xref: /linux-6.15/include/linux/netdevice.h (revision ff999149)
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 /**
486  * napi_is_scheduled - test if NAPI is scheduled
487  * @n: NAPI context
488  *
489  * This check is "best-effort". With no locking implemented,
490  * a NAPI can be scheduled or terminate right after this check
491  * and produce not precise results.
492  *
493  * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
494  * should not be used normally and napi_schedule should be
495  * used instead.
496  *
497  * Use only if the driver really needs to check if a NAPI
498  * is scheduled for example in the context of delayed timer
499  * that can be skipped if a NAPI is already scheduled.
500  *
501  * Return True if NAPI is scheduled, False otherwise.
502  */
503 static inline bool napi_is_scheduled(struct napi_struct *n)
504 {
505 	return test_bit(NAPI_STATE_SCHED, &n->state);
506 }
507 
508 bool napi_schedule_prep(struct napi_struct *n);
509 
510 /**
511  *	napi_schedule - schedule NAPI poll
512  *	@n: NAPI context
513  *
514  * Schedule NAPI poll routine to be called if it is not already
515  * running.
516  * Return true if we schedule a NAPI or false if not.
517  * Refer to napi_schedule_prep() for additional reason on why
518  * a NAPI might not be scheduled.
519  */
520 static inline bool napi_schedule(struct napi_struct *n)
521 {
522 	if (napi_schedule_prep(n)) {
523 		__napi_schedule(n);
524 		return true;
525 	}
526 
527 	return false;
528 }
529 
530 /**
531  *	napi_schedule_irqoff - schedule NAPI poll
532  *	@n: NAPI context
533  *
534  * Variant of napi_schedule(), assuming hard irqs are masked.
535  */
536 static inline void napi_schedule_irqoff(struct napi_struct *n)
537 {
538 	if (napi_schedule_prep(n))
539 		__napi_schedule_irqoff(n);
540 }
541 
542 /**
543  * napi_complete_done - NAPI processing complete
544  * @n: NAPI context
545  * @work_done: number of packets processed
546  *
547  * Mark NAPI processing as complete. Should only be called if poll budget
548  * has not been completely consumed.
549  * Prefer over napi_complete().
550  * Return false if device should avoid rearming interrupts.
551  */
552 bool napi_complete_done(struct napi_struct *n, int work_done);
553 
554 static inline bool napi_complete(struct napi_struct *n)
555 {
556 	return napi_complete_done(n, 0);
557 }
558 
559 int dev_set_threaded(struct net_device *dev, bool threaded);
560 
561 /**
562  *	napi_disable - prevent NAPI from scheduling
563  *	@n: NAPI context
564  *
565  * Stop NAPI from being scheduled on this context.
566  * Waits till any outstanding processing completes.
567  */
568 void napi_disable(struct napi_struct *n);
569 
570 void napi_enable(struct napi_struct *n);
571 
572 /**
573  *	napi_synchronize - wait until NAPI is not running
574  *	@n: NAPI context
575  *
576  * Wait until NAPI is done being scheduled on this context.
577  * Waits till any outstanding processing completes but
578  * does not disable future activations.
579  */
580 static inline void napi_synchronize(const struct napi_struct *n)
581 {
582 	if (IS_ENABLED(CONFIG_SMP))
583 		while (test_bit(NAPI_STATE_SCHED, &n->state))
584 			msleep(1);
585 	else
586 		barrier();
587 }
588 
589 /**
590  *	napi_if_scheduled_mark_missed - if napi is running, set the
591  *	NAPIF_STATE_MISSED
592  *	@n: NAPI context
593  *
594  * If napi is running, set the NAPIF_STATE_MISSED, and return true if
595  * NAPI is scheduled.
596  **/
597 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
598 {
599 	unsigned long val, new;
600 
601 	val = READ_ONCE(n->state);
602 	do {
603 		if (val & NAPIF_STATE_DISABLE)
604 			return true;
605 
606 		if (!(val & NAPIF_STATE_SCHED))
607 			return false;
608 
609 		new = val | NAPIF_STATE_MISSED;
610 	} while (!try_cmpxchg(&n->state, &val, new));
611 
612 	return true;
613 }
614 
615 enum netdev_queue_state_t {
616 	__QUEUE_STATE_DRV_XOFF,
617 	__QUEUE_STATE_STACK_XOFF,
618 	__QUEUE_STATE_FROZEN,
619 };
620 
621 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
622 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
623 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
624 
625 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
626 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
627 					QUEUE_STATE_FROZEN)
628 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
629 					QUEUE_STATE_FROZEN)
630 
631 /*
632  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
633  * netif_tx_* functions below are used to manipulate this flag.  The
634  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
635  * queue independently.  The netif_xmit_*stopped functions below are called
636  * to check if the queue has been stopped by the driver or stack (either
637  * of the XOFF bits are set in the state).  Drivers should not need to call
638  * netif_xmit*stopped functions, they should only be using netif_tx_*.
639  */
640 
641 struct netdev_queue {
642 /*
643  * read-mostly part
644  */
645 	struct net_device	*dev;
646 	netdevice_tracker	dev_tracker;
647 
648 	struct Qdisc __rcu	*qdisc;
649 	struct Qdisc __rcu	*qdisc_sleeping;
650 #ifdef CONFIG_SYSFS
651 	struct kobject		kobj;
652 #endif
653 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
654 	int			numa_node;
655 #endif
656 	unsigned long		tx_maxrate;
657 	/*
658 	 * Number of TX timeouts for this queue
659 	 * (/sys/class/net/DEV/Q/trans_timeout)
660 	 */
661 	atomic_long_t		trans_timeout;
662 
663 	/* Subordinate device that the queue has been assigned to */
664 	struct net_device	*sb_dev;
665 #ifdef CONFIG_XDP_SOCKETS
666 	struct xsk_buff_pool    *pool;
667 #endif
668 	/* NAPI instance for the queue
669 	 * Readers and writers must hold RTNL
670 	 */
671 	struct napi_struct      *napi;
672 /*
673  * write-mostly part
674  */
675 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
676 	int			xmit_lock_owner;
677 	/*
678 	 * Time (in jiffies) of last Tx
679 	 */
680 	unsigned long		trans_start;
681 
682 	unsigned long		state;
683 
684 #ifdef CONFIG_BQL
685 	struct dql		dql;
686 #endif
687 } ____cacheline_aligned_in_smp;
688 
689 extern int sysctl_fb_tunnels_only_for_init_net;
690 extern int sysctl_devconf_inherit_init_net;
691 
692 /*
693  * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
694  *                                     == 1 : For initns only
695  *                                     == 2 : For none.
696  */
697 static inline bool net_has_fallback_tunnels(const struct net *net)
698 {
699 #if IS_ENABLED(CONFIG_SYSCTL)
700 	int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
701 
702 	return !fb_tunnels_only_for_init_net ||
703 		(net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
704 #else
705 	return true;
706 #endif
707 }
708 
709 static inline int net_inherit_devconf(void)
710 {
711 #if IS_ENABLED(CONFIG_SYSCTL)
712 	return READ_ONCE(sysctl_devconf_inherit_init_net);
713 #else
714 	return 0;
715 #endif
716 }
717 
718 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
719 {
720 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
721 	return q->numa_node;
722 #else
723 	return NUMA_NO_NODE;
724 #endif
725 }
726 
727 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
728 {
729 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
730 	q->numa_node = node;
731 #endif
732 }
733 
734 #ifdef CONFIG_RPS
735 /*
736  * This structure holds an RPS map which can be of variable length.  The
737  * map is an array of CPUs.
738  */
739 struct rps_map {
740 	unsigned int len;
741 	struct rcu_head rcu;
742 	u16 cpus[];
743 };
744 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
745 
746 /*
747  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
748  * tail pointer for that CPU's input queue at the time of last enqueue, and
749  * a hardware filter index.
750  */
751 struct rps_dev_flow {
752 	u16 cpu;
753 	u16 filter;
754 	unsigned int last_qtail;
755 };
756 #define RPS_NO_FILTER 0xffff
757 
758 /*
759  * The rps_dev_flow_table structure contains a table of flow mappings.
760  */
761 struct rps_dev_flow_table {
762 	unsigned int mask;
763 	struct rcu_head rcu;
764 	struct rps_dev_flow flows[];
765 };
766 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
767     ((_num) * sizeof(struct rps_dev_flow)))
768 
769 /*
770  * The rps_sock_flow_table contains mappings of flows to the last CPU
771  * on which they were processed by the application (set in recvmsg).
772  * Each entry is a 32bit value. Upper part is the high-order bits
773  * of flow hash, lower part is CPU number.
774  * rps_cpu_mask is used to partition the space, depending on number of
775  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
776  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
777  * meaning we use 32-6=26 bits for the hash.
778  */
779 struct rps_sock_flow_table {
780 	u32	mask;
781 
782 	u32	ents[] ____cacheline_aligned_in_smp;
783 };
784 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
785 
786 #define RPS_NO_CPU 0xffff
787 
788 extern u32 rps_cpu_mask;
789 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
790 
791 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
792 					u32 hash)
793 {
794 	if (table && hash) {
795 		unsigned int index = hash & table->mask;
796 		u32 val = hash & ~rps_cpu_mask;
797 
798 		/* We only give a hint, preemption can change CPU under us */
799 		val |= raw_smp_processor_id();
800 
801 		/* The following WRITE_ONCE() is paired with the READ_ONCE()
802 		 * here, and another one in get_rps_cpu().
803 		 */
804 		if (READ_ONCE(table->ents[index]) != val)
805 			WRITE_ONCE(table->ents[index], val);
806 	}
807 }
808 
809 #ifdef CONFIG_RFS_ACCEL
810 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
811 			 u16 filter_id);
812 #endif
813 #endif /* CONFIG_RPS */
814 
815 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
816 enum xps_map_type {
817 	XPS_CPUS = 0,
818 	XPS_RXQS,
819 	XPS_MAPS_MAX,
820 };
821 
822 #ifdef CONFIG_XPS
823 /*
824  * This structure holds an XPS map which can be of variable length.  The
825  * map is an array of queues.
826  */
827 struct xps_map {
828 	unsigned int len;
829 	unsigned int alloc_len;
830 	struct rcu_head rcu;
831 	u16 queues[];
832 };
833 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
834 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
835        - sizeof(struct xps_map)) / sizeof(u16))
836 
837 /*
838  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
839  *
840  * We keep track of the number of cpus/rxqs used when the struct is allocated,
841  * in nr_ids. This will help not accessing out-of-bound memory.
842  *
843  * We keep track of the number of traffic classes used when the struct is
844  * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
845  * not crossing its upper bound, as the original dev->num_tc can be updated in
846  * the meantime.
847  */
848 struct xps_dev_maps {
849 	struct rcu_head rcu;
850 	unsigned int nr_ids;
851 	s16 num_tc;
852 	struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
853 };
854 
855 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +	\
856 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
857 
858 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
859 	(_rxqs * (_tcs) * sizeof(struct xps_map *)))
860 
861 #endif /* CONFIG_XPS */
862 
863 #define TC_MAX_QUEUE	16
864 #define TC_BITMASK	15
865 /* HW offloaded queuing disciplines txq count and offset maps */
866 struct netdev_tc_txq {
867 	u16 count;
868 	u16 offset;
869 };
870 
871 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
872 /*
873  * This structure is to hold information about the device
874  * configured to run FCoE protocol stack.
875  */
876 struct netdev_fcoe_hbainfo {
877 	char	manufacturer[64];
878 	char	serial_number[64];
879 	char	hardware_version[64];
880 	char	driver_version[64];
881 	char	optionrom_version[64];
882 	char	firmware_version[64];
883 	char	model[256];
884 	char	model_description[256];
885 };
886 #endif
887 
888 #define MAX_PHYS_ITEM_ID_LEN 32
889 
890 /* This structure holds a unique identifier to identify some
891  * physical item (port for example) used by a netdevice.
892  */
893 struct netdev_phys_item_id {
894 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
895 	unsigned char id_len;
896 };
897 
898 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
899 					    struct netdev_phys_item_id *b)
900 {
901 	return a->id_len == b->id_len &&
902 	       memcmp(a->id, b->id, a->id_len) == 0;
903 }
904 
905 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
906 				       struct sk_buff *skb,
907 				       struct net_device *sb_dev);
908 
909 enum net_device_path_type {
910 	DEV_PATH_ETHERNET = 0,
911 	DEV_PATH_VLAN,
912 	DEV_PATH_BRIDGE,
913 	DEV_PATH_PPPOE,
914 	DEV_PATH_DSA,
915 	DEV_PATH_MTK_WDMA,
916 };
917 
918 struct net_device_path {
919 	enum net_device_path_type	type;
920 	const struct net_device		*dev;
921 	union {
922 		struct {
923 			u16		id;
924 			__be16		proto;
925 			u8		h_dest[ETH_ALEN];
926 		} encap;
927 		struct {
928 			enum {
929 				DEV_PATH_BR_VLAN_KEEP,
930 				DEV_PATH_BR_VLAN_TAG,
931 				DEV_PATH_BR_VLAN_UNTAG,
932 				DEV_PATH_BR_VLAN_UNTAG_HW,
933 			}		vlan_mode;
934 			u16		vlan_id;
935 			__be16		vlan_proto;
936 		} bridge;
937 		struct {
938 			int port;
939 			u16 proto;
940 		} dsa;
941 		struct {
942 			u8 wdma_idx;
943 			u8 queue;
944 			u16 wcid;
945 			u8 bss;
946 			u8 amsdu;
947 		} mtk_wdma;
948 	};
949 };
950 
951 #define NET_DEVICE_PATH_STACK_MAX	5
952 #define NET_DEVICE_PATH_VLAN_MAX	2
953 
954 struct net_device_path_stack {
955 	int			num_paths;
956 	struct net_device_path	path[NET_DEVICE_PATH_STACK_MAX];
957 };
958 
959 struct net_device_path_ctx {
960 	const struct net_device *dev;
961 	u8			daddr[ETH_ALEN];
962 
963 	int			num_vlans;
964 	struct {
965 		u16		id;
966 		__be16		proto;
967 	} vlan[NET_DEVICE_PATH_VLAN_MAX];
968 };
969 
970 enum tc_setup_type {
971 	TC_QUERY_CAPS,
972 	TC_SETUP_QDISC_MQPRIO,
973 	TC_SETUP_CLSU32,
974 	TC_SETUP_CLSFLOWER,
975 	TC_SETUP_CLSMATCHALL,
976 	TC_SETUP_CLSBPF,
977 	TC_SETUP_BLOCK,
978 	TC_SETUP_QDISC_CBS,
979 	TC_SETUP_QDISC_RED,
980 	TC_SETUP_QDISC_PRIO,
981 	TC_SETUP_QDISC_MQ,
982 	TC_SETUP_QDISC_ETF,
983 	TC_SETUP_ROOT_QDISC,
984 	TC_SETUP_QDISC_GRED,
985 	TC_SETUP_QDISC_TAPRIO,
986 	TC_SETUP_FT,
987 	TC_SETUP_QDISC_ETS,
988 	TC_SETUP_QDISC_TBF,
989 	TC_SETUP_QDISC_FIFO,
990 	TC_SETUP_QDISC_HTB,
991 	TC_SETUP_ACT,
992 };
993 
994 /* These structures hold the attributes of bpf state that are being passed
995  * to the netdevice through the bpf op.
996  */
997 enum bpf_netdev_command {
998 	/* Set or clear a bpf program used in the earliest stages of packet
999 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
1000 	 * is responsible for calling bpf_prog_put on any old progs that are
1001 	 * stored. In case of error, the callee need not release the new prog
1002 	 * reference, but on success it takes ownership and must bpf_prog_put
1003 	 * when it is no longer used.
1004 	 */
1005 	XDP_SETUP_PROG,
1006 	XDP_SETUP_PROG_HW,
1007 	/* BPF program for offload callbacks, invoked at program load time. */
1008 	BPF_OFFLOAD_MAP_ALLOC,
1009 	BPF_OFFLOAD_MAP_FREE,
1010 	XDP_SETUP_XSK_POOL,
1011 };
1012 
1013 struct bpf_prog_offload_ops;
1014 struct netlink_ext_ack;
1015 struct xdp_umem;
1016 struct xdp_dev_bulk_queue;
1017 struct bpf_xdp_link;
1018 
1019 enum bpf_xdp_mode {
1020 	XDP_MODE_SKB = 0,
1021 	XDP_MODE_DRV = 1,
1022 	XDP_MODE_HW = 2,
1023 	__MAX_XDP_MODE
1024 };
1025 
1026 struct bpf_xdp_entity {
1027 	struct bpf_prog *prog;
1028 	struct bpf_xdp_link *link;
1029 };
1030 
1031 struct netdev_bpf {
1032 	enum bpf_netdev_command command;
1033 	union {
1034 		/* XDP_SETUP_PROG */
1035 		struct {
1036 			u32 flags;
1037 			struct bpf_prog *prog;
1038 			struct netlink_ext_ack *extack;
1039 		};
1040 		/* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1041 		struct {
1042 			struct bpf_offloaded_map *offmap;
1043 		};
1044 		/* XDP_SETUP_XSK_POOL */
1045 		struct {
1046 			struct xsk_buff_pool *pool;
1047 			u16 queue_id;
1048 		} xsk;
1049 	};
1050 };
1051 
1052 /* Flags for ndo_xsk_wakeup. */
1053 #define XDP_WAKEUP_RX (1 << 0)
1054 #define XDP_WAKEUP_TX (1 << 1)
1055 
1056 #ifdef CONFIG_XFRM_OFFLOAD
1057 struct xfrmdev_ops {
1058 	int	(*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1059 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
1060 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
1061 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
1062 				       struct xfrm_state *x);
1063 	void	(*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1064 	void	(*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1065 	int	(*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1066 	void	(*xdo_dev_policy_delete) (struct xfrm_policy *x);
1067 	void	(*xdo_dev_policy_free) (struct xfrm_policy *x);
1068 };
1069 #endif
1070 
1071 struct dev_ifalias {
1072 	struct rcu_head rcuhead;
1073 	char ifalias[];
1074 };
1075 
1076 struct devlink;
1077 struct tlsdev_ops;
1078 
1079 struct netdev_net_notifier {
1080 	struct list_head list;
1081 	struct notifier_block *nb;
1082 };
1083 
1084 /*
1085  * This structure defines the management hooks for network devices.
1086  * The following hooks can be defined; unless noted otherwise, they are
1087  * optional and can be filled with a null pointer.
1088  *
1089  * int (*ndo_init)(struct net_device *dev);
1090  *     This function is called once when a network device is registered.
1091  *     The network device can use this for any late stage initialization
1092  *     or semantic validation. It can fail with an error code which will
1093  *     be propagated back to register_netdev.
1094  *
1095  * void (*ndo_uninit)(struct net_device *dev);
1096  *     This function is called when device is unregistered or when registration
1097  *     fails. It is not called if init fails.
1098  *
1099  * int (*ndo_open)(struct net_device *dev);
1100  *     This function is called when a network device transitions to the up
1101  *     state.
1102  *
1103  * int (*ndo_stop)(struct net_device *dev);
1104  *     This function is called when a network device transitions to the down
1105  *     state.
1106  *
1107  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1108  *                               struct net_device *dev);
1109  *	Called when a packet needs to be transmitted.
1110  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
1111  *	the queue before that can happen; it's for obsolete devices and weird
1112  *	corner cases, but the stack really does a non-trivial amount
1113  *	of useless work if you return NETDEV_TX_BUSY.
1114  *	Required; cannot be NULL.
1115  *
1116  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1117  *					   struct net_device *dev
1118  *					   netdev_features_t features);
1119  *	Called by core transmit path to determine if device is capable of
1120  *	performing offload operations on a given packet. This is to give
1121  *	the device an opportunity to implement any restrictions that cannot
1122  *	be otherwise expressed by feature flags. The check is called with
1123  *	the set of features that the stack has calculated and it returns
1124  *	those the driver believes to be appropriate.
1125  *
1126  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1127  *                         struct net_device *sb_dev);
1128  *	Called to decide which queue to use when device supports multiple
1129  *	transmit queues.
1130  *
1131  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1132  *	This function is called to allow device receiver to make
1133  *	changes to configuration when multicast or promiscuous is enabled.
1134  *
1135  * void (*ndo_set_rx_mode)(struct net_device *dev);
1136  *	This function is called device changes address list filtering.
1137  *	If driver handles unicast address filtering, it should set
1138  *	IFF_UNICAST_FLT in its priv_flags.
1139  *
1140  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1141  *	This function  is called when the Media Access Control address
1142  *	needs to be changed. If this interface is not defined, the
1143  *	MAC address can not be changed.
1144  *
1145  * int (*ndo_validate_addr)(struct net_device *dev);
1146  *	Test if Media Access Control address is valid for the device.
1147  *
1148  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1149  *	Old-style ioctl entry point. This is used internally by the
1150  *	appletalk and ieee802154 subsystems but is no longer called by
1151  *	the device ioctl handler.
1152  *
1153  * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1154  *	Used by the bonding driver for its device specific ioctls:
1155  *	SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1156  *	SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1157  *
1158  * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1159  *	Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1160  *	SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1161  *
1162  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1163  *	Used to set network devices bus interface parameters. This interface
1164  *	is retained for legacy reasons; new devices should use the bus
1165  *	interface (PCI) for low level management.
1166  *
1167  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1168  *	Called when a user wants to change the Maximum Transfer Unit
1169  *	of a device.
1170  *
1171  * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1172  *	Callback used when the transmitter has not made any progress
1173  *	for dev->watchdog ticks.
1174  *
1175  * void (*ndo_get_stats64)(struct net_device *dev,
1176  *                         struct rtnl_link_stats64 *storage);
1177  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1178  *	Called when a user wants to get the network device usage
1179  *	statistics. Drivers must do one of the following:
1180  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
1181  *	   rtnl_link_stats64 structure passed by the caller.
1182  *	2. Define @ndo_get_stats to update a net_device_stats structure
1183  *	   (which should normally be dev->stats) and return a pointer to
1184  *	   it. The structure may be changed asynchronously only if each
1185  *	   field is written atomically.
1186  *	3. Update dev->stats asynchronously and atomically, and define
1187  *	   neither operation.
1188  *
1189  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1190  *	Return true if this device supports offload stats of this attr_id.
1191  *
1192  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1193  *	void *attr_data)
1194  *	Get statistics for offload operations by attr_id. Write it into the
1195  *	attr_data pointer.
1196  *
1197  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1198  *	If device supports VLAN filtering this function is called when a
1199  *	VLAN id is registered.
1200  *
1201  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1202  *	If device supports VLAN filtering this function is called when a
1203  *	VLAN id is unregistered.
1204  *
1205  * void (*ndo_poll_controller)(struct net_device *dev);
1206  *
1207  *	SR-IOV management functions.
1208  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1209  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1210  *			  u8 qos, __be16 proto);
1211  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1212  *			  int max_tx_rate);
1213  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1214  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1215  * int (*ndo_get_vf_config)(struct net_device *dev,
1216  *			    int vf, struct ifla_vf_info *ivf);
1217  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1218  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1219  *			  struct nlattr *port[]);
1220  *
1221  *      Enable or disable the VF ability to query its RSS Redirection Table and
1222  *      Hash Key. This is needed since on some devices VF share this information
1223  *      with PF and querying it may introduce a theoretical security risk.
1224  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1225  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1226  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1227  *		       void *type_data);
1228  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
1229  *	This is always called from the stack with the rtnl lock held and netif
1230  *	tx queues stopped. This allows the netdevice to perform queue
1231  *	management safely.
1232  *
1233  *	Fiber Channel over Ethernet (FCoE) offload functions.
1234  * int (*ndo_fcoe_enable)(struct net_device *dev);
1235  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
1236  *	so the underlying device can perform whatever needed configuration or
1237  *	initialization to support acceleration of FCoE traffic.
1238  *
1239  * int (*ndo_fcoe_disable)(struct net_device *dev);
1240  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
1241  *	so the underlying device can perform whatever needed clean-ups to
1242  *	stop supporting acceleration of FCoE traffic.
1243  *
1244  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1245  *			     struct scatterlist *sgl, unsigned int sgc);
1246  *	Called when the FCoE Initiator wants to initialize an I/O that
1247  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1248  *	perform necessary setup and returns 1 to indicate the device is set up
1249  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1250  *
1251  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1252  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1253  *	indicated by the FC exchange id 'xid', so the underlying device can
1254  *	clean up and reuse resources for later DDP requests.
1255  *
1256  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1257  *			      struct scatterlist *sgl, unsigned int sgc);
1258  *	Called when the FCoE Target wants to initialize an I/O that
1259  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1260  *	perform necessary setup and returns 1 to indicate the device is set up
1261  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1262  *
1263  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1264  *			       struct netdev_fcoe_hbainfo *hbainfo);
1265  *	Called when the FCoE Protocol stack wants information on the underlying
1266  *	device. This information is utilized by the FCoE protocol stack to
1267  *	register attributes with Fiber Channel management service as per the
1268  *	FC-GS Fabric Device Management Information(FDMI) specification.
1269  *
1270  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1271  *	Called when the underlying device wants to override default World Wide
1272  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1273  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1274  *	protocol stack to use.
1275  *
1276  *	RFS acceleration.
1277  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1278  *			    u16 rxq_index, u32 flow_id);
1279  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1280  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1281  *	Return the filter ID on success, or a negative error code.
1282  *
1283  *	Slave management functions (for bridge, bonding, etc).
1284  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1285  *	Called to make another netdev an underling.
1286  *
1287  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1288  *	Called to release previously enslaved netdev.
1289  *
1290  * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1291  *					    struct sk_buff *skb,
1292  *					    bool all_slaves);
1293  *	Get the xmit slave of master device. If all_slaves is true, function
1294  *	assume all the slaves can transmit.
1295  *
1296  *      Feature/offload setting functions.
1297  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1298  *		netdev_features_t features);
1299  *	Adjusts the requested feature flags according to device-specific
1300  *	constraints, and returns the resulting flags. Must not modify
1301  *	the device state.
1302  *
1303  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1304  *	Called to update device configuration to new features. Passed
1305  *	feature set might be less than what was returned by ndo_fix_features()).
1306  *	Must return >0 or -errno if it changed dev->features itself.
1307  *
1308  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1309  *		      struct net_device *dev,
1310  *		      const unsigned char *addr, u16 vid, u16 flags,
1311  *		      struct netlink_ext_ack *extack);
1312  *	Adds an FDB entry to dev for addr.
1313  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1314  *		      struct net_device *dev,
1315  *		      const unsigned char *addr, u16 vid)
1316  *	Deletes the FDB entry from dev coresponding to addr.
1317  * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1318  *			   struct netlink_ext_ack *extack);
1319  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1320  *		       struct net_device *dev, struct net_device *filter_dev,
1321  *		       int *idx)
1322  *	Used to add FDB entries to dump requests. Implementers should add
1323  *	entries to skb and update idx with the number of entries.
1324  *
1325  * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1326  *		      u16 nlmsg_flags, struct netlink_ext_ack *extack);
1327  *	Adds an MDB entry to dev.
1328  * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1329  *		      struct netlink_ext_ack *extack);
1330  *	Deletes the MDB entry from dev.
1331  * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1332  *		       struct netlink_callback *cb);
1333  *	Dumps MDB entries from dev. The first argument (marker) in the netlink
1334  *	callback is used by core rtnetlink code.
1335  *
1336  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1337  *			     u16 flags, struct netlink_ext_ack *extack)
1338  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1339  *			     struct net_device *dev, u32 filter_mask,
1340  *			     int nlflags)
1341  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1342  *			     u16 flags);
1343  *
1344  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1345  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1346  *	which do not represent real hardware may define this to allow their
1347  *	userspace components to manage their virtual carrier state. Devices
1348  *	that determine carrier state from physical hardware properties (eg
1349  *	network cables) or protocol-dependent mechanisms (eg
1350  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1351  *
1352  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1353  *			       struct netdev_phys_item_id *ppid);
1354  *	Called to get ID of physical port of this device. If driver does
1355  *	not implement this, it is assumed that the hw is not able to have
1356  *	multiple net devices on single physical port.
1357  *
1358  * int (*ndo_get_port_parent_id)(struct net_device *dev,
1359  *				 struct netdev_phys_item_id *ppid)
1360  *	Called to get the parent ID of the physical port of this device.
1361  *
1362  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1363  *				 struct net_device *dev)
1364  *	Called by upper layer devices to accelerate switching or other
1365  *	station functionality into hardware. 'pdev is the lowerdev
1366  *	to use for the offload and 'dev' is the net device that will
1367  *	back the offload. Returns a pointer to the private structure
1368  *	the upper layer will maintain.
1369  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1370  *	Called by upper layer device to delete the station created
1371  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1372  *	the station and priv is the structure returned by the add
1373  *	operation.
1374  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1375  *			     int queue_index, u32 maxrate);
1376  *	Called when a user wants to set a max-rate limitation of specific
1377  *	TX queue.
1378  * int (*ndo_get_iflink)(const struct net_device *dev);
1379  *	Called to get the iflink value of this device.
1380  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1381  *	This function is used to get egress tunnel information for given skb.
1382  *	This is useful for retrieving outer tunnel header parameters while
1383  *	sampling packet.
1384  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1385  *	This function is used to specify the headroom that the skb must
1386  *	consider when allocation skb during packet reception. Setting
1387  *	appropriate rx headroom value allows avoiding skb head copy on
1388  *	forward. Setting a negative value resets the rx headroom to the
1389  *	default value.
1390  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1391  *	This function is used to set or query state related to XDP on the
1392  *	netdevice and manage BPF offload. See definition of
1393  *	enum bpf_netdev_command for details.
1394  * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1395  *			u32 flags);
1396  *	This function is used to submit @n XDP packets for transmit on a
1397  *	netdevice. Returns number of frames successfully transmitted, frames
1398  *	that got dropped are freed/returned via xdp_return_frame().
1399  *	Returns negative number, means general error invoking ndo, meaning
1400  *	no frames were xmit'ed and core-caller will free all frames.
1401  * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1402  *					        struct xdp_buff *xdp);
1403  *      Get the xmit slave of master device based on the xdp_buff.
1404  * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1405  *      This function is used to wake up the softirq, ksoftirqd or kthread
1406  *	responsible for sending and/or receiving packets on a specific
1407  *	queue id bound to an AF_XDP socket. The flags field specifies if
1408  *	only RX, only Tx, or both should be woken up using the flags
1409  *	XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1410  * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1411  *			 int cmd);
1412  *	Add, change, delete or get information on an IPv4 tunnel.
1413  * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1414  *	If a device is paired with a peer device, return the peer instance.
1415  *	The caller must be under RCU read context.
1416  * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1417  *     Get the forwarding path to reach the real device from the HW destination address
1418  * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1419  *			     const struct skb_shared_hwtstamps *hwtstamps,
1420  *			     bool cycles);
1421  *	Get hardware timestamp based on normal/adjustable time or free running
1422  *	cycle counter. This function is required if physical clock supports a
1423  *	free running cycle counter.
1424  *
1425  * int (*ndo_hwtstamp_get)(struct net_device *dev,
1426  *			   struct kernel_hwtstamp_config *kernel_config);
1427  *	Get the currently configured hardware timestamping parameters for the
1428  *	NIC device.
1429  *
1430  * int (*ndo_hwtstamp_set)(struct net_device *dev,
1431  *			   struct kernel_hwtstamp_config *kernel_config,
1432  *			   struct netlink_ext_ack *extack);
1433  *	Change the hardware timestamping parameters for NIC device.
1434  */
1435 struct net_device_ops {
1436 	int			(*ndo_init)(struct net_device *dev);
1437 	void			(*ndo_uninit)(struct net_device *dev);
1438 	int			(*ndo_open)(struct net_device *dev);
1439 	int			(*ndo_stop)(struct net_device *dev);
1440 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1441 						  struct net_device *dev);
1442 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1443 						      struct net_device *dev,
1444 						      netdev_features_t features);
1445 	u16			(*ndo_select_queue)(struct net_device *dev,
1446 						    struct sk_buff *skb,
1447 						    struct net_device *sb_dev);
1448 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1449 						       int flags);
1450 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1451 	int			(*ndo_set_mac_address)(struct net_device *dev,
1452 						       void *addr);
1453 	int			(*ndo_validate_addr)(struct net_device *dev);
1454 	int			(*ndo_do_ioctl)(struct net_device *dev,
1455 					        struct ifreq *ifr, int cmd);
1456 	int			(*ndo_eth_ioctl)(struct net_device *dev,
1457 						 struct ifreq *ifr, int cmd);
1458 	int			(*ndo_siocbond)(struct net_device *dev,
1459 						struct ifreq *ifr, int cmd);
1460 	int			(*ndo_siocwandev)(struct net_device *dev,
1461 						  struct if_settings *ifs);
1462 	int			(*ndo_siocdevprivate)(struct net_device *dev,
1463 						      struct ifreq *ifr,
1464 						      void __user *data, int cmd);
1465 	int			(*ndo_set_config)(struct net_device *dev,
1466 					          struct ifmap *map);
1467 	int			(*ndo_change_mtu)(struct net_device *dev,
1468 						  int new_mtu);
1469 	int			(*ndo_neigh_setup)(struct net_device *dev,
1470 						   struct neigh_parms *);
1471 	void			(*ndo_tx_timeout) (struct net_device *dev,
1472 						   unsigned int txqueue);
1473 
1474 	void			(*ndo_get_stats64)(struct net_device *dev,
1475 						   struct rtnl_link_stats64 *storage);
1476 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1477 	int			(*ndo_get_offload_stats)(int attr_id,
1478 							 const struct net_device *dev,
1479 							 void *attr_data);
1480 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1481 
1482 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1483 						       __be16 proto, u16 vid);
1484 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1485 						        __be16 proto, u16 vid);
1486 #ifdef CONFIG_NET_POLL_CONTROLLER
1487 	void                    (*ndo_poll_controller)(struct net_device *dev);
1488 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1489 						     struct netpoll_info *info);
1490 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1491 #endif
1492 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1493 						  int queue, u8 *mac);
1494 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1495 						   int queue, u16 vlan,
1496 						   u8 qos, __be16 proto);
1497 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1498 						   int vf, int min_tx_rate,
1499 						   int max_tx_rate);
1500 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1501 						       int vf, bool setting);
1502 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1503 						    int vf, bool setting);
1504 	int			(*ndo_get_vf_config)(struct net_device *dev,
1505 						     int vf,
1506 						     struct ifla_vf_info *ivf);
1507 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1508 							 int vf, int link_state);
1509 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1510 						    int vf,
1511 						    struct ifla_vf_stats
1512 						    *vf_stats);
1513 	int			(*ndo_set_vf_port)(struct net_device *dev,
1514 						   int vf,
1515 						   struct nlattr *port[]);
1516 	int			(*ndo_get_vf_port)(struct net_device *dev,
1517 						   int vf, struct sk_buff *skb);
1518 	int			(*ndo_get_vf_guid)(struct net_device *dev,
1519 						   int vf,
1520 						   struct ifla_vf_guid *node_guid,
1521 						   struct ifla_vf_guid *port_guid);
1522 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1523 						   int vf, u64 guid,
1524 						   int guid_type);
1525 	int			(*ndo_set_vf_rss_query_en)(
1526 						   struct net_device *dev,
1527 						   int vf, bool setting);
1528 	int			(*ndo_setup_tc)(struct net_device *dev,
1529 						enum tc_setup_type type,
1530 						void *type_data);
1531 #if IS_ENABLED(CONFIG_FCOE)
1532 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1533 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1534 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1535 						      u16 xid,
1536 						      struct scatterlist *sgl,
1537 						      unsigned int sgc);
1538 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1539 						     u16 xid);
1540 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1541 						       u16 xid,
1542 						       struct scatterlist *sgl,
1543 						       unsigned int sgc);
1544 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1545 							struct netdev_fcoe_hbainfo *hbainfo);
1546 #endif
1547 
1548 #if IS_ENABLED(CONFIG_LIBFCOE)
1549 #define NETDEV_FCOE_WWNN 0
1550 #define NETDEV_FCOE_WWPN 1
1551 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1552 						    u64 *wwn, int type);
1553 #endif
1554 
1555 #ifdef CONFIG_RFS_ACCEL
1556 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1557 						     const struct sk_buff *skb,
1558 						     u16 rxq_index,
1559 						     u32 flow_id);
1560 #endif
1561 	int			(*ndo_add_slave)(struct net_device *dev,
1562 						 struct net_device *slave_dev,
1563 						 struct netlink_ext_ack *extack);
1564 	int			(*ndo_del_slave)(struct net_device *dev,
1565 						 struct net_device *slave_dev);
1566 	struct net_device*	(*ndo_get_xmit_slave)(struct net_device *dev,
1567 						      struct sk_buff *skb,
1568 						      bool all_slaves);
1569 	struct net_device*	(*ndo_sk_get_lower_dev)(struct net_device *dev,
1570 							struct sock *sk);
1571 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1572 						    netdev_features_t features);
1573 	int			(*ndo_set_features)(struct net_device *dev,
1574 						    netdev_features_t features);
1575 	int			(*ndo_neigh_construct)(struct net_device *dev,
1576 						       struct neighbour *n);
1577 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1578 						     struct neighbour *n);
1579 
1580 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1581 					       struct nlattr *tb[],
1582 					       struct net_device *dev,
1583 					       const unsigned char *addr,
1584 					       u16 vid,
1585 					       u16 flags,
1586 					       struct netlink_ext_ack *extack);
1587 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1588 					       struct nlattr *tb[],
1589 					       struct net_device *dev,
1590 					       const unsigned char *addr,
1591 					       u16 vid, struct netlink_ext_ack *extack);
1592 	int			(*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1593 						    struct net_device *dev,
1594 						    struct netlink_ext_ack *extack);
1595 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1596 						struct netlink_callback *cb,
1597 						struct net_device *dev,
1598 						struct net_device *filter_dev,
1599 						int *idx);
1600 	int			(*ndo_fdb_get)(struct sk_buff *skb,
1601 					       struct nlattr *tb[],
1602 					       struct net_device *dev,
1603 					       const unsigned char *addr,
1604 					       u16 vid, u32 portid, u32 seq,
1605 					       struct netlink_ext_ack *extack);
1606 	int			(*ndo_mdb_add)(struct net_device *dev,
1607 					       struct nlattr *tb[],
1608 					       u16 nlmsg_flags,
1609 					       struct netlink_ext_ack *extack);
1610 	int			(*ndo_mdb_del)(struct net_device *dev,
1611 					       struct nlattr *tb[],
1612 					       struct netlink_ext_ack *extack);
1613 	int			(*ndo_mdb_dump)(struct net_device *dev,
1614 						struct sk_buff *skb,
1615 						struct netlink_callback *cb);
1616 	int			(*ndo_mdb_get)(struct net_device *dev,
1617 					       struct nlattr *tb[], u32 portid,
1618 					       u32 seq,
1619 					       struct netlink_ext_ack *extack);
1620 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1621 						      struct nlmsghdr *nlh,
1622 						      u16 flags,
1623 						      struct netlink_ext_ack *extack);
1624 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1625 						      u32 pid, u32 seq,
1626 						      struct net_device *dev,
1627 						      u32 filter_mask,
1628 						      int nlflags);
1629 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1630 						      struct nlmsghdr *nlh,
1631 						      u16 flags);
1632 	int			(*ndo_change_carrier)(struct net_device *dev,
1633 						      bool new_carrier);
1634 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1635 							struct netdev_phys_item_id *ppid);
1636 	int			(*ndo_get_port_parent_id)(struct net_device *dev,
1637 							  struct netdev_phys_item_id *ppid);
1638 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1639 							  char *name, size_t len);
1640 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1641 							struct net_device *dev);
1642 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1643 							void *priv);
1644 
1645 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1646 						      int queue_index,
1647 						      u32 maxrate);
1648 	int			(*ndo_get_iflink)(const struct net_device *dev);
1649 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1650 						       struct sk_buff *skb);
1651 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1652 						       int needed_headroom);
1653 	int			(*ndo_bpf)(struct net_device *dev,
1654 					   struct netdev_bpf *bpf);
1655 	int			(*ndo_xdp_xmit)(struct net_device *dev, int n,
1656 						struct xdp_frame **xdp,
1657 						u32 flags);
1658 	struct net_device *	(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1659 							  struct xdp_buff *xdp);
1660 	int			(*ndo_xsk_wakeup)(struct net_device *dev,
1661 						  u32 queue_id, u32 flags);
1662 	int			(*ndo_tunnel_ctl)(struct net_device *dev,
1663 						  struct ip_tunnel_parm *p, int cmd);
1664 	struct net_device *	(*ndo_get_peer_dev)(struct net_device *dev);
1665 	int                     (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1666                                                          struct net_device_path *path);
1667 	ktime_t			(*ndo_get_tstamp)(struct net_device *dev,
1668 						  const struct skb_shared_hwtstamps *hwtstamps,
1669 						  bool cycles);
1670 	int			(*ndo_hwtstamp_get)(struct net_device *dev,
1671 						    struct kernel_hwtstamp_config *kernel_config);
1672 	int			(*ndo_hwtstamp_set)(struct net_device *dev,
1673 						    struct kernel_hwtstamp_config *kernel_config,
1674 						    struct netlink_ext_ack *extack);
1675 };
1676 
1677 /**
1678  * enum netdev_priv_flags - &struct net_device priv_flags
1679  *
1680  * These are the &struct net_device, they are only set internally
1681  * by drivers and used in the kernel. These flags are invisible to
1682  * userspace; this means that the order of these flags can change
1683  * during any kernel release.
1684  *
1685  * You should have a pretty good reason to be extending these flags.
1686  *
1687  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1688  * @IFF_EBRIDGE: Ethernet bridging device
1689  * @IFF_BONDING: bonding master or slave
1690  * @IFF_ISATAP: ISATAP interface (RFC4214)
1691  * @IFF_WAN_HDLC: WAN HDLC device
1692  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1693  *	release skb->dst
1694  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1695  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1696  * @IFF_MACVLAN_PORT: device used as macvlan port
1697  * @IFF_BRIDGE_PORT: device used as bridge port
1698  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1699  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1700  * @IFF_UNICAST_FLT: Supports unicast filtering
1701  * @IFF_TEAM_PORT: device used as team port
1702  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1703  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1704  *	change when it's running
1705  * @IFF_MACVLAN: Macvlan device
1706  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1707  *	underlying stacked devices
1708  * @IFF_L3MDEV_MASTER: device is an L3 master device
1709  * @IFF_NO_QUEUE: device can run without qdisc attached
1710  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1711  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1712  * @IFF_TEAM: device is a team device
1713  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1714  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1715  *	entity (i.e. the master device for bridged veth)
1716  * @IFF_MACSEC: device is a MACsec device
1717  * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1718  * @IFF_FAILOVER: device is a failover master device
1719  * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1720  * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1721  * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1722  * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1723  *	skb_headlen(skb) == 0 (data starts from frag0)
1724  * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1725  * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1726  *	ndo_hwtstamp_set() for all timestamp requests regardless of source,
1727  *	even if those aren't HWTSTAMP_SOURCE_NETDEV.
1728  */
1729 enum netdev_priv_flags {
1730 	IFF_802_1Q_VLAN			= 1<<0,
1731 	IFF_EBRIDGE			= 1<<1,
1732 	IFF_BONDING			= 1<<2,
1733 	IFF_ISATAP			= 1<<3,
1734 	IFF_WAN_HDLC			= 1<<4,
1735 	IFF_XMIT_DST_RELEASE		= 1<<5,
1736 	IFF_DONT_BRIDGE			= 1<<6,
1737 	IFF_DISABLE_NETPOLL		= 1<<7,
1738 	IFF_MACVLAN_PORT		= 1<<8,
1739 	IFF_BRIDGE_PORT			= 1<<9,
1740 	IFF_OVS_DATAPATH		= 1<<10,
1741 	IFF_TX_SKB_SHARING		= 1<<11,
1742 	IFF_UNICAST_FLT			= 1<<12,
1743 	IFF_TEAM_PORT			= 1<<13,
1744 	IFF_SUPP_NOFCS			= 1<<14,
1745 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1746 	IFF_MACVLAN			= 1<<16,
1747 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1748 	IFF_L3MDEV_MASTER		= 1<<18,
1749 	IFF_NO_QUEUE			= 1<<19,
1750 	IFF_OPENVSWITCH			= 1<<20,
1751 	IFF_L3MDEV_SLAVE		= 1<<21,
1752 	IFF_TEAM			= 1<<22,
1753 	IFF_RXFH_CONFIGURED		= 1<<23,
1754 	IFF_PHONY_HEADROOM		= 1<<24,
1755 	IFF_MACSEC			= 1<<25,
1756 	IFF_NO_RX_HANDLER		= 1<<26,
1757 	IFF_FAILOVER			= 1<<27,
1758 	IFF_FAILOVER_SLAVE		= 1<<28,
1759 	IFF_L3MDEV_RX_HANDLER		= 1<<29,
1760 	IFF_NO_ADDRCONF			= BIT_ULL(30),
1761 	IFF_TX_SKB_NO_LINEAR		= BIT_ULL(31),
1762 	IFF_CHANGE_PROTO_DOWN		= BIT_ULL(32),
1763 	IFF_SEE_ALL_HWTSTAMP_REQUESTS	= BIT_ULL(33),
1764 };
1765 
1766 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1767 #define IFF_EBRIDGE			IFF_EBRIDGE
1768 #define IFF_BONDING			IFF_BONDING
1769 #define IFF_ISATAP			IFF_ISATAP
1770 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1771 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1772 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1773 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1774 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1775 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1776 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1777 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1778 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1779 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1780 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1781 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1782 #define IFF_MACVLAN			IFF_MACVLAN
1783 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1784 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1785 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1786 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1787 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1788 #define IFF_TEAM			IFF_TEAM
1789 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1790 #define IFF_PHONY_HEADROOM		IFF_PHONY_HEADROOM
1791 #define IFF_MACSEC			IFF_MACSEC
1792 #define IFF_NO_RX_HANDLER		IFF_NO_RX_HANDLER
1793 #define IFF_FAILOVER			IFF_FAILOVER
1794 #define IFF_FAILOVER_SLAVE		IFF_FAILOVER_SLAVE
1795 #define IFF_L3MDEV_RX_HANDLER		IFF_L3MDEV_RX_HANDLER
1796 #define IFF_TX_SKB_NO_LINEAR		IFF_TX_SKB_NO_LINEAR
1797 
1798 /* Specifies the type of the struct net_device::ml_priv pointer */
1799 enum netdev_ml_priv_type {
1800 	ML_PRIV_NONE,
1801 	ML_PRIV_CAN,
1802 };
1803 
1804 enum netdev_stat_type {
1805 	NETDEV_PCPU_STAT_NONE,
1806 	NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1807 	NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1808 	NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1809 };
1810 
1811 /**
1812  *	struct net_device - The DEVICE structure.
1813  *
1814  *	Actually, this whole structure is a big mistake.  It mixes I/O
1815  *	data with strictly "high-level" data, and it has to know about
1816  *	almost every data structure used in the INET module.
1817  *
1818  *	@name:	This is the first field of the "visible" part of this structure
1819  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1820  *		of the interface.
1821  *
1822  *	@name_node:	Name hashlist node
1823  *	@ifalias:	SNMP alias
1824  *	@mem_end:	Shared memory end
1825  *	@mem_start:	Shared memory start
1826  *	@base_addr:	Device I/O address
1827  *	@irq:		Device IRQ number
1828  *
1829  *	@state:		Generic network queuing layer state, see netdev_state_t
1830  *	@dev_list:	The global list of network devices
1831  *	@napi_list:	List entry used for polling NAPI devices
1832  *	@unreg_list:	List entry  when we are unregistering the
1833  *			device; see the function unregister_netdev
1834  *	@close_list:	List entry used when we are closing the device
1835  *	@ptype_all:     Device-specific packet handlers for all protocols
1836  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1837  *
1838  *	@adj_list:	Directly linked devices, like slaves for bonding
1839  *	@features:	Currently active device features
1840  *	@hw_features:	User-changeable features
1841  *
1842  *	@wanted_features:	User-requested features
1843  *	@vlan_features:		Mask of features inheritable by VLAN devices
1844  *
1845  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1846  *				This field indicates what encapsulation
1847  *				offloads the hardware is capable of doing,
1848  *				and drivers will need to set them appropriately.
1849  *
1850  *	@mpls_features:	Mask of features inheritable by MPLS
1851  *	@gso_partial_features: value(s) from NETIF_F_GSO\*
1852  *
1853  *	@ifindex:	interface index
1854  *	@group:		The group the device belongs to
1855  *
1856  *	@stats:		Statistics struct, which was left as a legacy, use
1857  *			rtnl_link_stats64 instead
1858  *
1859  *	@core_stats:	core networking counters,
1860  *			do not use this in drivers
1861  *	@carrier_up_count:	Number of times the carrier has been up
1862  *	@carrier_down_count:	Number of times the carrier has been down
1863  *
1864  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1865  *				instead of ioctl,
1866  *				see <net/iw_handler.h> for details.
1867  *	@wireless_data:	Instance data managed by the core of wireless extensions
1868  *
1869  *	@netdev_ops:	Includes several pointers to callbacks,
1870  *			if one wants to override the ndo_*() functions
1871  *	@xdp_metadata_ops:	Includes pointers to XDP metadata callbacks.
1872  *	@xsk_tx_metadata_ops:	Includes pointers to AF_XDP TX metadata callbacks.
1873  *	@ethtool_ops:	Management operations
1874  *	@l3mdev_ops:	Layer 3 master device operations
1875  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1876  *			discovery handling. Necessary for e.g. 6LoWPAN.
1877  *	@xfrmdev_ops:	Transformation offload operations
1878  *	@tlsdev_ops:	Transport Layer Security offload operations
1879  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1880  *			of Layer 2 headers.
1881  *
1882  *	@flags:		Interface flags (a la BSD)
1883  *	@xdp_features:	XDP capability supported by the device
1884  *	@priv_flags:	Like 'flags' but invisible to userspace,
1885  *			see if.h for the definitions
1886  *	@gflags:	Global flags ( kept as legacy )
1887  *	@padded:	How much padding added by alloc_netdev()
1888  *	@operstate:	RFC2863 operstate
1889  *	@link_mode:	Mapping policy to operstate
1890  *	@if_port:	Selectable AUI, TP, ...
1891  *	@dma:		DMA channel
1892  *	@mtu:		Interface MTU value
1893  *	@min_mtu:	Interface Minimum MTU value
1894  *	@max_mtu:	Interface Maximum MTU value
1895  *	@type:		Interface hardware type
1896  *	@hard_header_len: Maximum hardware header length.
1897  *	@min_header_len:  Minimum hardware header length
1898  *
1899  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1900  *			  cases can this be guaranteed
1901  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1902  *			  cases can this be guaranteed. Some cases also use
1903  *			  LL_MAX_HEADER instead to allocate the skb
1904  *
1905  *	interface address info:
1906  *
1907  * 	@perm_addr:		Permanent hw address
1908  * 	@addr_assign_type:	Hw address assignment type
1909  * 	@addr_len:		Hardware address length
1910  *	@upper_level:		Maximum depth level of upper devices.
1911  *	@lower_level:		Maximum depth level of lower devices.
1912  *	@neigh_priv_len:	Used in neigh_alloc()
1913  * 	@dev_id:		Used to differentiate devices that share
1914  * 				the same link layer address
1915  * 	@dev_port:		Used to differentiate devices that share
1916  * 				the same function
1917  *	@addr_list_lock:	XXX: need comments on this one
1918  *	@name_assign_type:	network interface name assignment type
1919  *	@uc_promisc:		Counter that indicates promiscuous mode
1920  *				has been enabled due to the need to listen to
1921  *				additional unicast addresses in a device that
1922  *				does not implement ndo_set_rx_mode()
1923  *	@uc:			unicast mac addresses
1924  *	@mc:			multicast mac addresses
1925  *	@dev_addrs:		list of device hw addresses
1926  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1927  *	@promiscuity:		Number of times the NIC is told to work in
1928  *				promiscuous mode; if it becomes 0 the NIC will
1929  *				exit promiscuous mode
1930  *	@allmulti:		Counter, enables or disables allmulticast mode
1931  *
1932  *	@vlan_info:	VLAN info
1933  *	@dsa_ptr:	dsa specific data
1934  *	@tipc_ptr:	TIPC specific data
1935  *	@atalk_ptr:	AppleTalk link
1936  *	@ip_ptr:	IPv4 specific data
1937  *	@ip6_ptr:	IPv6 specific data
1938  *	@ax25_ptr:	AX.25 specific data
1939  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1940  *	@ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1941  *			 device struct
1942  *	@mpls_ptr:	mpls_dev struct pointer
1943  *	@mctp_ptr:	MCTP specific data
1944  *
1945  *	@dev_addr:	Hw address (before bcast,
1946  *			because most packets are unicast)
1947  *
1948  *	@_rx:			Array of RX queues
1949  *	@num_rx_queues:		Number of RX queues
1950  *				allocated at register_netdev() time
1951  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1952  *	@xdp_prog:		XDP sockets filter program pointer
1953  *	@gro_flush_timeout:	timeout for GRO layer in NAPI
1954  *	@napi_defer_hard_irqs:	If not zero, provides a counter that would
1955  *				allow to avoid NIC hard IRQ, on busy queues.
1956  *
1957  *	@rx_handler:		handler for received packets
1958  *	@rx_handler_data: 	XXX: need comments on this one
1959  *	@tcx_ingress:		BPF & clsact qdisc specific data for ingress processing
1960  *	@ingress_queue:		XXX: need comments on this one
1961  *	@nf_hooks_ingress:	netfilter hooks executed for ingress packets
1962  *	@broadcast:		hw bcast address
1963  *
1964  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1965  *			indexed by RX queue number. Assigned by driver.
1966  *			This must only be set if the ndo_rx_flow_steer
1967  *			operation is defined
1968  *	@index_hlist:		Device index hash chain
1969  *
1970  *	@_tx:			Array of TX queues
1971  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1972  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1973  *	@qdisc:			Root qdisc from userspace point of view
1974  *	@tx_queue_len:		Max frames per queue allowed
1975  *	@tx_global_lock: 	XXX: need comments on this one
1976  *	@xdp_bulkq:		XDP device bulk queue
1977  *	@xps_maps:		all CPUs/RXQs maps for XPS device
1978  *
1979  *	@xps_maps:	XXX: need comments on this one
1980  *	@tcx_egress:		BPF & clsact qdisc specific data for egress processing
1981  *	@nf_hooks_egress:	netfilter hooks executed for egress packets
1982  *	@qdisc_hash:		qdisc hash table
1983  *	@watchdog_timeo:	Represents the timeout that is used by
1984  *				the watchdog (see dev_watchdog())
1985  *	@watchdog_timer:	List of timers
1986  *
1987  *	@proto_down_reason:	reason a netdev interface is held down
1988  *	@pcpu_refcnt:		Number of references to this device
1989  *	@dev_refcnt:		Number of references to this device
1990  *	@refcnt_tracker:	Tracker directory for tracked references to this device
1991  *	@todo_list:		Delayed register/unregister
1992  *	@link_watch_list:	XXX: need comments on this one
1993  *
1994  *	@reg_state:		Register/unregister state machine
1995  *	@dismantle:		Device is going to be freed
1996  *	@rtnl_link_state:	This enum represents the phases of creating
1997  *				a new link
1998  *
1999  *	@needs_free_netdev:	Should unregister perform free_netdev?
2000  *	@priv_destructor:	Called from unregister
2001  *	@npinfo:		XXX: need comments on this one
2002  * 	@nd_net:		Network namespace this network device is inside
2003  *
2004  * 	@ml_priv:	Mid-layer private
2005  *	@ml_priv_type:  Mid-layer private type
2006  *
2007  *	@pcpu_stat_type:	Type of device statistics which the core should
2008  *				allocate/free: none, lstats, tstats, dstats. none
2009  *				means the driver is handling statistics allocation/
2010  *				freeing internally.
2011  *	@lstats:		Loopback statistics: packets, bytes
2012  *	@tstats:		Tunnel statistics: RX/TX packets, RX/TX bytes
2013  *	@dstats:		Dummy statistics: RX/TX/drop packets, RX/TX bytes
2014  *
2015  *	@garp_port:	GARP
2016  *	@mrp_port:	MRP
2017  *
2018  *	@dm_private:	Drop monitor private
2019  *
2020  *	@dev:		Class/net/name entry
2021  *	@sysfs_groups:	Space for optional device, statistics and wireless
2022  *			sysfs groups
2023  *
2024  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
2025  *	@rtnl_link_ops:	Rtnl_link_ops
2026  *
2027  *	@gso_max_size:	Maximum size of generic segmentation offload
2028  *	@tso_max_size:	Device (as in HW) limit on the max TSO request size
2029  *	@gso_max_segs:	Maximum number of segments that can be passed to the
2030  *			NIC for GSO
2031  *	@tso_max_segs:	Device (as in HW) limit on the max TSO segment count
2032  * 	@gso_ipv4_max_size:	Maximum size of generic segmentation offload,
2033  * 				for IPv4.
2034  *
2035  *	@dcbnl_ops:	Data Center Bridging netlink ops
2036  *	@num_tc:	Number of traffic classes in the net device
2037  *	@tc_to_txq:	XXX: need comments on this one
2038  *	@prio_tc_map:	XXX: need comments on this one
2039  *
2040  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
2041  *
2042  *	@priomap:	XXX: need comments on this one
2043  *	@phydev:	Physical device may attach itself
2044  *			for hardware timestamping
2045  *	@sfp_bus:	attached &struct sfp_bus structure.
2046  *
2047  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2048  *
2049  *	@proto_down:	protocol port state information can be sent to the
2050  *			switch driver and used to set the phys state of the
2051  *			switch port.
2052  *
2053  *	@wol_enabled:	Wake-on-LAN is enabled
2054  *
2055  *	@threaded:	napi threaded mode is enabled
2056  *
2057  *	@net_notifier_list:	List of per-net netdev notifier block
2058  *				that follow this device when it is moved
2059  *				to another network namespace.
2060  *
2061  *	@macsec_ops:    MACsec offloading ops
2062  *
2063  *	@udp_tunnel_nic_info:	static structure describing the UDP tunnel
2064  *				offload capabilities of the device
2065  *	@udp_tunnel_nic:	UDP tunnel offload state
2066  *	@xdp_state:		stores info on attached XDP BPF programs
2067  *
2068  *	@nested_level:	Used as a parameter of spin_lock_nested() of
2069  *			dev->addr_list_lock.
2070  *	@unlink_list:	As netif_addr_lock() can be called recursively,
2071  *			keep a list of interfaces to be deleted.
2072  *	@gro_max_size:	Maximum size of aggregated packet in generic
2073  *			receive offload (GRO)
2074  * 	@gro_ipv4_max_size:	Maximum size of aggregated packet in generic
2075  * 				receive offload (GRO), for IPv4.
2076  *	@xdp_zc_max_segs:	Maximum number of segments supported by AF_XDP
2077  *				zero copy driver
2078  *
2079  *	@dev_addr_shadow:	Copy of @dev_addr to catch direct writes.
2080  *	@linkwatch_dev_tracker:	refcount tracker used by linkwatch.
2081  *	@watchdog_dev_tracker:	refcount tracker used by watchdog.
2082  *	@dev_registered_tracker:	tracker for reference held while
2083  *					registered
2084  *	@offload_xstats_l3:	L3 HW stats for this netdevice.
2085  *
2086  *	@devlink_port:	Pointer to related devlink port structure.
2087  *			Assigned by a driver before netdev registration using
2088  *			SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2089  *			during the time netdevice is registered.
2090  *
2091  *	@dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2092  *		   where the clock is recovered.
2093  *
2094  *	FIXME: cleanup struct net_device such that network protocol info
2095  *	moves out.
2096  */
2097 
2098 struct net_device {
2099 	char			name[IFNAMSIZ];
2100 	struct netdev_name_node	*name_node;
2101 	struct dev_ifalias	__rcu *ifalias;
2102 	/*
2103 	 *	I/O specific fields
2104 	 *	FIXME: Merge these and struct ifmap into one
2105 	 */
2106 	unsigned long		mem_end;
2107 	unsigned long		mem_start;
2108 	unsigned long		base_addr;
2109 
2110 	/*
2111 	 *	Some hardware also needs these fields (state,dev_list,
2112 	 *	napi_list,unreg_list,close_list) but they are not
2113 	 *	part of the usual set specified in Space.c.
2114 	 */
2115 
2116 	unsigned long		state;
2117 
2118 	struct list_head	dev_list;
2119 	struct list_head	napi_list;
2120 	struct list_head	unreg_list;
2121 	struct list_head	close_list;
2122 	struct list_head	ptype_all;
2123 	struct list_head	ptype_specific;
2124 
2125 	struct {
2126 		struct list_head upper;
2127 		struct list_head lower;
2128 	} adj_list;
2129 
2130 	/* Read-mostly cache-line for fast-path access */
2131 	unsigned int		flags;
2132 	xdp_features_t		xdp_features;
2133 	unsigned long long	priv_flags;
2134 	const struct net_device_ops *netdev_ops;
2135 	const struct xdp_metadata_ops *xdp_metadata_ops;
2136 	const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2137 	int			ifindex;
2138 	unsigned short		gflags;
2139 	unsigned short		hard_header_len;
2140 
2141 	/* Note : dev->mtu is often read without holding a lock.
2142 	 * Writers usually hold RTNL.
2143 	 * It is recommended to use READ_ONCE() to annotate the reads,
2144 	 * and to use WRITE_ONCE() to annotate the writes.
2145 	 */
2146 	unsigned int		mtu;
2147 	unsigned short		needed_headroom;
2148 	unsigned short		needed_tailroom;
2149 
2150 	netdev_features_t	features;
2151 	netdev_features_t	hw_features;
2152 	netdev_features_t	wanted_features;
2153 	netdev_features_t	vlan_features;
2154 	netdev_features_t	hw_enc_features;
2155 	netdev_features_t	mpls_features;
2156 	netdev_features_t	gso_partial_features;
2157 
2158 	unsigned int		min_mtu;
2159 	unsigned int		max_mtu;
2160 	unsigned short		type;
2161 	unsigned char		min_header_len;
2162 	unsigned char		name_assign_type;
2163 
2164 	int			group;
2165 
2166 	struct net_device_stats	stats; /* not used by modern drivers */
2167 
2168 	struct net_device_core_stats __percpu *core_stats;
2169 
2170 	/* Stats to monitor link on/off, flapping */
2171 	atomic_t		carrier_up_count;
2172 	atomic_t		carrier_down_count;
2173 
2174 #ifdef CONFIG_WIRELESS_EXT
2175 	const struct iw_handler_def *wireless_handlers;
2176 	struct iw_public_data	*wireless_data;
2177 #endif
2178 	const struct ethtool_ops *ethtool_ops;
2179 #ifdef CONFIG_NET_L3_MASTER_DEV
2180 	const struct l3mdev_ops	*l3mdev_ops;
2181 #endif
2182 #if IS_ENABLED(CONFIG_IPV6)
2183 	const struct ndisc_ops *ndisc_ops;
2184 #endif
2185 
2186 #ifdef CONFIG_XFRM_OFFLOAD
2187 	const struct xfrmdev_ops *xfrmdev_ops;
2188 #endif
2189 
2190 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2191 	const struct tlsdev_ops *tlsdev_ops;
2192 #endif
2193 
2194 	const struct header_ops *header_ops;
2195 
2196 	unsigned char		operstate;
2197 	unsigned char		link_mode;
2198 
2199 	unsigned char		if_port;
2200 	unsigned char		dma;
2201 
2202 	/* Interface address info. */
2203 	unsigned char		perm_addr[MAX_ADDR_LEN];
2204 	unsigned char		addr_assign_type;
2205 	unsigned char		addr_len;
2206 	unsigned char		upper_level;
2207 	unsigned char		lower_level;
2208 
2209 	unsigned short		neigh_priv_len;
2210 	unsigned short          dev_id;
2211 	unsigned short          dev_port;
2212 	unsigned short		padded;
2213 
2214 	spinlock_t		addr_list_lock;
2215 	int			irq;
2216 
2217 	struct netdev_hw_addr_list	uc;
2218 	struct netdev_hw_addr_list	mc;
2219 	struct netdev_hw_addr_list	dev_addrs;
2220 
2221 #ifdef CONFIG_SYSFS
2222 	struct kset		*queues_kset;
2223 #endif
2224 #ifdef CONFIG_LOCKDEP
2225 	struct list_head	unlink_list;
2226 #endif
2227 	unsigned int		promiscuity;
2228 	unsigned int		allmulti;
2229 	bool			uc_promisc;
2230 #ifdef CONFIG_LOCKDEP
2231 	unsigned char		nested_level;
2232 #endif
2233 
2234 
2235 	/* Protocol-specific pointers */
2236 
2237 	struct in_device __rcu	*ip_ptr;
2238 	struct inet6_dev __rcu	*ip6_ptr;
2239 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2240 	struct vlan_info __rcu	*vlan_info;
2241 #endif
2242 #if IS_ENABLED(CONFIG_NET_DSA)
2243 	struct dsa_port		*dsa_ptr;
2244 #endif
2245 #if IS_ENABLED(CONFIG_TIPC)
2246 	struct tipc_bearer __rcu *tipc_ptr;
2247 #endif
2248 #if IS_ENABLED(CONFIG_ATALK)
2249 	void 			*atalk_ptr;
2250 #endif
2251 #if IS_ENABLED(CONFIG_AX25)
2252 	void			*ax25_ptr;
2253 #endif
2254 #if IS_ENABLED(CONFIG_CFG80211)
2255 	struct wireless_dev	*ieee80211_ptr;
2256 #endif
2257 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2258 	struct wpan_dev		*ieee802154_ptr;
2259 #endif
2260 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2261 	struct mpls_dev __rcu	*mpls_ptr;
2262 #endif
2263 #if IS_ENABLED(CONFIG_MCTP)
2264 	struct mctp_dev __rcu	*mctp_ptr;
2265 #endif
2266 
2267 /*
2268  * Cache lines mostly used on receive path (including eth_type_trans())
2269  */
2270 	/* Interface address info used in eth_type_trans() */
2271 	const unsigned char	*dev_addr;
2272 
2273 	struct netdev_rx_queue	*_rx;
2274 	unsigned int		num_rx_queues;
2275 	unsigned int		real_num_rx_queues;
2276 
2277 	struct bpf_prog __rcu	*xdp_prog;
2278 	unsigned long		gro_flush_timeout;
2279 	int			napi_defer_hard_irqs;
2280 #define GRO_LEGACY_MAX_SIZE	65536u
2281 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2282  * and shinfo->gso_segs is a 16bit field.
2283  */
2284 #define GRO_MAX_SIZE		(8 * 65535u)
2285 	unsigned int		gro_max_size;
2286 	unsigned int		gro_ipv4_max_size;
2287 	unsigned int		xdp_zc_max_segs;
2288 	rx_handler_func_t __rcu	*rx_handler;
2289 	void __rcu		*rx_handler_data;
2290 #ifdef CONFIG_NET_XGRESS
2291 	struct bpf_mprog_entry __rcu *tcx_ingress;
2292 #endif
2293 	struct netdev_queue __rcu *ingress_queue;
2294 #ifdef CONFIG_NETFILTER_INGRESS
2295 	struct nf_hook_entries __rcu *nf_hooks_ingress;
2296 #endif
2297 
2298 	unsigned char		broadcast[MAX_ADDR_LEN];
2299 #ifdef CONFIG_RFS_ACCEL
2300 	struct cpu_rmap		*rx_cpu_rmap;
2301 #endif
2302 	struct hlist_node	index_hlist;
2303 
2304 /*
2305  * Cache lines mostly used on transmit path
2306  */
2307 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
2308 	unsigned int		num_tx_queues;
2309 	unsigned int		real_num_tx_queues;
2310 	struct Qdisc __rcu	*qdisc;
2311 	unsigned int		tx_queue_len;
2312 	spinlock_t		tx_global_lock;
2313 
2314 	struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2315 
2316 #ifdef CONFIG_XPS
2317 	struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2318 #endif
2319 #ifdef CONFIG_NET_XGRESS
2320 	struct bpf_mprog_entry __rcu *tcx_egress;
2321 #endif
2322 #ifdef CONFIG_NETFILTER_EGRESS
2323 	struct nf_hook_entries __rcu *nf_hooks_egress;
2324 #endif
2325 
2326 #ifdef CONFIG_NET_SCHED
2327 	DECLARE_HASHTABLE	(qdisc_hash, 4);
2328 #endif
2329 	/* These may be needed for future network-power-down code. */
2330 	struct timer_list	watchdog_timer;
2331 	int			watchdog_timeo;
2332 
2333 	u32                     proto_down_reason;
2334 
2335 	struct list_head	todo_list;
2336 
2337 #ifdef CONFIG_PCPU_DEV_REFCNT
2338 	int __percpu		*pcpu_refcnt;
2339 #else
2340 	refcount_t		dev_refcnt;
2341 #endif
2342 	struct ref_tracker_dir	refcnt_tracker;
2343 
2344 	struct list_head	link_watch_list;
2345 
2346 	enum { NETREG_UNINITIALIZED=0,
2347 	       NETREG_REGISTERED,	/* completed register_netdevice */
2348 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
2349 	       NETREG_UNREGISTERED,	/* completed unregister todo */
2350 	       NETREG_RELEASED,		/* called free_netdev */
2351 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
2352 	} reg_state:8;
2353 
2354 	bool dismantle;
2355 
2356 	enum {
2357 		RTNL_LINK_INITIALIZED,
2358 		RTNL_LINK_INITIALIZING,
2359 	} rtnl_link_state:16;
2360 
2361 	bool needs_free_netdev;
2362 	void (*priv_destructor)(struct net_device *dev);
2363 
2364 #ifdef CONFIG_NETPOLL
2365 	struct netpoll_info __rcu	*npinfo;
2366 #endif
2367 
2368 	possible_net_t			nd_net;
2369 
2370 	/* mid-layer private */
2371 	void				*ml_priv;
2372 	enum netdev_ml_priv_type	ml_priv_type;
2373 
2374 	enum netdev_stat_type		pcpu_stat_type:8;
2375 	union {
2376 		struct pcpu_lstats __percpu		*lstats;
2377 		struct pcpu_sw_netstats __percpu	*tstats;
2378 		struct pcpu_dstats __percpu		*dstats;
2379 	};
2380 
2381 #if IS_ENABLED(CONFIG_GARP)
2382 	struct garp_port __rcu	*garp_port;
2383 #endif
2384 #if IS_ENABLED(CONFIG_MRP)
2385 	struct mrp_port __rcu	*mrp_port;
2386 #endif
2387 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2388 	struct dm_hw_stat_delta __rcu *dm_private;
2389 #endif
2390 	struct device		dev;
2391 	const struct attribute_group *sysfs_groups[4];
2392 	const struct attribute_group *sysfs_rx_queue_group;
2393 
2394 	const struct rtnl_link_ops *rtnl_link_ops;
2395 
2396 	/* for setting kernel sock attribute on TCP connection setup */
2397 #define GSO_MAX_SEGS		65535u
2398 #define GSO_LEGACY_MAX_SIZE	65536u
2399 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2400  * and shinfo->gso_segs is a 16bit field.
2401  */
2402 #define GSO_MAX_SIZE		(8 * GSO_MAX_SEGS)
2403 
2404 	unsigned int		gso_max_size;
2405 #define TSO_LEGACY_MAX_SIZE	65536
2406 #define TSO_MAX_SIZE		UINT_MAX
2407 	unsigned int		tso_max_size;
2408 	u16			gso_max_segs;
2409 #define TSO_MAX_SEGS		U16_MAX
2410 	u16			tso_max_segs;
2411 	unsigned int		gso_ipv4_max_size;
2412 
2413 #ifdef CONFIG_DCB
2414 	const struct dcbnl_rtnl_ops *dcbnl_ops;
2415 #endif
2416 	s16			num_tc;
2417 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
2418 	u8			prio_tc_map[TC_BITMASK + 1];
2419 
2420 #if IS_ENABLED(CONFIG_FCOE)
2421 	unsigned int		fcoe_ddp_xid;
2422 #endif
2423 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2424 	struct netprio_map __rcu *priomap;
2425 #endif
2426 	struct phy_device	*phydev;
2427 	struct sfp_bus		*sfp_bus;
2428 	struct lock_class_key	*qdisc_tx_busylock;
2429 	bool			proto_down;
2430 	unsigned		wol_enabled:1;
2431 	unsigned		threaded:1;
2432 
2433 	struct list_head	net_notifier_list;
2434 
2435 #if IS_ENABLED(CONFIG_MACSEC)
2436 	/* MACsec management functions */
2437 	const struct macsec_ops *macsec_ops;
2438 #endif
2439 	const struct udp_tunnel_nic_info	*udp_tunnel_nic_info;
2440 	struct udp_tunnel_nic	*udp_tunnel_nic;
2441 
2442 	/* protected by rtnl_lock */
2443 	struct bpf_xdp_entity	xdp_state[__MAX_XDP_MODE];
2444 
2445 	u8 dev_addr_shadow[MAX_ADDR_LEN];
2446 	netdevice_tracker	linkwatch_dev_tracker;
2447 	netdevice_tracker	watchdog_dev_tracker;
2448 	netdevice_tracker	dev_registered_tracker;
2449 	struct rtnl_hw_stats64	*offload_xstats_l3;
2450 
2451 	struct devlink_port	*devlink_port;
2452 
2453 #if IS_ENABLED(CONFIG_DPLL)
2454 	struct dpll_pin		*dpll_pin;
2455 #endif
2456 #if IS_ENABLED(CONFIG_PAGE_POOL)
2457 	/** @page_pools: page pools created for this netdevice */
2458 	struct hlist_head	page_pools;
2459 #endif
2460 };
2461 #define to_net_dev(d) container_of(d, struct net_device, dev)
2462 
2463 /*
2464  * Driver should use this to assign devlink port instance to a netdevice
2465  * before it registers the netdevice. Therefore devlink_port is static
2466  * during the netdev lifetime after it is registered.
2467  */
2468 #define SET_NETDEV_DEVLINK_PORT(dev, port)			\
2469 ({								\
2470 	WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED);	\
2471 	((dev)->devlink_port = (port));				\
2472 })
2473 
2474 static inline bool netif_elide_gro(const struct net_device *dev)
2475 {
2476 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2477 		return true;
2478 	return false;
2479 }
2480 
2481 #define	NETDEV_ALIGN		32
2482 
2483 static inline
2484 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2485 {
2486 	return dev->prio_tc_map[prio & TC_BITMASK];
2487 }
2488 
2489 static inline
2490 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2491 {
2492 	if (tc >= dev->num_tc)
2493 		return -EINVAL;
2494 
2495 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2496 	return 0;
2497 }
2498 
2499 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2500 void netdev_reset_tc(struct net_device *dev);
2501 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2502 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2503 
2504 static inline
2505 int netdev_get_num_tc(struct net_device *dev)
2506 {
2507 	return dev->num_tc;
2508 }
2509 
2510 static inline void net_prefetch(void *p)
2511 {
2512 	prefetch(p);
2513 #if L1_CACHE_BYTES < 128
2514 	prefetch((u8 *)p + L1_CACHE_BYTES);
2515 #endif
2516 }
2517 
2518 static inline void net_prefetchw(void *p)
2519 {
2520 	prefetchw(p);
2521 #if L1_CACHE_BYTES < 128
2522 	prefetchw((u8 *)p + L1_CACHE_BYTES);
2523 #endif
2524 }
2525 
2526 void netdev_unbind_sb_channel(struct net_device *dev,
2527 			      struct net_device *sb_dev);
2528 int netdev_bind_sb_channel_queue(struct net_device *dev,
2529 				 struct net_device *sb_dev,
2530 				 u8 tc, u16 count, u16 offset);
2531 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2532 static inline int netdev_get_sb_channel(struct net_device *dev)
2533 {
2534 	return max_t(int, -dev->num_tc, 0);
2535 }
2536 
2537 static inline
2538 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2539 					 unsigned int index)
2540 {
2541 	DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2542 	return &dev->_tx[index];
2543 }
2544 
2545 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2546 						    const struct sk_buff *skb)
2547 {
2548 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2549 }
2550 
2551 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2552 					    void (*f)(struct net_device *,
2553 						      struct netdev_queue *,
2554 						      void *),
2555 					    void *arg)
2556 {
2557 	unsigned int i;
2558 
2559 	for (i = 0; i < dev->num_tx_queues; i++)
2560 		f(dev, &dev->_tx[i], arg);
2561 }
2562 
2563 #define netdev_lockdep_set_classes(dev)				\
2564 {								\
2565 	static struct lock_class_key qdisc_tx_busylock_key;	\
2566 	static struct lock_class_key qdisc_xmit_lock_key;	\
2567 	static struct lock_class_key dev_addr_list_lock_key;	\
2568 	unsigned int i;						\
2569 								\
2570 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
2571 	lockdep_set_class(&(dev)->addr_list_lock,		\
2572 			  &dev_addr_list_lock_key);		\
2573 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
2574 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
2575 				  &qdisc_xmit_lock_key);	\
2576 }
2577 
2578 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2579 		     struct net_device *sb_dev);
2580 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2581 					 struct sk_buff *skb,
2582 					 struct net_device *sb_dev);
2583 
2584 /* returns the headroom that the master device needs to take in account
2585  * when forwarding to this dev
2586  */
2587 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2588 {
2589 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2590 }
2591 
2592 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2593 {
2594 	if (dev->netdev_ops->ndo_set_rx_headroom)
2595 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2596 }
2597 
2598 /* set the device rx headroom to the dev's default */
2599 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2600 {
2601 	netdev_set_rx_headroom(dev, -1);
2602 }
2603 
2604 static inline void *netdev_get_ml_priv(struct net_device *dev,
2605 				       enum netdev_ml_priv_type type)
2606 {
2607 	if (dev->ml_priv_type != type)
2608 		return NULL;
2609 
2610 	return dev->ml_priv;
2611 }
2612 
2613 static inline void netdev_set_ml_priv(struct net_device *dev,
2614 				      void *ml_priv,
2615 				      enum netdev_ml_priv_type type)
2616 {
2617 	WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2618 	     "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2619 	     dev->ml_priv_type, type);
2620 	WARN(!dev->ml_priv_type && dev->ml_priv,
2621 	     "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2622 
2623 	dev->ml_priv = ml_priv;
2624 	dev->ml_priv_type = type;
2625 }
2626 
2627 /*
2628  * Net namespace inlines
2629  */
2630 static inline
2631 struct net *dev_net(const struct net_device *dev)
2632 {
2633 	return read_pnet(&dev->nd_net);
2634 }
2635 
2636 static inline
2637 void dev_net_set(struct net_device *dev, struct net *net)
2638 {
2639 	write_pnet(&dev->nd_net, net);
2640 }
2641 
2642 /**
2643  *	netdev_priv - access network device private data
2644  *	@dev: network device
2645  *
2646  * Get network device private data
2647  */
2648 static inline void *netdev_priv(const struct net_device *dev)
2649 {
2650 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2651 }
2652 
2653 /* Set the sysfs physical device reference for the network logical device
2654  * if set prior to registration will cause a symlink during initialization.
2655  */
2656 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2657 
2658 /* Set the sysfs device type for the network logical device to allow
2659  * fine-grained identification of different network device types. For
2660  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2661  */
2662 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2663 
2664 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2665 			  enum netdev_queue_type type,
2666 			  struct napi_struct *napi);
2667 
2668 /* Default NAPI poll() weight
2669  * Device drivers are strongly advised to not use bigger value
2670  */
2671 #define NAPI_POLL_WEIGHT 64
2672 
2673 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2674 			   int (*poll)(struct napi_struct *, int), int weight);
2675 
2676 /**
2677  * netif_napi_add() - initialize a NAPI context
2678  * @dev:  network device
2679  * @napi: NAPI context
2680  * @poll: polling function
2681  *
2682  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2683  * *any* of the other NAPI-related functions.
2684  */
2685 static inline void
2686 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2687 	       int (*poll)(struct napi_struct *, int))
2688 {
2689 	netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2690 }
2691 
2692 static inline void
2693 netif_napi_add_tx_weight(struct net_device *dev,
2694 			 struct napi_struct *napi,
2695 			 int (*poll)(struct napi_struct *, int),
2696 			 int weight)
2697 {
2698 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2699 	netif_napi_add_weight(dev, napi, poll, weight);
2700 }
2701 
2702 /**
2703  * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2704  * @dev:  network device
2705  * @napi: NAPI context
2706  * @poll: polling function
2707  *
2708  * This variant of netif_napi_add() should be used from drivers using NAPI
2709  * to exclusively poll a TX queue.
2710  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2711  */
2712 static inline void netif_napi_add_tx(struct net_device *dev,
2713 				     struct napi_struct *napi,
2714 				     int (*poll)(struct napi_struct *, int))
2715 {
2716 	netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2717 }
2718 
2719 /**
2720  *  __netif_napi_del - remove a NAPI context
2721  *  @napi: NAPI context
2722  *
2723  * Warning: caller must observe RCU grace period before freeing memory
2724  * containing @napi. Drivers might want to call this helper to combine
2725  * all the needed RCU grace periods into a single one.
2726  */
2727 void __netif_napi_del(struct napi_struct *napi);
2728 
2729 /**
2730  *  netif_napi_del - remove a NAPI context
2731  *  @napi: NAPI context
2732  *
2733  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2734  */
2735 static inline void netif_napi_del(struct napi_struct *napi)
2736 {
2737 	__netif_napi_del(napi);
2738 	synchronize_net();
2739 }
2740 
2741 struct packet_type {
2742 	__be16			type;	/* This is really htons(ether_type). */
2743 	bool			ignore_outgoing;
2744 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2745 	netdevice_tracker	dev_tracker;
2746 	int			(*func) (struct sk_buff *,
2747 					 struct net_device *,
2748 					 struct packet_type *,
2749 					 struct net_device *);
2750 	void			(*list_func) (struct list_head *,
2751 					      struct packet_type *,
2752 					      struct net_device *);
2753 	bool			(*id_match)(struct packet_type *ptype,
2754 					    struct sock *sk);
2755 	struct net		*af_packet_net;
2756 	void			*af_packet_priv;
2757 	struct list_head	list;
2758 };
2759 
2760 struct offload_callbacks {
2761 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2762 						netdev_features_t features);
2763 	struct sk_buff		*(*gro_receive)(struct list_head *head,
2764 						struct sk_buff *skb);
2765 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2766 };
2767 
2768 struct packet_offload {
2769 	__be16			 type;	/* This is really htons(ether_type). */
2770 	u16			 priority;
2771 	struct offload_callbacks callbacks;
2772 	struct list_head	 list;
2773 };
2774 
2775 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2776 struct pcpu_sw_netstats {
2777 	u64_stats_t		rx_packets;
2778 	u64_stats_t		rx_bytes;
2779 	u64_stats_t		tx_packets;
2780 	u64_stats_t		tx_bytes;
2781 	struct u64_stats_sync   syncp;
2782 } __aligned(4 * sizeof(u64));
2783 
2784 struct pcpu_dstats {
2785 	u64			rx_packets;
2786 	u64			rx_bytes;
2787 	u64			rx_drops;
2788 	u64			tx_packets;
2789 	u64			tx_bytes;
2790 	u64			tx_drops;
2791 	struct u64_stats_sync	syncp;
2792 } __aligned(8 * sizeof(u64));
2793 
2794 struct pcpu_lstats {
2795 	u64_stats_t packets;
2796 	u64_stats_t bytes;
2797 	struct u64_stats_sync syncp;
2798 } __aligned(2 * sizeof(u64));
2799 
2800 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2801 
2802 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2803 {
2804 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2805 
2806 	u64_stats_update_begin(&tstats->syncp);
2807 	u64_stats_add(&tstats->rx_bytes, len);
2808 	u64_stats_inc(&tstats->rx_packets);
2809 	u64_stats_update_end(&tstats->syncp);
2810 }
2811 
2812 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2813 					  unsigned int packets,
2814 					  unsigned int len)
2815 {
2816 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2817 
2818 	u64_stats_update_begin(&tstats->syncp);
2819 	u64_stats_add(&tstats->tx_bytes, len);
2820 	u64_stats_add(&tstats->tx_packets, packets);
2821 	u64_stats_update_end(&tstats->syncp);
2822 }
2823 
2824 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2825 {
2826 	struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2827 
2828 	u64_stats_update_begin(&lstats->syncp);
2829 	u64_stats_add(&lstats->bytes, len);
2830 	u64_stats_inc(&lstats->packets);
2831 	u64_stats_update_end(&lstats->syncp);
2832 }
2833 
2834 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2835 ({									\
2836 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2837 	if (pcpu_stats)	{						\
2838 		int __cpu;						\
2839 		for_each_possible_cpu(__cpu) {				\
2840 			typeof(type) *stat;				\
2841 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2842 			u64_stats_init(&stat->syncp);			\
2843 		}							\
2844 	}								\
2845 	pcpu_stats;							\
2846 })
2847 
2848 #define netdev_alloc_pcpu_stats(type)					\
2849 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2850 
2851 #define devm_netdev_alloc_pcpu_stats(dev, type)				\
2852 ({									\
2853 	typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2854 	if (pcpu_stats) {						\
2855 		int __cpu;						\
2856 		for_each_possible_cpu(__cpu) {				\
2857 			typeof(type) *stat;				\
2858 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2859 			u64_stats_init(&stat->syncp);			\
2860 		}							\
2861 	}								\
2862 	pcpu_stats;							\
2863 })
2864 
2865 enum netdev_lag_tx_type {
2866 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2867 	NETDEV_LAG_TX_TYPE_RANDOM,
2868 	NETDEV_LAG_TX_TYPE_BROADCAST,
2869 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2870 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2871 	NETDEV_LAG_TX_TYPE_HASH,
2872 };
2873 
2874 enum netdev_lag_hash {
2875 	NETDEV_LAG_HASH_NONE,
2876 	NETDEV_LAG_HASH_L2,
2877 	NETDEV_LAG_HASH_L34,
2878 	NETDEV_LAG_HASH_L23,
2879 	NETDEV_LAG_HASH_E23,
2880 	NETDEV_LAG_HASH_E34,
2881 	NETDEV_LAG_HASH_VLAN_SRCMAC,
2882 	NETDEV_LAG_HASH_UNKNOWN,
2883 };
2884 
2885 struct netdev_lag_upper_info {
2886 	enum netdev_lag_tx_type tx_type;
2887 	enum netdev_lag_hash hash_type;
2888 };
2889 
2890 struct netdev_lag_lower_state_info {
2891 	u8 link_up : 1,
2892 	   tx_enabled : 1;
2893 };
2894 
2895 #include <linux/notifier.h>
2896 
2897 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2898  * and the rtnetlink notification exclusion list in rtnetlink_event() when
2899  * adding new types.
2900  */
2901 enum netdev_cmd {
2902 	NETDEV_UP	= 1,	/* For now you can't veto a device up/down */
2903 	NETDEV_DOWN,
2904 	NETDEV_REBOOT,		/* Tell a protocol stack a network interface
2905 				   detected a hardware crash and restarted
2906 				   - we can use this eg to kick tcp sessions
2907 				   once done */
2908 	NETDEV_CHANGE,		/* Notify device state change */
2909 	NETDEV_REGISTER,
2910 	NETDEV_UNREGISTER,
2911 	NETDEV_CHANGEMTU,	/* notify after mtu change happened */
2912 	NETDEV_CHANGEADDR,	/* notify after the address change */
2913 	NETDEV_PRE_CHANGEADDR,	/* notify before the address change */
2914 	NETDEV_GOING_DOWN,
2915 	NETDEV_CHANGENAME,
2916 	NETDEV_FEAT_CHANGE,
2917 	NETDEV_BONDING_FAILOVER,
2918 	NETDEV_PRE_UP,
2919 	NETDEV_PRE_TYPE_CHANGE,
2920 	NETDEV_POST_TYPE_CHANGE,
2921 	NETDEV_POST_INIT,
2922 	NETDEV_PRE_UNINIT,
2923 	NETDEV_RELEASE,
2924 	NETDEV_NOTIFY_PEERS,
2925 	NETDEV_JOIN,
2926 	NETDEV_CHANGEUPPER,
2927 	NETDEV_RESEND_IGMP,
2928 	NETDEV_PRECHANGEMTU,	/* notify before mtu change happened */
2929 	NETDEV_CHANGEINFODATA,
2930 	NETDEV_BONDING_INFO,
2931 	NETDEV_PRECHANGEUPPER,
2932 	NETDEV_CHANGELOWERSTATE,
2933 	NETDEV_UDP_TUNNEL_PUSH_INFO,
2934 	NETDEV_UDP_TUNNEL_DROP_INFO,
2935 	NETDEV_CHANGE_TX_QUEUE_LEN,
2936 	NETDEV_CVLAN_FILTER_PUSH_INFO,
2937 	NETDEV_CVLAN_FILTER_DROP_INFO,
2938 	NETDEV_SVLAN_FILTER_PUSH_INFO,
2939 	NETDEV_SVLAN_FILTER_DROP_INFO,
2940 	NETDEV_OFFLOAD_XSTATS_ENABLE,
2941 	NETDEV_OFFLOAD_XSTATS_DISABLE,
2942 	NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2943 	NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2944 	NETDEV_XDP_FEAT_CHANGE,
2945 };
2946 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2947 
2948 int register_netdevice_notifier(struct notifier_block *nb);
2949 int unregister_netdevice_notifier(struct notifier_block *nb);
2950 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2951 int unregister_netdevice_notifier_net(struct net *net,
2952 				      struct notifier_block *nb);
2953 int register_netdevice_notifier_dev_net(struct net_device *dev,
2954 					struct notifier_block *nb,
2955 					struct netdev_net_notifier *nn);
2956 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2957 					  struct notifier_block *nb,
2958 					  struct netdev_net_notifier *nn);
2959 
2960 struct netdev_notifier_info {
2961 	struct net_device	*dev;
2962 	struct netlink_ext_ack	*extack;
2963 };
2964 
2965 struct netdev_notifier_info_ext {
2966 	struct netdev_notifier_info info; /* must be first */
2967 	union {
2968 		u32 mtu;
2969 	} ext;
2970 };
2971 
2972 struct netdev_notifier_change_info {
2973 	struct netdev_notifier_info info; /* must be first */
2974 	unsigned int flags_changed;
2975 };
2976 
2977 struct netdev_notifier_changeupper_info {
2978 	struct netdev_notifier_info info; /* must be first */
2979 	struct net_device *upper_dev; /* new upper dev */
2980 	bool master; /* is upper dev master */
2981 	bool linking; /* is the notification for link or unlink */
2982 	void *upper_info; /* upper dev info */
2983 };
2984 
2985 struct netdev_notifier_changelowerstate_info {
2986 	struct netdev_notifier_info info; /* must be first */
2987 	void *lower_state_info; /* is lower dev state */
2988 };
2989 
2990 struct netdev_notifier_pre_changeaddr_info {
2991 	struct netdev_notifier_info info; /* must be first */
2992 	const unsigned char *dev_addr;
2993 };
2994 
2995 enum netdev_offload_xstats_type {
2996 	NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2997 };
2998 
2999 struct netdev_notifier_offload_xstats_info {
3000 	struct netdev_notifier_info info; /* must be first */
3001 	enum netdev_offload_xstats_type type;
3002 
3003 	union {
3004 		/* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
3005 		struct netdev_notifier_offload_xstats_rd *report_delta;
3006 		/* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
3007 		struct netdev_notifier_offload_xstats_ru *report_used;
3008 	};
3009 };
3010 
3011 int netdev_offload_xstats_enable(struct net_device *dev,
3012 				 enum netdev_offload_xstats_type type,
3013 				 struct netlink_ext_ack *extack);
3014 int netdev_offload_xstats_disable(struct net_device *dev,
3015 				  enum netdev_offload_xstats_type type);
3016 bool netdev_offload_xstats_enabled(const struct net_device *dev,
3017 				   enum netdev_offload_xstats_type type);
3018 int netdev_offload_xstats_get(struct net_device *dev,
3019 			      enum netdev_offload_xstats_type type,
3020 			      struct rtnl_hw_stats64 *stats, bool *used,
3021 			      struct netlink_ext_ack *extack);
3022 void
3023 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3024 				   const struct rtnl_hw_stats64 *stats);
3025 void
3026 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3027 void netdev_offload_xstats_push_delta(struct net_device *dev,
3028 				      enum netdev_offload_xstats_type type,
3029 				      const struct rtnl_hw_stats64 *stats);
3030 
3031 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3032 					     struct net_device *dev)
3033 {
3034 	info->dev = dev;
3035 	info->extack = NULL;
3036 }
3037 
3038 static inline struct net_device *
3039 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3040 {
3041 	return info->dev;
3042 }
3043 
3044 static inline struct netlink_ext_ack *
3045 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3046 {
3047 	return info->extack;
3048 }
3049 
3050 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3051 int call_netdevice_notifiers_info(unsigned long val,
3052 				  struct netdev_notifier_info *info);
3053 
3054 extern rwlock_t				dev_base_lock;		/* Device list lock */
3055 
3056 #define for_each_netdev(net, d)		\
3057 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3058 #define for_each_netdev_reverse(net, d)	\
3059 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3060 #define for_each_netdev_rcu(net, d)		\
3061 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3062 #define for_each_netdev_safe(net, d, n)	\
3063 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3064 #define for_each_netdev_continue(net, d)		\
3065 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3066 #define for_each_netdev_continue_reverse(net, d)		\
3067 		list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3068 						     dev_list)
3069 #define for_each_netdev_continue_rcu(net, d)		\
3070 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3071 #define for_each_netdev_in_bond_rcu(bond, slave)	\
3072 		for_each_netdev_rcu(&init_net, slave)	\
3073 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3074 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
3075 
3076 #define for_each_netdev_dump(net, d, ifindex)				\
3077 	xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3078 
3079 static inline struct net_device *next_net_device(struct net_device *dev)
3080 {
3081 	struct list_head *lh;
3082 	struct net *net;
3083 
3084 	net = dev_net(dev);
3085 	lh = dev->dev_list.next;
3086 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3087 }
3088 
3089 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3090 {
3091 	struct list_head *lh;
3092 	struct net *net;
3093 
3094 	net = dev_net(dev);
3095 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3096 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3097 }
3098 
3099 static inline struct net_device *first_net_device(struct net *net)
3100 {
3101 	return list_empty(&net->dev_base_head) ? NULL :
3102 		net_device_entry(net->dev_base_head.next);
3103 }
3104 
3105 static inline struct net_device *first_net_device_rcu(struct net *net)
3106 {
3107 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3108 
3109 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3110 }
3111 
3112 int netdev_boot_setup_check(struct net_device *dev);
3113 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3114 				       const char *hwaddr);
3115 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3116 void dev_add_pack(struct packet_type *pt);
3117 void dev_remove_pack(struct packet_type *pt);
3118 void __dev_remove_pack(struct packet_type *pt);
3119 void dev_add_offload(struct packet_offload *po);
3120 void dev_remove_offload(struct packet_offload *po);
3121 
3122 int dev_get_iflink(const struct net_device *dev);
3123 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3124 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3125 			  struct net_device_path_stack *stack);
3126 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3127 				      unsigned short mask);
3128 struct net_device *dev_get_by_name(struct net *net, const char *name);
3129 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3130 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3131 bool netdev_name_in_use(struct net *net, const char *name);
3132 int dev_alloc_name(struct net_device *dev, const char *name);
3133 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3134 void dev_close(struct net_device *dev);
3135 void dev_close_many(struct list_head *head, bool unlink);
3136 void dev_disable_lro(struct net_device *dev);
3137 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3138 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3139 		     struct net_device *sb_dev);
3140 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3141 		       struct net_device *sb_dev);
3142 
3143 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3144 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3145 
3146 static inline int dev_queue_xmit(struct sk_buff *skb)
3147 {
3148 	return __dev_queue_xmit(skb, NULL);
3149 }
3150 
3151 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3152 				       struct net_device *sb_dev)
3153 {
3154 	return __dev_queue_xmit(skb, sb_dev);
3155 }
3156 
3157 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3158 {
3159 	int ret;
3160 
3161 	ret = __dev_direct_xmit(skb, queue_id);
3162 	if (!dev_xmit_complete(ret))
3163 		kfree_skb(skb);
3164 	return ret;
3165 }
3166 
3167 int register_netdevice(struct net_device *dev);
3168 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3169 void unregister_netdevice_many(struct list_head *head);
3170 static inline void unregister_netdevice(struct net_device *dev)
3171 {
3172 	unregister_netdevice_queue(dev, NULL);
3173 }
3174 
3175 int netdev_refcnt_read(const struct net_device *dev);
3176 void free_netdev(struct net_device *dev);
3177 void netdev_freemem(struct net_device *dev);
3178 int init_dummy_netdev(struct net_device *dev);
3179 
3180 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3181 					 struct sk_buff *skb,
3182 					 bool all_slaves);
3183 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3184 					    struct sock *sk);
3185 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3186 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3187 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3188 				       netdevice_tracker *tracker, gfp_t gfp);
3189 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3190 				      netdevice_tracker *tracker, gfp_t gfp);
3191 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3192 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3193 
3194 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3195 				  unsigned short type,
3196 				  const void *daddr, const void *saddr,
3197 				  unsigned int len)
3198 {
3199 	if (!dev->header_ops || !dev->header_ops->create)
3200 		return 0;
3201 
3202 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3203 }
3204 
3205 static inline int dev_parse_header(const struct sk_buff *skb,
3206 				   unsigned char *haddr)
3207 {
3208 	const struct net_device *dev = skb->dev;
3209 
3210 	if (!dev->header_ops || !dev->header_ops->parse)
3211 		return 0;
3212 	return dev->header_ops->parse(skb, haddr);
3213 }
3214 
3215 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3216 {
3217 	const struct net_device *dev = skb->dev;
3218 
3219 	if (!dev->header_ops || !dev->header_ops->parse_protocol)
3220 		return 0;
3221 	return dev->header_ops->parse_protocol(skb);
3222 }
3223 
3224 /* ll_header must have at least hard_header_len allocated */
3225 static inline bool dev_validate_header(const struct net_device *dev,
3226 				       char *ll_header, int len)
3227 {
3228 	if (likely(len >= dev->hard_header_len))
3229 		return true;
3230 	if (len < dev->min_header_len)
3231 		return false;
3232 
3233 	if (capable(CAP_SYS_RAWIO)) {
3234 		memset(ll_header + len, 0, dev->hard_header_len - len);
3235 		return true;
3236 	}
3237 
3238 	if (dev->header_ops && dev->header_ops->validate)
3239 		return dev->header_ops->validate(ll_header, len);
3240 
3241 	return false;
3242 }
3243 
3244 static inline bool dev_has_header(const struct net_device *dev)
3245 {
3246 	return dev->header_ops && dev->header_ops->create;
3247 }
3248 
3249 /*
3250  * Incoming packets are placed on per-CPU queues
3251  */
3252 struct softnet_data {
3253 	struct list_head	poll_list;
3254 	struct sk_buff_head	process_queue;
3255 
3256 	/* stats */
3257 	unsigned int		processed;
3258 	unsigned int		time_squeeze;
3259 #ifdef CONFIG_RPS
3260 	struct softnet_data	*rps_ipi_list;
3261 #endif
3262 
3263 	bool			in_net_rx_action;
3264 	bool			in_napi_threaded_poll;
3265 
3266 #ifdef CONFIG_NET_FLOW_LIMIT
3267 	struct sd_flow_limit __rcu *flow_limit;
3268 #endif
3269 	struct Qdisc		*output_queue;
3270 	struct Qdisc		**output_queue_tailp;
3271 	struct sk_buff		*completion_queue;
3272 #ifdef CONFIG_XFRM_OFFLOAD
3273 	struct sk_buff_head	xfrm_backlog;
3274 #endif
3275 	/* written and read only by owning cpu: */
3276 	struct {
3277 		u16 recursion;
3278 		u8  more;
3279 #ifdef CONFIG_NET_EGRESS
3280 		u8  skip_txqueue;
3281 #endif
3282 	} xmit;
3283 #ifdef CONFIG_RPS
3284 	/* input_queue_head should be written by cpu owning this struct,
3285 	 * and only read by other cpus. Worth using a cache line.
3286 	 */
3287 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
3288 
3289 	/* Elements below can be accessed between CPUs for RPS/RFS */
3290 	call_single_data_t	csd ____cacheline_aligned_in_smp;
3291 	struct softnet_data	*rps_ipi_next;
3292 	unsigned int		cpu;
3293 	unsigned int		input_queue_tail;
3294 #endif
3295 	unsigned int		received_rps;
3296 	unsigned int		dropped;
3297 	struct sk_buff_head	input_pkt_queue;
3298 	struct napi_struct	backlog;
3299 
3300 	/* Another possibly contended cache line */
3301 	spinlock_t		defer_lock ____cacheline_aligned_in_smp;
3302 	int			defer_count;
3303 	int			defer_ipi_scheduled;
3304 	struct sk_buff		*defer_list;
3305 	call_single_data_t	defer_csd;
3306 };
3307 
3308 static inline void input_queue_head_incr(struct softnet_data *sd)
3309 {
3310 #ifdef CONFIG_RPS
3311 	sd->input_queue_head++;
3312 #endif
3313 }
3314 
3315 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3316 					      unsigned int *qtail)
3317 {
3318 #ifdef CONFIG_RPS
3319 	*qtail = ++sd->input_queue_tail;
3320 #endif
3321 }
3322 
3323 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3324 
3325 static inline int dev_recursion_level(void)
3326 {
3327 	return this_cpu_read(softnet_data.xmit.recursion);
3328 }
3329 
3330 #define XMIT_RECURSION_LIMIT	8
3331 static inline bool dev_xmit_recursion(void)
3332 {
3333 	return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3334 			XMIT_RECURSION_LIMIT);
3335 }
3336 
3337 static inline void dev_xmit_recursion_inc(void)
3338 {
3339 	__this_cpu_inc(softnet_data.xmit.recursion);
3340 }
3341 
3342 static inline void dev_xmit_recursion_dec(void)
3343 {
3344 	__this_cpu_dec(softnet_data.xmit.recursion);
3345 }
3346 
3347 void __netif_schedule(struct Qdisc *q);
3348 void netif_schedule_queue(struct netdev_queue *txq);
3349 
3350 static inline void netif_tx_schedule_all(struct net_device *dev)
3351 {
3352 	unsigned int i;
3353 
3354 	for (i = 0; i < dev->num_tx_queues; i++)
3355 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
3356 }
3357 
3358 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3359 {
3360 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3361 }
3362 
3363 /**
3364  *	netif_start_queue - allow transmit
3365  *	@dev: network device
3366  *
3367  *	Allow upper layers to call the device hard_start_xmit routine.
3368  */
3369 static inline void netif_start_queue(struct net_device *dev)
3370 {
3371 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3372 }
3373 
3374 static inline void netif_tx_start_all_queues(struct net_device *dev)
3375 {
3376 	unsigned int i;
3377 
3378 	for (i = 0; i < dev->num_tx_queues; i++) {
3379 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3380 		netif_tx_start_queue(txq);
3381 	}
3382 }
3383 
3384 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3385 
3386 /**
3387  *	netif_wake_queue - restart transmit
3388  *	@dev: network device
3389  *
3390  *	Allow upper layers to call the device hard_start_xmit routine.
3391  *	Used for flow control when transmit resources are available.
3392  */
3393 static inline void netif_wake_queue(struct net_device *dev)
3394 {
3395 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3396 }
3397 
3398 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3399 {
3400 	unsigned int i;
3401 
3402 	for (i = 0; i < dev->num_tx_queues; i++) {
3403 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3404 		netif_tx_wake_queue(txq);
3405 	}
3406 }
3407 
3408 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3409 {
3410 	/* Must be an atomic op see netif_txq_try_stop() */
3411 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3412 }
3413 
3414 /**
3415  *	netif_stop_queue - stop transmitted packets
3416  *	@dev: network device
3417  *
3418  *	Stop upper layers calling the device hard_start_xmit routine.
3419  *	Used for flow control when transmit resources are unavailable.
3420  */
3421 static inline void netif_stop_queue(struct net_device *dev)
3422 {
3423 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3424 }
3425 
3426 void netif_tx_stop_all_queues(struct net_device *dev);
3427 
3428 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3429 {
3430 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3431 }
3432 
3433 /**
3434  *	netif_queue_stopped - test if transmit queue is flowblocked
3435  *	@dev: network device
3436  *
3437  *	Test if transmit queue on device is currently unable to send.
3438  */
3439 static inline bool netif_queue_stopped(const struct net_device *dev)
3440 {
3441 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3442 }
3443 
3444 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3445 {
3446 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3447 }
3448 
3449 static inline bool
3450 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3451 {
3452 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3453 }
3454 
3455 static inline bool
3456 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3457 {
3458 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3459 }
3460 
3461 /**
3462  *	netdev_queue_set_dql_min_limit - set dql minimum limit
3463  *	@dev_queue: pointer to transmit queue
3464  *	@min_limit: dql minimum limit
3465  *
3466  * Forces xmit_more() to return true until the minimum threshold
3467  * defined by @min_limit is reached (or until the tx queue is
3468  * empty). Warning: to be use with care, misuse will impact the
3469  * latency.
3470  */
3471 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3472 						  unsigned int min_limit)
3473 {
3474 #ifdef CONFIG_BQL
3475 	dev_queue->dql.min_limit = min_limit;
3476 #endif
3477 }
3478 
3479 /**
3480  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3481  *	@dev_queue: pointer to transmit queue
3482  *
3483  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3484  * to give appropriate hint to the CPU.
3485  */
3486 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3487 {
3488 #ifdef CONFIG_BQL
3489 	prefetchw(&dev_queue->dql.num_queued);
3490 #endif
3491 }
3492 
3493 /**
3494  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3495  *	@dev_queue: pointer to transmit queue
3496  *
3497  * BQL enabled drivers might use this helper in their TX completion path,
3498  * to give appropriate hint to the CPU.
3499  */
3500 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3501 {
3502 #ifdef CONFIG_BQL
3503 	prefetchw(&dev_queue->dql.limit);
3504 #endif
3505 }
3506 
3507 /**
3508  *	netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3509  *	@dev_queue: network device queue
3510  *	@bytes: number of bytes queued to the device queue
3511  *
3512  *	Report the number of bytes queued for sending/completion to the network
3513  *	device hardware queue. @bytes should be a good approximation and should
3514  *	exactly match netdev_completed_queue() @bytes.
3515  *	This is typically called once per packet, from ndo_start_xmit().
3516  */
3517 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3518 					unsigned int bytes)
3519 {
3520 #ifdef CONFIG_BQL
3521 	dql_queued(&dev_queue->dql, bytes);
3522 
3523 	if (likely(dql_avail(&dev_queue->dql) >= 0))
3524 		return;
3525 
3526 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3527 
3528 	/*
3529 	 * The XOFF flag must be set before checking the dql_avail below,
3530 	 * because in netdev_tx_completed_queue we update the dql_completed
3531 	 * before checking the XOFF flag.
3532 	 */
3533 	smp_mb();
3534 
3535 	/* check again in case another CPU has just made room avail */
3536 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3537 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3538 #endif
3539 }
3540 
3541 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3542  * that they should not test BQL status themselves.
3543  * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3544  * skb of a batch.
3545  * Returns true if the doorbell must be used to kick the NIC.
3546  */
3547 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3548 					  unsigned int bytes,
3549 					  bool xmit_more)
3550 {
3551 	if (xmit_more) {
3552 #ifdef CONFIG_BQL
3553 		dql_queued(&dev_queue->dql, bytes);
3554 #endif
3555 		return netif_tx_queue_stopped(dev_queue);
3556 	}
3557 	netdev_tx_sent_queue(dev_queue, bytes);
3558 	return true;
3559 }
3560 
3561 /**
3562  *	netdev_sent_queue - report the number of bytes queued to hardware
3563  *	@dev: network device
3564  *	@bytes: number of bytes queued to the hardware device queue
3565  *
3566  *	Report the number of bytes queued for sending/completion to the network
3567  *	device hardware queue#0. @bytes should be a good approximation and should
3568  *	exactly match netdev_completed_queue() @bytes.
3569  *	This is typically called once per packet, from ndo_start_xmit().
3570  */
3571 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3572 {
3573 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3574 }
3575 
3576 static inline bool __netdev_sent_queue(struct net_device *dev,
3577 				       unsigned int bytes,
3578 				       bool xmit_more)
3579 {
3580 	return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3581 				      xmit_more);
3582 }
3583 
3584 /**
3585  *	netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3586  *	@dev_queue: network device queue
3587  *	@pkts: number of packets (currently ignored)
3588  *	@bytes: number of bytes dequeued from the device queue
3589  *
3590  *	Must be called at most once per TX completion round (and not per
3591  *	individual packet), so that BQL can adjust its limits appropriately.
3592  */
3593 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3594 					     unsigned int pkts, unsigned int bytes)
3595 {
3596 #ifdef CONFIG_BQL
3597 	if (unlikely(!bytes))
3598 		return;
3599 
3600 	dql_completed(&dev_queue->dql, bytes);
3601 
3602 	/*
3603 	 * Without the memory barrier there is a small possiblity that
3604 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3605 	 * be stopped forever
3606 	 */
3607 	smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3608 
3609 	if (unlikely(dql_avail(&dev_queue->dql) < 0))
3610 		return;
3611 
3612 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3613 		netif_schedule_queue(dev_queue);
3614 #endif
3615 }
3616 
3617 /**
3618  * 	netdev_completed_queue - report bytes and packets completed by device
3619  * 	@dev: network device
3620  * 	@pkts: actual number of packets sent over the medium
3621  * 	@bytes: actual number of bytes sent over the medium
3622  *
3623  * 	Report the number of bytes and packets transmitted by the network device
3624  * 	hardware queue over the physical medium, @bytes must exactly match the
3625  * 	@bytes amount passed to netdev_sent_queue()
3626  */
3627 static inline void netdev_completed_queue(struct net_device *dev,
3628 					  unsigned int pkts, unsigned int bytes)
3629 {
3630 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3631 }
3632 
3633 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3634 {
3635 #ifdef CONFIG_BQL
3636 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3637 	dql_reset(&q->dql);
3638 #endif
3639 }
3640 
3641 /**
3642  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3643  * 	@dev_queue: network device
3644  *
3645  * 	Reset the bytes and packet count of a network device and clear the
3646  * 	software flow control OFF bit for this network device
3647  */
3648 static inline void netdev_reset_queue(struct net_device *dev_queue)
3649 {
3650 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3651 }
3652 
3653 /**
3654  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3655  * 	@dev: network device
3656  * 	@queue_index: given tx queue index
3657  *
3658  * 	Returns 0 if given tx queue index >= number of device tx queues,
3659  * 	otherwise returns the originally passed tx queue index.
3660  */
3661 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3662 {
3663 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3664 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3665 				     dev->name, queue_index,
3666 				     dev->real_num_tx_queues);
3667 		return 0;
3668 	}
3669 
3670 	return queue_index;
3671 }
3672 
3673 /**
3674  *	netif_running - test if up
3675  *	@dev: network device
3676  *
3677  *	Test if the device has been brought up.
3678  */
3679 static inline bool netif_running(const struct net_device *dev)
3680 {
3681 	return test_bit(__LINK_STATE_START, &dev->state);
3682 }
3683 
3684 /*
3685  * Routines to manage the subqueues on a device.  We only need start,
3686  * stop, and a check if it's stopped.  All other device management is
3687  * done at the overall netdevice level.
3688  * Also test the device if we're multiqueue.
3689  */
3690 
3691 /**
3692  *	netif_start_subqueue - allow sending packets on subqueue
3693  *	@dev: network device
3694  *	@queue_index: sub queue index
3695  *
3696  * Start individual transmit queue of a device with multiple transmit queues.
3697  */
3698 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3699 {
3700 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3701 
3702 	netif_tx_start_queue(txq);
3703 }
3704 
3705 /**
3706  *	netif_stop_subqueue - stop sending packets on subqueue
3707  *	@dev: network device
3708  *	@queue_index: sub queue index
3709  *
3710  * Stop individual transmit queue of a device with multiple transmit queues.
3711  */
3712 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3713 {
3714 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3715 	netif_tx_stop_queue(txq);
3716 }
3717 
3718 /**
3719  *	__netif_subqueue_stopped - test status of subqueue
3720  *	@dev: network device
3721  *	@queue_index: sub queue index
3722  *
3723  * Check individual transmit queue of a device with multiple transmit queues.
3724  */
3725 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3726 					    u16 queue_index)
3727 {
3728 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3729 
3730 	return netif_tx_queue_stopped(txq);
3731 }
3732 
3733 /**
3734  *	netif_subqueue_stopped - test status of subqueue
3735  *	@dev: network device
3736  *	@skb: sub queue buffer pointer
3737  *
3738  * Check individual transmit queue of a device with multiple transmit queues.
3739  */
3740 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3741 					  struct sk_buff *skb)
3742 {
3743 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3744 }
3745 
3746 /**
3747  *	netif_wake_subqueue - allow sending packets on subqueue
3748  *	@dev: network device
3749  *	@queue_index: sub queue index
3750  *
3751  * Resume individual transmit queue of a device with multiple transmit queues.
3752  */
3753 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3754 {
3755 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3756 
3757 	netif_tx_wake_queue(txq);
3758 }
3759 
3760 #ifdef CONFIG_XPS
3761 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3762 			u16 index);
3763 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3764 			  u16 index, enum xps_map_type type);
3765 
3766 /**
3767  *	netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3768  *	@j: CPU/Rx queue index
3769  *	@mask: bitmask of all cpus/rx queues
3770  *	@nr_bits: number of bits in the bitmask
3771  *
3772  * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3773  */
3774 static inline bool netif_attr_test_mask(unsigned long j,
3775 					const unsigned long *mask,
3776 					unsigned int nr_bits)
3777 {
3778 	cpu_max_bits_warn(j, nr_bits);
3779 	return test_bit(j, mask);
3780 }
3781 
3782 /**
3783  *	netif_attr_test_online - Test for online CPU/Rx queue
3784  *	@j: CPU/Rx queue index
3785  *	@online_mask: bitmask for CPUs/Rx queues that are online
3786  *	@nr_bits: number of bits in the bitmask
3787  *
3788  * Returns true if a CPU/Rx queue is online.
3789  */
3790 static inline bool netif_attr_test_online(unsigned long j,
3791 					  const unsigned long *online_mask,
3792 					  unsigned int nr_bits)
3793 {
3794 	cpu_max_bits_warn(j, nr_bits);
3795 
3796 	if (online_mask)
3797 		return test_bit(j, online_mask);
3798 
3799 	return (j < nr_bits);
3800 }
3801 
3802 /**
3803  *	netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3804  *	@n: CPU/Rx queue index
3805  *	@srcp: the cpumask/Rx queue mask pointer
3806  *	@nr_bits: number of bits in the bitmask
3807  *
3808  * Returns >= nr_bits if no further CPUs/Rx queues set.
3809  */
3810 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3811 					       unsigned int nr_bits)
3812 {
3813 	/* -1 is a legal arg here. */
3814 	if (n != -1)
3815 		cpu_max_bits_warn(n, nr_bits);
3816 
3817 	if (srcp)
3818 		return find_next_bit(srcp, nr_bits, n + 1);
3819 
3820 	return n + 1;
3821 }
3822 
3823 /**
3824  *	netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3825  *	@n: CPU/Rx queue index
3826  *	@src1p: the first CPUs/Rx queues mask pointer
3827  *	@src2p: the second CPUs/Rx queues mask pointer
3828  *	@nr_bits: number of bits in the bitmask
3829  *
3830  * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3831  */
3832 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3833 					  const unsigned long *src2p,
3834 					  unsigned int nr_bits)
3835 {
3836 	/* -1 is a legal arg here. */
3837 	if (n != -1)
3838 		cpu_max_bits_warn(n, nr_bits);
3839 
3840 	if (src1p && src2p)
3841 		return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3842 	else if (src1p)
3843 		return find_next_bit(src1p, nr_bits, n + 1);
3844 	else if (src2p)
3845 		return find_next_bit(src2p, nr_bits, n + 1);
3846 
3847 	return n + 1;
3848 }
3849 #else
3850 static inline int netif_set_xps_queue(struct net_device *dev,
3851 				      const struct cpumask *mask,
3852 				      u16 index)
3853 {
3854 	return 0;
3855 }
3856 
3857 static inline int __netif_set_xps_queue(struct net_device *dev,
3858 					const unsigned long *mask,
3859 					u16 index, enum xps_map_type type)
3860 {
3861 	return 0;
3862 }
3863 #endif
3864 
3865 /**
3866  *	netif_is_multiqueue - test if device has multiple transmit queues
3867  *	@dev: network device
3868  *
3869  * Check if device has multiple transmit queues
3870  */
3871 static inline bool netif_is_multiqueue(const struct net_device *dev)
3872 {
3873 	return dev->num_tx_queues > 1;
3874 }
3875 
3876 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3877 
3878 #ifdef CONFIG_SYSFS
3879 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3880 #else
3881 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3882 						unsigned int rxqs)
3883 {
3884 	dev->real_num_rx_queues = rxqs;
3885 	return 0;
3886 }
3887 #endif
3888 int netif_set_real_num_queues(struct net_device *dev,
3889 			      unsigned int txq, unsigned int rxq);
3890 
3891 int netif_get_num_default_rss_queues(void);
3892 
3893 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3894 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3895 
3896 /*
3897  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3898  * interrupt context or with hardware interrupts being disabled.
3899  * (in_hardirq() || irqs_disabled())
3900  *
3901  * We provide four helpers that can be used in following contexts :
3902  *
3903  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3904  *  replacing kfree_skb(skb)
3905  *
3906  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3907  *  Typically used in place of consume_skb(skb) in TX completion path
3908  *
3909  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3910  *  replacing kfree_skb(skb)
3911  *
3912  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3913  *  and consumed a packet. Used in place of consume_skb(skb)
3914  */
3915 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3916 {
3917 	dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3918 }
3919 
3920 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3921 {
3922 	dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3923 }
3924 
3925 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3926 {
3927 	dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3928 }
3929 
3930 static inline void dev_consume_skb_any(struct sk_buff *skb)
3931 {
3932 	dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3933 }
3934 
3935 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3936 			     struct bpf_prog *xdp_prog);
3937 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3938 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3939 int netif_rx(struct sk_buff *skb);
3940 int __netif_rx(struct sk_buff *skb);
3941 
3942 int netif_receive_skb(struct sk_buff *skb);
3943 int netif_receive_skb_core(struct sk_buff *skb);
3944 void netif_receive_skb_list_internal(struct list_head *head);
3945 void netif_receive_skb_list(struct list_head *head);
3946 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3947 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3948 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3949 void napi_get_frags_check(struct napi_struct *napi);
3950 gro_result_t napi_gro_frags(struct napi_struct *napi);
3951 struct packet_offload *gro_find_receive_by_type(__be16 type);
3952 struct packet_offload *gro_find_complete_by_type(__be16 type);
3953 
3954 static inline void napi_free_frags(struct napi_struct *napi)
3955 {
3956 	kfree_skb(napi->skb);
3957 	napi->skb = NULL;
3958 }
3959 
3960 bool netdev_is_rx_handler_busy(struct net_device *dev);
3961 int netdev_rx_handler_register(struct net_device *dev,
3962 			       rx_handler_func_t *rx_handler,
3963 			       void *rx_handler_data);
3964 void netdev_rx_handler_unregister(struct net_device *dev);
3965 
3966 bool dev_valid_name(const char *name);
3967 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3968 {
3969 	return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3970 }
3971 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3972 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3973 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3974 		void __user *data, bool *need_copyout);
3975 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3976 int generic_hwtstamp_get_lower(struct net_device *dev,
3977 			       struct kernel_hwtstamp_config *kernel_cfg);
3978 int generic_hwtstamp_set_lower(struct net_device *dev,
3979 			       struct kernel_hwtstamp_config *kernel_cfg,
3980 			       struct netlink_ext_ack *extack);
3981 int dev_set_hwtstamp_phylib(struct net_device *dev,
3982 			    struct kernel_hwtstamp_config *cfg,
3983 			    struct netlink_ext_ack *extack);
3984 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3985 unsigned int dev_get_flags(const struct net_device *);
3986 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3987 		       struct netlink_ext_ack *extack);
3988 int dev_change_flags(struct net_device *dev, unsigned int flags,
3989 		     struct netlink_ext_ack *extack);
3990 int dev_set_alias(struct net_device *, const char *, size_t);
3991 int dev_get_alias(const struct net_device *, char *, size_t);
3992 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3993 			       const char *pat, int new_ifindex);
3994 static inline
3995 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3996 			     const char *pat)
3997 {
3998 	return __dev_change_net_namespace(dev, net, pat, 0);
3999 }
4000 int __dev_set_mtu(struct net_device *, int);
4001 int dev_set_mtu(struct net_device *, int);
4002 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4003 			      struct netlink_ext_ack *extack);
4004 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4005 			struct netlink_ext_ack *extack);
4006 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4007 			     struct netlink_ext_ack *extack);
4008 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4009 int dev_get_port_parent_id(struct net_device *dev,
4010 			   struct netdev_phys_item_id *ppid, bool recurse);
4011 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4012 void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin);
4013 void netdev_dpll_pin_clear(struct net_device *dev);
4014 
4015 static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev)
4016 {
4017 #if IS_ENABLED(CONFIG_DPLL)
4018 	return dev->dpll_pin;
4019 #else
4020 	return NULL;
4021 #endif
4022 }
4023 
4024 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4025 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4026 				    struct netdev_queue *txq, int *ret);
4027 
4028 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4029 u8 dev_xdp_prog_count(struct net_device *dev);
4030 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4031 
4032 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4033 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4034 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4035 bool is_skb_forwardable(const struct net_device *dev,
4036 			const struct sk_buff *skb);
4037 
4038 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4039 						 const struct sk_buff *skb,
4040 						 const bool check_mtu)
4041 {
4042 	const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4043 	unsigned int len;
4044 
4045 	if (!(dev->flags & IFF_UP))
4046 		return false;
4047 
4048 	if (!check_mtu)
4049 		return true;
4050 
4051 	len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4052 	if (skb->len <= len)
4053 		return true;
4054 
4055 	/* if TSO is enabled, we don't care about the length as the packet
4056 	 * could be forwarded without being segmented before
4057 	 */
4058 	if (skb_is_gso(skb))
4059 		return true;
4060 
4061 	return false;
4062 }
4063 
4064 void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4065 
4066 #define DEV_CORE_STATS_INC(FIELD)						\
4067 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev)		\
4068 {										\
4069 	netdev_core_stats_inc(dev,						\
4070 			offsetof(struct net_device_core_stats, FIELD));		\
4071 }
4072 DEV_CORE_STATS_INC(rx_dropped)
4073 DEV_CORE_STATS_INC(tx_dropped)
4074 DEV_CORE_STATS_INC(rx_nohandler)
4075 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4076 #undef DEV_CORE_STATS_INC
4077 
4078 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4079 					       struct sk_buff *skb,
4080 					       const bool check_mtu)
4081 {
4082 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4083 	    unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4084 		dev_core_stats_rx_dropped_inc(dev);
4085 		kfree_skb(skb);
4086 		return NET_RX_DROP;
4087 	}
4088 
4089 	skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4090 	skb->priority = 0;
4091 	return 0;
4092 }
4093 
4094 bool dev_nit_active(struct net_device *dev);
4095 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4096 
4097 static inline void __dev_put(struct net_device *dev)
4098 {
4099 	if (dev) {
4100 #ifdef CONFIG_PCPU_DEV_REFCNT
4101 		this_cpu_dec(*dev->pcpu_refcnt);
4102 #else
4103 		refcount_dec(&dev->dev_refcnt);
4104 #endif
4105 	}
4106 }
4107 
4108 static inline void __dev_hold(struct net_device *dev)
4109 {
4110 	if (dev) {
4111 #ifdef CONFIG_PCPU_DEV_REFCNT
4112 		this_cpu_inc(*dev->pcpu_refcnt);
4113 #else
4114 		refcount_inc(&dev->dev_refcnt);
4115 #endif
4116 	}
4117 }
4118 
4119 static inline void __netdev_tracker_alloc(struct net_device *dev,
4120 					  netdevice_tracker *tracker,
4121 					  gfp_t gfp)
4122 {
4123 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4124 	ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4125 #endif
4126 }
4127 
4128 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4129  * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4130  */
4131 static inline void netdev_tracker_alloc(struct net_device *dev,
4132 					netdevice_tracker *tracker, gfp_t gfp)
4133 {
4134 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4135 	refcount_dec(&dev->refcnt_tracker.no_tracker);
4136 	__netdev_tracker_alloc(dev, tracker, gfp);
4137 #endif
4138 }
4139 
4140 static inline void netdev_tracker_free(struct net_device *dev,
4141 				       netdevice_tracker *tracker)
4142 {
4143 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4144 	ref_tracker_free(&dev->refcnt_tracker, tracker);
4145 #endif
4146 }
4147 
4148 static inline void netdev_hold(struct net_device *dev,
4149 			       netdevice_tracker *tracker, gfp_t gfp)
4150 {
4151 	if (dev) {
4152 		__dev_hold(dev);
4153 		__netdev_tracker_alloc(dev, tracker, gfp);
4154 	}
4155 }
4156 
4157 static inline void netdev_put(struct net_device *dev,
4158 			      netdevice_tracker *tracker)
4159 {
4160 	if (dev) {
4161 		netdev_tracker_free(dev, tracker);
4162 		__dev_put(dev);
4163 	}
4164 }
4165 
4166 /**
4167  *	dev_hold - get reference to device
4168  *	@dev: network device
4169  *
4170  * Hold reference to device to keep it from being freed.
4171  * Try using netdev_hold() instead.
4172  */
4173 static inline void dev_hold(struct net_device *dev)
4174 {
4175 	netdev_hold(dev, NULL, GFP_ATOMIC);
4176 }
4177 
4178 /**
4179  *	dev_put - release reference to device
4180  *	@dev: network device
4181  *
4182  * Release reference to device to allow it to be freed.
4183  * Try using netdev_put() instead.
4184  */
4185 static inline void dev_put(struct net_device *dev)
4186 {
4187 	netdev_put(dev, NULL);
4188 }
4189 
4190 static inline void netdev_ref_replace(struct net_device *odev,
4191 				      struct net_device *ndev,
4192 				      netdevice_tracker *tracker,
4193 				      gfp_t gfp)
4194 {
4195 	if (odev)
4196 		netdev_tracker_free(odev, tracker);
4197 
4198 	__dev_hold(ndev);
4199 	__dev_put(odev);
4200 
4201 	if (ndev)
4202 		__netdev_tracker_alloc(ndev, tracker, gfp);
4203 }
4204 
4205 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4206  * and _off may be called from IRQ context, but it is caller
4207  * who is responsible for serialization of these calls.
4208  *
4209  * The name carrier is inappropriate, these functions should really be
4210  * called netif_lowerlayer_*() because they represent the state of any
4211  * kind of lower layer not just hardware media.
4212  */
4213 void linkwatch_fire_event(struct net_device *dev);
4214 
4215 /**
4216  *	netif_carrier_ok - test if carrier present
4217  *	@dev: network device
4218  *
4219  * Check if carrier is present on device
4220  */
4221 static inline bool netif_carrier_ok(const struct net_device *dev)
4222 {
4223 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4224 }
4225 
4226 unsigned long dev_trans_start(struct net_device *dev);
4227 
4228 void __netdev_watchdog_up(struct net_device *dev);
4229 
4230 void netif_carrier_on(struct net_device *dev);
4231 void netif_carrier_off(struct net_device *dev);
4232 void netif_carrier_event(struct net_device *dev);
4233 
4234 /**
4235  *	netif_dormant_on - mark device as dormant.
4236  *	@dev: network device
4237  *
4238  * Mark device as dormant (as per RFC2863).
4239  *
4240  * The dormant state indicates that the relevant interface is not
4241  * actually in a condition to pass packets (i.e., it is not 'up') but is
4242  * in a "pending" state, waiting for some external event.  For "on-
4243  * demand" interfaces, this new state identifies the situation where the
4244  * interface is waiting for events to place it in the up state.
4245  */
4246 static inline void netif_dormant_on(struct net_device *dev)
4247 {
4248 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4249 		linkwatch_fire_event(dev);
4250 }
4251 
4252 /**
4253  *	netif_dormant_off - set device as not dormant.
4254  *	@dev: network device
4255  *
4256  * Device is not in dormant state.
4257  */
4258 static inline void netif_dormant_off(struct net_device *dev)
4259 {
4260 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4261 		linkwatch_fire_event(dev);
4262 }
4263 
4264 /**
4265  *	netif_dormant - test if device is dormant
4266  *	@dev: network device
4267  *
4268  * Check if device is dormant.
4269  */
4270 static inline bool netif_dormant(const struct net_device *dev)
4271 {
4272 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
4273 }
4274 
4275 
4276 /**
4277  *	netif_testing_on - mark device as under test.
4278  *	@dev: network device
4279  *
4280  * Mark device as under test (as per RFC2863).
4281  *
4282  * The testing state indicates that some test(s) must be performed on
4283  * the interface. After completion, of the test, the interface state
4284  * will change to up, dormant, or down, as appropriate.
4285  */
4286 static inline void netif_testing_on(struct net_device *dev)
4287 {
4288 	if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4289 		linkwatch_fire_event(dev);
4290 }
4291 
4292 /**
4293  *	netif_testing_off - set device as not under test.
4294  *	@dev: network device
4295  *
4296  * Device is not in testing state.
4297  */
4298 static inline void netif_testing_off(struct net_device *dev)
4299 {
4300 	if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4301 		linkwatch_fire_event(dev);
4302 }
4303 
4304 /**
4305  *	netif_testing - test if device is under test
4306  *	@dev: network device
4307  *
4308  * Check if device is under test
4309  */
4310 static inline bool netif_testing(const struct net_device *dev)
4311 {
4312 	return test_bit(__LINK_STATE_TESTING, &dev->state);
4313 }
4314 
4315 
4316 /**
4317  *	netif_oper_up - test if device is operational
4318  *	@dev: network device
4319  *
4320  * Check if carrier is operational
4321  */
4322 static inline bool netif_oper_up(const struct net_device *dev)
4323 {
4324 	return (dev->operstate == IF_OPER_UP ||
4325 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4326 }
4327 
4328 /**
4329  *	netif_device_present - is device available or removed
4330  *	@dev: network device
4331  *
4332  * Check if device has not been removed from system.
4333  */
4334 static inline bool netif_device_present(const struct net_device *dev)
4335 {
4336 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
4337 }
4338 
4339 void netif_device_detach(struct net_device *dev);
4340 
4341 void netif_device_attach(struct net_device *dev);
4342 
4343 /*
4344  * Network interface message level settings
4345  */
4346 
4347 enum {
4348 	NETIF_MSG_DRV_BIT,
4349 	NETIF_MSG_PROBE_BIT,
4350 	NETIF_MSG_LINK_BIT,
4351 	NETIF_MSG_TIMER_BIT,
4352 	NETIF_MSG_IFDOWN_BIT,
4353 	NETIF_MSG_IFUP_BIT,
4354 	NETIF_MSG_RX_ERR_BIT,
4355 	NETIF_MSG_TX_ERR_BIT,
4356 	NETIF_MSG_TX_QUEUED_BIT,
4357 	NETIF_MSG_INTR_BIT,
4358 	NETIF_MSG_TX_DONE_BIT,
4359 	NETIF_MSG_RX_STATUS_BIT,
4360 	NETIF_MSG_PKTDATA_BIT,
4361 	NETIF_MSG_HW_BIT,
4362 	NETIF_MSG_WOL_BIT,
4363 
4364 	/* When you add a new bit above, update netif_msg_class_names array
4365 	 * in net/ethtool/common.c
4366 	 */
4367 	NETIF_MSG_CLASS_COUNT,
4368 };
4369 /* Both ethtool_ops interface and internal driver implementation use u32 */
4370 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4371 
4372 #define __NETIF_MSG_BIT(bit)	((u32)1 << (bit))
4373 #define __NETIF_MSG(name)	__NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4374 
4375 #define NETIF_MSG_DRV		__NETIF_MSG(DRV)
4376 #define NETIF_MSG_PROBE		__NETIF_MSG(PROBE)
4377 #define NETIF_MSG_LINK		__NETIF_MSG(LINK)
4378 #define NETIF_MSG_TIMER		__NETIF_MSG(TIMER)
4379 #define NETIF_MSG_IFDOWN	__NETIF_MSG(IFDOWN)
4380 #define NETIF_MSG_IFUP		__NETIF_MSG(IFUP)
4381 #define NETIF_MSG_RX_ERR	__NETIF_MSG(RX_ERR)
4382 #define NETIF_MSG_TX_ERR	__NETIF_MSG(TX_ERR)
4383 #define NETIF_MSG_TX_QUEUED	__NETIF_MSG(TX_QUEUED)
4384 #define NETIF_MSG_INTR		__NETIF_MSG(INTR)
4385 #define NETIF_MSG_TX_DONE	__NETIF_MSG(TX_DONE)
4386 #define NETIF_MSG_RX_STATUS	__NETIF_MSG(RX_STATUS)
4387 #define NETIF_MSG_PKTDATA	__NETIF_MSG(PKTDATA)
4388 #define NETIF_MSG_HW		__NETIF_MSG(HW)
4389 #define NETIF_MSG_WOL		__NETIF_MSG(WOL)
4390 
4391 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
4392 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
4393 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
4394 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
4395 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
4396 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
4397 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
4398 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
4399 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4400 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
4401 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
4402 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
4403 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
4404 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
4405 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
4406 
4407 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4408 {
4409 	/* use default */
4410 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4411 		return default_msg_enable_bits;
4412 	if (debug_value == 0)	/* no output */
4413 		return 0;
4414 	/* set low N bits */
4415 	return (1U << debug_value) - 1;
4416 }
4417 
4418 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4419 {
4420 	spin_lock(&txq->_xmit_lock);
4421 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4422 	WRITE_ONCE(txq->xmit_lock_owner, cpu);
4423 }
4424 
4425 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4426 {
4427 	__acquire(&txq->_xmit_lock);
4428 	return true;
4429 }
4430 
4431 static inline void __netif_tx_release(struct netdev_queue *txq)
4432 {
4433 	__release(&txq->_xmit_lock);
4434 }
4435 
4436 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4437 {
4438 	spin_lock_bh(&txq->_xmit_lock);
4439 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4440 	WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4441 }
4442 
4443 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4444 {
4445 	bool ok = spin_trylock(&txq->_xmit_lock);
4446 
4447 	if (likely(ok)) {
4448 		/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4449 		WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4450 	}
4451 	return ok;
4452 }
4453 
4454 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4455 {
4456 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4457 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4458 	spin_unlock(&txq->_xmit_lock);
4459 }
4460 
4461 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4462 {
4463 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4464 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4465 	spin_unlock_bh(&txq->_xmit_lock);
4466 }
4467 
4468 /*
4469  * txq->trans_start can be read locklessly from dev_watchdog()
4470  */
4471 static inline void txq_trans_update(struct netdev_queue *txq)
4472 {
4473 	if (txq->xmit_lock_owner != -1)
4474 		WRITE_ONCE(txq->trans_start, jiffies);
4475 }
4476 
4477 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4478 {
4479 	unsigned long now = jiffies;
4480 
4481 	if (READ_ONCE(txq->trans_start) != now)
4482 		WRITE_ONCE(txq->trans_start, now);
4483 }
4484 
4485 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4486 static inline void netif_trans_update(struct net_device *dev)
4487 {
4488 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4489 
4490 	txq_trans_cond_update(txq);
4491 }
4492 
4493 /**
4494  *	netif_tx_lock - grab network device transmit lock
4495  *	@dev: network device
4496  *
4497  * Get network device transmit lock
4498  */
4499 void netif_tx_lock(struct net_device *dev);
4500 
4501 static inline void netif_tx_lock_bh(struct net_device *dev)
4502 {
4503 	local_bh_disable();
4504 	netif_tx_lock(dev);
4505 }
4506 
4507 void netif_tx_unlock(struct net_device *dev);
4508 
4509 static inline void netif_tx_unlock_bh(struct net_device *dev)
4510 {
4511 	netif_tx_unlock(dev);
4512 	local_bh_enable();
4513 }
4514 
4515 #define HARD_TX_LOCK(dev, txq, cpu) {			\
4516 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4517 		__netif_tx_lock(txq, cpu);		\
4518 	} else {					\
4519 		__netif_tx_acquire(txq);		\
4520 	}						\
4521 }
4522 
4523 #define HARD_TX_TRYLOCK(dev, txq)			\
4524 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
4525 		__netif_tx_trylock(txq) :		\
4526 		__netif_tx_acquire(txq))
4527 
4528 #define HARD_TX_UNLOCK(dev, txq) {			\
4529 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4530 		__netif_tx_unlock(txq);			\
4531 	} else {					\
4532 		__netif_tx_release(txq);		\
4533 	}						\
4534 }
4535 
4536 static inline void netif_tx_disable(struct net_device *dev)
4537 {
4538 	unsigned int i;
4539 	int cpu;
4540 
4541 	local_bh_disable();
4542 	cpu = smp_processor_id();
4543 	spin_lock(&dev->tx_global_lock);
4544 	for (i = 0; i < dev->num_tx_queues; i++) {
4545 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4546 
4547 		__netif_tx_lock(txq, cpu);
4548 		netif_tx_stop_queue(txq);
4549 		__netif_tx_unlock(txq);
4550 	}
4551 	spin_unlock(&dev->tx_global_lock);
4552 	local_bh_enable();
4553 }
4554 
4555 static inline void netif_addr_lock(struct net_device *dev)
4556 {
4557 	unsigned char nest_level = 0;
4558 
4559 #ifdef CONFIG_LOCKDEP
4560 	nest_level = dev->nested_level;
4561 #endif
4562 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4563 }
4564 
4565 static inline void netif_addr_lock_bh(struct net_device *dev)
4566 {
4567 	unsigned char nest_level = 0;
4568 
4569 #ifdef CONFIG_LOCKDEP
4570 	nest_level = dev->nested_level;
4571 #endif
4572 	local_bh_disable();
4573 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4574 }
4575 
4576 static inline void netif_addr_unlock(struct net_device *dev)
4577 {
4578 	spin_unlock(&dev->addr_list_lock);
4579 }
4580 
4581 static inline void netif_addr_unlock_bh(struct net_device *dev)
4582 {
4583 	spin_unlock_bh(&dev->addr_list_lock);
4584 }
4585 
4586 /*
4587  * dev_addrs walker. Should be used only for read access. Call with
4588  * rcu_read_lock held.
4589  */
4590 #define for_each_dev_addr(dev, ha) \
4591 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4592 
4593 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4594 
4595 void ether_setup(struct net_device *dev);
4596 
4597 /* Support for loadable net-drivers */
4598 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4599 				    unsigned char name_assign_type,
4600 				    void (*setup)(struct net_device *),
4601 				    unsigned int txqs, unsigned int rxqs);
4602 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4603 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4604 
4605 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4606 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4607 			 count)
4608 
4609 int register_netdev(struct net_device *dev);
4610 void unregister_netdev(struct net_device *dev);
4611 
4612 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4613 
4614 /* General hardware address lists handling functions */
4615 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4616 		   struct netdev_hw_addr_list *from_list, int addr_len);
4617 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4618 		      struct netdev_hw_addr_list *from_list, int addr_len);
4619 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4620 		       struct net_device *dev,
4621 		       int (*sync)(struct net_device *, const unsigned char *),
4622 		       int (*unsync)(struct net_device *,
4623 				     const unsigned char *));
4624 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4625 			   struct net_device *dev,
4626 			   int (*sync)(struct net_device *,
4627 				       const unsigned char *, int),
4628 			   int (*unsync)(struct net_device *,
4629 					 const unsigned char *, int));
4630 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4631 			      struct net_device *dev,
4632 			      int (*unsync)(struct net_device *,
4633 					    const unsigned char *, int));
4634 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4635 			  struct net_device *dev,
4636 			  int (*unsync)(struct net_device *,
4637 					const unsigned char *));
4638 void __hw_addr_init(struct netdev_hw_addr_list *list);
4639 
4640 /* Functions used for device addresses handling */
4641 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4642 		  const void *addr, size_t len);
4643 
4644 static inline void
4645 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4646 {
4647 	dev_addr_mod(dev, 0, addr, len);
4648 }
4649 
4650 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4651 {
4652 	__dev_addr_set(dev, addr, dev->addr_len);
4653 }
4654 
4655 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4656 		 unsigned char addr_type);
4657 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4658 		 unsigned char addr_type);
4659 
4660 /* Functions used for unicast addresses handling */
4661 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4662 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4663 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4664 int dev_uc_sync(struct net_device *to, struct net_device *from);
4665 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4666 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4667 void dev_uc_flush(struct net_device *dev);
4668 void dev_uc_init(struct net_device *dev);
4669 
4670 /**
4671  *  __dev_uc_sync - Synchonize device's unicast list
4672  *  @dev:  device to sync
4673  *  @sync: function to call if address should be added
4674  *  @unsync: function to call if address should be removed
4675  *
4676  *  Add newly added addresses to the interface, and release
4677  *  addresses that have been deleted.
4678  */
4679 static inline int __dev_uc_sync(struct net_device *dev,
4680 				int (*sync)(struct net_device *,
4681 					    const unsigned char *),
4682 				int (*unsync)(struct net_device *,
4683 					      const unsigned char *))
4684 {
4685 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4686 }
4687 
4688 /**
4689  *  __dev_uc_unsync - Remove synchronized addresses from device
4690  *  @dev:  device to sync
4691  *  @unsync: function to call if address should be removed
4692  *
4693  *  Remove all addresses that were added to the device by dev_uc_sync().
4694  */
4695 static inline void __dev_uc_unsync(struct net_device *dev,
4696 				   int (*unsync)(struct net_device *,
4697 						 const unsigned char *))
4698 {
4699 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
4700 }
4701 
4702 /* Functions used for multicast addresses handling */
4703 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4704 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4705 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4706 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4707 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4708 int dev_mc_sync(struct net_device *to, struct net_device *from);
4709 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4710 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4711 void dev_mc_flush(struct net_device *dev);
4712 void dev_mc_init(struct net_device *dev);
4713 
4714 /**
4715  *  __dev_mc_sync - Synchonize device's multicast list
4716  *  @dev:  device to sync
4717  *  @sync: function to call if address should be added
4718  *  @unsync: function to call if address should be removed
4719  *
4720  *  Add newly added addresses to the interface, and release
4721  *  addresses that have been deleted.
4722  */
4723 static inline int __dev_mc_sync(struct net_device *dev,
4724 				int (*sync)(struct net_device *,
4725 					    const unsigned char *),
4726 				int (*unsync)(struct net_device *,
4727 					      const unsigned char *))
4728 {
4729 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4730 }
4731 
4732 /**
4733  *  __dev_mc_unsync - Remove synchronized addresses from device
4734  *  @dev:  device to sync
4735  *  @unsync: function to call if address should be removed
4736  *
4737  *  Remove all addresses that were added to the device by dev_mc_sync().
4738  */
4739 static inline void __dev_mc_unsync(struct net_device *dev,
4740 				   int (*unsync)(struct net_device *,
4741 						 const unsigned char *))
4742 {
4743 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
4744 }
4745 
4746 /* Functions used for secondary unicast and multicast support */
4747 void dev_set_rx_mode(struct net_device *dev);
4748 int dev_set_promiscuity(struct net_device *dev, int inc);
4749 int dev_set_allmulti(struct net_device *dev, int inc);
4750 void netdev_state_change(struct net_device *dev);
4751 void __netdev_notify_peers(struct net_device *dev);
4752 void netdev_notify_peers(struct net_device *dev);
4753 void netdev_features_change(struct net_device *dev);
4754 /* Load a device via the kmod */
4755 void dev_load(struct net *net, const char *name);
4756 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4757 					struct rtnl_link_stats64 *storage);
4758 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4759 			     const struct net_device_stats *netdev_stats);
4760 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4761 			   const struct pcpu_sw_netstats __percpu *netstats);
4762 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4763 
4764 extern int		netdev_max_backlog;
4765 extern int		dev_rx_weight;
4766 extern int		dev_tx_weight;
4767 extern int		gro_normal_batch;
4768 
4769 enum {
4770 	NESTED_SYNC_IMM_BIT,
4771 	NESTED_SYNC_TODO_BIT,
4772 };
4773 
4774 #define __NESTED_SYNC_BIT(bit)	((u32)1 << (bit))
4775 #define __NESTED_SYNC(name)	__NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4776 
4777 #define NESTED_SYNC_IMM		__NESTED_SYNC(IMM)
4778 #define NESTED_SYNC_TODO	__NESTED_SYNC(TODO)
4779 
4780 struct netdev_nested_priv {
4781 	unsigned char flags;
4782 	void *data;
4783 };
4784 
4785 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4786 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4787 						     struct list_head **iter);
4788 
4789 /* iterate through upper list, must be called under RCU read lock */
4790 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4791 	for (iter = &(dev)->adj_list.upper, \
4792 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4793 	     updev; \
4794 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4795 
4796 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4797 				  int (*fn)(struct net_device *upper_dev,
4798 					    struct netdev_nested_priv *priv),
4799 				  struct netdev_nested_priv *priv);
4800 
4801 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4802 				  struct net_device *upper_dev);
4803 
4804 bool netdev_has_any_upper_dev(struct net_device *dev);
4805 
4806 void *netdev_lower_get_next_private(struct net_device *dev,
4807 				    struct list_head **iter);
4808 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4809 					struct list_head **iter);
4810 
4811 #define netdev_for_each_lower_private(dev, priv, iter) \
4812 	for (iter = (dev)->adj_list.lower.next, \
4813 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
4814 	     priv; \
4815 	     priv = netdev_lower_get_next_private(dev, &(iter)))
4816 
4817 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4818 	for (iter = &(dev)->adj_list.lower, \
4819 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4820 	     priv; \
4821 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4822 
4823 void *netdev_lower_get_next(struct net_device *dev,
4824 				struct list_head **iter);
4825 
4826 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4827 	for (iter = (dev)->adj_list.lower.next, \
4828 	     ldev = netdev_lower_get_next(dev, &(iter)); \
4829 	     ldev; \
4830 	     ldev = netdev_lower_get_next(dev, &(iter)))
4831 
4832 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4833 					     struct list_head **iter);
4834 int netdev_walk_all_lower_dev(struct net_device *dev,
4835 			      int (*fn)(struct net_device *lower_dev,
4836 					struct netdev_nested_priv *priv),
4837 			      struct netdev_nested_priv *priv);
4838 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4839 				  int (*fn)(struct net_device *lower_dev,
4840 					    struct netdev_nested_priv *priv),
4841 				  struct netdev_nested_priv *priv);
4842 
4843 void *netdev_adjacent_get_private(struct list_head *adj_list);
4844 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4845 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4846 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4847 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4848 			  struct netlink_ext_ack *extack);
4849 int netdev_master_upper_dev_link(struct net_device *dev,
4850 				 struct net_device *upper_dev,
4851 				 void *upper_priv, void *upper_info,
4852 				 struct netlink_ext_ack *extack);
4853 void netdev_upper_dev_unlink(struct net_device *dev,
4854 			     struct net_device *upper_dev);
4855 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4856 				   struct net_device *new_dev,
4857 				   struct net_device *dev,
4858 				   struct netlink_ext_ack *extack);
4859 void netdev_adjacent_change_commit(struct net_device *old_dev,
4860 				   struct net_device *new_dev,
4861 				   struct net_device *dev);
4862 void netdev_adjacent_change_abort(struct net_device *old_dev,
4863 				  struct net_device *new_dev,
4864 				  struct net_device *dev);
4865 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4866 void *netdev_lower_dev_get_private(struct net_device *dev,
4867 				   struct net_device *lower_dev);
4868 void netdev_lower_state_changed(struct net_device *lower_dev,
4869 				void *lower_state_info);
4870 
4871 /* RSS keys are 40 or 52 bytes long */
4872 #define NETDEV_RSS_KEY_LEN 52
4873 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4874 void netdev_rss_key_fill(void *buffer, size_t len);
4875 
4876 int skb_checksum_help(struct sk_buff *skb);
4877 int skb_crc32c_csum_help(struct sk_buff *skb);
4878 int skb_csum_hwoffload_help(struct sk_buff *skb,
4879 			    const netdev_features_t features);
4880 
4881 struct netdev_bonding_info {
4882 	ifslave	slave;
4883 	ifbond	master;
4884 };
4885 
4886 struct netdev_notifier_bonding_info {
4887 	struct netdev_notifier_info info; /* must be first */
4888 	struct netdev_bonding_info  bonding_info;
4889 };
4890 
4891 void netdev_bonding_info_change(struct net_device *dev,
4892 				struct netdev_bonding_info *bonding_info);
4893 
4894 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4895 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4896 #else
4897 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4898 				  const void *data)
4899 {
4900 }
4901 #endif
4902 
4903 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4904 
4905 static inline bool can_checksum_protocol(netdev_features_t features,
4906 					 __be16 protocol)
4907 {
4908 	if (protocol == htons(ETH_P_FCOE))
4909 		return !!(features & NETIF_F_FCOE_CRC);
4910 
4911 	/* Assume this is an IP checksum (not SCTP CRC) */
4912 
4913 	if (features & NETIF_F_HW_CSUM) {
4914 		/* Can checksum everything */
4915 		return true;
4916 	}
4917 
4918 	switch (protocol) {
4919 	case htons(ETH_P_IP):
4920 		return !!(features & NETIF_F_IP_CSUM);
4921 	case htons(ETH_P_IPV6):
4922 		return !!(features & NETIF_F_IPV6_CSUM);
4923 	default:
4924 		return false;
4925 	}
4926 }
4927 
4928 #ifdef CONFIG_BUG
4929 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4930 #else
4931 static inline void netdev_rx_csum_fault(struct net_device *dev,
4932 					struct sk_buff *skb)
4933 {
4934 }
4935 #endif
4936 /* rx skb timestamps */
4937 void net_enable_timestamp(void);
4938 void net_disable_timestamp(void);
4939 
4940 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4941 					const struct skb_shared_hwtstamps *hwtstamps,
4942 					bool cycles)
4943 {
4944 	const struct net_device_ops *ops = dev->netdev_ops;
4945 
4946 	if (ops->ndo_get_tstamp)
4947 		return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4948 
4949 	return hwtstamps->hwtstamp;
4950 }
4951 
4952 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4953 					      struct sk_buff *skb, struct net_device *dev,
4954 					      bool more)
4955 {
4956 	__this_cpu_write(softnet_data.xmit.more, more);
4957 	return ops->ndo_start_xmit(skb, dev);
4958 }
4959 
4960 static inline bool netdev_xmit_more(void)
4961 {
4962 	return __this_cpu_read(softnet_data.xmit.more);
4963 }
4964 
4965 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4966 					    struct netdev_queue *txq, bool more)
4967 {
4968 	const struct net_device_ops *ops = dev->netdev_ops;
4969 	netdev_tx_t rc;
4970 
4971 	rc = __netdev_start_xmit(ops, skb, dev, more);
4972 	if (rc == NETDEV_TX_OK)
4973 		txq_trans_update(txq);
4974 
4975 	return rc;
4976 }
4977 
4978 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4979 				const void *ns);
4980 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4981 				 const void *ns);
4982 
4983 extern const struct kobj_ns_type_operations net_ns_type_operations;
4984 
4985 const char *netdev_drivername(const struct net_device *dev);
4986 
4987 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4988 							  netdev_features_t f2)
4989 {
4990 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4991 		if (f1 & NETIF_F_HW_CSUM)
4992 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4993 		else
4994 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4995 	}
4996 
4997 	return f1 & f2;
4998 }
4999 
5000 static inline netdev_features_t netdev_get_wanted_features(
5001 	struct net_device *dev)
5002 {
5003 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
5004 }
5005 netdev_features_t netdev_increment_features(netdev_features_t all,
5006 	netdev_features_t one, netdev_features_t mask);
5007 
5008 /* Allow TSO being used on stacked device :
5009  * Performing the GSO segmentation before last device
5010  * is a performance improvement.
5011  */
5012 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5013 							netdev_features_t mask)
5014 {
5015 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5016 }
5017 
5018 int __netdev_update_features(struct net_device *dev);
5019 void netdev_update_features(struct net_device *dev);
5020 void netdev_change_features(struct net_device *dev);
5021 
5022 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5023 					struct net_device *dev);
5024 
5025 netdev_features_t passthru_features_check(struct sk_buff *skb,
5026 					  struct net_device *dev,
5027 					  netdev_features_t features);
5028 netdev_features_t netif_skb_features(struct sk_buff *skb);
5029 void skb_warn_bad_offload(const struct sk_buff *skb);
5030 
5031 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5032 {
5033 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5034 
5035 	/* check flags correspondence */
5036 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5037 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5038 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5039 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5040 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5041 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5042 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5043 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5044 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5045 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5046 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5047 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5048 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5049 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5050 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5051 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5052 	BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5053 	BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5054 	BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5055 
5056 	return (features & feature) == feature;
5057 }
5058 
5059 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5060 {
5061 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5062 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5063 }
5064 
5065 static inline bool netif_needs_gso(struct sk_buff *skb,
5066 				   netdev_features_t features)
5067 {
5068 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5069 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5070 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5071 }
5072 
5073 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5074 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5075 void netif_inherit_tso_max(struct net_device *to,
5076 			   const struct net_device *from);
5077 
5078 static inline bool netif_is_macsec(const struct net_device *dev)
5079 {
5080 	return dev->priv_flags & IFF_MACSEC;
5081 }
5082 
5083 static inline bool netif_is_macvlan(const struct net_device *dev)
5084 {
5085 	return dev->priv_flags & IFF_MACVLAN;
5086 }
5087 
5088 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5089 {
5090 	return dev->priv_flags & IFF_MACVLAN_PORT;
5091 }
5092 
5093 static inline bool netif_is_bond_master(const struct net_device *dev)
5094 {
5095 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5096 }
5097 
5098 static inline bool netif_is_bond_slave(const struct net_device *dev)
5099 {
5100 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5101 }
5102 
5103 static inline bool netif_supports_nofcs(struct net_device *dev)
5104 {
5105 	return dev->priv_flags & IFF_SUPP_NOFCS;
5106 }
5107 
5108 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5109 {
5110 	return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5111 }
5112 
5113 static inline bool netif_is_l3_master(const struct net_device *dev)
5114 {
5115 	return dev->priv_flags & IFF_L3MDEV_MASTER;
5116 }
5117 
5118 static inline bool netif_is_l3_slave(const struct net_device *dev)
5119 {
5120 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
5121 }
5122 
5123 static inline int dev_sdif(const struct net_device *dev)
5124 {
5125 #ifdef CONFIG_NET_L3_MASTER_DEV
5126 	if (netif_is_l3_slave(dev))
5127 		return dev->ifindex;
5128 #endif
5129 	return 0;
5130 }
5131 
5132 static inline bool netif_is_bridge_master(const struct net_device *dev)
5133 {
5134 	return dev->priv_flags & IFF_EBRIDGE;
5135 }
5136 
5137 static inline bool netif_is_bridge_port(const struct net_device *dev)
5138 {
5139 	return dev->priv_flags & IFF_BRIDGE_PORT;
5140 }
5141 
5142 static inline bool netif_is_ovs_master(const struct net_device *dev)
5143 {
5144 	return dev->priv_flags & IFF_OPENVSWITCH;
5145 }
5146 
5147 static inline bool netif_is_ovs_port(const struct net_device *dev)
5148 {
5149 	return dev->priv_flags & IFF_OVS_DATAPATH;
5150 }
5151 
5152 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5153 {
5154 	return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5155 }
5156 
5157 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5158 {
5159 	return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5160 }
5161 
5162 static inline bool netif_is_team_master(const struct net_device *dev)
5163 {
5164 	return dev->priv_flags & IFF_TEAM;
5165 }
5166 
5167 static inline bool netif_is_team_port(const struct net_device *dev)
5168 {
5169 	return dev->priv_flags & IFF_TEAM_PORT;
5170 }
5171 
5172 static inline bool netif_is_lag_master(const struct net_device *dev)
5173 {
5174 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
5175 }
5176 
5177 static inline bool netif_is_lag_port(const struct net_device *dev)
5178 {
5179 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5180 }
5181 
5182 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5183 {
5184 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
5185 }
5186 
5187 static inline bool netif_is_failover(const struct net_device *dev)
5188 {
5189 	return dev->priv_flags & IFF_FAILOVER;
5190 }
5191 
5192 static inline bool netif_is_failover_slave(const struct net_device *dev)
5193 {
5194 	return dev->priv_flags & IFF_FAILOVER_SLAVE;
5195 }
5196 
5197 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5198 static inline void netif_keep_dst(struct net_device *dev)
5199 {
5200 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5201 }
5202 
5203 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5204 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5205 {
5206 	/* TODO: reserve and use an additional IFF bit, if we get more users */
5207 	return netif_is_macsec(dev);
5208 }
5209 
5210 extern struct pernet_operations __net_initdata loopback_net_ops;
5211 
5212 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5213 
5214 /* netdev_printk helpers, similar to dev_printk */
5215 
5216 static inline const char *netdev_name(const struct net_device *dev)
5217 {
5218 	if (!dev->name[0] || strchr(dev->name, '%'))
5219 		return "(unnamed net_device)";
5220 	return dev->name;
5221 }
5222 
5223 static inline const char *netdev_reg_state(const struct net_device *dev)
5224 {
5225 	switch (dev->reg_state) {
5226 	case NETREG_UNINITIALIZED: return " (uninitialized)";
5227 	case NETREG_REGISTERED: return "";
5228 	case NETREG_UNREGISTERING: return " (unregistering)";
5229 	case NETREG_UNREGISTERED: return " (unregistered)";
5230 	case NETREG_RELEASED: return " (released)";
5231 	case NETREG_DUMMY: return " (dummy)";
5232 	}
5233 
5234 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5235 	return " (unknown)";
5236 }
5237 
5238 #define MODULE_ALIAS_NETDEV(device) \
5239 	MODULE_ALIAS("netdev-" device)
5240 
5241 /*
5242  * netdev_WARN() acts like dev_printk(), but with the key difference
5243  * of using a WARN/WARN_ON to get the message out, including the
5244  * file/line information and a backtrace.
5245  */
5246 #define netdev_WARN(dev, format, args...)			\
5247 	WARN(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5248 	     netdev_reg_state(dev), ##args)
5249 
5250 #define netdev_WARN_ONCE(dev, format, args...)				\
5251 	WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5252 		  netdev_reg_state(dev), ##args)
5253 
5254 /*
5255  *	The list of packet types we will receive (as opposed to discard)
5256  *	and the routines to invoke.
5257  *
5258  *	Why 16. Because with 16 the only overlap we get on a hash of the
5259  *	low nibble of the protocol value is RARP/SNAP/X.25.
5260  *
5261  *		0800	IP
5262  *		0001	802.3
5263  *		0002	AX.25
5264  *		0004	802.2
5265  *		8035	RARP
5266  *		0005	SNAP
5267  *		0805	X.25
5268  *		0806	ARP
5269  *		8137	IPX
5270  *		0009	Localtalk
5271  *		86DD	IPv6
5272  */
5273 #define PTYPE_HASH_SIZE	(16)
5274 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
5275 
5276 extern struct list_head ptype_all __read_mostly;
5277 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5278 
5279 extern struct net_device *blackhole_netdev;
5280 
5281 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5282 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5283 #define DEV_STATS_ADD(DEV, FIELD, VAL) 	\
5284 		atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5285 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5286 
5287 #endif	/* _LINUX_NETDEVICE_H */
5288