xref: /linux-6.15/include/linux/netdevice.h (revision 2ff80cef)
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 #include <net/neighbour_tables.h>
56 
57 struct netpoll_info;
58 struct device;
59 struct ethtool_ops;
60 struct kernel_hwtstamp_config;
61 struct phy_device;
62 struct dsa_port;
63 struct ip_tunnel_parm_kern;
64 struct macsec_context;
65 struct macsec_ops;
66 struct netdev_name_node;
67 struct sd_flow_limit;
68 struct sfp_bus;
69 /* 802.11 specific */
70 struct wireless_dev;
71 /* 802.15.4 specific */
72 struct wpan_dev;
73 struct mpls_dev;
74 /* UDP Tunnel offloads */
75 struct udp_tunnel_info;
76 struct udp_tunnel_nic_info;
77 struct udp_tunnel_nic;
78 struct bpf_prog;
79 struct xdp_buff;
80 struct xdp_frame;
81 struct xdp_metadata_ops;
82 struct xdp_md;
83 struct ethtool_netdev_state;
84 struct phy_link_topology;
85 struct hwtstamp_provider;
86 
87 typedef u32 xdp_features_t;
88 
89 void synchronize_net(void);
90 void netdev_set_default_ethtool_ops(struct net_device *dev,
91 				    const struct ethtool_ops *ops);
92 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
93 
94 /* Backlog congestion levels */
95 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
96 #define NET_RX_DROP		1	/* packet dropped */
97 
98 #define MAX_NEST_DEV 8
99 
100 /*
101  * Transmit return codes: transmit return codes originate from three different
102  * namespaces:
103  *
104  * - qdisc return codes
105  * - driver transmit return codes
106  * - errno values
107  *
108  * Drivers are allowed to return any one of those in their hard_start_xmit()
109  * function. Real network devices commonly used with qdiscs should only return
110  * the driver transmit return codes though - when qdiscs are used, the actual
111  * transmission happens asynchronously, so the value is not propagated to
112  * higher layers. Virtual network devices transmit synchronously; in this case
113  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
114  * others are propagated to higher layers.
115  */
116 
117 /* qdisc ->enqueue() return codes. */
118 #define NET_XMIT_SUCCESS	0x00
119 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
120 #define NET_XMIT_CN		0x02	/* congestion notification	*/
121 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
122 
123 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
124  * indicates that the device will soon be dropping packets, or already drops
125  * some packets of the same priority; prompting us to send less aggressively. */
126 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
127 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
128 
129 /* Driver transmit return codes */
130 #define NETDEV_TX_MASK		0xf0
131 
132 enum netdev_tx {
133 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
134 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
135 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
136 };
137 typedef enum netdev_tx netdev_tx_t;
138 
139 /*
140  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
141  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
142  */
143 static inline bool dev_xmit_complete(int rc)
144 {
145 	/*
146 	 * Positive cases with an skb consumed by a driver:
147 	 * - successful transmission (rc == NETDEV_TX_OK)
148 	 * - error while transmitting (rc < 0)
149 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
150 	 */
151 	if (likely(rc < NET_XMIT_MASK))
152 		return true;
153 
154 	return false;
155 }
156 
157 /*
158  *	Compute the worst-case header length according to the protocols
159  *	used.
160  */
161 
162 #if defined(CONFIG_HYPERV_NET)
163 # define LL_MAX_HEADER 128
164 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
165 # if defined(CONFIG_MAC80211_MESH)
166 #  define LL_MAX_HEADER 128
167 # else
168 #  define LL_MAX_HEADER 96
169 # endif
170 #else
171 # define LL_MAX_HEADER 32
172 #endif
173 
174 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
175     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
176 #define MAX_HEADER LL_MAX_HEADER
177 #else
178 #define MAX_HEADER (LL_MAX_HEADER + 48)
179 #endif
180 
181 /*
182  *	Old network device statistics. Fields are native words
183  *	(unsigned long) so they can be read and written atomically.
184  */
185 
186 #define NET_DEV_STAT(FIELD)			\
187 	union {					\
188 		unsigned long FIELD;		\
189 		atomic_long_t __##FIELD;	\
190 	}
191 
192 struct net_device_stats {
193 	NET_DEV_STAT(rx_packets);
194 	NET_DEV_STAT(tx_packets);
195 	NET_DEV_STAT(rx_bytes);
196 	NET_DEV_STAT(tx_bytes);
197 	NET_DEV_STAT(rx_errors);
198 	NET_DEV_STAT(tx_errors);
199 	NET_DEV_STAT(rx_dropped);
200 	NET_DEV_STAT(tx_dropped);
201 	NET_DEV_STAT(multicast);
202 	NET_DEV_STAT(collisions);
203 	NET_DEV_STAT(rx_length_errors);
204 	NET_DEV_STAT(rx_over_errors);
205 	NET_DEV_STAT(rx_crc_errors);
206 	NET_DEV_STAT(rx_frame_errors);
207 	NET_DEV_STAT(rx_fifo_errors);
208 	NET_DEV_STAT(rx_missed_errors);
209 	NET_DEV_STAT(tx_aborted_errors);
210 	NET_DEV_STAT(tx_carrier_errors);
211 	NET_DEV_STAT(tx_fifo_errors);
212 	NET_DEV_STAT(tx_heartbeat_errors);
213 	NET_DEV_STAT(tx_window_errors);
214 	NET_DEV_STAT(rx_compressed);
215 	NET_DEV_STAT(tx_compressed);
216 };
217 #undef NET_DEV_STAT
218 
219 /* per-cpu stats, allocated on demand.
220  * Try to fit them in a single cache line, for dev_get_stats() sake.
221  */
222 struct net_device_core_stats {
223 	unsigned long	rx_dropped;
224 	unsigned long	tx_dropped;
225 	unsigned long	rx_nohandler;
226 	unsigned long	rx_otherhost_dropped;
227 } __aligned(4 * sizeof(unsigned long));
228 
229 #include <linux/cache.h>
230 #include <linux/skbuff.h>
231 
232 struct neighbour;
233 struct neigh_parms;
234 struct sk_buff;
235 
236 struct netdev_hw_addr {
237 	struct list_head	list;
238 	struct rb_node		node;
239 	unsigned char		addr[MAX_ADDR_LEN];
240 	unsigned char		type;
241 #define NETDEV_HW_ADDR_T_LAN		1
242 #define NETDEV_HW_ADDR_T_SAN		2
243 #define NETDEV_HW_ADDR_T_UNICAST	3
244 #define NETDEV_HW_ADDR_T_MULTICAST	4
245 	bool			global_use;
246 	int			sync_cnt;
247 	int			refcount;
248 	int			synced;
249 	struct rcu_head		rcu_head;
250 };
251 
252 struct netdev_hw_addr_list {
253 	struct list_head	list;
254 	int			count;
255 
256 	/* Auxiliary tree for faster lookup on addition and deletion */
257 	struct rb_root		tree;
258 };
259 
260 #define netdev_hw_addr_list_count(l) ((l)->count)
261 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
262 #define netdev_hw_addr_list_for_each(ha, l) \
263 	list_for_each_entry(ha, &(l)->list, list)
264 
265 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
266 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
267 #define netdev_for_each_uc_addr(ha, dev) \
268 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
269 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
270 	netdev_for_each_uc_addr((_ha), (_dev)) \
271 		if ((_ha)->sync_cnt)
272 
273 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
274 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
275 #define netdev_for_each_mc_addr(ha, dev) \
276 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
277 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
278 	netdev_for_each_mc_addr((_ha), (_dev)) \
279 		if ((_ha)->sync_cnt)
280 
281 struct hh_cache {
282 	unsigned int	hh_len;
283 	seqlock_t	hh_lock;
284 
285 	/* cached hardware header; allow for machine alignment needs.        */
286 #define HH_DATA_MOD	16
287 #define HH_DATA_OFF(__len) \
288 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
289 #define HH_DATA_ALIGN(__len) \
290 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
291 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
292 };
293 
294 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
295  * Alternative is:
296  *   dev->hard_header_len ? (dev->hard_header_len +
297  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
298  *
299  * We could use other alignment values, but we must maintain the
300  * relationship HH alignment <= LL alignment.
301  */
302 #define LL_RESERVED_SPACE(dev) \
303 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
304 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
305 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
306 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
307 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
308 
309 struct header_ops {
310 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
311 			   unsigned short type, const void *daddr,
312 			   const void *saddr, unsigned int len);
313 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
314 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
315 	void	(*cache_update)(struct hh_cache *hh,
316 				const struct net_device *dev,
317 				const unsigned char *haddr);
318 	bool	(*validate)(const char *ll_header, unsigned int len);
319 	__be16	(*parse_protocol)(const struct sk_buff *skb);
320 };
321 
322 /* These flag bits are private to the generic network queueing
323  * layer; they may not be explicitly referenced by any other
324  * code.
325  */
326 
327 enum netdev_state_t {
328 	__LINK_STATE_START,
329 	__LINK_STATE_PRESENT,
330 	__LINK_STATE_NOCARRIER,
331 	__LINK_STATE_LINKWATCH_PENDING,
332 	__LINK_STATE_DORMANT,
333 	__LINK_STATE_TESTING,
334 };
335 
336 struct gro_list {
337 	struct list_head	list;
338 	int			count;
339 };
340 
341 /*
342  * size of gro hash buckets, must less than bit number of
343  * napi_struct::gro_bitmask
344  */
345 #define GRO_HASH_BUCKETS	8
346 
347 /*
348  * Structure for per-NAPI config
349  */
350 struct napi_config {
351 	u64 gro_flush_timeout;
352 	u64 irq_suspend_timeout;
353 	u32 defer_hard_irqs;
354 	unsigned int napi_id;
355 };
356 
357 /*
358  * Structure for NAPI scheduling similar to tasklet but with weighting
359  */
360 struct napi_struct {
361 	/* The poll_list must only be managed by the entity which
362 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
363 	 * whoever atomically sets that bit can add this napi_struct
364 	 * to the per-CPU poll_list, and whoever clears that bit
365 	 * can remove from the list right before clearing the bit.
366 	 */
367 	struct list_head	poll_list;
368 
369 	unsigned long		state;
370 	int			weight;
371 	u32			defer_hard_irqs_count;
372 	unsigned long		gro_bitmask;
373 	int			(*poll)(struct napi_struct *, int);
374 #ifdef CONFIG_NETPOLL
375 	/* CPU actively polling if netpoll is configured */
376 	int			poll_owner;
377 #endif
378 	/* CPU on which NAPI has been scheduled for processing */
379 	int			list_owner;
380 	struct net_device	*dev;
381 	struct gro_list		gro_hash[GRO_HASH_BUCKETS];
382 	struct sk_buff		*skb;
383 	struct list_head	rx_list; /* Pending GRO_NORMAL skbs */
384 	int			rx_count; /* length of rx_list */
385 	unsigned int		napi_id;
386 	struct hrtimer		timer;
387 	struct task_struct	*thread;
388 	unsigned long		gro_flush_timeout;
389 	unsigned long		irq_suspend_timeout;
390 	u32			defer_hard_irqs;
391 	/* control-path-only fields follow */
392 	struct list_head	dev_list;
393 	struct hlist_node	napi_hash_node;
394 	int			irq;
395 	int			index;
396 	struct napi_config	*config;
397 };
398 
399 enum {
400 	NAPI_STATE_SCHED,		/* Poll is scheduled */
401 	NAPI_STATE_MISSED,		/* reschedule a napi */
402 	NAPI_STATE_DISABLE,		/* Disable pending */
403 	NAPI_STATE_NPSVC,		/* Netpoll - don't dequeue from poll_list */
404 	NAPI_STATE_LISTED,		/* NAPI added to system lists */
405 	NAPI_STATE_NO_BUSY_POLL,	/* Do not add in napi_hash, no busy polling */
406 	NAPI_STATE_IN_BUSY_POLL,	/* sk_busy_loop() owns this NAPI */
407 	NAPI_STATE_PREFER_BUSY_POLL,	/* prefer busy-polling over softirq processing*/
408 	NAPI_STATE_THREADED,		/* The poll is performed inside its own thread*/
409 	NAPI_STATE_SCHED_THREADED,	/* Napi is currently scheduled in threaded mode */
410 };
411 
412 enum {
413 	NAPIF_STATE_SCHED		= BIT(NAPI_STATE_SCHED),
414 	NAPIF_STATE_MISSED		= BIT(NAPI_STATE_MISSED),
415 	NAPIF_STATE_DISABLE		= BIT(NAPI_STATE_DISABLE),
416 	NAPIF_STATE_NPSVC		= BIT(NAPI_STATE_NPSVC),
417 	NAPIF_STATE_LISTED		= BIT(NAPI_STATE_LISTED),
418 	NAPIF_STATE_NO_BUSY_POLL	= BIT(NAPI_STATE_NO_BUSY_POLL),
419 	NAPIF_STATE_IN_BUSY_POLL	= BIT(NAPI_STATE_IN_BUSY_POLL),
420 	NAPIF_STATE_PREFER_BUSY_POLL	= BIT(NAPI_STATE_PREFER_BUSY_POLL),
421 	NAPIF_STATE_THREADED		= BIT(NAPI_STATE_THREADED),
422 	NAPIF_STATE_SCHED_THREADED	= BIT(NAPI_STATE_SCHED_THREADED),
423 };
424 
425 enum gro_result {
426 	GRO_MERGED,
427 	GRO_MERGED_FREE,
428 	GRO_HELD,
429 	GRO_NORMAL,
430 	GRO_CONSUMED,
431 };
432 typedef enum gro_result gro_result_t;
433 
434 /*
435  * enum rx_handler_result - Possible return values for rx_handlers.
436  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
437  * further.
438  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
439  * case skb->dev was changed by rx_handler.
440  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
441  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
442  *
443  * rx_handlers are functions called from inside __netif_receive_skb(), to do
444  * special processing of the skb, prior to delivery to protocol handlers.
445  *
446  * Currently, a net_device can only have a single rx_handler registered. Trying
447  * to register a second rx_handler will return -EBUSY.
448  *
449  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
450  * To unregister a rx_handler on a net_device, use
451  * netdev_rx_handler_unregister().
452  *
453  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
454  * do with the skb.
455  *
456  * If the rx_handler consumed the skb in some way, it should return
457  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
458  * the skb to be delivered in some other way.
459  *
460  * If the rx_handler changed skb->dev, to divert the skb to another
461  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
462  * new device will be called if it exists.
463  *
464  * If the rx_handler decides the skb should be ignored, it should return
465  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
466  * are registered on exact device (ptype->dev == skb->dev).
467  *
468  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
469  * delivered, it should return RX_HANDLER_PASS.
470  *
471  * A device without a registered rx_handler will behave as if rx_handler
472  * returned RX_HANDLER_PASS.
473  */
474 
475 enum rx_handler_result {
476 	RX_HANDLER_CONSUMED,
477 	RX_HANDLER_ANOTHER,
478 	RX_HANDLER_EXACT,
479 	RX_HANDLER_PASS,
480 };
481 typedef enum rx_handler_result rx_handler_result_t;
482 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
483 
484 void __napi_schedule(struct napi_struct *n);
485 void __napi_schedule_irqoff(struct napi_struct *n);
486 
487 static inline bool napi_disable_pending(struct napi_struct *n)
488 {
489 	return test_bit(NAPI_STATE_DISABLE, &n->state);
490 }
491 
492 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
493 {
494 	return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
495 }
496 
497 /**
498  * napi_is_scheduled - test if NAPI is scheduled
499  * @n: NAPI context
500  *
501  * This check is "best-effort". With no locking implemented,
502  * a NAPI can be scheduled or terminate right after this check
503  * and produce not precise results.
504  *
505  * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
506  * should not be used normally and napi_schedule should be
507  * used instead.
508  *
509  * Use only if the driver really needs to check if a NAPI
510  * is scheduled for example in the context of delayed timer
511  * that can be skipped if a NAPI is already scheduled.
512  *
513  * Return: True if NAPI is scheduled, False otherwise.
514  */
515 static inline bool napi_is_scheduled(struct napi_struct *n)
516 {
517 	return test_bit(NAPI_STATE_SCHED, &n->state);
518 }
519 
520 bool napi_schedule_prep(struct napi_struct *n);
521 
522 /**
523  *	napi_schedule - schedule NAPI poll
524  *	@n: NAPI context
525  *
526  * Schedule NAPI poll routine to be called if it is not already
527  * running.
528  * Return: true if we schedule a NAPI or false if not.
529  * Refer to napi_schedule_prep() for additional reason on why
530  * a NAPI might not be scheduled.
531  */
532 static inline bool napi_schedule(struct napi_struct *n)
533 {
534 	if (napi_schedule_prep(n)) {
535 		__napi_schedule(n);
536 		return true;
537 	}
538 
539 	return false;
540 }
541 
542 /**
543  *	napi_schedule_irqoff - schedule NAPI poll
544  *	@n: NAPI context
545  *
546  * Variant of napi_schedule(), assuming hard irqs are masked.
547  */
548 static inline void napi_schedule_irqoff(struct napi_struct *n)
549 {
550 	if (napi_schedule_prep(n))
551 		__napi_schedule_irqoff(n);
552 }
553 
554 /**
555  * napi_complete_done - NAPI processing complete
556  * @n: NAPI context
557  * @work_done: number of packets processed
558  *
559  * Mark NAPI processing as complete. Should only be called if poll budget
560  * has not been completely consumed.
561  * Prefer over napi_complete().
562  * Return: false if device should avoid rearming interrupts.
563  */
564 bool napi_complete_done(struct napi_struct *n, int work_done);
565 
566 static inline bool napi_complete(struct napi_struct *n)
567 {
568 	return napi_complete_done(n, 0);
569 }
570 
571 int dev_set_threaded(struct net_device *dev, bool threaded);
572 
573 /**
574  *	napi_disable - prevent NAPI from scheduling
575  *	@n: NAPI context
576  *
577  * Stop NAPI from being scheduled on this context.
578  * Waits till any outstanding processing completes.
579  */
580 void napi_disable(struct napi_struct *n);
581 
582 void napi_enable(struct napi_struct *n);
583 
584 /**
585  *	napi_synchronize - wait until NAPI is not running
586  *	@n: NAPI context
587  *
588  * Wait until NAPI is done being scheduled on this context.
589  * Waits till any outstanding processing completes but
590  * does not disable future activations.
591  */
592 static inline void napi_synchronize(const struct napi_struct *n)
593 {
594 	if (IS_ENABLED(CONFIG_SMP))
595 		while (test_bit(NAPI_STATE_SCHED, &n->state))
596 			msleep(1);
597 	else
598 		barrier();
599 }
600 
601 /**
602  *	napi_if_scheduled_mark_missed - if napi is running, set the
603  *	NAPIF_STATE_MISSED
604  *	@n: NAPI context
605  *
606  * If napi is running, set the NAPIF_STATE_MISSED, and return true if
607  * NAPI is scheduled.
608  **/
609 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
610 {
611 	unsigned long val, new;
612 
613 	val = READ_ONCE(n->state);
614 	do {
615 		if (val & NAPIF_STATE_DISABLE)
616 			return true;
617 
618 		if (!(val & NAPIF_STATE_SCHED))
619 			return false;
620 
621 		new = val | NAPIF_STATE_MISSED;
622 	} while (!try_cmpxchg(&n->state, &val, new));
623 
624 	return true;
625 }
626 
627 enum netdev_queue_state_t {
628 	__QUEUE_STATE_DRV_XOFF,
629 	__QUEUE_STATE_STACK_XOFF,
630 	__QUEUE_STATE_FROZEN,
631 };
632 
633 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
634 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
635 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
636 
637 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
638 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
639 					QUEUE_STATE_FROZEN)
640 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
641 					QUEUE_STATE_FROZEN)
642 
643 /*
644  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
645  * netif_tx_* functions below are used to manipulate this flag.  The
646  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
647  * queue independently.  The netif_xmit_*stopped functions below are called
648  * to check if the queue has been stopped by the driver or stack (either
649  * of the XOFF bits are set in the state).  Drivers should not need to call
650  * netif_xmit*stopped functions, they should only be using netif_tx_*.
651  */
652 
653 struct netdev_queue {
654 /*
655  * read-mostly part
656  */
657 	struct net_device	*dev;
658 	netdevice_tracker	dev_tracker;
659 
660 	struct Qdisc __rcu	*qdisc;
661 	struct Qdisc __rcu	*qdisc_sleeping;
662 #ifdef CONFIG_SYSFS
663 	struct kobject		kobj;
664 #endif
665 	unsigned long		tx_maxrate;
666 	/*
667 	 * Number of TX timeouts for this queue
668 	 * (/sys/class/net/DEV/Q/trans_timeout)
669 	 */
670 	atomic_long_t		trans_timeout;
671 
672 	/* Subordinate device that the queue has been assigned to */
673 	struct net_device	*sb_dev;
674 #ifdef CONFIG_XDP_SOCKETS
675 	struct xsk_buff_pool    *pool;
676 #endif
677 
678 /*
679  * write-mostly part
680  */
681 #ifdef CONFIG_BQL
682 	struct dql		dql;
683 #endif
684 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
685 	int			xmit_lock_owner;
686 	/*
687 	 * Time (in jiffies) of last Tx
688 	 */
689 	unsigned long		trans_start;
690 
691 	unsigned long		state;
692 
693 /*
694  * slow- / control-path part
695  */
696 	/* NAPI instance for the queue
697 	 * Readers and writers must hold RTNL
698 	 */
699 	struct napi_struct	*napi;
700 
701 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
702 	int			numa_node;
703 #endif
704 } ____cacheline_aligned_in_smp;
705 
706 extern int sysctl_fb_tunnels_only_for_init_net;
707 extern int sysctl_devconf_inherit_init_net;
708 
709 /*
710  * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
711  *                                     == 1 : For initns only
712  *                                     == 2 : For none.
713  */
714 static inline bool net_has_fallback_tunnels(const struct net *net)
715 {
716 #if IS_ENABLED(CONFIG_SYSCTL)
717 	int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
718 
719 	return !fb_tunnels_only_for_init_net ||
720 		(net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
721 #else
722 	return true;
723 #endif
724 }
725 
726 static inline int net_inherit_devconf(void)
727 {
728 #if IS_ENABLED(CONFIG_SYSCTL)
729 	return READ_ONCE(sysctl_devconf_inherit_init_net);
730 #else
731 	return 0;
732 #endif
733 }
734 
735 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
736 {
737 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
738 	return q->numa_node;
739 #else
740 	return NUMA_NO_NODE;
741 #endif
742 }
743 
744 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
745 {
746 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
747 	q->numa_node = node;
748 #endif
749 }
750 
751 #ifdef CONFIG_RFS_ACCEL
752 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
753 			 u16 filter_id);
754 #endif
755 
756 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
757 enum xps_map_type {
758 	XPS_CPUS = 0,
759 	XPS_RXQS,
760 	XPS_MAPS_MAX,
761 };
762 
763 #ifdef CONFIG_XPS
764 /*
765  * This structure holds an XPS map which can be of variable length.  The
766  * map is an array of queues.
767  */
768 struct xps_map {
769 	unsigned int len;
770 	unsigned int alloc_len;
771 	struct rcu_head rcu;
772 	u16 queues[];
773 };
774 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
775 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
776        - sizeof(struct xps_map)) / sizeof(u16))
777 
778 /*
779  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
780  *
781  * We keep track of the number of cpus/rxqs used when the struct is allocated,
782  * in nr_ids. This will help not accessing out-of-bound memory.
783  *
784  * We keep track of the number of traffic classes used when the struct is
785  * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
786  * not crossing its upper bound, as the original dev->num_tc can be updated in
787  * the meantime.
788  */
789 struct xps_dev_maps {
790 	struct rcu_head rcu;
791 	unsigned int nr_ids;
792 	s16 num_tc;
793 	struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
794 };
795 
796 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +	\
797 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
798 
799 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
800 	(_rxqs * (_tcs) * sizeof(struct xps_map *)))
801 
802 #endif /* CONFIG_XPS */
803 
804 #define TC_MAX_QUEUE	16
805 #define TC_BITMASK	15
806 /* HW offloaded queuing disciplines txq count and offset maps */
807 struct netdev_tc_txq {
808 	u16 count;
809 	u16 offset;
810 };
811 
812 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
813 /*
814  * This structure is to hold information about the device
815  * configured to run FCoE protocol stack.
816  */
817 struct netdev_fcoe_hbainfo {
818 	char	manufacturer[64];
819 	char	serial_number[64];
820 	char	hardware_version[64];
821 	char	driver_version[64];
822 	char	optionrom_version[64];
823 	char	firmware_version[64];
824 	char	model[256];
825 	char	model_description[256];
826 };
827 #endif
828 
829 #define MAX_PHYS_ITEM_ID_LEN 32
830 
831 /* This structure holds a unique identifier to identify some
832  * physical item (port for example) used by a netdevice.
833  */
834 struct netdev_phys_item_id {
835 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
836 	unsigned char id_len;
837 };
838 
839 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
840 					    struct netdev_phys_item_id *b)
841 {
842 	return a->id_len == b->id_len &&
843 	       memcmp(a->id, b->id, a->id_len) == 0;
844 }
845 
846 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
847 				       struct sk_buff *skb,
848 				       struct net_device *sb_dev);
849 
850 enum net_device_path_type {
851 	DEV_PATH_ETHERNET = 0,
852 	DEV_PATH_VLAN,
853 	DEV_PATH_BRIDGE,
854 	DEV_PATH_PPPOE,
855 	DEV_PATH_DSA,
856 	DEV_PATH_MTK_WDMA,
857 };
858 
859 struct net_device_path {
860 	enum net_device_path_type	type;
861 	const struct net_device		*dev;
862 	union {
863 		struct {
864 			u16		id;
865 			__be16		proto;
866 			u8		h_dest[ETH_ALEN];
867 		} encap;
868 		struct {
869 			enum {
870 				DEV_PATH_BR_VLAN_KEEP,
871 				DEV_PATH_BR_VLAN_TAG,
872 				DEV_PATH_BR_VLAN_UNTAG,
873 				DEV_PATH_BR_VLAN_UNTAG_HW,
874 			}		vlan_mode;
875 			u16		vlan_id;
876 			__be16		vlan_proto;
877 		} bridge;
878 		struct {
879 			int port;
880 			u16 proto;
881 		} dsa;
882 		struct {
883 			u8 wdma_idx;
884 			u8 queue;
885 			u16 wcid;
886 			u8 bss;
887 			u8 amsdu;
888 		} mtk_wdma;
889 	};
890 };
891 
892 #define NET_DEVICE_PATH_STACK_MAX	5
893 #define NET_DEVICE_PATH_VLAN_MAX	2
894 
895 struct net_device_path_stack {
896 	int			num_paths;
897 	struct net_device_path	path[NET_DEVICE_PATH_STACK_MAX];
898 };
899 
900 struct net_device_path_ctx {
901 	const struct net_device *dev;
902 	u8			daddr[ETH_ALEN];
903 
904 	int			num_vlans;
905 	struct {
906 		u16		id;
907 		__be16		proto;
908 	} vlan[NET_DEVICE_PATH_VLAN_MAX];
909 };
910 
911 enum tc_setup_type {
912 	TC_QUERY_CAPS,
913 	TC_SETUP_QDISC_MQPRIO,
914 	TC_SETUP_CLSU32,
915 	TC_SETUP_CLSFLOWER,
916 	TC_SETUP_CLSMATCHALL,
917 	TC_SETUP_CLSBPF,
918 	TC_SETUP_BLOCK,
919 	TC_SETUP_QDISC_CBS,
920 	TC_SETUP_QDISC_RED,
921 	TC_SETUP_QDISC_PRIO,
922 	TC_SETUP_QDISC_MQ,
923 	TC_SETUP_QDISC_ETF,
924 	TC_SETUP_ROOT_QDISC,
925 	TC_SETUP_QDISC_GRED,
926 	TC_SETUP_QDISC_TAPRIO,
927 	TC_SETUP_FT,
928 	TC_SETUP_QDISC_ETS,
929 	TC_SETUP_QDISC_TBF,
930 	TC_SETUP_QDISC_FIFO,
931 	TC_SETUP_QDISC_HTB,
932 	TC_SETUP_ACT,
933 };
934 
935 /* These structures hold the attributes of bpf state that are being passed
936  * to the netdevice through the bpf op.
937  */
938 enum bpf_netdev_command {
939 	/* Set or clear a bpf program used in the earliest stages of packet
940 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
941 	 * is responsible for calling bpf_prog_put on any old progs that are
942 	 * stored. In case of error, the callee need not release the new prog
943 	 * reference, but on success it takes ownership and must bpf_prog_put
944 	 * when it is no longer used.
945 	 */
946 	XDP_SETUP_PROG,
947 	XDP_SETUP_PROG_HW,
948 	/* BPF program for offload callbacks, invoked at program load time. */
949 	BPF_OFFLOAD_MAP_ALLOC,
950 	BPF_OFFLOAD_MAP_FREE,
951 	XDP_SETUP_XSK_POOL,
952 };
953 
954 struct bpf_prog_offload_ops;
955 struct netlink_ext_ack;
956 struct xdp_umem;
957 struct xdp_dev_bulk_queue;
958 struct bpf_xdp_link;
959 
960 enum bpf_xdp_mode {
961 	XDP_MODE_SKB = 0,
962 	XDP_MODE_DRV = 1,
963 	XDP_MODE_HW = 2,
964 	__MAX_XDP_MODE
965 };
966 
967 struct bpf_xdp_entity {
968 	struct bpf_prog *prog;
969 	struct bpf_xdp_link *link;
970 };
971 
972 struct netdev_bpf {
973 	enum bpf_netdev_command command;
974 	union {
975 		/* XDP_SETUP_PROG */
976 		struct {
977 			u32 flags;
978 			struct bpf_prog *prog;
979 			struct netlink_ext_ack *extack;
980 		};
981 		/* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
982 		struct {
983 			struct bpf_offloaded_map *offmap;
984 		};
985 		/* XDP_SETUP_XSK_POOL */
986 		struct {
987 			struct xsk_buff_pool *pool;
988 			u16 queue_id;
989 		} xsk;
990 	};
991 };
992 
993 /* Flags for ndo_xsk_wakeup. */
994 #define XDP_WAKEUP_RX (1 << 0)
995 #define XDP_WAKEUP_TX (1 << 1)
996 
997 #ifdef CONFIG_XFRM_OFFLOAD
998 struct xfrmdev_ops {
999 	int	(*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1000 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
1001 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
1002 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
1003 				       struct xfrm_state *x);
1004 	void	(*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1005 	void	(*xdo_dev_state_update_stats) (struct xfrm_state *x);
1006 	int	(*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1007 	void	(*xdo_dev_policy_delete) (struct xfrm_policy *x);
1008 	void	(*xdo_dev_policy_free) (struct xfrm_policy *x);
1009 };
1010 #endif
1011 
1012 struct dev_ifalias {
1013 	struct rcu_head rcuhead;
1014 	char ifalias[];
1015 };
1016 
1017 struct devlink;
1018 struct tlsdev_ops;
1019 
1020 struct netdev_net_notifier {
1021 	struct list_head list;
1022 	struct notifier_block *nb;
1023 };
1024 
1025 /*
1026  * This structure defines the management hooks for network devices.
1027  * The following hooks can be defined; unless noted otherwise, they are
1028  * optional and can be filled with a null pointer.
1029  *
1030  * int (*ndo_init)(struct net_device *dev);
1031  *     This function is called once when a network device is registered.
1032  *     The network device can use this for any late stage initialization
1033  *     or semantic validation. It can fail with an error code which will
1034  *     be propagated back to register_netdev.
1035  *
1036  * void (*ndo_uninit)(struct net_device *dev);
1037  *     This function is called when device is unregistered or when registration
1038  *     fails. It is not called if init fails.
1039  *
1040  * int (*ndo_open)(struct net_device *dev);
1041  *     This function is called when a network device transitions to the up
1042  *     state.
1043  *
1044  * int (*ndo_stop)(struct net_device *dev);
1045  *     This function is called when a network device transitions to the down
1046  *     state.
1047  *
1048  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1049  *                               struct net_device *dev);
1050  *	Called when a packet needs to be transmitted.
1051  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
1052  *	the queue before that can happen; it's for obsolete devices and weird
1053  *	corner cases, but the stack really does a non-trivial amount
1054  *	of useless work if you return NETDEV_TX_BUSY.
1055  *	Required; cannot be NULL.
1056  *
1057  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1058  *					   struct net_device *dev
1059  *					   netdev_features_t features);
1060  *	Called by core transmit path to determine if device is capable of
1061  *	performing offload operations on a given packet. This is to give
1062  *	the device an opportunity to implement any restrictions that cannot
1063  *	be otherwise expressed by feature flags. The check is called with
1064  *	the set of features that the stack has calculated and it returns
1065  *	those the driver believes to be appropriate.
1066  *
1067  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1068  *                         struct net_device *sb_dev);
1069  *	Called to decide which queue to use when device supports multiple
1070  *	transmit queues.
1071  *
1072  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1073  *	This function is called to allow device receiver to make
1074  *	changes to configuration when multicast or promiscuous is enabled.
1075  *
1076  * void (*ndo_set_rx_mode)(struct net_device *dev);
1077  *	This function is called device changes address list filtering.
1078  *	If driver handles unicast address filtering, it should set
1079  *	IFF_UNICAST_FLT in its priv_flags.
1080  *
1081  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1082  *	This function  is called when the Media Access Control address
1083  *	needs to be changed. If this interface is not defined, the
1084  *	MAC address can not be changed.
1085  *
1086  * int (*ndo_validate_addr)(struct net_device *dev);
1087  *	Test if Media Access Control address is valid for the device.
1088  *
1089  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1090  *	Old-style ioctl entry point. This is used internally by the
1091  *	appletalk and ieee802154 subsystems but is no longer called by
1092  *	the device ioctl handler.
1093  *
1094  * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1095  *	Used by the bonding driver for its device specific ioctls:
1096  *	SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1097  *	SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1098  *
1099  * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1100  *	Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1101  *	SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1102  *
1103  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1104  *	Used to set network devices bus interface parameters. This interface
1105  *	is retained for legacy reasons; new devices should use the bus
1106  *	interface (PCI) for low level management.
1107  *
1108  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1109  *	Called when a user wants to change the Maximum Transfer Unit
1110  *	of a device.
1111  *
1112  * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1113  *	Callback used when the transmitter has not made any progress
1114  *	for dev->watchdog ticks.
1115  *
1116  * void (*ndo_get_stats64)(struct net_device *dev,
1117  *                         struct rtnl_link_stats64 *storage);
1118  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1119  *	Called when a user wants to get the network device usage
1120  *	statistics. Drivers must do one of the following:
1121  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
1122  *	   rtnl_link_stats64 structure passed by the caller.
1123  *	2. Define @ndo_get_stats to update a net_device_stats structure
1124  *	   (which should normally be dev->stats) and return a pointer to
1125  *	   it. The structure may be changed asynchronously only if each
1126  *	   field is written atomically.
1127  *	3. Update dev->stats asynchronously and atomically, and define
1128  *	   neither operation.
1129  *
1130  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1131  *	Return true if this device supports offload stats of this attr_id.
1132  *
1133  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1134  *	void *attr_data)
1135  *	Get statistics for offload operations by attr_id. Write it into the
1136  *	attr_data pointer.
1137  *
1138  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1139  *	If device supports VLAN filtering this function is called when a
1140  *	VLAN id is registered.
1141  *
1142  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1143  *	If device supports VLAN filtering this function is called when a
1144  *	VLAN id is unregistered.
1145  *
1146  * void (*ndo_poll_controller)(struct net_device *dev);
1147  *
1148  *	SR-IOV management functions.
1149  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1150  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1151  *			  u8 qos, __be16 proto);
1152  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1153  *			  int max_tx_rate);
1154  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1155  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1156  * int (*ndo_get_vf_config)(struct net_device *dev,
1157  *			    int vf, struct ifla_vf_info *ivf);
1158  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1159  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1160  *			  struct nlattr *port[]);
1161  *
1162  *      Enable or disable the VF ability to query its RSS Redirection Table and
1163  *      Hash Key. This is needed since on some devices VF share this information
1164  *      with PF and querying it may introduce a theoretical security risk.
1165  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1166  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1167  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1168  *		       void *type_data);
1169  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
1170  *	This is always called from the stack with the rtnl lock held and netif
1171  *	tx queues stopped. This allows the netdevice to perform queue
1172  *	management safely.
1173  *
1174  *	Fiber Channel over Ethernet (FCoE) offload functions.
1175  * int (*ndo_fcoe_enable)(struct net_device *dev);
1176  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
1177  *	so the underlying device can perform whatever needed configuration or
1178  *	initialization to support acceleration of FCoE traffic.
1179  *
1180  * int (*ndo_fcoe_disable)(struct net_device *dev);
1181  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
1182  *	so the underlying device can perform whatever needed clean-ups to
1183  *	stop supporting acceleration of FCoE traffic.
1184  *
1185  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1186  *			     struct scatterlist *sgl, unsigned int sgc);
1187  *	Called when the FCoE Initiator wants to initialize an I/O that
1188  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1189  *	perform necessary setup and returns 1 to indicate the device is set up
1190  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1191  *
1192  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1193  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1194  *	indicated by the FC exchange id 'xid', so the underlying device can
1195  *	clean up and reuse resources for later DDP requests.
1196  *
1197  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1198  *			      struct scatterlist *sgl, unsigned int sgc);
1199  *	Called when the FCoE Target wants to initialize an I/O that
1200  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1201  *	perform necessary setup and returns 1 to indicate the device is set up
1202  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1203  *
1204  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1205  *			       struct netdev_fcoe_hbainfo *hbainfo);
1206  *	Called when the FCoE Protocol stack wants information on the underlying
1207  *	device. This information is utilized by the FCoE protocol stack to
1208  *	register attributes with Fiber Channel management service as per the
1209  *	FC-GS Fabric Device Management Information(FDMI) specification.
1210  *
1211  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1212  *	Called when the underlying device wants to override default World Wide
1213  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1214  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1215  *	protocol stack to use.
1216  *
1217  *	RFS acceleration.
1218  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1219  *			    u16 rxq_index, u32 flow_id);
1220  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1221  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1222  *	Return the filter ID on success, or a negative error code.
1223  *
1224  *	Slave management functions (for bridge, bonding, etc).
1225  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1226  *	Called to make another netdev an underling.
1227  *
1228  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1229  *	Called to release previously enslaved netdev.
1230  *
1231  * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1232  *					    struct sk_buff *skb,
1233  *					    bool all_slaves);
1234  *	Get the xmit slave of master device. If all_slaves is true, function
1235  *	assume all the slaves can transmit.
1236  *
1237  *      Feature/offload setting functions.
1238  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1239  *		netdev_features_t features);
1240  *	Adjusts the requested feature flags according to device-specific
1241  *	constraints, and returns the resulting flags. Must not modify
1242  *	the device state.
1243  *
1244  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1245  *	Called to update device configuration to new features. Passed
1246  *	feature set might be less than what was returned by ndo_fix_features()).
1247  *	Must return >0 or -errno if it changed dev->features itself.
1248  *
1249  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1250  *		      struct net_device *dev,
1251  *		      const unsigned char *addr, u16 vid, u16 flags,
1252  *		      bool *notified, struct netlink_ext_ack *extack);
1253  *	Adds an FDB entry to dev for addr.
1254  *	Callee shall set *notified to true if it sent any appropriate
1255  *	notification(s). Otherwise core will send a generic one.
1256  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1257  *		      struct net_device *dev,
1258  *		      const unsigned char *addr, u16 vid
1259  *		      bool *notified, struct netlink_ext_ack *extack);
1260  *	Deletes the FDB entry from dev corresponding to addr.
1261  *	Callee shall set *notified to true if it sent any appropriate
1262  *	notification(s). Otherwise core will send a generic one.
1263  * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1264  *			   struct netlink_ext_ack *extack);
1265  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1266  *		       struct net_device *dev, struct net_device *filter_dev,
1267  *		       int *idx)
1268  *	Used to add FDB entries to dump requests. Implementers should add
1269  *	entries to skb and update idx with the number of entries.
1270  *
1271  * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1272  *		      u16 nlmsg_flags, struct netlink_ext_ack *extack);
1273  *	Adds an MDB entry to dev.
1274  * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1275  *		      struct netlink_ext_ack *extack);
1276  *	Deletes the MDB entry from dev.
1277  * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1278  *			   struct netlink_ext_ack *extack);
1279  *	Bulk deletes MDB entries from dev.
1280  * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1281  *		       struct netlink_callback *cb);
1282  *	Dumps MDB entries from dev. The first argument (marker) in the netlink
1283  *	callback is used by core rtnetlink code.
1284  *
1285  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1286  *			     u16 flags, struct netlink_ext_ack *extack)
1287  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1288  *			     struct net_device *dev, u32 filter_mask,
1289  *			     int nlflags)
1290  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1291  *			     u16 flags);
1292  *
1293  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1294  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1295  *	which do not represent real hardware may define this to allow their
1296  *	userspace components to manage their virtual carrier state. Devices
1297  *	that determine carrier state from physical hardware properties (eg
1298  *	network cables) or protocol-dependent mechanisms (eg
1299  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1300  *
1301  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1302  *			       struct netdev_phys_item_id *ppid);
1303  *	Called to get ID of physical port of this device. If driver does
1304  *	not implement this, it is assumed that the hw is not able to have
1305  *	multiple net devices on single physical port.
1306  *
1307  * int (*ndo_get_port_parent_id)(struct net_device *dev,
1308  *				 struct netdev_phys_item_id *ppid)
1309  *	Called to get the parent ID of the physical port of this device.
1310  *
1311  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1312  *				 struct net_device *dev)
1313  *	Called by upper layer devices to accelerate switching or other
1314  *	station functionality into hardware. 'pdev is the lowerdev
1315  *	to use for the offload and 'dev' is the net device that will
1316  *	back the offload. Returns a pointer to the private structure
1317  *	the upper layer will maintain.
1318  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1319  *	Called by upper layer device to delete the station created
1320  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1321  *	the station and priv is the structure returned by the add
1322  *	operation.
1323  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1324  *			     int queue_index, u32 maxrate);
1325  *	Called when a user wants to set a max-rate limitation of specific
1326  *	TX queue.
1327  * int (*ndo_get_iflink)(const struct net_device *dev);
1328  *	Called to get the iflink value of this device.
1329  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1330  *	This function is used to get egress tunnel information for given skb.
1331  *	This is useful for retrieving outer tunnel header parameters while
1332  *	sampling packet.
1333  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1334  *	This function is used to specify the headroom that the skb must
1335  *	consider when allocation skb during packet reception. Setting
1336  *	appropriate rx headroom value allows avoiding skb head copy on
1337  *	forward. Setting a negative value resets the rx headroom to the
1338  *	default value.
1339  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1340  *	This function is used to set or query state related to XDP on the
1341  *	netdevice and manage BPF offload. See definition of
1342  *	enum bpf_netdev_command for details.
1343  * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1344  *			u32 flags);
1345  *	This function is used to submit @n XDP packets for transmit on a
1346  *	netdevice. Returns number of frames successfully transmitted, frames
1347  *	that got dropped are freed/returned via xdp_return_frame().
1348  *	Returns negative number, means general error invoking ndo, meaning
1349  *	no frames were xmit'ed and core-caller will free all frames.
1350  * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1351  *					        struct xdp_buff *xdp);
1352  *      Get the xmit slave of master device based on the xdp_buff.
1353  * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1354  *      This function is used to wake up the softirq, ksoftirqd or kthread
1355  *	responsible for sending and/or receiving packets on a specific
1356  *	queue id bound to an AF_XDP socket. The flags field specifies if
1357  *	only RX, only Tx, or both should be woken up using the flags
1358  *	XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1359  * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1360  *			 int cmd);
1361  *	Add, change, delete or get information on an IPv4 tunnel.
1362  * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1363  *	If a device is paired with a peer device, return the peer instance.
1364  *	The caller must be under RCU read context.
1365  * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1366  *     Get the forwarding path to reach the real device from the HW destination address
1367  * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1368  *			     const struct skb_shared_hwtstamps *hwtstamps,
1369  *			     bool cycles);
1370  *	Get hardware timestamp based on normal/adjustable time or free running
1371  *	cycle counter. This function is required if physical clock supports a
1372  *	free running cycle counter.
1373  *
1374  * int (*ndo_hwtstamp_get)(struct net_device *dev,
1375  *			   struct kernel_hwtstamp_config *kernel_config);
1376  *	Get the currently configured hardware timestamping parameters for the
1377  *	NIC device.
1378  *
1379  * int (*ndo_hwtstamp_set)(struct net_device *dev,
1380  *			   struct kernel_hwtstamp_config *kernel_config,
1381  *			   struct netlink_ext_ack *extack);
1382  *	Change the hardware timestamping parameters for NIC device.
1383  */
1384 struct net_device_ops {
1385 	int			(*ndo_init)(struct net_device *dev);
1386 	void			(*ndo_uninit)(struct net_device *dev);
1387 	int			(*ndo_open)(struct net_device *dev);
1388 	int			(*ndo_stop)(struct net_device *dev);
1389 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1390 						  struct net_device *dev);
1391 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1392 						      struct net_device *dev,
1393 						      netdev_features_t features);
1394 	u16			(*ndo_select_queue)(struct net_device *dev,
1395 						    struct sk_buff *skb,
1396 						    struct net_device *sb_dev);
1397 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1398 						       int flags);
1399 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1400 	int			(*ndo_set_mac_address)(struct net_device *dev,
1401 						       void *addr);
1402 	int			(*ndo_validate_addr)(struct net_device *dev);
1403 	int			(*ndo_do_ioctl)(struct net_device *dev,
1404 					        struct ifreq *ifr, int cmd);
1405 	int			(*ndo_eth_ioctl)(struct net_device *dev,
1406 						 struct ifreq *ifr, int cmd);
1407 	int			(*ndo_siocbond)(struct net_device *dev,
1408 						struct ifreq *ifr, int cmd);
1409 	int			(*ndo_siocwandev)(struct net_device *dev,
1410 						  struct if_settings *ifs);
1411 	int			(*ndo_siocdevprivate)(struct net_device *dev,
1412 						      struct ifreq *ifr,
1413 						      void __user *data, int cmd);
1414 	int			(*ndo_set_config)(struct net_device *dev,
1415 					          struct ifmap *map);
1416 	int			(*ndo_change_mtu)(struct net_device *dev,
1417 						  int new_mtu);
1418 	int			(*ndo_neigh_setup)(struct net_device *dev,
1419 						   struct neigh_parms *);
1420 	void			(*ndo_tx_timeout) (struct net_device *dev,
1421 						   unsigned int txqueue);
1422 
1423 	void			(*ndo_get_stats64)(struct net_device *dev,
1424 						   struct rtnl_link_stats64 *storage);
1425 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1426 	int			(*ndo_get_offload_stats)(int attr_id,
1427 							 const struct net_device *dev,
1428 							 void *attr_data);
1429 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1430 
1431 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1432 						       __be16 proto, u16 vid);
1433 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1434 						        __be16 proto, u16 vid);
1435 #ifdef CONFIG_NET_POLL_CONTROLLER
1436 	void                    (*ndo_poll_controller)(struct net_device *dev);
1437 	int			(*ndo_netpoll_setup)(struct net_device *dev);
1438 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1439 #endif
1440 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1441 						  int queue, u8 *mac);
1442 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1443 						   int queue, u16 vlan,
1444 						   u8 qos, __be16 proto);
1445 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1446 						   int vf, int min_tx_rate,
1447 						   int max_tx_rate);
1448 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1449 						       int vf, bool setting);
1450 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1451 						    int vf, bool setting);
1452 	int			(*ndo_get_vf_config)(struct net_device *dev,
1453 						     int vf,
1454 						     struct ifla_vf_info *ivf);
1455 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1456 							 int vf, int link_state);
1457 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1458 						    int vf,
1459 						    struct ifla_vf_stats
1460 						    *vf_stats);
1461 	int			(*ndo_set_vf_port)(struct net_device *dev,
1462 						   int vf,
1463 						   struct nlattr *port[]);
1464 	int			(*ndo_get_vf_port)(struct net_device *dev,
1465 						   int vf, struct sk_buff *skb);
1466 	int			(*ndo_get_vf_guid)(struct net_device *dev,
1467 						   int vf,
1468 						   struct ifla_vf_guid *node_guid,
1469 						   struct ifla_vf_guid *port_guid);
1470 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1471 						   int vf, u64 guid,
1472 						   int guid_type);
1473 	int			(*ndo_set_vf_rss_query_en)(
1474 						   struct net_device *dev,
1475 						   int vf, bool setting);
1476 	int			(*ndo_setup_tc)(struct net_device *dev,
1477 						enum tc_setup_type type,
1478 						void *type_data);
1479 #if IS_ENABLED(CONFIG_FCOE)
1480 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1481 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1482 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1483 						      u16 xid,
1484 						      struct scatterlist *sgl,
1485 						      unsigned int sgc);
1486 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1487 						     u16 xid);
1488 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1489 						       u16 xid,
1490 						       struct scatterlist *sgl,
1491 						       unsigned int sgc);
1492 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1493 							struct netdev_fcoe_hbainfo *hbainfo);
1494 #endif
1495 
1496 #if IS_ENABLED(CONFIG_LIBFCOE)
1497 #define NETDEV_FCOE_WWNN 0
1498 #define NETDEV_FCOE_WWPN 1
1499 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1500 						    u64 *wwn, int type);
1501 #endif
1502 
1503 #ifdef CONFIG_RFS_ACCEL
1504 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1505 						     const struct sk_buff *skb,
1506 						     u16 rxq_index,
1507 						     u32 flow_id);
1508 #endif
1509 	int			(*ndo_add_slave)(struct net_device *dev,
1510 						 struct net_device *slave_dev,
1511 						 struct netlink_ext_ack *extack);
1512 	int			(*ndo_del_slave)(struct net_device *dev,
1513 						 struct net_device *slave_dev);
1514 	struct net_device*	(*ndo_get_xmit_slave)(struct net_device *dev,
1515 						      struct sk_buff *skb,
1516 						      bool all_slaves);
1517 	struct net_device*	(*ndo_sk_get_lower_dev)(struct net_device *dev,
1518 							struct sock *sk);
1519 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1520 						    netdev_features_t features);
1521 	int			(*ndo_set_features)(struct net_device *dev,
1522 						    netdev_features_t features);
1523 	int			(*ndo_neigh_construct)(struct net_device *dev,
1524 						       struct neighbour *n);
1525 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1526 						     struct neighbour *n);
1527 
1528 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1529 					       struct nlattr *tb[],
1530 					       struct net_device *dev,
1531 					       const unsigned char *addr,
1532 					       u16 vid,
1533 					       u16 flags,
1534 					       bool *notified,
1535 					       struct netlink_ext_ack *extack);
1536 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1537 					       struct nlattr *tb[],
1538 					       struct net_device *dev,
1539 					       const unsigned char *addr,
1540 					       u16 vid,
1541 					       bool *notified,
1542 					       struct netlink_ext_ack *extack);
1543 	int			(*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1544 						    struct net_device *dev,
1545 						    struct netlink_ext_ack *extack);
1546 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1547 						struct netlink_callback *cb,
1548 						struct net_device *dev,
1549 						struct net_device *filter_dev,
1550 						int *idx);
1551 	int			(*ndo_fdb_get)(struct sk_buff *skb,
1552 					       struct nlattr *tb[],
1553 					       struct net_device *dev,
1554 					       const unsigned char *addr,
1555 					       u16 vid, u32 portid, u32 seq,
1556 					       struct netlink_ext_ack *extack);
1557 	int			(*ndo_mdb_add)(struct net_device *dev,
1558 					       struct nlattr *tb[],
1559 					       u16 nlmsg_flags,
1560 					       struct netlink_ext_ack *extack);
1561 	int			(*ndo_mdb_del)(struct net_device *dev,
1562 					       struct nlattr *tb[],
1563 					       struct netlink_ext_ack *extack);
1564 	int			(*ndo_mdb_del_bulk)(struct net_device *dev,
1565 						    struct nlattr *tb[],
1566 						    struct netlink_ext_ack *extack);
1567 	int			(*ndo_mdb_dump)(struct net_device *dev,
1568 						struct sk_buff *skb,
1569 						struct netlink_callback *cb);
1570 	int			(*ndo_mdb_get)(struct net_device *dev,
1571 					       struct nlattr *tb[], u32 portid,
1572 					       u32 seq,
1573 					       struct netlink_ext_ack *extack);
1574 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1575 						      struct nlmsghdr *nlh,
1576 						      u16 flags,
1577 						      struct netlink_ext_ack *extack);
1578 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1579 						      u32 pid, u32 seq,
1580 						      struct net_device *dev,
1581 						      u32 filter_mask,
1582 						      int nlflags);
1583 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1584 						      struct nlmsghdr *nlh,
1585 						      u16 flags);
1586 	int			(*ndo_change_carrier)(struct net_device *dev,
1587 						      bool new_carrier);
1588 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1589 							struct netdev_phys_item_id *ppid);
1590 	int			(*ndo_get_port_parent_id)(struct net_device *dev,
1591 							  struct netdev_phys_item_id *ppid);
1592 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1593 							  char *name, size_t len);
1594 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1595 							struct net_device *dev);
1596 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1597 							void *priv);
1598 
1599 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1600 						      int queue_index,
1601 						      u32 maxrate);
1602 	int			(*ndo_get_iflink)(const struct net_device *dev);
1603 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1604 						       struct sk_buff *skb);
1605 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1606 						       int needed_headroom);
1607 	int			(*ndo_bpf)(struct net_device *dev,
1608 					   struct netdev_bpf *bpf);
1609 	int			(*ndo_xdp_xmit)(struct net_device *dev, int n,
1610 						struct xdp_frame **xdp,
1611 						u32 flags);
1612 	struct net_device *	(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1613 							  struct xdp_buff *xdp);
1614 	int			(*ndo_xsk_wakeup)(struct net_device *dev,
1615 						  u32 queue_id, u32 flags);
1616 	int			(*ndo_tunnel_ctl)(struct net_device *dev,
1617 						  struct ip_tunnel_parm_kern *p,
1618 						  int cmd);
1619 	struct net_device *	(*ndo_get_peer_dev)(struct net_device *dev);
1620 	int                     (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1621                                                          struct net_device_path *path);
1622 	ktime_t			(*ndo_get_tstamp)(struct net_device *dev,
1623 						  const struct skb_shared_hwtstamps *hwtstamps,
1624 						  bool cycles);
1625 	int			(*ndo_hwtstamp_get)(struct net_device *dev,
1626 						    struct kernel_hwtstamp_config *kernel_config);
1627 	int			(*ndo_hwtstamp_set)(struct net_device *dev,
1628 						    struct kernel_hwtstamp_config *kernel_config,
1629 						    struct netlink_ext_ack *extack);
1630 
1631 #if IS_ENABLED(CONFIG_NET_SHAPER)
1632 	/**
1633 	 * @net_shaper_ops: Device shaping offload operations
1634 	 * see include/net/net_shapers.h
1635 	 */
1636 	const struct net_shaper_ops *net_shaper_ops;
1637 #endif
1638 };
1639 
1640 /**
1641  * enum netdev_priv_flags - &struct net_device priv_flags
1642  *
1643  * These are the &struct net_device, they are only set internally
1644  * by drivers and used in the kernel. These flags are invisible to
1645  * userspace; this means that the order of these flags can change
1646  * during any kernel release.
1647  *
1648  * You should add bitfield booleans after either net_device::priv_flags
1649  * (hotpath) or ::threaded (slowpath) instead of extending these flags.
1650  *
1651  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1652  * @IFF_EBRIDGE: Ethernet bridging device
1653  * @IFF_BONDING: bonding master or slave
1654  * @IFF_ISATAP: ISATAP interface (RFC4214)
1655  * @IFF_WAN_HDLC: WAN HDLC device
1656  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1657  *	release skb->dst
1658  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1659  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1660  * @IFF_MACVLAN_PORT: device used as macvlan port
1661  * @IFF_BRIDGE_PORT: device used as bridge port
1662  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1663  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1664  * @IFF_UNICAST_FLT: Supports unicast filtering
1665  * @IFF_TEAM_PORT: device used as team port
1666  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1667  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1668  *	change when it's running
1669  * @IFF_MACVLAN: Macvlan device
1670  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1671  *	underlying stacked devices
1672  * @IFF_L3MDEV_MASTER: device is an L3 master device
1673  * @IFF_NO_QUEUE: device can run without qdisc attached
1674  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1675  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1676  * @IFF_TEAM: device is a team device
1677  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1678  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1679  *	entity (i.e. the master device for bridged veth)
1680  * @IFF_MACSEC: device is a MACsec device
1681  * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1682  * @IFF_FAILOVER: device is a failover master device
1683  * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1684  * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1685  * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1686  * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1687  *	skb_headlen(skb) == 0 (data starts from frag0)
1688  */
1689 enum netdev_priv_flags {
1690 	IFF_802_1Q_VLAN			= 1<<0,
1691 	IFF_EBRIDGE			= 1<<1,
1692 	IFF_BONDING			= 1<<2,
1693 	IFF_ISATAP			= 1<<3,
1694 	IFF_WAN_HDLC			= 1<<4,
1695 	IFF_XMIT_DST_RELEASE		= 1<<5,
1696 	IFF_DONT_BRIDGE			= 1<<6,
1697 	IFF_DISABLE_NETPOLL		= 1<<7,
1698 	IFF_MACVLAN_PORT		= 1<<8,
1699 	IFF_BRIDGE_PORT			= 1<<9,
1700 	IFF_OVS_DATAPATH		= 1<<10,
1701 	IFF_TX_SKB_SHARING		= 1<<11,
1702 	IFF_UNICAST_FLT			= 1<<12,
1703 	IFF_TEAM_PORT			= 1<<13,
1704 	IFF_SUPP_NOFCS			= 1<<14,
1705 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1706 	IFF_MACVLAN			= 1<<16,
1707 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1708 	IFF_L3MDEV_MASTER		= 1<<18,
1709 	IFF_NO_QUEUE			= 1<<19,
1710 	IFF_OPENVSWITCH			= 1<<20,
1711 	IFF_L3MDEV_SLAVE		= 1<<21,
1712 	IFF_TEAM			= 1<<22,
1713 	IFF_RXFH_CONFIGURED		= 1<<23,
1714 	IFF_PHONY_HEADROOM		= 1<<24,
1715 	IFF_MACSEC			= 1<<25,
1716 	IFF_NO_RX_HANDLER		= 1<<26,
1717 	IFF_FAILOVER			= 1<<27,
1718 	IFF_FAILOVER_SLAVE		= 1<<28,
1719 	IFF_L3MDEV_RX_HANDLER		= 1<<29,
1720 	IFF_NO_ADDRCONF			= BIT_ULL(30),
1721 	IFF_TX_SKB_NO_LINEAR		= BIT_ULL(31),
1722 };
1723 
1724 /* Specifies the type of the struct net_device::ml_priv pointer */
1725 enum netdev_ml_priv_type {
1726 	ML_PRIV_NONE,
1727 	ML_PRIV_CAN,
1728 };
1729 
1730 enum netdev_stat_type {
1731 	NETDEV_PCPU_STAT_NONE,
1732 	NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1733 	NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1734 	NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1735 };
1736 
1737 enum netdev_reg_state {
1738 	NETREG_UNINITIALIZED = 0,
1739 	NETREG_REGISTERED,	/* completed register_netdevice */
1740 	NETREG_UNREGISTERING,	/* called unregister_netdevice */
1741 	NETREG_UNREGISTERED,	/* completed unregister todo */
1742 	NETREG_RELEASED,	/* called free_netdev */
1743 	NETREG_DUMMY,		/* dummy device for NAPI poll */
1744 };
1745 
1746 /**
1747  *	struct net_device - The DEVICE structure.
1748  *
1749  *	Actually, this whole structure is a big mistake.  It mixes I/O
1750  *	data with strictly "high-level" data, and it has to know about
1751  *	almost every data structure used in the INET module.
1752  *
1753  *	@priv_flags:	flags invisible to userspace defined as bits, see
1754  *			enum netdev_priv_flags for the definitions
1755  *	@lltx:		device supports lockless Tx. Deprecated for real HW
1756  *			drivers. Mainly used by logical interfaces, such as
1757  *			bonding and tunnels
1758  *
1759  *	@name:	This is the first field of the "visible" part of this structure
1760  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1761  *		of the interface.
1762  *
1763  *	@name_node:	Name hashlist node
1764  *	@ifalias:	SNMP alias
1765  *	@mem_end:	Shared memory end
1766  *	@mem_start:	Shared memory start
1767  *	@base_addr:	Device I/O address
1768  *	@irq:		Device IRQ number
1769  *
1770  *	@state:		Generic network queuing layer state, see netdev_state_t
1771  *	@dev_list:	The global list of network devices
1772  *	@napi_list:	List entry used for polling NAPI devices
1773  *	@unreg_list:	List entry  when we are unregistering the
1774  *			device; see the function unregister_netdev
1775  *	@close_list:	List entry used when we are closing the device
1776  *	@ptype_all:     Device-specific packet handlers for all protocols
1777  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1778  *
1779  *	@adj_list:	Directly linked devices, like slaves for bonding
1780  *	@features:	Currently active device features
1781  *	@hw_features:	User-changeable features
1782  *
1783  *	@wanted_features:	User-requested features
1784  *	@vlan_features:		Mask of features inheritable by VLAN devices
1785  *
1786  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1787  *				This field indicates what encapsulation
1788  *				offloads the hardware is capable of doing,
1789  *				and drivers will need to set them appropriately.
1790  *
1791  *	@mpls_features:	Mask of features inheritable by MPLS
1792  *	@gso_partial_features: value(s) from NETIF_F_GSO\*
1793  *
1794  *	@ifindex:	interface index
1795  *	@group:		The group the device belongs to
1796  *
1797  *	@stats:		Statistics struct, which was left as a legacy, use
1798  *			rtnl_link_stats64 instead
1799  *
1800  *	@core_stats:	core networking counters,
1801  *			do not use this in drivers
1802  *	@carrier_up_count:	Number of times the carrier has been up
1803  *	@carrier_down_count:	Number of times the carrier has been down
1804  *
1805  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1806  *				instead of ioctl,
1807  *				see <net/iw_handler.h> for details.
1808  *
1809  *	@netdev_ops:	Includes several pointers to callbacks,
1810  *			if one wants to override the ndo_*() functions
1811  *	@xdp_metadata_ops:	Includes pointers to XDP metadata callbacks.
1812  *	@xsk_tx_metadata_ops:	Includes pointers to AF_XDP TX metadata callbacks.
1813  *	@ethtool_ops:	Management operations
1814  *	@l3mdev_ops:	Layer 3 master device operations
1815  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1816  *			discovery handling. Necessary for e.g. 6LoWPAN.
1817  *	@xfrmdev_ops:	Transformation offload operations
1818  *	@tlsdev_ops:	Transport Layer Security offload operations
1819  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1820  *			of Layer 2 headers.
1821  *
1822  *	@flags:		Interface flags (a la BSD)
1823  *	@xdp_features:	XDP capability supported by the device
1824  *	@gflags:	Global flags ( kept as legacy )
1825  *	@priv_len:	Size of the ->priv flexible array
1826  *	@priv:		Flexible array containing private data
1827  *	@operstate:	RFC2863 operstate
1828  *	@link_mode:	Mapping policy to operstate
1829  *	@if_port:	Selectable AUI, TP, ...
1830  *	@dma:		DMA channel
1831  *	@mtu:		Interface MTU value
1832  *	@min_mtu:	Interface Minimum MTU value
1833  *	@max_mtu:	Interface Maximum MTU value
1834  *	@type:		Interface hardware type
1835  *	@hard_header_len: Maximum hardware header length.
1836  *	@min_header_len:  Minimum hardware header length
1837  *
1838  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1839  *			  cases can this be guaranteed
1840  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1841  *			  cases can this be guaranteed. Some cases also use
1842  *			  LL_MAX_HEADER instead to allocate the skb
1843  *
1844  *	interface address info:
1845  *
1846  * 	@perm_addr:		Permanent hw address
1847  * 	@addr_assign_type:	Hw address assignment type
1848  * 	@addr_len:		Hardware address length
1849  *	@upper_level:		Maximum depth level of upper devices.
1850  *	@lower_level:		Maximum depth level of lower devices.
1851  *	@neigh_priv_len:	Used in neigh_alloc()
1852  * 	@dev_id:		Used to differentiate devices that share
1853  * 				the same link layer address
1854  * 	@dev_port:		Used to differentiate devices that share
1855  * 				the same function
1856  *	@addr_list_lock:	XXX: need comments on this one
1857  *	@name_assign_type:	network interface name assignment type
1858  *	@uc_promisc:		Counter that indicates promiscuous mode
1859  *				has been enabled due to the need to listen to
1860  *				additional unicast addresses in a device that
1861  *				does not implement ndo_set_rx_mode()
1862  *	@uc:			unicast mac addresses
1863  *	@mc:			multicast mac addresses
1864  *	@dev_addrs:		list of device hw addresses
1865  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1866  *	@promiscuity:		Number of times the NIC is told to work in
1867  *				promiscuous mode; if it becomes 0 the NIC will
1868  *				exit promiscuous mode
1869  *	@allmulti:		Counter, enables or disables allmulticast mode
1870  *
1871  *	@vlan_info:	VLAN info
1872  *	@dsa_ptr:	dsa specific data
1873  *	@tipc_ptr:	TIPC specific data
1874  *	@atalk_ptr:	AppleTalk link
1875  *	@ip_ptr:	IPv4 specific data
1876  *	@ip6_ptr:	IPv6 specific data
1877  *	@ax25_ptr:	AX.25 specific data
1878  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1879  *	@ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1880  *			 device struct
1881  *	@mpls_ptr:	mpls_dev struct pointer
1882  *	@mctp_ptr:	MCTP specific data
1883  *
1884  *	@dev_addr:	Hw address (before bcast,
1885  *			because most packets are unicast)
1886  *
1887  *	@_rx:			Array of RX queues
1888  *	@num_rx_queues:		Number of RX queues
1889  *				allocated at register_netdev() time
1890  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1891  *	@xdp_prog:		XDP sockets filter program pointer
1892  *
1893  *	@rx_handler:		handler for received packets
1894  *	@rx_handler_data: 	XXX: need comments on this one
1895  *	@tcx_ingress:		BPF & clsact qdisc specific data for ingress processing
1896  *	@ingress_queue:		XXX: need comments on this one
1897  *	@nf_hooks_ingress:	netfilter hooks executed for ingress packets
1898  *	@broadcast:		hw bcast address
1899  *
1900  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1901  *			indexed by RX queue number. Assigned by driver.
1902  *			This must only be set if the ndo_rx_flow_steer
1903  *			operation is defined
1904  *	@index_hlist:		Device index hash chain
1905  *
1906  *	@_tx:			Array of TX queues
1907  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1908  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1909  *	@qdisc:			Root qdisc from userspace point of view
1910  *	@tx_queue_len:		Max frames per queue allowed
1911  *	@tx_global_lock: 	XXX: need comments on this one
1912  *	@xdp_bulkq:		XDP device bulk queue
1913  *	@xps_maps:		all CPUs/RXQs maps for XPS device
1914  *
1915  *	@xps_maps:	XXX: need comments on this one
1916  *	@tcx_egress:		BPF & clsact qdisc specific data for egress processing
1917  *	@nf_hooks_egress:	netfilter hooks executed for egress packets
1918  *	@qdisc_hash:		qdisc hash table
1919  *	@watchdog_timeo:	Represents the timeout that is used by
1920  *				the watchdog (see dev_watchdog())
1921  *	@watchdog_timer:	List of timers
1922  *
1923  *	@proto_down_reason:	reason a netdev interface is held down
1924  *	@pcpu_refcnt:		Number of references to this device
1925  *	@dev_refcnt:		Number of references to this device
1926  *	@refcnt_tracker:	Tracker directory for tracked references to this device
1927  *	@todo_list:		Delayed register/unregister
1928  *	@link_watch_list:	XXX: need comments on this one
1929  *
1930  *	@reg_state:		Register/unregister state machine
1931  *	@dismantle:		Device is going to be freed
1932  *	@rtnl_link_state:	This enum represents the phases of creating
1933  *				a new link
1934  *
1935  *	@needs_free_netdev:	Should unregister perform free_netdev?
1936  *	@priv_destructor:	Called from unregister
1937  *	@npinfo:		XXX: need comments on this one
1938  * 	@nd_net:		Network namespace this network device is inside
1939  *
1940  * 	@ml_priv:	Mid-layer private
1941  *	@ml_priv_type:  Mid-layer private type
1942  *
1943  *	@pcpu_stat_type:	Type of device statistics which the core should
1944  *				allocate/free: none, lstats, tstats, dstats. none
1945  *				means the driver is handling statistics allocation/
1946  *				freeing internally.
1947  *	@lstats:		Loopback statistics: packets, bytes
1948  *	@tstats:		Tunnel statistics: RX/TX packets, RX/TX bytes
1949  *	@dstats:		Dummy statistics: RX/TX/drop packets, RX/TX bytes
1950  *
1951  *	@garp_port:	GARP
1952  *	@mrp_port:	MRP
1953  *
1954  *	@dm_private:	Drop monitor private
1955  *
1956  *	@dev:		Class/net/name entry
1957  *	@sysfs_groups:	Space for optional device, statistics and wireless
1958  *			sysfs groups
1959  *
1960  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1961  *	@rtnl_link_ops:	Rtnl_link_ops
1962  *	@stat_ops:	Optional ops for queue-aware statistics
1963  *	@queue_mgmt_ops:	Optional ops for queue management
1964  *
1965  *	@gso_max_size:	Maximum size of generic segmentation offload
1966  *	@tso_max_size:	Device (as in HW) limit on the max TSO request size
1967  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1968  *			NIC for GSO
1969  *	@tso_max_segs:	Device (as in HW) limit on the max TSO segment count
1970  * 	@gso_ipv4_max_size:	Maximum size of generic segmentation offload,
1971  * 				for IPv4.
1972  *
1973  *	@dcbnl_ops:	Data Center Bridging netlink ops
1974  *	@num_tc:	Number of traffic classes in the net device
1975  *	@tc_to_txq:	XXX: need comments on this one
1976  *	@prio_tc_map:	XXX: need comments on this one
1977  *
1978  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1979  *
1980  *	@priomap:	XXX: need comments on this one
1981  *	@link_topo:	Physical link topology tracking attached PHYs
1982  *	@phydev:	Physical device may attach itself
1983  *			for hardware timestamping
1984  *	@sfp_bus:	attached &struct sfp_bus structure.
1985  *
1986  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1987  *
1988  *	@proto_down:	protocol port state information can be sent to the
1989  *			switch driver and used to set the phys state of the
1990  *			switch port.
1991  *
1992  *	@threaded:	napi threaded mode is enabled
1993  *
1994  *	@see_all_hwtstamp_requests: device wants to see calls to
1995  *			ndo_hwtstamp_set() for all timestamp requests
1996  *			regardless of source, even if those aren't
1997  *			HWTSTAMP_SOURCE_NETDEV
1998  *	@change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN
1999  *	@netns_local: interface can't change network namespaces
2000  *	@fcoe_mtu:	device supports maximum FCoE MTU, 2158 bytes
2001  *
2002  *	@net_notifier_list:	List of per-net netdev notifier block
2003  *				that follow this device when it is moved
2004  *				to another network namespace.
2005  *
2006  *	@macsec_ops:    MACsec offloading ops
2007  *
2008  *	@udp_tunnel_nic_info:	static structure describing the UDP tunnel
2009  *				offload capabilities of the device
2010  *	@udp_tunnel_nic:	UDP tunnel offload state
2011  *	@ethtool:	ethtool related state
2012  *	@xdp_state:		stores info on attached XDP BPF programs
2013  *
2014  *	@nested_level:	Used as a parameter of spin_lock_nested() of
2015  *			dev->addr_list_lock.
2016  *	@unlink_list:	As netif_addr_lock() can be called recursively,
2017  *			keep a list of interfaces to be deleted.
2018  *	@gro_max_size:	Maximum size of aggregated packet in generic
2019  *			receive offload (GRO)
2020  * 	@gro_ipv4_max_size:	Maximum size of aggregated packet in generic
2021  * 				receive offload (GRO), for IPv4.
2022  *	@xdp_zc_max_segs:	Maximum number of segments supported by AF_XDP
2023  *				zero copy driver
2024  *
2025  *	@dev_addr_shadow:	Copy of @dev_addr to catch direct writes.
2026  *	@linkwatch_dev_tracker:	refcount tracker used by linkwatch.
2027  *	@watchdog_dev_tracker:	refcount tracker used by watchdog.
2028  *	@dev_registered_tracker:	tracker for reference held while
2029  *					registered
2030  *	@offload_xstats_l3:	L3 HW stats for this netdevice.
2031  *
2032  *	@devlink_port:	Pointer to related devlink port structure.
2033  *			Assigned by a driver before netdev registration using
2034  *			SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2035  *			during the time netdevice is registered.
2036  *
2037  *	@dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2038  *		   where the clock is recovered.
2039  *
2040  *	@max_pacing_offload_horizon: max EDT offload horizon in nsec.
2041  *	@napi_config: An array of napi_config structures containing per-NAPI
2042  *		      settings.
2043  *	@gro_flush_timeout:	timeout for GRO layer in NAPI
2044  *	@napi_defer_hard_irqs:	If not zero, provides a counter that would
2045  *				allow to avoid NIC hard IRQ, on busy queues.
2046  *
2047  *	@neighbours:	List heads pointing to this device's neighbours'
2048  *			dev_list, one per address-family.
2049  *	@hwprov: Tracks which PTP performs hardware packet time stamping.
2050  *
2051  *	FIXME: cleanup struct net_device such that network protocol info
2052  *	moves out.
2053  */
2054 
2055 struct net_device {
2056 	/* Cacheline organization can be found documented in
2057 	 * Documentation/networking/net_cachelines/net_device.rst.
2058 	 * Please update the document when adding new fields.
2059 	 */
2060 
2061 	/* TX read-mostly hotpath */
2062 	__cacheline_group_begin(net_device_read_tx);
2063 	struct_group(priv_flags_fast,
2064 		unsigned long		priv_flags:32;
2065 		unsigned long		lltx:1;
2066 	);
2067 	const struct net_device_ops *netdev_ops;
2068 	const struct header_ops *header_ops;
2069 	struct netdev_queue	*_tx;
2070 	netdev_features_t	gso_partial_features;
2071 	unsigned int		real_num_tx_queues;
2072 	unsigned int		gso_max_size;
2073 	unsigned int		gso_ipv4_max_size;
2074 	u16			gso_max_segs;
2075 	s16			num_tc;
2076 	/* Note : dev->mtu is often read without holding a lock.
2077 	 * Writers usually hold RTNL.
2078 	 * It is recommended to use READ_ONCE() to annotate the reads,
2079 	 * and to use WRITE_ONCE() to annotate the writes.
2080 	 */
2081 	unsigned int		mtu;
2082 	unsigned short		needed_headroom;
2083 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
2084 #ifdef CONFIG_XPS
2085 	struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2086 #endif
2087 #ifdef CONFIG_NETFILTER_EGRESS
2088 	struct nf_hook_entries __rcu *nf_hooks_egress;
2089 #endif
2090 #ifdef CONFIG_NET_XGRESS
2091 	struct bpf_mprog_entry __rcu *tcx_egress;
2092 #endif
2093 	__cacheline_group_end(net_device_read_tx);
2094 
2095 	/* TXRX read-mostly hotpath */
2096 	__cacheline_group_begin(net_device_read_txrx);
2097 	union {
2098 		struct pcpu_lstats __percpu		*lstats;
2099 		struct pcpu_sw_netstats __percpu	*tstats;
2100 		struct pcpu_dstats __percpu		*dstats;
2101 	};
2102 	unsigned long		state;
2103 	unsigned int		flags;
2104 	unsigned short		hard_header_len;
2105 	netdev_features_t	features;
2106 	struct inet6_dev __rcu	*ip6_ptr;
2107 	__cacheline_group_end(net_device_read_txrx);
2108 
2109 	/* RX read-mostly hotpath */
2110 	__cacheline_group_begin(net_device_read_rx);
2111 	struct bpf_prog __rcu	*xdp_prog;
2112 	struct list_head	ptype_specific;
2113 	int			ifindex;
2114 	unsigned int		real_num_rx_queues;
2115 	struct netdev_rx_queue	*_rx;
2116 	unsigned int		gro_max_size;
2117 	unsigned int		gro_ipv4_max_size;
2118 	rx_handler_func_t __rcu	*rx_handler;
2119 	void __rcu		*rx_handler_data;
2120 	possible_net_t			nd_net;
2121 #ifdef CONFIG_NETPOLL
2122 	struct netpoll_info __rcu	*npinfo;
2123 #endif
2124 #ifdef CONFIG_NET_XGRESS
2125 	struct bpf_mprog_entry __rcu *tcx_ingress;
2126 #endif
2127 	__cacheline_group_end(net_device_read_rx);
2128 
2129 	char			name[IFNAMSIZ];
2130 	struct netdev_name_node	*name_node;
2131 	struct dev_ifalias	__rcu *ifalias;
2132 	/*
2133 	 *	I/O specific fields
2134 	 *	FIXME: Merge these and struct ifmap into one
2135 	 */
2136 	unsigned long		mem_end;
2137 	unsigned long		mem_start;
2138 	unsigned long		base_addr;
2139 
2140 	/*
2141 	 *	Some hardware also needs these fields (state,dev_list,
2142 	 *	napi_list,unreg_list,close_list) but they are not
2143 	 *	part of the usual set specified in Space.c.
2144 	 */
2145 
2146 
2147 	struct list_head	dev_list;
2148 	struct list_head	napi_list;
2149 	struct list_head	unreg_list;
2150 	struct list_head	close_list;
2151 	struct list_head	ptype_all;
2152 
2153 	struct {
2154 		struct list_head upper;
2155 		struct list_head lower;
2156 	} adj_list;
2157 
2158 	/* Read-mostly cache-line for fast-path access */
2159 	xdp_features_t		xdp_features;
2160 	const struct xdp_metadata_ops *xdp_metadata_ops;
2161 	const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2162 	unsigned short		gflags;
2163 
2164 	unsigned short		needed_tailroom;
2165 
2166 	netdev_features_t	hw_features;
2167 	netdev_features_t	wanted_features;
2168 	netdev_features_t	vlan_features;
2169 	netdev_features_t	hw_enc_features;
2170 	netdev_features_t	mpls_features;
2171 
2172 	unsigned int		min_mtu;
2173 	unsigned int		max_mtu;
2174 	unsigned short		type;
2175 	unsigned char		min_header_len;
2176 	unsigned char		name_assign_type;
2177 
2178 	int			group;
2179 
2180 	struct net_device_stats	stats; /* not used by modern drivers */
2181 
2182 	struct net_device_core_stats __percpu *core_stats;
2183 
2184 	/* Stats to monitor link on/off, flapping */
2185 	atomic_t		carrier_up_count;
2186 	atomic_t		carrier_down_count;
2187 
2188 #ifdef CONFIG_WIRELESS_EXT
2189 	const struct iw_handler_def *wireless_handlers;
2190 #endif
2191 	const struct ethtool_ops *ethtool_ops;
2192 #ifdef CONFIG_NET_L3_MASTER_DEV
2193 	const struct l3mdev_ops	*l3mdev_ops;
2194 #endif
2195 #if IS_ENABLED(CONFIG_IPV6)
2196 	const struct ndisc_ops *ndisc_ops;
2197 #endif
2198 
2199 #ifdef CONFIG_XFRM_OFFLOAD
2200 	const struct xfrmdev_ops *xfrmdev_ops;
2201 #endif
2202 
2203 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2204 	const struct tlsdev_ops *tlsdev_ops;
2205 #endif
2206 
2207 	unsigned int		operstate;
2208 	unsigned char		link_mode;
2209 
2210 	unsigned char		if_port;
2211 	unsigned char		dma;
2212 
2213 	/* Interface address info. */
2214 	unsigned char		perm_addr[MAX_ADDR_LEN];
2215 	unsigned char		addr_assign_type;
2216 	unsigned char		addr_len;
2217 	unsigned char		upper_level;
2218 	unsigned char		lower_level;
2219 
2220 	unsigned short		neigh_priv_len;
2221 	unsigned short          dev_id;
2222 	unsigned short          dev_port;
2223 	int			irq;
2224 	u32			priv_len;
2225 
2226 	spinlock_t		addr_list_lock;
2227 
2228 	struct netdev_hw_addr_list	uc;
2229 	struct netdev_hw_addr_list	mc;
2230 	struct netdev_hw_addr_list	dev_addrs;
2231 
2232 #ifdef CONFIG_SYSFS
2233 	struct kset		*queues_kset;
2234 #endif
2235 #ifdef CONFIG_LOCKDEP
2236 	struct list_head	unlink_list;
2237 #endif
2238 	unsigned int		promiscuity;
2239 	unsigned int		allmulti;
2240 	bool			uc_promisc;
2241 #ifdef CONFIG_LOCKDEP
2242 	unsigned char		nested_level;
2243 #endif
2244 
2245 
2246 	/* Protocol-specific pointers */
2247 	struct in_device __rcu	*ip_ptr;
2248 	/** @fib_nh_head: nexthops associated with this netdev */
2249 	struct hlist_head	fib_nh_head;
2250 
2251 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2252 	struct vlan_info __rcu	*vlan_info;
2253 #endif
2254 #if IS_ENABLED(CONFIG_NET_DSA)
2255 	struct dsa_port		*dsa_ptr;
2256 #endif
2257 #if IS_ENABLED(CONFIG_TIPC)
2258 	struct tipc_bearer __rcu *tipc_ptr;
2259 #endif
2260 #if IS_ENABLED(CONFIG_ATALK)
2261 	void 			*atalk_ptr;
2262 #endif
2263 #if IS_ENABLED(CONFIG_AX25)
2264 	struct ax25_dev	__rcu	*ax25_ptr;
2265 #endif
2266 #if IS_ENABLED(CONFIG_CFG80211)
2267 	struct wireless_dev	*ieee80211_ptr;
2268 #endif
2269 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2270 	struct wpan_dev		*ieee802154_ptr;
2271 #endif
2272 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2273 	struct mpls_dev __rcu	*mpls_ptr;
2274 #endif
2275 #if IS_ENABLED(CONFIG_MCTP)
2276 	struct mctp_dev __rcu	*mctp_ptr;
2277 #endif
2278 
2279 /*
2280  * Cache lines mostly used on receive path (including eth_type_trans())
2281  */
2282 	/* Interface address info used in eth_type_trans() */
2283 	const unsigned char	*dev_addr;
2284 
2285 	unsigned int		num_rx_queues;
2286 #define GRO_LEGACY_MAX_SIZE	65536u
2287 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2288  * and shinfo->gso_segs is a 16bit field.
2289  */
2290 #define GRO_MAX_SIZE		(8 * 65535u)
2291 	unsigned int		xdp_zc_max_segs;
2292 	struct netdev_queue __rcu *ingress_queue;
2293 #ifdef CONFIG_NETFILTER_INGRESS
2294 	struct nf_hook_entries __rcu *nf_hooks_ingress;
2295 #endif
2296 
2297 	unsigned char		broadcast[MAX_ADDR_LEN];
2298 #ifdef CONFIG_RFS_ACCEL
2299 	struct cpu_rmap		*rx_cpu_rmap;
2300 #endif
2301 	struct hlist_node	index_hlist;
2302 
2303 /*
2304  * Cache lines mostly used on transmit path
2305  */
2306 	unsigned int		num_tx_queues;
2307 	struct Qdisc __rcu	*qdisc;
2308 	unsigned int		tx_queue_len;
2309 	spinlock_t		tx_global_lock;
2310 
2311 	struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2312 
2313 #ifdef CONFIG_NET_SCHED
2314 	DECLARE_HASHTABLE	(qdisc_hash, 4);
2315 #endif
2316 	/* These may be needed for future network-power-down code. */
2317 	struct timer_list	watchdog_timer;
2318 	int			watchdog_timeo;
2319 
2320 	u32                     proto_down_reason;
2321 
2322 	struct list_head	todo_list;
2323 
2324 #ifdef CONFIG_PCPU_DEV_REFCNT
2325 	int __percpu		*pcpu_refcnt;
2326 #else
2327 	refcount_t		dev_refcnt;
2328 #endif
2329 	struct ref_tracker_dir	refcnt_tracker;
2330 
2331 	struct list_head	link_watch_list;
2332 
2333 	u8 reg_state;
2334 
2335 	bool dismantle;
2336 
2337 	enum {
2338 		RTNL_LINK_INITIALIZED,
2339 		RTNL_LINK_INITIALIZING,
2340 	} rtnl_link_state:16;
2341 
2342 	bool needs_free_netdev;
2343 	void (*priv_destructor)(struct net_device *dev);
2344 
2345 	/* mid-layer private */
2346 	void				*ml_priv;
2347 	enum netdev_ml_priv_type	ml_priv_type;
2348 
2349 	enum netdev_stat_type		pcpu_stat_type:8;
2350 
2351 #if IS_ENABLED(CONFIG_GARP)
2352 	struct garp_port __rcu	*garp_port;
2353 #endif
2354 #if IS_ENABLED(CONFIG_MRP)
2355 	struct mrp_port __rcu	*mrp_port;
2356 #endif
2357 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2358 	struct dm_hw_stat_delta __rcu *dm_private;
2359 #endif
2360 	struct device		dev;
2361 	const struct attribute_group *sysfs_groups[4];
2362 	const struct attribute_group *sysfs_rx_queue_group;
2363 
2364 	const struct rtnl_link_ops *rtnl_link_ops;
2365 
2366 	const struct netdev_stat_ops *stat_ops;
2367 
2368 	const struct netdev_queue_mgmt_ops *queue_mgmt_ops;
2369 
2370 	/* for setting kernel sock attribute on TCP connection setup */
2371 #define GSO_MAX_SEGS		65535u
2372 #define GSO_LEGACY_MAX_SIZE	65536u
2373 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2374  * and shinfo->gso_segs is a 16bit field.
2375  */
2376 #define GSO_MAX_SIZE		(8 * GSO_MAX_SEGS)
2377 
2378 #define TSO_LEGACY_MAX_SIZE	65536
2379 #define TSO_MAX_SIZE		UINT_MAX
2380 	unsigned int		tso_max_size;
2381 #define TSO_MAX_SEGS		U16_MAX
2382 	u16			tso_max_segs;
2383 
2384 #ifdef CONFIG_DCB
2385 	const struct dcbnl_rtnl_ops *dcbnl_ops;
2386 #endif
2387 	u8			prio_tc_map[TC_BITMASK + 1];
2388 
2389 #if IS_ENABLED(CONFIG_FCOE)
2390 	unsigned int		fcoe_ddp_xid;
2391 #endif
2392 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2393 	struct netprio_map __rcu *priomap;
2394 #endif
2395 	struct phy_link_topology	*link_topo;
2396 	struct phy_device	*phydev;
2397 	struct sfp_bus		*sfp_bus;
2398 	struct lock_class_key	*qdisc_tx_busylock;
2399 	bool			proto_down;
2400 	bool			threaded;
2401 
2402 	/* priv_flags_slow, ungrouped to save space */
2403 	unsigned long		see_all_hwtstamp_requests:1;
2404 	unsigned long		change_proto_down:1;
2405 	unsigned long		netns_local:1;
2406 	unsigned long		fcoe_mtu:1;
2407 
2408 	struct list_head	net_notifier_list;
2409 
2410 #if IS_ENABLED(CONFIG_MACSEC)
2411 	/* MACsec management functions */
2412 	const struct macsec_ops *macsec_ops;
2413 #endif
2414 	const struct udp_tunnel_nic_info	*udp_tunnel_nic_info;
2415 	struct udp_tunnel_nic	*udp_tunnel_nic;
2416 
2417 	struct ethtool_netdev_state *ethtool;
2418 
2419 	/* protected by rtnl_lock */
2420 	struct bpf_xdp_entity	xdp_state[__MAX_XDP_MODE];
2421 
2422 	u8 dev_addr_shadow[MAX_ADDR_LEN];
2423 	netdevice_tracker	linkwatch_dev_tracker;
2424 	netdevice_tracker	watchdog_dev_tracker;
2425 	netdevice_tracker	dev_registered_tracker;
2426 	struct rtnl_hw_stats64	*offload_xstats_l3;
2427 
2428 	struct devlink_port	*devlink_port;
2429 
2430 #if IS_ENABLED(CONFIG_DPLL)
2431 	struct dpll_pin	__rcu	*dpll_pin;
2432 #endif
2433 #if IS_ENABLED(CONFIG_PAGE_POOL)
2434 	/** @page_pools: page pools created for this netdevice */
2435 	struct hlist_head	page_pools;
2436 #endif
2437 
2438 	/** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */
2439 	struct dim_irq_moder	*irq_moder;
2440 
2441 	u64			max_pacing_offload_horizon;
2442 	struct napi_config	*napi_config;
2443 	unsigned long		gro_flush_timeout;
2444 	u32			napi_defer_hard_irqs;
2445 
2446 	/**
2447 	 * @lock: protects @net_shaper_hierarchy, feel free to use for other
2448 	 * netdev-scope protection. Ordering: take after rtnl_lock.
2449 	 */
2450 	struct mutex		lock;
2451 
2452 #if IS_ENABLED(CONFIG_NET_SHAPER)
2453 	/**
2454 	 * @net_shaper_hierarchy: data tracking the current shaper status
2455 	 *  see include/net/net_shapers.h
2456 	 */
2457 	struct net_shaper_hierarchy *net_shaper_hierarchy;
2458 #endif
2459 
2460 	struct hlist_head neighbours[NEIGH_NR_TABLES];
2461 
2462 	struct hwtstamp_provider __rcu	*hwprov;
2463 
2464 	u8			priv[] ____cacheline_aligned
2465 				       __counted_by(priv_len);
2466 } ____cacheline_aligned;
2467 #define to_net_dev(d) container_of(d, struct net_device, dev)
2468 
2469 /*
2470  * Driver should use this to assign devlink port instance to a netdevice
2471  * before it registers the netdevice. Therefore devlink_port is static
2472  * during the netdev lifetime after it is registered.
2473  */
2474 #define SET_NETDEV_DEVLINK_PORT(dev, port)			\
2475 ({								\
2476 	WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED);	\
2477 	((dev)->devlink_port = (port));				\
2478 })
2479 
2480 static inline bool netif_elide_gro(const struct net_device *dev)
2481 {
2482 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2483 		return true;
2484 	return false;
2485 }
2486 
2487 #define	NETDEV_ALIGN		32
2488 
2489 static inline
2490 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2491 {
2492 	return dev->prio_tc_map[prio & TC_BITMASK];
2493 }
2494 
2495 static inline
2496 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2497 {
2498 	if (tc >= dev->num_tc)
2499 		return -EINVAL;
2500 
2501 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2502 	return 0;
2503 }
2504 
2505 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2506 void netdev_reset_tc(struct net_device *dev);
2507 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2508 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2509 
2510 static inline
2511 int netdev_get_num_tc(struct net_device *dev)
2512 {
2513 	return dev->num_tc;
2514 }
2515 
2516 static inline void net_prefetch(void *p)
2517 {
2518 	prefetch(p);
2519 #if L1_CACHE_BYTES < 128
2520 	prefetch((u8 *)p + L1_CACHE_BYTES);
2521 #endif
2522 }
2523 
2524 static inline void net_prefetchw(void *p)
2525 {
2526 	prefetchw(p);
2527 #if L1_CACHE_BYTES < 128
2528 	prefetchw((u8 *)p + L1_CACHE_BYTES);
2529 #endif
2530 }
2531 
2532 void netdev_unbind_sb_channel(struct net_device *dev,
2533 			      struct net_device *sb_dev);
2534 int netdev_bind_sb_channel_queue(struct net_device *dev,
2535 				 struct net_device *sb_dev,
2536 				 u8 tc, u16 count, u16 offset);
2537 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2538 static inline int netdev_get_sb_channel(struct net_device *dev)
2539 {
2540 	return max_t(int, -dev->num_tc, 0);
2541 }
2542 
2543 static inline
2544 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2545 					 unsigned int index)
2546 {
2547 	DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2548 	return &dev->_tx[index];
2549 }
2550 
2551 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2552 						    const struct sk_buff *skb)
2553 {
2554 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2555 }
2556 
2557 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2558 					    void (*f)(struct net_device *,
2559 						      struct netdev_queue *,
2560 						      void *),
2561 					    void *arg)
2562 {
2563 	unsigned int i;
2564 
2565 	for (i = 0; i < dev->num_tx_queues; i++)
2566 		f(dev, &dev->_tx[i], arg);
2567 }
2568 
2569 #define netdev_lockdep_set_classes(dev)				\
2570 {								\
2571 	static struct lock_class_key qdisc_tx_busylock_key;	\
2572 	static struct lock_class_key qdisc_xmit_lock_key;	\
2573 	static struct lock_class_key dev_addr_list_lock_key;	\
2574 	unsigned int i;						\
2575 								\
2576 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
2577 	lockdep_set_class(&(dev)->addr_list_lock,		\
2578 			  &dev_addr_list_lock_key);		\
2579 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
2580 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
2581 				  &qdisc_xmit_lock_key);	\
2582 }
2583 
2584 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2585 		     struct net_device *sb_dev);
2586 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2587 					 struct sk_buff *skb,
2588 					 struct net_device *sb_dev);
2589 
2590 /* returns the headroom that the master device needs to take in account
2591  * when forwarding to this dev
2592  */
2593 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2594 {
2595 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2596 }
2597 
2598 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2599 {
2600 	if (dev->netdev_ops->ndo_set_rx_headroom)
2601 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2602 }
2603 
2604 /* set the device rx headroom to the dev's default */
2605 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2606 {
2607 	netdev_set_rx_headroom(dev, -1);
2608 }
2609 
2610 static inline void *netdev_get_ml_priv(struct net_device *dev,
2611 				       enum netdev_ml_priv_type type)
2612 {
2613 	if (dev->ml_priv_type != type)
2614 		return NULL;
2615 
2616 	return dev->ml_priv;
2617 }
2618 
2619 static inline void netdev_set_ml_priv(struct net_device *dev,
2620 				      void *ml_priv,
2621 				      enum netdev_ml_priv_type type)
2622 {
2623 	WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2624 	     "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2625 	     dev->ml_priv_type, type);
2626 	WARN(!dev->ml_priv_type && dev->ml_priv,
2627 	     "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2628 
2629 	dev->ml_priv = ml_priv;
2630 	dev->ml_priv_type = type;
2631 }
2632 
2633 /*
2634  * Net namespace inlines
2635  */
2636 static inline
2637 struct net *dev_net(const struct net_device *dev)
2638 {
2639 	return read_pnet(&dev->nd_net);
2640 }
2641 
2642 static inline
2643 void dev_net_set(struct net_device *dev, struct net *net)
2644 {
2645 	write_pnet(&dev->nd_net, net);
2646 }
2647 
2648 /**
2649  *	netdev_priv - access network device private data
2650  *	@dev: network device
2651  *
2652  * Get network device private data
2653  */
2654 static inline void *netdev_priv(const struct net_device *dev)
2655 {
2656 	return (void *)dev->priv;
2657 }
2658 
2659 /* Set the sysfs physical device reference for the network logical device
2660  * if set prior to registration will cause a symlink during initialization.
2661  */
2662 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2663 
2664 /* Set the sysfs device type for the network logical device to allow
2665  * fine-grained identification of different network device types. For
2666  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2667  */
2668 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2669 
2670 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2671 			  enum netdev_queue_type type,
2672 			  struct napi_struct *napi);
2673 
2674 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2675 {
2676 	napi->irq = irq;
2677 }
2678 
2679 /* Default NAPI poll() weight
2680  * Device drivers are strongly advised to not use bigger value
2681  */
2682 #define NAPI_POLL_WEIGHT 64
2683 
2684 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2685 			   int (*poll)(struct napi_struct *, int), int weight);
2686 
2687 /**
2688  * netif_napi_add() - initialize a NAPI context
2689  * @dev:  network device
2690  * @napi: NAPI context
2691  * @poll: polling function
2692  *
2693  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2694  * *any* of the other NAPI-related functions.
2695  */
2696 static inline void
2697 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2698 	       int (*poll)(struct napi_struct *, int))
2699 {
2700 	netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2701 }
2702 
2703 static inline void
2704 netif_napi_add_tx_weight(struct net_device *dev,
2705 			 struct napi_struct *napi,
2706 			 int (*poll)(struct napi_struct *, int),
2707 			 int weight)
2708 {
2709 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2710 	netif_napi_add_weight(dev, napi, poll, weight);
2711 }
2712 
2713 /**
2714  * netif_napi_add_config - initialize a NAPI context with persistent config
2715  * @dev: network device
2716  * @napi: NAPI context
2717  * @poll: polling function
2718  * @index: the NAPI index
2719  */
2720 static inline void
2721 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi,
2722 		      int (*poll)(struct napi_struct *, int), int index)
2723 {
2724 	napi->index = index;
2725 	napi->config = &dev->napi_config[index];
2726 	netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2727 }
2728 
2729 /**
2730  * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2731  * @dev:  network device
2732  * @napi: NAPI context
2733  * @poll: polling function
2734  *
2735  * This variant of netif_napi_add() should be used from drivers using NAPI
2736  * to exclusively poll a TX queue.
2737  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2738  */
2739 static inline void netif_napi_add_tx(struct net_device *dev,
2740 				     struct napi_struct *napi,
2741 				     int (*poll)(struct napi_struct *, int))
2742 {
2743 	netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2744 }
2745 
2746 /**
2747  *  __netif_napi_del - remove a NAPI context
2748  *  @napi: NAPI context
2749  *
2750  * Warning: caller must observe RCU grace period before freeing memory
2751  * containing @napi. Drivers might want to call this helper to combine
2752  * all the needed RCU grace periods into a single one.
2753  */
2754 void __netif_napi_del(struct napi_struct *napi);
2755 
2756 /**
2757  *  netif_napi_del - remove a NAPI context
2758  *  @napi: NAPI context
2759  *
2760  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2761  */
2762 static inline void netif_napi_del(struct napi_struct *napi)
2763 {
2764 	__netif_napi_del(napi);
2765 	synchronize_net();
2766 }
2767 
2768 struct packet_type {
2769 	__be16			type;	/* This is really htons(ether_type). */
2770 	bool			ignore_outgoing;
2771 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2772 	netdevice_tracker	dev_tracker;
2773 	int			(*func) (struct sk_buff *,
2774 					 struct net_device *,
2775 					 struct packet_type *,
2776 					 struct net_device *);
2777 	void			(*list_func) (struct list_head *,
2778 					      struct packet_type *,
2779 					      struct net_device *);
2780 	bool			(*id_match)(struct packet_type *ptype,
2781 					    struct sock *sk);
2782 	struct net		*af_packet_net;
2783 	void			*af_packet_priv;
2784 	struct list_head	list;
2785 };
2786 
2787 struct offload_callbacks {
2788 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2789 						netdev_features_t features);
2790 	struct sk_buff		*(*gro_receive)(struct list_head *head,
2791 						struct sk_buff *skb);
2792 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2793 };
2794 
2795 struct packet_offload {
2796 	__be16			 type;	/* This is really htons(ether_type). */
2797 	u16			 priority;
2798 	struct offload_callbacks callbacks;
2799 	struct list_head	 list;
2800 };
2801 
2802 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2803 struct pcpu_sw_netstats {
2804 	u64_stats_t		rx_packets;
2805 	u64_stats_t		rx_bytes;
2806 	u64_stats_t		tx_packets;
2807 	u64_stats_t		tx_bytes;
2808 	struct u64_stats_sync   syncp;
2809 } __aligned(4 * sizeof(u64));
2810 
2811 struct pcpu_dstats {
2812 	u64_stats_t		rx_packets;
2813 	u64_stats_t		rx_bytes;
2814 	u64_stats_t		rx_drops;
2815 	u64_stats_t		tx_packets;
2816 	u64_stats_t		tx_bytes;
2817 	u64_stats_t		tx_drops;
2818 	struct u64_stats_sync	syncp;
2819 } __aligned(8 * sizeof(u64));
2820 
2821 struct pcpu_lstats {
2822 	u64_stats_t packets;
2823 	u64_stats_t bytes;
2824 	struct u64_stats_sync syncp;
2825 } __aligned(2 * sizeof(u64));
2826 
2827 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2828 
2829 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2830 {
2831 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2832 
2833 	u64_stats_update_begin(&tstats->syncp);
2834 	u64_stats_add(&tstats->rx_bytes, len);
2835 	u64_stats_inc(&tstats->rx_packets);
2836 	u64_stats_update_end(&tstats->syncp);
2837 }
2838 
2839 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2840 					  unsigned int packets,
2841 					  unsigned int len)
2842 {
2843 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2844 
2845 	u64_stats_update_begin(&tstats->syncp);
2846 	u64_stats_add(&tstats->tx_bytes, len);
2847 	u64_stats_add(&tstats->tx_packets, packets);
2848 	u64_stats_update_end(&tstats->syncp);
2849 }
2850 
2851 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2852 {
2853 	struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2854 
2855 	u64_stats_update_begin(&lstats->syncp);
2856 	u64_stats_add(&lstats->bytes, len);
2857 	u64_stats_inc(&lstats->packets);
2858 	u64_stats_update_end(&lstats->syncp);
2859 }
2860 
2861 static inline void dev_dstats_rx_add(struct net_device *dev,
2862 				     unsigned int len)
2863 {
2864 	struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats);
2865 
2866 	u64_stats_update_begin(&dstats->syncp);
2867 	u64_stats_inc(&dstats->rx_packets);
2868 	u64_stats_add(&dstats->rx_bytes, len);
2869 	u64_stats_update_end(&dstats->syncp);
2870 }
2871 
2872 static inline void dev_dstats_rx_dropped(struct net_device *dev)
2873 {
2874 	struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats);
2875 
2876 	u64_stats_update_begin(&dstats->syncp);
2877 	u64_stats_inc(&dstats->rx_drops);
2878 	u64_stats_update_end(&dstats->syncp);
2879 }
2880 
2881 static inline void dev_dstats_tx_add(struct net_device *dev,
2882 				     unsigned int len)
2883 {
2884 	struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats);
2885 
2886 	u64_stats_update_begin(&dstats->syncp);
2887 	u64_stats_inc(&dstats->tx_packets);
2888 	u64_stats_add(&dstats->tx_bytes, len);
2889 	u64_stats_update_end(&dstats->syncp);
2890 }
2891 
2892 static inline void dev_dstats_tx_dropped(struct net_device *dev)
2893 {
2894 	struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats);
2895 
2896 	u64_stats_update_begin(&dstats->syncp);
2897 	u64_stats_inc(&dstats->tx_drops);
2898 	u64_stats_update_end(&dstats->syncp);
2899 }
2900 
2901 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2902 ({									\
2903 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2904 	if (pcpu_stats)	{						\
2905 		int __cpu;						\
2906 		for_each_possible_cpu(__cpu) {				\
2907 			typeof(type) *stat;				\
2908 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2909 			u64_stats_init(&stat->syncp);			\
2910 		}							\
2911 	}								\
2912 	pcpu_stats;							\
2913 })
2914 
2915 #define netdev_alloc_pcpu_stats(type)					\
2916 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2917 
2918 #define devm_netdev_alloc_pcpu_stats(dev, type)				\
2919 ({									\
2920 	typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2921 	if (pcpu_stats) {						\
2922 		int __cpu;						\
2923 		for_each_possible_cpu(__cpu) {				\
2924 			typeof(type) *stat;				\
2925 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2926 			u64_stats_init(&stat->syncp);			\
2927 		}							\
2928 	}								\
2929 	pcpu_stats;							\
2930 })
2931 
2932 enum netdev_lag_tx_type {
2933 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2934 	NETDEV_LAG_TX_TYPE_RANDOM,
2935 	NETDEV_LAG_TX_TYPE_BROADCAST,
2936 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2937 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2938 	NETDEV_LAG_TX_TYPE_HASH,
2939 };
2940 
2941 enum netdev_lag_hash {
2942 	NETDEV_LAG_HASH_NONE,
2943 	NETDEV_LAG_HASH_L2,
2944 	NETDEV_LAG_HASH_L34,
2945 	NETDEV_LAG_HASH_L23,
2946 	NETDEV_LAG_HASH_E23,
2947 	NETDEV_LAG_HASH_E34,
2948 	NETDEV_LAG_HASH_VLAN_SRCMAC,
2949 	NETDEV_LAG_HASH_UNKNOWN,
2950 };
2951 
2952 struct netdev_lag_upper_info {
2953 	enum netdev_lag_tx_type tx_type;
2954 	enum netdev_lag_hash hash_type;
2955 };
2956 
2957 struct netdev_lag_lower_state_info {
2958 	u8 link_up : 1,
2959 	   tx_enabled : 1;
2960 };
2961 
2962 #include <linux/notifier.h>
2963 
2964 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2965  * and the rtnetlink notification exclusion list in rtnetlink_event() when
2966  * adding new types.
2967  */
2968 enum netdev_cmd {
2969 	NETDEV_UP	= 1,	/* For now you can't veto a device up/down */
2970 	NETDEV_DOWN,
2971 	NETDEV_REBOOT,		/* Tell a protocol stack a network interface
2972 				   detected a hardware crash and restarted
2973 				   - we can use this eg to kick tcp sessions
2974 				   once done */
2975 	NETDEV_CHANGE,		/* Notify device state change */
2976 	NETDEV_REGISTER,
2977 	NETDEV_UNREGISTER,
2978 	NETDEV_CHANGEMTU,	/* notify after mtu change happened */
2979 	NETDEV_CHANGEADDR,	/* notify after the address change */
2980 	NETDEV_PRE_CHANGEADDR,	/* notify before the address change */
2981 	NETDEV_GOING_DOWN,
2982 	NETDEV_CHANGENAME,
2983 	NETDEV_FEAT_CHANGE,
2984 	NETDEV_BONDING_FAILOVER,
2985 	NETDEV_PRE_UP,
2986 	NETDEV_PRE_TYPE_CHANGE,
2987 	NETDEV_POST_TYPE_CHANGE,
2988 	NETDEV_POST_INIT,
2989 	NETDEV_PRE_UNINIT,
2990 	NETDEV_RELEASE,
2991 	NETDEV_NOTIFY_PEERS,
2992 	NETDEV_JOIN,
2993 	NETDEV_CHANGEUPPER,
2994 	NETDEV_RESEND_IGMP,
2995 	NETDEV_PRECHANGEMTU,	/* notify before mtu change happened */
2996 	NETDEV_CHANGEINFODATA,
2997 	NETDEV_BONDING_INFO,
2998 	NETDEV_PRECHANGEUPPER,
2999 	NETDEV_CHANGELOWERSTATE,
3000 	NETDEV_UDP_TUNNEL_PUSH_INFO,
3001 	NETDEV_UDP_TUNNEL_DROP_INFO,
3002 	NETDEV_CHANGE_TX_QUEUE_LEN,
3003 	NETDEV_CVLAN_FILTER_PUSH_INFO,
3004 	NETDEV_CVLAN_FILTER_DROP_INFO,
3005 	NETDEV_SVLAN_FILTER_PUSH_INFO,
3006 	NETDEV_SVLAN_FILTER_DROP_INFO,
3007 	NETDEV_OFFLOAD_XSTATS_ENABLE,
3008 	NETDEV_OFFLOAD_XSTATS_DISABLE,
3009 	NETDEV_OFFLOAD_XSTATS_REPORT_USED,
3010 	NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
3011 	NETDEV_XDP_FEAT_CHANGE,
3012 };
3013 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
3014 
3015 int register_netdevice_notifier(struct notifier_block *nb);
3016 int unregister_netdevice_notifier(struct notifier_block *nb);
3017 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
3018 int unregister_netdevice_notifier_net(struct net *net,
3019 				      struct notifier_block *nb);
3020 int register_netdevice_notifier_dev_net(struct net_device *dev,
3021 					struct notifier_block *nb,
3022 					struct netdev_net_notifier *nn);
3023 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
3024 					  struct notifier_block *nb,
3025 					  struct netdev_net_notifier *nn);
3026 
3027 struct netdev_notifier_info {
3028 	struct net_device	*dev;
3029 	struct netlink_ext_ack	*extack;
3030 };
3031 
3032 struct netdev_notifier_info_ext {
3033 	struct netdev_notifier_info info; /* must be first */
3034 	union {
3035 		u32 mtu;
3036 	} ext;
3037 };
3038 
3039 struct netdev_notifier_change_info {
3040 	struct netdev_notifier_info info; /* must be first */
3041 	unsigned int flags_changed;
3042 };
3043 
3044 struct netdev_notifier_changeupper_info {
3045 	struct netdev_notifier_info info; /* must be first */
3046 	struct net_device *upper_dev; /* new upper dev */
3047 	bool master; /* is upper dev master */
3048 	bool linking; /* is the notification for link or unlink */
3049 	void *upper_info; /* upper dev info */
3050 };
3051 
3052 struct netdev_notifier_changelowerstate_info {
3053 	struct netdev_notifier_info info; /* must be first */
3054 	void *lower_state_info; /* is lower dev state */
3055 };
3056 
3057 struct netdev_notifier_pre_changeaddr_info {
3058 	struct netdev_notifier_info info; /* must be first */
3059 	const unsigned char *dev_addr;
3060 };
3061 
3062 enum netdev_offload_xstats_type {
3063 	NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
3064 };
3065 
3066 struct netdev_notifier_offload_xstats_info {
3067 	struct netdev_notifier_info info; /* must be first */
3068 	enum netdev_offload_xstats_type type;
3069 
3070 	union {
3071 		/* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
3072 		struct netdev_notifier_offload_xstats_rd *report_delta;
3073 		/* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
3074 		struct netdev_notifier_offload_xstats_ru *report_used;
3075 	};
3076 };
3077 
3078 int netdev_offload_xstats_enable(struct net_device *dev,
3079 				 enum netdev_offload_xstats_type type,
3080 				 struct netlink_ext_ack *extack);
3081 int netdev_offload_xstats_disable(struct net_device *dev,
3082 				  enum netdev_offload_xstats_type type);
3083 bool netdev_offload_xstats_enabled(const struct net_device *dev,
3084 				   enum netdev_offload_xstats_type type);
3085 int netdev_offload_xstats_get(struct net_device *dev,
3086 			      enum netdev_offload_xstats_type type,
3087 			      struct rtnl_hw_stats64 *stats, bool *used,
3088 			      struct netlink_ext_ack *extack);
3089 void
3090 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3091 				   const struct rtnl_hw_stats64 *stats);
3092 void
3093 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3094 void netdev_offload_xstats_push_delta(struct net_device *dev,
3095 				      enum netdev_offload_xstats_type type,
3096 				      const struct rtnl_hw_stats64 *stats);
3097 
3098 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3099 					     struct net_device *dev)
3100 {
3101 	info->dev = dev;
3102 	info->extack = NULL;
3103 }
3104 
3105 static inline struct net_device *
3106 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3107 {
3108 	return info->dev;
3109 }
3110 
3111 static inline struct netlink_ext_ack *
3112 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3113 {
3114 	return info->extack;
3115 }
3116 
3117 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3118 int call_netdevice_notifiers_info(unsigned long val,
3119 				  struct netdev_notifier_info *info);
3120 
3121 #define for_each_netdev(net, d)		\
3122 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3123 #define for_each_netdev_reverse(net, d)	\
3124 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3125 #define for_each_netdev_rcu(net, d)		\
3126 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3127 #define for_each_netdev_safe(net, d, n)	\
3128 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3129 #define for_each_netdev_continue(net, d)		\
3130 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3131 #define for_each_netdev_continue_reverse(net, d)		\
3132 		list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3133 						     dev_list)
3134 #define for_each_netdev_continue_rcu(net, d)		\
3135 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3136 #define for_each_netdev_in_bond_rcu(bond, slave)	\
3137 		for_each_netdev_rcu(&init_net, slave)	\
3138 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3139 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
3140 
3141 #define for_each_netdev_dump(net, d, ifindex)				\
3142 	for (; (d = xa_find(&(net)->dev_by_index, &ifindex,		\
3143 			    ULONG_MAX, XA_PRESENT)); ifindex++)
3144 
3145 static inline struct net_device *next_net_device(struct net_device *dev)
3146 {
3147 	struct list_head *lh;
3148 	struct net *net;
3149 
3150 	net = dev_net(dev);
3151 	lh = dev->dev_list.next;
3152 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3153 }
3154 
3155 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3156 {
3157 	struct list_head *lh;
3158 	struct net *net;
3159 
3160 	net = dev_net(dev);
3161 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3162 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3163 }
3164 
3165 static inline struct net_device *first_net_device(struct net *net)
3166 {
3167 	return list_empty(&net->dev_base_head) ? NULL :
3168 		net_device_entry(net->dev_base_head.next);
3169 }
3170 
3171 static inline struct net_device *first_net_device_rcu(struct net *net)
3172 {
3173 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3174 
3175 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3176 }
3177 
3178 int netdev_boot_setup_check(struct net_device *dev);
3179 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3180 				       const char *hwaddr);
3181 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3182 void dev_add_pack(struct packet_type *pt);
3183 void dev_remove_pack(struct packet_type *pt);
3184 void __dev_remove_pack(struct packet_type *pt);
3185 void dev_add_offload(struct packet_offload *po);
3186 void dev_remove_offload(struct packet_offload *po);
3187 
3188 int dev_get_iflink(const struct net_device *dev);
3189 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3190 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3191 			  struct net_device_path_stack *stack);
3192 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3193 				      unsigned short mask);
3194 struct net_device *dev_get_by_name(struct net *net, const char *name);
3195 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3196 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3197 bool netdev_name_in_use(struct net *net, const char *name);
3198 int dev_alloc_name(struct net_device *dev, const char *name);
3199 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3200 void dev_close(struct net_device *dev);
3201 void dev_close_many(struct list_head *head, bool unlink);
3202 void dev_disable_lro(struct net_device *dev);
3203 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3204 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3205 		     struct net_device *sb_dev);
3206 
3207 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3208 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3209 
3210 static inline int dev_queue_xmit(struct sk_buff *skb)
3211 {
3212 	return __dev_queue_xmit(skb, NULL);
3213 }
3214 
3215 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3216 				       struct net_device *sb_dev)
3217 {
3218 	return __dev_queue_xmit(skb, sb_dev);
3219 }
3220 
3221 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3222 {
3223 	int ret;
3224 
3225 	ret = __dev_direct_xmit(skb, queue_id);
3226 	if (!dev_xmit_complete(ret))
3227 		kfree_skb(skb);
3228 	return ret;
3229 }
3230 
3231 int register_netdevice(struct net_device *dev);
3232 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3233 void unregister_netdevice_many(struct list_head *head);
3234 static inline void unregister_netdevice(struct net_device *dev)
3235 {
3236 	unregister_netdevice_queue(dev, NULL);
3237 }
3238 
3239 int netdev_refcnt_read(const struct net_device *dev);
3240 void free_netdev(struct net_device *dev);
3241 void init_dummy_netdev(struct net_device *dev);
3242 
3243 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3244 					 struct sk_buff *skb,
3245 					 bool all_slaves);
3246 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3247 					    struct sock *sk);
3248 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3249 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3250 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3251 				       netdevice_tracker *tracker, gfp_t gfp);
3252 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3253 				      netdevice_tracker *tracker, gfp_t gfp);
3254 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3255 void netdev_copy_name(struct net_device *dev, char *name);
3256 
3257 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3258 				  unsigned short type,
3259 				  const void *daddr, const void *saddr,
3260 				  unsigned int len)
3261 {
3262 	if (!dev->header_ops || !dev->header_ops->create)
3263 		return 0;
3264 
3265 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3266 }
3267 
3268 static inline int dev_parse_header(const struct sk_buff *skb,
3269 				   unsigned char *haddr)
3270 {
3271 	const struct net_device *dev = skb->dev;
3272 
3273 	if (!dev->header_ops || !dev->header_ops->parse)
3274 		return 0;
3275 	return dev->header_ops->parse(skb, haddr);
3276 }
3277 
3278 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3279 {
3280 	const struct net_device *dev = skb->dev;
3281 
3282 	if (!dev->header_ops || !dev->header_ops->parse_protocol)
3283 		return 0;
3284 	return dev->header_ops->parse_protocol(skb);
3285 }
3286 
3287 /* ll_header must have at least hard_header_len allocated */
3288 static inline bool dev_validate_header(const struct net_device *dev,
3289 				       char *ll_header, int len)
3290 {
3291 	if (likely(len >= dev->hard_header_len))
3292 		return true;
3293 	if (len < dev->min_header_len)
3294 		return false;
3295 
3296 	if (capable(CAP_SYS_RAWIO)) {
3297 		memset(ll_header + len, 0, dev->hard_header_len - len);
3298 		return true;
3299 	}
3300 
3301 	if (dev->header_ops && dev->header_ops->validate)
3302 		return dev->header_ops->validate(ll_header, len);
3303 
3304 	return false;
3305 }
3306 
3307 static inline bool dev_has_header(const struct net_device *dev)
3308 {
3309 	return dev->header_ops && dev->header_ops->create;
3310 }
3311 
3312 /*
3313  * Incoming packets are placed on per-CPU queues
3314  */
3315 struct softnet_data {
3316 	struct list_head	poll_list;
3317 	struct sk_buff_head	process_queue;
3318 	local_lock_t		process_queue_bh_lock;
3319 
3320 	/* stats */
3321 	unsigned int		processed;
3322 	unsigned int		time_squeeze;
3323 #ifdef CONFIG_RPS
3324 	struct softnet_data	*rps_ipi_list;
3325 #endif
3326 
3327 	unsigned int		received_rps;
3328 	bool			in_net_rx_action;
3329 	bool			in_napi_threaded_poll;
3330 
3331 #ifdef CONFIG_NET_FLOW_LIMIT
3332 	struct sd_flow_limit __rcu *flow_limit;
3333 #endif
3334 	struct Qdisc		*output_queue;
3335 	struct Qdisc		**output_queue_tailp;
3336 	struct sk_buff		*completion_queue;
3337 #ifdef CONFIG_XFRM_OFFLOAD
3338 	struct sk_buff_head	xfrm_backlog;
3339 #endif
3340 	/* written and read only by owning cpu: */
3341 	struct netdev_xmit xmit;
3342 #ifdef CONFIG_RPS
3343 	/* input_queue_head should be written by cpu owning this struct,
3344 	 * and only read by other cpus. Worth using a cache line.
3345 	 */
3346 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
3347 
3348 	/* Elements below can be accessed between CPUs for RPS/RFS */
3349 	call_single_data_t	csd ____cacheline_aligned_in_smp;
3350 	struct softnet_data	*rps_ipi_next;
3351 	unsigned int		cpu;
3352 	unsigned int		input_queue_tail;
3353 #endif
3354 	struct sk_buff_head	input_pkt_queue;
3355 	struct napi_struct	backlog;
3356 
3357 	atomic_t		dropped ____cacheline_aligned_in_smp;
3358 
3359 	/* Another possibly contended cache line */
3360 	spinlock_t		defer_lock ____cacheline_aligned_in_smp;
3361 	int			defer_count;
3362 	int			defer_ipi_scheduled;
3363 	struct sk_buff		*defer_list;
3364 	call_single_data_t	defer_csd;
3365 };
3366 
3367 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3368 DECLARE_PER_CPU(struct page_pool *, system_page_pool);
3369 
3370 #ifndef CONFIG_PREEMPT_RT
3371 static inline int dev_recursion_level(void)
3372 {
3373 	return this_cpu_read(softnet_data.xmit.recursion);
3374 }
3375 #else
3376 static inline int dev_recursion_level(void)
3377 {
3378 	return current->net_xmit.recursion;
3379 }
3380 
3381 #endif
3382 
3383 void __netif_schedule(struct Qdisc *q);
3384 void netif_schedule_queue(struct netdev_queue *txq);
3385 
3386 static inline void netif_tx_schedule_all(struct net_device *dev)
3387 {
3388 	unsigned int i;
3389 
3390 	for (i = 0; i < dev->num_tx_queues; i++)
3391 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
3392 }
3393 
3394 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3395 {
3396 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3397 }
3398 
3399 /**
3400  *	netif_start_queue - allow transmit
3401  *	@dev: network device
3402  *
3403  *	Allow upper layers to call the device hard_start_xmit routine.
3404  */
3405 static inline void netif_start_queue(struct net_device *dev)
3406 {
3407 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3408 }
3409 
3410 static inline void netif_tx_start_all_queues(struct net_device *dev)
3411 {
3412 	unsigned int i;
3413 
3414 	for (i = 0; i < dev->num_tx_queues; i++) {
3415 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3416 		netif_tx_start_queue(txq);
3417 	}
3418 }
3419 
3420 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3421 
3422 /**
3423  *	netif_wake_queue - restart transmit
3424  *	@dev: network device
3425  *
3426  *	Allow upper layers to call the device hard_start_xmit routine.
3427  *	Used for flow control when transmit resources are available.
3428  */
3429 static inline void netif_wake_queue(struct net_device *dev)
3430 {
3431 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3432 }
3433 
3434 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3435 {
3436 	unsigned int i;
3437 
3438 	for (i = 0; i < dev->num_tx_queues; i++) {
3439 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3440 		netif_tx_wake_queue(txq);
3441 	}
3442 }
3443 
3444 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3445 {
3446 	/* Paired with READ_ONCE() from dev_watchdog() */
3447 	WRITE_ONCE(dev_queue->trans_start, jiffies);
3448 
3449 	/* This barrier is paired with smp_mb() from dev_watchdog() */
3450 	smp_mb__before_atomic();
3451 
3452 	/* Must be an atomic op see netif_txq_try_stop() */
3453 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3454 }
3455 
3456 /**
3457  *	netif_stop_queue - stop transmitted packets
3458  *	@dev: network device
3459  *
3460  *	Stop upper layers calling the device hard_start_xmit routine.
3461  *	Used for flow control when transmit resources are unavailable.
3462  */
3463 static inline void netif_stop_queue(struct net_device *dev)
3464 {
3465 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3466 }
3467 
3468 void netif_tx_stop_all_queues(struct net_device *dev);
3469 
3470 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3471 {
3472 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3473 }
3474 
3475 /**
3476  *	netif_queue_stopped - test if transmit queue is flowblocked
3477  *	@dev: network device
3478  *
3479  *	Test if transmit queue on device is currently unable to send.
3480  */
3481 static inline bool netif_queue_stopped(const struct net_device *dev)
3482 {
3483 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3484 }
3485 
3486 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3487 {
3488 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3489 }
3490 
3491 static inline bool
3492 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3493 {
3494 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3495 }
3496 
3497 static inline bool
3498 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3499 {
3500 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3501 }
3502 
3503 /**
3504  *	netdev_queue_set_dql_min_limit - set dql minimum limit
3505  *	@dev_queue: pointer to transmit queue
3506  *	@min_limit: dql minimum limit
3507  *
3508  * Forces xmit_more() to return true until the minimum threshold
3509  * defined by @min_limit is reached (or until the tx queue is
3510  * empty). Warning: to be use with care, misuse will impact the
3511  * latency.
3512  */
3513 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3514 						  unsigned int min_limit)
3515 {
3516 #ifdef CONFIG_BQL
3517 	dev_queue->dql.min_limit = min_limit;
3518 #endif
3519 }
3520 
3521 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq)
3522 {
3523 #ifdef CONFIG_BQL
3524 	/* Non-BQL migrated drivers will return 0, too. */
3525 	return dql_avail(&txq->dql);
3526 #else
3527 	return 0;
3528 #endif
3529 }
3530 
3531 /**
3532  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3533  *	@dev_queue: pointer to transmit queue
3534  *
3535  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3536  * to give appropriate hint to the CPU.
3537  */
3538 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3539 {
3540 #ifdef CONFIG_BQL
3541 	prefetchw(&dev_queue->dql.num_queued);
3542 #endif
3543 }
3544 
3545 /**
3546  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3547  *	@dev_queue: pointer to transmit queue
3548  *
3549  * BQL enabled drivers might use this helper in their TX completion path,
3550  * to give appropriate hint to the CPU.
3551  */
3552 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3553 {
3554 #ifdef CONFIG_BQL
3555 	prefetchw(&dev_queue->dql.limit);
3556 #endif
3557 }
3558 
3559 /**
3560  *	netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3561  *	@dev_queue: network device queue
3562  *	@bytes: number of bytes queued to the device queue
3563  *
3564  *	Report the number of bytes queued for sending/completion to the network
3565  *	device hardware queue. @bytes should be a good approximation and should
3566  *	exactly match netdev_completed_queue() @bytes.
3567  *	This is typically called once per packet, from ndo_start_xmit().
3568  */
3569 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3570 					unsigned int bytes)
3571 {
3572 #ifdef CONFIG_BQL
3573 	dql_queued(&dev_queue->dql, bytes);
3574 
3575 	if (likely(dql_avail(&dev_queue->dql) >= 0))
3576 		return;
3577 
3578 	/* Paired with READ_ONCE() from dev_watchdog() */
3579 	WRITE_ONCE(dev_queue->trans_start, jiffies);
3580 
3581 	/* This barrier is paired with smp_mb() from dev_watchdog() */
3582 	smp_mb__before_atomic();
3583 
3584 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3585 
3586 	/*
3587 	 * The XOFF flag must be set before checking the dql_avail below,
3588 	 * because in netdev_tx_completed_queue we update the dql_completed
3589 	 * before checking the XOFF flag.
3590 	 */
3591 	smp_mb__after_atomic();
3592 
3593 	/* check again in case another CPU has just made room avail */
3594 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3595 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3596 #endif
3597 }
3598 
3599 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3600  * that they should not test BQL status themselves.
3601  * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3602  * skb of a batch.
3603  * Returns true if the doorbell must be used to kick the NIC.
3604  */
3605 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3606 					  unsigned int bytes,
3607 					  bool xmit_more)
3608 {
3609 	if (xmit_more) {
3610 #ifdef CONFIG_BQL
3611 		dql_queued(&dev_queue->dql, bytes);
3612 #endif
3613 		return netif_tx_queue_stopped(dev_queue);
3614 	}
3615 	netdev_tx_sent_queue(dev_queue, bytes);
3616 	return true;
3617 }
3618 
3619 /**
3620  *	netdev_sent_queue - report the number of bytes queued to hardware
3621  *	@dev: network device
3622  *	@bytes: number of bytes queued to the hardware device queue
3623  *
3624  *	Report the number of bytes queued for sending/completion to the network
3625  *	device hardware queue#0. @bytes should be a good approximation and should
3626  *	exactly match netdev_completed_queue() @bytes.
3627  *	This is typically called once per packet, from ndo_start_xmit().
3628  */
3629 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3630 {
3631 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3632 }
3633 
3634 static inline bool __netdev_sent_queue(struct net_device *dev,
3635 				       unsigned int bytes,
3636 				       bool xmit_more)
3637 {
3638 	return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3639 				      xmit_more);
3640 }
3641 
3642 /**
3643  *	netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3644  *	@dev_queue: network device queue
3645  *	@pkts: number of packets (currently ignored)
3646  *	@bytes: number of bytes dequeued from the device queue
3647  *
3648  *	Must be called at most once per TX completion round (and not per
3649  *	individual packet), so that BQL can adjust its limits appropriately.
3650  */
3651 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3652 					     unsigned int pkts, unsigned int bytes)
3653 {
3654 #ifdef CONFIG_BQL
3655 	if (unlikely(!bytes))
3656 		return;
3657 
3658 	dql_completed(&dev_queue->dql, bytes);
3659 
3660 	/*
3661 	 * Without the memory barrier there is a small possibility that
3662 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3663 	 * be stopped forever
3664 	 */
3665 	smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3666 
3667 	if (unlikely(dql_avail(&dev_queue->dql) < 0))
3668 		return;
3669 
3670 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3671 		netif_schedule_queue(dev_queue);
3672 #endif
3673 }
3674 
3675 /**
3676  * 	netdev_completed_queue - report bytes and packets completed by device
3677  * 	@dev: network device
3678  * 	@pkts: actual number of packets sent over the medium
3679  * 	@bytes: actual number of bytes sent over the medium
3680  *
3681  * 	Report the number of bytes and packets transmitted by the network device
3682  * 	hardware queue over the physical medium, @bytes must exactly match the
3683  * 	@bytes amount passed to netdev_sent_queue()
3684  */
3685 static inline void netdev_completed_queue(struct net_device *dev,
3686 					  unsigned int pkts, unsigned int bytes)
3687 {
3688 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3689 }
3690 
3691 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3692 {
3693 #ifdef CONFIG_BQL
3694 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3695 	dql_reset(&q->dql);
3696 #endif
3697 }
3698 
3699 /**
3700  * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue
3701  * @dev: network device
3702  * @qid: stack index of the queue to reset
3703  */
3704 static inline void netdev_tx_reset_subqueue(const struct net_device *dev,
3705 					    u32 qid)
3706 {
3707 	netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid));
3708 }
3709 
3710 /**
3711  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3712  * 	@dev_queue: network device
3713  *
3714  * 	Reset the bytes and packet count of a network device and clear the
3715  * 	software flow control OFF bit for this network device
3716  */
3717 static inline void netdev_reset_queue(struct net_device *dev_queue)
3718 {
3719 	netdev_tx_reset_subqueue(dev_queue, 0);
3720 }
3721 
3722 /**
3723  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3724  * 	@dev: network device
3725  * 	@queue_index: given tx queue index
3726  *
3727  * 	Returns 0 if given tx queue index >= number of device tx queues,
3728  * 	otherwise returns the originally passed tx queue index.
3729  */
3730 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3731 {
3732 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3733 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3734 				     dev->name, queue_index,
3735 				     dev->real_num_tx_queues);
3736 		return 0;
3737 	}
3738 
3739 	return queue_index;
3740 }
3741 
3742 /**
3743  *	netif_running - test if up
3744  *	@dev: network device
3745  *
3746  *	Test if the device has been brought up.
3747  */
3748 static inline bool netif_running(const struct net_device *dev)
3749 {
3750 	return test_bit(__LINK_STATE_START, &dev->state);
3751 }
3752 
3753 /*
3754  * Routines to manage the subqueues on a device.  We only need start,
3755  * stop, and a check if it's stopped.  All other device management is
3756  * done at the overall netdevice level.
3757  * Also test the device if we're multiqueue.
3758  */
3759 
3760 /**
3761  *	netif_start_subqueue - allow sending packets on subqueue
3762  *	@dev: network device
3763  *	@queue_index: sub queue index
3764  *
3765  * Start individual transmit queue of a device with multiple transmit queues.
3766  */
3767 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3768 {
3769 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3770 
3771 	netif_tx_start_queue(txq);
3772 }
3773 
3774 /**
3775  *	netif_stop_subqueue - stop sending packets on subqueue
3776  *	@dev: network device
3777  *	@queue_index: sub queue index
3778  *
3779  * Stop individual transmit queue of a device with multiple transmit queues.
3780  */
3781 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3782 {
3783 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3784 	netif_tx_stop_queue(txq);
3785 }
3786 
3787 /**
3788  *	__netif_subqueue_stopped - test status of subqueue
3789  *	@dev: network device
3790  *	@queue_index: sub queue index
3791  *
3792  * Check individual transmit queue of a device with multiple transmit queues.
3793  */
3794 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3795 					    u16 queue_index)
3796 {
3797 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3798 
3799 	return netif_tx_queue_stopped(txq);
3800 }
3801 
3802 /**
3803  *	netif_subqueue_stopped - test status of subqueue
3804  *	@dev: network device
3805  *	@skb: sub queue buffer pointer
3806  *
3807  * Check individual transmit queue of a device with multiple transmit queues.
3808  */
3809 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3810 					  struct sk_buff *skb)
3811 {
3812 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3813 }
3814 
3815 /**
3816  *	netif_wake_subqueue - allow sending packets on subqueue
3817  *	@dev: network device
3818  *	@queue_index: sub queue index
3819  *
3820  * Resume individual transmit queue of a device with multiple transmit queues.
3821  */
3822 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3823 {
3824 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3825 
3826 	netif_tx_wake_queue(txq);
3827 }
3828 
3829 #ifdef CONFIG_XPS
3830 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3831 			u16 index);
3832 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3833 			  u16 index, enum xps_map_type type);
3834 
3835 /**
3836  *	netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3837  *	@j: CPU/Rx queue index
3838  *	@mask: bitmask of all cpus/rx queues
3839  *	@nr_bits: number of bits in the bitmask
3840  *
3841  * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3842  */
3843 static inline bool netif_attr_test_mask(unsigned long j,
3844 					const unsigned long *mask,
3845 					unsigned int nr_bits)
3846 {
3847 	cpu_max_bits_warn(j, nr_bits);
3848 	return test_bit(j, mask);
3849 }
3850 
3851 /**
3852  *	netif_attr_test_online - Test for online CPU/Rx queue
3853  *	@j: CPU/Rx queue index
3854  *	@online_mask: bitmask for CPUs/Rx queues that are online
3855  *	@nr_bits: number of bits in the bitmask
3856  *
3857  * Returns: true if a CPU/Rx queue is online.
3858  */
3859 static inline bool netif_attr_test_online(unsigned long j,
3860 					  const unsigned long *online_mask,
3861 					  unsigned int nr_bits)
3862 {
3863 	cpu_max_bits_warn(j, nr_bits);
3864 
3865 	if (online_mask)
3866 		return test_bit(j, online_mask);
3867 
3868 	return (j < nr_bits);
3869 }
3870 
3871 /**
3872  *	netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3873  *	@n: CPU/Rx queue index
3874  *	@srcp: the cpumask/Rx queue mask pointer
3875  *	@nr_bits: number of bits in the bitmask
3876  *
3877  * Returns: next (after n) CPU/Rx queue index in the mask;
3878  * >= nr_bits if no further CPUs/Rx queues set.
3879  */
3880 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3881 					       unsigned int nr_bits)
3882 {
3883 	/* -1 is a legal arg here. */
3884 	if (n != -1)
3885 		cpu_max_bits_warn(n, nr_bits);
3886 
3887 	if (srcp)
3888 		return find_next_bit(srcp, nr_bits, n + 1);
3889 
3890 	return n + 1;
3891 }
3892 
3893 /**
3894  *	netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3895  *	@n: CPU/Rx queue index
3896  *	@src1p: the first CPUs/Rx queues mask pointer
3897  *	@src2p: the second CPUs/Rx queues mask pointer
3898  *	@nr_bits: number of bits in the bitmask
3899  *
3900  * Returns: next (after n) CPU/Rx queue index set in both masks;
3901  * >= nr_bits if no further CPUs/Rx queues set in both.
3902  */
3903 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3904 					  const unsigned long *src2p,
3905 					  unsigned int nr_bits)
3906 {
3907 	/* -1 is a legal arg here. */
3908 	if (n != -1)
3909 		cpu_max_bits_warn(n, nr_bits);
3910 
3911 	if (src1p && src2p)
3912 		return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3913 	else if (src1p)
3914 		return find_next_bit(src1p, nr_bits, n + 1);
3915 	else if (src2p)
3916 		return find_next_bit(src2p, nr_bits, n + 1);
3917 
3918 	return n + 1;
3919 }
3920 #else
3921 static inline int netif_set_xps_queue(struct net_device *dev,
3922 				      const struct cpumask *mask,
3923 				      u16 index)
3924 {
3925 	return 0;
3926 }
3927 
3928 static inline int __netif_set_xps_queue(struct net_device *dev,
3929 					const unsigned long *mask,
3930 					u16 index, enum xps_map_type type)
3931 {
3932 	return 0;
3933 }
3934 #endif
3935 
3936 /**
3937  *	netif_is_multiqueue - test if device has multiple transmit queues
3938  *	@dev: network device
3939  *
3940  * Check if device has multiple transmit queues
3941  */
3942 static inline bool netif_is_multiqueue(const struct net_device *dev)
3943 {
3944 	return dev->num_tx_queues > 1;
3945 }
3946 
3947 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3948 
3949 #ifdef CONFIG_SYSFS
3950 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3951 #else
3952 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3953 						unsigned int rxqs)
3954 {
3955 	dev->real_num_rx_queues = rxqs;
3956 	return 0;
3957 }
3958 #endif
3959 int netif_set_real_num_queues(struct net_device *dev,
3960 			      unsigned int txq, unsigned int rxq);
3961 
3962 int netif_get_num_default_rss_queues(void);
3963 
3964 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3965 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3966 
3967 /*
3968  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3969  * interrupt context or with hardware interrupts being disabled.
3970  * (in_hardirq() || irqs_disabled())
3971  *
3972  * We provide four helpers that can be used in following contexts :
3973  *
3974  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3975  *  replacing kfree_skb(skb)
3976  *
3977  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3978  *  Typically used in place of consume_skb(skb) in TX completion path
3979  *
3980  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3981  *  replacing kfree_skb(skb)
3982  *
3983  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3984  *  and consumed a packet. Used in place of consume_skb(skb)
3985  */
3986 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3987 {
3988 	dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3989 }
3990 
3991 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3992 {
3993 	dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3994 }
3995 
3996 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3997 {
3998 	dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3999 }
4000 
4001 static inline void dev_consume_skb_any(struct sk_buff *skb)
4002 {
4003 	dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
4004 }
4005 
4006 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
4007 			     const struct bpf_prog *xdp_prog);
4008 void generic_xdp_tx(struct sk_buff *skb, const struct bpf_prog *xdp_prog);
4009 int do_xdp_generic(const struct bpf_prog *xdp_prog, struct sk_buff **pskb);
4010 int netif_rx(struct sk_buff *skb);
4011 int __netif_rx(struct sk_buff *skb);
4012 
4013 int netif_receive_skb(struct sk_buff *skb);
4014 int netif_receive_skb_core(struct sk_buff *skb);
4015 void netif_receive_skb_list_internal(struct list_head *head);
4016 void netif_receive_skb_list(struct list_head *head);
4017 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
4018 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
4019 struct sk_buff *napi_get_frags(struct napi_struct *napi);
4020 void napi_get_frags_check(struct napi_struct *napi);
4021 gro_result_t napi_gro_frags(struct napi_struct *napi);
4022 
4023 static inline void napi_free_frags(struct napi_struct *napi)
4024 {
4025 	kfree_skb(napi->skb);
4026 	napi->skb = NULL;
4027 }
4028 
4029 bool netdev_is_rx_handler_busy(struct net_device *dev);
4030 int netdev_rx_handler_register(struct net_device *dev,
4031 			       rx_handler_func_t *rx_handler,
4032 			       void *rx_handler_data);
4033 void netdev_rx_handler_unregister(struct net_device *dev);
4034 
4035 bool dev_valid_name(const char *name);
4036 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
4037 {
4038 	return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
4039 }
4040 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
4041 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
4042 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4043 		void __user *data, bool *need_copyout);
4044 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4045 int generic_hwtstamp_get_lower(struct net_device *dev,
4046 			       struct kernel_hwtstamp_config *kernel_cfg);
4047 int generic_hwtstamp_set_lower(struct net_device *dev,
4048 			       struct kernel_hwtstamp_config *kernel_cfg,
4049 			       struct netlink_ext_ack *extack);
4050 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4051 unsigned int dev_get_flags(const struct net_device *);
4052 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4053 		       struct netlink_ext_ack *extack);
4054 int dev_change_flags(struct net_device *dev, unsigned int flags,
4055 		     struct netlink_ext_ack *extack);
4056 int dev_set_alias(struct net_device *, const char *, size_t);
4057 int dev_get_alias(const struct net_device *, char *, size_t);
4058 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4059 			       const char *pat, int new_ifindex);
4060 static inline
4061 int dev_change_net_namespace(struct net_device *dev, struct net *net,
4062 			     const char *pat)
4063 {
4064 	return __dev_change_net_namespace(dev, net, pat, 0);
4065 }
4066 int __dev_set_mtu(struct net_device *, int);
4067 int dev_set_mtu(struct net_device *, int);
4068 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4069 			      struct netlink_ext_ack *extack);
4070 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4071 			struct netlink_ext_ack *extack);
4072 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4073 			     struct netlink_ext_ack *extack);
4074 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4075 int dev_get_port_parent_id(struct net_device *dev,
4076 			   struct netdev_phys_item_id *ppid, bool recurse);
4077 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4078 
4079 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4080 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4081 				    struct netdev_queue *txq, int *ret);
4082 
4083 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4084 u8 dev_xdp_prog_count(struct net_device *dev);
4085 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf);
4086 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4087 
4088 u32 dev_get_min_mp_channel_count(const struct net_device *dev);
4089 
4090 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4091 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4092 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4093 bool is_skb_forwardable(const struct net_device *dev,
4094 			const struct sk_buff *skb);
4095 
4096 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4097 						 const struct sk_buff *skb,
4098 						 const bool check_mtu)
4099 {
4100 	const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4101 	unsigned int len;
4102 
4103 	if (!(dev->flags & IFF_UP))
4104 		return false;
4105 
4106 	if (!check_mtu)
4107 		return true;
4108 
4109 	len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4110 	if (skb->len <= len)
4111 		return true;
4112 
4113 	/* if TSO is enabled, we don't care about the length as the packet
4114 	 * could be forwarded without being segmented before
4115 	 */
4116 	if (skb_is_gso(skb))
4117 		return true;
4118 
4119 	return false;
4120 }
4121 
4122 void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4123 
4124 #define DEV_CORE_STATS_INC(FIELD)						\
4125 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev)		\
4126 {										\
4127 	netdev_core_stats_inc(dev,						\
4128 			offsetof(struct net_device_core_stats, FIELD));		\
4129 }
4130 DEV_CORE_STATS_INC(rx_dropped)
4131 DEV_CORE_STATS_INC(tx_dropped)
4132 DEV_CORE_STATS_INC(rx_nohandler)
4133 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4134 #undef DEV_CORE_STATS_INC
4135 
4136 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4137 					       struct sk_buff *skb,
4138 					       const bool check_mtu)
4139 {
4140 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4141 	    unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4142 		dev_core_stats_rx_dropped_inc(dev);
4143 		kfree_skb(skb);
4144 		return NET_RX_DROP;
4145 	}
4146 
4147 	skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4148 	skb->priority = 0;
4149 	return 0;
4150 }
4151 
4152 bool dev_nit_active(struct net_device *dev);
4153 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4154 
4155 static inline void __dev_put(struct net_device *dev)
4156 {
4157 	if (dev) {
4158 #ifdef CONFIG_PCPU_DEV_REFCNT
4159 		this_cpu_dec(*dev->pcpu_refcnt);
4160 #else
4161 		refcount_dec(&dev->dev_refcnt);
4162 #endif
4163 	}
4164 }
4165 
4166 static inline void __dev_hold(struct net_device *dev)
4167 {
4168 	if (dev) {
4169 #ifdef CONFIG_PCPU_DEV_REFCNT
4170 		this_cpu_inc(*dev->pcpu_refcnt);
4171 #else
4172 		refcount_inc(&dev->dev_refcnt);
4173 #endif
4174 	}
4175 }
4176 
4177 static inline void __netdev_tracker_alloc(struct net_device *dev,
4178 					  netdevice_tracker *tracker,
4179 					  gfp_t gfp)
4180 {
4181 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4182 	ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4183 #endif
4184 }
4185 
4186 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4187  * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4188  */
4189 static inline void netdev_tracker_alloc(struct net_device *dev,
4190 					netdevice_tracker *tracker, gfp_t gfp)
4191 {
4192 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4193 	refcount_dec(&dev->refcnt_tracker.no_tracker);
4194 	__netdev_tracker_alloc(dev, tracker, gfp);
4195 #endif
4196 }
4197 
4198 static inline void netdev_tracker_free(struct net_device *dev,
4199 				       netdevice_tracker *tracker)
4200 {
4201 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4202 	ref_tracker_free(&dev->refcnt_tracker, tracker);
4203 #endif
4204 }
4205 
4206 static inline void netdev_hold(struct net_device *dev,
4207 			       netdevice_tracker *tracker, gfp_t gfp)
4208 {
4209 	if (dev) {
4210 		__dev_hold(dev);
4211 		__netdev_tracker_alloc(dev, tracker, gfp);
4212 	}
4213 }
4214 
4215 static inline void netdev_put(struct net_device *dev,
4216 			      netdevice_tracker *tracker)
4217 {
4218 	if (dev) {
4219 		netdev_tracker_free(dev, tracker);
4220 		__dev_put(dev);
4221 	}
4222 }
4223 
4224 /**
4225  *	dev_hold - get reference to device
4226  *	@dev: network device
4227  *
4228  * Hold reference to device to keep it from being freed.
4229  * Try using netdev_hold() instead.
4230  */
4231 static inline void dev_hold(struct net_device *dev)
4232 {
4233 	netdev_hold(dev, NULL, GFP_ATOMIC);
4234 }
4235 
4236 /**
4237  *	dev_put - release reference to device
4238  *	@dev: network device
4239  *
4240  * Release reference to device to allow it to be freed.
4241  * Try using netdev_put() instead.
4242  */
4243 static inline void dev_put(struct net_device *dev)
4244 {
4245 	netdev_put(dev, NULL);
4246 }
4247 
4248 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T))
4249 
4250 static inline void netdev_ref_replace(struct net_device *odev,
4251 				      struct net_device *ndev,
4252 				      netdevice_tracker *tracker,
4253 				      gfp_t gfp)
4254 {
4255 	if (odev)
4256 		netdev_tracker_free(odev, tracker);
4257 
4258 	__dev_hold(ndev);
4259 	__dev_put(odev);
4260 
4261 	if (ndev)
4262 		__netdev_tracker_alloc(ndev, tracker, gfp);
4263 }
4264 
4265 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4266  * and _off may be called from IRQ context, but it is caller
4267  * who is responsible for serialization of these calls.
4268  *
4269  * The name carrier is inappropriate, these functions should really be
4270  * called netif_lowerlayer_*() because they represent the state of any
4271  * kind of lower layer not just hardware media.
4272  */
4273 void linkwatch_fire_event(struct net_device *dev);
4274 
4275 /**
4276  * linkwatch_sync_dev - sync linkwatch for the given device
4277  * @dev: network device to sync linkwatch for
4278  *
4279  * Sync linkwatch for the given device, removing it from the
4280  * pending work list (if queued).
4281  */
4282 void linkwatch_sync_dev(struct net_device *dev);
4283 
4284 /**
4285  *	netif_carrier_ok - test if carrier present
4286  *	@dev: network device
4287  *
4288  * Check if carrier is present on device
4289  */
4290 static inline bool netif_carrier_ok(const struct net_device *dev)
4291 {
4292 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4293 }
4294 
4295 unsigned long dev_trans_start(struct net_device *dev);
4296 
4297 void netdev_watchdog_up(struct net_device *dev);
4298 
4299 void netif_carrier_on(struct net_device *dev);
4300 void netif_carrier_off(struct net_device *dev);
4301 void netif_carrier_event(struct net_device *dev);
4302 
4303 /**
4304  *	netif_dormant_on - mark device as dormant.
4305  *	@dev: network device
4306  *
4307  * Mark device as dormant (as per RFC2863).
4308  *
4309  * The dormant state indicates that the relevant interface is not
4310  * actually in a condition to pass packets (i.e., it is not 'up') but is
4311  * in a "pending" state, waiting for some external event.  For "on-
4312  * demand" interfaces, this new state identifies the situation where the
4313  * interface is waiting for events to place it in the up state.
4314  */
4315 static inline void netif_dormant_on(struct net_device *dev)
4316 {
4317 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4318 		linkwatch_fire_event(dev);
4319 }
4320 
4321 /**
4322  *	netif_dormant_off - set device as not dormant.
4323  *	@dev: network device
4324  *
4325  * Device is not in dormant state.
4326  */
4327 static inline void netif_dormant_off(struct net_device *dev)
4328 {
4329 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4330 		linkwatch_fire_event(dev);
4331 }
4332 
4333 /**
4334  *	netif_dormant - test if device is dormant
4335  *	@dev: network device
4336  *
4337  * Check if device is dormant.
4338  */
4339 static inline bool netif_dormant(const struct net_device *dev)
4340 {
4341 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
4342 }
4343 
4344 
4345 /**
4346  *	netif_testing_on - mark device as under test.
4347  *	@dev: network device
4348  *
4349  * Mark device as under test (as per RFC2863).
4350  *
4351  * The testing state indicates that some test(s) must be performed on
4352  * the interface. After completion, of the test, the interface state
4353  * will change to up, dormant, or down, as appropriate.
4354  */
4355 static inline void netif_testing_on(struct net_device *dev)
4356 {
4357 	if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4358 		linkwatch_fire_event(dev);
4359 }
4360 
4361 /**
4362  *	netif_testing_off - set device as not under test.
4363  *	@dev: network device
4364  *
4365  * Device is not in testing state.
4366  */
4367 static inline void netif_testing_off(struct net_device *dev)
4368 {
4369 	if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4370 		linkwatch_fire_event(dev);
4371 }
4372 
4373 /**
4374  *	netif_testing - test if device is under test
4375  *	@dev: network device
4376  *
4377  * Check if device is under test
4378  */
4379 static inline bool netif_testing(const struct net_device *dev)
4380 {
4381 	return test_bit(__LINK_STATE_TESTING, &dev->state);
4382 }
4383 
4384 
4385 /**
4386  *	netif_oper_up - test if device is operational
4387  *	@dev: network device
4388  *
4389  * Check if carrier is operational
4390  */
4391 static inline bool netif_oper_up(const struct net_device *dev)
4392 {
4393 	unsigned int operstate = READ_ONCE(dev->operstate);
4394 
4395 	return	operstate == IF_OPER_UP ||
4396 		operstate == IF_OPER_UNKNOWN /* backward compat */;
4397 }
4398 
4399 /**
4400  *	netif_device_present - is device available or removed
4401  *	@dev: network device
4402  *
4403  * Check if device has not been removed from system.
4404  */
4405 static inline bool netif_device_present(const struct net_device *dev)
4406 {
4407 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
4408 }
4409 
4410 void netif_device_detach(struct net_device *dev);
4411 
4412 void netif_device_attach(struct net_device *dev);
4413 
4414 /*
4415  * Network interface message level settings
4416  */
4417 
4418 enum {
4419 	NETIF_MSG_DRV_BIT,
4420 	NETIF_MSG_PROBE_BIT,
4421 	NETIF_MSG_LINK_BIT,
4422 	NETIF_MSG_TIMER_BIT,
4423 	NETIF_MSG_IFDOWN_BIT,
4424 	NETIF_MSG_IFUP_BIT,
4425 	NETIF_MSG_RX_ERR_BIT,
4426 	NETIF_MSG_TX_ERR_BIT,
4427 	NETIF_MSG_TX_QUEUED_BIT,
4428 	NETIF_MSG_INTR_BIT,
4429 	NETIF_MSG_TX_DONE_BIT,
4430 	NETIF_MSG_RX_STATUS_BIT,
4431 	NETIF_MSG_PKTDATA_BIT,
4432 	NETIF_MSG_HW_BIT,
4433 	NETIF_MSG_WOL_BIT,
4434 
4435 	/* When you add a new bit above, update netif_msg_class_names array
4436 	 * in net/ethtool/common.c
4437 	 */
4438 	NETIF_MSG_CLASS_COUNT,
4439 };
4440 /* Both ethtool_ops interface and internal driver implementation use u32 */
4441 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4442 
4443 #define __NETIF_MSG_BIT(bit)	((u32)1 << (bit))
4444 #define __NETIF_MSG(name)	__NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4445 
4446 #define NETIF_MSG_DRV		__NETIF_MSG(DRV)
4447 #define NETIF_MSG_PROBE		__NETIF_MSG(PROBE)
4448 #define NETIF_MSG_LINK		__NETIF_MSG(LINK)
4449 #define NETIF_MSG_TIMER		__NETIF_MSG(TIMER)
4450 #define NETIF_MSG_IFDOWN	__NETIF_MSG(IFDOWN)
4451 #define NETIF_MSG_IFUP		__NETIF_MSG(IFUP)
4452 #define NETIF_MSG_RX_ERR	__NETIF_MSG(RX_ERR)
4453 #define NETIF_MSG_TX_ERR	__NETIF_MSG(TX_ERR)
4454 #define NETIF_MSG_TX_QUEUED	__NETIF_MSG(TX_QUEUED)
4455 #define NETIF_MSG_INTR		__NETIF_MSG(INTR)
4456 #define NETIF_MSG_TX_DONE	__NETIF_MSG(TX_DONE)
4457 #define NETIF_MSG_RX_STATUS	__NETIF_MSG(RX_STATUS)
4458 #define NETIF_MSG_PKTDATA	__NETIF_MSG(PKTDATA)
4459 #define NETIF_MSG_HW		__NETIF_MSG(HW)
4460 #define NETIF_MSG_WOL		__NETIF_MSG(WOL)
4461 
4462 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
4463 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
4464 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
4465 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
4466 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
4467 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
4468 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
4469 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
4470 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4471 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
4472 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
4473 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
4474 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
4475 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
4476 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
4477 
4478 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4479 {
4480 	/* use default */
4481 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4482 		return default_msg_enable_bits;
4483 	if (debug_value == 0)	/* no output */
4484 		return 0;
4485 	/* set low N bits */
4486 	return (1U << debug_value) - 1;
4487 }
4488 
4489 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4490 {
4491 	spin_lock(&txq->_xmit_lock);
4492 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4493 	WRITE_ONCE(txq->xmit_lock_owner, cpu);
4494 }
4495 
4496 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4497 {
4498 	__acquire(&txq->_xmit_lock);
4499 	return true;
4500 }
4501 
4502 static inline void __netif_tx_release(struct netdev_queue *txq)
4503 {
4504 	__release(&txq->_xmit_lock);
4505 }
4506 
4507 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4508 {
4509 	spin_lock_bh(&txq->_xmit_lock);
4510 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4511 	WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4512 }
4513 
4514 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4515 {
4516 	bool ok = spin_trylock(&txq->_xmit_lock);
4517 
4518 	if (likely(ok)) {
4519 		/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4520 		WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4521 	}
4522 	return ok;
4523 }
4524 
4525 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4526 {
4527 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4528 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4529 	spin_unlock(&txq->_xmit_lock);
4530 }
4531 
4532 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4533 {
4534 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4535 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4536 	spin_unlock_bh(&txq->_xmit_lock);
4537 }
4538 
4539 /*
4540  * txq->trans_start can be read locklessly from dev_watchdog()
4541  */
4542 static inline void txq_trans_update(struct netdev_queue *txq)
4543 {
4544 	if (txq->xmit_lock_owner != -1)
4545 		WRITE_ONCE(txq->trans_start, jiffies);
4546 }
4547 
4548 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4549 {
4550 	unsigned long now = jiffies;
4551 
4552 	if (READ_ONCE(txq->trans_start) != now)
4553 		WRITE_ONCE(txq->trans_start, now);
4554 }
4555 
4556 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4557 static inline void netif_trans_update(struct net_device *dev)
4558 {
4559 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4560 
4561 	txq_trans_cond_update(txq);
4562 }
4563 
4564 /**
4565  *	netif_tx_lock - grab network device transmit lock
4566  *	@dev: network device
4567  *
4568  * Get network device transmit lock
4569  */
4570 void netif_tx_lock(struct net_device *dev);
4571 
4572 static inline void netif_tx_lock_bh(struct net_device *dev)
4573 {
4574 	local_bh_disable();
4575 	netif_tx_lock(dev);
4576 }
4577 
4578 void netif_tx_unlock(struct net_device *dev);
4579 
4580 static inline void netif_tx_unlock_bh(struct net_device *dev)
4581 {
4582 	netif_tx_unlock(dev);
4583 	local_bh_enable();
4584 }
4585 
4586 #define HARD_TX_LOCK(dev, txq, cpu) {			\
4587 	if (!(dev)->lltx) {				\
4588 		__netif_tx_lock(txq, cpu);		\
4589 	} else {					\
4590 		__netif_tx_acquire(txq);		\
4591 	}						\
4592 }
4593 
4594 #define HARD_TX_TRYLOCK(dev, txq)			\
4595 	(!(dev)->lltx ?					\
4596 		__netif_tx_trylock(txq) :		\
4597 		__netif_tx_acquire(txq))
4598 
4599 #define HARD_TX_UNLOCK(dev, txq) {			\
4600 	if (!(dev)->lltx) {				\
4601 		__netif_tx_unlock(txq);			\
4602 	} else {					\
4603 		__netif_tx_release(txq);		\
4604 	}						\
4605 }
4606 
4607 static inline void netif_tx_disable(struct net_device *dev)
4608 {
4609 	unsigned int i;
4610 	int cpu;
4611 
4612 	local_bh_disable();
4613 	cpu = smp_processor_id();
4614 	spin_lock(&dev->tx_global_lock);
4615 	for (i = 0; i < dev->num_tx_queues; i++) {
4616 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4617 
4618 		__netif_tx_lock(txq, cpu);
4619 		netif_tx_stop_queue(txq);
4620 		__netif_tx_unlock(txq);
4621 	}
4622 	spin_unlock(&dev->tx_global_lock);
4623 	local_bh_enable();
4624 }
4625 
4626 static inline void netif_addr_lock(struct net_device *dev)
4627 {
4628 	unsigned char nest_level = 0;
4629 
4630 #ifdef CONFIG_LOCKDEP
4631 	nest_level = dev->nested_level;
4632 #endif
4633 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4634 }
4635 
4636 static inline void netif_addr_lock_bh(struct net_device *dev)
4637 {
4638 	unsigned char nest_level = 0;
4639 
4640 #ifdef CONFIG_LOCKDEP
4641 	nest_level = dev->nested_level;
4642 #endif
4643 	local_bh_disable();
4644 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4645 }
4646 
4647 static inline void netif_addr_unlock(struct net_device *dev)
4648 {
4649 	spin_unlock(&dev->addr_list_lock);
4650 }
4651 
4652 static inline void netif_addr_unlock_bh(struct net_device *dev)
4653 {
4654 	spin_unlock_bh(&dev->addr_list_lock);
4655 }
4656 
4657 /*
4658  * dev_addrs walker. Should be used only for read access. Call with
4659  * rcu_read_lock held.
4660  */
4661 #define for_each_dev_addr(dev, ha) \
4662 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4663 
4664 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4665 
4666 void ether_setup(struct net_device *dev);
4667 
4668 /* Allocate dummy net_device */
4669 struct net_device *alloc_netdev_dummy(int sizeof_priv);
4670 
4671 /* Support for loadable net-drivers */
4672 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4673 				    unsigned char name_assign_type,
4674 				    void (*setup)(struct net_device *),
4675 				    unsigned int txqs, unsigned int rxqs);
4676 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4677 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4678 
4679 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4680 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4681 			 count)
4682 
4683 int register_netdev(struct net_device *dev);
4684 void unregister_netdev(struct net_device *dev);
4685 
4686 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4687 
4688 /* General hardware address lists handling functions */
4689 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4690 		   struct netdev_hw_addr_list *from_list, int addr_len);
4691 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4692 		      struct netdev_hw_addr_list *from_list, int addr_len);
4693 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4694 		       struct net_device *dev,
4695 		       int (*sync)(struct net_device *, const unsigned char *),
4696 		       int (*unsync)(struct net_device *,
4697 				     const unsigned char *));
4698 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4699 			   struct net_device *dev,
4700 			   int (*sync)(struct net_device *,
4701 				       const unsigned char *, int),
4702 			   int (*unsync)(struct net_device *,
4703 					 const unsigned char *, int));
4704 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4705 			      struct net_device *dev,
4706 			      int (*unsync)(struct net_device *,
4707 					    const unsigned char *, int));
4708 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4709 			  struct net_device *dev,
4710 			  int (*unsync)(struct net_device *,
4711 					const unsigned char *));
4712 void __hw_addr_init(struct netdev_hw_addr_list *list);
4713 
4714 /* Functions used for device addresses handling */
4715 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4716 		  const void *addr, size_t len);
4717 
4718 static inline void
4719 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4720 {
4721 	dev_addr_mod(dev, 0, addr, len);
4722 }
4723 
4724 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4725 {
4726 	__dev_addr_set(dev, addr, dev->addr_len);
4727 }
4728 
4729 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4730 		 unsigned char addr_type);
4731 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4732 		 unsigned char addr_type);
4733 
4734 /* Functions used for unicast addresses handling */
4735 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4736 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4737 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4738 int dev_uc_sync(struct net_device *to, struct net_device *from);
4739 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4740 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4741 void dev_uc_flush(struct net_device *dev);
4742 void dev_uc_init(struct net_device *dev);
4743 
4744 /**
4745  *  __dev_uc_sync - Synchronize device's unicast list
4746  *  @dev:  device to sync
4747  *  @sync: function to call if address should be added
4748  *  @unsync: function to call if address should be removed
4749  *
4750  *  Add newly added addresses to the interface, and release
4751  *  addresses that have been deleted.
4752  */
4753 static inline int __dev_uc_sync(struct net_device *dev,
4754 				int (*sync)(struct net_device *,
4755 					    const unsigned char *),
4756 				int (*unsync)(struct net_device *,
4757 					      const unsigned char *))
4758 {
4759 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4760 }
4761 
4762 /**
4763  *  __dev_uc_unsync - Remove synchronized addresses from device
4764  *  @dev:  device to sync
4765  *  @unsync: function to call if address should be removed
4766  *
4767  *  Remove all addresses that were added to the device by dev_uc_sync().
4768  */
4769 static inline void __dev_uc_unsync(struct net_device *dev,
4770 				   int (*unsync)(struct net_device *,
4771 						 const unsigned char *))
4772 {
4773 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
4774 }
4775 
4776 /* Functions used for multicast addresses handling */
4777 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4778 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4779 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4780 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4781 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4782 int dev_mc_sync(struct net_device *to, struct net_device *from);
4783 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4784 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4785 void dev_mc_flush(struct net_device *dev);
4786 void dev_mc_init(struct net_device *dev);
4787 
4788 /**
4789  *  __dev_mc_sync - Synchronize device's multicast list
4790  *  @dev:  device to sync
4791  *  @sync: function to call if address should be added
4792  *  @unsync: function to call if address should be removed
4793  *
4794  *  Add newly added addresses to the interface, and release
4795  *  addresses that have been deleted.
4796  */
4797 static inline int __dev_mc_sync(struct net_device *dev,
4798 				int (*sync)(struct net_device *,
4799 					    const unsigned char *),
4800 				int (*unsync)(struct net_device *,
4801 					      const unsigned char *))
4802 {
4803 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4804 }
4805 
4806 /**
4807  *  __dev_mc_unsync - Remove synchronized addresses from device
4808  *  @dev:  device to sync
4809  *  @unsync: function to call if address should be removed
4810  *
4811  *  Remove all addresses that were added to the device by dev_mc_sync().
4812  */
4813 static inline void __dev_mc_unsync(struct net_device *dev,
4814 				   int (*unsync)(struct net_device *,
4815 						 const unsigned char *))
4816 {
4817 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
4818 }
4819 
4820 /* Functions used for secondary unicast and multicast support */
4821 void dev_set_rx_mode(struct net_device *dev);
4822 int dev_set_promiscuity(struct net_device *dev, int inc);
4823 int dev_set_allmulti(struct net_device *dev, int inc);
4824 void netdev_state_change(struct net_device *dev);
4825 void __netdev_notify_peers(struct net_device *dev);
4826 void netdev_notify_peers(struct net_device *dev);
4827 void netdev_features_change(struct net_device *dev);
4828 /* Load a device via the kmod */
4829 void dev_load(struct net *net, const char *name);
4830 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4831 					struct rtnl_link_stats64 *storage);
4832 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4833 			     const struct net_device_stats *netdev_stats);
4834 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4835 			   const struct pcpu_sw_netstats __percpu *netstats);
4836 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4837 
4838 enum {
4839 	NESTED_SYNC_IMM_BIT,
4840 	NESTED_SYNC_TODO_BIT,
4841 };
4842 
4843 #define __NESTED_SYNC_BIT(bit)	((u32)1 << (bit))
4844 #define __NESTED_SYNC(name)	__NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4845 
4846 #define NESTED_SYNC_IMM		__NESTED_SYNC(IMM)
4847 #define NESTED_SYNC_TODO	__NESTED_SYNC(TODO)
4848 
4849 struct netdev_nested_priv {
4850 	unsigned char flags;
4851 	void *data;
4852 };
4853 
4854 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4855 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4856 						     struct list_head **iter);
4857 
4858 /* iterate through upper list, must be called under RCU read lock */
4859 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4860 	for (iter = &(dev)->adj_list.upper, \
4861 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4862 	     updev; \
4863 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4864 
4865 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4866 				  int (*fn)(struct net_device *upper_dev,
4867 					    struct netdev_nested_priv *priv),
4868 				  struct netdev_nested_priv *priv);
4869 
4870 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4871 				  struct net_device *upper_dev);
4872 
4873 bool netdev_has_any_upper_dev(struct net_device *dev);
4874 
4875 void *netdev_lower_get_next_private(struct net_device *dev,
4876 				    struct list_head **iter);
4877 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4878 					struct list_head **iter);
4879 
4880 #define netdev_for_each_lower_private(dev, priv, iter) \
4881 	for (iter = (dev)->adj_list.lower.next, \
4882 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
4883 	     priv; \
4884 	     priv = netdev_lower_get_next_private(dev, &(iter)))
4885 
4886 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4887 	for (iter = &(dev)->adj_list.lower, \
4888 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4889 	     priv; \
4890 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4891 
4892 void *netdev_lower_get_next(struct net_device *dev,
4893 				struct list_head **iter);
4894 
4895 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4896 	for (iter = (dev)->adj_list.lower.next, \
4897 	     ldev = netdev_lower_get_next(dev, &(iter)); \
4898 	     ldev; \
4899 	     ldev = netdev_lower_get_next(dev, &(iter)))
4900 
4901 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4902 					     struct list_head **iter);
4903 int netdev_walk_all_lower_dev(struct net_device *dev,
4904 			      int (*fn)(struct net_device *lower_dev,
4905 					struct netdev_nested_priv *priv),
4906 			      struct netdev_nested_priv *priv);
4907 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4908 				  int (*fn)(struct net_device *lower_dev,
4909 					    struct netdev_nested_priv *priv),
4910 				  struct netdev_nested_priv *priv);
4911 
4912 void *netdev_adjacent_get_private(struct list_head *adj_list);
4913 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4914 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4915 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4916 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4917 			  struct netlink_ext_ack *extack);
4918 int netdev_master_upper_dev_link(struct net_device *dev,
4919 				 struct net_device *upper_dev,
4920 				 void *upper_priv, void *upper_info,
4921 				 struct netlink_ext_ack *extack);
4922 void netdev_upper_dev_unlink(struct net_device *dev,
4923 			     struct net_device *upper_dev);
4924 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4925 				   struct net_device *new_dev,
4926 				   struct net_device *dev,
4927 				   struct netlink_ext_ack *extack);
4928 void netdev_adjacent_change_commit(struct net_device *old_dev,
4929 				   struct net_device *new_dev,
4930 				   struct net_device *dev);
4931 void netdev_adjacent_change_abort(struct net_device *old_dev,
4932 				  struct net_device *new_dev,
4933 				  struct net_device *dev);
4934 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4935 void *netdev_lower_dev_get_private(struct net_device *dev,
4936 				   struct net_device *lower_dev);
4937 void netdev_lower_state_changed(struct net_device *lower_dev,
4938 				void *lower_state_info);
4939 
4940 /* RSS keys are 40 or 52 bytes long */
4941 #define NETDEV_RSS_KEY_LEN 52
4942 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4943 void netdev_rss_key_fill(void *buffer, size_t len);
4944 
4945 int skb_checksum_help(struct sk_buff *skb);
4946 int skb_crc32c_csum_help(struct sk_buff *skb);
4947 int skb_csum_hwoffload_help(struct sk_buff *skb,
4948 			    const netdev_features_t features);
4949 
4950 struct netdev_bonding_info {
4951 	ifslave	slave;
4952 	ifbond	master;
4953 };
4954 
4955 struct netdev_notifier_bonding_info {
4956 	struct netdev_notifier_info info; /* must be first */
4957 	struct netdev_bonding_info  bonding_info;
4958 };
4959 
4960 void netdev_bonding_info_change(struct net_device *dev,
4961 				struct netdev_bonding_info *bonding_info);
4962 
4963 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4964 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4965 #else
4966 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4967 				  const void *data)
4968 {
4969 }
4970 #endif
4971 
4972 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4973 
4974 static inline bool can_checksum_protocol(netdev_features_t features,
4975 					 __be16 protocol)
4976 {
4977 	if (protocol == htons(ETH_P_FCOE))
4978 		return !!(features & NETIF_F_FCOE_CRC);
4979 
4980 	/* Assume this is an IP checksum (not SCTP CRC) */
4981 
4982 	if (features & NETIF_F_HW_CSUM) {
4983 		/* Can checksum everything */
4984 		return true;
4985 	}
4986 
4987 	switch (protocol) {
4988 	case htons(ETH_P_IP):
4989 		return !!(features & NETIF_F_IP_CSUM);
4990 	case htons(ETH_P_IPV6):
4991 		return !!(features & NETIF_F_IPV6_CSUM);
4992 	default:
4993 		return false;
4994 	}
4995 }
4996 
4997 #ifdef CONFIG_BUG
4998 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4999 #else
5000 static inline void netdev_rx_csum_fault(struct net_device *dev,
5001 					struct sk_buff *skb)
5002 {
5003 }
5004 #endif
5005 /* rx skb timestamps */
5006 void net_enable_timestamp(void);
5007 void net_disable_timestamp(void);
5008 
5009 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
5010 					const struct skb_shared_hwtstamps *hwtstamps,
5011 					bool cycles)
5012 {
5013 	const struct net_device_ops *ops = dev->netdev_ops;
5014 
5015 	if (ops->ndo_get_tstamp)
5016 		return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
5017 
5018 	return hwtstamps->hwtstamp;
5019 }
5020 
5021 #ifndef CONFIG_PREEMPT_RT
5022 static inline void netdev_xmit_set_more(bool more)
5023 {
5024 	__this_cpu_write(softnet_data.xmit.more, more);
5025 }
5026 
5027 static inline bool netdev_xmit_more(void)
5028 {
5029 	return __this_cpu_read(softnet_data.xmit.more);
5030 }
5031 #else
5032 static inline void netdev_xmit_set_more(bool more)
5033 {
5034 	current->net_xmit.more = more;
5035 }
5036 
5037 static inline bool netdev_xmit_more(void)
5038 {
5039 	return current->net_xmit.more;
5040 }
5041 #endif
5042 
5043 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
5044 					      struct sk_buff *skb, struct net_device *dev,
5045 					      bool more)
5046 {
5047 	netdev_xmit_set_more(more);
5048 	return ops->ndo_start_xmit(skb, dev);
5049 }
5050 
5051 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
5052 					    struct netdev_queue *txq, bool more)
5053 {
5054 	const struct net_device_ops *ops = dev->netdev_ops;
5055 	netdev_tx_t rc;
5056 
5057 	rc = __netdev_start_xmit(ops, skb, dev, more);
5058 	if (rc == NETDEV_TX_OK)
5059 		txq_trans_update(txq);
5060 
5061 	return rc;
5062 }
5063 
5064 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
5065 				const void *ns);
5066 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
5067 				 const void *ns);
5068 
5069 extern const struct kobj_ns_type_operations net_ns_type_operations;
5070 
5071 const char *netdev_drivername(const struct net_device *dev);
5072 
5073 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
5074 							  netdev_features_t f2)
5075 {
5076 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
5077 		if (f1 & NETIF_F_HW_CSUM)
5078 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5079 		else
5080 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5081 	}
5082 
5083 	return f1 & f2;
5084 }
5085 
5086 static inline netdev_features_t netdev_get_wanted_features(
5087 	struct net_device *dev)
5088 {
5089 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
5090 }
5091 netdev_features_t netdev_increment_features(netdev_features_t all,
5092 	netdev_features_t one, netdev_features_t mask);
5093 
5094 /* Allow TSO being used on stacked device :
5095  * Performing the GSO segmentation before last device
5096  * is a performance improvement.
5097  */
5098 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5099 							netdev_features_t mask)
5100 {
5101 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5102 }
5103 
5104 int __netdev_update_features(struct net_device *dev);
5105 void netdev_update_features(struct net_device *dev);
5106 void netdev_change_features(struct net_device *dev);
5107 
5108 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5109 					struct net_device *dev);
5110 
5111 netdev_features_t passthru_features_check(struct sk_buff *skb,
5112 					  struct net_device *dev,
5113 					  netdev_features_t features);
5114 netdev_features_t netif_skb_features(struct sk_buff *skb);
5115 void skb_warn_bad_offload(const struct sk_buff *skb);
5116 
5117 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5118 {
5119 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5120 
5121 	/* check flags correspondence */
5122 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5123 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5124 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5125 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5126 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5127 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5128 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5129 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5130 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5131 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5132 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5133 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5134 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5135 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5136 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5137 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5138 	BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5139 	BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5140 	BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5141 
5142 	return (features & feature) == feature;
5143 }
5144 
5145 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5146 {
5147 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5148 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5149 }
5150 
5151 static inline bool netif_needs_gso(struct sk_buff *skb,
5152 				   netdev_features_t features)
5153 {
5154 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5155 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5156 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5157 }
5158 
5159 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5160 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5161 void netif_inherit_tso_max(struct net_device *to,
5162 			   const struct net_device *from);
5163 
5164 static inline unsigned int
5165 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb)
5166 {
5167 	/* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */
5168 	return skb->protocol == htons(ETH_P_IPV6) ?
5169 	       READ_ONCE(dev->gro_max_size) :
5170 	       READ_ONCE(dev->gro_ipv4_max_size);
5171 }
5172 
5173 static inline unsigned int
5174 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb)
5175 {
5176 	/* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */
5177 	return skb->protocol == htons(ETH_P_IPV6) ?
5178 	       READ_ONCE(dev->gso_max_size) :
5179 	       READ_ONCE(dev->gso_ipv4_max_size);
5180 }
5181 
5182 static inline bool netif_is_macsec(const struct net_device *dev)
5183 {
5184 	return dev->priv_flags & IFF_MACSEC;
5185 }
5186 
5187 static inline bool netif_is_macvlan(const struct net_device *dev)
5188 {
5189 	return dev->priv_flags & IFF_MACVLAN;
5190 }
5191 
5192 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5193 {
5194 	return dev->priv_flags & IFF_MACVLAN_PORT;
5195 }
5196 
5197 static inline bool netif_is_bond_master(const struct net_device *dev)
5198 {
5199 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5200 }
5201 
5202 static inline bool netif_is_bond_slave(const struct net_device *dev)
5203 {
5204 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5205 }
5206 
5207 static inline bool netif_supports_nofcs(struct net_device *dev)
5208 {
5209 	return dev->priv_flags & IFF_SUPP_NOFCS;
5210 }
5211 
5212 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5213 {
5214 	return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5215 }
5216 
5217 static inline bool netif_is_l3_master(const struct net_device *dev)
5218 {
5219 	return dev->priv_flags & IFF_L3MDEV_MASTER;
5220 }
5221 
5222 static inline bool netif_is_l3_slave(const struct net_device *dev)
5223 {
5224 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
5225 }
5226 
5227 static inline int dev_sdif(const struct net_device *dev)
5228 {
5229 #ifdef CONFIG_NET_L3_MASTER_DEV
5230 	if (netif_is_l3_slave(dev))
5231 		return dev->ifindex;
5232 #endif
5233 	return 0;
5234 }
5235 
5236 static inline bool netif_is_bridge_master(const struct net_device *dev)
5237 {
5238 	return dev->priv_flags & IFF_EBRIDGE;
5239 }
5240 
5241 static inline bool netif_is_bridge_port(const struct net_device *dev)
5242 {
5243 	return dev->priv_flags & IFF_BRIDGE_PORT;
5244 }
5245 
5246 static inline bool netif_is_ovs_master(const struct net_device *dev)
5247 {
5248 	return dev->priv_flags & IFF_OPENVSWITCH;
5249 }
5250 
5251 static inline bool netif_is_ovs_port(const struct net_device *dev)
5252 {
5253 	return dev->priv_flags & IFF_OVS_DATAPATH;
5254 }
5255 
5256 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5257 {
5258 	return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5259 }
5260 
5261 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5262 {
5263 	return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5264 }
5265 
5266 static inline bool netif_is_team_master(const struct net_device *dev)
5267 {
5268 	return dev->priv_flags & IFF_TEAM;
5269 }
5270 
5271 static inline bool netif_is_team_port(const struct net_device *dev)
5272 {
5273 	return dev->priv_flags & IFF_TEAM_PORT;
5274 }
5275 
5276 static inline bool netif_is_lag_master(const struct net_device *dev)
5277 {
5278 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
5279 }
5280 
5281 static inline bool netif_is_lag_port(const struct net_device *dev)
5282 {
5283 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5284 }
5285 
5286 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5287 {
5288 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
5289 }
5290 
5291 static inline bool netif_is_failover(const struct net_device *dev)
5292 {
5293 	return dev->priv_flags & IFF_FAILOVER;
5294 }
5295 
5296 static inline bool netif_is_failover_slave(const struct net_device *dev)
5297 {
5298 	return dev->priv_flags & IFF_FAILOVER_SLAVE;
5299 }
5300 
5301 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5302 static inline void netif_keep_dst(struct net_device *dev)
5303 {
5304 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5305 }
5306 
5307 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5308 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5309 {
5310 	/* TODO: reserve and use an additional IFF bit, if we get more users */
5311 	return netif_is_macsec(dev);
5312 }
5313 
5314 extern struct pernet_operations __net_initdata loopback_net_ops;
5315 
5316 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5317 
5318 /* netdev_printk helpers, similar to dev_printk */
5319 
5320 static inline const char *netdev_name(const struct net_device *dev)
5321 {
5322 	if (!dev->name[0] || strchr(dev->name, '%'))
5323 		return "(unnamed net_device)";
5324 	return dev->name;
5325 }
5326 
5327 static inline const char *netdev_reg_state(const struct net_device *dev)
5328 {
5329 	u8 reg_state = READ_ONCE(dev->reg_state);
5330 
5331 	switch (reg_state) {
5332 	case NETREG_UNINITIALIZED: return " (uninitialized)";
5333 	case NETREG_REGISTERED: return "";
5334 	case NETREG_UNREGISTERING: return " (unregistering)";
5335 	case NETREG_UNREGISTERED: return " (unregistered)";
5336 	case NETREG_RELEASED: return " (released)";
5337 	case NETREG_DUMMY: return " (dummy)";
5338 	}
5339 
5340 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state);
5341 	return " (unknown)";
5342 }
5343 
5344 #define MODULE_ALIAS_NETDEV(device) \
5345 	MODULE_ALIAS("netdev-" device)
5346 
5347 /*
5348  * netdev_WARN() acts like dev_printk(), but with the key difference
5349  * of using a WARN/WARN_ON to get the message out, including the
5350  * file/line information and a backtrace.
5351  */
5352 #define netdev_WARN(dev, format, args...)			\
5353 	WARN(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5354 	     netdev_reg_state(dev), ##args)
5355 
5356 #define netdev_WARN_ONCE(dev, format, args...)				\
5357 	WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5358 		  netdev_reg_state(dev), ##args)
5359 
5360 /*
5361  *	The list of packet types we will receive (as opposed to discard)
5362  *	and the routines to invoke.
5363  *
5364  *	Why 16. Because with 16 the only overlap we get on a hash of the
5365  *	low nibble of the protocol value is RARP/SNAP/X.25.
5366  *
5367  *		0800	IP
5368  *		0001	802.3
5369  *		0002	AX.25
5370  *		0004	802.2
5371  *		8035	RARP
5372  *		0005	SNAP
5373  *		0805	X.25
5374  *		0806	ARP
5375  *		8137	IPX
5376  *		0009	Localtalk
5377  *		86DD	IPv6
5378  */
5379 #define PTYPE_HASH_SIZE	(16)
5380 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
5381 
5382 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5383 
5384 extern struct net_device *blackhole_netdev;
5385 
5386 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5387 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5388 #define DEV_STATS_ADD(DEV, FIELD, VAL) 	\
5389 		atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5390 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5391 
5392 #endif	/* _LINUX_NETDEVICE_H */
5393