xref: /linux-6.15/include/linux/netdevice.h (revision 98cfbe42)
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		Definitions for the Interfaces handler.
8  *
9  * Version:	@(#)dev.h	1.0.10	08/12/93
10  *
11  * Authors:	Ross Biro
12  *		Fred N. van Kempen, <[email protected]>
13  *		Corey Minyard <[email protected]>
14  *		Donald J. Becker, <[email protected]>
15  *		Alan Cox, <[email protected]>
16  *		Bjorn Ekwall. <[email protected]>
17  *              Pekka Riikonen <[email protected]>
18  *
19  *		Moved to /usr/include/linux for NET3
20  */
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
23 
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
31 #include <asm/local.h>
32 
33 #include <linux/percpu.h>
34 #include <linux/rculist.h>
35 #include <linux/workqueue.h>
36 #include <linux/dynamic_queue_limits.h>
37 
38 #include <net/net_namespace.h>
39 #ifdef CONFIG_DCB
40 #include <net/dcbnl.h>
41 #endif
42 #include <net/netprio_cgroup.h>
43 
44 #include <linux/netdev_features.h>
45 #include <linux/neighbour.h>
46 #include <uapi/linux/netdevice.h>
47 #include <uapi/linux/if_bonding.h>
48 #include <uapi/linux/pkt_cls.h>
49 #include <uapi/linux/netdev.h>
50 #include <linux/hashtable.h>
51 #include <linux/rbtree.h>
52 #include <net/net_trackers.h>
53 #include <net/net_debug.h>
54 #include <net/dropreason-core.h>
55 
56 struct netpoll_info;
57 struct device;
58 struct ethtool_ops;
59 struct kernel_hwtstamp_config;
60 struct phy_device;
61 struct dsa_port;
62 struct ip_tunnel_parm;
63 struct macsec_context;
64 struct macsec_ops;
65 struct netdev_name_node;
66 struct sd_flow_limit;
67 struct sfp_bus;
68 /* 802.11 specific */
69 struct wireless_dev;
70 /* 802.15.4 specific */
71 struct wpan_dev;
72 struct mpls_dev;
73 /* UDP Tunnel offloads */
74 struct udp_tunnel_info;
75 struct udp_tunnel_nic_info;
76 struct udp_tunnel_nic;
77 struct bpf_prog;
78 struct xdp_buff;
79 struct xdp_frame;
80 struct xdp_metadata_ops;
81 struct xdp_md;
82 /* DPLL specific */
83 struct dpll_pin;
84 
85 typedef u32 xdp_features_t;
86 
87 void synchronize_net(void);
88 void netdev_set_default_ethtool_ops(struct net_device *dev,
89 				    const struct ethtool_ops *ops);
90 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
91 
92 /* Backlog congestion levels */
93 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
94 #define NET_RX_DROP		1	/* packet dropped */
95 
96 #define MAX_NEST_DEV 8
97 
98 /*
99  * Transmit return codes: transmit return codes originate from three different
100  * namespaces:
101  *
102  * - qdisc return codes
103  * - driver transmit return codes
104  * - errno values
105  *
106  * Drivers are allowed to return any one of those in their hard_start_xmit()
107  * function. Real network devices commonly used with qdiscs should only return
108  * the driver transmit return codes though - when qdiscs are used, the actual
109  * transmission happens asynchronously, so the value is not propagated to
110  * higher layers. Virtual network devices transmit synchronously; in this case
111  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
112  * others are propagated to higher layers.
113  */
114 
115 /* qdisc ->enqueue() return codes. */
116 #define NET_XMIT_SUCCESS	0x00
117 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
118 #define NET_XMIT_CN		0x02	/* congestion notification	*/
119 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
120 
121 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
122  * indicates that the device will soon be dropping packets, or already drops
123  * some packets of the same priority; prompting us to send less aggressively. */
124 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
125 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
126 
127 /* Driver transmit return codes */
128 #define NETDEV_TX_MASK		0xf0
129 
130 enum netdev_tx {
131 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
132 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
133 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
134 };
135 typedef enum netdev_tx netdev_tx_t;
136 
137 /*
138  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
139  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
140  */
141 static inline bool dev_xmit_complete(int rc)
142 {
143 	/*
144 	 * Positive cases with an skb consumed by a driver:
145 	 * - successful transmission (rc == NETDEV_TX_OK)
146 	 * - error while transmitting (rc < 0)
147 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
148 	 */
149 	if (likely(rc < NET_XMIT_MASK))
150 		return true;
151 
152 	return false;
153 }
154 
155 /*
156  *	Compute the worst-case header length according to the protocols
157  *	used.
158  */
159 
160 #if defined(CONFIG_HYPERV_NET)
161 # define LL_MAX_HEADER 128
162 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
163 # if defined(CONFIG_MAC80211_MESH)
164 #  define LL_MAX_HEADER 128
165 # else
166 #  define LL_MAX_HEADER 96
167 # endif
168 #else
169 # define LL_MAX_HEADER 32
170 #endif
171 
172 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
173     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
174 #define MAX_HEADER LL_MAX_HEADER
175 #else
176 #define MAX_HEADER (LL_MAX_HEADER + 48)
177 #endif
178 
179 /*
180  *	Old network device statistics. Fields are native words
181  *	(unsigned long) so they can be read and written atomically.
182  */
183 
184 #define NET_DEV_STAT(FIELD)			\
185 	union {					\
186 		unsigned long FIELD;		\
187 		atomic_long_t __##FIELD;	\
188 	}
189 
190 struct net_device_stats {
191 	NET_DEV_STAT(rx_packets);
192 	NET_DEV_STAT(tx_packets);
193 	NET_DEV_STAT(rx_bytes);
194 	NET_DEV_STAT(tx_bytes);
195 	NET_DEV_STAT(rx_errors);
196 	NET_DEV_STAT(tx_errors);
197 	NET_DEV_STAT(rx_dropped);
198 	NET_DEV_STAT(tx_dropped);
199 	NET_DEV_STAT(multicast);
200 	NET_DEV_STAT(collisions);
201 	NET_DEV_STAT(rx_length_errors);
202 	NET_DEV_STAT(rx_over_errors);
203 	NET_DEV_STAT(rx_crc_errors);
204 	NET_DEV_STAT(rx_frame_errors);
205 	NET_DEV_STAT(rx_fifo_errors);
206 	NET_DEV_STAT(rx_missed_errors);
207 	NET_DEV_STAT(tx_aborted_errors);
208 	NET_DEV_STAT(tx_carrier_errors);
209 	NET_DEV_STAT(tx_fifo_errors);
210 	NET_DEV_STAT(tx_heartbeat_errors);
211 	NET_DEV_STAT(tx_window_errors);
212 	NET_DEV_STAT(rx_compressed);
213 	NET_DEV_STAT(tx_compressed);
214 };
215 #undef NET_DEV_STAT
216 
217 /* per-cpu stats, allocated on demand.
218  * Try to fit them in a single cache line, for dev_get_stats() sake.
219  */
220 struct net_device_core_stats {
221 	unsigned long	rx_dropped;
222 	unsigned long	tx_dropped;
223 	unsigned long	rx_nohandler;
224 	unsigned long	rx_otherhost_dropped;
225 } __aligned(4 * sizeof(unsigned long));
226 
227 #include <linux/cache.h>
228 #include <linux/skbuff.h>
229 
230 #ifdef CONFIG_RPS
231 #include <linux/static_key.h>
232 extern struct static_key_false rps_needed;
233 extern struct static_key_false rfs_needed;
234 #endif
235 
236 struct neighbour;
237 struct neigh_parms;
238 struct sk_buff;
239 
240 struct netdev_hw_addr {
241 	struct list_head	list;
242 	struct rb_node		node;
243 	unsigned char		addr[MAX_ADDR_LEN];
244 	unsigned char		type;
245 #define NETDEV_HW_ADDR_T_LAN		1
246 #define NETDEV_HW_ADDR_T_SAN		2
247 #define NETDEV_HW_ADDR_T_UNICAST	3
248 #define NETDEV_HW_ADDR_T_MULTICAST	4
249 	bool			global_use;
250 	int			sync_cnt;
251 	int			refcount;
252 	int			synced;
253 	struct rcu_head		rcu_head;
254 };
255 
256 struct netdev_hw_addr_list {
257 	struct list_head	list;
258 	int			count;
259 
260 	/* Auxiliary tree for faster lookup on addition and deletion */
261 	struct rb_root		tree;
262 };
263 
264 #define netdev_hw_addr_list_count(l) ((l)->count)
265 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
266 #define netdev_hw_addr_list_for_each(ha, l) \
267 	list_for_each_entry(ha, &(l)->list, list)
268 
269 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
270 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
271 #define netdev_for_each_uc_addr(ha, dev) \
272 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
273 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
274 	netdev_for_each_uc_addr((_ha), (_dev)) \
275 		if ((_ha)->sync_cnt)
276 
277 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
278 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
279 #define netdev_for_each_mc_addr(ha, dev) \
280 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
281 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
282 	netdev_for_each_mc_addr((_ha), (_dev)) \
283 		if ((_ha)->sync_cnt)
284 
285 struct hh_cache {
286 	unsigned int	hh_len;
287 	seqlock_t	hh_lock;
288 
289 	/* cached hardware header; allow for machine alignment needs.        */
290 #define HH_DATA_MOD	16
291 #define HH_DATA_OFF(__len) \
292 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
293 #define HH_DATA_ALIGN(__len) \
294 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
295 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
296 };
297 
298 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
299  * Alternative is:
300  *   dev->hard_header_len ? (dev->hard_header_len +
301  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
302  *
303  * We could use other alignment values, but we must maintain the
304  * relationship HH alignment <= LL alignment.
305  */
306 #define LL_RESERVED_SPACE(dev) \
307 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
308 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
309 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
310 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
311 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
312 
313 struct header_ops {
314 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
315 			   unsigned short type, const void *daddr,
316 			   const void *saddr, unsigned int len);
317 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
318 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
319 	void	(*cache_update)(struct hh_cache *hh,
320 				const struct net_device *dev,
321 				const unsigned char *haddr);
322 	bool	(*validate)(const char *ll_header, unsigned int len);
323 	__be16	(*parse_protocol)(const struct sk_buff *skb);
324 };
325 
326 /* These flag bits are private to the generic network queueing
327  * layer; they may not be explicitly referenced by any other
328  * code.
329  */
330 
331 enum netdev_state_t {
332 	__LINK_STATE_START,
333 	__LINK_STATE_PRESENT,
334 	__LINK_STATE_NOCARRIER,
335 	__LINK_STATE_LINKWATCH_PENDING,
336 	__LINK_STATE_DORMANT,
337 	__LINK_STATE_TESTING,
338 };
339 
340 struct gro_list {
341 	struct list_head	list;
342 	int			count;
343 };
344 
345 /*
346  * size of gro hash buckets, must less than bit number of
347  * napi_struct::gro_bitmask
348  */
349 #define GRO_HASH_BUCKETS	8
350 
351 /*
352  * Structure for NAPI scheduling similar to tasklet but with weighting
353  */
354 struct napi_struct {
355 	/* The poll_list must only be managed by the entity which
356 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
357 	 * whoever atomically sets that bit can add this napi_struct
358 	 * to the per-CPU poll_list, and whoever clears that bit
359 	 * can remove from the list right before clearing the bit.
360 	 */
361 	struct list_head	poll_list;
362 
363 	unsigned long		state;
364 	int			weight;
365 	int			defer_hard_irqs_count;
366 	unsigned long		gro_bitmask;
367 	int			(*poll)(struct napi_struct *, int);
368 #ifdef CONFIG_NETPOLL
369 	/* CPU actively polling if netpoll is configured */
370 	int			poll_owner;
371 #endif
372 	/* CPU on which NAPI has been scheduled for processing */
373 	int			list_owner;
374 	struct net_device	*dev;
375 	struct gro_list		gro_hash[GRO_HASH_BUCKETS];
376 	struct sk_buff		*skb;
377 	struct list_head	rx_list; /* Pending GRO_NORMAL skbs */
378 	int			rx_count; /* length of rx_list */
379 	unsigned int		napi_id;
380 	struct hrtimer		timer;
381 	struct task_struct	*thread;
382 	/* control-path-only fields follow */
383 	struct list_head	dev_list;
384 	struct hlist_node	napi_hash_node;
385 };
386 
387 enum {
388 	NAPI_STATE_SCHED,		/* Poll is scheduled */
389 	NAPI_STATE_MISSED,		/* reschedule a napi */
390 	NAPI_STATE_DISABLE,		/* Disable pending */
391 	NAPI_STATE_NPSVC,		/* Netpoll - don't dequeue from poll_list */
392 	NAPI_STATE_LISTED,		/* NAPI added to system lists */
393 	NAPI_STATE_NO_BUSY_POLL,	/* Do not add in napi_hash, no busy polling */
394 	NAPI_STATE_IN_BUSY_POLL,	/* sk_busy_loop() owns this NAPI */
395 	NAPI_STATE_PREFER_BUSY_POLL,	/* prefer busy-polling over softirq processing*/
396 	NAPI_STATE_THREADED,		/* The poll is performed inside its own thread*/
397 	NAPI_STATE_SCHED_THREADED,	/* Napi is currently scheduled in threaded mode */
398 };
399 
400 enum {
401 	NAPIF_STATE_SCHED		= BIT(NAPI_STATE_SCHED),
402 	NAPIF_STATE_MISSED		= BIT(NAPI_STATE_MISSED),
403 	NAPIF_STATE_DISABLE		= BIT(NAPI_STATE_DISABLE),
404 	NAPIF_STATE_NPSVC		= BIT(NAPI_STATE_NPSVC),
405 	NAPIF_STATE_LISTED		= BIT(NAPI_STATE_LISTED),
406 	NAPIF_STATE_NO_BUSY_POLL	= BIT(NAPI_STATE_NO_BUSY_POLL),
407 	NAPIF_STATE_IN_BUSY_POLL	= BIT(NAPI_STATE_IN_BUSY_POLL),
408 	NAPIF_STATE_PREFER_BUSY_POLL	= BIT(NAPI_STATE_PREFER_BUSY_POLL),
409 	NAPIF_STATE_THREADED		= BIT(NAPI_STATE_THREADED),
410 	NAPIF_STATE_SCHED_THREADED	= BIT(NAPI_STATE_SCHED_THREADED),
411 };
412 
413 enum gro_result {
414 	GRO_MERGED,
415 	GRO_MERGED_FREE,
416 	GRO_HELD,
417 	GRO_NORMAL,
418 	GRO_CONSUMED,
419 };
420 typedef enum gro_result gro_result_t;
421 
422 /*
423  * enum rx_handler_result - Possible return values for rx_handlers.
424  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
425  * further.
426  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
427  * case skb->dev was changed by rx_handler.
428  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
429  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
430  *
431  * rx_handlers are functions called from inside __netif_receive_skb(), to do
432  * special processing of the skb, prior to delivery to protocol handlers.
433  *
434  * Currently, a net_device can only have a single rx_handler registered. Trying
435  * to register a second rx_handler will return -EBUSY.
436  *
437  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
438  * To unregister a rx_handler on a net_device, use
439  * netdev_rx_handler_unregister().
440  *
441  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
442  * do with the skb.
443  *
444  * If the rx_handler consumed the skb in some way, it should return
445  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
446  * the skb to be delivered in some other way.
447  *
448  * If the rx_handler changed skb->dev, to divert the skb to another
449  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
450  * new device will be called if it exists.
451  *
452  * If the rx_handler decides the skb should be ignored, it should return
453  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
454  * are registered on exact device (ptype->dev == skb->dev).
455  *
456  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
457  * delivered, it should return RX_HANDLER_PASS.
458  *
459  * A device without a registered rx_handler will behave as if rx_handler
460  * returned RX_HANDLER_PASS.
461  */
462 
463 enum rx_handler_result {
464 	RX_HANDLER_CONSUMED,
465 	RX_HANDLER_ANOTHER,
466 	RX_HANDLER_EXACT,
467 	RX_HANDLER_PASS,
468 };
469 typedef enum rx_handler_result rx_handler_result_t;
470 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
471 
472 void __napi_schedule(struct napi_struct *n);
473 void __napi_schedule_irqoff(struct napi_struct *n);
474 
475 static inline bool napi_disable_pending(struct napi_struct *n)
476 {
477 	return test_bit(NAPI_STATE_DISABLE, &n->state);
478 }
479 
480 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
481 {
482 	return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
483 }
484 
485 bool napi_schedule_prep(struct napi_struct *n);
486 
487 /**
488  *	napi_schedule - schedule NAPI poll
489  *	@n: NAPI context
490  *
491  * Schedule NAPI poll routine to be called if it is not already
492  * running.
493  */
494 static inline void napi_schedule(struct napi_struct *n)
495 {
496 	if (napi_schedule_prep(n))
497 		__napi_schedule(n);
498 }
499 
500 /**
501  *	napi_schedule_irqoff - schedule NAPI poll
502  *	@n: NAPI context
503  *
504  * Variant of napi_schedule(), assuming hard irqs are masked.
505  */
506 static inline void napi_schedule_irqoff(struct napi_struct *n)
507 {
508 	if (napi_schedule_prep(n))
509 		__napi_schedule_irqoff(n);
510 }
511 
512 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
513 static inline bool napi_reschedule(struct napi_struct *napi)
514 {
515 	if (napi_schedule_prep(napi)) {
516 		__napi_schedule(napi);
517 		return true;
518 	}
519 	return false;
520 }
521 
522 /**
523  * napi_complete_done - NAPI processing complete
524  * @n: NAPI context
525  * @work_done: number of packets processed
526  *
527  * Mark NAPI processing as complete. Should only be called if poll budget
528  * has not been completely consumed.
529  * Prefer over napi_complete().
530  * Return false if device should avoid rearming interrupts.
531  */
532 bool napi_complete_done(struct napi_struct *n, int work_done);
533 
534 static inline bool napi_complete(struct napi_struct *n)
535 {
536 	return napi_complete_done(n, 0);
537 }
538 
539 int dev_set_threaded(struct net_device *dev, bool threaded);
540 
541 /**
542  *	napi_disable - prevent NAPI from scheduling
543  *	@n: NAPI context
544  *
545  * Stop NAPI from being scheduled on this context.
546  * Waits till any outstanding processing completes.
547  */
548 void napi_disable(struct napi_struct *n);
549 
550 void napi_enable(struct napi_struct *n);
551 
552 /**
553  *	napi_synchronize - wait until NAPI is not running
554  *	@n: NAPI context
555  *
556  * Wait until NAPI is done being scheduled on this context.
557  * Waits till any outstanding processing completes but
558  * does not disable future activations.
559  */
560 static inline void napi_synchronize(const struct napi_struct *n)
561 {
562 	if (IS_ENABLED(CONFIG_SMP))
563 		while (test_bit(NAPI_STATE_SCHED, &n->state))
564 			msleep(1);
565 	else
566 		barrier();
567 }
568 
569 /**
570  *	napi_if_scheduled_mark_missed - if napi is running, set the
571  *	NAPIF_STATE_MISSED
572  *	@n: NAPI context
573  *
574  * If napi is running, set the NAPIF_STATE_MISSED, and return true if
575  * NAPI is scheduled.
576  **/
577 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
578 {
579 	unsigned long val, new;
580 
581 	val = READ_ONCE(n->state);
582 	do {
583 		if (val & NAPIF_STATE_DISABLE)
584 			return true;
585 
586 		if (!(val & NAPIF_STATE_SCHED))
587 			return false;
588 
589 		new = val | NAPIF_STATE_MISSED;
590 	} while (!try_cmpxchg(&n->state, &val, new));
591 
592 	return true;
593 }
594 
595 enum netdev_queue_state_t {
596 	__QUEUE_STATE_DRV_XOFF,
597 	__QUEUE_STATE_STACK_XOFF,
598 	__QUEUE_STATE_FROZEN,
599 };
600 
601 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
602 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
603 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
604 
605 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
606 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
607 					QUEUE_STATE_FROZEN)
608 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
609 					QUEUE_STATE_FROZEN)
610 
611 /*
612  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
613  * netif_tx_* functions below are used to manipulate this flag.  The
614  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
615  * queue independently.  The netif_xmit_*stopped functions below are called
616  * to check if the queue has been stopped by the driver or stack (either
617  * of the XOFF bits are set in the state).  Drivers should not need to call
618  * netif_xmit*stopped functions, they should only be using netif_tx_*.
619  */
620 
621 struct netdev_queue {
622 /*
623  * read-mostly part
624  */
625 	struct net_device	*dev;
626 	netdevice_tracker	dev_tracker;
627 
628 	struct Qdisc __rcu	*qdisc;
629 	struct Qdisc __rcu	*qdisc_sleeping;
630 #ifdef CONFIG_SYSFS
631 	struct kobject		kobj;
632 #endif
633 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
634 	int			numa_node;
635 #endif
636 	unsigned long		tx_maxrate;
637 	/*
638 	 * Number of TX timeouts for this queue
639 	 * (/sys/class/net/DEV/Q/trans_timeout)
640 	 */
641 	atomic_long_t		trans_timeout;
642 
643 	/* Subordinate device that the queue has been assigned to */
644 	struct net_device	*sb_dev;
645 #ifdef CONFIG_XDP_SOCKETS
646 	struct xsk_buff_pool    *pool;
647 #endif
648 /*
649  * write-mostly part
650  */
651 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
652 	int			xmit_lock_owner;
653 	/*
654 	 * Time (in jiffies) of last Tx
655 	 */
656 	unsigned long		trans_start;
657 
658 	unsigned long		state;
659 
660 #ifdef CONFIG_BQL
661 	struct dql		dql;
662 #endif
663 } ____cacheline_aligned_in_smp;
664 
665 extern int sysctl_fb_tunnels_only_for_init_net;
666 extern int sysctl_devconf_inherit_init_net;
667 
668 /*
669  * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
670  *                                     == 1 : For initns only
671  *                                     == 2 : For none.
672  */
673 static inline bool net_has_fallback_tunnels(const struct net *net)
674 {
675 #if IS_ENABLED(CONFIG_SYSCTL)
676 	int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
677 
678 	return !fb_tunnels_only_for_init_net ||
679 		(net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
680 #else
681 	return true;
682 #endif
683 }
684 
685 static inline int net_inherit_devconf(void)
686 {
687 #if IS_ENABLED(CONFIG_SYSCTL)
688 	return READ_ONCE(sysctl_devconf_inherit_init_net);
689 #else
690 	return 0;
691 #endif
692 }
693 
694 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
695 {
696 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
697 	return q->numa_node;
698 #else
699 	return NUMA_NO_NODE;
700 #endif
701 }
702 
703 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
704 {
705 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
706 	q->numa_node = node;
707 #endif
708 }
709 
710 #ifdef CONFIG_RPS
711 /*
712  * This structure holds an RPS map which can be of variable length.  The
713  * map is an array of CPUs.
714  */
715 struct rps_map {
716 	unsigned int len;
717 	struct rcu_head rcu;
718 	u16 cpus[];
719 };
720 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
721 
722 /*
723  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
724  * tail pointer for that CPU's input queue at the time of last enqueue, and
725  * a hardware filter index.
726  */
727 struct rps_dev_flow {
728 	u16 cpu;
729 	u16 filter;
730 	unsigned int last_qtail;
731 };
732 #define RPS_NO_FILTER 0xffff
733 
734 /*
735  * The rps_dev_flow_table structure contains a table of flow mappings.
736  */
737 struct rps_dev_flow_table {
738 	unsigned int mask;
739 	struct rcu_head rcu;
740 	struct rps_dev_flow flows[];
741 };
742 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
743     ((_num) * sizeof(struct rps_dev_flow)))
744 
745 /*
746  * The rps_sock_flow_table contains mappings of flows to the last CPU
747  * on which they were processed by the application (set in recvmsg).
748  * Each entry is a 32bit value. Upper part is the high-order bits
749  * of flow hash, lower part is CPU number.
750  * rps_cpu_mask is used to partition the space, depending on number of
751  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
752  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
753  * meaning we use 32-6=26 bits for the hash.
754  */
755 struct rps_sock_flow_table {
756 	u32	mask;
757 
758 	u32	ents[] ____cacheline_aligned_in_smp;
759 };
760 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
761 
762 #define RPS_NO_CPU 0xffff
763 
764 extern u32 rps_cpu_mask;
765 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
766 
767 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
768 					u32 hash)
769 {
770 	if (table && hash) {
771 		unsigned int index = hash & table->mask;
772 		u32 val = hash & ~rps_cpu_mask;
773 
774 		/* We only give a hint, preemption can change CPU under us */
775 		val |= raw_smp_processor_id();
776 
777 		/* The following WRITE_ONCE() is paired with the READ_ONCE()
778 		 * here, and another one in get_rps_cpu().
779 		 */
780 		if (READ_ONCE(table->ents[index]) != val)
781 			WRITE_ONCE(table->ents[index], val);
782 	}
783 }
784 
785 #ifdef CONFIG_RFS_ACCEL
786 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
787 			 u16 filter_id);
788 #endif
789 #endif /* CONFIG_RPS */
790 
791 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
792 enum xps_map_type {
793 	XPS_CPUS = 0,
794 	XPS_RXQS,
795 	XPS_MAPS_MAX,
796 };
797 
798 #ifdef CONFIG_XPS
799 /*
800  * This structure holds an XPS map which can be of variable length.  The
801  * map is an array of queues.
802  */
803 struct xps_map {
804 	unsigned int len;
805 	unsigned int alloc_len;
806 	struct rcu_head rcu;
807 	u16 queues[];
808 };
809 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
810 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
811        - sizeof(struct xps_map)) / sizeof(u16))
812 
813 /*
814  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
815  *
816  * We keep track of the number of cpus/rxqs used when the struct is allocated,
817  * in nr_ids. This will help not accessing out-of-bound memory.
818  *
819  * We keep track of the number of traffic classes used when the struct is
820  * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
821  * not crossing its upper bound, as the original dev->num_tc can be updated in
822  * the meantime.
823  */
824 struct xps_dev_maps {
825 	struct rcu_head rcu;
826 	unsigned int nr_ids;
827 	s16 num_tc;
828 	struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
829 };
830 
831 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +	\
832 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
833 
834 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
835 	(_rxqs * (_tcs) * sizeof(struct xps_map *)))
836 
837 #endif /* CONFIG_XPS */
838 
839 #define TC_MAX_QUEUE	16
840 #define TC_BITMASK	15
841 /* HW offloaded queuing disciplines txq count and offset maps */
842 struct netdev_tc_txq {
843 	u16 count;
844 	u16 offset;
845 };
846 
847 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
848 /*
849  * This structure is to hold information about the device
850  * configured to run FCoE protocol stack.
851  */
852 struct netdev_fcoe_hbainfo {
853 	char	manufacturer[64];
854 	char	serial_number[64];
855 	char	hardware_version[64];
856 	char	driver_version[64];
857 	char	optionrom_version[64];
858 	char	firmware_version[64];
859 	char	model[256];
860 	char	model_description[256];
861 };
862 #endif
863 
864 #define MAX_PHYS_ITEM_ID_LEN 32
865 
866 /* This structure holds a unique identifier to identify some
867  * physical item (port for example) used by a netdevice.
868  */
869 struct netdev_phys_item_id {
870 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
871 	unsigned char id_len;
872 };
873 
874 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
875 					    struct netdev_phys_item_id *b)
876 {
877 	return a->id_len == b->id_len &&
878 	       memcmp(a->id, b->id, a->id_len) == 0;
879 }
880 
881 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
882 				       struct sk_buff *skb,
883 				       struct net_device *sb_dev);
884 
885 enum net_device_path_type {
886 	DEV_PATH_ETHERNET = 0,
887 	DEV_PATH_VLAN,
888 	DEV_PATH_BRIDGE,
889 	DEV_PATH_PPPOE,
890 	DEV_PATH_DSA,
891 	DEV_PATH_MTK_WDMA,
892 };
893 
894 struct net_device_path {
895 	enum net_device_path_type	type;
896 	const struct net_device		*dev;
897 	union {
898 		struct {
899 			u16		id;
900 			__be16		proto;
901 			u8		h_dest[ETH_ALEN];
902 		} encap;
903 		struct {
904 			enum {
905 				DEV_PATH_BR_VLAN_KEEP,
906 				DEV_PATH_BR_VLAN_TAG,
907 				DEV_PATH_BR_VLAN_UNTAG,
908 				DEV_PATH_BR_VLAN_UNTAG_HW,
909 			}		vlan_mode;
910 			u16		vlan_id;
911 			__be16		vlan_proto;
912 		} bridge;
913 		struct {
914 			int port;
915 			u16 proto;
916 		} dsa;
917 		struct {
918 			u8 wdma_idx;
919 			u8 queue;
920 			u16 wcid;
921 			u8 bss;
922 			u8 amsdu;
923 		} mtk_wdma;
924 	};
925 };
926 
927 #define NET_DEVICE_PATH_STACK_MAX	5
928 #define NET_DEVICE_PATH_VLAN_MAX	2
929 
930 struct net_device_path_stack {
931 	int			num_paths;
932 	struct net_device_path	path[NET_DEVICE_PATH_STACK_MAX];
933 };
934 
935 struct net_device_path_ctx {
936 	const struct net_device *dev;
937 	u8			daddr[ETH_ALEN];
938 
939 	int			num_vlans;
940 	struct {
941 		u16		id;
942 		__be16		proto;
943 	} vlan[NET_DEVICE_PATH_VLAN_MAX];
944 };
945 
946 enum tc_setup_type {
947 	TC_QUERY_CAPS,
948 	TC_SETUP_QDISC_MQPRIO,
949 	TC_SETUP_CLSU32,
950 	TC_SETUP_CLSFLOWER,
951 	TC_SETUP_CLSMATCHALL,
952 	TC_SETUP_CLSBPF,
953 	TC_SETUP_BLOCK,
954 	TC_SETUP_QDISC_CBS,
955 	TC_SETUP_QDISC_RED,
956 	TC_SETUP_QDISC_PRIO,
957 	TC_SETUP_QDISC_MQ,
958 	TC_SETUP_QDISC_ETF,
959 	TC_SETUP_ROOT_QDISC,
960 	TC_SETUP_QDISC_GRED,
961 	TC_SETUP_QDISC_TAPRIO,
962 	TC_SETUP_FT,
963 	TC_SETUP_QDISC_ETS,
964 	TC_SETUP_QDISC_TBF,
965 	TC_SETUP_QDISC_FIFO,
966 	TC_SETUP_QDISC_HTB,
967 	TC_SETUP_ACT,
968 };
969 
970 /* These structures hold the attributes of bpf state that are being passed
971  * to the netdevice through the bpf op.
972  */
973 enum bpf_netdev_command {
974 	/* Set or clear a bpf program used in the earliest stages of packet
975 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
976 	 * is responsible for calling bpf_prog_put on any old progs that are
977 	 * stored. In case of error, the callee need not release the new prog
978 	 * reference, but on success it takes ownership and must bpf_prog_put
979 	 * when it is no longer used.
980 	 */
981 	XDP_SETUP_PROG,
982 	XDP_SETUP_PROG_HW,
983 	/* BPF program for offload callbacks, invoked at program load time. */
984 	BPF_OFFLOAD_MAP_ALLOC,
985 	BPF_OFFLOAD_MAP_FREE,
986 	XDP_SETUP_XSK_POOL,
987 };
988 
989 struct bpf_prog_offload_ops;
990 struct netlink_ext_ack;
991 struct xdp_umem;
992 struct xdp_dev_bulk_queue;
993 struct bpf_xdp_link;
994 
995 enum bpf_xdp_mode {
996 	XDP_MODE_SKB = 0,
997 	XDP_MODE_DRV = 1,
998 	XDP_MODE_HW = 2,
999 	__MAX_XDP_MODE
1000 };
1001 
1002 struct bpf_xdp_entity {
1003 	struct bpf_prog *prog;
1004 	struct bpf_xdp_link *link;
1005 };
1006 
1007 struct netdev_bpf {
1008 	enum bpf_netdev_command command;
1009 	union {
1010 		/* XDP_SETUP_PROG */
1011 		struct {
1012 			u32 flags;
1013 			struct bpf_prog *prog;
1014 			struct netlink_ext_ack *extack;
1015 		};
1016 		/* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1017 		struct {
1018 			struct bpf_offloaded_map *offmap;
1019 		};
1020 		/* XDP_SETUP_XSK_POOL */
1021 		struct {
1022 			struct xsk_buff_pool *pool;
1023 			u16 queue_id;
1024 		} xsk;
1025 	};
1026 };
1027 
1028 /* Flags for ndo_xsk_wakeup. */
1029 #define XDP_WAKEUP_RX (1 << 0)
1030 #define XDP_WAKEUP_TX (1 << 1)
1031 
1032 #ifdef CONFIG_XFRM_OFFLOAD
1033 struct xfrmdev_ops {
1034 	int	(*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1035 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
1036 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
1037 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
1038 				       struct xfrm_state *x);
1039 	void	(*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1040 	void	(*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1041 	int	(*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1042 	void	(*xdo_dev_policy_delete) (struct xfrm_policy *x);
1043 	void	(*xdo_dev_policy_free) (struct xfrm_policy *x);
1044 };
1045 #endif
1046 
1047 struct dev_ifalias {
1048 	struct rcu_head rcuhead;
1049 	char ifalias[];
1050 };
1051 
1052 struct devlink;
1053 struct tlsdev_ops;
1054 
1055 struct netdev_net_notifier {
1056 	struct list_head list;
1057 	struct notifier_block *nb;
1058 };
1059 
1060 /*
1061  * This structure defines the management hooks for network devices.
1062  * The following hooks can be defined; unless noted otherwise, they are
1063  * optional and can be filled with a null pointer.
1064  *
1065  * int (*ndo_init)(struct net_device *dev);
1066  *     This function is called once when a network device is registered.
1067  *     The network device can use this for any late stage initialization
1068  *     or semantic validation. It can fail with an error code which will
1069  *     be propagated back to register_netdev.
1070  *
1071  * void (*ndo_uninit)(struct net_device *dev);
1072  *     This function is called when device is unregistered or when registration
1073  *     fails. It is not called if init fails.
1074  *
1075  * int (*ndo_open)(struct net_device *dev);
1076  *     This function is called when a network device transitions to the up
1077  *     state.
1078  *
1079  * int (*ndo_stop)(struct net_device *dev);
1080  *     This function is called when a network device transitions to the down
1081  *     state.
1082  *
1083  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1084  *                               struct net_device *dev);
1085  *	Called when a packet needs to be transmitted.
1086  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
1087  *	the queue before that can happen; it's for obsolete devices and weird
1088  *	corner cases, but the stack really does a non-trivial amount
1089  *	of useless work if you return NETDEV_TX_BUSY.
1090  *	Required; cannot be NULL.
1091  *
1092  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1093  *					   struct net_device *dev
1094  *					   netdev_features_t features);
1095  *	Called by core transmit path to determine if device is capable of
1096  *	performing offload operations on a given packet. This is to give
1097  *	the device an opportunity to implement any restrictions that cannot
1098  *	be otherwise expressed by feature flags. The check is called with
1099  *	the set of features that the stack has calculated and it returns
1100  *	those the driver believes to be appropriate.
1101  *
1102  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1103  *                         struct net_device *sb_dev);
1104  *	Called to decide which queue to use when device supports multiple
1105  *	transmit queues.
1106  *
1107  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1108  *	This function is called to allow device receiver to make
1109  *	changes to configuration when multicast or promiscuous is enabled.
1110  *
1111  * void (*ndo_set_rx_mode)(struct net_device *dev);
1112  *	This function is called device changes address list filtering.
1113  *	If driver handles unicast address filtering, it should set
1114  *	IFF_UNICAST_FLT in its priv_flags.
1115  *
1116  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1117  *	This function  is called when the Media Access Control address
1118  *	needs to be changed. If this interface is not defined, the
1119  *	MAC address can not be changed.
1120  *
1121  * int (*ndo_validate_addr)(struct net_device *dev);
1122  *	Test if Media Access Control address is valid for the device.
1123  *
1124  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1125  *	Old-style ioctl entry point. This is used internally by the
1126  *	appletalk and ieee802154 subsystems but is no longer called by
1127  *	the device ioctl handler.
1128  *
1129  * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1130  *	Used by the bonding driver for its device specific ioctls:
1131  *	SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1132  *	SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1133  *
1134  * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1135  *	Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1136  *	SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1137  *
1138  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1139  *	Used to set network devices bus interface parameters. This interface
1140  *	is retained for legacy reasons; new devices should use the bus
1141  *	interface (PCI) for low level management.
1142  *
1143  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1144  *	Called when a user wants to change the Maximum Transfer Unit
1145  *	of a device.
1146  *
1147  * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1148  *	Callback used when the transmitter has not made any progress
1149  *	for dev->watchdog ticks.
1150  *
1151  * void (*ndo_get_stats64)(struct net_device *dev,
1152  *                         struct rtnl_link_stats64 *storage);
1153  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1154  *	Called when a user wants to get the network device usage
1155  *	statistics. Drivers must do one of the following:
1156  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
1157  *	   rtnl_link_stats64 structure passed by the caller.
1158  *	2. Define @ndo_get_stats to update a net_device_stats structure
1159  *	   (which should normally be dev->stats) and return a pointer to
1160  *	   it. The structure may be changed asynchronously only if each
1161  *	   field is written atomically.
1162  *	3. Update dev->stats asynchronously and atomically, and define
1163  *	   neither operation.
1164  *
1165  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1166  *	Return true if this device supports offload stats of this attr_id.
1167  *
1168  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1169  *	void *attr_data)
1170  *	Get statistics for offload operations by attr_id. Write it into the
1171  *	attr_data pointer.
1172  *
1173  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1174  *	If device supports VLAN filtering this function is called when a
1175  *	VLAN id is registered.
1176  *
1177  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1178  *	If device supports VLAN filtering this function is called when a
1179  *	VLAN id is unregistered.
1180  *
1181  * void (*ndo_poll_controller)(struct net_device *dev);
1182  *
1183  *	SR-IOV management functions.
1184  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1185  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1186  *			  u8 qos, __be16 proto);
1187  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1188  *			  int max_tx_rate);
1189  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1190  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1191  * int (*ndo_get_vf_config)(struct net_device *dev,
1192  *			    int vf, struct ifla_vf_info *ivf);
1193  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1194  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1195  *			  struct nlattr *port[]);
1196  *
1197  *      Enable or disable the VF ability to query its RSS Redirection Table and
1198  *      Hash Key. This is needed since on some devices VF share this information
1199  *      with PF and querying it may introduce a theoretical security risk.
1200  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1201  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1202  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1203  *		       void *type_data);
1204  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
1205  *	This is always called from the stack with the rtnl lock held and netif
1206  *	tx queues stopped. This allows the netdevice to perform queue
1207  *	management safely.
1208  *
1209  *	Fiber Channel over Ethernet (FCoE) offload functions.
1210  * int (*ndo_fcoe_enable)(struct net_device *dev);
1211  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
1212  *	so the underlying device can perform whatever needed configuration or
1213  *	initialization to support acceleration of FCoE traffic.
1214  *
1215  * int (*ndo_fcoe_disable)(struct net_device *dev);
1216  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
1217  *	so the underlying device can perform whatever needed clean-ups to
1218  *	stop supporting acceleration of FCoE traffic.
1219  *
1220  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1221  *			     struct scatterlist *sgl, unsigned int sgc);
1222  *	Called when the FCoE Initiator wants to initialize an I/O that
1223  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1224  *	perform necessary setup and returns 1 to indicate the device is set up
1225  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1226  *
1227  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1228  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1229  *	indicated by the FC exchange id 'xid', so the underlying device can
1230  *	clean up and reuse resources for later DDP requests.
1231  *
1232  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1233  *			      struct scatterlist *sgl, unsigned int sgc);
1234  *	Called when the FCoE Target wants to initialize an I/O that
1235  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1236  *	perform necessary setup and returns 1 to indicate the device is set up
1237  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1238  *
1239  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1240  *			       struct netdev_fcoe_hbainfo *hbainfo);
1241  *	Called when the FCoE Protocol stack wants information on the underlying
1242  *	device. This information is utilized by the FCoE protocol stack to
1243  *	register attributes with Fiber Channel management service as per the
1244  *	FC-GS Fabric Device Management Information(FDMI) specification.
1245  *
1246  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1247  *	Called when the underlying device wants to override default World Wide
1248  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1249  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1250  *	protocol stack to use.
1251  *
1252  *	RFS acceleration.
1253  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1254  *			    u16 rxq_index, u32 flow_id);
1255  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1256  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1257  *	Return the filter ID on success, or a negative error code.
1258  *
1259  *	Slave management functions (for bridge, bonding, etc).
1260  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1261  *	Called to make another netdev an underling.
1262  *
1263  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1264  *	Called to release previously enslaved netdev.
1265  *
1266  * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1267  *					    struct sk_buff *skb,
1268  *					    bool all_slaves);
1269  *	Get the xmit slave of master device. If all_slaves is true, function
1270  *	assume all the slaves can transmit.
1271  *
1272  *      Feature/offload setting functions.
1273  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1274  *		netdev_features_t features);
1275  *	Adjusts the requested feature flags according to device-specific
1276  *	constraints, and returns the resulting flags. Must not modify
1277  *	the device state.
1278  *
1279  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1280  *	Called to update device configuration to new features. Passed
1281  *	feature set might be less than what was returned by ndo_fix_features()).
1282  *	Must return >0 or -errno if it changed dev->features itself.
1283  *
1284  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1285  *		      struct net_device *dev,
1286  *		      const unsigned char *addr, u16 vid, u16 flags,
1287  *		      struct netlink_ext_ack *extack);
1288  *	Adds an FDB entry to dev for addr.
1289  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1290  *		      struct net_device *dev,
1291  *		      const unsigned char *addr, u16 vid)
1292  *	Deletes the FDB entry from dev coresponding to addr.
1293  * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1294  *			   struct net_device *dev,
1295  *			   u16 vid,
1296  *			   struct netlink_ext_ack *extack);
1297  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1298  *		       struct net_device *dev, struct net_device *filter_dev,
1299  *		       int *idx)
1300  *	Used to add FDB entries to dump requests. Implementers should add
1301  *	entries to skb and update idx with the number of entries.
1302  *
1303  * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1304  *		      u16 nlmsg_flags, struct netlink_ext_ack *extack);
1305  *	Adds an MDB entry to dev.
1306  * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1307  *		      struct netlink_ext_ack *extack);
1308  *	Deletes the MDB entry from dev.
1309  * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1310  *		       struct netlink_callback *cb);
1311  *	Dumps MDB entries from dev. The first argument (marker) in the netlink
1312  *	callback is used by core rtnetlink code.
1313  *
1314  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1315  *			     u16 flags, struct netlink_ext_ack *extack)
1316  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1317  *			     struct net_device *dev, u32 filter_mask,
1318  *			     int nlflags)
1319  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1320  *			     u16 flags);
1321  *
1322  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1323  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1324  *	which do not represent real hardware may define this to allow their
1325  *	userspace components to manage their virtual carrier state. Devices
1326  *	that determine carrier state from physical hardware properties (eg
1327  *	network cables) or protocol-dependent mechanisms (eg
1328  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1329  *
1330  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1331  *			       struct netdev_phys_item_id *ppid);
1332  *	Called to get ID of physical port of this device. If driver does
1333  *	not implement this, it is assumed that the hw is not able to have
1334  *	multiple net devices on single physical port.
1335  *
1336  * int (*ndo_get_port_parent_id)(struct net_device *dev,
1337  *				 struct netdev_phys_item_id *ppid)
1338  *	Called to get the parent ID of the physical port of this device.
1339  *
1340  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1341  *				 struct net_device *dev)
1342  *	Called by upper layer devices to accelerate switching or other
1343  *	station functionality into hardware. 'pdev is the lowerdev
1344  *	to use for the offload and 'dev' is the net device that will
1345  *	back the offload. Returns a pointer to the private structure
1346  *	the upper layer will maintain.
1347  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1348  *	Called by upper layer device to delete the station created
1349  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1350  *	the station and priv is the structure returned by the add
1351  *	operation.
1352  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1353  *			     int queue_index, u32 maxrate);
1354  *	Called when a user wants to set a max-rate limitation of specific
1355  *	TX queue.
1356  * int (*ndo_get_iflink)(const struct net_device *dev);
1357  *	Called to get the iflink value of this device.
1358  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1359  *	This function is used to get egress tunnel information for given skb.
1360  *	This is useful for retrieving outer tunnel header parameters while
1361  *	sampling packet.
1362  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1363  *	This function is used to specify the headroom that the skb must
1364  *	consider when allocation skb during packet reception. Setting
1365  *	appropriate rx headroom value allows avoiding skb head copy on
1366  *	forward. Setting a negative value resets the rx headroom to the
1367  *	default value.
1368  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1369  *	This function is used to set or query state related to XDP on the
1370  *	netdevice and manage BPF offload. See definition of
1371  *	enum bpf_netdev_command for details.
1372  * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1373  *			u32 flags);
1374  *	This function is used to submit @n XDP packets for transmit on a
1375  *	netdevice. Returns number of frames successfully transmitted, frames
1376  *	that got dropped are freed/returned via xdp_return_frame().
1377  *	Returns negative number, means general error invoking ndo, meaning
1378  *	no frames were xmit'ed and core-caller will free all frames.
1379  * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1380  *					        struct xdp_buff *xdp);
1381  *      Get the xmit slave of master device based on the xdp_buff.
1382  * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1383  *      This function is used to wake up the softirq, ksoftirqd or kthread
1384  *	responsible for sending and/or receiving packets on a specific
1385  *	queue id bound to an AF_XDP socket. The flags field specifies if
1386  *	only RX, only Tx, or both should be woken up using the flags
1387  *	XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1388  * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1389  *			 int cmd);
1390  *	Add, change, delete or get information on an IPv4 tunnel.
1391  * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1392  *	If a device is paired with a peer device, return the peer instance.
1393  *	The caller must be under RCU read context.
1394  * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1395  *     Get the forwarding path to reach the real device from the HW destination address
1396  * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1397  *			     const struct skb_shared_hwtstamps *hwtstamps,
1398  *			     bool cycles);
1399  *	Get hardware timestamp based on normal/adjustable time or free running
1400  *	cycle counter. This function is required if physical clock supports a
1401  *	free running cycle counter.
1402  *
1403  * int (*ndo_hwtstamp_get)(struct net_device *dev,
1404  *			   struct kernel_hwtstamp_config *kernel_config);
1405  *	Get the currently configured hardware timestamping parameters for the
1406  *	NIC device.
1407  *
1408  * int (*ndo_hwtstamp_set)(struct net_device *dev,
1409  *			   struct kernel_hwtstamp_config *kernel_config,
1410  *			   struct netlink_ext_ack *extack);
1411  *	Change the hardware timestamping parameters for NIC device.
1412  */
1413 struct net_device_ops {
1414 	int			(*ndo_init)(struct net_device *dev);
1415 	void			(*ndo_uninit)(struct net_device *dev);
1416 	int			(*ndo_open)(struct net_device *dev);
1417 	int			(*ndo_stop)(struct net_device *dev);
1418 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1419 						  struct net_device *dev);
1420 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1421 						      struct net_device *dev,
1422 						      netdev_features_t features);
1423 	u16			(*ndo_select_queue)(struct net_device *dev,
1424 						    struct sk_buff *skb,
1425 						    struct net_device *sb_dev);
1426 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1427 						       int flags);
1428 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1429 	int			(*ndo_set_mac_address)(struct net_device *dev,
1430 						       void *addr);
1431 	int			(*ndo_validate_addr)(struct net_device *dev);
1432 	int			(*ndo_do_ioctl)(struct net_device *dev,
1433 					        struct ifreq *ifr, int cmd);
1434 	int			(*ndo_eth_ioctl)(struct net_device *dev,
1435 						 struct ifreq *ifr, int cmd);
1436 	int			(*ndo_siocbond)(struct net_device *dev,
1437 						struct ifreq *ifr, int cmd);
1438 	int			(*ndo_siocwandev)(struct net_device *dev,
1439 						  struct if_settings *ifs);
1440 	int			(*ndo_siocdevprivate)(struct net_device *dev,
1441 						      struct ifreq *ifr,
1442 						      void __user *data, int cmd);
1443 	int			(*ndo_set_config)(struct net_device *dev,
1444 					          struct ifmap *map);
1445 	int			(*ndo_change_mtu)(struct net_device *dev,
1446 						  int new_mtu);
1447 	int			(*ndo_neigh_setup)(struct net_device *dev,
1448 						   struct neigh_parms *);
1449 	void			(*ndo_tx_timeout) (struct net_device *dev,
1450 						   unsigned int txqueue);
1451 
1452 	void			(*ndo_get_stats64)(struct net_device *dev,
1453 						   struct rtnl_link_stats64 *storage);
1454 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1455 	int			(*ndo_get_offload_stats)(int attr_id,
1456 							 const struct net_device *dev,
1457 							 void *attr_data);
1458 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1459 
1460 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1461 						       __be16 proto, u16 vid);
1462 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1463 						        __be16 proto, u16 vid);
1464 #ifdef CONFIG_NET_POLL_CONTROLLER
1465 	void                    (*ndo_poll_controller)(struct net_device *dev);
1466 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1467 						     struct netpoll_info *info);
1468 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1469 #endif
1470 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1471 						  int queue, u8 *mac);
1472 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1473 						   int queue, u16 vlan,
1474 						   u8 qos, __be16 proto);
1475 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1476 						   int vf, int min_tx_rate,
1477 						   int max_tx_rate);
1478 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1479 						       int vf, bool setting);
1480 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1481 						    int vf, bool setting);
1482 	int			(*ndo_get_vf_config)(struct net_device *dev,
1483 						     int vf,
1484 						     struct ifla_vf_info *ivf);
1485 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1486 							 int vf, int link_state);
1487 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1488 						    int vf,
1489 						    struct ifla_vf_stats
1490 						    *vf_stats);
1491 	int			(*ndo_set_vf_port)(struct net_device *dev,
1492 						   int vf,
1493 						   struct nlattr *port[]);
1494 	int			(*ndo_get_vf_port)(struct net_device *dev,
1495 						   int vf, struct sk_buff *skb);
1496 	int			(*ndo_get_vf_guid)(struct net_device *dev,
1497 						   int vf,
1498 						   struct ifla_vf_guid *node_guid,
1499 						   struct ifla_vf_guid *port_guid);
1500 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1501 						   int vf, u64 guid,
1502 						   int guid_type);
1503 	int			(*ndo_set_vf_rss_query_en)(
1504 						   struct net_device *dev,
1505 						   int vf, bool setting);
1506 	int			(*ndo_setup_tc)(struct net_device *dev,
1507 						enum tc_setup_type type,
1508 						void *type_data);
1509 #if IS_ENABLED(CONFIG_FCOE)
1510 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1511 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1512 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1513 						      u16 xid,
1514 						      struct scatterlist *sgl,
1515 						      unsigned int sgc);
1516 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1517 						     u16 xid);
1518 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1519 						       u16 xid,
1520 						       struct scatterlist *sgl,
1521 						       unsigned int sgc);
1522 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1523 							struct netdev_fcoe_hbainfo *hbainfo);
1524 #endif
1525 
1526 #if IS_ENABLED(CONFIG_LIBFCOE)
1527 #define NETDEV_FCOE_WWNN 0
1528 #define NETDEV_FCOE_WWPN 1
1529 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1530 						    u64 *wwn, int type);
1531 #endif
1532 
1533 #ifdef CONFIG_RFS_ACCEL
1534 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1535 						     const struct sk_buff *skb,
1536 						     u16 rxq_index,
1537 						     u32 flow_id);
1538 #endif
1539 	int			(*ndo_add_slave)(struct net_device *dev,
1540 						 struct net_device *slave_dev,
1541 						 struct netlink_ext_ack *extack);
1542 	int			(*ndo_del_slave)(struct net_device *dev,
1543 						 struct net_device *slave_dev);
1544 	struct net_device*	(*ndo_get_xmit_slave)(struct net_device *dev,
1545 						      struct sk_buff *skb,
1546 						      bool all_slaves);
1547 	struct net_device*	(*ndo_sk_get_lower_dev)(struct net_device *dev,
1548 							struct sock *sk);
1549 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1550 						    netdev_features_t features);
1551 	int			(*ndo_set_features)(struct net_device *dev,
1552 						    netdev_features_t features);
1553 	int			(*ndo_neigh_construct)(struct net_device *dev,
1554 						       struct neighbour *n);
1555 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1556 						     struct neighbour *n);
1557 
1558 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1559 					       struct nlattr *tb[],
1560 					       struct net_device *dev,
1561 					       const unsigned char *addr,
1562 					       u16 vid,
1563 					       u16 flags,
1564 					       struct netlink_ext_ack *extack);
1565 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1566 					       struct nlattr *tb[],
1567 					       struct net_device *dev,
1568 					       const unsigned char *addr,
1569 					       u16 vid, struct netlink_ext_ack *extack);
1570 	int			(*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1571 						    struct nlattr *tb[],
1572 						    struct net_device *dev,
1573 						    u16 vid,
1574 						    struct netlink_ext_ack *extack);
1575 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1576 						struct netlink_callback *cb,
1577 						struct net_device *dev,
1578 						struct net_device *filter_dev,
1579 						int *idx);
1580 	int			(*ndo_fdb_get)(struct sk_buff *skb,
1581 					       struct nlattr *tb[],
1582 					       struct net_device *dev,
1583 					       const unsigned char *addr,
1584 					       u16 vid, u32 portid, u32 seq,
1585 					       struct netlink_ext_ack *extack);
1586 	int			(*ndo_mdb_add)(struct net_device *dev,
1587 					       struct nlattr *tb[],
1588 					       u16 nlmsg_flags,
1589 					       struct netlink_ext_ack *extack);
1590 	int			(*ndo_mdb_del)(struct net_device *dev,
1591 					       struct nlattr *tb[],
1592 					       struct netlink_ext_ack *extack);
1593 	int			(*ndo_mdb_dump)(struct net_device *dev,
1594 						struct sk_buff *skb,
1595 						struct netlink_callback *cb);
1596 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1597 						      struct nlmsghdr *nlh,
1598 						      u16 flags,
1599 						      struct netlink_ext_ack *extack);
1600 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1601 						      u32 pid, u32 seq,
1602 						      struct net_device *dev,
1603 						      u32 filter_mask,
1604 						      int nlflags);
1605 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1606 						      struct nlmsghdr *nlh,
1607 						      u16 flags);
1608 	int			(*ndo_change_carrier)(struct net_device *dev,
1609 						      bool new_carrier);
1610 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1611 							struct netdev_phys_item_id *ppid);
1612 	int			(*ndo_get_port_parent_id)(struct net_device *dev,
1613 							  struct netdev_phys_item_id *ppid);
1614 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1615 							  char *name, size_t len);
1616 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1617 							struct net_device *dev);
1618 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1619 							void *priv);
1620 
1621 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1622 						      int queue_index,
1623 						      u32 maxrate);
1624 	int			(*ndo_get_iflink)(const struct net_device *dev);
1625 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1626 						       struct sk_buff *skb);
1627 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1628 						       int needed_headroom);
1629 	int			(*ndo_bpf)(struct net_device *dev,
1630 					   struct netdev_bpf *bpf);
1631 	int			(*ndo_xdp_xmit)(struct net_device *dev, int n,
1632 						struct xdp_frame **xdp,
1633 						u32 flags);
1634 	struct net_device *	(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1635 							  struct xdp_buff *xdp);
1636 	int			(*ndo_xsk_wakeup)(struct net_device *dev,
1637 						  u32 queue_id, u32 flags);
1638 	int			(*ndo_tunnel_ctl)(struct net_device *dev,
1639 						  struct ip_tunnel_parm *p, int cmd);
1640 	struct net_device *	(*ndo_get_peer_dev)(struct net_device *dev);
1641 	int                     (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1642                                                          struct net_device_path *path);
1643 	ktime_t			(*ndo_get_tstamp)(struct net_device *dev,
1644 						  const struct skb_shared_hwtstamps *hwtstamps,
1645 						  bool cycles);
1646 	int			(*ndo_hwtstamp_get)(struct net_device *dev,
1647 						    struct kernel_hwtstamp_config *kernel_config);
1648 	int			(*ndo_hwtstamp_set)(struct net_device *dev,
1649 						    struct kernel_hwtstamp_config *kernel_config,
1650 						    struct netlink_ext_ack *extack);
1651 };
1652 
1653 /**
1654  * enum netdev_priv_flags - &struct net_device priv_flags
1655  *
1656  * These are the &struct net_device, they are only set internally
1657  * by drivers and used in the kernel. These flags are invisible to
1658  * userspace; this means that the order of these flags can change
1659  * during any kernel release.
1660  *
1661  * You should have a pretty good reason to be extending these flags.
1662  *
1663  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1664  * @IFF_EBRIDGE: Ethernet bridging device
1665  * @IFF_BONDING: bonding master or slave
1666  * @IFF_ISATAP: ISATAP interface (RFC4214)
1667  * @IFF_WAN_HDLC: WAN HDLC device
1668  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1669  *	release skb->dst
1670  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1671  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1672  * @IFF_MACVLAN_PORT: device used as macvlan port
1673  * @IFF_BRIDGE_PORT: device used as bridge port
1674  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1675  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1676  * @IFF_UNICAST_FLT: Supports unicast filtering
1677  * @IFF_TEAM_PORT: device used as team port
1678  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1679  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1680  *	change when it's running
1681  * @IFF_MACVLAN: Macvlan device
1682  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1683  *	underlying stacked devices
1684  * @IFF_L3MDEV_MASTER: device is an L3 master device
1685  * @IFF_NO_QUEUE: device can run without qdisc attached
1686  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1687  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1688  * @IFF_TEAM: device is a team device
1689  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1690  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1691  *	entity (i.e. the master device for bridged veth)
1692  * @IFF_MACSEC: device is a MACsec device
1693  * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1694  * @IFF_FAILOVER: device is a failover master device
1695  * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1696  * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1697  * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1698  * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1699  *	skb_headlen(skb) == 0 (data starts from frag0)
1700  * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1701  * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1702  *	ndo_hwtstamp_set() for all timestamp requests regardless of source,
1703  *	even if those aren't HWTSTAMP_SOURCE_NETDEV.
1704  */
1705 enum netdev_priv_flags {
1706 	IFF_802_1Q_VLAN			= 1<<0,
1707 	IFF_EBRIDGE			= 1<<1,
1708 	IFF_BONDING			= 1<<2,
1709 	IFF_ISATAP			= 1<<3,
1710 	IFF_WAN_HDLC			= 1<<4,
1711 	IFF_XMIT_DST_RELEASE		= 1<<5,
1712 	IFF_DONT_BRIDGE			= 1<<6,
1713 	IFF_DISABLE_NETPOLL		= 1<<7,
1714 	IFF_MACVLAN_PORT		= 1<<8,
1715 	IFF_BRIDGE_PORT			= 1<<9,
1716 	IFF_OVS_DATAPATH		= 1<<10,
1717 	IFF_TX_SKB_SHARING		= 1<<11,
1718 	IFF_UNICAST_FLT			= 1<<12,
1719 	IFF_TEAM_PORT			= 1<<13,
1720 	IFF_SUPP_NOFCS			= 1<<14,
1721 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1722 	IFF_MACVLAN			= 1<<16,
1723 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1724 	IFF_L3MDEV_MASTER		= 1<<18,
1725 	IFF_NO_QUEUE			= 1<<19,
1726 	IFF_OPENVSWITCH			= 1<<20,
1727 	IFF_L3MDEV_SLAVE		= 1<<21,
1728 	IFF_TEAM			= 1<<22,
1729 	IFF_RXFH_CONFIGURED		= 1<<23,
1730 	IFF_PHONY_HEADROOM		= 1<<24,
1731 	IFF_MACSEC			= 1<<25,
1732 	IFF_NO_RX_HANDLER		= 1<<26,
1733 	IFF_FAILOVER			= 1<<27,
1734 	IFF_FAILOVER_SLAVE		= 1<<28,
1735 	IFF_L3MDEV_RX_HANDLER		= 1<<29,
1736 	IFF_NO_ADDRCONF			= BIT_ULL(30),
1737 	IFF_TX_SKB_NO_LINEAR		= BIT_ULL(31),
1738 	IFF_CHANGE_PROTO_DOWN		= BIT_ULL(32),
1739 	IFF_SEE_ALL_HWTSTAMP_REQUESTS	= BIT_ULL(33),
1740 };
1741 
1742 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1743 #define IFF_EBRIDGE			IFF_EBRIDGE
1744 #define IFF_BONDING			IFF_BONDING
1745 #define IFF_ISATAP			IFF_ISATAP
1746 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1747 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1748 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1749 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1750 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1751 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1752 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1753 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1754 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1755 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1756 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1757 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1758 #define IFF_MACVLAN			IFF_MACVLAN
1759 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1760 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1761 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1762 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1763 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1764 #define IFF_TEAM			IFF_TEAM
1765 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1766 #define IFF_PHONY_HEADROOM		IFF_PHONY_HEADROOM
1767 #define IFF_MACSEC			IFF_MACSEC
1768 #define IFF_NO_RX_HANDLER		IFF_NO_RX_HANDLER
1769 #define IFF_FAILOVER			IFF_FAILOVER
1770 #define IFF_FAILOVER_SLAVE		IFF_FAILOVER_SLAVE
1771 #define IFF_L3MDEV_RX_HANDLER		IFF_L3MDEV_RX_HANDLER
1772 #define IFF_TX_SKB_NO_LINEAR		IFF_TX_SKB_NO_LINEAR
1773 
1774 /* Specifies the type of the struct net_device::ml_priv pointer */
1775 enum netdev_ml_priv_type {
1776 	ML_PRIV_NONE,
1777 	ML_PRIV_CAN,
1778 };
1779 
1780 /**
1781  *	struct net_device - The DEVICE structure.
1782  *
1783  *	Actually, this whole structure is a big mistake.  It mixes I/O
1784  *	data with strictly "high-level" data, and it has to know about
1785  *	almost every data structure used in the INET module.
1786  *
1787  *	@name:	This is the first field of the "visible" part of this structure
1788  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1789  *		of the interface.
1790  *
1791  *	@name_node:	Name hashlist node
1792  *	@ifalias:	SNMP alias
1793  *	@mem_end:	Shared memory end
1794  *	@mem_start:	Shared memory start
1795  *	@base_addr:	Device I/O address
1796  *	@irq:		Device IRQ number
1797  *
1798  *	@state:		Generic network queuing layer state, see netdev_state_t
1799  *	@dev_list:	The global list of network devices
1800  *	@napi_list:	List entry used for polling NAPI devices
1801  *	@unreg_list:	List entry  when we are unregistering the
1802  *			device; see the function unregister_netdev
1803  *	@close_list:	List entry used when we are closing the device
1804  *	@ptype_all:     Device-specific packet handlers for all protocols
1805  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1806  *
1807  *	@adj_list:	Directly linked devices, like slaves for bonding
1808  *	@features:	Currently active device features
1809  *	@hw_features:	User-changeable features
1810  *
1811  *	@wanted_features:	User-requested features
1812  *	@vlan_features:		Mask of features inheritable by VLAN devices
1813  *
1814  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1815  *				This field indicates what encapsulation
1816  *				offloads the hardware is capable of doing,
1817  *				and drivers will need to set them appropriately.
1818  *
1819  *	@mpls_features:	Mask of features inheritable by MPLS
1820  *	@gso_partial_features: value(s) from NETIF_F_GSO\*
1821  *
1822  *	@ifindex:	interface index
1823  *	@group:		The group the device belongs to
1824  *
1825  *	@stats:		Statistics struct, which was left as a legacy, use
1826  *			rtnl_link_stats64 instead
1827  *
1828  *	@core_stats:	core networking counters,
1829  *			do not use this in drivers
1830  *	@carrier_up_count:	Number of times the carrier has been up
1831  *	@carrier_down_count:	Number of times the carrier has been down
1832  *
1833  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1834  *				instead of ioctl,
1835  *				see <net/iw_handler.h> for details.
1836  *	@wireless_data:	Instance data managed by the core of wireless extensions
1837  *
1838  *	@netdev_ops:	Includes several pointers to callbacks,
1839  *			if one wants to override the ndo_*() functions
1840  *	@xdp_metadata_ops:	Includes pointers to XDP metadata callbacks.
1841  *	@ethtool_ops:	Management operations
1842  *	@l3mdev_ops:	Layer 3 master device operations
1843  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1844  *			discovery handling. Necessary for e.g. 6LoWPAN.
1845  *	@xfrmdev_ops:	Transformation offload operations
1846  *	@tlsdev_ops:	Transport Layer Security offload operations
1847  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1848  *			of Layer 2 headers.
1849  *
1850  *	@flags:		Interface flags (a la BSD)
1851  *	@xdp_features:	XDP capability supported by the device
1852  *	@priv_flags:	Like 'flags' but invisible to userspace,
1853  *			see if.h for the definitions
1854  *	@gflags:	Global flags ( kept as legacy )
1855  *	@padded:	How much padding added by alloc_netdev()
1856  *	@operstate:	RFC2863 operstate
1857  *	@link_mode:	Mapping policy to operstate
1858  *	@if_port:	Selectable AUI, TP, ...
1859  *	@dma:		DMA channel
1860  *	@mtu:		Interface MTU value
1861  *	@min_mtu:	Interface Minimum MTU value
1862  *	@max_mtu:	Interface Maximum MTU value
1863  *	@type:		Interface hardware type
1864  *	@hard_header_len: Maximum hardware header length.
1865  *	@min_header_len:  Minimum hardware header length
1866  *
1867  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1868  *			  cases can this be guaranteed
1869  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1870  *			  cases can this be guaranteed. Some cases also use
1871  *			  LL_MAX_HEADER instead to allocate the skb
1872  *
1873  *	interface address info:
1874  *
1875  * 	@perm_addr:		Permanent hw address
1876  * 	@addr_assign_type:	Hw address assignment type
1877  * 	@addr_len:		Hardware address length
1878  *	@upper_level:		Maximum depth level of upper devices.
1879  *	@lower_level:		Maximum depth level of lower devices.
1880  *	@neigh_priv_len:	Used in neigh_alloc()
1881  * 	@dev_id:		Used to differentiate devices that share
1882  * 				the same link layer address
1883  * 	@dev_port:		Used to differentiate devices that share
1884  * 				the same function
1885  *	@addr_list_lock:	XXX: need comments on this one
1886  *	@name_assign_type:	network interface name assignment type
1887  *	@uc_promisc:		Counter that indicates promiscuous mode
1888  *				has been enabled due to the need to listen to
1889  *				additional unicast addresses in a device that
1890  *				does not implement ndo_set_rx_mode()
1891  *	@uc:			unicast mac addresses
1892  *	@mc:			multicast mac addresses
1893  *	@dev_addrs:		list of device hw addresses
1894  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1895  *	@promiscuity:		Number of times the NIC is told to work in
1896  *				promiscuous mode; if it becomes 0 the NIC will
1897  *				exit promiscuous mode
1898  *	@allmulti:		Counter, enables or disables allmulticast mode
1899  *
1900  *	@vlan_info:	VLAN info
1901  *	@dsa_ptr:	dsa specific data
1902  *	@tipc_ptr:	TIPC specific data
1903  *	@atalk_ptr:	AppleTalk link
1904  *	@ip_ptr:	IPv4 specific data
1905  *	@ip6_ptr:	IPv6 specific data
1906  *	@ax25_ptr:	AX.25 specific data
1907  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1908  *	@ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1909  *			 device struct
1910  *	@mpls_ptr:	mpls_dev struct pointer
1911  *	@mctp_ptr:	MCTP specific data
1912  *
1913  *	@dev_addr:	Hw address (before bcast,
1914  *			because most packets are unicast)
1915  *
1916  *	@_rx:			Array of RX queues
1917  *	@num_rx_queues:		Number of RX queues
1918  *				allocated at register_netdev() time
1919  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1920  *	@xdp_prog:		XDP sockets filter program pointer
1921  *	@gro_flush_timeout:	timeout for GRO layer in NAPI
1922  *	@napi_defer_hard_irqs:	If not zero, provides a counter that would
1923  *				allow to avoid NIC hard IRQ, on busy queues.
1924  *
1925  *	@rx_handler:		handler for received packets
1926  *	@rx_handler_data: 	XXX: need comments on this one
1927  *	@tcx_ingress:		BPF & clsact qdisc specific data for ingress processing
1928  *	@ingress_queue:		XXX: need comments on this one
1929  *	@nf_hooks_ingress:	netfilter hooks executed for ingress packets
1930  *	@broadcast:		hw bcast address
1931  *
1932  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1933  *			indexed by RX queue number. Assigned by driver.
1934  *			This must only be set if the ndo_rx_flow_steer
1935  *			operation is defined
1936  *	@index_hlist:		Device index hash chain
1937  *
1938  *	@_tx:			Array of TX queues
1939  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1940  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1941  *	@qdisc:			Root qdisc from userspace point of view
1942  *	@tx_queue_len:		Max frames per queue allowed
1943  *	@tx_global_lock: 	XXX: need comments on this one
1944  *	@xdp_bulkq:		XDP device bulk queue
1945  *	@xps_maps:		all CPUs/RXQs maps for XPS device
1946  *
1947  *	@xps_maps:	XXX: need comments on this one
1948  *	@tcx_egress:		BPF & clsact qdisc specific data for egress processing
1949  *	@nf_hooks_egress:	netfilter hooks executed for egress packets
1950  *	@qdisc_hash:		qdisc hash table
1951  *	@watchdog_timeo:	Represents the timeout that is used by
1952  *				the watchdog (see dev_watchdog())
1953  *	@watchdog_timer:	List of timers
1954  *
1955  *	@proto_down_reason:	reason a netdev interface is held down
1956  *	@pcpu_refcnt:		Number of references to this device
1957  *	@dev_refcnt:		Number of references to this device
1958  *	@refcnt_tracker:	Tracker directory for tracked references to this device
1959  *	@todo_list:		Delayed register/unregister
1960  *	@link_watch_list:	XXX: need comments on this one
1961  *
1962  *	@reg_state:		Register/unregister state machine
1963  *	@dismantle:		Device is going to be freed
1964  *	@rtnl_link_state:	This enum represents the phases of creating
1965  *				a new link
1966  *
1967  *	@needs_free_netdev:	Should unregister perform free_netdev?
1968  *	@priv_destructor:	Called from unregister
1969  *	@npinfo:		XXX: need comments on this one
1970  * 	@nd_net:		Network namespace this network device is inside
1971  *
1972  * 	@ml_priv:	Mid-layer private
1973  *	@ml_priv_type:  Mid-layer private type
1974  * 	@lstats:	Loopback statistics
1975  * 	@tstats:	Tunnel statistics
1976  * 	@dstats:	Dummy statistics
1977  * 	@vstats:	Virtual ethernet statistics
1978  *
1979  *	@garp_port:	GARP
1980  *	@mrp_port:	MRP
1981  *
1982  *	@dm_private:	Drop monitor private
1983  *
1984  *	@dev:		Class/net/name entry
1985  *	@sysfs_groups:	Space for optional device, statistics and wireless
1986  *			sysfs groups
1987  *
1988  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1989  *	@rtnl_link_ops:	Rtnl_link_ops
1990  *
1991  *	@gso_max_size:	Maximum size of generic segmentation offload
1992  *	@tso_max_size:	Device (as in HW) limit on the max TSO request size
1993  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1994  *			NIC for GSO
1995  *	@tso_max_segs:	Device (as in HW) limit on the max TSO segment count
1996  * 	@gso_ipv4_max_size:	Maximum size of generic segmentation offload,
1997  * 				for IPv4.
1998  *
1999  *	@dcbnl_ops:	Data Center Bridging netlink ops
2000  *	@num_tc:	Number of traffic classes in the net device
2001  *	@tc_to_txq:	XXX: need comments on this one
2002  *	@prio_tc_map:	XXX: need comments on this one
2003  *
2004  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
2005  *
2006  *	@priomap:	XXX: need comments on this one
2007  *	@phydev:	Physical device may attach itself
2008  *			for hardware timestamping
2009  *	@sfp_bus:	attached &struct sfp_bus structure.
2010  *
2011  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2012  *
2013  *	@proto_down:	protocol port state information can be sent to the
2014  *			switch driver and used to set the phys state of the
2015  *			switch port.
2016  *
2017  *	@wol_enabled:	Wake-on-LAN is enabled
2018  *
2019  *	@threaded:	napi threaded mode is enabled
2020  *
2021  *	@net_notifier_list:	List of per-net netdev notifier block
2022  *				that follow this device when it is moved
2023  *				to another network namespace.
2024  *
2025  *	@macsec_ops:    MACsec offloading ops
2026  *
2027  *	@udp_tunnel_nic_info:	static structure describing the UDP tunnel
2028  *				offload capabilities of the device
2029  *	@udp_tunnel_nic:	UDP tunnel offload state
2030  *	@xdp_state:		stores info on attached XDP BPF programs
2031  *
2032  *	@nested_level:	Used as a parameter of spin_lock_nested() of
2033  *			dev->addr_list_lock.
2034  *	@unlink_list:	As netif_addr_lock() can be called recursively,
2035  *			keep a list of interfaces to be deleted.
2036  *	@gro_max_size:	Maximum size of aggregated packet in generic
2037  *			receive offload (GRO)
2038  * 	@gro_ipv4_max_size:	Maximum size of aggregated packet in generic
2039  * 				receive offload (GRO), for IPv4.
2040  *	@xdp_zc_max_segs:	Maximum number of segments supported by AF_XDP
2041  *				zero copy driver
2042  *
2043  *	@dev_addr_shadow:	Copy of @dev_addr to catch direct writes.
2044  *	@linkwatch_dev_tracker:	refcount tracker used by linkwatch.
2045  *	@watchdog_dev_tracker:	refcount tracker used by watchdog.
2046  *	@dev_registered_tracker:	tracker for reference held while
2047  *					registered
2048  *	@offload_xstats_l3:	L3 HW stats for this netdevice.
2049  *
2050  *	@devlink_port:	Pointer to related devlink port structure.
2051  *			Assigned by a driver before netdev registration using
2052  *			SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2053  *			during the time netdevice is registered.
2054  *
2055  *	@dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2056  *		   where the clock is recovered.
2057  *
2058  *	FIXME: cleanup struct net_device such that network protocol info
2059  *	moves out.
2060  */
2061 
2062 struct net_device {
2063 	char			name[IFNAMSIZ];
2064 	struct netdev_name_node	*name_node;
2065 	struct dev_ifalias	__rcu *ifalias;
2066 	/*
2067 	 *	I/O specific fields
2068 	 *	FIXME: Merge these and struct ifmap into one
2069 	 */
2070 	unsigned long		mem_end;
2071 	unsigned long		mem_start;
2072 	unsigned long		base_addr;
2073 
2074 	/*
2075 	 *	Some hardware also needs these fields (state,dev_list,
2076 	 *	napi_list,unreg_list,close_list) but they are not
2077 	 *	part of the usual set specified in Space.c.
2078 	 */
2079 
2080 	unsigned long		state;
2081 
2082 	struct list_head	dev_list;
2083 	struct list_head	napi_list;
2084 	struct list_head	unreg_list;
2085 	struct list_head	close_list;
2086 	struct list_head	ptype_all;
2087 	struct list_head	ptype_specific;
2088 
2089 	struct {
2090 		struct list_head upper;
2091 		struct list_head lower;
2092 	} adj_list;
2093 
2094 	/* Read-mostly cache-line for fast-path access */
2095 	unsigned int		flags;
2096 	xdp_features_t		xdp_features;
2097 	unsigned long long	priv_flags;
2098 	const struct net_device_ops *netdev_ops;
2099 	const struct xdp_metadata_ops *xdp_metadata_ops;
2100 	int			ifindex;
2101 	unsigned short		gflags;
2102 	unsigned short		hard_header_len;
2103 
2104 	/* Note : dev->mtu is often read without holding a lock.
2105 	 * Writers usually hold RTNL.
2106 	 * It is recommended to use READ_ONCE() to annotate the reads,
2107 	 * and to use WRITE_ONCE() to annotate the writes.
2108 	 */
2109 	unsigned int		mtu;
2110 	unsigned short		needed_headroom;
2111 	unsigned short		needed_tailroom;
2112 
2113 	netdev_features_t	features;
2114 	netdev_features_t	hw_features;
2115 	netdev_features_t	wanted_features;
2116 	netdev_features_t	vlan_features;
2117 	netdev_features_t	hw_enc_features;
2118 	netdev_features_t	mpls_features;
2119 	netdev_features_t	gso_partial_features;
2120 
2121 	unsigned int		min_mtu;
2122 	unsigned int		max_mtu;
2123 	unsigned short		type;
2124 	unsigned char		min_header_len;
2125 	unsigned char		name_assign_type;
2126 
2127 	int			group;
2128 
2129 	struct net_device_stats	stats; /* not used by modern drivers */
2130 
2131 	struct net_device_core_stats __percpu *core_stats;
2132 
2133 	/* Stats to monitor link on/off, flapping */
2134 	atomic_t		carrier_up_count;
2135 	atomic_t		carrier_down_count;
2136 
2137 #ifdef CONFIG_WIRELESS_EXT
2138 	const struct iw_handler_def *wireless_handlers;
2139 	struct iw_public_data	*wireless_data;
2140 #endif
2141 	const struct ethtool_ops *ethtool_ops;
2142 #ifdef CONFIG_NET_L3_MASTER_DEV
2143 	const struct l3mdev_ops	*l3mdev_ops;
2144 #endif
2145 #if IS_ENABLED(CONFIG_IPV6)
2146 	const struct ndisc_ops *ndisc_ops;
2147 #endif
2148 
2149 #ifdef CONFIG_XFRM_OFFLOAD
2150 	const struct xfrmdev_ops *xfrmdev_ops;
2151 #endif
2152 
2153 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2154 	const struct tlsdev_ops *tlsdev_ops;
2155 #endif
2156 
2157 	const struct header_ops *header_ops;
2158 
2159 	unsigned char		operstate;
2160 	unsigned char		link_mode;
2161 
2162 	unsigned char		if_port;
2163 	unsigned char		dma;
2164 
2165 	/* Interface address info. */
2166 	unsigned char		perm_addr[MAX_ADDR_LEN];
2167 	unsigned char		addr_assign_type;
2168 	unsigned char		addr_len;
2169 	unsigned char		upper_level;
2170 	unsigned char		lower_level;
2171 
2172 	unsigned short		neigh_priv_len;
2173 	unsigned short          dev_id;
2174 	unsigned short          dev_port;
2175 	unsigned short		padded;
2176 
2177 	spinlock_t		addr_list_lock;
2178 	int			irq;
2179 
2180 	struct netdev_hw_addr_list	uc;
2181 	struct netdev_hw_addr_list	mc;
2182 	struct netdev_hw_addr_list	dev_addrs;
2183 
2184 #ifdef CONFIG_SYSFS
2185 	struct kset		*queues_kset;
2186 #endif
2187 #ifdef CONFIG_LOCKDEP
2188 	struct list_head	unlink_list;
2189 #endif
2190 	unsigned int		promiscuity;
2191 	unsigned int		allmulti;
2192 	bool			uc_promisc;
2193 #ifdef CONFIG_LOCKDEP
2194 	unsigned char		nested_level;
2195 #endif
2196 
2197 
2198 	/* Protocol-specific pointers */
2199 
2200 	struct in_device __rcu	*ip_ptr;
2201 	struct inet6_dev __rcu	*ip6_ptr;
2202 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2203 	struct vlan_info __rcu	*vlan_info;
2204 #endif
2205 #if IS_ENABLED(CONFIG_NET_DSA)
2206 	struct dsa_port		*dsa_ptr;
2207 #endif
2208 #if IS_ENABLED(CONFIG_TIPC)
2209 	struct tipc_bearer __rcu *tipc_ptr;
2210 #endif
2211 #if IS_ENABLED(CONFIG_ATALK)
2212 	void 			*atalk_ptr;
2213 #endif
2214 #if IS_ENABLED(CONFIG_AX25)
2215 	void			*ax25_ptr;
2216 #endif
2217 #if IS_ENABLED(CONFIG_CFG80211)
2218 	struct wireless_dev	*ieee80211_ptr;
2219 #endif
2220 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2221 	struct wpan_dev		*ieee802154_ptr;
2222 #endif
2223 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2224 	struct mpls_dev __rcu	*mpls_ptr;
2225 #endif
2226 #if IS_ENABLED(CONFIG_MCTP)
2227 	struct mctp_dev __rcu	*mctp_ptr;
2228 #endif
2229 
2230 /*
2231  * Cache lines mostly used on receive path (including eth_type_trans())
2232  */
2233 	/* Interface address info used in eth_type_trans() */
2234 	const unsigned char	*dev_addr;
2235 
2236 	struct netdev_rx_queue	*_rx;
2237 	unsigned int		num_rx_queues;
2238 	unsigned int		real_num_rx_queues;
2239 
2240 	struct bpf_prog __rcu	*xdp_prog;
2241 	unsigned long		gro_flush_timeout;
2242 	int			napi_defer_hard_irqs;
2243 #define GRO_LEGACY_MAX_SIZE	65536u
2244 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2245  * and shinfo->gso_segs is a 16bit field.
2246  */
2247 #define GRO_MAX_SIZE		(8 * 65535u)
2248 	unsigned int		gro_max_size;
2249 	unsigned int		gro_ipv4_max_size;
2250 	unsigned int		xdp_zc_max_segs;
2251 	rx_handler_func_t __rcu	*rx_handler;
2252 	void __rcu		*rx_handler_data;
2253 #ifdef CONFIG_NET_XGRESS
2254 	struct bpf_mprog_entry __rcu *tcx_ingress;
2255 #endif
2256 	struct netdev_queue __rcu *ingress_queue;
2257 #ifdef CONFIG_NETFILTER_INGRESS
2258 	struct nf_hook_entries __rcu *nf_hooks_ingress;
2259 #endif
2260 
2261 	unsigned char		broadcast[MAX_ADDR_LEN];
2262 #ifdef CONFIG_RFS_ACCEL
2263 	struct cpu_rmap		*rx_cpu_rmap;
2264 #endif
2265 	struct hlist_node	index_hlist;
2266 
2267 /*
2268  * Cache lines mostly used on transmit path
2269  */
2270 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
2271 	unsigned int		num_tx_queues;
2272 	unsigned int		real_num_tx_queues;
2273 	struct Qdisc __rcu	*qdisc;
2274 	unsigned int		tx_queue_len;
2275 	spinlock_t		tx_global_lock;
2276 
2277 	struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2278 
2279 #ifdef CONFIG_XPS
2280 	struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2281 #endif
2282 #ifdef CONFIG_NET_XGRESS
2283 	struct bpf_mprog_entry __rcu *tcx_egress;
2284 #endif
2285 #ifdef CONFIG_NETFILTER_EGRESS
2286 	struct nf_hook_entries __rcu *nf_hooks_egress;
2287 #endif
2288 
2289 #ifdef CONFIG_NET_SCHED
2290 	DECLARE_HASHTABLE	(qdisc_hash, 4);
2291 #endif
2292 	/* These may be needed for future network-power-down code. */
2293 	struct timer_list	watchdog_timer;
2294 	int			watchdog_timeo;
2295 
2296 	u32                     proto_down_reason;
2297 
2298 	struct list_head	todo_list;
2299 
2300 #ifdef CONFIG_PCPU_DEV_REFCNT
2301 	int __percpu		*pcpu_refcnt;
2302 #else
2303 	refcount_t		dev_refcnt;
2304 #endif
2305 	struct ref_tracker_dir	refcnt_tracker;
2306 
2307 	struct list_head	link_watch_list;
2308 
2309 	enum { NETREG_UNINITIALIZED=0,
2310 	       NETREG_REGISTERED,	/* completed register_netdevice */
2311 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
2312 	       NETREG_UNREGISTERED,	/* completed unregister todo */
2313 	       NETREG_RELEASED,		/* called free_netdev */
2314 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
2315 	} reg_state:8;
2316 
2317 	bool dismantle;
2318 
2319 	enum {
2320 		RTNL_LINK_INITIALIZED,
2321 		RTNL_LINK_INITIALIZING,
2322 	} rtnl_link_state:16;
2323 
2324 	bool needs_free_netdev;
2325 	void (*priv_destructor)(struct net_device *dev);
2326 
2327 #ifdef CONFIG_NETPOLL
2328 	struct netpoll_info __rcu	*npinfo;
2329 #endif
2330 
2331 	possible_net_t			nd_net;
2332 
2333 	/* mid-layer private */
2334 	void				*ml_priv;
2335 	enum netdev_ml_priv_type	ml_priv_type;
2336 
2337 	union {
2338 		struct pcpu_lstats __percpu		*lstats;
2339 		struct pcpu_sw_netstats __percpu	*tstats;
2340 		struct pcpu_dstats __percpu		*dstats;
2341 	};
2342 
2343 #if IS_ENABLED(CONFIG_GARP)
2344 	struct garp_port __rcu	*garp_port;
2345 #endif
2346 #if IS_ENABLED(CONFIG_MRP)
2347 	struct mrp_port __rcu	*mrp_port;
2348 #endif
2349 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2350 	struct dm_hw_stat_delta __rcu *dm_private;
2351 #endif
2352 	struct device		dev;
2353 	const struct attribute_group *sysfs_groups[4];
2354 	const struct attribute_group *sysfs_rx_queue_group;
2355 
2356 	const struct rtnl_link_ops *rtnl_link_ops;
2357 
2358 	/* for setting kernel sock attribute on TCP connection setup */
2359 #define GSO_MAX_SEGS		65535u
2360 #define GSO_LEGACY_MAX_SIZE	65536u
2361 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2362  * and shinfo->gso_segs is a 16bit field.
2363  */
2364 #define GSO_MAX_SIZE		(8 * GSO_MAX_SEGS)
2365 
2366 	unsigned int		gso_max_size;
2367 #define TSO_LEGACY_MAX_SIZE	65536
2368 #define TSO_MAX_SIZE		UINT_MAX
2369 	unsigned int		tso_max_size;
2370 	u16			gso_max_segs;
2371 #define TSO_MAX_SEGS		U16_MAX
2372 	u16			tso_max_segs;
2373 	unsigned int		gso_ipv4_max_size;
2374 
2375 #ifdef CONFIG_DCB
2376 	const struct dcbnl_rtnl_ops *dcbnl_ops;
2377 #endif
2378 	s16			num_tc;
2379 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
2380 	u8			prio_tc_map[TC_BITMASK + 1];
2381 
2382 #if IS_ENABLED(CONFIG_FCOE)
2383 	unsigned int		fcoe_ddp_xid;
2384 #endif
2385 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2386 	struct netprio_map __rcu *priomap;
2387 #endif
2388 	struct phy_device	*phydev;
2389 	struct sfp_bus		*sfp_bus;
2390 	struct lock_class_key	*qdisc_tx_busylock;
2391 	bool			proto_down;
2392 	unsigned		wol_enabled:1;
2393 	unsigned		threaded:1;
2394 
2395 	struct list_head	net_notifier_list;
2396 
2397 #if IS_ENABLED(CONFIG_MACSEC)
2398 	/* MACsec management functions */
2399 	const struct macsec_ops *macsec_ops;
2400 #endif
2401 	const struct udp_tunnel_nic_info	*udp_tunnel_nic_info;
2402 	struct udp_tunnel_nic	*udp_tunnel_nic;
2403 
2404 	/* protected by rtnl_lock */
2405 	struct bpf_xdp_entity	xdp_state[__MAX_XDP_MODE];
2406 
2407 	u8 dev_addr_shadow[MAX_ADDR_LEN];
2408 	netdevice_tracker	linkwatch_dev_tracker;
2409 	netdevice_tracker	watchdog_dev_tracker;
2410 	netdevice_tracker	dev_registered_tracker;
2411 	struct rtnl_hw_stats64	*offload_xstats_l3;
2412 
2413 	struct devlink_port	*devlink_port;
2414 
2415 #if IS_ENABLED(CONFIG_DPLL)
2416 	struct dpll_pin		*dpll_pin;
2417 #endif
2418 };
2419 #define to_net_dev(d) container_of(d, struct net_device, dev)
2420 
2421 /*
2422  * Driver should use this to assign devlink port instance to a netdevice
2423  * before it registers the netdevice. Therefore devlink_port is static
2424  * during the netdev lifetime after it is registered.
2425  */
2426 #define SET_NETDEV_DEVLINK_PORT(dev, port)			\
2427 ({								\
2428 	WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED);	\
2429 	((dev)->devlink_port = (port));				\
2430 })
2431 
2432 static inline bool netif_elide_gro(const struct net_device *dev)
2433 {
2434 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2435 		return true;
2436 	return false;
2437 }
2438 
2439 #define	NETDEV_ALIGN		32
2440 
2441 static inline
2442 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2443 {
2444 	return dev->prio_tc_map[prio & TC_BITMASK];
2445 }
2446 
2447 static inline
2448 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2449 {
2450 	if (tc >= dev->num_tc)
2451 		return -EINVAL;
2452 
2453 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2454 	return 0;
2455 }
2456 
2457 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2458 void netdev_reset_tc(struct net_device *dev);
2459 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2460 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2461 
2462 static inline
2463 int netdev_get_num_tc(struct net_device *dev)
2464 {
2465 	return dev->num_tc;
2466 }
2467 
2468 static inline void net_prefetch(void *p)
2469 {
2470 	prefetch(p);
2471 #if L1_CACHE_BYTES < 128
2472 	prefetch((u8 *)p + L1_CACHE_BYTES);
2473 #endif
2474 }
2475 
2476 static inline void net_prefetchw(void *p)
2477 {
2478 	prefetchw(p);
2479 #if L1_CACHE_BYTES < 128
2480 	prefetchw((u8 *)p + L1_CACHE_BYTES);
2481 #endif
2482 }
2483 
2484 void netdev_unbind_sb_channel(struct net_device *dev,
2485 			      struct net_device *sb_dev);
2486 int netdev_bind_sb_channel_queue(struct net_device *dev,
2487 				 struct net_device *sb_dev,
2488 				 u8 tc, u16 count, u16 offset);
2489 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2490 static inline int netdev_get_sb_channel(struct net_device *dev)
2491 {
2492 	return max_t(int, -dev->num_tc, 0);
2493 }
2494 
2495 static inline
2496 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2497 					 unsigned int index)
2498 {
2499 	DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2500 	return &dev->_tx[index];
2501 }
2502 
2503 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2504 						    const struct sk_buff *skb)
2505 {
2506 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2507 }
2508 
2509 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2510 					    void (*f)(struct net_device *,
2511 						      struct netdev_queue *,
2512 						      void *),
2513 					    void *arg)
2514 {
2515 	unsigned int i;
2516 
2517 	for (i = 0; i < dev->num_tx_queues; i++)
2518 		f(dev, &dev->_tx[i], arg);
2519 }
2520 
2521 #define netdev_lockdep_set_classes(dev)				\
2522 {								\
2523 	static struct lock_class_key qdisc_tx_busylock_key;	\
2524 	static struct lock_class_key qdisc_xmit_lock_key;	\
2525 	static struct lock_class_key dev_addr_list_lock_key;	\
2526 	unsigned int i;						\
2527 								\
2528 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
2529 	lockdep_set_class(&(dev)->addr_list_lock,		\
2530 			  &dev_addr_list_lock_key);		\
2531 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
2532 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
2533 				  &qdisc_xmit_lock_key);	\
2534 }
2535 
2536 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2537 		     struct net_device *sb_dev);
2538 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2539 					 struct sk_buff *skb,
2540 					 struct net_device *sb_dev);
2541 
2542 /* returns the headroom that the master device needs to take in account
2543  * when forwarding to this dev
2544  */
2545 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2546 {
2547 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2548 }
2549 
2550 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2551 {
2552 	if (dev->netdev_ops->ndo_set_rx_headroom)
2553 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2554 }
2555 
2556 /* set the device rx headroom to the dev's default */
2557 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2558 {
2559 	netdev_set_rx_headroom(dev, -1);
2560 }
2561 
2562 static inline void *netdev_get_ml_priv(struct net_device *dev,
2563 				       enum netdev_ml_priv_type type)
2564 {
2565 	if (dev->ml_priv_type != type)
2566 		return NULL;
2567 
2568 	return dev->ml_priv;
2569 }
2570 
2571 static inline void netdev_set_ml_priv(struct net_device *dev,
2572 				      void *ml_priv,
2573 				      enum netdev_ml_priv_type type)
2574 {
2575 	WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2576 	     "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2577 	     dev->ml_priv_type, type);
2578 	WARN(!dev->ml_priv_type && dev->ml_priv,
2579 	     "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2580 
2581 	dev->ml_priv = ml_priv;
2582 	dev->ml_priv_type = type;
2583 }
2584 
2585 /*
2586  * Net namespace inlines
2587  */
2588 static inline
2589 struct net *dev_net(const struct net_device *dev)
2590 {
2591 	return read_pnet(&dev->nd_net);
2592 }
2593 
2594 static inline
2595 void dev_net_set(struct net_device *dev, struct net *net)
2596 {
2597 	write_pnet(&dev->nd_net, net);
2598 }
2599 
2600 /**
2601  *	netdev_priv - access network device private data
2602  *	@dev: network device
2603  *
2604  * Get network device private data
2605  */
2606 static inline void *netdev_priv(const struct net_device *dev)
2607 {
2608 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2609 }
2610 
2611 /* Set the sysfs physical device reference for the network logical device
2612  * if set prior to registration will cause a symlink during initialization.
2613  */
2614 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2615 
2616 /* Set the sysfs device type for the network logical device to allow
2617  * fine-grained identification of different network device types. For
2618  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2619  */
2620 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2621 
2622 /* Default NAPI poll() weight
2623  * Device drivers are strongly advised to not use bigger value
2624  */
2625 #define NAPI_POLL_WEIGHT 64
2626 
2627 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2628 			   int (*poll)(struct napi_struct *, int), int weight);
2629 
2630 /**
2631  * netif_napi_add() - initialize a NAPI context
2632  * @dev:  network device
2633  * @napi: NAPI context
2634  * @poll: polling function
2635  *
2636  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2637  * *any* of the other NAPI-related functions.
2638  */
2639 static inline void
2640 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2641 	       int (*poll)(struct napi_struct *, int))
2642 {
2643 	netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2644 }
2645 
2646 static inline void
2647 netif_napi_add_tx_weight(struct net_device *dev,
2648 			 struct napi_struct *napi,
2649 			 int (*poll)(struct napi_struct *, int),
2650 			 int weight)
2651 {
2652 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2653 	netif_napi_add_weight(dev, napi, poll, weight);
2654 }
2655 
2656 /**
2657  * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2658  * @dev:  network device
2659  * @napi: NAPI context
2660  * @poll: polling function
2661  *
2662  * This variant of netif_napi_add() should be used from drivers using NAPI
2663  * to exclusively poll a TX queue.
2664  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2665  */
2666 static inline void netif_napi_add_tx(struct net_device *dev,
2667 				     struct napi_struct *napi,
2668 				     int (*poll)(struct napi_struct *, int))
2669 {
2670 	netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2671 }
2672 
2673 /**
2674  *  __netif_napi_del - remove a NAPI context
2675  *  @napi: NAPI context
2676  *
2677  * Warning: caller must observe RCU grace period before freeing memory
2678  * containing @napi. Drivers might want to call this helper to combine
2679  * all the needed RCU grace periods into a single one.
2680  */
2681 void __netif_napi_del(struct napi_struct *napi);
2682 
2683 /**
2684  *  netif_napi_del - remove a NAPI context
2685  *  @napi: NAPI context
2686  *
2687  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2688  */
2689 static inline void netif_napi_del(struct napi_struct *napi)
2690 {
2691 	__netif_napi_del(napi);
2692 	synchronize_net();
2693 }
2694 
2695 struct packet_type {
2696 	__be16			type;	/* This is really htons(ether_type). */
2697 	bool			ignore_outgoing;
2698 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2699 	netdevice_tracker	dev_tracker;
2700 	int			(*func) (struct sk_buff *,
2701 					 struct net_device *,
2702 					 struct packet_type *,
2703 					 struct net_device *);
2704 	void			(*list_func) (struct list_head *,
2705 					      struct packet_type *,
2706 					      struct net_device *);
2707 	bool			(*id_match)(struct packet_type *ptype,
2708 					    struct sock *sk);
2709 	struct net		*af_packet_net;
2710 	void			*af_packet_priv;
2711 	struct list_head	list;
2712 };
2713 
2714 struct offload_callbacks {
2715 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2716 						netdev_features_t features);
2717 	struct sk_buff		*(*gro_receive)(struct list_head *head,
2718 						struct sk_buff *skb);
2719 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2720 };
2721 
2722 struct packet_offload {
2723 	__be16			 type;	/* This is really htons(ether_type). */
2724 	u16			 priority;
2725 	struct offload_callbacks callbacks;
2726 	struct list_head	 list;
2727 };
2728 
2729 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2730 struct pcpu_sw_netstats {
2731 	u64_stats_t		rx_packets;
2732 	u64_stats_t		rx_bytes;
2733 	u64_stats_t		tx_packets;
2734 	u64_stats_t		tx_bytes;
2735 	struct u64_stats_sync   syncp;
2736 } __aligned(4 * sizeof(u64));
2737 
2738 struct pcpu_lstats {
2739 	u64_stats_t packets;
2740 	u64_stats_t bytes;
2741 	struct u64_stats_sync syncp;
2742 } __aligned(2 * sizeof(u64));
2743 
2744 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2745 
2746 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2747 {
2748 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2749 
2750 	u64_stats_update_begin(&tstats->syncp);
2751 	u64_stats_add(&tstats->rx_bytes, len);
2752 	u64_stats_inc(&tstats->rx_packets);
2753 	u64_stats_update_end(&tstats->syncp);
2754 }
2755 
2756 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2757 					  unsigned int packets,
2758 					  unsigned int len)
2759 {
2760 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2761 
2762 	u64_stats_update_begin(&tstats->syncp);
2763 	u64_stats_add(&tstats->tx_bytes, len);
2764 	u64_stats_add(&tstats->tx_packets, packets);
2765 	u64_stats_update_end(&tstats->syncp);
2766 }
2767 
2768 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2769 {
2770 	struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2771 
2772 	u64_stats_update_begin(&lstats->syncp);
2773 	u64_stats_add(&lstats->bytes, len);
2774 	u64_stats_inc(&lstats->packets);
2775 	u64_stats_update_end(&lstats->syncp);
2776 }
2777 
2778 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2779 ({									\
2780 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2781 	if (pcpu_stats)	{						\
2782 		int __cpu;						\
2783 		for_each_possible_cpu(__cpu) {				\
2784 			typeof(type) *stat;				\
2785 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2786 			u64_stats_init(&stat->syncp);			\
2787 		}							\
2788 	}								\
2789 	pcpu_stats;							\
2790 })
2791 
2792 #define netdev_alloc_pcpu_stats(type)					\
2793 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2794 
2795 #define devm_netdev_alloc_pcpu_stats(dev, type)				\
2796 ({									\
2797 	typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2798 	if (pcpu_stats) {						\
2799 		int __cpu;						\
2800 		for_each_possible_cpu(__cpu) {				\
2801 			typeof(type) *stat;				\
2802 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2803 			u64_stats_init(&stat->syncp);			\
2804 		}							\
2805 	}								\
2806 	pcpu_stats;							\
2807 })
2808 
2809 enum netdev_lag_tx_type {
2810 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2811 	NETDEV_LAG_TX_TYPE_RANDOM,
2812 	NETDEV_LAG_TX_TYPE_BROADCAST,
2813 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2814 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2815 	NETDEV_LAG_TX_TYPE_HASH,
2816 };
2817 
2818 enum netdev_lag_hash {
2819 	NETDEV_LAG_HASH_NONE,
2820 	NETDEV_LAG_HASH_L2,
2821 	NETDEV_LAG_HASH_L34,
2822 	NETDEV_LAG_HASH_L23,
2823 	NETDEV_LAG_HASH_E23,
2824 	NETDEV_LAG_HASH_E34,
2825 	NETDEV_LAG_HASH_VLAN_SRCMAC,
2826 	NETDEV_LAG_HASH_UNKNOWN,
2827 };
2828 
2829 struct netdev_lag_upper_info {
2830 	enum netdev_lag_tx_type tx_type;
2831 	enum netdev_lag_hash hash_type;
2832 };
2833 
2834 struct netdev_lag_lower_state_info {
2835 	u8 link_up : 1,
2836 	   tx_enabled : 1;
2837 };
2838 
2839 #include <linux/notifier.h>
2840 
2841 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2842  * and the rtnetlink notification exclusion list in rtnetlink_event() when
2843  * adding new types.
2844  */
2845 enum netdev_cmd {
2846 	NETDEV_UP	= 1,	/* For now you can't veto a device up/down */
2847 	NETDEV_DOWN,
2848 	NETDEV_REBOOT,		/* Tell a protocol stack a network interface
2849 				   detected a hardware crash and restarted
2850 				   - we can use this eg to kick tcp sessions
2851 				   once done */
2852 	NETDEV_CHANGE,		/* Notify device state change */
2853 	NETDEV_REGISTER,
2854 	NETDEV_UNREGISTER,
2855 	NETDEV_CHANGEMTU,	/* notify after mtu change happened */
2856 	NETDEV_CHANGEADDR,	/* notify after the address change */
2857 	NETDEV_PRE_CHANGEADDR,	/* notify before the address change */
2858 	NETDEV_GOING_DOWN,
2859 	NETDEV_CHANGENAME,
2860 	NETDEV_FEAT_CHANGE,
2861 	NETDEV_BONDING_FAILOVER,
2862 	NETDEV_PRE_UP,
2863 	NETDEV_PRE_TYPE_CHANGE,
2864 	NETDEV_POST_TYPE_CHANGE,
2865 	NETDEV_POST_INIT,
2866 	NETDEV_PRE_UNINIT,
2867 	NETDEV_RELEASE,
2868 	NETDEV_NOTIFY_PEERS,
2869 	NETDEV_JOIN,
2870 	NETDEV_CHANGEUPPER,
2871 	NETDEV_RESEND_IGMP,
2872 	NETDEV_PRECHANGEMTU,	/* notify before mtu change happened */
2873 	NETDEV_CHANGEINFODATA,
2874 	NETDEV_BONDING_INFO,
2875 	NETDEV_PRECHANGEUPPER,
2876 	NETDEV_CHANGELOWERSTATE,
2877 	NETDEV_UDP_TUNNEL_PUSH_INFO,
2878 	NETDEV_UDP_TUNNEL_DROP_INFO,
2879 	NETDEV_CHANGE_TX_QUEUE_LEN,
2880 	NETDEV_CVLAN_FILTER_PUSH_INFO,
2881 	NETDEV_CVLAN_FILTER_DROP_INFO,
2882 	NETDEV_SVLAN_FILTER_PUSH_INFO,
2883 	NETDEV_SVLAN_FILTER_DROP_INFO,
2884 	NETDEV_OFFLOAD_XSTATS_ENABLE,
2885 	NETDEV_OFFLOAD_XSTATS_DISABLE,
2886 	NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2887 	NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2888 	NETDEV_XDP_FEAT_CHANGE,
2889 };
2890 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2891 
2892 int register_netdevice_notifier(struct notifier_block *nb);
2893 int unregister_netdevice_notifier(struct notifier_block *nb);
2894 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2895 int unregister_netdevice_notifier_net(struct net *net,
2896 				      struct notifier_block *nb);
2897 int register_netdevice_notifier_dev_net(struct net_device *dev,
2898 					struct notifier_block *nb,
2899 					struct netdev_net_notifier *nn);
2900 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2901 					  struct notifier_block *nb,
2902 					  struct netdev_net_notifier *nn);
2903 
2904 struct netdev_notifier_info {
2905 	struct net_device	*dev;
2906 	struct netlink_ext_ack	*extack;
2907 };
2908 
2909 struct netdev_notifier_info_ext {
2910 	struct netdev_notifier_info info; /* must be first */
2911 	union {
2912 		u32 mtu;
2913 	} ext;
2914 };
2915 
2916 struct netdev_notifier_change_info {
2917 	struct netdev_notifier_info info; /* must be first */
2918 	unsigned int flags_changed;
2919 };
2920 
2921 struct netdev_notifier_changeupper_info {
2922 	struct netdev_notifier_info info; /* must be first */
2923 	struct net_device *upper_dev; /* new upper dev */
2924 	bool master; /* is upper dev master */
2925 	bool linking; /* is the notification for link or unlink */
2926 	void *upper_info; /* upper dev info */
2927 };
2928 
2929 struct netdev_notifier_changelowerstate_info {
2930 	struct netdev_notifier_info info; /* must be first */
2931 	void *lower_state_info; /* is lower dev state */
2932 };
2933 
2934 struct netdev_notifier_pre_changeaddr_info {
2935 	struct netdev_notifier_info info; /* must be first */
2936 	const unsigned char *dev_addr;
2937 };
2938 
2939 enum netdev_offload_xstats_type {
2940 	NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2941 };
2942 
2943 struct netdev_notifier_offload_xstats_info {
2944 	struct netdev_notifier_info info; /* must be first */
2945 	enum netdev_offload_xstats_type type;
2946 
2947 	union {
2948 		/* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2949 		struct netdev_notifier_offload_xstats_rd *report_delta;
2950 		/* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2951 		struct netdev_notifier_offload_xstats_ru *report_used;
2952 	};
2953 };
2954 
2955 int netdev_offload_xstats_enable(struct net_device *dev,
2956 				 enum netdev_offload_xstats_type type,
2957 				 struct netlink_ext_ack *extack);
2958 int netdev_offload_xstats_disable(struct net_device *dev,
2959 				  enum netdev_offload_xstats_type type);
2960 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2961 				   enum netdev_offload_xstats_type type);
2962 int netdev_offload_xstats_get(struct net_device *dev,
2963 			      enum netdev_offload_xstats_type type,
2964 			      struct rtnl_hw_stats64 *stats, bool *used,
2965 			      struct netlink_ext_ack *extack);
2966 void
2967 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2968 				   const struct rtnl_hw_stats64 *stats);
2969 void
2970 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2971 void netdev_offload_xstats_push_delta(struct net_device *dev,
2972 				      enum netdev_offload_xstats_type type,
2973 				      const struct rtnl_hw_stats64 *stats);
2974 
2975 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2976 					     struct net_device *dev)
2977 {
2978 	info->dev = dev;
2979 	info->extack = NULL;
2980 }
2981 
2982 static inline struct net_device *
2983 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2984 {
2985 	return info->dev;
2986 }
2987 
2988 static inline struct netlink_ext_ack *
2989 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2990 {
2991 	return info->extack;
2992 }
2993 
2994 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2995 int call_netdevice_notifiers_info(unsigned long val,
2996 				  struct netdev_notifier_info *info);
2997 
2998 extern rwlock_t				dev_base_lock;		/* Device list lock */
2999 
3000 #define for_each_netdev(net, d)		\
3001 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3002 #define for_each_netdev_reverse(net, d)	\
3003 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3004 #define for_each_netdev_rcu(net, d)		\
3005 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3006 #define for_each_netdev_safe(net, d, n)	\
3007 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3008 #define for_each_netdev_continue(net, d)		\
3009 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3010 #define for_each_netdev_continue_reverse(net, d)		\
3011 		list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3012 						     dev_list)
3013 #define for_each_netdev_continue_rcu(net, d)		\
3014 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3015 #define for_each_netdev_in_bond_rcu(bond, slave)	\
3016 		for_each_netdev_rcu(&init_net, slave)	\
3017 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3018 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
3019 
3020 #define for_each_netdev_dump(net, d, ifindex)				\
3021 	xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3022 
3023 static inline struct net_device *next_net_device(struct net_device *dev)
3024 {
3025 	struct list_head *lh;
3026 	struct net *net;
3027 
3028 	net = dev_net(dev);
3029 	lh = dev->dev_list.next;
3030 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3031 }
3032 
3033 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3034 {
3035 	struct list_head *lh;
3036 	struct net *net;
3037 
3038 	net = dev_net(dev);
3039 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3040 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3041 }
3042 
3043 static inline struct net_device *first_net_device(struct net *net)
3044 {
3045 	return list_empty(&net->dev_base_head) ? NULL :
3046 		net_device_entry(net->dev_base_head.next);
3047 }
3048 
3049 static inline struct net_device *first_net_device_rcu(struct net *net)
3050 {
3051 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3052 
3053 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3054 }
3055 
3056 int netdev_boot_setup_check(struct net_device *dev);
3057 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3058 				       const char *hwaddr);
3059 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3060 void dev_add_pack(struct packet_type *pt);
3061 void dev_remove_pack(struct packet_type *pt);
3062 void __dev_remove_pack(struct packet_type *pt);
3063 void dev_add_offload(struct packet_offload *po);
3064 void dev_remove_offload(struct packet_offload *po);
3065 
3066 int dev_get_iflink(const struct net_device *dev);
3067 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3068 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3069 			  struct net_device_path_stack *stack);
3070 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3071 				      unsigned short mask);
3072 struct net_device *dev_get_by_name(struct net *net, const char *name);
3073 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3074 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3075 bool netdev_name_in_use(struct net *net, const char *name);
3076 int dev_alloc_name(struct net_device *dev, const char *name);
3077 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3078 void dev_close(struct net_device *dev);
3079 void dev_close_many(struct list_head *head, bool unlink);
3080 void dev_disable_lro(struct net_device *dev);
3081 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3082 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3083 		     struct net_device *sb_dev);
3084 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3085 		       struct net_device *sb_dev);
3086 
3087 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3088 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3089 
3090 static inline int dev_queue_xmit(struct sk_buff *skb)
3091 {
3092 	return __dev_queue_xmit(skb, NULL);
3093 }
3094 
3095 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3096 				       struct net_device *sb_dev)
3097 {
3098 	return __dev_queue_xmit(skb, sb_dev);
3099 }
3100 
3101 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3102 {
3103 	int ret;
3104 
3105 	ret = __dev_direct_xmit(skb, queue_id);
3106 	if (!dev_xmit_complete(ret))
3107 		kfree_skb(skb);
3108 	return ret;
3109 }
3110 
3111 int register_netdevice(struct net_device *dev);
3112 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3113 void unregister_netdevice_many(struct list_head *head);
3114 static inline void unregister_netdevice(struct net_device *dev)
3115 {
3116 	unregister_netdevice_queue(dev, NULL);
3117 }
3118 
3119 int netdev_refcnt_read(const struct net_device *dev);
3120 void free_netdev(struct net_device *dev);
3121 void netdev_freemem(struct net_device *dev);
3122 int init_dummy_netdev(struct net_device *dev);
3123 
3124 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3125 					 struct sk_buff *skb,
3126 					 bool all_slaves);
3127 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3128 					    struct sock *sk);
3129 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3130 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3131 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3132 				       netdevice_tracker *tracker, gfp_t gfp);
3133 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3134 				      netdevice_tracker *tracker, gfp_t gfp);
3135 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3136 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3137 
3138 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3139 				  unsigned short type,
3140 				  const void *daddr, const void *saddr,
3141 				  unsigned int len)
3142 {
3143 	if (!dev->header_ops || !dev->header_ops->create)
3144 		return 0;
3145 
3146 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3147 }
3148 
3149 static inline int dev_parse_header(const struct sk_buff *skb,
3150 				   unsigned char *haddr)
3151 {
3152 	const struct net_device *dev = skb->dev;
3153 
3154 	if (!dev->header_ops || !dev->header_ops->parse)
3155 		return 0;
3156 	return dev->header_ops->parse(skb, haddr);
3157 }
3158 
3159 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3160 {
3161 	const struct net_device *dev = skb->dev;
3162 
3163 	if (!dev->header_ops || !dev->header_ops->parse_protocol)
3164 		return 0;
3165 	return dev->header_ops->parse_protocol(skb);
3166 }
3167 
3168 /* ll_header must have at least hard_header_len allocated */
3169 static inline bool dev_validate_header(const struct net_device *dev,
3170 				       char *ll_header, int len)
3171 {
3172 	if (likely(len >= dev->hard_header_len))
3173 		return true;
3174 	if (len < dev->min_header_len)
3175 		return false;
3176 
3177 	if (capable(CAP_SYS_RAWIO)) {
3178 		memset(ll_header + len, 0, dev->hard_header_len - len);
3179 		return true;
3180 	}
3181 
3182 	if (dev->header_ops && dev->header_ops->validate)
3183 		return dev->header_ops->validate(ll_header, len);
3184 
3185 	return false;
3186 }
3187 
3188 static inline bool dev_has_header(const struct net_device *dev)
3189 {
3190 	return dev->header_ops && dev->header_ops->create;
3191 }
3192 
3193 /*
3194  * Incoming packets are placed on per-CPU queues
3195  */
3196 struct softnet_data {
3197 	struct list_head	poll_list;
3198 	struct sk_buff_head	process_queue;
3199 
3200 	/* stats */
3201 	unsigned int		processed;
3202 	unsigned int		time_squeeze;
3203 #ifdef CONFIG_RPS
3204 	struct softnet_data	*rps_ipi_list;
3205 #endif
3206 
3207 	bool			in_net_rx_action;
3208 	bool			in_napi_threaded_poll;
3209 
3210 #ifdef CONFIG_NET_FLOW_LIMIT
3211 	struct sd_flow_limit __rcu *flow_limit;
3212 #endif
3213 	struct Qdisc		*output_queue;
3214 	struct Qdisc		**output_queue_tailp;
3215 	struct sk_buff		*completion_queue;
3216 #ifdef CONFIG_XFRM_OFFLOAD
3217 	struct sk_buff_head	xfrm_backlog;
3218 #endif
3219 	/* written and read only by owning cpu: */
3220 	struct {
3221 		u16 recursion;
3222 		u8  more;
3223 #ifdef CONFIG_NET_EGRESS
3224 		u8  skip_txqueue;
3225 #endif
3226 	} xmit;
3227 #ifdef CONFIG_RPS
3228 	/* input_queue_head should be written by cpu owning this struct,
3229 	 * and only read by other cpus. Worth using a cache line.
3230 	 */
3231 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
3232 
3233 	/* Elements below can be accessed between CPUs for RPS/RFS */
3234 	call_single_data_t	csd ____cacheline_aligned_in_smp;
3235 	struct softnet_data	*rps_ipi_next;
3236 	unsigned int		cpu;
3237 	unsigned int		input_queue_tail;
3238 #endif
3239 	unsigned int		received_rps;
3240 	unsigned int		dropped;
3241 	struct sk_buff_head	input_pkt_queue;
3242 	struct napi_struct	backlog;
3243 
3244 	/* Another possibly contended cache line */
3245 	spinlock_t		defer_lock ____cacheline_aligned_in_smp;
3246 	int			defer_count;
3247 	int			defer_ipi_scheduled;
3248 	struct sk_buff		*defer_list;
3249 	call_single_data_t	defer_csd;
3250 };
3251 
3252 static inline void input_queue_head_incr(struct softnet_data *sd)
3253 {
3254 #ifdef CONFIG_RPS
3255 	sd->input_queue_head++;
3256 #endif
3257 }
3258 
3259 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3260 					      unsigned int *qtail)
3261 {
3262 #ifdef CONFIG_RPS
3263 	*qtail = ++sd->input_queue_tail;
3264 #endif
3265 }
3266 
3267 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3268 
3269 static inline int dev_recursion_level(void)
3270 {
3271 	return this_cpu_read(softnet_data.xmit.recursion);
3272 }
3273 
3274 #define XMIT_RECURSION_LIMIT	8
3275 static inline bool dev_xmit_recursion(void)
3276 {
3277 	return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3278 			XMIT_RECURSION_LIMIT);
3279 }
3280 
3281 static inline void dev_xmit_recursion_inc(void)
3282 {
3283 	__this_cpu_inc(softnet_data.xmit.recursion);
3284 }
3285 
3286 static inline void dev_xmit_recursion_dec(void)
3287 {
3288 	__this_cpu_dec(softnet_data.xmit.recursion);
3289 }
3290 
3291 void __netif_schedule(struct Qdisc *q);
3292 void netif_schedule_queue(struct netdev_queue *txq);
3293 
3294 static inline void netif_tx_schedule_all(struct net_device *dev)
3295 {
3296 	unsigned int i;
3297 
3298 	for (i = 0; i < dev->num_tx_queues; i++)
3299 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
3300 }
3301 
3302 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3303 {
3304 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3305 }
3306 
3307 /**
3308  *	netif_start_queue - allow transmit
3309  *	@dev: network device
3310  *
3311  *	Allow upper layers to call the device hard_start_xmit routine.
3312  */
3313 static inline void netif_start_queue(struct net_device *dev)
3314 {
3315 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3316 }
3317 
3318 static inline void netif_tx_start_all_queues(struct net_device *dev)
3319 {
3320 	unsigned int i;
3321 
3322 	for (i = 0; i < dev->num_tx_queues; i++) {
3323 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3324 		netif_tx_start_queue(txq);
3325 	}
3326 }
3327 
3328 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3329 
3330 /**
3331  *	netif_wake_queue - restart transmit
3332  *	@dev: network device
3333  *
3334  *	Allow upper layers to call the device hard_start_xmit routine.
3335  *	Used for flow control when transmit resources are available.
3336  */
3337 static inline void netif_wake_queue(struct net_device *dev)
3338 {
3339 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3340 }
3341 
3342 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3343 {
3344 	unsigned int i;
3345 
3346 	for (i = 0; i < dev->num_tx_queues; i++) {
3347 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3348 		netif_tx_wake_queue(txq);
3349 	}
3350 }
3351 
3352 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3353 {
3354 	/* Must be an atomic op see netif_txq_try_stop() */
3355 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3356 }
3357 
3358 /**
3359  *	netif_stop_queue - stop transmitted packets
3360  *	@dev: network device
3361  *
3362  *	Stop upper layers calling the device hard_start_xmit routine.
3363  *	Used for flow control when transmit resources are unavailable.
3364  */
3365 static inline void netif_stop_queue(struct net_device *dev)
3366 {
3367 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3368 }
3369 
3370 void netif_tx_stop_all_queues(struct net_device *dev);
3371 
3372 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3373 {
3374 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3375 }
3376 
3377 /**
3378  *	netif_queue_stopped - test if transmit queue is flowblocked
3379  *	@dev: network device
3380  *
3381  *	Test if transmit queue on device is currently unable to send.
3382  */
3383 static inline bool netif_queue_stopped(const struct net_device *dev)
3384 {
3385 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3386 }
3387 
3388 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3389 {
3390 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3391 }
3392 
3393 static inline bool
3394 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3395 {
3396 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3397 }
3398 
3399 static inline bool
3400 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3401 {
3402 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3403 }
3404 
3405 /**
3406  *	netdev_queue_set_dql_min_limit - set dql minimum limit
3407  *	@dev_queue: pointer to transmit queue
3408  *	@min_limit: dql minimum limit
3409  *
3410  * Forces xmit_more() to return true until the minimum threshold
3411  * defined by @min_limit is reached (or until the tx queue is
3412  * empty). Warning: to be use with care, misuse will impact the
3413  * latency.
3414  */
3415 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3416 						  unsigned int min_limit)
3417 {
3418 #ifdef CONFIG_BQL
3419 	dev_queue->dql.min_limit = min_limit;
3420 #endif
3421 }
3422 
3423 /**
3424  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3425  *	@dev_queue: pointer to transmit queue
3426  *
3427  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3428  * to give appropriate hint to the CPU.
3429  */
3430 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3431 {
3432 #ifdef CONFIG_BQL
3433 	prefetchw(&dev_queue->dql.num_queued);
3434 #endif
3435 }
3436 
3437 /**
3438  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3439  *	@dev_queue: pointer to transmit queue
3440  *
3441  * BQL enabled drivers might use this helper in their TX completion path,
3442  * to give appropriate hint to the CPU.
3443  */
3444 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3445 {
3446 #ifdef CONFIG_BQL
3447 	prefetchw(&dev_queue->dql.limit);
3448 #endif
3449 }
3450 
3451 /**
3452  *	netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3453  *	@dev_queue: network device queue
3454  *	@bytes: number of bytes queued to the device queue
3455  *
3456  *	Report the number of bytes queued for sending/completion to the network
3457  *	device hardware queue. @bytes should be a good approximation and should
3458  *	exactly match netdev_completed_queue() @bytes.
3459  *	This is typically called once per packet, from ndo_start_xmit().
3460  */
3461 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3462 					unsigned int bytes)
3463 {
3464 #ifdef CONFIG_BQL
3465 	dql_queued(&dev_queue->dql, bytes);
3466 
3467 	if (likely(dql_avail(&dev_queue->dql) >= 0))
3468 		return;
3469 
3470 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3471 
3472 	/*
3473 	 * The XOFF flag must be set before checking the dql_avail below,
3474 	 * because in netdev_tx_completed_queue we update the dql_completed
3475 	 * before checking the XOFF flag.
3476 	 */
3477 	smp_mb();
3478 
3479 	/* check again in case another CPU has just made room avail */
3480 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3481 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3482 #endif
3483 }
3484 
3485 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3486  * that they should not test BQL status themselves.
3487  * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3488  * skb of a batch.
3489  * Returns true if the doorbell must be used to kick the NIC.
3490  */
3491 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3492 					  unsigned int bytes,
3493 					  bool xmit_more)
3494 {
3495 	if (xmit_more) {
3496 #ifdef CONFIG_BQL
3497 		dql_queued(&dev_queue->dql, bytes);
3498 #endif
3499 		return netif_tx_queue_stopped(dev_queue);
3500 	}
3501 	netdev_tx_sent_queue(dev_queue, bytes);
3502 	return true;
3503 }
3504 
3505 /**
3506  *	netdev_sent_queue - report the number of bytes queued to hardware
3507  *	@dev: network device
3508  *	@bytes: number of bytes queued to the hardware device queue
3509  *
3510  *	Report the number of bytes queued for sending/completion to the network
3511  *	device hardware queue#0. @bytes should be a good approximation and should
3512  *	exactly match netdev_completed_queue() @bytes.
3513  *	This is typically called once per packet, from ndo_start_xmit().
3514  */
3515 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3516 {
3517 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3518 }
3519 
3520 static inline bool __netdev_sent_queue(struct net_device *dev,
3521 				       unsigned int bytes,
3522 				       bool xmit_more)
3523 {
3524 	return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3525 				      xmit_more);
3526 }
3527 
3528 /**
3529  *	netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3530  *	@dev_queue: network device queue
3531  *	@pkts: number of packets (currently ignored)
3532  *	@bytes: number of bytes dequeued from the device queue
3533  *
3534  *	Must be called at most once per TX completion round (and not per
3535  *	individual packet), so that BQL can adjust its limits appropriately.
3536  */
3537 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3538 					     unsigned int pkts, unsigned int bytes)
3539 {
3540 #ifdef CONFIG_BQL
3541 	if (unlikely(!bytes))
3542 		return;
3543 
3544 	dql_completed(&dev_queue->dql, bytes);
3545 
3546 	/*
3547 	 * Without the memory barrier there is a small possiblity that
3548 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3549 	 * be stopped forever
3550 	 */
3551 	smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3552 
3553 	if (unlikely(dql_avail(&dev_queue->dql) < 0))
3554 		return;
3555 
3556 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3557 		netif_schedule_queue(dev_queue);
3558 #endif
3559 }
3560 
3561 /**
3562  * 	netdev_completed_queue - report bytes and packets completed by device
3563  * 	@dev: network device
3564  * 	@pkts: actual number of packets sent over the medium
3565  * 	@bytes: actual number of bytes sent over the medium
3566  *
3567  * 	Report the number of bytes and packets transmitted by the network device
3568  * 	hardware queue over the physical medium, @bytes must exactly match the
3569  * 	@bytes amount passed to netdev_sent_queue()
3570  */
3571 static inline void netdev_completed_queue(struct net_device *dev,
3572 					  unsigned int pkts, unsigned int bytes)
3573 {
3574 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3575 }
3576 
3577 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3578 {
3579 #ifdef CONFIG_BQL
3580 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3581 	dql_reset(&q->dql);
3582 #endif
3583 }
3584 
3585 /**
3586  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3587  * 	@dev_queue: network device
3588  *
3589  * 	Reset the bytes and packet count of a network device and clear the
3590  * 	software flow control OFF bit for this network device
3591  */
3592 static inline void netdev_reset_queue(struct net_device *dev_queue)
3593 {
3594 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3595 }
3596 
3597 /**
3598  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3599  * 	@dev: network device
3600  * 	@queue_index: given tx queue index
3601  *
3602  * 	Returns 0 if given tx queue index >= number of device tx queues,
3603  * 	otherwise returns the originally passed tx queue index.
3604  */
3605 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3606 {
3607 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3608 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3609 				     dev->name, queue_index,
3610 				     dev->real_num_tx_queues);
3611 		return 0;
3612 	}
3613 
3614 	return queue_index;
3615 }
3616 
3617 /**
3618  *	netif_running - test if up
3619  *	@dev: network device
3620  *
3621  *	Test if the device has been brought up.
3622  */
3623 static inline bool netif_running(const struct net_device *dev)
3624 {
3625 	return test_bit(__LINK_STATE_START, &dev->state);
3626 }
3627 
3628 /*
3629  * Routines to manage the subqueues on a device.  We only need start,
3630  * stop, and a check if it's stopped.  All other device management is
3631  * done at the overall netdevice level.
3632  * Also test the device if we're multiqueue.
3633  */
3634 
3635 /**
3636  *	netif_start_subqueue - allow sending packets on subqueue
3637  *	@dev: network device
3638  *	@queue_index: sub queue index
3639  *
3640  * Start individual transmit queue of a device with multiple transmit queues.
3641  */
3642 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3643 {
3644 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3645 
3646 	netif_tx_start_queue(txq);
3647 }
3648 
3649 /**
3650  *	netif_stop_subqueue - stop sending packets on subqueue
3651  *	@dev: network device
3652  *	@queue_index: sub queue index
3653  *
3654  * Stop individual transmit queue of a device with multiple transmit queues.
3655  */
3656 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3657 {
3658 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3659 	netif_tx_stop_queue(txq);
3660 }
3661 
3662 /**
3663  *	__netif_subqueue_stopped - test status of subqueue
3664  *	@dev: network device
3665  *	@queue_index: sub queue index
3666  *
3667  * Check individual transmit queue of a device with multiple transmit queues.
3668  */
3669 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3670 					    u16 queue_index)
3671 {
3672 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3673 
3674 	return netif_tx_queue_stopped(txq);
3675 }
3676 
3677 /**
3678  *	netif_subqueue_stopped - test status of subqueue
3679  *	@dev: network device
3680  *	@skb: sub queue buffer pointer
3681  *
3682  * Check individual transmit queue of a device with multiple transmit queues.
3683  */
3684 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3685 					  struct sk_buff *skb)
3686 {
3687 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3688 }
3689 
3690 /**
3691  *	netif_wake_subqueue - allow sending packets on subqueue
3692  *	@dev: network device
3693  *	@queue_index: sub queue index
3694  *
3695  * Resume individual transmit queue of a device with multiple transmit queues.
3696  */
3697 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3698 {
3699 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3700 
3701 	netif_tx_wake_queue(txq);
3702 }
3703 
3704 #ifdef CONFIG_XPS
3705 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3706 			u16 index);
3707 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3708 			  u16 index, enum xps_map_type type);
3709 
3710 /**
3711  *	netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3712  *	@j: CPU/Rx queue index
3713  *	@mask: bitmask of all cpus/rx queues
3714  *	@nr_bits: number of bits in the bitmask
3715  *
3716  * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3717  */
3718 static inline bool netif_attr_test_mask(unsigned long j,
3719 					const unsigned long *mask,
3720 					unsigned int nr_bits)
3721 {
3722 	cpu_max_bits_warn(j, nr_bits);
3723 	return test_bit(j, mask);
3724 }
3725 
3726 /**
3727  *	netif_attr_test_online - Test for online CPU/Rx queue
3728  *	@j: CPU/Rx queue index
3729  *	@online_mask: bitmask for CPUs/Rx queues that are online
3730  *	@nr_bits: number of bits in the bitmask
3731  *
3732  * Returns true if a CPU/Rx queue is online.
3733  */
3734 static inline bool netif_attr_test_online(unsigned long j,
3735 					  const unsigned long *online_mask,
3736 					  unsigned int nr_bits)
3737 {
3738 	cpu_max_bits_warn(j, nr_bits);
3739 
3740 	if (online_mask)
3741 		return test_bit(j, online_mask);
3742 
3743 	return (j < nr_bits);
3744 }
3745 
3746 /**
3747  *	netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3748  *	@n: CPU/Rx queue index
3749  *	@srcp: the cpumask/Rx queue mask pointer
3750  *	@nr_bits: number of bits in the bitmask
3751  *
3752  * Returns >= nr_bits if no further CPUs/Rx queues set.
3753  */
3754 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3755 					       unsigned int nr_bits)
3756 {
3757 	/* -1 is a legal arg here. */
3758 	if (n != -1)
3759 		cpu_max_bits_warn(n, nr_bits);
3760 
3761 	if (srcp)
3762 		return find_next_bit(srcp, nr_bits, n + 1);
3763 
3764 	return n + 1;
3765 }
3766 
3767 /**
3768  *	netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3769  *	@n: CPU/Rx queue index
3770  *	@src1p: the first CPUs/Rx queues mask pointer
3771  *	@src2p: the second CPUs/Rx queues mask pointer
3772  *	@nr_bits: number of bits in the bitmask
3773  *
3774  * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3775  */
3776 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3777 					  const unsigned long *src2p,
3778 					  unsigned int nr_bits)
3779 {
3780 	/* -1 is a legal arg here. */
3781 	if (n != -1)
3782 		cpu_max_bits_warn(n, nr_bits);
3783 
3784 	if (src1p && src2p)
3785 		return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3786 	else if (src1p)
3787 		return find_next_bit(src1p, nr_bits, n + 1);
3788 	else if (src2p)
3789 		return find_next_bit(src2p, nr_bits, n + 1);
3790 
3791 	return n + 1;
3792 }
3793 #else
3794 static inline int netif_set_xps_queue(struct net_device *dev,
3795 				      const struct cpumask *mask,
3796 				      u16 index)
3797 {
3798 	return 0;
3799 }
3800 
3801 static inline int __netif_set_xps_queue(struct net_device *dev,
3802 					const unsigned long *mask,
3803 					u16 index, enum xps_map_type type)
3804 {
3805 	return 0;
3806 }
3807 #endif
3808 
3809 /**
3810  *	netif_is_multiqueue - test if device has multiple transmit queues
3811  *	@dev: network device
3812  *
3813  * Check if device has multiple transmit queues
3814  */
3815 static inline bool netif_is_multiqueue(const struct net_device *dev)
3816 {
3817 	return dev->num_tx_queues > 1;
3818 }
3819 
3820 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3821 
3822 #ifdef CONFIG_SYSFS
3823 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3824 #else
3825 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3826 						unsigned int rxqs)
3827 {
3828 	dev->real_num_rx_queues = rxqs;
3829 	return 0;
3830 }
3831 #endif
3832 int netif_set_real_num_queues(struct net_device *dev,
3833 			      unsigned int txq, unsigned int rxq);
3834 
3835 int netif_get_num_default_rss_queues(void);
3836 
3837 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3838 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3839 
3840 /*
3841  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3842  * interrupt context or with hardware interrupts being disabled.
3843  * (in_hardirq() || irqs_disabled())
3844  *
3845  * We provide four helpers that can be used in following contexts :
3846  *
3847  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3848  *  replacing kfree_skb(skb)
3849  *
3850  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3851  *  Typically used in place of consume_skb(skb) in TX completion path
3852  *
3853  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3854  *  replacing kfree_skb(skb)
3855  *
3856  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3857  *  and consumed a packet. Used in place of consume_skb(skb)
3858  */
3859 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3860 {
3861 	dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3862 }
3863 
3864 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3865 {
3866 	dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3867 }
3868 
3869 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3870 {
3871 	dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3872 }
3873 
3874 static inline void dev_consume_skb_any(struct sk_buff *skb)
3875 {
3876 	dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3877 }
3878 
3879 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3880 			     struct bpf_prog *xdp_prog);
3881 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3882 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3883 int netif_rx(struct sk_buff *skb);
3884 int __netif_rx(struct sk_buff *skb);
3885 
3886 int netif_receive_skb(struct sk_buff *skb);
3887 int netif_receive_skb_core(struct sk_buff *skb);
3888 void netif_receive_skb_list_internal(struct list_head *head);
3889 void netif_receive_skb_list(struct list_head *head);
3890 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3891 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3892 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3893 void napi_get_frags_check(struct napi_struct *napi);
3894 gro_result_t napi_gro_frags(struct napi_struct *napi);
3895 struct packet_offload *gro_find_receive_by_type(__be16 type);
3896 struct packet_offload *gro_find_complete_by_type(__be16 type);
3897 
3898 static inline void napi_free_frags(struct napi_struct *napi)
3899 {
3900 	kfree_skb(napi->skb);
3901 	napi->skb = NULL;
3902 }
3903 
3904 bool netdev_is_rx_handler_busy(struct net_device *dev);
3905 int netdev_rx_handler_register(struct net_device *dev,
3906 			       rx_handler_func_t *rx_handler,
3907 			       void *rx_handler_data);
3908 void netdev_rx_handler_unregister(struct net_device *dev);
3909 
3910 bool dev_valid_name(const char *name);
3911 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3912 {
3913 	return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3914 }
3915 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3916 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3917 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3918 		void __user *data, bool *need_copyout);
3919 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3920 int generic_hwtstamp_get_lower(struct net_device *dev,
3921 			       struct kernel_hwtstamp_config *kernel_cfg);
3922 int generic_hwtstamp_set_lower(struct net_device *dev,
3923 			       struct kernel_hwtstamp_config *kernel_cfg,
3924 			       struct netlink_ext_ack *extack);
3925 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3926 unsigned int dev_get_flags(const struct net_device *);
3927 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3928 		       struct netlink_ext_ack *extack);
3929 int dev_change_flags(struct net_device *dev, unsigned int flags,
3930 		     struct netlink_ext_ack *extack);
3931 int dev_set_alias(struct net_device *, const char *, size_t);
3932 int dev_get_alias(const struct net_device *, char *, size_t);
3933 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3934 			       const char *pat, int new_ifindex);
3935 static inline
3936 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3937 			     const char *pat)
3938 {
3939 	return __dev_change_net_namespace(dev, net, pat, 0);
3940 }
3941 int __dev_set_mtu(struct net_device *, int);
3942 int dev_set_mtu(struct net_device *, int);
3943 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3944 			      struct netlink_ext_ack *extack);
3945 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3946 			struct netlink_ext_ack *extack);
3947 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3948 			     struct netlink_ext_ack *extack);
3949 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3950 int dev_get_port_parent_id(struct net_device *dev,
3951 			   struct netdev_phys_item_id *ppid, bool recurse);
3952 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3953 void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin);
3954 void netdev_dpll_pin_clear(struct net_device *dev);
3955 
3956 static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev)
3957 {
3958 #if IS_ENABLED(CONFIG_DPLL)
3959 	return dev->dpll_pin;
3960 #else
3961 	return NULL;
3962 #endif
3963 }
3964 
3965 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3966 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3967 				    struct netdev_queue *txq, int *ret);
3968 
3969 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3970 u8 dev_xdp_prog_count(struct net_device *dev);
3971 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3972 
3973 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3974 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3975 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3976 bool is_skb_forwardable(const struct net_device *dev,
3977 			const struct sk_buff *skb);
3978 
3979 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3980 						 const struct sk_buff *skb,
3981 						 const bool check_mtu)
3982 {
3983 	const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3984 	unsigned int len;
3985 
3986 	if (!(dev->flags & IFF_UP))
3987 		return false;
3988 
3989 	if (!check_mtu)
3990 		return true;
3991 
3992 	len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3993 	if (skb->len <= len)
3994 		return true;
3995 
3996 	/* if TSO is enabled, we don't care about the length as the packet
3997 	 * could be forwarded without being segmented before
3998 	 */
3999 	if (skb_is_gso(skb))
4000 		return true;
4001 
4002 	return false;
4003 }
4004 
4005 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
4006 
4007 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
4008 {
4009 	/* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
4010 	struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
4011 
4012 	if (likely(p))
4013 		return p;
4014 
4015 	return netdev_core_stats_alloc(dev);
4016 }
4017 
4018 #define DEV_CORE_STATS_INC(FIELD)						\
4019 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev)		\
4020 {										\
4021 	struct net_device_core_stats __percpu *p;				\
4022 										\
4023 	p = dev_core_stats(dev);						\
4024 	if (p)									\
4025 		this_cpu_inc(p->FIELD);						\
4026 }
4027 DEV_CORE_STATS_INC(rx_dropped)
4028 DEV_CORE_STATS_INC(tx_dropped)
4029 DEV_CORE_STATS_INC(rx_nohandler)
4030 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4031 
4032 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4033 					       struct sk_buff *skb,
4034 					       const bool check_mtu)
4035 {
4036 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4037 	    unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4038 		dev_core_stats_rx_dropped_inc(dev);
4039 		kfree_skb(skb);
4040 		return NET_RX_DROP;
4041 	}
4042 
4043 	skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4044 	skb->priority = 0;
4045 	return 0;
4046 }
4047 
4048 bool dev_nit_active(struct net_device *dev);
4049 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4050 
4051 static inline void __dev_put(struct net_device *dev)
4052 {
4053 	if (dev) {
4054 #ifdef CONFIG_PCPU_DEV_REFCNT
4055 		this_cpu_dec(*dev->pcpu_refcnt);
4056 #else
4057 		refcount_dec(&dev->dev_refcnt);
4058 #endif
4059 	}
4060 }
4061 
4062 static inline void __dev_hold(struct net_device *dev)
4063 {
4064 	if (dev) {
4065 #ifdef CONFIG_PCPU_DEV_REFCNT
4066 		this_cpu_inc(*dev->pcpu_refcnt);
4067 #else
4068 		refcount_inc(&dev->dev_refcnt);
4069 #endif
4070 	}
4071 }
4072 
4073 static inline void __netdev_tracker_alloc(struct net_device *dev,
4074 					  netdevice_tracker *tracker,
4075 					  gfp_t gfp)
4076 {
4077 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4078 	ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4079 #endif
4080 }
4081 
4082 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4083  * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4084  */
4085 static inline void netdev_tracker_alloc(struct net_device *dev,
4086 					netdevice_tracker *tracker, gfp_t gfp)
4087 {
4088 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4089 	refcount_dec(&dev->refcnt_tracker.no_tracker);
4090 	__netdev_tracker_alloc(dev, tracker, gfp);
4091 #endif
4092 }
4093 
4094 static inline void netdev_tracker_free(struct net_device *dev,
4095 				       netdevice_tracker *tracker)
4096 {
4097 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4098 	ref_tracker_free(&dev->refcnt_tracker, tracker);
4099 #endif
4100 }
4101 
4102 static inline void netdev_hold(struct net_device *dev,
4103 			       netdevice_tracker *tracker, gfp_t gfp)
4104 {
4105 	if (dev) {
4106 		__dev_hold(dev);
4107 		__netdev_tracker_alloc(dev, tracker, gfp);
4108 	}
4109 }
4110 
4111 static inline void netdev_put(struct net_device *dev,
4112 			      netdevice_tracker *tracker)
4113 {
4114 	if (dev) {
4115 		netdev_tracker_free(dev, tracker);
4116 		__dev_put(dev);
4117 	}
4118 }
4119 
4120 /**
4121  *	dev_hold - get reference to device
4122  *	@dev: network device
4123  *
4124  * Hold reference to device to keep it from being freed.
4125  * Try using netdev_hold() instead.
4126  */
4127 static inline void dev_hold(struct net_device *dev)
4128 {
4129 	netdev_hold(dev, NULL, GFP_ATOMIC);
4130 }
4131 
4132 /**
4133  *	dev_put - release reference to device
4134  *	@dev: network device
4135  *
4136  * Release reference to device to allow it to be freed.
4137  * Try using netdev_put() instead.
4138  */
4139 static inline void dev_put(struct net_device *dev)
4140 {
4141 	netdev_put(dev, NULL);
4142 }
4143 
4144 static inline void netdev_ref_replace(struct net_device *odev,
4145 				      struct net_device *ndev,
4146 				      netdevice_tracker *tracker,
4147 				      gfp_t gfp)
4148 {
4149 	if (odev)
4150 		netdev_tracker_free(odev, tracker);
4151 
4152 	__dev_hold(ndev);
4153 	__dev_put(odev);
4154 
4155 	if (ndev)
4156 		__netdev_tracker_alloc(ndev, tracker, gfp);
4157 }
4158 
4159 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4160  * and _off may be called from IRQ context, but it is caller
4161  * who is responsible for serialization of these calls.
4162  *
4163  * The name carrier is inappropriate, these functions should really be
4164  * called netif_lowerlayer_*() because they represent the state of any
4165  * kind of lower layer not just hardware media.
4166  */
4167 void linkwatch_fire_event(struct net_device *dev);
4168 
4169 /**
4170  *	netif_carrier_ok - test if carrier present
4171  *	@dev: network device
4172  *
4173  * Check if carrier is present on device
4174  */
4175 static inline bool netif_carrier_ok(const struct net_device *dev)
4176 {
4177 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4178 }
4179 
4180 unsigned long dev_trans_start(struct net_device *dev);
4181 
4182 void __netdev_watchdog_up(struct net_device *dev);
4183 
4184 void netif_carrier_on(struct net_device *dev);
4185 void netif_carrier_off(struct net_device *dev);
4186 void netif_carrier_event(struct net_device *dev);
4187 
4188 /**
4189  *	netif_dormant_on - mark device as dormant.
4190  *	@dev: network device
4191  *
4192  * Mark device as dormant (as per RFC2863).
4193  *
4194  * The dormant state indicates that the relevant interface is not
4195  * actually in a condition to pass packets (i.e., it is not 'up') but is
4196  * in a "pending" state, waiting for some external event.  For "on-
4197  * demand" interfaces, this new state identifies the situation where the
4198  * interface is waiting for events to place it in the up state.
4199  */
4200 static inline void netif_dormant_on(struct net_device *dev)
4201 {
4202 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4203 		linkwatch_fire_event(dev);
4204 }
4205 
4206 /**
4207  *	netif_dormant_off - set device as not dormant.
4208  *	@dev: network device
4209  *
4210  * Device is not in dormant state.
4211  */
4212 static inline void netif_dormant_off(struct net_device *dev)
4213 {
4214 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4215 		linkwatch_fire_event(dev);
4216 }
4217 
4218 /**
4219  *	netif_dormant - test if device is dormant
4220  *	@dev: network device
4221  *
4222  * Check if device is dormant.
4223  */
4224 static inline bool netif_dormant(const struct net_device *dev)
4225 {
4226 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
4227 }
4228 
4229 
4230 /**
4231  *	netif_testing_on - mark device as under test.
4232  *	@dev: network device
4233  *
4234  * Mark device as under test (as per RFC2863).
4235  *
4236  * The testing state indicates that some test(s) must be performed on
4237  * the interface. After completion, of the test, the interface state
4238  * will change to up, dormant, or down, as appropriate.
4239  */
4240 static inline void netif_testing_on(struct net_device *dev)
4241 {
4242 	if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4243 		linkwatch_fire_event(dev);
4244 }
4245 
4246 /**
4247  *	netif_testing_off - set device as not under test.
4248  *	@dev: network device
4249  *
4250  * Device is not in testing state.
4251  */
4252 static inline void netif_testing_off(struct net_device *dev)
4253 {
4254 	if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4255 		linkwatch_fire_event(dev);
4256 }
4257 
4258 /**
4259  *	netif_testing - test if device is under test
4260  *	@dev: network device
4261  *
4262  * Check if device is under test
4263  */
4264 static inline bool netif_testing(const struct net_device *dev)
4265 {
4266 	return test_bit(__LINK_STATE_TESTING, &dev->state);
4267 }
4268 
4269 
4270 /**
4271  *	netif_oper_up - test if device is operational
4272  *	@dev: network device
4273  *
4274  * Check if carrier is operational
4275  */
4276 static inline bool netif_oper_up(const struct net_device *dev)
4277 {
4278 	return (dev->operstate == IF_OPER_UP ||
4279 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4280 }
4281 
4282 /**
4283  *	netif_device_present - is device available or removed
4284  *	@dev: network device
4285  *
4286  * Check if device has not been removed from system.
4287  */
4288 static inline bool netif_device_present(const struct net_device *dev)
4289 {
4290 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
4291 }
4292 
4293 void netif_device_detach(struct net_device *dev);
4294 
4295 void netif_device_attach(struct net_device *dev);
4296 
4297 /*
4298  * Network interface message level settings
4299  */
4300 
4301 enum {
4302 	NETIF_MSG_DRV_BIT,
4303 	NETIF_MSG_PROBE_BIT,
4304 	NETIF_MSG_LINK_BIT,
4305 	NETIF_MSG_TIMER_BIT,
4306 	NETIF_MSG_IFDOWN_BIT,
4307 	NETIF_MSG_IFUP_BIT,
4308 	NETIF_MSG_RX_ERR_BIT,
4309 	NETIF_MSG_TX_ERR_BIT,
4310 	NETIF_MSG_TX_QUEUED_BIT,
4311 	NETIF_MSG_INTR_BIT,
4312 	NETIF_MSG_TX_DONE_BIT,
4313 	NETIF_MSG_RX_STATUS_BIT,
4314 	NETIF_MSG_PKTDATA_BIT,
4315 	NETIF_MSG_HW_BIT,
4316 	NETIF_MSG_WOL_BIT,
4317 
4318 	/* When you add a new bit above, update netif_msg_class_names array
4319 	 * in net/ethtool/common.c
4320 	 */
4321 	NETIF_MSG_CLASS_COUNT,
4322 };
4323 /* Both ethtool_ops interface and internal driver implementation use u32 */
4324 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4325 
4326 #define __NETIF_MSG_BIT(bit)	((u32)1 << (bit))
4327 #define __NETIF_MSG(name)	__NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4328 
4329 #define NETIF_MSG_DRV		__NETIF_MSG(DRV)
4330 #define NETIF_MSG_PROBE		__NETIF_MSG(PROBE)
4331 #define NETIF_MSG_LINK		__NETIF_MSG(LINK)
4332 #define NETIF_MSG_TIMER		__NETIF_MSG(TIMER)
4333 #define NETIF_MSG_IFDOWN	__NETIF_MSG(IFDOWN)
4334 #define NETIF_MSG_IFUP		__NETIF_MSG(IFUP)
4335 #define NETIF_MSG_RX_ERR	__NETIF_MSG(RX_ERR)
4336 #define NETIF_MSG_TX_ERR	__NETIF_MSG(TX_ERR)
4337 #define NETIF_MSG_TX_QUEUED	__NETIF_MSG(TX_QUEUED)
4338 #define NETIF_MSG_INTR		__NETIF_MSG(INTR)
4339 #define NETIF_MSG_TX_DONE	__NETIF_MSG(TX_DONE)
4340 #define NETIF_MSG_RX_STATUS	__NETIF_MSG(RX_STATUS)
4341 #define NETIF_MSG_PKTDATA	__NETIF_MSG(PKTDATA)
4342 #define NETIF_MSG_HW		__NETIF_MSG(HW)
4343 #define NETIF_MSG_WOL		__NETIF_MSG(WOL)
4344 
4345 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
4346 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
4347 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
4348 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
4349 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
4350 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
4351 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
4352 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
4353 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4354 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
4355 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
4356 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
4357 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
4358 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
4359 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
4360 
4361 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4362 {
4363 	/* use default */
4364 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4365 		return default_msg_enable_bits;
4366 	if (debug_value == 0)	/* no output */
4367 		return 0;
4368 	/* set low N bits */
4369 	return (1U << debug_value) - 1;
4370 }
4371 
4372 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4373 {
4374 	spin_lock(&txq->_xmit_lock);
4375 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4376 	WRITE_ONCE(txq->xmit_lock_owner, cpu);
4377 }
4378 
4379 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4380 {
4381 	__acquire(&txq->_xmit_lock);
4382 	return true;
4383 }
4384 
4385 static inline void __netif_tx_release(struct netdev_queue *txq)
4386 {
4387 	__release(&txq->_xmit_lock);
4388 }
4389 
4390 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4391 {
4392 	spin_lock_bh(&txq->_xmit_lock);
4393 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4394 	WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4395 }
4396 
4397 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4398 {
4399 	bool ok = spin_trylock(&txq->_xmit_lock);
4400 
4401 	if (likely(ok)) {
4402 		/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4403 		WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4404 	}
4405 	return ok;
4406 }
4407 
4408 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4409 {
4410 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4411 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4412 	spin_unlock(&txq->_xmit_lock);
4413 }
4414 
4415 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4416 {
4417 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4418 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4419 	spin_unlock_bh(&txq->_xmit_lock);
4420 }
4421 
4422 /*
4423  * txq->trans_start can be read locklessly from dev_watchdog()
4424  */
4425 static inline void txq_trans_update(struct netdev_queue *txq)
4426 {
4427 	if (txq->xmit_lock_owner != -1)
4428 		WRITE_ONCE(txq->trans_start, jiffies);
4429 }
4430 
4431 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4432 {
4433 	unsigned long now = jiffies;
4434 
4435 	if (READ_ONCE(txq->trans_start) != now)
4436 		WRITE_ONCE(txq->trans_start, now);
4437 }
4438 
4439 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4440 static inline void netif_trans_update(struct net_device *dev)
4441 {
4442 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4443 
4444 	txq_trans_cond_update(txq);
4445 }
4446 
4447 /**
4448  *	netif_tx_lock - grab network device transmit lock
4449  *	@dev: network device
4450  *
4451  * Get network device transmit lock
4452  */
4453 void netif_tx_lock(struct net_device *dev);
4454 
4455 static inline void netif_tx_lock_bh(struct net_device *dev)
4456 {
4457 	local_bh_disable();
4458 	netif_tx_lock(dev);
4459 }
4460 
4461 void netif_tx_unlock(struct net_device *dev);
4462 
4463 static inline void netif_tx_unlock_bh(struct net_device *dev)
4464 {
4465 	netif_tx_unlock(dev);
4466 	local_bh_enable();
4467 }
4468 
4469 #define HARD_TX_LOCK(dev, txq, cpu) {			\
4470 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4471 		__netif_tx_lock(txq, cpu);		\
4472 	} else {					\
4473 		__netif_tx_acquire(txq);		\
4474 	}						\
4475 }
4476 
4477 #define HARD_TX_TRYLOCK(dev, txq)			\
4478 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
4479 		__netif_tx_trylock(txq) :		\
4480 		__netif_tx_acquire(txq))
4481 
4482 #define HARD_TX_UNLOCK(dev, txq) {			\
4483 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4484 		__netif_tx_unlock(txq);			\
4485 	} else {					\
4486 		__netif_tx_release(txq);		\
4487 	}						\
4488 }
4489 
4490 static inline void netif_tx_disable(struct net_device *dev)
4491 {
4492 	unsigned int i;
4493 	int cpu;
4494 
4495 	local_bh_disable();
4496 	cpu = smp_processor_id();
4497 	spin_lock(&dev->tx_global_lock);
4498 	for (i = 0; i < dev->num_tx_queues; i++) {
4499 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4500 
4501 		__netif_tx_lock(txq, cpu);
4502 		netif_tx_stop_queue(txq);
4503 		__netif_tx_unlock(txq);
4504 	}
4505 	spin_unlock(&dev->tx_global_lock);
4506 	local_bh_enable();
4507 }
4508 
4509 static inline void netif_addr_lock(struct net_device *dev)
4510 {
4511 	unsigned char nest_level = 0;
4512 
4513 #ifdef CONFIG_LOCKDEP
4514 	nest_level = dev->nested_level;
4515 #endif
4516 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4517 }
4518 
4519 static inline void netif_addr_lock_bh(struct net_device *dev)
4520 {
4521 	unsigned char nest_level = 0;
4522 
4523 #ifdef CONFIG_LOCKDEP
4524 	nest_level = dev->nested_level;
4525 #endif
4526 	local_bh_disable();
4527 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4528 }
4529 
4530 static inline void netif_addr_unlock(struct net_device *dev)
4531 {
4532 	spin_unlock(&dev->addr_list_lock);
4533 }
4534 
4535 static inline void netif_addr_unlock_bh(struct net_device *dev)
4536 {
4537 	spin_unlock_bh(&dev->addr_list_lock);
4538 }
4539 
4540 /*
4541  * dev_addrs walker. Should be used only for read access. Call with
4542  * rcu_read_lock held.
4543  */
4544 #define for_each_dev_addr(dev, ha) \
4545 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4546 
4547 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4548 
4549 void ether_setup(struct net_device *dev);
4550 
4551 /* Support for loadable net-drivers */
4552 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4553 				    unsigned char name_assign_type,
4554 				    void (*setup)(struct net_device *),
4555 				    unsigned int txqs, unsigned int rxqs);
4556 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4557 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4558 
4559 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4560 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4561 			 count)
4562 
4563 int register_netdev(struct net_device *dev);
4564 void unregister_netdev(struct net_device *dev);
4565 
4566 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4567 
4568 /* General hardware address lists handling functions */
4569 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4570 		   struct netdev_hw_addr_list *from_list, int addr_len);
4571 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4572 		      struct netdev_hw_addr_list *from_list, int addr_len);
4573 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4574 		       struct net_device *dev,
4575 		       int (*sync)(struct net_device *, const unsigned char *),
4576 		       int (*unsync)(struct net_device *,
4577 				     const unsigned char *));
4578 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4579 			   struct net_device *dev,
4580 			   int (*sync)(struct net_device *,
4581 				       const unsigned char *, int),
4582 			   int (*unsync)(struct net_device *,
4583 					 const unsigned char *, int));
4584 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4585 			      struct net_device *dev,
4586 			      int (*unsync)(struct net_device *,
4587 					    const unsigned char *, int));
4588 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4589 			  struct net_device *dev,
4590 			  int (*unsync)(struct net_device *,
4591 					const unsigned char *));
4592 void __hw_addr_init(struct netdev_hw_addr_list *list);
4593 
4594 /* Functions used for device addresses handling */
4595 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4596 		  const void *addr, size_t len);
4597 
4598 static inline void
4599 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4600 {
4601 	dev_addr_mod(dev, 0, addr, len);
4602 }
4603 
4604 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4605 {
4606 	__dev_addr_set(dev, addr, dev->addr_len);
4607 }
4608 
4609 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4610 		 unsigned char addr_type);
4611 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4612 		 unsigned char addr_type);
4613 
4614 /* Functions used for unicast addresses handling */
4615 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4616 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4617 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4618 int dev_uc_sync(struct net_device *to, struct net_device *from);
4619 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4620 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4621 void dev_uc_flush(struct net_device *dev);
4622 void dev_uc_init(struct net_device *dev);
4623 
4624 /**
4625  *  __dev_uc_sync - Synchonize device's unicast list
4626  *  @dev:  device to sync
4627  *  @sync: function to call if address should be added
4628  *  @unsync: function to call if address should be removed
4629  *
4630  *  Add newly added addresses to the interface, and release
4631  *  addresses that have been deleted.
4632  */
4633 static inline int __dev_uc_sync(struct net_device *dev,
4634 				int (*sync)(struct net_device *,
4635 					    const unsigned char *),
4636 				int (*unsync)(struct net_device *,
4637 					      const unsigned char *))
4638 {
4639 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4640 }
4641 
4642 /**
4643  *  __dev_uc_unsync - Remove synchronized addresses from device
4644  *  @dev:  device to sync
4645  *  @unsync: function to call if address should be removed
4646  *
4647  *  Remove all addresses that were added to the device by dev_uc_sync().
4648  */
4649 static inline void __dev_uc_unsync(struct net_device *dev,
4650 				   int (*unsync)(struct net_device *,
4651 						 const unsigned char *))
4652 {
4653 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
4654 }
4655 
4656 /* Functions used for multicast addresses handling */
4657 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4658 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4659 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4660 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4661 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4662 int dev_mc_sync(struct net_device *to, struct net_device *from);
4663 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4664 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4665 void dev_mc_flush(struct net_device *dev);
4666 void dev_mc_init(struct net_device *dev);
4667 
4668 /**
4669  *  __dev_mc_sync - Synchonize device's multicast list
4670  *  @dev:  device to sync
4671  *  @sync: function to call if address should be added
4672  *  @unsync: function to call if address should be removed
4673  *
4674  *  Add newly added addresses to the interface, and release
4675  *  addresses that have been deleted.
4676  */
4677 static inline int __dev_mc_sync(struct net_device *dev,
4678 				int (*sync)(struct net_device *,
4679 					    const unsigned char *),
4680 				int (*unsync)(struct net_device *,
4681 					      const unsigned char *))
4682 {
4683 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4684 }
4685 
4686 /**
4687  *  __dev_mc_unsync - Remove synchronized addresses from device
4688  *  @dev:  device to sync
4689  *  @unsync: function to call if address should be removed
4690  *
4691  *  Remove all addresses that were added to the device by dev_mc_sync().
4692  */
4693 static inline void __dev_mc_unsync(struct net_device *dev,
4694 				   int (*unsync)(struct net_device *,
4695 						 const unsigned char *))
4696 {
4697 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
4698 }
4699 
4700 /* Functions used for secondary unicast and multicast support */
4701 void dev_set_rx_mode(struct net_device *dev);
4702 int dev_set_promiscuity(struct net_device *dev, int inc);
4703 int dev_set_allmulti(struct net_device *dev, int inc);
4704 void netdev_state_change(struct net_device *dev);
4705 void __netdev_notify_peers(struct net_device *dev);
4706 void netdev_notify_peers(struct net_device *dev);
4707 void netdev_features_change(struct net_device *dev);
4708 /* Load a device via the kmod */
4709 void dev_load(struct net *net, const char *name);
4710 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4711 					struct rtnl_link_stats64 *storage);
4712 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4713 			     const struct net_device_stats *netdev_stats);
4714 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4715 			   const struct pcpu_sw_netstats __percpu *netstats);
4716 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4717 
4718 extern int		netdev_max_backlog;
4719 extern int		dev_rx_weight;
4720 extern int		dev_tx_weight;
4721 extern int		gro_normal_batch;
4722 
4723 enum {
4724 	NESTED_SYNC_IMM_BIT,
4725 	NESTED_SYNC_TODO_BIT,
4726 };
4727 
4728 #define __NESTED_SYNC_BIT(bit)	((u32)1 << (bit))
4729 #define __NESTED_SYNC(name)	__NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4730 
4731 #define NESTED_SYNC_IMM		__NESTED_SYNC(IMM)
4732 #define NESTED_SYNC_TODO	__NESTED_SYNC(TODO)
4733 
4734 struct netdev_nested_priv {
4735 	unsigned char flags;
4736 	void *data;
4737 };
4738 
4739 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4740 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4741 						     struct list_head **iter);
4742 
4743 /* iterate through upper list, must be called under RCU read lock */
4744 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4745 	for (iter = &(dev)->adj_list.upper, \
4746 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4747 	     updev; \
4748 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4749 
4750 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4751 				  int (*fn)(struct net_device *upper_dev,
4752 					    struct netdev_nested_priv *priv),
4753 				  struct netdev_nested_priv *priv);
4754 
4755 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4756 				  struct net_device *upper_dev);
4757 
4758 bool netdev_has_any_upper_dev(struct net_device *dev);
4759 
4760 void *netdev_lower_get_next_private(struct net_device *dev,
4761 				    struct list_head **iter);
4762 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4763 					struct list_head **iter);
4764 
4765 #define netdev_for_each_lower_private(dev, priv, iter) \
4766 	for (iter = (dev)->adj_list.lower.next, \
4767 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
4768 	     priv; \
4769 	     priv = netdev_lower_get_next_private(dev, &(iter)))
4770 
4771 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4772 	for (iter = &(dev)->adj_list.lower, \
4773 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4774 	     priv; \
4775 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4776 
4777 void *netdev_lower_get_next(struct net_device *dev,
4778 				struct list_head **iter);
4779 
4780 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4781 	for (iter = (dev)->adj_list.lower.next, \
4782 	     ldev = netdev_lower_get_next(dev, &(iter)); \
4783 	     ldev; \
4784 	     ldev = netdev_lower_get_next(dev, &(iter)))
4785 
4786 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4787 					     struct list_head **iter);
4788 int netdev_walk_all_lower_dev(struct net_device *dev,
4789 			      int (*fn)(struct net_device *lower_dev,
4790 					struct netdev_nested_priv *priv),
4791 			      struct netdev_nested_priv *priv);
4792 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4793 				  int (*fn)(struct net_device *lower_dev,
4794 					    struct netdev_nested_priv *priv),
4795 				  struct netdev_nested_priv *priv);
4796 
4797 void *netdev_adjacent_get_private(struct list_head *adj_list);
4798 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4799 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4800 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4801 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4802 			  struct netlink_ext_ack *extack);
4803 int netdev_master_upper_dev_link(struct net_device *dev,
4804 				 struct net_device *upper_dev,
4805 				 void *upper_priv, void *upper_info,
4806 				 struct netlink_ext_ack *extack);
4807 void netdev_upper_dev_unlink(struct net_device *dev,
4808 			     struct net_device *upper_dev);
4809 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4810 				   struct net_device *new_dev,
4811 				   struct net_device *dev,
4812 				   struct netlink_ext_ack *extack);
4813 void netdev_adjacent_change_commit(struct net_device *old_dev,
4814 				   struct net_device *new_dev,
4815 				   struct net_device *dev);
4816 void netdev_adjacent_change_abort(struct net_device *old_dev,
4817 				  struct net_device *new_dev,
4818 				  struct net_device *dev);
4819 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4820 void *netdev_lower_dev_get_private(struct net_device *dev,
4821 				   struct net_device *lower_dev);
4822 void netdev_lower_state_changed(struct net_device *lower_dev,
4823 				void *lower_state_info);
4824 
4825 /* RSS keys are 40 or 52 bytes long */
4826 #define NETDEV_RSS_KEY_LEN 52
4827 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4828 void netdev_rss_key_fill(void *buffer, size_t len);
4829 
4830 int skb_checksum_help(struct sk_buff *skb);
4831 int skb_crc32c_csum_help(struct sk_buff *skb);
4832 int skb_csum_hwoffload_help(struct sk_buff *skb,
4833 			    const netdev_features_t features);
4834 
4835 struct netdev_bonding_info {
4836 	ifslave	slave;
4837 	ifbond	master;
4838 };
4839 
4840 struct netdev_notifier_bonding_info {
4841 	struct netdev_notifier_info info; /* must be first */
4842 	struct netdev_bonding_info  bonding_info;
4843 };
4844 
4845 void netdev_bonding_info_change(struct net_device *dev,
4846 				struct netdev_bonding_info *bonding_info);
4847 
4848 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4849 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4850 #else
4851 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4852 				  const void *data)
4853 {
4854 }
4855 #endif
4856 
4857 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4858 
4859 static inline bool can_checksum_protocol(netdev_features_t features,
4860 					 __be16 protocol)
4861 {
4862 	if (protocol == htons(ETH_P_FCOE))
4863 		return !!(features & NETIF_F_FCOE_CRC);
4864 
4865 	/* Assume this is an IP checksum (not SCTP CRC) */
4866 
4867 	if (features & NETIF_F_HW_CSUM) {
4868 		/* Can checksum everything */
4869 		return true;
4870 	}
4871 
4872 	switch (protocol) {
4873 	case htons(ETH_P_IP):
4874 		return !!(features & NETIF_F_IP_CSUM);
4875 	case htons(ETH_P_IPV6):
4876 		return !!(features & NETIF_F_IPV6_CSUM);
4877 	default:
4878 		return false;
4879 	}
4880 }
4881 
4882 #ifdef CONFIG_BUG
4883 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4884 #else
4885 static inline void netdev_rx_csum_fault(struct net_device *dev,
4886 					struct sk_buff *skb)
4887 {
4888 }
4889 #endif
4890 /* rx skb timestamps */
4891 void net_enable_timestamp(void);
4892 void net_disable_timestamp(void);
4893 
4894 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4895 					const struct skb_shared_hwtstamps *hwtstamps,
4896 					bool cycles)
4897 {
4898 	const struct net_device_ops *ops = dev->netdev_ops;
4899 
4900 	if (ops->ndo_get_tstamp)
4901 		return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4902 
4903 	return hwtstamps->hwtstamp;
4904 }
4905 
4906 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4907 					      struct sk_buff *skb, struct net_device *dev,
4908 					      bool more)
4909 {
4910 	__this_cpu_write(softnet_data.xmit.more, more);
4911 	return ops->ndo_start_xmit(skb, dev);
4912 }
4913 
4914 static inline bool netdev_xmit_more(void)
4915 {
4916 	return __this_cpu_read(softnet_data.xmit.more);
4917 }
4918 
4919 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4920 					    struct netdev_queue *txq, bool more)
4921 {
4922 	const struct net_device_ops *ops = dev->netdev_ops;
4923 	netdev_tx_t rc;
4924 
4925 	rc = __netdev_start_xmit(ops, skb, dev, more);
4926 	if (rc == NETDEV_TX_OK)
4927 		txq_trans_update(txq);
4928 
4929 	return rc;
4930 }
4931 
4932 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4933 				const void *ns);
4934 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4935 				 const void *ns);
4936 
4937 extern const struct kobj_ns_type_operations net_ns_type_operations;
4938 
4939 const char *netdev_drivername(const struct net_device *dev);
4940 
4941 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4942 							  netdev_features_t f2)
4943 {
4944 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4945 		if (f1 & NETIF_F_HW_CSUM)
4946 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4947 		else
4948 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4949 	}
4950 
4951 	return f1 & f2;
4952 }
4953 
4954 static inline netdev_features_t netdev_get_wanted_features(
4955 	struct net_device *dev)
4956 {
4957 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4958 }
4959 netdev_features_t netdev_increment_features(netdev_features_t all,
4960 	netdev_features_t one, netdev_features_t mask);
4961 
4962 /* Allow TSO being used on stacked device :
4963  * Performing the GSO segmentation before last device
4964  * is a performance improvement.
4965  */
4966 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4967 							netdev_features_t mask)
4968 {
4969 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4970 }
4971 
4972 int __netdev_update_features(struct net_device *dev);
4973 void netdev_update_features(struct net_device *dev);
4974 void netdev_change_features(struct net_device *dev);
4975 
4976 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4977 					struct net_device *dev);
4978 
4979 netdev_features_t passthru_features_check(struct sk_buff *skb,
4980 					  struct net_device *dev,
4981 					  netdev_features_t features);
4982 netdev_features_t netif_skb_features(struct sk_buff *skb);
4983 void skb_warn_bad_offload(const struct sk_buff *skb);
4984 
4985 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4986 {
4987 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4988 
4989 	/* check flags correspondence */
4990 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4991 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4992 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4993 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4994 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4995 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4996 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4997 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4998 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4999 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5000 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5001 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5002 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5003 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5004 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5005 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5006 	BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5007 	BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5008 	BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5009 
5010 	return (features & feature) == feature;
5011 }
5012 
5013 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5014 {
5015 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5016 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5017 }
5018 
5019 static inline bool netif_needs_gso(struct sk_buff *skb,
5020 				   netdev_features_t features)
5021 {
5022 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5023 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5024 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5025 }
5026 
5027 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5028 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5029 void netif_inherit_tso_max(struct net_device *to,
5030 			   const struct net_device *from);
5031 
5032 static inline bool netif_is_macsec(const struct net_device *dev)
5033 {
5034 	return dev->priv_flags & IFF_MACSEC;
5035 }
5036 
5037 static inline bool netif_is_macvlan(const struct net_device *dev)
5038 {
5039 	return dev->priv_flags & IFF_MACVLAN;
5040 }
5041 
5042 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5043 {
5044 	return dev->priv_flags & IFF_MACVLAN_PORT;
5045 }
5046 
5047 static inline bool netif_is_bond_master(const struct net_device *dev)
5048 {
5049 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5050 }
5051 
5052 static inline bool netif_is_bond_slave(const struct net_device *dev)
5053 {
5054 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5055 }
5056 
5057 static inline bool netif_supports_nofcs(struct net_device *dev)
5058 {
5059 	return dev->priv_flags & IFF_SUPP_NOFCS;
5060 }
5061 
5062 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5063 {
5064 	return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5065 }
5066 
5067 static inline bool netif_is_l3_master(const struct net_device *dev)
5068 {
5069 	return dev->priv_flags & IFF_L3MDEV_MASTER;
5070 }
5071 
5072 static inline bool netif_is_l3_slave(const struct net_device *dev)
5073 {
5074 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
5075 }
5076 
5077 static inline int dev_sdif(const struct net_device *dev)
5078 {
5079 #ifdef CONFIG_NET_L3_MASTER_DEV
5080 	if (netif_is_l3_slave(dev))
5081 		return dev->ifindex;
5082 #endif
5083 	return 0;
5084 }
5085 
5086 static inline bool netif_is_bridge_master(const struct net_device *dev)
5087 {
5088 	return dev->priv_flags & IFF_EBRIDGE;
5089 }
5090 
5091 static inline bool netif_is_bridge_port(const struct net_device *dev)
5092 {
5093 	return dev->priv_flags & IFF_BRIDGE_PORT;
5094 }
5095 
5096 static inline bool netif_is_ovs_master(const struct net_device *dev)
5097 {
5098 	return dev->priv_flags & IFF_OPENVSWITCH;
5099 }
5100 
5101 static inline bool netif_is_ovs_port(const struct net_device *dev)
5102 {
5103 	return dev->priv_flags & IFF_OVS_DATAPATH;
5104 }
5105 
5106 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5107 {
5108 	return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5109 }
5110 
5111 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5112 {
5113 	return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5114 }
5115 
5116 static inline bool netif_is_team_master(const struct net_device *dev)
5117 {
5118 	return dev->priv_flags & IFF_TEAM;
5119 }
5120 
5121 static inline bool netif_is_team_port(const struct net_device *dev)
5122 {
5123 	return dev->priv_flags & IFF_TEAM_PORT;
5124 }
5125 
5126 static inline bool netif_is_lag_master(const struct net_device *dev)
5127 {
5128 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
5129 }
5130 
5131 static inline bool netif_is_lag_port(const struct net_device *dev)
5132 {
5133 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5134 }
5135 
5136 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5137 {
5138 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
5139 }
5140 
5141 static inline bool netif_is_failover(const struct net_device *dev)
5142 {
5143 	return dev->priv_flags & IFF_FAILOVER;
5144 }
5145 
5146 static inline bool netif_is_failover_slave(const struct net_device *dev)
5147 {
5148 	return dev->priv_flags & IFF_FAILOVER_SLAVE;
5149 }
5150 
5151 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5152 static inline void netif_keep_dst(struct net_device *dev)
5153 {
5154 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5155 }
5156 
5157 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5158 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5159 {
5160 	/* TODO: reserve and use an additional IFF bit, if we get more users */
5161 	return netif_is_macsec(dev);
5162 }
5163 
5164 extern struct pernet_operations __net_initdata loopback_net_ops;
5165 
5166 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5167 
5168 /* netdev_printk helpers, similar to dev_printk */
5169 
5170 static inline const char *netdev_name(const struct net_device *dev)
5171 {
5172 	if (!dev->name[0] || strchr(dev->name, '%'))
5173 		return "(unnamed net_device)";
5174 	return dev->name;
5175 }
5176 
5177 static inline const char *netdev_reg_state(const struct net_device *dev)
5178 {
5179 	switch (dev->reg_state) {
5180 	case NETREG_UNINITIALIZED: return " (uninitialized)";
5181 	case NETREG_REGISTERED: return "";
5182 	case NETREG_UNREGISTERING: return " (unregistering)";
5183 	case NETREG_UNREGISTERED: return " (unregistered)";
5184 	case NETREG_RELEASED: return " (released)";
5185 	case NETREG_DUMMY: return " (dummy)";
5186 	}
5187 
5188 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5189 	return " (unknown)";
5190 }
5191 
5192 #define MODULE_ALIAS_NETDEV(device) \
5193 	MODULE_ALIAS("netdev-" device)
5194 
5195 /*
5196  * netdev_WARN() acts like dev_printk(), but with the key difference
5197  * of using a WARN/WARN_ON to get the message out, including the
5198  * file/line information and a backtrace.
5199  */
5200 #define netdev_WARN(dev, format, args...)			\
5201 	WARN(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5202 	     netdev_reg_state(dev), ##args)
5203 
5204 #define netdev_WARN_ONCE(dev, format, args...)				\
5205 	WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5206 		  netdev_reg_state(dev), ##args)
5207 
5208 /*
5209  *	The list of packet types we will receive (as opposed to discard)
5210  *	and the routines to invoke.
5211  *
5212  *	Why 16. Because with 16 the only overlap we get on a hash of the
5213  *	low nibble of the protocol value is RARP/SNAP/X.25.
5214  *
5215  *		0800	IP
5216  *		0001	802.3
5217  *		0002	AX.25
5218  *		0004	802.2
5219  *		8035	RARP
5220  *		0005	SNAP
5221  *		0805	X.25
5222  *		0806	ARP
5223  *		8137	IPX
5224  *		0009	Localtalk
5225  *		86DD	IPv6
5226  */
5227 #define PTYPE_HASH_SIZE	(16)
5228 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
5229 
5230 extern struct list_head ptype_all __read_mostly;
5231 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5232 
5233 extern struct net_device *blackhole_netdev;
5234 
5235 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5236 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5237 #define DEV_STATS_ADD(DEV, FIELD, VAL) 	\
5238 		atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5239 
5240 #endif	/* _LINUX_NETDEVICE_H */
5241