(net, d)		\
3045 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3046 #define for_each_netdev_in_bond_rcu(bond, slave)	\
3047 		for_each_netdev_rcu(&init_net, slave)	\
3048 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3049 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
3050 
3051 #define for_each_netdev_dump(net, d, ifindex)				\
3052 	for (; (d = xa_find(&(net)->dev_by_index, &ifindex,		\
3053 			    ULONG_MAX, XA_PRESENT)); ifindex++)
3054 
3055 static inline struct net_device *next_net_device(struct net_device *dev)
3056 {
3057 	struct list_head *lh;
3058 	struct net *net;
3059 
3060 	net = dev_net(dev);
3061 	lh = dev->dev_list.next;
3062 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3063 }
3064 
3065 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3066 {
3067 	struct list_head *lh;
3068 	struct net *net;
3069 
3070 	net = dev_net(dev);
3071 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3072 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3073 }
3074 
3075 static inline struct net_device *first_net_device(struct net *net)
3076 {
3077 	return list_empty(&net->dev_base_head) ? NULL :
3078 		net_device_entry(net->dev_base_head.next);
3079 }
3080 
3081 static inline struct net_device *first_net_device_rcu(struct net *net)
3082 {
3083 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3084 
3085 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3086 }
3087 
3088 int netdev_boot_setup_check(struct net_device *dev);
3089 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3090 				       const char *hwaddr);
3091 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3092 void dev_add_pack(struct packet_type *pt);
3093 void dev_remove_pack(struct packet_type *pt);
3094 void __dev_remove_pack(struct packet_type *pt);
3095 void dev_add_offload(struct packet_offload *po);
3096 void dev_remove_offload(struct packet_offload *po);
3097 
3098 int dev_get_iflink(const struct net_device *dev);
3099 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3100 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3101 			  struct net_device_path_stack *stack);
3102 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3103 				      unsigned short mask);
3104 struct net_device *dev_get_by_name(struct net *net, const char *name);
3105 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3106 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3107 bool netdev_name_in_use(struct net *net, const char *name);
3108 int dev_alloc_name(struct net_device *dev, const char *name);
3109 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3110 void dev_close(struct net_device *dev);
3111 void dev_close_many(struct list_head *head, bool unlink);
3112 void dev_disable_lro(struct net_device *dev);
3113 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3114 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3115 		     struct net_device *sb_dev);
3116 
3117 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3118 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3119 
3120 static inline int dev_queue_xmit(struct sk_buff *skb)
3121 {
3122 	return __dev_queue_xmit(skb, NULL);
3123 }
3124 
3125 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3126 				       struct net_device *sb_dev)
3127 {
3128 	return __dev_queue_xmit(skb, sb_dev);
3129 }
3130 
3131 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3132 {
3133 	int ret;
3134 
3135 	ret = __dev_direct_xmit(skb, queue_id);
3136 	if (!dev_xmit_complete(ret))
3137 		kfree_skb(skb);
3138 	return ret;
3139 }
3140 
3141 int register_netdevice(struct net_device *dev);
3142 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3143 void unregister_netdevice_many(struct list_head *head);
3144 static inline void unregister_netdevice(struct net_device *dev)
3145 {
3146 	unregister_netdevice_queue(dev, NULL);
3147 }
3148 
3149 int netdev_refcnt_read(const struct net_device *dev);
3150 void free_netdev(struct net_device *dev);
3151 void init_dummy_netdev(struct net_device *dev);
3152 
3153 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3154 					 struct sk_buff *skb,
3155 					 bool all_slaves);
3156 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3157 					    struct sock *sk);
3158 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3159 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3160 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3161 				       netdevice_tracker *tracker, gfp_t gfp);
3162 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3163 				      netdevice_tracker *tracker, gfp_t gfp);
3164 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3165 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3166 void netdev_copy_name(struct net_device *dev, char *name);
3167 
3168 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3169 				  unsigned short type,
3170 				  const void *daddr, const void *saddr,
3171 				  unsigned int len)
3172 {
3173 	if (!dev->header_ops || !dev->header_ops->create)
3174 		return 0;
3175 
3176 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3177 }
3178 
3179 static inline int dev_parse_header(const struct sk_buff *skb,
3180 				   unsigned char *haddr)
3181 {
3182 	const struct net_device *dev = skb->dev;
3183 
3184 	if (!dev->header_ops || !dev->header_ops->parse)
3185 		return 0;
3186 	return dev->header_ops->parse(skb, haddr);
3187 }
3188 
3189 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3190 {
3191 	const struct net_device *dev = skb->dev;
3192 
3193 	if (!dev->header_ops || !dev->header_ops->parse_protocol)
3194 		return 0;
3195 	return dev->header_ops->parse_protocol(skb);
3196 }
3197 
3198 /* ll_header must have at least hard_header_len allocated */
3199 static inline bool dev_validate_header(const struct net_device *dev,
3200 				       char *ll_header, int len)
3201 {
3202 	if (likely(len >= dev->hard_header_len))
3203 		return true;
3204 	if (len < dev->min_header_len)
3205 		return false;
3206 
3207 	if (capable(CAP_SYS_RAWIO)) {
3208 		memset(ll_header + len, 0, dev->hard_header_len - len);
3209 		return true;
3210 	}
3211 
3212 	if (dev->header_ops && dev->header_ops->validate)
3213 		return dev->header_ops->validate(ll_header, len);
3214 
3215 	return false;
3216 }
3217 
3218 static inline bool dev_has_header(const struct net_device *dev)
3219 {
3220 	return dev->header_ops && dev->header_ops->create;
3221 }
3222 
3223 /*
3224  * Incoming packets are placed on per-CPU queues
3225  */
3226 struct softnet_data {
3227 	struct list_head	poll_list;
3228 	struct sk_buff_head	process_queue;
3229 	local_lock_t		process_queue_bh_lock;
3230 
3231 	/* stats */
3232 	unsigned int		processed;
3233 	unsigned int		time_squeeze;
3234 #ifdef CONFIG_RPS
3235 	struct softnet_data	*rps_ipi_list;
3236 #endif
3237 
3238 	unsigned int		received_rps;
3239 	bool			in_net_rx_action;
3240 	bool			in_napi_threaded_poll;
3241 
3242 #ifdef CONFIG_NET_FLOW_LIMIT
3243 	struct sd_flow_limit __rcu *flow_limit;
3244 #endif
3245 	struct Qdisc		*output_queue;
3246 	struct Qdisc		**output_queue_tailp;
3247 	struct sk_buff		*completion_queue;
3248 #ifdef CONFIG_XFRM_OFFLOAD
3249 	struct sk_buff_head	xfrm_backlog;
3250 #endif
3251 	/* written and read only by owning cpu: */
3252 	struct netdev_xmit xmit;
3253 #ifdef CONFIG_RPS
3254 	/* input_queue_head should be written by cpu owning this struct,
3255 	 * and only read by other cpus. Worth using a cache line.
3256 	 */
3257 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
3258 
3259 	/* Elements below can be accessed between CPUs for RPS/RFS */
3260 	call_single_data_t	csd ____cacheline_aligned_in_smp;
3261 	struct softnet_data	*rps_ipi_next;
3262 	unsigned int		cpu;
3263 	unsigned int		input_queue_tail;
3264 #endif
3265 	struct sk_buff_head	input_pkt_queue;
3266 	struct napi_struct	backlog;
3267 
3268 	atomic_t		dropped ____cacheline_aligned_in_smp;
3269 
3270 	/* Another possibly contended cache line */
3271 	spinlock_t		defer_lock ____cacheline_aligned_in_smp;
3272 	int			defer_count;
3273 	int			defer_ipi_scheduled;
3274 	struct sk_buff		*defer_list;
3275 	call_single_data_t	defer_csd;
3276 };
3277 
3278 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3279 
3280 #ifndef CONFIG_PREEMPT_RT
3281 static inline int dev_recursion_level(void)
3282 {
3283 	return this_cpu_read(softnet_data.xmit.recursion);
3284 }
3285 #else
3286 static inline int dev_recursion_level(void)
3287 {
3288 	return current->net_xmit.recursion;
3289 }
3290 
3291 #endif
3292 
3293 void __netif_schedule(struct Qdisc *q);
3294 void netif_schedule_queue(struct netdev_queue *txq);
3295 
3296 static inline void netif_tx_schedule_all(struct net_device *dev)
3297 {
3298 	unsigned int i;
3299 
3300 	for (i = 0; i < dev->num_tx_queues; i++)
3301 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
3302 }
3303 
3304 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3305 {
3306 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3307 }
3308 
3309 /**
3310  *	netif_start_queue - allow transmit
3311  *	@dev: network device
3312  *
3313  *	Allow upper layers to call the device hard_start_xmit routine.
3314  */
3315 static inline void netif_start_queue(struct net_device *dev)
3316 {
3317 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3318 }
3319 
3320 static inline void netif_tx_start_all_queues(struct net_device *dev)
3321 {
3322 	unsigned int i;
3323 
3324 	for (i = 0; i < dev->num_tx_queues; i++) {
3325 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3326 		netif_tx_start_queue(txq);
3327 	}
3328 }
3329 
3330 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3331 
3332 /**
3333  *	netif_wake_queue - restart transmit
3334  *	@dev: network device
3335  *
3336  *	Allow upper layers to call the device hard_start_xmit routine.
3337  *	Used for flow control when transmit resources are available.
3338  */
3339 static inline void netif_wake_queue(struct net_device *dev)
3340 {
3341 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3342 }
3343 
3344 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3345 {
3346 	unsigned int i;
3347 
3348 	for (i = 0; i < dev->num_tx_queues; i++) {
3349 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3350 		netif_tx_wake_queue(txq);
3351 	}
3352 }
3353 
3354 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3355 {
3356 	/* Must be an atomic op see netif_txq_try_stop() */
3357 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3358 }
3359 
3360 /**
3361  *	netif_stop_queue - stop transmitted packets
3362  *	@dev: network device
3363  *
3364  *	Stop upper layers calling the device hard_start_xmit routine.
3365  *	Used for flow control when transmit resources are unavailable.
3366  */
3367 static inline void netif_stop_queue(struct net_device *dev)
3368 {
3369 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3370 }
3371 
3372 void netif_tx_stop_all_queues(struct net_device *dev);
3373 
3374 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3375 {
3376 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3377 }
3378 
3379 /**
3380  *	netif_queue_stopped - test if transmit queue is flowblocked
3381  *	@dev: network device
3382  *
3383  *	Test if transmit queue on device is currently unable to send.
3384  */
3385 static inline bool netif_queue_stopped(const struct net_device *dev)
3386 {
3387 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3388 }
3389 
3390 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3391 {
3392 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3393 }
3394 
3395 static inline bool
3396 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3397 {
3398 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3399 }
3400 
3401 static inline bool
3402 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3403 {
3404 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3405 }
3406 
3407 /**
3408  *	netdev_queue_set_dql_min_limit - set dql minimum limit
3409  *	@dev_queue: pointer to transmit queue
3410  *	@min_limit: dql minimum limit
3411  *
3412  * Forces xmit_more() to return true until the minimum threshold
3413  * defined by @min_limit is reached (or until the tx queue is
3414  * empty). Warning: to be use with care, misuse will impact the
3415  * latency.
3416  */
3417 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3418 						  unsigned int min_limit)
3419 {
3420 #ifdef CONFIG_BQL
3421 	dev_queue->dql.min_limit = min_limit;
3422 #endif
3423 }
3424 
3425 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq)
3426 {
3427 #ifdef CONFIG_BQL
3428 	/* Non-BQL migrated drivers will return 0, too. */
3429 	return dql_avail(&txq->dql);
3430 #else
3431 	return 0;
3432 #endif
3433 }
3434 
3435 /**
3436  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3437  *	@dev_queue: pointer to transmit queue
3438  *
3439  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3440  * to give appropriate hint to the CPU.
3441  */
3442 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3443 {
3444 #ifdef CONFIG_BQL
3445 	prefetchw(&dev_queue->dql.num_queued);
3446 #endif
3447 }
3448 
3449 /**
3450  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3451  *	@dev_queue: pointer to transmit queue
3452  *
3453  * BQL enabled drivers might use this helper in their TX completion path,
3454  * to give appropriate hint to the CPU.
3455  */
3456 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3457 {
3458 #ifdef CONFIG_BQL
3459 	prefetchw(&dev_queue->dql.limit);
3460 #endif
3461 }
3462 
3463 /**
3464  *	netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3465  *	@dev_queue: network device queue
3466  *	@bytes: number of bytes queued to the device queue
3467  *
3468  *	Report the number of bytes queued for sending/completion to the network
3469  *	device hardware queue. @bytes should be a good approximation and should
3470  *	exactly match netdev_completed_queue() @bytes.
3471  *	This is typically called once per packet, from ndo_start_xmit().
3472  */
3473 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3474 					unsigned int bytes)
3475 {
3476 #ifdef CONFIG_BQL
3477 	dql_queued(&dev_queue->dql, bytes);
3478 
3479 	if (likely(dql_avail(&dev_queue->dql) >= 0))
3480 		return;
3481 
3482 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3483 
3484 	/*
3485 	 * The XOFF flag must be set before checking the dql_avail below,
3486 	 * because in netdev_tx_completed_queue we update the dql_completed
3487 	 * before checking the XOFF flag.
3488 	 */
3489 	smp_mb();
3490 
3491 	/* check again in case another CPU has just made room avail */
3492 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3493 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3494 #endif
3495 }
3496 
3497 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3498  * that they should not test BQL status themselves.
3499  * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3500  * skb of a batch.
3501  * Returns true if the doorbell must be used to kick the NIC.
3502  */
3503 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3504 					  unsigned int bytes,
3505 					  bool xmit_more)
3506 {
3507 	if (xmit_more) {
3508 #ifdef CONFIG_BQL
3509 		dql_queued(&dev_queue->dql, bytes);
3510 #endif
3511 		return netif_tx_queue_stopped(dev_queue);
3512 	}
3513 	netdev_tx_sent_queue(dev_queue, bytes);
3514 	return true;
3515 }
3516 
3517 /**
3518  *	netdev_sent_queue - report the number of bytes queued to hardware
3519  *	@dev: network device
3520  *	@bytes: number of bytes queued to the hardware device queue
3521  *
3522  *	Report the number of bytes queued for sending/completion to the network
3523  *	device hardware queue#0. @bytes should be a good approximation and should
3524  *	exactly match netdev_completed_queue() @bytes.
3525  *	This is typically called once per packet, from ndo_start_xmit().
3526  */
3527 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3528 {
3529 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3530 }
3531 
3532 static inline bool __netdev_sent_queue(struct net_device *dev,
3533 				       unsigned int bytes,
3534 				       bool xmit_more)
3535 {
3536 	return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3537 				      xmit_more);
3538 }
3539 
3540 /**
3541  *	netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3542  *	@dev_queue: network device queue
3543  *	@pkts: number of packets (currently ignored)
3544  *	@bytes: number of bytes dequeued from the device queue
3545  *
3546  *	Must be called at most once per TX completion round (and not per
3547  *	individual packet), so that BQL can adjust its limits appropriately.
3548  */
3549 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3550 					     unsigned int pkts, unsigned int bytes)
3551 {
3552 #ifdef CONFIG_BQL
3553 	if (unlikely(!bytes))
3554 		return;
3555 
3556 	dql_completed(&dev_queue->dql, bytes);
3557 
3558 	/*
3559 	 * Without the memory barrier there is a small possibility that
3560 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3561 	 * be stopped forever
3562 	 */
3563 	smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3564 
3565 	if (unlikely(dql_avail(&dev_queue->dql) < 0))
3566 		return;
3567 
3568 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3569 		netif_schedule_queue(dev_queue);
3570 #endif
3571 }
3572 
3573 /**
3574  * 	netdev_completed_queue - report bytes and packets completed by device
3575  * 	@dev: network device
3576  * 	@pkts: actual number of packets sent over the medium
3577  * 	@bytes: actual number of bytes sent over the medium
3578  *
3579  * 	Report the number of bytes and packets transmitted by the network device
3580  * 	hardware queue over the physical medium, @bytes must exactly match the
3581  * 	@bytes amount passed to netdev_sent_queue()
3582  */
3583 static inline void netdev_completed_queue(struct net_device *dev,
3584 					  unsigned int pkts, unsigned int bytes)
3585 {
3586 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3587 }
3588 
3589 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3590 {
3591 #ifdef CONFIG_BQL
3592 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3593 	dql_reset(&q->dql);
3594 #endif
3595 }
3596 
3597 /**
3598  * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue
3599  * @dev: network device
3600  * @qid: stack index of the queue to reset
3601  */
3602 static inline void netdev_tx_reset_subqueue(const struct net_device *dev,
3603 					    u32 qid)
3604 {
3605 	netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid));
3606 }
3607 
3608 /**
3609  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3610  * 	@dev_queue: network device
3611  *
3612  * 	Reset the bytes and packet count of a network device and clear the
3613  * 	software flow control OFF bit for this network device
3614  */
3615 static inline void netdev_reset_queue(struct net_device *dev_queue)
3616 {
3617 	netdev_tx_reset_subqueue(dev_queue, 0);
3618 }
3619 
3620 /**
3621  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3622  * 	@dev: network device
3623  * 	@queue_index: given tx queue index
3624  *
3625  * 	Returns 0 if given tx queue index >= number of device tx queues,
3626  * 	otherwise returns the originally passed tx queue index.
3627  */
3628 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3629 {
3630 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3631 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3632 				     dev->name, queue_index,
3633 				     dev->real_num_tx_queues);
3634 		return 0;
3635 	}
3636 
3637 	return queue_index;
3638 }
3639 
3640 /**
3641  *	netif_running - test if up
3642  *	@dev: network device
3643  *
3644  *	Test if the device has been brought up.
3645  */
3646 static inline bool netif_running(const struct net_device *dev)
3647 {
3648 	return test_bit(__LINK_STATE_START, &dev->state);
3649 }
3650 
3651 /*
3652  * Routines to manage the subqueues on a device.  We only need start,
3653  * stop, and a check if it's stopped.  All other device management is
3654  * done at the overall netdevice level.
3655  * Also test the device if we're multiqueue.
3656  */
3657 
3658 /**
3659  *	netif_start_subqueue - allow sending packets on subqueue
3660  *	@dev: network device
3661  *	@queue_index: sub queue index
3662  *
3663  * Start individual transmit queue of a device with multiple transmit queues.
3664  */
3665 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3666 {
3667 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3668 
3669 	netif_tx_start_queue(txq);
3670 }
3671 
3672 /**
3673  *	netif_stop_subqueue - stop sending packets on subqueue
3674  *	@dev: network device
3675  *	@queue_index: sub queue index
3676  *
3677  * Stop individual transmit queue of a device with multiple transmit queues.
3678  */
3679 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3680 {
3681 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3682 	netif_tx_stop_queue(txq);
3683 }
3684 
3685 /**
3686  *	__netif_subqueue_stopped - test status of subqueue
3687  *	@dev: network device
3688  *	@queue_index: sub queue index
3689  *
3690  * Check individual transmit queue of a device with multiple transmit queues.
3691  */
3692 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3693 					    u16 queue_index)
3694 {
3695 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3696 
3697 	return netif_tx_queue_stopped(txq);
3698 }
3699 
3700 /**
3701  *	netif_subqueue_stopped - test status of subqueue
3702  *	@dev: network device
3703  *	@skb: sub queue buffer pointer
3704  *
3705  * Check individual transmit queue of a device with multiple transmit queues.
3706  */
3707 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3708 					  struct sk_buff *skb)
3709 {
3710 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3711 }
3712 
3713 /**
3714  *	netif_wake_subqueue - allow sending packets on subqueue
3715  *	@dev: network device
3716  *	@queue_index: sub queue index
3717  *
3718  * Resume individual transmit queue of a device with multiple transmit queues.
3719  */
3720 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3721 {
3722 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3723 
3724 	netif_tx_wake_queue(txq);
3725 }
3726 
3727 #ifdef CONFIG_XPS
3728 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3729 			u16 index);
3730 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3731 			  u16 index, enum xps_map_type type);
3732 
3733 /**
3734  *	netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3735  *	@j: CPU/Rx queue index
3736  *	@mask: bitmask of all cpus/rx queues
3737  *	@nr_bits: number of bits in the bitmask
3738  *
3739  * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3740  */
3741 static inline bool netif_attr_test_mask(unsigned long j,
3742 					const unsigned long *mask,
3743 					unsigned int nr_bits)
3744 {
3745 	cpu_max_bits_warn(j, nr_bits);
3746 	return test_bit(j, mask);
3747 }
3748 
3749 /**
3750  *	netif_attr_test_online - Test for online CPU/Rx queue
3751  *	@j: CPU/Rx queue index
3752  *	@online_mask: bitmask for CPUs/Rx queues that are online
3753  *	@nr_bits: number of bits in the bitmask
3754  *
3755  * Returns true if a CPU/Rx queue is online.
3756  */
3757 static inline bool netif_attr_test_online(unsigned long j,
3758 					  const unsigned long *online_mask,
3759 					  unsigned int nr_bits)
3760 {
3761 	cpu_max_bits_warn(j, nr_bits);
3762 
3763 	if (online_mask)
3764 		return test_bit(j, online_mask);
3765 
3766 	return (j < nr_bits);
3767 }
3768 
3769 /**
3770  *	netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3771  *	@n: CPU/Rx queue index
3772  *	@srcp: the cpumask/Rx queue mask pointer
3773  *	@nr_bits: number of bits in the bitmask
3774  *
3775  * Returns >= nr_bits if no further CPUs/Rx queues set.
3776  */
3777 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3778 					       unsigned int nr_bits)
3779 {
3780 	/* -1 is a legal arg here. */
3781 	if (n != -1)
3782 		cpu_max_bits_warn(n, nr_bits);
3783 
3784 	if (srcp)
3785 		return find_next_bit(srcp, nr_bits, n + 1);
3786 
3787 	return n + 1;
3788 }
3789 
3790 /**
3791  *	netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3792  *	@n: CPU/Rx queue index
3793  *	@src1p: the first CPUs/Rx queues mask pointer
3794  *	@src2p: the second CPUs/Rx queues mask pointer
3795  *	@nr_bits: number of bits in the bitmask
3796  *
3797  * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3798  */
3799 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3800 					  const unsigned long *src2p,
3801 					  unsigned int nr_bits)
3802 {
3803 	/* -1 is a legal arg here. */
3804 	if (n != -1)
3805 		cpu_max_bits_warn(n, nr_bits);
3806 
3807 	if (src1p && src2p)
3808 		return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3809 	else if (src1p)
3810 		return find_next_bit(src1p, nr_bits, n + 1);
3811 	else if (src2p)
3812 		return find_next_bit(src2p, nr_bits, n + 1);
3813 
3814 	return n + 1;
3815 }
3816 #else
3817 static inline int netif_set_xps_queue(struct net_device *dev,
3818 				      const struct cpumask *mask,
3819 				      u16 index)
3820 {
3821 	return 0;
3822 }
3823 
3824 static inline int __netif_set_xps_queue(struct net_device *dev,
3825 					const unsigned long *mask,
3826 					u16 index, enum xps_map_type type)
3827 {
3828 	return 0;
3829 }
3830 #endif
3831 
3832 /**
3833  *	netif_is_multiqueue - test if device has multiple transmit queues
3834  *	@dev: network device
3835  *
3836  * Check if device has multiple transmit queues
3837  */
3838 static inline bool netif_is_multiqueue(const struct net_device *dev)
3839 {
3840 	return dev->num_tx_queues > 1;
3841 }
3842 
3843 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3844 
3845 #ifdef CONFIG_SYSFS
3846 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3847 #else
3848 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3849 						unsigned int rxqs)
3850 {
3851 	dev->real_num_rx_queues = rxqs;
3852 	return 0;
3853 }
3854 #endif
3855 int netif_set_real_num_queues(struct net_device *dev,
3856 			      unsigned int txq, unsigned int rxq);
3857 
3858 int netif_get_num_default_rss_queues(void);
3859 
3860 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3861 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3862 
3863 /*
3864  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3865  * interrupt context or with hardware interrupts being disabled.
3866  * (in_hardirq() || irqs_disabled())
3867  *
3868  * We provide four helpers that can be used in following contexts :
3869  *
3870  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3871  *  replacing kfree_skb(skb)
3872  *
3873  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3874  *  Typically used in place of consume_skb(skb) in TX completion path
3875  *
3876  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3877  *  replacing kfree_skb(skb)
3878  *
3879  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3880  *  and consumed a packet. Used in place of consume_skb(skb)
3881  */
3882 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3883 {
3884 	dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3885 }
3886 
3887 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3888 {
3889 	dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3890 }
3891 
3892 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3893 {
3894 	dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3895 }
3896 
3897 static inline void dev_consume_skb_any(struct sk_buff *skb)
3898 {
3899 	dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3900 }
3901 
3902 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3903 			     struct bpf_prog *xdp_prog);
3904 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3905 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb);
3906 int netif_rx(struct sk_buff *skb);
3907 int __netif_rx(struct sk_buff *skb);
3908 
3909 int netif_receive_skb(struct sk_buff *skb);
3910 int netif_receive_skb_core(struct sk_buff *skb);
3911 void netif_receive_skb_list_internal(struct list_head *head);
3912 void netif_receive_skb_list(struct list_head *head);
3913 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3914 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3915 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3916 void napi_get_frags_check(struct napi_struct *napi);
3917 gro_result_t napi_gro_frags(struct napi_struct *napi);
3918 
3919 static inline void napi_free_frags(struct napi_struct *napi)
3920 {
3921 	kfree_skb(napi->skb);
3922 	napi->skb = NULL;
3923 }
3924 
3925 bool netdev_is_rx_handler_busy(struct net_device *dev);
3926 int netdev_rx_handler_register(struct net_device *dev,
3927 			       rx_handler_func_t *rx_handler,
3928 			       void *rx_handler_data);
3929 void netdev_rx_handler_unregister(struct net_device *dev);
3930 
3931 bool dev_valid_name(const char *name);
3932 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3933 {
3934 	return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3935 }
3936 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3937 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3938 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3939 		void __user *data, bool *need_copyout);
3940 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3941 int generic_hwtstamp_get_lower(struct net_device *dev,
3942 			       struct kernel_hwtstamp_config *kernel_cfg);
3943 int generic_hwtstamp_set_lower(struct net_device *dev,
3944 			       struct kernel_hwtstamp_config *kernel_cfg,
3945 			       struct netlink_ext_ack *extack);
3946 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3947 unsigned int dev_get_flags(const struct net_device *);
3948 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3949 		       struct netlink_ext_ack *extack);
3950 int dev_change_flags(struct net_device *dev, unsigned int flags,
3951 		     struct netlink_ext_ack *extack);
3952 int dev_set_alias(struct net_device *, const char *, size_t);
3953 int dev_get_alias(const struct net_device *, char *, size_t);
3954 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3955 			       const char *pat, int new_ifindex);
3956 static inline
3957 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3958 			     const char *pat)
3959 {
3960 	return __dev_change_net_namespace(dev, net, pat, 0);
3961 }
3962 int __dev_set_mtu(struct net_device *, int);
3963 int dev_set_mtu(struct net_device *, int);
3964 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3965 			      struct netlink_ext_ack *extack);
3966 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3967 			struct netlink_ext_ack *extack);
3968 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3969 			     struct netlink_ext_ack *extack);
3970 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3971 int dev_get_port_parent_id(struct net_device *dev,
3972 			   struct netdev_phys_item_id *ppid, bool recurse);
3973 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3974 
3975 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3976 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3977 				    struct netdev_queue *txq, int *ret);
3978 
3979 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3980 u8 dev_xdp_prog_count(struct net_device *dev);
3981 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf);
3982 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3983 
3984 u32 dev_get_min_mp_channel_count(const struct net_device *dev);
3985 
3986 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3987 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3988 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3989 bool is_skb_forwardable(const struct net_device *dev,
3990 			const struct sk_buff *skb);
3991 
3992 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3993 						 const struct sk_buff *skb,
3994 						 const bool check_mtu)
3995 {
3996 	const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3997 	unsigned int len;
3998 
3999 	if (!(dev->flags & IFF_UP))
4000 		return false;
4001 
4002 	if (!check_mtu)
4003 		return true;
4004 
4005 	len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4006 	if (skb->len <= len)
4007 		return true;
4008 
4009 	/* if TSO is enabled, we don't care about the length as the packet
4010 	 * could be forwarded without being segmented before
4011 	 */
4012 	if (skb_is_gso(skb))
4013 		return true;
4014 
4015 	return false;
4016 }
4017 
4018 void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4019 
4020 #define DEV_CORE_STATS_INC(FIELD)						\
4021 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev)		\
4022 {										\
4023 	netdev_core_stats_inc(dev,						\
4024 			offsetof(struct net_device_core_stats, FIELD));		\
4025 }
4026 DEV_CORE_STATS_INC(rx_dropped)
4027 DEV_CORE_STATS_INC(tx_dropped)
4028 DEV_CORE_STATS_INC(rx_nohandler)
4029 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4030 #undef DEV_CORE_STATS_INC
4031 
4032 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4033 					       struct sk_buff *skb,
4034 					       const bool check_mtu)
4035 {
4036 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4037 	    unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4038 		dev_core_stats_rx_dropped_inc(dev);
4039 		kfree_skb(skb);
4040 		return NET_RX_DROP;
4041 	}
4042 
4043 	skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4044 	skb->priority = 0;
4045 	return 0;
4046 }
4047 
4048 bool dev_nit_active(struct net_device *dev);
4049 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4050 
4051 static inline void __dev_put(struct net_device *dev)
4052 {
4053 	if (dev) {
4054 #ifdef CONFIG_PCPU_DEV_REFCNT
4055 		this_cpu_dec(*dev->pcpu_refcnt);
4056 #else
4057 		refcount_dec(&dev->dev_refcnt);
4058 #endif
4059 	}
4060 }
4061 
4062 static inline void __dev_hold(struct net_device *dev)
4063 {
4064 	if (dev) {
4065 #ifdef CONFIG_PCPU_DEV_REFCNT
4066 		this_cpu_inc(*dev->pcpu_refcnt);
4067 #else
4068 		refcount_inc(&dev->dev_refcnt);
4069 #endif
4070 	}
4071 }
4072 
4073 static inline void __netdev_tracker_alloc(struct net_device *dev,
4074 					  netdevice_tracker *tracker,
4075 					  gfp_t gfp)
4076 {
4077 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4078 	ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4079 #endif
4080 }
4081 
4082 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4083  * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4084  */
4085 static inline void netdev_tracker_alloc(struct net_device *dev,
4086 					netdevice_tracker *tracker, gfp_t gfp)
4087 {
4088 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4089 	refcount_dec(&dev->refcnt_tracker.no_tracker);
4090 	__netdev_tracker_alloc(dev, tracker, gfp);
4091 #endif
4092 }
4093 
4094 static inline void netdev_tracker_free(struct net_device *dev,
4095 				       netdevice_tracker *tracker)
4096 {
4097 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4098 	ref_tracker_free(&dev->refcnt_tracker, tracker);
4099 #endif
4100 }
4101 
4102 static inline void netdev_hold(struct net_device *dev,
4103 			       netdevice_tracker *tracker, gfp_t gfp)
4104 {
4105 	if (dev) {
4106 		__dev_hold(dev);
4107 		__netdev_tracker_alloc(dev, tracker, gfp);
4108 	}
4109 }
4110 
4111 static inline void netdev_put(struct net_device *dev,
4112 			      netdevice_tracker *tracker)
4113 {
4114 	if (dev) {
4115 		netdev_tracker_free(dev, tracker);
4116 		__dev_put(dev);
4117 	}
4118 }
4119 
4120 /**
4121  *	dev_hold - get reference to device
4122  *	@dev: network device
4123  *
4124  * Hold reference to device to keep it from being freed.
4125  * Try using netdev_hold() instead.
4126  */
4127 static inline void dev_hold(struct net_device *dev)
4128 {
4129 	netdev_hold(dev, NULL, GFP_ATOMIC);
4130 }
4131 
4132 /**
4133  *	dev_put - release reference to device
4134  *	@dev: network device
4135  *
4136  * Release reference to device to allow it to be freed.
4137  * Try using netdev_put() instead.
4138  */
4139 static inline void dev_put(struct net_device *dev)
4140 {
4141 	netdev_put(dev, NULL);
4142 }
4143 
4144 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T))
4145 
4146 static inline void netdev_ref_replace(struct net_device *odev,
4147 				      struct net_device *ndev,
4148 				      netdevice_tracker *tracker,
4149 				      gfp_t gfp)
4150 {
4151 	if (odev)
4152 		netdev_tracker_free(odev, tracker);
4153 
4154 	__dev_hold(ndev);
4155 	__dev_put(odev);
4156 
4157 	if (ndev)
4158 		__netdev_tracker_alloc(ndev, tracker, gfp);
4159 }
4160 
4161 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4162  * and _off may be called from IRQ context, but it is caller
4163  * who is responsible for serialization of these calls.
4164  *
4165  * The name carrier is inappropriate, these functions should really be
4166  * called netif_lowerlayer_*() because they represent the state of any
4167  * kind of lower layer not just hardware media.
4168  */
4169 void linkwatch_fire_event(struct net_device *dev);
4170 
4171 /**
4172  * linkwatch_sync_dev - sync linkwatch for the given device
4173  * @dev: network device to sync linkwatch for
4174  *
4175  * Sync linkwatch for the given device, removing it from the
4176  * pending work list (if queued).
4177  */
4178 void linkwatch_sync_dev(struct net_device *dev);
4179 
4180 /**
4181  *	netif_carrier_ok - test if carrier present
4182  *	@dev: network device
4183  *
4184  * Check if carrier is present on device
4185  */
4186 static inline bool netif_carrier_ok(const struct net_device *dev)
4187 {
4188 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4189 }
4190 
4191 unsigned long dev_trans_start(struct net_device *dev);
4192 
4193 void __netdev_watchdog_up(struct net_device *dev);
4194 
4195 void netif_carrier_on(struct net_device *dev);
4196 void netif_carrier_off(struct net_device *dev);
4197 void netif_carrier_event(struct net_device *dev);
4198 
4199 /**
4200  *	netif_dormant_on - mark device as dormant.
4201  *	@dev: network device
4202  *
4203  * Mark device as dormant (as per RFC2863).
4204  *
4205  * The dormant state indicates that the relevant interface is not
4206  * actually in a condition to pass packets (i.e., it is not 'up') but is
4207  * in a "pending" state, waiting for some external event.  For "on-
4208  * demand" interfaces, this new state identifies the situation where the
4209  * interface is waiting for events to place it in the up state.
4210  */
4211 static inline void netif_dormant_on(struct net_device *dev)
4212 {
4213 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4214 		linkwatch_fire_event(dev);
4215 }
4216 
4217 /**
4218  *	netif_dormant_off - set device as not dormant.
4219  *	@dev: network device
4220  *
4221  * Device is not in dormant state.
4222  */
4223 static inline void netif_dormant_off(struct net_device *dev)
4224 {
4225 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4226 		linkwatch_fire_event(dev);
4227 }
4228 
4229 /**
4230  *	netif_dormant - test if device is dormant
4231  *	@dev: network device
4232  *
4233  * Check if device is dormant.
4234  */
4235 static inline bool netif_dormant(const struct net_device *dev)
4236 {
4237 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
4238 }
4239 
4240 
4241 /**
4242  *	netif_testing_on - mark device as under test.
4243  *	@dev: network device
4244  *
4245  * Mark device as under test (as per RFC2863).
4246  *
4247  * The testing state indicates that some test(s) must be performed on
4248  * the interface. After completion, of the test, the interface state
4249  * will change to up, dormant, or down, as appropriate.
4250  */
4251 static inline void netif_testing_on(struct net_device *dev)
4252 {
4253 	if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4254 		linkwatch_fire_event(dev);
4255 }
4256 
4257 /**
4258  *	netif_testing_off - set device as not under test.
4259  *	@dev: network device
4260  *
4261  * Device is not in testing state.
4262  */
4263 static inline void netif_testing_off(struct net_device *dev)
4264 {
4265 	if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4266 		linkwatch_fire_event(dev);
4267 }
4268 
4269 /**
4270  *	netif_testing - test if device is under test
4271  *	@dev: network device
4272  *
4273  * Check if device is under test
4274  */
4275 static inline bool netif_testing(const struct net_device *dev)
4276 {
4277 	return test_bit(__LINK_STATE_TESTING, &dev->state);
4278 }
4279 
4280 
4281 /**
4282  *	netif_oper_up - test if device is operational
4283  *	@dev: network device
4284  *
4285  * Check if carrier is operational
4286  */
4287 static inline bool netif_oper_up(const struct net_device *dev)
4288 {
4289 	unsigned int operstate = READ_ONCE(dev->operstate);
4290 
4291 	return	operstate == IF_OPER_UP ||
4292 		operstate == IF_OPER_UNKNOWN /* backward compat */;
4293 }
4294 
4295 /**
4296  *	netif_device_present - is device available or removed
4297  *	@dev: network device
4298  *
4299  * Check if device has not been removed from system.
4300  */
4301 static inline bool netif_device_present(const struct net_device *dev)
4302 {
4303 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
4304 }
4305 
4306 void netif_device_detach(struct net_device *dev);
4307 
4308 void netif_device_attach(struct net_device *dev);
4309 
4310 /*
4311  * Network interface message level settings
4312  */
4313 
4314 enum {
4315 	NETIF_MSG_DRV_BIT,
4316 	NETIF_MSG_PROBE_BIT,
4317 	NETIF_MSG_LINK_BIT,
4318 	NETIF_MSG_TIMER_BIT,
4319 	NETIF_MSG_IFDOWN_BIT,
4320 	NETIF_MSG_IFUP_BIT,
4321 	NETIF_MSG_RX_ERR_BIT,
4322 	NETIF_MSG_TX_ERR_BIT,
4323 	NETIF_MSG_TX_QUEUED_BIT,
4324 	NETIF_MSG_INTR_BIT,
4325 	NETIF_MSG_TX_DONE_BIT,
4326 	NETIF_MSG_RX_STATUS_BIT,
4327 	NETIF_MSG_PKTDATA_BIT,
4328 	NETIF_MSG_HW_BIT,
4329 	NETIF_MSG_WOL_BIT,
4330 
4331 	/* When you add a new bit above, update netif_msg_class_names array
4332 	 * in net/ethtool/common.c
4333 	 */
4334 	NETIF_MSG_CLASS_COUNT,
4335 };
4336 /* Both ethtool_ops interface and internal driver implementation use u32 */
4337 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4338 
4339 #define __NETIF_MSG_BIT(bit)	((u32)1 << (bit))
4340 #define __NETIF_MSG(name)	__NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4341 
4342 #define NETIF_MSG_DRV		__NETIF_MSG(DRV)
4343 #define NETIF_MSG_PROBE		__NETIF_MSG(PROBE)
4344 #define NETIF_MSG_LINK		__NETIF_MSG(LINK)
4345 #define NETIF_MSG_TIMER		__NETIF_MSG(TIMER)
4346 #define NETIF_MSG_IFDOWN	__NETIF_MSG(IFDOWN)
4347 #define NETIF_MSG_IFUP		__NETIF_MSG(IFUP)
4348 #define NETIF_MSG_RX_ERR	__NETIF_MSG(RX_ERR)
4349 #define NETIF_MSG_TX_ERR	__NETIF_MSG(TX_ERR)
4350 #define NETIF_MSG_TX_QUEUED	__NETIF_MSG(TX_QUEUED)
4351 #define NETIF_MSG_INTR		__NETIF_MSG(INTR)
4352 #define NETIF_MSG_TX_DONE	__NETIF_MSG(TX_DONE)
4353 #define NETIF_MSG_RX_STATUS	__NETIF_MSG(RX_STATUS)
4354 #define NETIF_MSG_PKTDATA	__NETIF_MSG(PKTDATA)
4355 #define NETIF_MSG_HW		__NETIF_MSG(HW)
4356 #define NETIF_MSG_WOL		__NETIF_MSG(WOL)
4357 
4358 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
4359 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
4360 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
4361 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
4362 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
4363 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
4364 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
4365 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
4366 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4367 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
4368 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
4369 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
4370 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
4371 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
4372 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
4373 
4374 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4375 {
4376 	/* use default */
4377 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4378 		return default_msg_enable_bits;
4379 	if (debug_value == 0)	/* no output */
4380 		return 0;
4381 	/* set low N bits */
4382 	return (1U << debug_value) - 1;
4383 }
4384 
4385 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4386 {
4387 	spin_lock(&txq->_xmit_lock);
4388 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4389 	WRITE_ONCE(txq->xmit_lock_owner, cpu);
4390 }
4391 
4392 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4393 {
4394 	__acquire(&txq->_xmit_lock);
4395 	return true;
4396 }
4397 
4398 static inline void __netif_tx_release(struct netdev_queue *txq)
4399 {
4400 	__release(&txq->_xmit_lock);
4401 }
4402 
4403 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4404 {
4405 	spin_lock_bh(&txq->_xmit_lock);
4406 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4407 	WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4408 }
4409 
4410 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4411 {
4412 	bool ok = spin_trylock(&txq->_xmit_lock);
4413 
4414 	if (likely(ok)) {
4415 		/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4416 		WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4417 	}
4418 	return ok;
4419 }
4420 
4421 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4422 {
4423 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4424 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4425 	spin_unlock(&txq->_xmit_lock);
4426 }
4427 
4428 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4429 {
4430 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4431 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4432 	spin_unlock_bh(&txq->_xmit_lock);
4433 }
4434 
4435 /*
4436  * txq->trans_start can be read locklessly from dev_watchdog()
4437  */
4438 static inline void txq_trans_update(struct netdev_queue *txq)
4439 {
4440 	if (txq->xmit_lock_owner != -1)
4441 		WRITE_ONCE(txq->trans_start, jiffies);
4442 }
4443 
4444 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4445 {
4446 	unsigned long now = jiffies;
4447 
4448 	if (READ_ONCE(txq->trans_start) != now)
4449 		WRITE_ONCE(txq->trans_start, now);
4450 }
4451 
4452 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4453 static inline void netif_trans_update(struct net_device *dev)
4454 {
4455 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4456 
4457 	txq_trans_cond_update(txq);
4458 }
4459 
4460 /**
4461  *	netif_tx_lock - grab network device transmit lock
4462  *	@dev: network device
4463  *
4464  * Get network device transmit lock
4465  */
4466 void netif_tx_lock(struct net_device *dev);
4467 
4468 static inline void netif_tx_lock_bh(struct net_device *dev)
4469 {
4470 	local_bh_disable();
4471 	netif_tx_lock(dev);
4472 }
4473 
4474 void netif_tx_unlock(struct net_device *dev);
4475 
4476 static inline void netif_tx_unlock_bh(struct net_device *dev)
4477 {
4478 	netif_tx_unlock(dev);
4479 	local_bh_enable();
4480 }
4481 
4482 #define HARD_TX_LOCK(dev, txq, cpu) {			\
4483 	if (!(dev)->lltx) {				\
4484 		__netif_tx_lock(txq, cpu);		\
4485 	} else {					\
4486 		__netif_tx_acquire(txq);		\
4487 	}						\
4488 }
4489 
4490 #define HARD_TX_TRYLOCK(dev, txq)			\
4491 	(!(dev)->lltx ?					\
4492 		__netif_tx_trylock(txq) :		\
4493 		__netif_tx_acquire(txq))
4494 
4495 #define HARD_TX_UNLOCK(dev, txq) {			\
4496 	if (!(dev)->lltx) {				\
4497 		__netif_tx_unlock(txq);			\
4498 	} else {					\
4499 		__netif_tx_release(txq);		\
4500 	}						\
4501 }
4502 
4503 static inline void netif_tx_disable(struct net_device *dev)
4504 {
4505 	unsigned int i;
4506 	int cpu;
4507 
4508 	local_bh_disable();
4509 	cpu = smp_processor_id();
4510 	spin_lock(&dev->tx_global_lock);
4511 	for (i = 0; i < dev->num_tx_queues; i++) {
4512 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4513 
4514 		__netif_tx_lock(txq, cpu);
4515 		netif_tx_stop_queue(txq);
4516 		__netif_tx_unlock(txq);
4517 	}
4518 	spin_unlock(&dev->tx_global_lock);
4519 	local_bh_enable();
4520 }
4521 
4522 static inline void netif_addr_lock(struct net_device *dev)
4523 {
4524 	unsigned char nest_level = 0;
4525 
4526 #ifdef CONFIG_LOCKDEP
4527 	nest_level = dev->nested_level;
4528 #endif
4529 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4530 }
4531 
4532 static inline void netif_addr_lock_bh(struct net_device *dev)
4533 {
4534 	unsigned char nest_level = 0;
4535 
4536 #ifdef CONFIG_LOCKDEP
4537 	nest_level = dev->nested_level;
4538 #endif
4539 	local_bh_disable();
4540 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4541 }
4542 
4543 static inline void netif_addr_unlock(struct net_device *dev)
4544 {
4545 	spin_unlock(&dev->addr_list_lock);
4546 }
4547 
4548 static inline void netif_addr_unlock_bh(struct net_device *dev)
4549 {
4550 	spin_unlock_bh(&dev->addr_list_lock);
4551 }
4552 
4553 /*
4554  * dev_addrs walker. Should be used only for read access. Call with
4555  * rcu_read_lock held.
4556  */
4557 #define for_each_dev_addr(dev, ha) \
4558 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4559 
4560 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4561 
4562 void ether_setup(struct net_device *dev);
4563 
4564 /* Allocate dummy net_device */
4565 struct net_device *alloc_netdev_dummy(int sizeof_priv);
4566 
4567 /* Support for loadable net-drivers */
4568 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4569 				    unsigned char name_assign_type,
4570 				    void (*setup)(struct net_device *),
4571 				    unsigned int txqs, unsigned int rxqs);
4572 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4573 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4574 
4575 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4576 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4577 			 count)
4578 
4579 int register_netdev(struct net_device *dev);
4580 void unregister_netdev(struct net_device *dev);
4581 
4582 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4583 
4584 /* General hardware address lists handling functions */
4585 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4586 		   struct netdev_hw_addr_list *from_list, int addr_len);
4587 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4588 		      struct netdev_hw_addr_list *from_list, int addr_len);
4589 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4590 		       struct net_device *dev,
4591 		       int (*sync)(struct net_device *, const unsigned char *),
4592 		       int (*unsync)(struct net_device *,
4593 				     const unsigned char *));
4594 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4595 			   struct net_device *dev,
4596 			   int (*sync)(struct net_device *,
4597 				       const unsigned char *, int),
4598 			   int (*unsync)(struct net_device *,
4599 					 const unsigned char *, int));
4600 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4601 			      struct net_device *dev,
4602 			      int (*unsync)(struct net_device *,
4603 					    const unsigned char *, int));
4604 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4605 			  struct net_device *dev,
4606 			  int (*unsync)(struct net_device *,
4607 					const unsigned char *));
4608 void __hw_addr_init(struct netdev_hw_addr_list *list);
4609 
4610 /* Functions used for device addresses handling */
4611 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4612 		  const void *addr, size_t len);
4613 
4614 static inline void
4615 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4616 {
4617 	dev_addr_mod(dev, 0, addr, len);
4618 }
4619 
4620 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4621 {
4622 	__dev_addr_set(dev, addr, dev->addr_len);
4623 }
4624 
4625 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4626 		 unsigned char addr_type);
4627 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4628 		 unsigned char addr_type);
4629 
4630 /* Functions used for unicast addresses handling */
4631 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4632 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4633 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4634 int dev_uc_sync(struct net_device *to, struct net_device *from);
4635 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4636 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4637 void dev_uc_flush(struct net_device *dev);
4638 void dev_uc_init(struct net_device *dev);
4639 
4640 /**
4641  *  __dev_uc_sync - Synchronize device's unicast list
4642  *  @dev:  device to sync
4643  *  @sync: function to call if address should be added
4644  *  @unsync: function to call if address should be removed
4645  *
4646  *  Add newly added addresses to the interface, and release
4647  *  addresses that have been deleted.
4648  */
4649 static inline int __dev_uc_sync(struct net_device *dev,
4650 				int (*sync)(struct net_device *,
4651 					    const unsigned char *),
4652 				int (*unsync)(struct net_device *,
4653 					      const unsigned char *))
4654 {
4655 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4656 }
4657 
4658 /**
4659  *  __dev_uc_unsync - Remove synchronized addresses from device
4660  *  @dev:  device to sync
4661  *  @unsync: function to call if address should be removed
4662  *
4663  *  Remove all addresses that were added to the device by dev_uc_sync().
4664  */
4665 static inline void __dev_uc_unsync(struct net_device *dev,
4666 				   int (*unsync)(struct net_device *,
4667 						 const unsigned char *))
4668 {
4669 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
4670 }
4671 
4672 /* Functions used for multicast addresses handling */
4673 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4674 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4675 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4676 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4677 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4678 int dev_mc_sync(struct net_device *to, struct net_device *from);
4679 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4680 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4681 void dev_mc_flush(struct net_device *dev);
4682 void dev_mc_init(struct net_device *dev);
4683 
4684 /**
4685  *  __dev_mc_sync - Synchronize device's multicast list
4686  *  @dev:  device to sync
4687  *  @sync: function to call if address should be added
4688  *  @unsync: function to call if address should be removed
4689  *
4690  *  Add newly added addresses to the interface, and release
4691  *  addresses that have been deleted.
4692  */
4693 static inline int __dev_mc_sync(struct net_device *dev,
4694 				int (*sync)(struct net_device *,
4695 					    const unsigned char *),
4696 				int (*unsync)(struct net_device *,
4697 					      const unsigned char *))
4698 {
4699 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4700 }
4701 
4702 /**
4703  *  __dev_mc_unsync - Remove synchronized addresses from device
4704  *  @dev:  device to sync
4705  *  @unsync: function to call if address should be removed
4706  *
4707  *  Remove all addresses that were added to the device by dev_mc_sync().
4708  */
4709 static inline void __dev_mc_unsync(struct net_device *dev,
4710 				   int (*unsync)(struct net_device *,
4711 						 const unsigned char *))
4712 {
4713 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
4714 }
4715 
4716 /* Functions used for secondary unicast and multicast support */
4717 void dev_set_rx_mode(struct net_device *dev);
4718 int dev_set_promiscuity(struct net_device *dev, int inc);
4719 int dev_set_allmulti(struct net_device *dev, int inc);
4720 void netdev_state_change(struct net_device *dev);
4721 void __netdev_notify_peers(struct net_device *dev);
4722 void netdev_notify_peers(struct net_device *dev);
4723 void netdev_features_change(struct net_device *dev);
4724 /* Load a device via the kmod */
4725 void dev_load(struct net *net, const char *name);
4726 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4727 					struct rtnl_link_stats64 *storage);
4728 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4729 			     const struct net_device_stats *netdev_stats);
4730 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4731 			   const struct pcpu_sw_netstats __percpu *netstats);
4732 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4733 
4734 enum {
4735 	NESTED_SYNC_IMM_BIT,
4736 	NESTED_SYNC_TODO_BIT,
4737 };
4738 
4739 #define __NESTED_SYNC_BIT(bit)	((u32)1 << (bit))
4740 #define __NESTED_SYNC(name)	__NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4741 
4742 #define NESTED_SYNC_IMM		__NESTED_SYNC(IMM)
4743 #define NESTED_SYNC_TODO	__NESTED_SYNC(TODO)
4744 
4745 struct netdev_nested_priv {
4746 	unsigned char flags;
4747 	void *data;
4748 };
4749 
4750 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4751 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4752 						     struct list_head **iter);
4753 
4754 /* iterate through upper list, must be called under RCU read lock */
4755 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4756 	for (iter = &(dev)->adj_list.upper, \
4757 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4758 	     updev; \
4759 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4760 
4761 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4762 				  int (*fn)(struct net_device *upper_dev,
4763 					    struct netdev_nested_priv *priv),
4764 				  struct netdev_nested_priv *priv);
4765 
4766 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4767 				  struct net_device *upper_dev);
4768 
4769 bool netdev_has_any_upper_dev(struct net_device *dev);
4770 
4771 void *netdev_lower_get_next_private(struct net_device *dev,
4772 				    struct list_head **iter);
4773 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4774 					struct list_head **iter);
4775 
4776 #define netdev_for_each_lower_private(dev, priv, iter) \
4777 	for (iter = (dev)->adj_list.lower.next, \
4778 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
4779 	     priv; \
4780 	     priv = netdev_lower_get_next_private(dev, &(iter)))
4781 
4782 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4783 	for (iter = &(dev)->adj_list.lower, \
4784 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4785 	     priv; \
4786 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4787 
4788 void *netdev_lower_get_next(struct net_device *dev,
4789 				struct list_head **iter);
4790 
4791 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4792 	for (iter = (dev)->adj_list.lower.next, \
4793 	     ldev = netdev_lower_get_next(dev, &(iter)); \
4794 	     ldev; \
4795 	     ldev = netdev_lower_get_next(dev, &(iter)))
4796 
4797 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4798 					     struct list_head **iter);
4799 int netdev_walk_all_lower_dev(struct net_device *dev,
4800 			      int (*fn)(struct net_device *lower_dev,
4801 					struct netdev_nested_priv *priv),
4802 			      struct netdev_nested_priv *priv);
4803 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4804 				  int (*fn)(struct net_device *lower_dev,
4805 					    struct netdev_nested_priv *priv),
4806 				  struct netdev_nested_priv *priv);
4807 
4808 void *netdev_adjacent_get_private(struct list_head *adj_list);
4809 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4810 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4811 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4812 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4813 			  struct netlink_ext_ack *extack);
4814 int netdev_master_upper_dev_link(struct net_device *dev,
4815 				 struct net_device *upper_dev,
4816 				 void *upper_priv, void *upper_info,
4817 				 struct netlink_ext_ack *extack);
4818 void netdev_upper_dev_unlink(struct net_device *dev,
4819 			     struct net_device *upper_dev);
4820 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4821 				   struct net_device *new_dev,
4822 				   struct net_device *dev,
4823 				   struct netlink_ext_ack *extack);
4824 void netdev_adjacent_change_commit(struct net_device *old_dev,
4825 				   struct net_device *new_dev,
4826 				   struct net_device *dev);
4827 void netdev_adjacent_change_abort(struct net_device *old_dev,
4828 				  struct net_device *new_dev,
4829 				  struct net_device *dev);
4830 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4831 void *netdev_lower_dev_get_private(struct net_device *dev,
4832 				   struct net_device *lower_dev);
4833 void netdev_lower_state_changed(struct net_device *lower_dev,
4834 				void *lower_state_info);
4835 
4836 /* RSS keys are 40 or 52 bytes long */
4837 #define NETDEV_RSS_KEY_LEN 52
4838 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4839 void netdev_rss_key_fill(void *buffer, size_t len);
4840 
4841 int skb_checksum_help(struct sk_buff *skb);
4842 int skb_crc32c_csum_help(struct sk_buff *skb);
4843 int skb_csum_hwoffload_help(struct sk_buff *skb,
4844 			    const netdev_features_t features);
4845 
4846 struct netdev_bonding_info {
4847 	ifslave	slave;
4848 	ifbond	master;
4849 };
4850 
4851 struct netdev_notifier_bonding_info {
4852 	struct netdev_notifier_info info; /* must be first */
4853 	struct netdev_bonding_info  bonding_info;
4854 };
4855 
4856 void netdev_bonding_info_change(struct net_device *dev,
4857 				struct netdev_bonding_info *bonding_info);
4858 
4859 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4860 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4861 #else
4862 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4863 				  const void *data)
4864 {
4865 }
4866 #endif
4867 
4868 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4869 
4870 static inline bool can_checksum_protocol(netdev_features_t features,
4871 					 __be16 protocol)
4872 {
4873 	if (protocol == htons(ETH_P_FCOE))
4874 		return !!(features & NETIF_F_FCOE_CRC);
4875 
4876 	/* Assume this is an IP checksum (not SCTP CRC) */
4877 
4878 	if (features & NETIF_F_HW_CSUM) {
4879 		/* Can checksum everything */
4880 		return true;
4881 	}
4882 
4883 	switch (protocol) {
4884 	case htons(ETH_P_IP):
4885 		return !!(features & NETIF_F_IP_CSUM);
4886 	case htons(ETH_P_IPV6):
4887 		return !!(features & NETIF_F_IPV6_CSUM);
4888 	default:
4889 		return false;
4890 	}
4891 }
4892 
4893 #ifdef CONFIG_BUG
4894 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4895 #else
4896 static inline void netdev_rx_csum_fault(struct net_device *dev,
4897 					struct sk_buff *skb)
4898 {
4899 }
4900 #endif
4901 /* rx skb timestamps */
4902 void net_enable_timestamp(void);
4903 void net_disable_timestamp(void);
4904 
4905 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4906 					const struct skb_shared_hwtstamps *hwtstamps,
4907 					bool cycles)
4908 {
4909 	const struct net_device_ops *ops = dev->netdev_ops;
4910 
4911 	if (ops->ndo_get_tstamp)
4912 		return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4913 
4914 	return hwtstamps->hwtstamp;
4915 }
4916 
4917 #ifndef CONFIG_PREEMPT_RT
4918 static inline void netdev_xmit_set_more(bool more)
4919 {
4920 	__this_cpu_write(softnet_data.xmit.more, more);
4921 }
4922 
4923 static inline bool netdev_xmit_more(void)
4924 {
4925 	return __this_cpu_read(softnet_data.xmit.more);
4926 }
4927 #else
4928 static inline void netdev_xmit_set_more(bool more)
4929 {
4930 	current->net_xmit.more = more;
4931 }
4932 
4933 static inline bool netdev_xmit_more(void)
4934 {
4935 	return current->net_xmit.more;
4936 }
4937 #endif
4938 
4939 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4940 					      struct sk_buff *skb, struct net_device *dev,
4941 					      bool more)
4942 {
4943 	netdev_xmit_set_more(more);
4944 	return ops->ndo_start_xmit(skb, dev);
4945 }
4946 
4947 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4948 					    struct netdev_queue *txq, bool more)
4949 {
4950 	const struct net_device_ops *ops = dev->netdev_ops;
4951 	netdev_tx_t rc;
4952 
4953 	rc = __netdev_start_xmit(ops, skb, dev, more);
4954 	if (rc == NETDEV_TX_OK)
4955 		txq_trans_update(txq);
4956 
4957 	return rc;
4958 }
4959 
4960 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4961 				const void *ns);
4962 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4963 				 const void *ns);
4964 
4965 extern const struct kobj_ns_type_operations net_ns_type_operations;
4966 
4967 const char *netdev_drivername(const struct net_device *dev);
4968 
4969 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4970 							  netdev_features_t f2)
4971 {
4972 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4973 		if (f1 & NETIF_F_HW_CSUM)
4974 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4975 		else
4976 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4977 	}
4978 
4979 	return f1 & f2;
4980 }
4981 
4982 static inline netdev_features_t netdev_get_wanted_features(
4983 	struct net_device *dev)
4984 {
4985 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4986 }
4987 netdev_features_t netdev_increment_features(netdev_features_t all,
4988 	netdev_features_t one, netdev_features_t mask);
4989 
4990 /* Allow TSO being used on stacked device :
4991  * Performing the GSO segmentation before last device
4992  * is a performance improvement.
4993  */
4994 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4995 							netdev_features_t mask)
4996 {
4997 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4998 }
4999 
5000 int __netdev_update_features(struct net_device *dev);
5001 void netdev_update_features(struct net_device *dev);
5002 void netdev_change_features(struct net_device *dev);
5003 
5004 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5005 					struct net_device *dev);
5006 
5007 netdev_features_t passthru_features_check(struct sk_buff *skb,
5008 					  struct net_device *dev,
5009 					  netdev_features_t features);
5010 netdev_features_t netif_skb_features(struct sk_buff *skb);
5011 void skb_warn_bad_offload(const struct sk_buff *skb);
5012 
5013 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5014 {
5015 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5016 
5017 	/* check flags correspondence */
5018 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5019 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5020 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5021 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5022 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5023 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5024 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5025 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5026 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5027 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5028 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5029 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5030 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5031 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5032 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5033 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5034 	BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5035 	BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5036 	BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5037 
5038 	return (features & feature) == feature;
5039 }
5040 
5041 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5042 {
5043 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5044 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5045 }
5046 
5047 static inline bool netif_needs_gso(struct sk_buff *skb,
5048 				   netdev_features_t features)
5049 {
5050 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5051 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5052 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5053 }
5054 
5055 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5056 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5057 void netif_inherit_tso_max(struct net_device *to,
5058 			   const struct net_device *from);
5059 
5060 static inline unsigned int
5061 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb)
5062 {
5063 	/* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */
5064 	return skb->protocol == htons(ETH_P_IPV6) ?
5065 	       READ_ONCE(dev->gro_max_size) :
5066 	       READ_ONCE(dev->gro_ipv4_max_size);
5067 }
5068 
5069 static inline unsigned int
5070 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb)
5071 {
5072 	/* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */
5073 	return skb->protocol == htons(ETH_P_IPV6) ?
5074 	       READ_ONCE(dev->gso_max_size) :
5075 	       READ_ONCE(dev->gso_ipv4_max_size);
5076 }
5077 
5078 static inline bool netif_is_macsec(const struct net_device *dev)
5079 {
5080 	return dev->priv_flags & IFF_MACSEC;
5081 }
5082 
5083 static inline bool netif_is_macvlan(const struct net_device *dev)
5084 {
5085 	return dev->priv_flags & IFF_MACVLAN;
5086 }
5087 
5088 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5089 {
5090 	return dev->priv_flags & IFF_MACVLAN_PORT;
5091 }
5092 
5093 static inline bool netif_is_bond_master(const struct net_device *dev)
5094 {
5095 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5096 }
5097 
5098 static inline bool netif_is_bond_slave(const struct net_device *dev)
5099 {
5100 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5101 }
5102 
5103 static inline bool netif_supports_nofcs(struct net_device *dev)
5104 {
5105 	return dev->priv_flags & IFF_SUPP_NOFCS;
5106 }
5107 
5108 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5109 {
5110 	return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5111 }
5112 
5113 static inline bool netif_is_l3_master(const struct net_device *dev)
5114 {
5115 	return dev->priv_flags & IFF_L3MDEV_MASTER;
5116 }
5117 
5118 static inline bool netif_is_l3_slave(const struct net_device *dev)
5119 {
5120 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
5121 }
5122 
5123 static inline int dev_sdif(const struct net_device *dev)
5124 {
5125 #ifdef CONFIG_NET_L3_MASTER_DEV
5126 	if (netif_is_l3_slave(dev))
5127 		return dev->ifindex;
5128 #endif
5129 	return 0;
5130 }
5131 
5132 static inline bool netif_is_bridge_master(const struct net_device *dev)
5133 {
5134 	return dev->priv_flags & IFF_EBRIDGE;
5135 }
5136 
5137 static inline bool netif_is_bridge_port(const struct net_device *dev)
5138 {
5139 	return dev->priv_flags & IFF_BRIDGE_PORT;
5140 }
5141 
5142 static inline bool netif_is_ovs_master(const struct net_device *dev)
5143 {
5144 	return dev->priv_flags & IFF_OPENVSWITCH;
5145 }
5146 
5147 static inline bool netif_is_ovs_port(const struct net_device *dev)
5148 {
5149 	return dev->priv_flags & IFF_OVS_DATAPATH;
5150 }
5151 
5152 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5153 {
5154 	return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5155 }
5156 
5157 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5158 {
5159 	return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5160 }
5161 
5162 static inline bool netif_is_team_master(const struct net_device *dev)
5163 {
5164 	return dev->priv_flags & IFF_TEAM;
5165 }
5166 
5167 static inline bool netif_is_team_port(const struct net_device *dev)
5168 {
5169 	return dev->priv_flags & IFF_TEAM_PORT;
5170 }
5171 
5172 static inline bool netif_is_lag_master(const struct net_device *dev)
5173 {
5174 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
5175 }
5176 
5177 static inline bool netif_is_lag_port(const struct net_device *dev)
5178 {
5179 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5180 }
5181 
5182 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5183 {
5184 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
5185 }
5186 
5187 static inline bool netif_is_failover(const struct net_device *dev)
5188 {
5189 	return dev->priv_flags & IFF_FAILOVER;
5190 }
5191 
5192 static inline bool netif_is_failover_slave(const struct net_device *dev)
5193 {
5194 	return dev->priv_flags & IFF_FAILOVER_SLAVE;
5195 }
5196 
5197 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5198 static inline void netif_keep_dst(struct net_device *dev)
5199 {
5200 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5201 }
5202 
5203 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5204 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5205 {
5206 	/* TODO: reserve and use an additional IFF bit, if we get more users */
5207 	return netif_is_macsec(dev);
5208 }
5209 
5210 extern struct pernet_operations __net_initdata loopback_net_ops;
5211 
5212 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5213 
5214 /* netdev_printk helpers, similar to dev_printk */
5215 
5216 static inline const char *netdev_name(const struct net_device *dev)
5217 {
5218 	if (!dev->name[0] || strchr(dev->name, '%'))
5219 		return "(unnamed net_device)";
5220 	return dev->name;
5221 }
5222 
5223 static inline const char *netdev_reg_state(const struct net_device *dev)
5224 {
5225 	u8 reg_state = READ_ONCE(dev->reg_state);
5226 
5227 	switch (reg_state) {
5228 	case NETREG_UNINITIALIZED: return " (uninitialized)";
5229 	case NETREG_REGISTERED: return "";
5230 	case NETREG_UNREGISTERING: return " (unregistering)";
5231 	case NETREG_UNREGISTERED: return " (unregistered)";
5232 	case NETREG_RELEASED: return " (released)";
5233 	case NETREG_DUMMY: return " (dummy)";
5234 	}
5235 
5236 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state);
5237 	return " (unknown)";
5238 }
5239 
5240 #define MODULE_ALIAS_NETDEV(device) \
5241 	MODULE_ALIAS("netdev-" device)
5242 
5243 /*
5244  * netdev_WARN() acts like dev_printk(), but with the key difference
5245  * of using a WARN/WARN_ON to get the message out, including the
5246  * file/line information and a backtrace.
5247  */
5248 #define netdev_WARN(dev, format, args...)			\
5249 	WARN(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5250 	     netdev_reg_state(dev), ##args)
5251 
5252 #define netdev_WARN_ONCE(dev, format, args...)				\
5253 	WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5254 		  netdev_reg_state(dev), ##args)
5255 
5256 /*
5257  *	The list of packet types we will receive (as opposed to discard)
5258  *	and the routines to invoke.
5259  *
5260  *	Why 16. Because with 16 the only overlap we get on a hash of the
5261  *	low nibble of the protocol value is RARP/SNAP/X.25.
5262  *
5263  *		0800	IP
5264  *		0001	802.3
5265  *		0002	AX.25
5266  *		0004	802.2
5267  *		8035	RARP
5268  *		0005	SNAP
5269  *		0805	X.25
5270  *		0806	ARP
5271  *		8137	IPX
5272  *		0009	Localtalk
5273  *		86DD	IPv6
5274  */
5275 #define PTYPE_HASH_SIZE	(16)
5276 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
5277 
5278 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5279 
5280 extern struct net_device *blackhole_netdev;
5281 
5282 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5283 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5284 #define DEV_STATS_ADD(DEV, FIELD, VAL) 	\
5285 		atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5286 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5287 
5288 #endif	/* _LINUX_NETDEVICE_H */
5289