xref: /linux-6.15/include/linux/netdevice.h (revision 6b7a2114)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Definitions for the Interfaces handler.
7  *
8  * Version:	@(#)dev.h	1.0.10	08/12/93
9  *
10  * Authors:	Ross Biro
11  *		Fred N. van Kempen, <[email protected]>
12  *		Corey Minyard <[email protected]>
13  *		Donald J. Becker, <[email protected]>
14  *		Alan Cox, <[email protected]>
15  *		Bjorn Ekwall. <[email protected]>
16  *              Pekka Riikonen <[email protected]>
17  *
18  *		This program is free software; you can redistribute it and/or
19  *		modify it under the terms of the GNU General Public License
20  *		as published by the Free Software Foundation; either version
21  *		2 of the License, or (at your option) any later version.
22  *
23  *		Moved to /usr/include/linux for NET3
24  */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27 
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35 
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/workqueue.h>
39 #include <linux/dynamic_queue_limits.h>
40 
41 #include <linux/ethtool.h>
42 #include <net/net_namespace.h>
43 #ifdef CONFIG_DCB
44 #include <net/dcbnl.h>
45 #endif
46 #include <net/netprio_cgroup.h>
47 #include <net/xdp.h>
48 
49 #include <linux/netdev_features.h>
50 #include <linux/neighbour.h>
51 #include <uapi/linux/netdevice.h>
52 #include <uapi/linux/if_bonding.h>
53 #include <uapi/linux/pkt_cls.h>
54 #include <linux/hashtable.h>
55 
56 struct netpoll_info;
57 struct device;
58 struct phy_device;
59 struct dsa_port;
60 
61 struct sfp_bus;
62 /* 802.11 specific */
63 struct wireless_dev;
64 /* 802.15.4 specific */
65 struct wpan_dev;
66 struct mpls_dev;
67 /* UDP Tunnel offloads */
68 struct udp_tunnel_info;
69 struct bpf_prog;
70 struct xdp_buff;
71 
72 void netdev_set_default_ethtool_ops(struct net_device *dev,
73 				    const struct ethtool_ops *ops);
74 
75 /* Backlog congestion levels */
76 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
77 #define NET_RX_DROP		1	/* packet dropped */
78 
79 /*
80  * Transmit return codes: transmit return codes originate from three different
81  * namespaces:
82  *
83  * - qdisc return codes
84  * - driver transmit return codes
85  * - errno values
86  *
87  * Drivers are allowed to return any one of those in their hard_start_xmit()
88  * function. Real network devices commonly used with qdiscs should only return
89  * the driver transmit return codes though - when qdiscs are used, the actual
90  * transmission happens asynchronously, so the value is not propagated to
91  * higher layers. Virtual network devices transmit synchronously; in this case
92  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
93  * others are propagated to higher layers.
94  */
95 
96 /* qdisc ->enqueue() return codes. */
97 #define NET_XMIT_SUCCESS	0x00
98 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
99 #define NET_XMIT_CN		0x02	/* congestion notification	*/
100 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
101 
102 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
103  * indicates that the device will soon be dropping packets, or already drops
104  * some packets of the same priority; prompting us to send less aggressively. */
105 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
106 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
107 
108 /* Driver transmit return codes */
109 #define NETDEV_TX_MASK		0xf0
110 
111 enum netdev_tx {
112 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
113 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
114 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
115 };
116 typedef enum netdev_tx netdev_tx_t;
117 
118 /*
119  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121  */
122 static inline bool dev_xmit_complete(int rc)
123 {
124 	/*
125 	 * Positive cases with an skb consumed by a driver:
126 	 * - successful transmission (rc == NETDEV_TX_OK)
127 	 * - error while transmitting (rc < 0)
128 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 	 */
130 	if (likely(rc < NET_XMIT_MASK))
131 		return true;
132 
133 	return false;
134 }
135 
136 /*
137  *	Compute the worst-case header length according to the protocols
138  *	used.
139  */
140 
141 #if defined(CONFIG_HYPERV_NET)
142 # define LL_MAX_HEADER 128
143 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
144 # if defined(CONFIG_MAC80211_MESH)
145 #  define LL_MAX_HEADER 128
146 # else
147 #  define LL_MAX_HEADER 96
148 # endif
149 #else
150 # define LL_MAX_HEADER 32
151 #endif
152 
153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
154     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
155 #define MAX_HEADER LL_MAX_HEADER
156 #else
157 #define MAX_HEADER (LL_MAX_HEADER + 48)
158 #endif
159 
160 /*
161  *	Old network device statistics. Fields are native words
162  *	(unsigned long) so they can be read and written atomically.
163  */
164 
165 struct net_device_stats {
166 	unsigned long	rx_packets;
167 	unsigned long	tx_packets;
168 	unsigned long	rx_bytes;
169 	unsigned long	tx_bytes;
170 	unsigned long	rx_errors;
171 	unsigned long	tx_errors;
172 	unsigned long	rx_dropped;
173 	unsigned long	tx_dropped;
174 	unsigned long	multicast;
175 	unsigned long	collisions;
176 	unsigned long	rx_length_errors;
177 	unsigned long	rx_over_errors;
178 	unsigned long	rx_crc_errors;
179 	unsigned long	rx_frame_errors;
180 	unsigned long	rx_fifo_errors;
181 	unsigned long	rx_missed_errors;
182 	unsigned long	tx_aborted_errors;
183 	unsigned long	tx_carrier_errors;
184 	unsigned long	tx_fifo_errors;
185 	unsigned long	tx_heartbeat_errors;
186 	unsigned long	tx_window_errors;
187 	unsigned long	rx_compressed;
188 	unsigned long	tx_compressed;
189 };
190 
191 
192 #include <linux/cache.h>
193 #include <linux/skbuff.h>
194 
195 #ifdef CONFIG_RPS
196 #include <linux/static_key.h>
197 extern struct static_key_false rps_needed;
198 extern struct static_key_false rfs_needed;
199 #endif
200 
201 struct neighbour;
202 struct neigh_parms;
203 struct sk_buff;
204 
205 struct netdev_hw_addr {
206 	struct list_head	list;
207 	unsigned char		addr[MAX_ADDR_LEN];
208 	unsigned char		type;
209 #define NETDEV_HW_ADDR_T_LAN		1
210 #define NETDEV_HW_ADDR_T_SAN		2
211 #define NETDEV_HW_ADDR_T_SLAVE		3
212 #define NETDEV_HW_ADDR_T_UNICAST	4
213 #define NETDEV_HW_ADDR_T_MULTICAST	5
214 	bool			global_use;
215 	int			sync_cnt;
216 	int			refcount;
217 	int			synced;
218 	struct rcu_head		rcu_head;
219 };
220 
221 struct netdev_hw_addr_list {
222 	struct list_head	list;
223 	int			count;
224 };
225 
226 #define netdev_hw_addr_list_count(l) ((l)->count)
227 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
228 #define netdev_hw_addr_list_for_each(ha, l) \
229 	list_for_each_entry(ha, &(l)->list, list)
230 
231 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
232 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
233 #define netdev_for_each_uc_addr(ha, dev) \
234 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
235 
236 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
237 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
238 #define netdev_for_each_mc_addr(ha, dev) \
239 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
240 
241 struct hh_cache {
242 	unsigned int	hh_len;
243 	seqlock_t	hh_lock;
244 
245 	/* cached hardware header; allow for machine alignment needs.        */
246 #define HH_DATA_MOD	16
247 #define HH_DATA_OFF(__len) \
248 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
249 #define HH_DATA_ALIGN(__len) \
250 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
251 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
252 };
253 
254 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
255  * Alternative is:
256  *   dev->hard_header_len ? (dev->hard_header_len +
257  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
258  *
259  * We could use other alignment values, but we must maintain the
260  * relationship HH alignment <= LL alignment.
261  */
262 #define LL_RESERVED_SPACE(dev) \
263 	((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
265 	((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
266 
267 struct header_ops {
268 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
269 			   unsigned short type, const void *daddr,
270 			   const void *saddr, unsigned int len);
271 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
272 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
273 	void	(*cache_update)(struct hh_cache *hh,
274 				const struct net_device *dev,
275 				const unsigned char *haddr);
276 	bool	(*validate)(const char *ll_header, unsigned int len);
277 	__be16	(*parse_protocol)(const struct sk_buff *skb);
278 };
279 
280 /* These flag bits are private to the generic network queueing
281  * layer; they may not be explicitly referenced by any other
282  * code.
283  */
284 
285 enum netdev_state_t {
286 	__LINK_STATE_START,
287 	__LINK_STATE_PRESENT,
288 	__LINK_STATE_NOCARRIER,
289 	__LINK_STATE_LINKWATCH_PENDING,
290 	__LINK_STATE_DORMANT,
291 };
292 
293 
294 /*
295  * This structure holds boot-time configured netdevice settings. They
296  * are then used in the device probing.
297  */
298 struct netdev_boot_setup {
299 	char name[IFNAMSIZ];
300 	struct ifmap map;
301 };
302 #define NETDEV_BOOT_SETUP_MAX 8
303 
304 int __init netdev_boot_setup(char *str);
305 
306 struct gro_list {
307 	struct list_head	list;
308 	int			count;
309 };
310 
311 /*
312  * size of gro hash buckets, must less than bit number of
313  * napi_struct::gro_bitmask
314  */
315 #define GRO_HASH_BUCKETS	8
316 
317 /*
318  * Structure for NAPI scheduling similar to tasklet but with weighting
319  */
320 struct napi_struct {
321 	/* The poll_list must only be managed by the entity which
322 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
323 	 * whoever atomically sets that bit can add this napi_struct
324 	 * to the per-CPU poll_list, and whoever clears that bit
325 	 * can remove from the list right before clearing the bit.
326 	 */
327 	struct list_head	poll_list;
328 
329 	unsigned long		state;
330 	int			weight;
331 	unsigned long		gro_bitmask;
332 	int			(*poll)(struct napi_struct *, int);
333 #ifdef CONFIG_NETPOLL
334 	int			poll_owner;
335 #endif
336 	struct net_device	*dev;
337 	struct gro_list		gro_hash[GRO_HASH_BUCKETS];
338 	struct sk_buff		*skb;
339 	struct hrtimer		timer;
340 	struct list_head	dev_list;
341 	struct hlist_node	napi_hash_node;
342 	unsigned int		napi_id;
343 };
344 
345 enum {
346 	NAPI_STATE_SCHED,	/* Poll is scheduled */
347 	NAPI_STATE_MISSED,	/* reschedule a napi */
348 	NAPI_STATE_DISABLE,	/* Disable pending */
349 	NAPI_STATE_NPSVC,	/* Netpoll - don't dequeue from poll_list */
350 	NAPI_STATE_HASHED,	/* In NAPI hash (busy polling possible) */
351 	NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
352 	NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
353 };
354 
355 enum {
356 	NAPIF_STATE_SCHED	 = BIT(NAPI_STATE_SCHED),
357 	NAPIF_STATE_MISSED	 = BIT(NAPI_STATE_MISSED),
358 	NAPIF_STATE_DISABLE	 = BIT(NAPI_STATE_DISABLE),
359 	NAPIF_STATE_NPSVC	 = BIT(NAPI_STATE_NPSVC),
360 	NAPIF_STATE_HASHED	 = BIT(NAPI_STATE_HASHED),
361 	NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
362 	NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
363 };
364 
365 enum gro_result {
366 	GRO_MERGED,
367 	GRO_MERGED_FREE,
368 	GRO_HELD,
369 	GRO_NORMAL,
370 	GRO_DROP,
371 	GRO_CONSUMED,
372 };
373 typedef enum gro_result gro_result_t;
374 
375 /*
376  * enum rx_handler_result - Possible return values for rx_handlers.
377  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
378  * further.
379  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
380  * case skb->dev was changed by rx_handler.
381  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
382  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
383  *
384  * rx_handlers are functions called from inside __netif_receive_skb(), to do
385  * special processing of the skb, prior to delivery to protocol handlers.
386  *
387  * Currently, a net_device can only have a single rx_handler registered. Trying
388  * to register a second rx_handler will return -EBUSY.
389  *
390  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
391  * To unregister a rx_handler on a net_device, use
392  * netdev_rx_handler_unregister().
393  *
394  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
395  * do with the skb.
396  *
397  * If the rx_handler consumed the skb in some way, it should return
398  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
399  * the skb to be delivered in some other way.
400  *
401  * If the rx_handler changed skb->dev, to divert the skb to another
402  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
403  * new device will be called if it exists.
404  *
405  * If the rx_handler decides the skb should be ignored, it should return
406  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
407  * are registered on exact device (ptype->dev == skb->dev).
408  *
409  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
410  * delivered, it should return RX_HANDLER_PASS.
411  *
412  * A device without a registered rx_handler will behave as if rx_handler
413  * returned RX_HANDLER_PASS.
414  */
415 
416 enum rx_handler_result {
417 	RX_HANDLER_CONSUMED,
418 	RX_HANDLER_ANOTHER,
419 	RX_HANDLER_EXACT,
420 	RX_HANDLER_PASS,
421 };
422 typedef enum rx_handler_result rx_handler_result_t;
423 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
424 
425 void __napi_schedule(struct napi_struct *n);
426 void __napi_schedule_irqoff(struct napi_struct *n);
427 
428 static inline bool napi_disable_pending(struct napi_struct *n)
429 {
430 	return test_bit(NAPI_STATE_DISABLE, &n->state);
431 }
432 
433 bool napi_schedule_prep(struct napi_struct *n);
434 
435 /**
436  *	napi_schedule - schedule NAPI poll
437  *	@n: NAPI context
438  *
439  * Schedule NAPI poll routine to be called if it is not already
440  * running.
441  */
442 static inline void napi_schedule(struct napi_struct *n)
443 {
444 	if (napi_schedule_prep(n))
445 		__napi_schedule(n);
446 }
447 
448 /**
449  *	napi_schedule_irqoff - schedule NAPI poll
450  *	@n: NAPI context
451  *
452  * Variant of napi_schedule(), assuming hard irqs are masked.
453  */
454 static inline void napi_schedule_irqoff(struct napi_struct *n)
455 {
456 	if (napi_schedule_prep(n))
457 		__napi_schedule_irqoff(n);
458 }
459 
460 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
461 static inline bool napi_reschedule(struct napi_struct *napi)
462 {
463 	if (napi_schedule_prep(napi)) {
464 		__napi_schedule(napi);
465 		return true;
466 	}
467 	return false;
468 }
469 
470 bool napi_complete_done(struct napi_struct *n, int work_done);
471 /**
472  *	napi_complete - NAPI processing complete
473  *	@n: NAPI context
474  *
475  * Mark NAPI processing as complete.
476  * Consider using napi_complete_done() instead.
477  * Return false if device should avoid rearming interrupts.
478  */
479 static inline bool napi_complete(struct napi_struct *n)
480 {
481 	return napi_complete_done(n, 0);
482 }
483 
484 /**
485  *	napi_hash_del - remove a NAPI from global table
486  *	@napi: NAPI context
487  *
488  * Warning: caller must observe RCU grace period
489  * before freeing memory containing @napi, if
490  * this function returns true.
491  * Note: core networking stack automatically calls it
492  * from netif_napi_del().
493  * Drivers might want to call this helper to combine all
494  * the needed RCU grace periods into a single one.
495  */
496 bool napi_hash_del(struct napi_struct *napi);
497 
498 /**
499  *	napi_disable - prevent NAPI from scheduling
500  *	@n: NAPI context
501  *
502  * Stop NAPI from being scheduled on this context.
503  * Waits till any outstanding processing completes.
504  */
505 void napi_disable(struct napi_struct *n);
506 
507 /**
508  *	napi_enable - enable NAPI scheduling
509  *	@n: NAPI context
510  *
511  * Resume NAPI from being scheduled on this context.
512  * Must be paired with napi_disable.
513  */
514 static inline void napi_enable(struct napi_struct *n)
515 {
516 	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
517 	smp_mb__before_atomic();
518 	clear_bit(NAPI_STATE_SCHED, &n->state);
519 	clear_bit(NAPI_STATE_NPSVC, &n->state);
520 }
521 
522 /**
523  *	napi_synchronize - wait until NAPI is not running
524  *	@n: NAPI context
525  *
526  * Wait until NAPI is done being scheduled on this context.
527  * Waits till any outstanding processing completes but
528  * does not disable future activations.
529  */
530 static inline void napi_synchronize(const struct napi_struct *n)
531 {
532 	if (IS_ENABLED(CONFIG_SMP))
533 		while (test_bit(NAPI_STATE_SCHED, &n->state))
534 			msleep(1);
535 	else
536 		barrier();
537 }
538 
539 /**
540  *	napi_if_scheduled_mark_missed - if napi is running, set the
541  *	NAPIF_STATE_MISSED
542  *	@n: NAPI context
543  *
544  * If napi is running, set the NAPIF_STATE_MISSED, and return true if
545  * NAPI is scheduled.
546  **/
547 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
548 {
549 	unsigned long val, new;
550 
551 	do {
552 		val = READ_ONCE(n->state);
553 		if (val & NAPIF_STATE_DISABLE)
554 			return true;
555 
556 		if (!(val & NAPIF_STATE_SCHED))
557 			return false;
558 
559 		new = val | NAPIF_STATE_MISSED;
560 	} while (cmpxchg(&n->state, val, new) != val);
561 
562 	return true;
563 }
564 
565 enum netdev_queue_state_t {
566 	__QUEUE_STATE_DRV_XOFF,
567 	__QUEUE_STATE_STACK_XOFF,
568 	__QUEUE_STATE_FROZEN,
569 };
570 
571 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
572 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
573 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
574 
575 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
576 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
577 					QUEUE_STATE_FROZEN)
578 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
579 					QUEUE_STATE_FROZEN)
580 
581 /*
582  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
583  * netif_tx_* functions below are used to manipulate this flag.  The
584  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
585  * queue independently.  The netif_xmit_*stopped functions below are called
586  * to check if the queue has been stopped by the driver or stack (either
587  * of the XOFF bits are set in the state).  Drivers should not need to call
588  * netif_xmit*stopped functions, they should only be using netif_tx_*.
589  */
590 
591 struct netdev_queue {
592 /*
593  * read-mostly part
594  */
595 	struct net_device	*dev;
596 	struct Qdisc __rcu	*qdisc;
597 	struct Qdisc		*qdisc_sleeping;
598 #ifdef CONFIG_SYSFS
599 	struct kobject		kobj;
600 #endif
601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
602 	int			numa_node;
603 #endif
604 	unsigned long		tx_maxrate;
605 	/*
606 	 * Number of TX timeouts for this queue
607 	 * (/sys/class/net/DEV/Q/trans_timeout)
608 	 */
609 	unsigned long		trans_timeout;
610 
611 	/* Subordinate device that the queue has been assigned to */
612 	struct net_device	*sb_dev;
613 #ifdef CONFIG_XDP_SOCKETS
614 	struct xdp_umem         *umem;
615 #endif
616 /*
617  * write-mostly part
618  */
619 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
620 	int			xmit_lock_owner;
621 	/*
622 	 * Time (in jiffies) of last Tx
623 	 */
624 	unsigned long		trans_start;
625 
626 	unsigned long		state;
627 
628 #ifdef CONFIG_BQL
629 	struct dql		dql;
630 #endif
631 } ____cacheline_aligned_in_smp;
632 
633 extern int sysctl_fb_tunnels_only_for_init_net;
634 extern int sysctl_devconf_inherit_init_net;
635 
636 static inline bool net_has_fallback_tunnels(const struct net *net)
637 {
638 	return net == &init_net ||
639 	       !IS_ENABLED(CONFIG_SYSCTL) ||
640 	       !sysctl_fb_tunnels_only_for_init_net;
641 }
642 
643 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
644 {
645 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
646 	return q->numa_node;
647 #else
648 	return NUMA_NO_NODE;
649 #endif
650 }
651 
652 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
653 {
654 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
655 	q->numa_node = node;
656 #endif
657 }
658 
659 #ifdef CONFIG_RPS
660 /*
661  * This structure holds an RPS map which can be of variable length.  The
662  * map is an array of CPUs.
663  */
664 struct rps_map {
665 	unsigned int len;
666 	struct rcu_head rcu;
667 	u16 cpus[0];
668 };
669 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
670 
671 /*
672  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
673  * tail pointer for that CPU's input queue at the time of last enqueue, and
674  * a hardware filter index.
675  */
676 struct rps_dev_flow {
677 	u16 cpu;
678 	u16 filter;
679 	unsigned int last_qtail;
680 };
681 #define RPS_NO_FILTER 0xffff
682 
683 /*
684  * The rps_dev_flow_table structure contains a table of flow mappings.
685  */
686 struct rps_dev_flow_table {
687 	unsigned int mask;
688 	struct rcu_head rcu;
689 	struct rps_dev_flow flows[0];
690 };
691 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
692     ((_num) * sizeof(struct rps_dev_flow)))
693 
694 /*
695  * The rps_sock_flow_table contains mappings of flows to the last CPU
696  * on which they were processed by the application (set in recvmsg).
697  * Each entry is a 32bit value. Upper part is the high-order bits
698  * of flow hash, lower part is CPU number.
699  * rps_cpu_mask is used to partition the space, depending on number of
700  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
701  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
702  * meaning we use 32-6=26 bits for the hash.
703  */
704 struct rps_sock_flow_table {
705 	u32	mask;
706 
707 	u32	ents[0] ____cacheline_aligned_in_smp;
708 };
709 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
710 
711 #define RPS_NO_CPU 0xffff
712 
713 extern u32 rps_cpu_mask;
714 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
715 
716 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
717 					u32 hash)
718 {
719 	if (table && hash) {
720 		unsigned int index = hash & table->mask;
721 		u32 val = hash & ~rps_cpu_mask;
722 
723 		/* We only give a hint, preemption can change CPU under us */
724 		val |= raw_smp_processor_id();
725 
726 		if (table->ents[index] != val)
727 			table->ents[index] = val;
728 	}
729 }
730 
731 #ifdef CONFIG_RFS_ACCEL
732 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
733 			 u16 filter_id);
734 #endif
735 #endif /* CONFIG_RPS */
736 
737 /* This structure contains an instance of an RX queue. */
738 struct netdev_rx_queue {
739 #ifdef CONFIG_RPS
740 	struct rps_map __rcu		*rps_map;
741 	struct rps_dev_flow_table __rcu	*rps_flow_table;
742 #endif
743 	struct kobject			kobj;
744 	struct net_device		*dev;
745 	struct xdp_rxq_info		xdp_rxq;
746 #ifdef CONFIG_XDP_SOCKETS
747 	struct xdp_umem                 *umem;
748 #endif
749 } ____cacheline_aligned_in_smp;
750 
751 /*
752  * RX queue sysfs structures and functions.
753  */
754 struct rx_queue_attribute {
755 	struct attribute attr;
756 	ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
757 	ssize_t (*store)(struct netdev_rx_queue *queue,
758 			 const char *buf, size_t len);
759 };
760 
761 #ifdef CONFIG_XPS
762 /*
763  * This structure holds an XPS map which can be of variable length.  The
764  * map is an array of queues.
765  */
766 struct xps_map {
767 	unsigned int len;
768 	unsigned int alloc_len;
769 	struct rcu_head rcu;
770 	u16 queues[0];
771 };
772 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
773 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
774        - sizeof(struct xps_map)) / sizeof(u16))
775 
776 /*
777  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
778  */
779 struct xps_dev_maps {
780 	struct rcu_head rcu;
781 	struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
782 };
783 
784 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +	\
785 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
786 
787 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
788 	(_rxqs * (_tcs) * sizeof(struct xps_map *)))
789 
790 #endif /* CONFIG_XPS */
791 
792 #define TC_MAX_QUEUE	16
793 #define TC_BITMASK	15
794 /* HW offloaded queuing disciplines txq count and offset maps */
795 struct netdev_tc_txq {
796 	u16 count;
797 	u16 offset;
798 };
799 
800 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
801 /*
802  * This structure is to hold information about the device
803  * configured to run FCoE protocol stack.
804  */
805 struct netdev_fcoe_hbainfo {
806 	char	manufacturer[64];
807 	char	serial_number[64];
808 	char	hardware_version[64];
809 	char	driver_version[64];
810 	char	optionrom_version[64];
811 	char	firmware_version[64];
812 	char	model[256];
813 	char	model_description[256];
814 };
815 #endif
816 
817 #define MAX_PHYS_ITEM_ID_LEN 32
818 
819 /* This structure holds a unique identifier to identify some
820  * physical item (port for example) used by a netdevice.
821  */
822 struct netdev_phys_item_id {
823 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
824 	unsigned char id_len;
825 };
826 
827 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
828 					    struct netdev_phys_item_id *b)
829 {
830 	return a->id_len == b->id_len &&
831 	       memcmp(a->id, b->id, a->id_len) == 0;
832 }
833 
834 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
835 				       struct sk_buff *skb,
836 				       struct net_device *sb_dev);
837 
838 enum tc_setup_type {
839 	TC_SETUP_QDISC_MQPRIO,
840 	TC_SETUP_CLSU32,
841 	TC_SETUP_CLSFLOWER,
842 	TC_SETUP_CLSMATCHALL,
843 	TC_SETUP_CLSBPF,
844 	TC_SETUP_BLOCK,
845 	TC_SETUP_QDISC_CBS,
846 	TC_SETUP_QDISC_RED,
847 	TC_SETUP_QDISC_PRIO,
848 	TC_SETUP_QDISC_MQ,
849 	TC_SETUP_QDISC_ETF,
850 	TC_SETUP_ROOT_QDISC,
851 	TC_SETUP_QDISC_GRED,
852 };
853 
854 /* These structures hold the attributes of bpf state that are being passed
855  * to the netdevice through the bpf op.
856  */
857 enum bpf_netdev_command {
858 	/* Set or clear a bpf program used in the earliest stages of packet
859 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
860 	 * is responsible for calling bpf_prog_put on any old progs that are
861 	 * stored. In case of error, the callee need not release the new prog
862 	 * reference, but on success it takes ownership and must bpf_prog_put
863 	 * when it is no longer used.
864 	 */
865 	XDP_SETUP_PROG,
866 	XDP_SETUP_PROG_HW,
867 	XDP_QUERY_PROG,
868 	XDP_QUERY_PROG_HW,
869 	/* BPF program for offload callbacks, invoked at program load time. */
870 	BPF_OFFLOAD_MAP_ALLOC,
871 	BPF_OFFLOAD_MAP_FREE,
872 	XDP_SETUP_XSK_UMEM,
873 };
874 
875 struct bpf_prog_offload_ops;
876 struct netlink_ext_ack;
877 struct xdp_umem;
878 
879 struct netdev_bpf {
880 	enum bpf_netdev_command command;
881 	union {
882 		/* XDP_SETUP_PROG */
883 		struct {
884 			u32 flags;
885 			struct bpf_prog *prog;
886 			struct netlink_ext_ack *extack;
887 		};
888 		/* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
889 		struct {
890 			u32 prog_id;
891 			/* flags with which program was installed */
892 			u32 prog_flags;
893 		};
894 		/* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
895 		struct {
896 			struct bpf_offloaded_map *offmap;
897 		};
898 		/* XDP_SETUP_XSK_UMEM */
899 		struct {
900 			struct xdp_umem *umem;
901 			u16 queue_id;
902 		} xsk;
903 	};
904 };
905 
906 #ifdef CONFIG_XFRM_OFFLOAD
907 struct xfrmdev_ops {
908 	int	(*xdo_dev_state_add) (struct xfrm_state *x);
909 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
910 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
911 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
912 				       struct xfrm_state *x);
913 	void	(*xdo_dev_state_advance_esn) (struct xfrm_state *x);
914 };
915 #endif
916 
917 #if IS_ENABLED(CONFIG_TLS_DEVICE)
918 enum tls_offload_ctx_dir {
919 	TLS_OFFLOAD_CTX_DIR_RX,
920 	TLS_OFFLOAD_CTX_DIR_TX,
921 };
922 
923 struct tls_crypto_info;
924 struct tls_context;
925 
926 struct tlsdev_ops {
927 	int (*tls_dev_add)(struct net_device *netdev, struct sock *sk,
928 			   enum tls_offload_ctx_dir direction,
929 			   struct tls_crypto_info *crypto_info,
930 			   u32 start_offload_tcp_sn);
931 	void (*tls_dev_del)(struct net_device *netdev,
932 			    struct tls_context *ctx,
933 			    enum tls_offload_ctx_dir direction);
934 	void (*tls_dev_resync_rx)(struct net_device *netdev,
935 				  struct sock *sk, u32 seq, u64 rcd_sn);
936 };
937 #endif
938 
939 struct dev_ifalias {
940 	struct rcu_head rcuhead;
941 	char ifalias[];
942 };
943 
944 struct devlink;
945 
946 /*
947  * This structure defines the management hooks for network devices.
948  * The following hooks can be defined; unless noted otherwise, they are
949  * optional and can be filled with a null pointer.
950  *
951  * int (*ndo_init)(struct net_device *dev);
952  *     This function is called once when a network device is registered.
953  *     The network device can use this for any late stage initialization
954  *     or semantic validation. It can fail with an error code which will
955  *     be propagated back to register_netdev.
956  *
957  * void (*ndo_uninit)(struct net_device *dev);
958  *     This function is called when device is unregistered or when registration
959  *     fails. It is not called if init fails.
960  *
961  * int (*ndo_open)(struct net_device *dev);
962  *     This function is called when a network device transitions to the up
963  *     state.
964  *
965  * int (*ndo_stop)(struct net_device *dev);
966  *     This function is called when a network device transitions to the down
967  *     state.
968  *
969  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
970  *                               struct net_device *dev);
971  *	Called when a packet needs to be transmitted.
972  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
973  *	the queue before that can happen; it's for obsolete devices and weird
974  *	corner cases, but the stack really does a non-trivial amount
975  *	of useless work if you return NETDEV_TX_BUSY.
976  *	Required; cannot be NULL.
977  *
978  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
979  *					   struct net_device *dev
980  *					   netdev_features_t features);
981  *	Called by core transmit path to determine if device is capable of
982  *	performing offload operations on a given packet. This is to give
983  *	the device an opportunity to implement any restrictions that cannot
984  *	be otherwise expressed by feature flags. The check is called with
985  *	the set of features that the stack has calculated and it returns
986  *	those the driver believes to be appropriate.
987  *
988  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
989  *                         struct net_device *sb_dev);
990  *	Called to decide which queue to use when device supports multiple
991  *	transmit queues.
992  *
993  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
994  *	This function is called to allow device receiver to make
995  *	changes to configuration when multicast or promiscuous is enabled.
996  *
997  * void (*ndo_set_rx_mode)(struct net_device *dev);
998  *	This function is called device changes address list filtering.
999  *	If driver handles unicast address filtering, it should set
1000  *	IFF_UNICAST_FLT in its priv_flags.
1001  *
1002  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1003  *	This function  is called when the Media Access Control address
1004  *	needs to be changed. If this interface is not defined, the
1005  *	MAC address can not be changed.
1006  *
1007  * int (*ndo_validate_addr)(struct net_device *dev);
1008  *	Test if Media Access Control address is valid for the device.
1009  *
1010  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1011  *	Called when a user requests an ioctl which can't be handled by
1012  *	the generic interface code. If not defined ioctls return
1013  *	not supported error code.
1014  *
1015  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1016  *	Used to set network devices bus interface parameters. This interface
1017  *	is retained for legacy reasons; new devices should use the bus
1018  *	interface (PCI) for low level management.
1019  *
1020  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1021  *	Called when a user wants to change the Maximum Transfer Unit
1022  *	of a device.
1023  *
1024  * void (*ndo_tx_timeout)(struct net_device *dev);
1025  *	Callback used when the transmitter has not made any progress
1026  *	for dev->watchdog ticks.
1027  *
1028  * void (*ndo_get_stats64)(struct net_device *dev,
1029  *                         struct rtnl_link_stats64 *storage);
1030  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1031  *	Called when a user wants to get the network device usage
1032  *	statistics. Drivers must do one of the following:
1033  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
1034  *	   rtnl_link_stats64 structure passed by the caller.
1035  *	2. Define @ndo_get_stats to update a net_device_stats structure
1036  *	   (which should normally be dev->stats) and return a pointer to
1037  *	   it. The structure may be changed asynchronously only if each
1038  *	   field is written atomically.
1039  *	3. Update dev->stats asynchronously and atomically, and define
1040  *	   neither operation.
1041  *
1042  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1043  *	Return true if this device supports offload stats of this attr_id.
1044  *
1045  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1046  *	void *attr_data)
1047  *	Get statistics for offload operations by attr_id. Write it into the
1048  *	attr_data pointer.
1049  *
1050  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1051  *	If device supports VLAN filtering this function is called when a
1052  *	VLAN id is registered.
1053  *
1054  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1055  *	If device supports VLAN filtering this function is called when a
1056  *	VLAN id is unregistered.
1057  *
1058  * void (*ndo_poll_controller)(struct net_device *dev);
1059  *
1060  *	SR-IOV management functions.
1061  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1062  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1063  *			  u8 qos, __be16 proto);
1064  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1065  *			  int max_tx_rate);
1066  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1067  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1068  * int (*ndo_get_vf_config)(struct net_device *dev,
1069  *			    int vf, struct ifla_vf_info *ivf);
1070  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1071  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1072  *			  struct nlattr *port[]);
1073  *
1074  *      Enable or disable the VF ability to query its RSS Redirection Table and
1075  *      Hash Key. This is needed since on some devices VF share this information
1076  *      with PF and querying it may introduce a theoretical security risk.
1077  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1078  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1079  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1080  *		       void *type_data);
1081  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
1082  *	This is always called from the stack with the rtnl lock held and netif
1083  *	tx queues stopped. This allows the netdevice to perform queue
1084  *	management safely.
1085  *
1086  *	Fiber Channel over Ethernet (FCoE) offload functions.
1087  * int (*ndo_fcoe_enable)(struct net_device *dev);
1088  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
1089  *	so the underlying device can perform whatever needed configuration or
1090  *	initialization to support acceleration of FCoE traffic.
1091  *
1092  * int (*ndo_fcoe_disable)(struct net_device *dev);
1093  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
1094  *	so the underlying device can perform whatever needed clean-ups to
1095  *	stop supporting acceleration of FCoE traffic.
1096  *
1097  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1098  *			     struct scatterlist *sgl, unsigned int sgc);
1099  *	Called when the FCoE Initiator wants to initialize an I/O that
1100  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1101  *	perform necessary setup and returns 1 to indicate the device is set up
1102  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1103  *
1104  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1105  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1106  *	indicated by the FC exchange id 'xid', so the underlying device can
1107  *	clean up and reuse resources for later DDP requests.
1108  *
1109  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1110  *			      struct scatterlist *sgl, unsigned int sgc);
1111  *	Called when the FCoE Target wants to initialize an I/O that
1112  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1113  *	perform necessary setup and returns 1 to indicate the device is set up
1114  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1115  *
1116  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1117  *			       struct netdev_fcoe_hbainfo *hbainfo);
1118  *	Called when the FCoE Protocol stack wants information on the underlying
1119  *	device. This information is utilized by the FCoE protocol stack to
1120  *	register attributes with Fiber Channel management service as per the
1121  *	FC-GS Fabric Device Management Information(FDMI) specification.
1122  *
1123  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1124  *	Called when the underlying device wants to override default World Wide
1125  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1126  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1127  *	protocol stack to use.
1128  *
1129  *	RFS acceleration.
1130  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1131  *			    u16 rxq_index, u32 flow_id);
1132  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1133  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1134  *	Return the filter ID on success, or a negative error code.
1135  *
1136  *	Slave management functions (for bridge, bonding, etc).
1137  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1138  *	Called to make another netdev an underling.
1139  *
1140  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1141  *	Called to release previously enslaved netdev.
1142  *
1143  *      Feature/offload setting functions.
1144  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1145  *		netdev_features_t features);
1146  *	Adjusts the requested feature flags according to device-specific
1147  *	constraints, and returns the resulting flags. Must not modify
1148  *	the device state.
1149  *
1150  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1151  *	Called to update device configuration to new features. Passed
1152  *	feature set might be less than what was returned by ndo_fix_features()).
1153  *	Must return >0 or -errno if it changed dev->features itself.
1154  *
1155  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1156  *		      struct net_device *dev,
1157  *		      const unsigned char *addr, u16 vid, u16 flags,
1158  *		      struct netlink_ext_ack *extack);
1159  *	Adds an FDB entry to dev for addr.
1160  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1161  *		      struct net_device *dev,
1162  *		      const unsigned char *addr, u16 vid)
1163  *	Deletes the FDB entry from dev coresponding to addr.
1164  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1165  *		       struct net_device *dev, struct net_device *filter_dev,
1166  *		       int *idx)
1167  *	Used to add FDB entries to dump requests. Implementers should add
1168  *	entries to skb and update idx with the number of entries.
1169  *
1170  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1171  *			     u16 flags, struct netlink_ext_ack *extack)
1172  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1173  *			     struct net_device *dev, u32 filter_mask,
1174  *			     int nlflags)
1175  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1176  *			     u16 flags);
1177  *
1178  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1179  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1180  *	which do not represent real hardware may define this to allow their
1181  *	userspace components to manage their virtual carrier state. Devices
1182  *	that determine carrier state from physical hardware properties (eg
1183  *	network cables) or protocol-dependent mechanisms (eg
1184  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1185  *
1186  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1187  *			       struct netdev_phys_item_id *ppid);
1188  *	Called to get ID of physical port of this device. If driver does
1189  *	not implement this, it is assumed that the hw is not able to have
1190  *	multiple net devices on single physical port.
1191  *
1192  * int (*ndo_get_port_parent_id)(struct net_device *dev,
1193  *				 struct netdev_phys_item_id *ppid)
1194  *	Called to get the parent ID of the physical port of this device.
1195  *
1196  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1197  *			      struct udp_tunnel_info *ti);
1198  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1199  *	address family that a UDP tunnel is listnening to. It is called only
1200  *	when a new port starts listening. The operation is protected by the
1201  *	RTNL.
1202  *
1203  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1204  *			      struct udp_tunnel_info *ti);
1205  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1206  *	address family that the UDP tunnel is not listening to anymore. The
1207  *	operation is protected by the RTNL.
1208  *
1209  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1210  *				 struct net_device *dev)
1211  *	Called by upper layer devices to accelerate switching or other
1212  *	station functionality into hardware. 'pdev is the lowerdev
1213  *	to use for the offload and 'dev' is the net device that will
1214  *	back the offload. Returns a pointer to the private structure
1215  *	the upper layer will maintain.
1216  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1217  *	Called by upper layer device to delete the station created
1218  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1219  *	the station and priv is the structure returned by the add
1220  *	operation.
1221  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1222  *			     int queue_index, u32 maxrate);
1223  *	Called when a user wants to set a max-rate limitation of specific
1224  *	TX queue.
1225  * int (*ndo_get_iflink)(const struct net_device *dev);
1226  *	Called to get the iflink value of this device.
1227  * void (*ndo_change_proto_down)(struct net_device *dev,
1228  *				 bool proto_down);
1229  *	This function is used to pass protocol port error state information
1230  *	to the switch driver. The switch driver can react to the proto_down
1231  *      by doing a phys down on the associated switch port.
1232  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1233  *	This function is used to get egress tunnel information for given skb.
1234  *	This is useful for retrieving outer tunnel header parameters while
1235  *	sampling packet.
1236  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1237  *	This function is used to specify the headroom that the skb must
1238  *	consider when allocation skb during packet reception. Setting
1239  *	appropriate rx headroom value allows avoiding skb head copy on
1240  *	forward. Setting a negative value resets the rx headroom to the
1241  *	default value.
1242  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1243  *	This function is used to set or query state related to XDP on the
1244  *	netdevice and manage BPF offload. See definition of
1245  *	enum bpf_netdev_command for details.
1246  * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1247  *			u32 flags);
1248  *	This function is used to submit @n XDP packets for transmit on a
1249  *	netdevice. Returns number of frames successfully transmitted, frames
1250  *	that got dropped are freed/returned via xdp_return_frame().
1251  *	Returns negative number, means general error invoking ndo, meaning
1252  *	no frames were xmit'ed and core-caller will free all frames.
1253  * struct devlink *(*ndo_get_devlink)(struct net_device *dev);
1254  *	Get devlink instance associated with a given netdev.
1255  *	Called with a reference on the netdevice and devlink locks only,
1256  *	rtnl_lock is not held.
1257  */
1258 struct net_device_ops {
1259 	int			(*ndo_init)(struct net_device *dev);
1260 	void			(*ndo_uninit)(struct net_device *dev);
1261 	int			(*ndo_open)(struct net_device *dev);
1262 	int			(*ndo_stop)(struct net_device *dev);
1263 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1264 						  struct net_device *dev);
1265 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1266 						      struct net_device *dev,
1267 						      netdev_features_t features);
1268 	u16			(*ndo_select_queue)(struct net_device *dev,
1269 						    struct sk_buff *skb,
1270 						    struct net_device *sb_dev);
1271 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1272 						       int flags);
1273 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1274 	int			(*ndo_set_mac_address)(struct net_device *dev,
1275 						       void *addr);
1276 	int			(*ndo_validate_addr)(struct net_device *dev);
1277 	int			(*ndo_do_ioctl)(struct net_device *dev,
1278 					        struct ifreq *ifr, int cmd);
1279 	int			(*ndo_set_config)(struct net_device *dev,
1280 					          struct ifmap *map);
1281 	int			(*ndo_change_mtu)(struct net_device *dev,
1282 						  int new_mtu);
1283 	int			(*ndo_neigh_setup)(struct net_device *dev,
1284 						   struct neigh_parms *);
1285 	void			(*ndo_tx_timeout) (struct net_device *dev);
1286 
1287 	void			(*ndo_get_stats64)(struct net_device *dev,
1288 						   struct rtnl_link_stats64 *storage);
1289 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1290 	int			(*ndo_get_offload_stats)(int attr_id,
1291 							 const struct net_device *dev,
1292 							 void *attr_data);
1293 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1294 
1295 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1296 						       __be16 proto, u16 vid);
1297 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1298 						        __be16 proto, u16 vid);
1299 #ifdef CONFIG_NET_POLL_CONTROLLER
1300 	void                    (*ndo_poll_controller)(struct net_device *dev);
1301 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1302 						     struct netpoll_info *info);
1303 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1304 #endif
1305 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1306 						  int queue, u8 *mac);
1307 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1308 						   int queue, u16 vlan,
1309 						   u8 qos, __be16 proto);
1310 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1311 						   int vf, int min_tx_rate,
1312 						   int max_tx_rate);
1313 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1314 						       int vf, bool setting);
1315 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1316 						    int vf, bool setting);
1317 	int			(*ndo_get_vf_config)(struct net_device *dev,
1318 						     int vf,
1319 						     struct ifla_vf_info *ivf);
1320 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1321 							 int vf, int link_state);
1322 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1323 						    int vf,
1324 						    struct ifla_vf_stats
1325 						    *vf_stats);
1326 	int			(*ndo_set_vf_port)(struct net_device *dev,
1327 						   int vf,
1328 						   struct nlattr *port[]);
1329 	int			(*ndo_get_vf_port)(struct net_device *dev,
1330 						   int vf, struct sk_buff *skb);
1331 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1332 						   int vf, u64 guid,
1333 						   int guid_type);
1334 	int			(*ndo_set_vf_rss_query_en)(
1335 						   struct net_device *dev,
1336 						   int vf, bool setting);
1337 	int			(*ndo_setup_tc)(struct net_device *dev,
1338 						enum tc_setup_type type,
1339 						void *type_data);
1340 #if IS_ENABLED(CONFIG_FCOE)
1341 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1342 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1343 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1344 						      u16 xid,
1345 						      struct scatterlist *sgl,
1346 						      unsigned int sgc);
1347 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1348 						     u16 xid);
1349 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1350 						       u16 xid,
1351 						       struct scatterlist *sgl,
1352 						       unsigned int sgc);
1353 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1354 							struct netdev_fcoe_hbainfo *hbainfo);
1355 #endif
1356 
1357 #if IS_ENABLED(CONFIG_LIBFCOE)
1358 #define NETDEV_FCOE_WWNN 0
1359 #define NETDEV_FCOE_WWPN 1
1360 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1361 						    u64 *wwn, int type);
1362 #endif
1363 
1364 #ifdef CONFIG_RFS_ACCEL
1365 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1366 						     const struct sk_buff *skb,
1367 						     u16 rxq_index,
1368 						     u32 flow_id);
1369 #endif
1370 	int			(*ndo_add_slave)(struct net_device *dev,
1371 						 struct net_device *slave_dev,
1372 						 struct netlink_ext_ack *extack);
1373 	int			(*ndo_del_slave)(struct net_device *dev,
1374 						 struct net_device *slave_dev);
1375 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1376 						    netdev_features_t features);
1377 	int			(*ndo_set_features)(struct net_device *dev,
1378 						    netdev_features_t features);
1379 	int			(*ndo_neigh_construct)(struct net_device *dev,
1380 						       struct neighbour *n);
1381 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1382 						     struct neighbour *n);
1383 
1384 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1385 					       struct nlattr *tb[],
1386 					       struct net_device *dev,
1387 					       const unsigned char *addr,
1388 					       u16 vid,
1389 					       u16 flags,
1390 					       struct netlink_ext_ack *extack);
1391 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1392 					       struct nlattr *tb[],
1393 					       struct net_device *dev,
1394 					       const unsigned char *addr,
1395 					       u16 vid);
1396 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1397 						struct netlink_callback *cb,
1398 						struct net_device *dev,
1399 						struct net_device *filter_dev,
1400 						int *idx);
1401 	int			(*ndo_fdb_get)(struct sk_buff *skb,
1402 					       struct nlattr *tb[],
1403 					       struct net_device *dev,
1404 					       const unsigned char *addr,
1405 					       u16 vid, u32 portid, u32 seq,
1406 					       struct netlink_ext_ack *extack);
1407 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1408 						      struct nlmsghdr *nlh,
1409 						      u16 flags,
1410 						      struct netlink_ext_ack *extack);
1411 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1412 						      u32 pid, u32 seq,
1413 						      struct net_device *dev,
1414 						      u32 filter_mask,
1415 						      int nlflags);
1416 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1417 						      struct nlmsghdr *nlh,
1418 						      u16 flags);
1419 	int			(*ndo_change_carrier)(struct net_device *dev,
1420 						      bool new_carrier);
1421 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1422 							struct netdev_phys_item_id *ppid);
1423 	int			(*ndo_get_port_parent_id)(struct net_device *dev,
1424 							  struct netdev_phys_item_id *ppid);
1425 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1426 							  char *name, size_t len);
1427 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1428 						      struct udp_tunnel_info *ti);
1429 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1430 						      struct udp_tunnel_info *ti);
1431 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1432 							struct net_device *dev);
1433 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1434 							void *priv);
1435 
1436 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1437 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1438 						      int queue_index,
1439 						      u32 maxrate);
1440 	int			(*ndo_get_iflink)(const struct net_device *dev);
1441 	int			(*ndo_change_proto_down)(struct net_device *dev,
1442 							 bool proto_down);
1443 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1444 						       struct sk_buff *skb);
1445 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1446 						       int needed_headroom);
1447 	int			(*ndo_bpf)(struct net_device *dev,
1448 					   struct netdev_bpf *bpf);
1449 	int			(*ndo_xdp_xmit)(struct net_device *dev, int n,
1450 						struct xdp_frame **xdp,
1451 						u32 flags);
1452 	int			(*ndo_xsk_async_xmit)(struct net_device *dev,
1453 						      u32 queue_id);
1454 	struct devlink *	(*ndo_get_devlink)(struct net_device *dev);
1455 };
1456 
1457 /**
1458  * enum net_device_priv_flags - &struct net_device priv_flags
1459  *
1460  * These are the &struct net_device, they are only set internally
1461  * by drivers and used in the kernel. These flags are invisible to
1462  * userspace; this means that the order of these flags can change
1463  * during any kernel release.
1464  *
1465  * You should have a pretty good reason to be extending these flags.
1466  *
1467  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1468  * @IFF_EBRIDGE: Ethernet bridging device
1469  * @IFF_BONDING: bonding master or slave
1470  * @IFF_ISATAP: ISATAP interface (RFC4214)
1471  * @IFF_WAN_HDLC: WAN HDLC device
1472  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1473  *	release skb->dst
1474  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1475  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1476  * @IFF_MACVLAN_PORT: device used as macvlan port
1477  * @IFF_BRIDGE_PORT: device used as bridge port
1478  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1479  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1480  * @IFF_UNICAST_FLT: Supports unicast filtering
1481  * @IFF_TEAM_PORT: device used as team port
1482  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1483  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1484  *	change when it's running
1485  * @IFF_MACVLAN: Macvlan device
1486  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1487  *	underlying stacked devices
1488  * @IFF_L3MDEV_MASTER: device is an L3 master device
1489  * @IFF_NO_QUEUE: device can run without qdisc attached
1490  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1491  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1492  * @IFF_TEAM: device is a team device
1493  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1494  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1495  *	entity (i.e. the master device for bridged veth)
1496  * @IFF_MACSEC: device is a MACsec device
1497  * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1498  * @IFF_FAILOVER: device is a failover master device
1499  * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1500  * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1501  */
1502 enum netdev_priv_flags {
1503 	IFF_802_1Q_VLAN			= 1<<0,
1504 	IFF_EBRIDGE			= 1<<1,
1505 	IFF_BONDING			= 1<<2,
1506 	IFF_ISATAP			= 1<<3,
1507 	IFF_WAN_HDLC			= 1<<4,
1508 	IFF_XMIT_DST_RELEASE		= 1<<5,
1509 	IFF_DONT_BRIDGE			= 1<<6,
1510 	IFF_DISABLE_NETPOLL		= 1<<7,
1511 	IFF_MACVLAN_PORT		= 1<<8,
1512 	IFF_BRIDGE_PORT			= 1<<9,
1513 	IFF_OVS_DATAPATH		= 1<<10,
1514 	IFF_TX_SKB_SHARING		= 1<<11,
1515 	IFF_UNICAST_FLT			= 1<<12,
1516 	IFF_TEAM_PORT			= 1<<13,
1517 	IFF_SUPP_NOFCS			= 1<<14,
1518 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1519 	IFF_MACVLAN			= 1<<16,
1520 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1521 	IFF_L3MDEV_MASTER		= 1<<18,
1522 	IFF_NO_QUEUE			= 1<<19,
1523 	IFF_OPENVSWITCH			= 1<<20,
1524 	IFF_L3MDEV_SLAVE		= 1<<21,
1525 	IFF_TEAM			= 1<<22,
1526 	IFF_RXFH_CONFIGURED		= 1<<23,
1527 	IFF_PHONY_HEADROOM		= 1<<24,
1528 	IFF_MACSEC			= 1<<25,
1529 	IFF_NO_RX_HANDLER		= 1<<26,
1530 	IFF_FAILOVER			= 1<<27,
1531 	IFF_FAILOVER_SLAVE		= 1<<28,
1532 	IFF_L3MDEV_RX_HANDLER		= 1<<29,
1533 };
1534 
1535 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1536 #define IFF_EBRIDGE			IFF_EBRIDGE
1537 #define IFF_BONDING			IFF_BONDING
1538 #define IFF_ISATAP			IFF_ISATAP
1539 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1540 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1541 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1542 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1543 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1544 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1545 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1546 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1547 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1548 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1549 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1550 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1551 #define IFF_MACVLAN			IFF_MACVLAN
1552 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1553 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1554 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1555 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1556 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1557 #define IFF_TEAM			IFF_TEAM
1558 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1559 #define IFF_MACSEC			IFF_MACSEC
1560 #define IFF_NO_RX_HANDLER		IFF_NO_RX_HANDLER
1561 #define IFF_FAILOVER			IFF_FAILOVER
1562 #define IFF_FAILOVER_SLAVE		IFF_FAILOVER_SLAVE
1563 #define IFF_L3MDEV_RX_HANDLER		IFF_L3MDEV_RX_HANDLER
1564 
1565 /**
1566  *	struct net_device - The DEVICE structure.
1567  *
1568  *	Actually, this whole structure is a big mistake.  It mixes I/O
1569  *	data with strictly "high-level" data, and it has to know about
1570  *	almost every data structure used in the INET module.
1571  *
1572  *	@name:	This is the first field of the "visible" part of this structure
1573  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1574  *		of the interface.
1575  *
1576  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1577  *	@ifalias:	SNMP alias
1578  *	@mem_end:	Shared memory end
1579  *	@mem_start:	Shared memory start
1580  *	@base_addr:	Device I/O address
1581  *	@irq:		Device IRQ number
1582  *
1583  *	@state:		Generic network queuing layer state, see netdev_state_t
1584  *	@dev_list:	The global list of network devices
1585  *	@napi_list:	List entry used for polling NAPI devices
1586  *	@unreg_list:	List entry  when we are unregistering the
1587  *			device; see the function unregister_netdev
1588  *	@close_list:	List entry used when we are closing the device
1589  *	@ptype_all:     Device-specific packet handlers for all protocols
1590  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1591  *
1592  *	@adj_list:	Directly linked devices, like slaves for bonding
1593  *	@features:	Currently active device features
1594  *	@hw_features:	User-changeable features
1595  *
1596  *	@wanted_features:	User-requested features
1597  *	@vlan_features:		Mask of features inheritable by VLAN devices
1598  *
1599  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1600  *				This field indicates what encapsulation
1601  *				offloads the hardware is capable of doing,
1602  *				and drivers will need to set them appropriately.
1603  *
1604  *	@mpls_features:	Mask of features inheritable by MPLS
1605  *
1606  *	@ifindex:	interface index
1607  *	@group:		The group the device belongs to
1608  *
1609  *	@stats:		Statistics struct, which was left as a legacy, use
1610  *			rtnl_link_stats64 instead
1611  *
1612  *	@rx_dropped:	Dropped packets by core network,
1613  *			do not use this in drivers
1614  *	@tx_dropped:	Dropped packets by core network,
1615  *			do not use this in drivers
1616  *	@rx_nohandler:	nohandler dropped packets by core network on
1617  *			inactive devices, do not use this in drivers
1618  *	@carrier_up_count:	Number of times the carrier has been up
1619  *	@carrier_down_count:	Number of times the carrier has been down
1620  *
1621  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1622  *				instead of ioctl,
1623  *				see <net/iw_handler.h> for details.
1624  *	@wireless_data:	Instance data managed by the core of wireless extensions
1625  *
1626  *	@netdev_ops:	Includes several pointers to callbacks,
1627  *			if one wants to override the ndo_*() functions
1628  *	@ethtool_ops:	Management operations
1629  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1630  *			discovery handling. Necessary for e.g. 6LoWPAN.
1631  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1632  *			of Layer 2 headers.
1633  *
1634  *	@flags:		Interface flags (a la BSD)
1635  *	@priv_flags:	Like 'flags' but invisible to userspace,
1636  *			see if.h for the definitions
1637  *	@gflags:	Global flags ( kept as legacy )
1638  *	@padded:	How much padding added by alloc_netdev()
1639  *	@operstate:	RFC2863 operstate
1640  *	@link_mode:	Mapping policy to operstate
1641  *	@if_port:	Selectable AUI, TP, ...
1642  *	@dma:		DMA channel
1643  *	@mtu:		Interface MTU value
1644  *	@min_mtu:	Interface Minimum MTU value
1645  *	@max_mtu:	Interface Maximum MTU value
1646  *	@type:		Interface hardware type
1647  *	@hard_header_len: Maximum hardware header length.
1648  *	@min_header_len:  Minimum hardware header length
1649  *
1650  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1651  *			  cases can this be guaranteed
1652  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1653  *			  cases can this be guaranteed. Some cases also use
1654  *			  LL_MAX_HEADER instead to allocate the skb
1655  *
1656  *	interface address info:
1657  *
1658  * 	@perm_addr:		Permanent hw address
1659  * 	@addr_assign_type:	Hw address assignment type
1660  * 	@addr_len:		Hardware address length
1661  *	@neigh_priv_len:	Used in neigh_alloc()
1662  * 	@dev_id:		Used to differentiate devices that share
1663  * 				the same link layer address
1664  * 	@dev_port:		Used to differentiate devices that share
1665  * 				the same function
1666  *	@addr_list_lock:	XXX: need comments on this one
1667  *	@uc_promisc:		Counter that indicates promiscuous mode
1668  *				has been enabled due to the need to listen to
1669  *				additional unicast addresses in a device that
1670  *				does not implement ndo_set_rx_mode()
1671  *	@uc:			unicast mac addresses
1672  *	@mc:			multicast mac addresses
1673  *	@dev_addrs:		list of device hw addresses
1674  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1675  *	@promiscuity:		Number of times the NIC is told to work in
1676  *				promiscuous mode; if it becomes 0 the NIC will
1677  *				exit promiscuous mode
1678  *	@allmulti:		Counter, enables or disables allmulticast mode
1679  *
1680  *	@vlan_info:	VLAN info
1681  *	@dsa_ptr:	dsa specific data
1682  *	@tipc_ptr:	TIPC specific data
1683  *	@atalk_ptr:	AppleTalk link
1684  *	@ip_ptr:	IPv4 specific data
1685  *	@dn_ptr:	DECnet specific data
1686  *	@ip6_ptr:	IPv6 specific data
1687  *	@ax25_ptr:	AX.25 specific data
1688  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1689  *
1690  *	@dev_addr:	Hw address (before bcast,
1691  *			because most packets are unicast)
1692  *
1693  *	@_rx:			Array of RX queues
1694  *	@num_rx_queues:		Number of RX queues
1695  *				allocated at register_netdev() time
1696  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1697  *
1698  *	@rx_handler:		handler for received packets
1699  *	@rx_handler_data: 	XXX: need comments on this one
1700  *	@miniq_ingress:		ingress/clsact qdisc specific data for
1701  *				ingress processing
1702  *	@ingress_queue:		XXX: need comments on this one
1703  *	@broadcast:		hw bcast address
1704  *
1705  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1706  *			indexed by RX queue number. Assigned by driver.
1707  *			This must only be set if the ndo_rx_flow_steer
1708  *			operation is defined
1709  *	@index_hlist:		Device index hash chain
1710  *
1711  *	@_tx:			Array of TX queues
1712  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1713  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1714  *	@qdisc:			Root qdisc from userspace point of view
1715  *	@tx_queue_len:		Max frames per queue allowed
1716  *	@tx_global_lock: 	XXX: need comments on this one
1717  *
1718  *	@xps_maps:	XXX: need comments on this one
1719  *	@miniq_egress:		clsact qdisc specific data for
1720  *				egress processing
1721  *	@watchdog_timeo:	Represents the timeout that is used by
1722  *				the watchdog (see dev_watchdog())
1723  *	@watchdog_timer:	List of timers
1724  *
1725  *	@pcpu_refcnt:		Number of references to this device
1726  *	@todo_list:		Delayed register/unregister
1727  *	@link_watch_list:	XXX: need comments on this one
1728  *
1729  *	@reg_state:		Register/unregister state machine
1730  *	@dismantle:		Device is going to be freed
1731  *	@rtnl_link_state:	This enum represents the phases of creating
1732  *				a new link
1733  *
1734  *	@needs_free_netdev:	Should unregister perform free_netdev?
1735  *	@priv_destructor:	Called from unregister
1736  *	@npinfo:		XXX: need comments on this one
1737  * 	@nd_net:		Network namespace this network device is inside
1738  *
1739  * 	@ml_priv:	Mid-layer private
1740  * 	@lstats:	Loopback statistics
1741  * 	@tstats:	Tunnel statistics
1742  * 	@dstats:	Dummy statistics
1743  * 	@vstats:	Virtual ethernet statistics
1744  *
1745  *	@garp_port:	GARP
1746  *	@mrp_port:	MRP
1747  *
1748  *	@dev:		Class/net/name entry
1749  *	@sysfs_groups:	Space for optional device, statistics and wireless
1750  *			sysfs groups
1751  *
1752  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1753  *	@rtnl_link_ops:	Rtnl_link_ops
1754  *
1755  *	@gso_max_size:	Maximum size of generic segmentation offload
1756  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1757  *			NIC for GSO
1758  *
1759  *	@dcbnl_ops:	Data Center Bridging netlink ops
1760  *	@num_tc:	Number of traffic classes in the net device
1761  *	@tc_to_txq:	XXX: need comments on this one
1762  *	@prio_tc_map:	XXX: need comments on this one
1763  *
1764  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1765  *
1766  *	@priomap:	XXX: need comments on this one
1767  *	@phydev:	Physical device may attach itself
1768  *			for hardware timestamping
1769  *	@sfp_bus:	attached &struct sfp_bus structure.
1770  *
1771  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1772  *	@qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1773  *
1774  *	@proto_down:	protocol port state information can be sent to the
1775  *			switch driver and used to set the phys state of the
1776  *			switch port.
1777  *
1778  *	@wol_enabled:	Wake-on-LAN is enabled
1779  *
1780  *	FIXME: cleanup struct net_device such that network protocol info
1781  *	moves out.
1782  */
1783 
1784 struct net_device {
1785 	char			name[IFNAMSIZ];
1786 	struct hlist_node	name_hlist;
1787 	struct dev_ifalias	__rcu *ifalias;
1788 	/*
1789 	 *	I/O specific fields
1790 	 *	FIXME: Merge these and struct ifmap into one
1791 	 */
1792 	unsigned long		mem_end;
1793 	unsigned long		mem_start;
1794 	unsigned long		base_addr;
1795 	int			irq;
1796 
1797 	/*
1798 	 *	Some hardware also needs these fields (state,dev_list,
1799 	 *	napi_list,unreg_list,close_list) but they are not
1800 	 *	part of the usual set specified in Space.c.
1801 	 */
1802 
1803 	unsigned long		state;
1804 
1805 	struct list_head	dev_list;
1806 	struct list_head	napi_list;
1807 	struct list_head	unreg_list;
1808 	struct list_head	close_list;
1809 	struct list_head	ptype_all;
1810 	struct list_head	ptype_specific;
1811 
1812 	struct {
1813 		struct list_head upper;
1814 		struct list_head lower;
1815 	} adj_list;
1816 
1817 	netdev_features_t	features;
1818 	netdev_features_t	hw_features;
1819 	netdev_features_t	wanted_features;
1820 	netdev_features_t	vlan_features;
1821 	netdev_features_t	hw_enc_features;
1822 	netdev_features_t	mpls_features;
1823 	netdev_features_t	gso_partial_features;
1824 
1825 	int			ifindex;
1826 	int			group;
1827 
1828 	struct net_device_stats	stats;
1829 
1830 	atomic_long_t		rx_dropped;
1831 	atomic_long_t		tx_dropped;
1832 	atomic_long_t		rx_nohandler;
1833 
1834 	/* Stats to monitor link on/off, flapping */
1835 	atomic_t		carrier_up_count;
1836 	atomic_t		carrier_down_count;
1837 
1838 #ifdef CONFIG_WIRELESS_EXT
1839 	const struct iw_handler_def *wireless_handlers;
1840 	struct iw_public_data	*wireless_data;
1841 #endif
1842 	const struct net_device_ops *netdev_ops;
1843 	const struct ethtool_ops *ethtool_ops;
1844 #ifdef CONFIG_NET_L3_MASTER_DEV
1845 	const struct l3mdev_ops	*l3mdev_ops;
1846 #endif
1847 #if IS_ENABLED(CONFIG_IPV6)
1848 	const struct ndisc_ops *ndisc_ops;
1849 #endif
1850 
1851 #ifdef CONFIG_XFRM_OFFLOAD
1852 	const struct xfrmdev_ops *xfrmdev_ops;
1853 #endif
1854 
1855 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1856 	const struct tlsdev_ops *tlsdev_ops;
1857 #endif
1858 
1859 	const struct header_ops *header_ops;
1860 
1861 	unsigned int		flags;
1862 	unsigned int		priv_flags;
1863 
1864 	unsigned short		gflags;
1865 	unsigned short		padded;
1866 
1867 	unsigned char		operstate;
1868 	unsigned char		link_mode;
1869 
1870 	unsigned char		if_port;
1871 	unsigned char		dma;
1872 
1873 	unsigned int		mtu;
1874 	unsigned int		min_mtu;
1875 	unsigned int		max_mtu;
1876 	unsigned short		type;
1877 	unsigned short		hard_header_len;
1878 	unsigned char		min_header_len;
1879 
1880 	unsigned short		needed_headroom;
1881 	unsigned short		needed_tailroom;
1882 
1883 	/* Interface address info. */
1884 	unsigned char		perm_addr[MAX_ADDR_LEN];
1885 	unsigned char		addr_assign_type;
1886 	unsigned char		addr_len;
1887 	unsigned short		neigh_priv_len;
1888 	unsigned short          dev_id;
1889 	unsigned short          dev_port;
1890 	spinlock_t		addr_list_lock;
1891 	unsigned char		name_assign_type;
1892 	bool			uc_promisc;
1893 	struct netdev_hw_addr_list	uc;
1894 	struct netdev_hw_addr_list	mc;
1895 	struct netdev_hw_addr_list	dev_addrs;
1896 
1897 #ifdef CONFIG_SYSFS
1898 	struct kset		*queues_kset;
1899 #endif
1900 	unsigned int		promiscuity;
1901 	unsigned int		allmulti;
1902 
1903 
1904 	/* Protocol-specific pointers */
1905 
1906 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1907 	struct vlan_info __rcu	*vlan_info;
1908 #endif
1909 #if IS_ENABLED(CONFIG_NET_DSA)
1910 	struct dsa_port		*dsa_ptr;
1911 #endif
1912 #if IS_ENABLED(CONFIG_TIPC)
1913 	struct tipc_bearer __rcu *tipc_ptr;
1914 #endif
1915 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1916 	void 			*atalk_ptr;
1917 #endif
1918 	struct in_device __rcu	*ip_ptr;
1919 #if IS_ENABLED(CONFIG_DECNET)
1920 	struct dn_dev __rcu     *dn_ptr;
1921 #endif
1922 	struct inet6_dev __rcu	*ip6_ptr;
1923 #if IS_ENABLED(CONFIG_AX25)
1924 	void			*ax25_ptr;
1925 #endif
1926 	struct wireless_dev	*ieee80211_ptr;
1927 	struct wpan_dev		*ieee802154_ptr;
1928 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1929 	struct mpls_dev __rcu	*mpls_ptr;
1930 #endif
1931 
1932 /*
1933  * Cache lines mostly used on receive path (including eth_type_trans())
1934  */
1935 	/* Interface address info used in eth_type_trans() */
1936 	unsigned char		*dev_addr;
1937 
1938 	struct netdev_rx_queue	*_rx;
1939 	unsigned int		num_rx_queues;
1940 	unsigned int		real_num_rx_queues;
1941 
1942 	struct bpf_prog __rcu	*xdp_prog;
1943 	unsigned long		gro_flush_timeout;
1944 	rx_handler_func_t __rcu	*rx_handler;
1945 	void __rcu		*rx_handler_data;
1946 
1947 #ifdef CONFIG_NET_CLS_ACT
1948 	struct mini_Qdisc __rcu	*miniq_ingress;
1949 #endif
1950 	struct netdev_queue __rcu *ingress_queue;
1951 #ifdef CONFIG_NETFILTER_INGRESS
1952 	struct nf_hook_entries __rcu *nf_hooks_ingress;
1953 #endif
1954 
1955 	unsigned char		broadcast[MAX_ADDR_LEN];
1956 #ifdef CONFIG_RFS_ACCEL
1957 	struct cpu_rmap		*rx_cpu_rmap;
1958 #endif
1959 	struct hlist_node	index_hlist;
1960 
1961 /*
1962  * Cache lines mostly used on transmit path
1963  */
1964 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1965 	unsigned int		num_tx_queues;
1966 	unsigned int		real_num_tx_queues;
1967 	struct Qdisc		*qdisc;
1968 #ifdef CONFIG_NET_SCHED
1969 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1970 #endif
1971 	unsigned int		tx_queue_len;
1972 	spinlock_t		tx_global_lock;
1973 	int			watchdog_timeo;
1974 
1975 #ifdef CONFIG_XPS
1976 	struct xps_dev_maps __rcu *xps_cpus_map;
1977 	struct xps_dev_maps __rcu *xps_rxqs_map;
1978 #endif
1979 #ifdef CONFIG_NET_CLS_ACT
1980 	struct mini_Qdisc __rcu	*miniq_egress;
1981 #endif
1982 
1983 	/* These may be needed for future network-power-down code. */
1984 	struct timer_list	watchdog_timer;
1985 
1986 	int __percpu		*pcpu_refcnt;
1987 	struct list_head	todo_list;
1988 
1989 	struct list_head	link_watch_list;
1990 
1991 	enum { NETREG_UNINITIALIZED=0,
1992 	       NETREG_REGISTERED,	/* completed register_netdevice */
1993 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1994 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1995 	       NETREG_RELEASED,		/* called free_netdev */
1996 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1997 	} reg_state:8;
1998 
1999 	bool dismantle;
2000 
2001 	enum {
2002 		RTNL_LINK_INITIALIZED,
2003 		RTNL_LINK_INITIALIZING,
2004 	} rtnl_link_state:16;
2005 
2006 	bool needs_free_netdev;
2007 	void (*priv_destructor)(struct net_device *dev);
2008 
2009 #ifdef CONFIG_NETPOLL
2010 	struct netpoll_info __rcu	*npinfo;
2011 #endif
2012 
2013 	possible_net_t			nd_net;
2014 
2015 	/* mid-layer private */
2016 	union {
2017 		void					*ml_priv;
2018 		struct pcpu_lstats __percpu		*lstats;
2019 		struct pcpu_sw_netstats __percpu	*tstats;
2020 		struct pcpu_dstats __percpu		*dstats;
2021 	};
2022 
2023 #if IS_ENABLED(CONFIG_GARP)
2024 	struct garp_port __rcu	*garp_port;
2025 #endif
2026 #if IS_ENABLED(CONFIG_MRP)
2027 	struct mrp_port __rcu	*mrp_port;
2028 #endif
2029 
2030 	struct device		dev;
2031 	const struct attribute_group *sysfs_groups[4];
2032 	const struct attribute_group *sysfs_rx_queue_group;
2033 
2034 	const struct rtnl_link_ops *rtnl_link_ops;
2035 
2036 	/* for setting kernel sock attribute on TCP connection setup */
2037 #define GSO_MAX_SIZE		65536
2038 	unsigned int		gso_max_size;
2039 #define GSO_MAX_SEGS		65535
2040 	u16			gso_max_segs;
2041 
2042 #ifdef CONFIG_DCB
2043 	const struct dcbnl_rtnl_ops *dcbnl_ops;
2044 #endif
2045 	s16			num_tc;
2046 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
2047 	u8			prio_tc_map[TC_BITMASK + 1];
2048 
2049 #if IS_ENABLED(CONFIG_FCOE)
2050 	unsigned int		fcoe_ddp_xid;
2051 #endif
2052 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2053 	struct netprio_map __rcu *priomap;
2054 #endif
2055 	struct phy_device	*phydev;
2056 	struct sfp_bus		*sfp_bus;
2057 	struct lock_class_key	*qdisc_tx_busylock;
2058 	struct lock_class_key	*qdisc_running_key;
2059 	bool			proto_down;
2060 	unsigned		wol_enabled:1;
2061 };
2062 #define to_net_dev(d) container_of(d, struct net_device, dev)
2063 
2064 static inline bool netif_elide_gro(const struct net_device *dev)
2065 {
2066 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2067 		return true;
2068 	return false;
2069 }
2070 
2071 #define	NETDEV_ALIGN		32
2072 
2073 static inline
2074 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2075 {
2076 	return dev->prio_tc_map[prio & TC_BITMASK];
2077 }
2078 
2079 static inline
2080 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2081 {
2082 	if (tc >= dev->num_tc)
2083 		return -EINVAL;
2084 
2085 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2086 	return 0;
2087 }
2088 
2089 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2090 void netdev_reset_tc(struct net_device *dev);
2091 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2092 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2093 
2094 static inline
2095 int netdev_get_num_tc(struct net_device *dev)
2096 {
2097 	return dev->num_tc;
2098 }
2099 
2100 void netdev_unbind_sb_channel(struct net_device *dev,
2101 			      struct net_device *sb_dev);
2102 int netdev_bind_sb_channel_queue(struct net_device *dev,
2103 				 struct net_device *sb_dev,
2104 				 u8 tc, u16 count, u16 offset);
2105 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2106 static inline int netdev_get_sb_channel(struct net_device *dev)
2107 {
2108 	return max_t(int, -dev->num_tc, 0);
2109 }
2110 
2111 static inline
2112 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2113 					 unsigned int index)
2114 {
2115 	return &dev->_tx[index];
2116 }
2117 
2118 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2119 						    const struct sk_buff *skb)
2120 {
2121 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2122 }
2123 
2124 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2125 					    void (*f)(struct net_device *,
2126 						      struct netdev_queue *,
2127 						      void *),
2128 					    void *arg)
2129 {
2130 	unsigned int i;
2131 
2132 	for (i = 0; i < dev->num_tx_queues; i++)
2133 		f(dev, &dev->_tx[i], arg);
2134 }
2135 
2136 #define netdev_lockdep_set_classes(dev)				\
2137 {								\
2138 	static struct lock_class_key qdisc_tx_busylock_key;	\
2139 	static struct lock_class_key qdisc_running_key;		\
2140 	static struct lock_class_key qdisc_xmit_lock_key;	\
2141 	static struct lock_class_key dev_addr_list_lock_key;	\
2142 	unsigned int i;						\
2143 								\
2144 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
2145 	(dev)->qdisc_running_key = &qdisc_running_key;		\
2146 	lockdep_set_class(&(dev)->addr_list_lock,		\
2147 			  &dev_addr_list_lock_key); 		\
2148 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
2149 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
2150 				  &qdisc_xmit_lock_key);	\
2151 }
2152 
2153 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2154 		     struct net_device *sb_dev);
2155 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2156 					 struct sk_buff *skb,
2157 					 struct net_device *sb_dev);
2158 
2159 /* returns the headroom that the master device needs to take in account
2160  * when forwarding to this dev
2161  */
2162 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2163 {
2164 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2165 }
2166 
2167 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2168 {
2169 	if (dev->netdev_ops->ndo_set_rx_headroom)
2170 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2171 }
2172 
2173 /* set the device rx headroom to the dev's default */
2174 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2175 {
2176 	netdev_set_rx_headroom(dev, -1);
2177 }
2178 
2179 /*
2180  * Net namespace inlines
2181  */
2182 static inline
2183 struct net *dev_net(const struct net_device *dev)
2184 {
2185 	return read_pnet(&dev->nd_net);
2186 }
2187 
2188 static inline
2189 void dev_net_set(struct net_device *dev, struct net *net)
2190 {
2191 	write_pnet(&dev->nd_net, net);
2192 }
2193 
2194 /**
2195  *	netdev_priv - access network device private data
2196  *	@dev: network device
2197  *
2198  * Get network device private data
2199  */
2200 static inline void *netdev_priv(const struct net_device *dev)
2201 {
2202 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2203 }
2204 
2205 /* Set the sysfs physical device reference for the network logical device
2206  * if set prior to registration will cause a symlink during initialization.
2207  */
2208 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2209 
2210 /* Set the sysfs device type for the network logical device to allow
2211  * fine-grained identification of different network device types. For
2212  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2213  */
2214 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2215 
2216 /* Default NAPI poll() weight
2217  * Device drivers are strongly advised to not use bigger value
2218  */
2219 #define NAPI_POLL_WEIGHT 64
2220 
2221 /**
2222  *	netif_napi_add - initialize a NAPI context
2223  *	@dev:  network device
2224  *	@napi: NAPI context
2225  *	@poll: polling function
2226  *	@weight: default weight
2227  *
2228  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2229  * *any* of the other NAPI-related functions.
2230  */
2231 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2232 		    int (*poll)(struct napi_struct *, int), int weight);
2233 
2234 /**
2235  *	netif_tx_napi_add - initialize a NAPI context
2236  *	@dev:  network device
2237  *	@napi: NAPI context
2238  *	@poll: polling function
2239  *	@weight: default weight
2240  *
2241  * This variant of netif_napi_add() should be used from drivers using NAPI
2242  * to exclusively poll a TX queue.
2243  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2244  */
2245 static inline void netif_tx_napi_add(struct net_device *dev,
2246 				     struct napi_struct *napi,
2247 				     int (*poll)(struct napi_struct *, int),
2248 				     int weight)
2249 {
2250 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2251 	netif_napi_add(dev, napi, poll, weight);
2252 }
2253 
2254 /**
2255  *  netif_napi_del - remove a NAPI context
2256  *  @napi: NAPI context
2257  *
2258  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2259  */
2260 void netif_napi_del(struct napi_struct *napi);
2261 
2262 struct napi_gro_cb {
2263 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2264 	void	*frag0;
2265 
2266 	/* Length of frag0. */
2267 	unsigned int frag0_len;
2268 
2269 	/* This indicates where we are processing relative to skb->data. */
2270 	int	data_offset;
2271 
2272 	/* This is non-zero if the packet cannot be merged with the new skb. */
2273 	u16	flush;
2274 
2275 	/* Save the IP ID here and check when we get to the transport layer */
2276 	u16	flush_id;
2277 
2278 	/* Number of segments aggregated. */
2279 	u16	count;
2280 
2281 	/* Start offset for remote checksum offload */
2282 	u16	gro_remcsum_start;
2283 
2284 	/* jiffies when first packet was created/queued */
2285 	unsigned long age;
2286 
2287 	/* Used in ipv6_gro_receive() and foo-over-udp */
2288 	u16	proto;
2289 
2290 	/* This is non-zero if the packet may be of the same flow. */
2291 	u8	same_flow:1;
2292 
2293 	/* Used in tunnel GRO receive */
2294 	u8	encap_mark:1;
2295 
2296 	/* GRO checksum is valid */
2297 	u8	csum_valid:1;
2298 
2299 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2300 	u8	csum_cnt:3;
2301 
2302 	/* Free the skb? */
2303 	u8	free:2;
2304 #define NAPI_GRO_FREE		  1
2305 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2306 
2307 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2308 	u8	is_ipv6:1;
2309 
2310 	/* Used in GRE, set in fou/gue_gro_receive */
2311 	u8	is_fou:1;
2312 
2313 	/* Used to determine if flush_id can be ignored */
2314 	u8	is_atomic:1;
2315 
2316 	/* Number of gro_receive callbacks this packet already went through */
2317 	u8 recursion_counter:4;
2318 
2319 	/* 1 bit hole */
2320 
2321 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2322 	__wsum	csum;
2323 
2324 	/* used in skb_gro_receive() slow path */
2325 	struct sk_buff *last;
2326 };
2327 
2328 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2329 
2330 #define GRO_RECURSION_LIMIT 15
2331 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2332 {
2333 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2334 }
2335 
2336 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2337 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2338 					       struct list_head *head,
2339 					       struct sk_buff *skb)
2340 {
2341 	if (unlikely(gro_recursion_inc_test(skb))) {
2342 		NAPI_GRO_CB(skb)->flush |= 1;
2343 		return NULL;
2344 	}
2345 
2346 	return cb(head, skb);
2347 }
2348 
2349 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2350 					    struct sk_buff *);
2351 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2352 						  struct sock *sk,
2353 						  struct list_head *head,
2354 						  struct sk_buff *skb)
2355 {
2356 	if (unlikely(gro_recursion_inc_test(skb))) {
2357 		NAPI_GRO_CB(skb)->flush |= 1;
2358 		return NULL;
2359 	}
2360 
2361 	return cb(sk, head, skb);
2362 }
2363 
2364 struct packet_type {
2365 	__be16			type;	/* This is really htons(ether_type). */
2366 	bool			ignore_outgoing;
2367 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2368 	int			(*func) (struct sk_buff *,
2369 					 struct net_device *,
2370 					 struct packet_type *,
2371 					 struct net_device *);
2372 	void			(*list_func) (struct list_head *,
2373 					      struct packet_type *,
2374 					      struct net_device *);
2375 	bool			(*id_match)(struct packet_type *ptype,
2376 					    struct sock *sk);
2377 	void			*af_packet_priv;
2378 	struct list_head	list;
2379 };
2380 
2381 struct offload_callbacks {
2382 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2383 						netdev_features_t features);
2384 	struct sk_buff		*(*gro_receive)(struct list_head *head,
2385 						struct sk_buff *skb);
2386 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2387 };
2388 
2389 struct packet_offload {
2390 	__be16			 type;	/* This is really htons(ether_type). */
2391 	u16			 priority;
2392 	struct offload_callbacks callbacks;
2393 	struct list_head	 list;
2394 };
2395 
2396 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2397 struct pcpu_sw_netstats {
2398 	u64     rx_packets;
2399 	u64     rx_bytes;
2400 	u64     tx_packets;
2401 	u64     tx_bytes;
2402 	struct u64_stats_sync   syncp;
2403 } __aligned(4 * sizeof(u64));
2404 
2405 struct pcpu_lstats {
2406 	u64 packets;
2407 	u64 bytes;
2408 	struct u64_stats_sync syncp;
2409 } __aligned(2 * sizeof(u64));
2410 
2411 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2412 ({									\
2413 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2414 	if (pcpu_stats)	{						\
2415 		int __cpu;						\
2416 		for_each_possible_cpu(__cpu) {				\
2417 			typeof(type) *stat;				\
2418 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2419 			u64_stats_init(&stat->syncp);			\
2420 		}							\
2421 	}								\
2422 	pcpu_stats;							\
2423 })
2424 
2425 #define netdev_alloc_pcpu_stats(type)					\
2426 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2427 
2428 enum netdev_lag_tx_type {
2429 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2430 	NETDEV_LAG_TX_TYPE_RANDOM,
2431 	NETDEV_LAG_TX_TYPE_BROADCAST,
2432 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2433 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2434 	NETDEV_LAG_TX_TYPE_HASH,
2435 };
2436 
2437 enum netdev_lag_hash {
2438 	NETDEV_LAG_HASH_NONE,
2439 	NETDEV_LAG_HASH_L2,
2440 	NETDEV_LAG_HASH_L34,
2441 	NETDEV_LAG_HASH_L23,
2442 	NETDEV_LAG_HASH_E23,
2443 	NETDEV_LAG_HASH_E34,
2444 	NETDEV_LAG_HASH_UNKNOWN,
2445 };
2446 
2447 struct netdev_lag_upper_info {
2448 	enum netdev_lag_tx_type tx_type;
2449 	enum netdev_lag_hash hash_type;
2450 };
2451 
2452 struct netdev_lag_lower_state_info {
2453 	u8 link_up : 1,
2454 	   tx_enabled : 1;
2455 };
2456 
2457 #include <linux/notifier.h>
2458 
2459 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2460  * and the rtnetlink notification exclusion list in rtnetlink_event() when
2461  * adding new types.
2462  */
2463 enum netdev_cmd {
2464 	NETDEV_UP	= 1,	/* For now you can't veto a device up/down */
2465 	NETDEV_DOWN,
2466 	NETDEV_REBOOT,		/* Tell a protocol stack a network interface
2467 				   detected a hardware crash and restarted
2468 				   - we can use this eg to kick tcp sessions
2469 				   once done */
2470 	NETDEV_CHANGE,		/* Notify device state change */
2471 	NETDEV_REGISTER,
2472 	NETDEV_UNREGISTER,
2473 	NETDEV_CHANGEMTU,	/* notify after mtu change happened */
2474 	NETDEV_CHANGEADDR,	/* notify after the address change */
2475 	NETDEV_PRE_CHANGEADDR,	/* notify before the address change */
2476 	NETDEV_GOING_DOWN,
2477 	NETDEV_CHANGENAME,
2478 	NETDEV_FEAT_CHANGE,
2479 	NETDEV_BONDING_FAILOVER,
2480 	NETDEV_PRE_UP,
2481 	NETDEV_PRE_TYPE_CHANGE,
2482 	NETDEV_POST_TYPE_CHANGE,
2483 	NETDEV_POST_INIT,
2484 	NETDEV_RELEASE,
2485 	NETDEV_NOTIFY_PEERS,
2486 	NETDEV_JOIN,
2487 	NETDEV_CHANGEUPPER,
2488 	NETDEV_RESEND_IGMP,
2489 	NETDEV_PRECHANGEMTU,	/* notify before mtu change happened */
2490 	NETDEV_CHANGEINFODATA,
2491 	NETDEV_BONDING_INFO,
2492 	NETDEV_PRECHANGEUPPER,
2493 	NETDEV_CHANGELOWERSTATE,
2494 	NETDEV_UDP_TUNNEL_PUSH_INFO,
2495 	NETDEV_UDP_TUNNEL_DROP_INFO,
2496 	NETDEV_CHANGE_TX_QUEUE_LEN,
2497 	NETDEV_CVLAN_FILTER_PUSH_INFO,
2498 	NETDEV_CVLAN_FILTER_DROP_INFO,
2499 	NETDEV_SVLAN_FILTER_PUSH_INFO,
2500 	NETDEV_SVLAN_FILTER_DROP_INFO,
2501 };
2502 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2503 
2504 int register_netdevice_notifier(struct notifier_block *nb);
2505 int unregister_netdevice_notifier(struct notifier_block *nb);
2506 
2507 struct netdev_notifier_info {
2508 	struct net_device	*dev;
2509 	struct netlink_ext_ack	*extack;
2510 };
2511 
2512 struct netdev_notifier_info_ext {
2513 	struct netdev_notifier_info info; /* must be first */
2514 	union {
2515 		u32 mtu;
2516 	} ext;
2517 };
2518 
2519 struct netdev_notifier_change_info {
2520 	struct netdev_notifier_info info; /* must be first */
2521 	unsigned int flags_changed;
2522 };
2523 
2524 struct netdev_notifier_changeupper_info {
2525 	struct netdev_notifier_info info; /* must be first */
2526 	struct net_device *upper_dev; /* new upper dev */
2527 	bool master; /* is upper dev master */
2528 	bool linking; /* is the notification for link or unlink */
2529 	void *upper_info; /* upper dev info */
2530 };
2531 
2532 struct netdev_notifier_changelowerstate_info {
2533 	struct netdev_notifier_info info; /* must be first */
2534 	void *lower_state_info; /* is lower dev state */
2535 };
2536 
2537 struct netdev_notifier_pre_changeaddr_info {
2538 	struct netdev_notifier_info info; /* must be first */
2539 	const unsigned char *dev_addr;
2540 };
2541 
2542 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2543 					     struct net_device *dev)
2544 {
2545 	info->dev = dev;
2546 	info->extack = NULL;
2547 }
2548 
2549 static inline struct net_device *
2550 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2551 {
2552 	return info->dev;
2553 }
2554 
2555 static inline struct netlink_ext_ack *
2556 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2557 {
2558 	return info->extack;
2559 }
2560 
2561 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2562 
2563 
2564 extern rwlock_t				dev_base_lock;		/* Device list lock */
2565 
2566 #define for_each_netdev(net, d)		\
2567 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2568 #define for_each_netdev_reverse(net, d)	\
2569 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2570 #define for_each_netdev_rcu(net, d)		\
2571 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2572 #define for_each_netdev_safe(net, d, n)	\
2573 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2574 #define for_each_netdev_continue(net, d)		\
2575 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2576 #define for_each_netdev_continue_rcu(net, d)		\
2577 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2578 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2579 		for_each_netdev_rcu(&init_net, slave)	\
2580 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2581 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2582 
2583 static inline struct net_device *next_net_device(struct net_device *dev)
2584 {
2585 	struct list_head *lh;
2586 	struct net *net;
2587 
2588 	net = dev_net(dev);
2589 	lh = dev->dev_list.next;
2590 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2591 }
2592 
2593 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2594 {
2595 	struct list_head *lh;
2596 	struct net *net;
2597 
2598 	net = dev_net(dev);
2599 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2600 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2601 }
2602 
2603 static inline struct net_device *first_net_device(struct net *net)
2604 {
2605 	return list_empty(&net->dev_base_head) ? NULL :
2606 		net_device_entry(net->dev_base_head.next);
2607 }
2608 
2609 static inline struct net_device *first_net_device_rcu(struct net *net)
2610 {
2611 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2612 
2613 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2614 }
2615 
2616 int netdev_boot_setup_check(struct net_device *dev);
2617 unsigned long netdev_boot_base(const char *prefix, int unit);
2618 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2619 				       const char *hwaddr);
2620 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2621 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2622 void dev_add_pack(struct packet_type *pt);
2623 void dev_remove_pack(struct packet_type *pt);
2624 void __dev_remove_pack(struct packet_type *pt);
2625 void dev_add_offload(struct packet_offload *po);
2626 void dev_remove_offload(struct packet_offload *po);
2627 
2628 int dev_get_iflink(const struct net_device *dev);
2629 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2630 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2631 				      unsigned short mask);
2632 struct net_device *dev_get_by_name(struct net *net, const char *name);
2633 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2634 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2635 int dev_alloc_name(struct net_device *dev, const char *name);
2636 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2637 void dev_close(struct net_device *dev);
2638 void dev_close_many(struct list_head *head, bool unlink);
2639 void dev_disable_lro(struct net_device *dev);
2640 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2641 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2642 		     struct net_device *sb_dev);
2643 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2644 		       struct net_device *sb_dev);
2645 int dev_queue_xmit(struct sk_buff *skb);
2646 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2647 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2648 int register_netdevice(struct net_device *dev);
2649 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2650 void unregister_netdevice_many(struct list_head *head);
2651 static inline void unregister_netdevice(struct net_device *dev)
2652 {
2653 	unregister_netdevice_queue(dev, NULL);
2654 }
2655 
2656 int netdev_refcnt_read(const struct net_device *dev);
2657 void free_netdev(struct net_device *dev);
2658 void netdev_freemem(struct net_device *dev);
2659 void synchronize_net(void);
2660 int init_dummy_netdev(struct net_device *dev);
2661 
2662 DECLARE_PER_CPU(int, xmit_recursion);
2663 #define XMIT_RECURSION_LIMIT	10
2664 
2665 static inline int dev_recursion_level(void)
2666 {
2667 	return this_cpu_read(xmit_recursion);
2668 }
2669 
2670 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2671 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2672 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2673 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2674 int netdev_get_name(struct net *net, char *name, int ifindex);
2675 int dev_restart(struct net_device *dev);
2676 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2677 
2678 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2679 {
2680 	return NAPI_GRO_CB(skb)->data_offset;
2681 }
2682 
2683 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2684 {
2685 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2686 }
2687 
2688 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2689 {
2690 	NAPI_GRO_CB(skb)->data_offset += len;
2691 }
2692 
2693 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2694 					unsigned int offset)
2695 {
2696 	return NAPI_GRO_CB(skb)->frag0 + offset;
2697 }
2698 
2699 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2700 {
2701 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2702 }
2703 
2704 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2705 {
2706 	NAPI_GRO_CB(skb)->frag0 = NULL;
2707 	NAPI_GRO_CB(skb)->frag0_len = 0;
2708 }
2709 
2710 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2711 					unsigned int offset)
2712 {
2713 	if (!pskb_may_pull(skb, hlen))
2714 		return NULL;
2715 
2716 	skb_gro_frag0_invalidate(skb);
2717 	return skb->data + offset;
2718 }
2719 
2720 static inline void *skb_gro_network_header(struct sk_buff *skb)
2721 {
2722 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2723 	       skb_network_offset(skb);
2724 }
2725 
2726 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2727 					const void *start, unsigned int len)
2728 {
2729 	if (NAPI_GRO_CB(skb)->csum_valid)
2730 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2731 						  csum_partial(start, len, 0));
2732 }
2733 
2734 /* GRO checksum functions. These are logical equivalents of the normal
2735  * checksum functions (in skbuff.h) except that they operate on the GRO
2736  * offsets and fields in sk_buff.
2737  */
2738 
2739 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2740 
2741 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2742 {
2743 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2744 }
2745 
2746 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2747 						      bool zero_okay,
2748 						      __sum16 check)
2749 {
2750 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2751 		skb_checksum_start_offset(skb) <
2752 		 skb_gro_offset(skb)) &&
2753 		!skb_at_gro_remcsum_start(skb) &&
2754 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2755 		(!zero_okay || check));
2756 }
2757 
2758 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2759 							   __wsum psum)
2760 {
2761 	if (NAPI_GRO_CB(skb)->csum_valid &&
2762 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2763 		return 0;
2764 
2765 	NAPI_GRO_CB(skb)->csum = psum;
2766 
2767 	return __skb_gro_checksum_complete(skb);
2768 }
2769 
2770 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2771 {
2772 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2773 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2774 		NAPI_GRO_CB(skb)->csum_cnt--;
2775 	} else {
2776 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2777 		 * verified a new top level checksum or an encapsulated one
2778 		 * during GRO. This saves work if we fallback to normal path.
2779 		 */
2780 		__skb_incr_checksum_unnecessary(skb);
2781 	}
2782 }
2783 
2784 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2785 				    compute_pseudo)			\
2786 ({									\
2787 	__sum16 __ret = 0;						\
2788 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2789 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2790 				compute_pseudo(skb, proto));		\
2791 	if (!__ret)							\
2792 		skb_gro_incr_csum_unnecessary(skb);			\
2793 	__ret;								\
2794 })
2795 
2796 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2797 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2798 
2799 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2800 					     compute_pseudo)		\
2801 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2802 
2803 #define skb_gro_checksum_simple_validate(skb)				\
2804 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2805 
2806 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2807 {
2808 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2809 		!NAPI_GRO_CB(skb)->csum_valid);
2810 }
2811 
2812 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2813 					      __sum16 check, __wsum pseudo)
2814 {
2815 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2816 	NAPI_GRO_CB(skb)->csum_valid = 1;
2817 }
2818 
2819 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2820 do {									\
2821 	if (__skb_gro_checksum_convert_check(skb))			\
2822 		__skb_gro_checksum_convert(skb, check,			\
2823 					   compute_pseudo(skb, proto));	\
2824 } while (0)
2825 
2826 struct gro_remcsum {
2827 	int offset;
2828 	__wsum delta;
2829 };
2830 
2831 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2832 {
2833 	grc->offset = 0;
2834 	grc->delta = 0;
2835 }
2836 
2837 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2838 					    unsigned int off, size_t hdrlen,
2839 					    int start, int offset,
2840 					    struct gro_remcsum *grc,
2841 					    bool nopartial)
2842 {
2843 	__wsum delta;
2844 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2845 
2846 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2847 
2848 	if (!nopartial) {
2849 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2850 		return ptr;
2851 	}
2852 
2853 	ptr = skb_gro_header_fast(skb, off);
2854 	if (skb_gro_header_hard(skb, off + plen)) {
2855 		ptr = skb_gro_header_slow(skb, off + plen, off);
2856 		if (!ptr)
2857 			return NULL;
2858 	}
2859 
2860 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2861 			       start, offset);
2862 
2863 	/* Adjust skb->csum since we changed the packet */
2864 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2865 
2866 	grc->offset = off + hdrlen + offset;
2867 	grc->delta = delta;
2868 
2869 	return ptr;
2870 }
2871 
2872 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2873 					   struct gro_remcsum *grc)
2874 {
2875 	void *ptr;
2876 	size_t plen = grc->offset + sizeof(u16);
2877 
2878 	if (!grc->delta)
2879 		return;
2880 
2881 	ptr = skb_gro_header_fast(skb, grc->offset);
2882 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2883 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2884 		if (!ptr)
2885 			return;
2886 	}
2887 
2888 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2889 }
2890 
2891 #ifdef CONFIG_XFRM_OFFLOAD
2892 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2893 {
2894 	if (PTR_ERR(pp) != -EINPROGRESS)
2895 		NAPI_GRO_CB(skb)->flush |= flush;
2896 }
2897 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2898 					       struct sk_buff *pp,
2899 					       int flush,
2900 					       struct gro_remcsum *grc)
2901 {
2902 	if (PTR_ERR(pp) != -EINPROGRESS) {
2903 		NAPI_GRO_CB(skb)->flush |= flush;
2904 		skb_gro_remcsum_cleanup(skb, grc);
2905 		skb->remcsum_offload = 0;
2906 	}
2907 }
2908 #else
2909 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2910 {
2911 	NAPI_GRO_CB(skb)->flush |= flush;
2912 }
2913 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2914 					       struct sk_buff *pp,
2915 					       int flush,
2916 					       struct gro_remcsum *grc)
2917 {
2918 	NAPI_GRO_CB(skb)->flush |= flush;
2919 	skb_gro_remcsum_cleanup(skb, grc);
2920 	skb->remcsum_offload = 0;
2921 }
2922 #endif
2923 
2924 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2925 				  unsigned short type,
2926 				  const void *daddr, const void *saddr,
2927 				  unsigned int len)
2928 {
2929 	if (!dev->header_ops || !dev->header_ops->create)
2930 		return 0;
2931 
2932 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2933 }
2934 
2935 static inline int dev_parse_header(const struct sk_buff *skb,
2936 				   unsigned char *haddr)
2937 {
2938 	const struct net_device *dev = skb->dev;
2939 
2940 	if (!dev->header_ops || !dev->header_ops->parse)
2941 		return 0;
2942 	return dev->header_ops->parse(skb, haddr);
2943 }
2944 
2945 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2946 {
2947 	const struct net_device *dev = skb->dev;
2948 
2949 	if (!dev->header_ops || !dev->header_ops->parse_protocol)
2950 		return 0;
2951 	return dev->header_ops->parse_protocol(skb);
2952 }
2953 
2954 /* ll_header must have at least hard_header_len allocated */
2955 static inline bool dev_validate_header(const struct net_device *dev,
2956 				       char *ll_header, int len)
2957 {
2958 	if (likely(len >= dev->hard_header_len))
2959 		return true;
2960 	if (len < dev->min_header_len)
2961 		return false;
2962 
2963 	if (capable(CAP_SYS_RAWIO)) {
2964 		memset(ll_header + len, 0, dev->hard_header_len - len);
2965 		return true;
2966 	}
2967 
2968 	if (dev->header_ops && dev->header_ops->validate)
2969 		return dev->header_ops->validate(ll_header, len);
2970 
2971 	return false;
2972 }
2973 
2974 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2975 			   int len, int size);
2976 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2977 static inline int unregister_gifconf(unsigned int family)
2978 {
2979 	return register_gifconf(family, NULL);
2980 }
2981 
2982 #ifdef CONFIG_NET_FLOW_LIMIT
2983 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2984 struct sd_flow_limit {
2985 	u64			count;
2986 	unsigned int		num_buckets;
2987 	unsigned int		history_head;
2988 	u16			history[FLOW_LIMIT_HISTORY];
2989 	u8			buckets[];
2990 };
2991 
2992 extern int netdev_flow_limit_table_len;
2993 #endif /* CONFIG_NET_FLOW_LIMIT */
2994 
2995 /*
2996  * Incoming packets are placed on per-CPU queues
2997  */
2998 struct softnet_data {
2999 	struct list_head	poll_list;
3000 	struct sk_buff_head	process_queue;
3001 
3002 	/* stats */
3003 	unsigned int		processed;
3004 	unsigned int		time_squeeze;
3005 	unsigned int		received_rps;
3006 #ifdef CONFIG_RPS
3007 	struct softnet_data	*rps_ipi_list;
3008 #endif
3009 #ifdef CONFIG_NET_FLOW_LIMIT
3010 	struct sd_flow_limit __rcu *flow_limit;
3011 #endif
3012 	struct Qdisc		*output_queue;
3013 	struct Qdisc		**output_queue_tailp;
3014 	struct sk_buff		*completion_queue;
3015 #ifdef CONFIG_XFRM_OFFLOAD
3016 	struct sk_buff_head	xfrm_backlog;
3017 #endif
3018 #ifdef CONFIG_RPS
3019 	/* input_queue_head should be written by cpu owning this struct,
3020 	 * and only read by other cpus. Worth using a cache line.
3021 	 */
3022 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
3023 
3024 	/* Elements below can be accessed between CPUs for RPS/RFS */
3025 	call_single_data_t	csd ____cacheline_aligned_in_smp;
3026 	struct softnet_data	*rps_ipi_next;
3027 	unsigned int		cpu;
3028 	unsigned int		input_queue_tail;
3029 #endif
3030 	unsigned int		dropped;
3031 	struct sk_buff_head	input_pkt_queue;
3032 	struct napi_struct	backlog;
3033 
3034 };
3035 
3036 static inline void input_queue_head_incr(struct softnet_data *sd)
3037 {
3038 #ifdef CONFIG_RPS
3039 	sd->input_queue_head++;
3040 #endif
3041 }
3042 
3043 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3044 					      unsigned int *qtail)
3045 {
3046 #ifdef CONFIG_RPS
3047 	*qtail = ++sd->input_queue_tail;
3048 #endif
3049 }
3050 
3051 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3052 
3053 void __netif_schedule(struct Qdisc *q);
3054 void netif_schedule_queue(struct netdev_queue *txq);
3055 
3056 static inline void netif_tx_schedule_all(struct net_device *dev)
3057 {
3058 	unsigned int i;
3059 
3060 	for (i = 0; i < dev->num_tx_queues; i++)
3061 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
3062 }
3063 
3064 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3065 {
3066 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3067 }
3068 
3069 /**
3070  *	netif_start_queue - allow transmit
3071  *	@dev: network device
3072  *
3073  *	Allow upper layers to call the device hard_start_xmit routine.
3074  */
3075 static inline void netif_start_queue(struct net_device *dev)
3076 {
3077 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3078 }
3079 
3080 static inline void netif_tx_start_all_queues(struct net_device *dev)
3081 {
3082 	unsigned int i;
3083 
3084 	for (i = 0; i < dev->num_tx_queues; i++) {
3085 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3086 		netif_tx_start_queue(txq);
3087 	}
3088 }
3089 
3090 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3091 
3092 /**
3093  *	netif_wake_queue - restart transmit
3094  *	@dev: network device
3095  *
3096  *	Allow upper layers to call the device hard_start_xmit routine.
3097  *	Used for flow control when transmit resources are available.
3098  */
3099 static inline void netif_wake_queue(struct net_device *dev)
3100 {
3101 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3102 }
3103 
3104 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3105 {
3106 	unsigned int i;
3107 
3108 	for (i = 0; i < dev->num_tx_queues; i++) {
3109 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3110 		netif_tx_wake_queue(txq);
3111 	}
3112 }
3113 
3114 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3115 {
3116 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3117 }
3118 
3119 /**
3120  *	netif_stop_queue - stop transmitted packets
3121  *	@dev: network device
3122  *
3123  *	Stop upper layers calling the device hard_start_xmit routine.
3124  *	Used for flow control when transmit resources are unavailable.
3125  */
3126 static inline void netif_stop_queue(struct net_device *dev)
3127 {
3128 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3129 }
3130 
3131 void netif_tx_stop_all_queues(struct net_device *dev);
3132 
3133 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3134 {
3135 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3136 }
3137 
3138 /**
3139  *	netif_queue_stopped - test if transmit queue is flowblocked
3140  *	@dev: network device
3141  *
3142  *	Test if transmit queue on device is currently unable to send.
3143  */
3144 static inline bool netif_queue_stopped(const struct net_device *dev)
3145 {
3146 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3147 }
3148 
3149 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3150 {
3151 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3152 }
3153 
3154 static inline bool
3155 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3156 {
3157 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3158 }
3159 
3160 static inline bool
3161 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3162 {
3163 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3164 }
3165 
3166 /**
3167  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3168  *	@dev_queue: pointer to transmit queue
3169  *
3170  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3171  * to give appropriate hint to the CPU.
3172  */
3173 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3174 {
3175 #ifdef CONFIG_BQL
3176 	prefetchw(&dev_queue->dql.num_queued);
3177 #endif
3178 }
3179 
3180 /**
3181  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3182  *	@dev_queue: pointer to transmit queue
3183  *
3184  * BQL enabled drivers might use this helper in their TX completion path,
3185  * to give appropriate hint to the CPU.
3186  */
3187 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3188 {
3189 #ifdef CONFIG_BQL
3190 	prefetchw(&dev_queue->dql.limit);
3191 #endif
3192 }
3193 
3194 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3195 					unsigned int bytes)
3196 {
3197 #ifdef CONFIG_BQL
3198 	dql_queued(&dev_queue->dql, bytes);
3199 
3200 	if (likely(dql_avail(&dev_queue->dql) >= 0))
3201 		return;
3202 
3203 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3204 
3205 	/*
3206 	 * The XOFF flag must be set before checking the dql_avail below,
3207 	 * because in netdev_tx_completed_queue we update the dql_completed
3208 	 * before checking the XOFF flag.
3209 	 */
3210 	smp_mb();
3211 
3212 	/* check again in case another CPU has just made room avail */
3213 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3214 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3215 #endif
3216 }
3217 
3218 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3219  * that they should not test BQL status themselves.
3220  * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3221  * skb of a batch.
3222  * Returns true if the doorbell must be used to kick the NIC.
3223  */
3224 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3225 					  unsigned int bytes,
3226 					  bool xmit_more)
3227 {
3228 	if (xmit_more) {
3229 #ifdef CONFIG_BQL
3230 		dql_queued(&dev_queue->dql, bytes);
3231 #endif
3232 		return netif_tx_queue_stopped(dev_queue);
3233 	}
3234 	netdev_tx_sent_queue(dev_queue, bytes);
3235 	return true;
3236 }
3237 
3238 /**
3239  * 	netdev_sent_queue - report the number of bytes queued to hardware
3240  * 	@dev: network device
3241  * 	@bytes: number of bytes queued to the hardware device queue
3242  *
3243  * 	Report the number of bytes queued for sending/completion to the network
3244  * 	device hardware queue. @bytes should be a good approximation and should
3245  * 	exactly match netdev_completed_queue() @bytes
3246  */
3247 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3248 {
3249 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3250 }
3251 
3252 static inline bool __netdev_sent_queue(struct net_device *dev,
3253 				       unsigned int bytes,
3254 				       bool xmit_more)
3255 {
3256 	return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3257 				      xmit_more);
3258 }
3259 
3260 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3261 					     unsigned int pkts, unsigned int bytes)
3262 {
3263 #ifdef CONFIG_BQL
3264 	if (unlikely(!bytes))
3265 		return;
3266 
3267 	dql_completed(&dev_queue->dql, bytes);
3268 
3269 	/*
3270 	 * Without the memory barrier there is a small possiblity that
3271 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3272 	 * be stopped forever
3273 	 */
3274 	smp_mb();
3275 
3276 	if (dql_avail(&dev_queue->dql) < 0)
3277 		return;
3278 
3279 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3280 		netif_schedule_queue(dev_queue);
3281 #endif
3282 }
3283 
3284 /**
3285  * 	netdev_completed_queue - report bytes and packets completed by device
3286  * 	@dev: network device
3287  * 	@pkts: actual number of packets sent over the medium
3288  * 	@bytes: actual number of bytes sent over the medium
3289  *
3290  * 	Report the number of bytes and packets transmitted by the network device
3291  * 	hardware queue over the physical medium, @bytes must exactly match the
3292  * 	@bytes amount passed to netdev_sent_queue()
3293  */
3294 static inline void netdev_completed_queue(struct net_device *dev,
3295 					  unsigned int pkts, unsigned int bytes)
3296 {
3297 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3298 }
3299 
3300 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3301 {
3302 #ifdef CONFIG_BQL
3303 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3304 	dql_reset(&q->dql);
3305 #endif
3306 }
3307 
3308 /**
3309  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3310  * 	@dev_queue: network device
3311  *
3312  * 	Reset the bytes and packet count of a network device and clear the
3313  * 	software flow control OFF bit for this network device
3314  */
3315 static inline void netdev_reset_queue(struct net_device *dev_queue)
3316 {
3317 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3318 }
3319 
3320 /**
3321  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3322  * 	@dev: network device
3323  * 	@queue_index: given tx queue index
3324  *
3325  * 	Returns 0 if given tx queue index >= number of device tx queues,
3326  * 	otherwise returns the originally passed tx queue index.
3327  */
3328 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3329 {
3330 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3331 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3332 				     dev->name, queue_index,
3333 				     dev->real_num_tx_queues);
3334 		return 0;
3335 	}
3336 
3337 	return queue_index;
3338 }
3339 
3340 /**
3341  *	netif_running - test if up
3342  *	@dev: network device
3343  *
3344  *	Test if the device has been brought up.
3345  */
3346 static inline bool netif_running(const struct net_device *dev)
3347 {
3348 	return test_bit(__LINK_STATE_START, &dev->state);
3349 }
3350 
3351 /*
3352  * Routines to manage the subqueues on a device.  We only need start,
3353  * stop, and a check if it's stopped.  All other device management is
3354  * done at the overall netdevice level.
3355  * Also test the device if we're multiqueue.
3356  */
3357 
3358 /**
3359  *	netif_start_subqueue - allow sending packets on subqueue
3360  *	@dev: network device
3361  *	@queue_index: sub queue index
3362  *
3363  * Start individual transmit queue of a device with multiple transmit queues.
3364  */
3365 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3366 {
3367 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3368 
3369 	netif_tx_start_queue(txq);
3370 }
3371 
3372 /**
3373  *	netif_stop_subqueue - stop sending packets on subqueue
3374  *	@dev: network device
3375  *	@queue_index: sub queue index
3376  *
3377  * Stop individual transmit queue of a device with multiple transmit queues.
3378  */
3379 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3380 {
3381 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3382 	netif_tx_stop_queue(txq);
3383 }
3384 
3385 /**
3386  *	netif_subqueue_stopped - test status of subqueue
3387  *	@dev: network device
3388  *	@queue_index: sub queue index
3389  *
3390  * Check individual transmit queue of a device with multiple transmit queues.
3391  */
3392 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3393 					    u16 queue_index)
3394 {
3395 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3396 
3397 	return netif_tx_queue_stopped(txq);
3398 }
3399 
3400 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3401 					  struct sk_buff *skb)
3402 {
3403 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3404 }
3405 
3406 /**
3407  *	netif_wake_subqueue - allow sending packets on subqueue
3408  *	@dev: network device
3409  *	@queue_index: sub queue index
3410  *
3411  * Resume individual transmit queue of a device with multiple transmit queues.
3412  */
3413 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3414 {
3415 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3416 
3417 	netif_tx_wake_queue(txq);
3418 }
3419 
3420 #ifdef CONFIG_XPS
3421 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3422 			u16 index);
3423 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3424 			  u16 index, bool is_rxqs_map);
3425 
3426 /**
3427  *	netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3428  *	@j: CPU/Rx queue index
3429  *	@mask: bitmask of all cpus/rx queues
3430  *	@nr_bits: number of bits in the bitmask
3431  *
3432  * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3433  */
3434 static inline bool netif_attr_test_mask(unsigned long j,
3435 					const unsigned long *mask,
3436 					unsigned int nr_bits)
3437 {
3438 	cpu_max_bits_warn(j, nr_bits);
3439 	return test_bit(j, mask);
3440 }
3441 
3442 /**
3443  *	netif_attr_test_online - Test for online CPU/Rx queue
3444  *	@j: CPU/Rx queue index
3445  *	@online_mask: bitmask for CPUs/Rx queues that are online
3446  *	@nr_bits: number of bits in the bitmask
3447  *
3448  * Returns true if a CPU/Rx queue is online.
3449  */
3450 static inline bool netif_attr_test_online(unsigned long j,
3451 					  const unsigned long *online_mask,
3452 					  unsigned int nr_bits)
3453 {
3454 	cpu_max_bits_warn(j, nr_bits);
3455 
3456 	if (online_mask)
3457 		return test_bit(j, online_mask);
3458 
3459 	return (j < nr_bits);
3460 }
3461 
3462 /**
3463  *	netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3464  *	@n: CPU/Rx queue index
3465  *	@srcp: the cpumask/Rx queue mask pointer
3466  *	@nr_bits: number of bits in the bitmask
3467  *
3468  * Returns >= nr_bits if no further CPUs/Rx queues set.
3469  */
3470 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3471 					       unsigned int nr_bits)
3472 {
3473 	/* -1 is a legal arg here. */
3474 	if (n != -1)
3475 		cpu_max_bits_warn(n, nr_bits);
3476 
3477 	if (srcp)
3478 		return find_next_bit(srcp, nr_bits, n + 1);
3479 
3480 	return n + 1;
3481 }
3482 
3483 /**
3484  *	netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3485  *	@n: CPU/Rx queue index
3486  *	@src1p: the first CPUs/Rx queues mask pointer
3487  *	@src2p: the second CPUs/Rx queues mask pointer
3488  *	@nr_bits: number of bits in the bitmask
3489  *
3490  * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3491  */
3492 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3493 					  const unsigned long *src2p,
3494 					  unsigned int nr_bits)
3495 {
3496 	/* -1 is a legal arg here. */
3497 	if (n != -1)
3498 		cpu_max_bits_warn(n, nr_bits);
3499 
3500 	if (src1p && src2p)
3501 		return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3502 	else if (src1p)
3503 		return find_next_bit(src1p, nr_bits, n + 1);
3504 	else if (src2p)
3505 		return find_next_bit(src2p, nr_bits, n + 1);
3506 
3507 	return n + 1;
3508 }
3509 #else
3510 static inline int netif_set_xps_queue(struct net_device *dev,
3511 				      const struct cpumask *mask,
3512 				      u16 index)
3513 {
3514 	return 0;
3515 }
3516 
3517 static inline int __netif_set_xps_queue(struct net_device *dev,
3518 					const unsigned long *mask,
3519 					u16 index, bool is_rxqs_map)
3520 {
3521 	return 0;
3522 }
3523 #endif
3524 
3525 /**
3526  *	netif_is_multiqueue - test if device has multiple transmit queues
3527  *	@dev: network device
3528  *
3529  * Check if device has multiple transmit queues
3530  */
3531 static inline bool netif_is_multiqueue(const struct net_device *dev)
3532 {
3533 	return dev->num_tx_queues > 1;
3534 }
3535 
3536 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3537 
3538 #ifdef CONFIG_SYSFS
3539 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3540 #else
3541 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3542 						unsigned int rxqs)
3543 {
3544 	dev->real_num_rx_queues = rxqs;
3545 	return 0;
3546 }
3547 #endif
3548 
3549 static inline struct netdev_rx_queue *
3550 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3551 {
3552 	return dev->_rx + rxq;
3553 }
3554 
3555 #ifdef CONFIG_SYSFS
3556 static inline unsigned int get_netdev_rx_queue_index(
3557 		struct netdev_rx_queue *queue)
3558 {
3559 	struct net_device *dev = queue->dev;
3560 	int index = queue - dev->_rx;
3561 
3562 	BUG_ON(index >= dev->num_rx_queues);
3563 	return index;
3564 }
3565 #endif
3566 
3567 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3568 int netif_get_num_default_rss_queues(void);
3569 
3570 enum skb_free_reason {
3571 	SKB_REASON_CONSUMED,
3572 	SKB_REASON_DROPPED,
3573 };
3574 
3575 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3576 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3577 
3578 /*
3579  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3580  * interrupt context or with hardware interrupts being disabled.
3581  * (in_irq() || irqs_disabled())
3582  *
3583  * We provide four helpers that can be used in following contexts :
3584  *
3585  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3586  *  replacing kfree_skb(skb)
3587  *
3588  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3589  *  Typically used in place of consume_skb(skb) in TX completion path
3590  *
3591  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3592  *  replacing kfree_skb(skb)
3593  *
3594  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3595  *  and consumed a packet. Used in place of consume_skb(skb)
3596  */
3597 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3598 {
3599 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3600 }
3601 
3602 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3603 {
3604 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3605 }
3606 
3607 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3608 {
3609 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3610 }
3611 
3612 static inline void dev_consume_skb_any(struct sk_buff *skb)
3613 {
3614 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3615 }
3616 
3617 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3618 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3619 int netif_rx(struct sk_buff *skb);
3620 int netif_rx_ni(struct sk_buff *skb);
3621 int netif_receive_skb(struct sk_buff *skb);
3622 int netif_receive_skb_core(struct sk_buff *skb);
3623 void netif_receive_skb_list(struct list_head *head);
3624 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3625 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3626 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3627 gro_result_t napi_gro_frags(struct napi_struct *napi);
3628 struct packet_offload *gro_find_receive_by_type(__be16 type);
3629 struct packet_offload *gro_find_complete_by_type(__be16 type);
3630 
3631 static inline void napi_free_frags(struct napi_struct *napi)
3632 {
3633 	kfree_skb(napi->skb);
3634 	napi->skb = NULL;
3635 }
3636 
3637 bool netdev_is_rx_handler_busy(struct net_device *dev);
3638 int netdev_rx_handler_register(struct net_device *dev,
3639 			       rx_handler_func_t *rx_handler,
3640 			       void *rx_handler_data);
3641 void netdev_rx_handler_unregister(struct net_device *dev);
3642 
3643 bool dev_valid_name(const char *name);
3644 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3645 		bool *need_copyout);
3646 int dev_ifconf(struct net *net, struct ifconf *, int);
3647 int dev_ethtool(struct net *net, struct ifreq *);
3648 unsigned int dev_get_flags(const struct net_device *);
3649 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3650 		       struct netlink_ext_ack *extack);
3651 int dev_change_flags(struct net_device *dev, unsigned int flags,
3652 		     struct netlink_ext_ack *extack);
3653 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3654 			unsigned int gchanges);
3655 int dev_change_name(struct net_device *, const char *);
3656 int dev_set_alias(struct net_device *, const char *, size_t);
3657 int dev_get_alias(const struct net_device *, char *, size_t);
3658 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3659 int __dev_set_mtu(struct net_device *, int);
3660 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3661 		    struct netlink_ext_ack *extack);
3662 int dev_set_mtu(struct net_device *, int);
3663 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3664 void dev_set_group(struct net_device *, int);
3665 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3666 			      struct netlink_ext_ack *extack);
3667 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3668 			struct netlink_ext_ack *extack);
3669 int dev_change_carrier(struct net_device *, bool new_carrier);
3670 int dev_get_phys_port_id(struct net_device *dev,
3671 			 struct netdev_phys_item_id *ppid);
3672 int dev_get_phys_port_name(struct net_device *dev,
3673 			   char *name, size_t len);
3674 int dev_get_port_parent_id(struct net_device *dev,
3675 			   struct netdev_phys_item_id *ppid, bool recurse);
3676 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3677 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3678 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3679 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3680 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3681 				    struct netdev_queue *txq, int *ret);
3682 
3683 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3684 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3685 		      int fd, u32 flags);
3686 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3687 		    enum bpf_netdev_command cmd);
3688 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3689 
3690 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3691 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3692 bool is_skb_forwardable(const struct net_device *dev,
3693 			const struct sk_buff *skb);
3694 
3695 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3696 					       struct sk_buff *skb)
3697 {
3698 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3699 	    unlikely(!is_skb_forwardable(dev, skb))) {
3700 		atomic_long_inc(&dev->rx_dropped);
3701 		kfree_skb(skb);
3702 		return NET_RX_DROP;
3703 	}
3704 
3705 	skb_scrub_packet(skb, true);
3706 	skb->priority = 0;
3707 	return 0;
3708 }
3709 
3710 bool dev_nit_active(struct net_device *dev);
3711 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3712 
3713 extern int		netdev_budget;
3714 extern unsigned int	netdev_budget_usecs;
3715 
3716 /* Called by rtnetlink.c:rtnl_unlock() */
3717 void netdev_run_todo(void);
3718 
3719 /**
3720  *	dev_put - release reference to device
3721  *	@dev: network device
3722  *
3723  * Release reference to device to allow it to be freed.
3724  */
3725 static inline void dev_put(struct net_device *dev)
3726 {
3727 	this_cpu_dec(*dev->pcpu_refcnt);
3728 }
3729 
3730 /**
3731  *	dev_hold - get reference to device
3732  *	@dev: network device
3733  *
3734  * Hold reference to device to keep it from being freed.
3735  */
3736 static inline void dev_hold(struct net_device *dev)
3737 {
3738 	this_cpu_inc(*dev->pcpu_refcnt);
3739 }
3740 
3741 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3742  * and _off may be called from IRQ context, but it is caller
3743  * who is responsible for serialization of these calls.
3744  *
3745  * The name carrier is inappropriate, these functions should really be
3746  * called netif_lowerlayer_*() because they represent the state of any
3747  * kind of lower layer not just hardware media.
3748  */
3749 
3750 void linkwatch_init_dev(struct net_device *dev);
3751 void linkwatch_fire_event(struct net_device *dev);
3752 void linkwatch_forget_dev(struct net_device *dev);
3753 
3754 /**
3755  *	netif_carrier_ok - test if carrier present
3756  *	@dev: network device
3757  *
3758  * Check if carrier is present on device
3759  */
3760 static inline bool netif_carrier_ok(const struct net_device *dev)
3761 {
3762 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3763 }
3764 
3765 unsigned long dev_trans_start(struct net_device *dev);
3766 
3767 void __netdev_watchdog_up(struct net_device *dev);
3768 
3769 void netif_carrier_on(struct net_device *dev);
3770 
3771 void netif_carrier_off(struct net_device *dev);
3772 
3773 /**
3774  *	netif_dormant_on - mark device as dormant.
3775  *	@dev: network device
3776  *
3777  * Mark device as dormant (as per RFC2863).
3778  *
3779  * The dormant state indicates that the relevant interface is not
3780  * actually in a condition to pass packets (i.e., it is not 'up') but is
3781  * in a "pending" state, waiting for some external event.  For "on-
3782  * demand" interfaces, this new state identifies the situation where the
3783  * interface is waiting for events to place it in the up state.
3784  */
3785 static inline void netif_dormant_on(struct net_device *dev)
3786 {
3787 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3788 		linkwatch_fire_event(dev);
3789 }
3790 
3791 /**
3792  *	netif_dormant_off - set device as not dormant.
3793  *	@dev: network device
3794  *
3795  * Device is not in dormant state.
3796  */
3797 static inline void netif_dormant_off(struct net_device *dev)
3798 {
3799 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3800 		linkwatch_fire_event(dev);
3801 }
3802 
3803 /**
3804  *	netif_dormant - test if device is dormant
3805  *	@dev: network device
3806  *
3807  * Check if device is dormant.
3808  */
3809 static inline bool netif_dormant(const struct net_device *dev)
3810 {
3811 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3812 }
3813 
3814 
3815 /**
3816  *	netif_oper_up - test if device is operational
3817  *	@dev: network device
3818  *
3819  * Check if carrier is operational
3820  */
3821 static inline bool netif_oper_up(const struct net_device *dev)
3822 {
3823 	return (dev->operstate == IF_OPER_UP ||
3824 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3825 }
3826 
3827 /**
3828  *	netif_device_present - is device available or removed
3829  *	@dev: network device
3830  *
3831  * Check if device has not been removed from system.
3832  */
3833 static inline bool netif_device_present(struct net_device *dev)
3834 {
3835 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3836 }
3837 
3838 void netif_device_detach(struct net_device *dev);
3839 
3840 void netif_device_attach(struct net_device *dev);
3841 
3842 /*
3843  * Network interface message level settings
3844  */
3845 
3846 enum {
3847 	NETIF_MSG_DRV		= 0x0001,
3848 	NETIF_MSG_PROBE		= 0x0002,
3849 	NETIF_MSG_LINK		= 0x0004,
3850 	NETIF_MSG_TIMER		= 0x0008,
3851 	NETIF_MSG_IFDOWN	= 0x0010,
3852 	NETIF_MSG_IFUP		= 0x0020,
3853 	NETIF_MSG_RX_ERR	= 0x0040,
3854 	NETIF_MSG_TX_ERR	= 0x0080,
3855 	NETIF_MSG_TX_QUEUED	= 0x0100,
3856 	NETIF_MSG_INTR		= 0x0200,
3857 	NETIF_MSG_TX_DONE	= 0x0400,
3858 	NETIF_MSG_RX_STATUS	= 0x0800,
3859 	NETIF_MSG_PKTDATA	= 0x1000,
3860 	NETIF_MSG_HW		= 0x2000,
3861 	NETIF_MSG_WOL		= 0x4000,
3862 };
3863 
3864 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3865 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3866 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3867 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3868 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3869 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3870 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3871 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3872 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3873 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3874 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3875 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3876 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3877 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3878 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3879 
3880 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3881 {
3882 	/* use default */
3883 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3884 		return default_msg_enable_bits;
3885 	if (debug_value == 0)	/* no output */
3886 		return 0;
3887 	/* set low N bits */
3888 	return (1U << debug_value) - 1;
3889 }
3890 
3891 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3892 {
3893 	spin_lock(&txq->_xmit_lock);
3894 	txq->xmit_lock_owner = cpu;
3895 }
3896 
3897 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3898 {
3899 	__acquire(&txq->_xmit_lock);
3900 	return true;
3901 }
3902 
3903 static inline void __netif_tx_release(struct netdev_queue *txq)
3904 {
3905 	__release(&txq->_xmit_lock);
3906 }
3907 
3908 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3909 {
3910 	spin_lock_bh(&txq->_xmit_lock);
3911 	txq->xmit_lock_owner = smp_processor_id();
3912 }
3913 
3914 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3915 {
3916 	bool ok = spin_trylock(&txq->_xmit_lock);
3917 	if (likely(ok))
3918 		txq->xmit_lock_owner = smp_processor_id();
3919 	return ok;
3920 }
3921 
3922 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3923 {
3924 	txq->xmit_lock_owner = -1;
3925 	spin_unlock(&txq->_xmit_lock);
3926 }
3927 
3928 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3929 {
3930 	txq->xmit_lock_owner = -1;
3931 	spin_unlock_bh(&txq->_xmit_lock);
3932 }
3933 
3934 static inline void txq_trans_update(struct netdev_queue *txq)
3935 {
3936 	if (txq->xmit_lock_owner != -1)
3937 		txq->trans_start = jiffies;
3938 }
3939 
3940 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3941 static inline void netif_trans_update(struct net_device *dev)
3942 {
3943 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3944 
3945 	if (txq->trans_start != jiffies)
3946 		txq->trans_start = jiffies;
3947 }
3948 
3949 /**
3950  *	netif_tx_lock - grab network device transmit lock
3951  *	@dev: network device
3952  *
3953  * Get network device transmit lock
3954  */
3955 static inline void netif_tx_lock(struct net_device *dev)
3956 {
3957 	unsigned int i;
3958 	int cpu;
3959 
3960 	spin_lock(&dev->tx_global_lock);
3961 	cpu = smp_processor_id();
3962 	for (i = 0; i < dev->num_tx_queues; i++) {
3963 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3964 
3965 		/* We are the only thread of execution doing a
3966 		 * freeze, but we have to grab the _xmit_lock in
3967 		 * order to synchronize with threads which are in
3968 		 * the ->hard_start_xmit() handler and already
3969 		 * checked the frozen bit.
3970 		 */
3971 		__netif_tx_lock(txq, cpu);
3972 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3973 		__netif_tx_unlock(txq);
3974 	}
3975 }
3976 
3977 static inline void netif_tx_lock_bh(struct net_device *dev)
3978 {
3979 	local_bh_disable();
3980 	netif_tx_lock(dev);
3981 }
3982 
3983 static inline void netif_tx_unlock(struct net_device *dev)
3984 {
3985 	unsigned int i;
3986 
3987 	for (i = 0; i < dev->num_tx_queues; i++) {
3988 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3989 
3990 		/* No need to grab the _xmit_lock here.  If the
3991 		 * queue is not stopped for another reason, we
3992 		 * force a schedule.
3993 		 */
3994 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3995 		netif_schedule_queue(txq);
3996 	}
3997 	spin_unlock(&dev->tx_global_lock);
3998 }
3999 
4000 static inline void netif_tx_unlock_bh(struct net_device *dev)
4001 {
4002 	netif_tx_unlock(dev);
4003 	local_bh_enable();
4004 }
4005 
4006 #define HARD_TX_LOCK(dev, txq, cpu) {			\
4007 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4008 		__netif_tx_lock(txq, cpu);		\
4009 	} else {					\
4010 		__netif_tx_acquire(txq);		\
4011 	}						\
4012 }
4013 
4014 #define HARD_TX_TRYLOCK(dev, txq)			\
4015 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
4016 		__netif_tx_trylock(txq) :		\
4017 		__netif_tx_acquire(txq))
4018 
4019 #define HARD_TX_UNLOCK(dev, txq) {			\
4020 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4021 		__netif_tx_unlock(txq);			\
4022 	} else {					\
4023 		__netif_tx_release(txq);		\
4024 	}						\
4025 }
4026 
4027 static inline void netif_tx_disable(struct net_device *dev)
4028 {
4029 	unsigned int i;
4030 	int cpu;
4031 
4032 	local_bh_disable();
4033 	cpu = smp_processor_id();
4034 	for (i = 0; i < dev->num_tx_queues; i++) {
4035 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4036 
4037 		__netif_tx_lock(txq, cpu);
4038 		netif_tx_stop_queue(txq);
4039 		__netif_tx_unlock(txq);
4040 	}
4041 	local_bh_enable();
4042 }
4043 
4044 static inline void netif_addr_lock(struct net_device *dev)
4045 {
4046 	spin_lock(&dev->addr_list_lock);
4047 }
4048 
4049 static inline void netif_addr_lock_nested(struct net_device *dev)
4050 {
4051 	int subclass = SINGLE_DEPTH_NESTING;
4052 
4053 	if (dev->netdev_ops->ndo_get_lock_subclass)
4054 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
4055 
4056 	spin_lock_nested(&dev->addr_list_lock, subclass);
4057 }
4058 
4059 static inline void netif_addr_lock_bh(struct net_device *dev)
4060 {
4061 	spin_lock_bh(&dev->addr_list_lock);
4062 }
4063 
4064 static inline void netif_addr_unlock(struct net_device *dev)
4065 {
4066 	spin_unlock(&dev->addr_list_lock);
4067 }
4068 
4069 static inline void netif_addr_unlock_bh(struct net_device *dev)
4070 {
4071 	spin_unlock_bh(&dev->addr_list_lock);
4072 }
4073 
4074 /*
4075  * dev_addrs walker. Should be used only for read access. Call with
4076  * rcu_read_lock held.
4077  */
4078 #define for_each_dev_addr(dev, ha) \
4079 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4080 
4081 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4082 
4083 void ether_setup(struct net_device *dev);
4084 
4085 /* Support for loadable net-drivers */
4086 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4087 				    unsigned char name_assign_type,
4088 				    void (*setup)(struct net_device *),
4089 				    unsigned int txqs, unsigned int rxqs);
4090 int dev_get_valid_name(struct net *net, struct net_device *dev,
4091 		       const char *name);
4092 
4093 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4094 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4095 
4096 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4097 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4098 			 count)
4099 
4100 int register_netdev(struct net_device *dev);
4101 void unregister_netdev(struct net_device *dev);
4102 
4103 /* General hardware address lists handling functions */
4104 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4105 		   struct netdev_hw_addr_list *from_list, int addr_len);
4106 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4107 		      struct netdev_hw_addr_list *from_list, int addr_len);
4108 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4109 		       struct net_device *dev,
4110 		       int (*sync)(struct net_device *, const unsigned char *),
4111 		       int (*unsync)(struct net_device *,
4112 				     const unsigned char *));
4113 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4114 			   struct net_device *dev,
4115 			   int (*sync)(struct net_device *,
4116 				       const unsigned char *, int),
4117 			   int (*unsync)(struct net_device *,
4118 					 const unsigned char *, int));
4119 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4120 			      struct net_device *dev,
4121 			      int (*unsync)(struct net_device *,
4122 					    const unsigned char *, int));
4123 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4124 			  struct net_device *dev,
4125 			  int (*unsync)(struct net_device *,
4126 					const unsigned char *));
4127 void __hw_addr_init(struct netdev_hw_addr_list *list);
4128 
4129 /* Functions used for device addresses handling */
4130 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4131 		 unsigned char addr_type);
4132 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4133 		 unsigned char addr_type);
4134 void dev_addr_flush(struct net_device *dev);
4135 int dev_addr_init(struct net_device *dev);
4136 
4137 /* Functions used for unicast addresses handling */
4138 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4139 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4140 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4141 int dev_uc_sync(struct net_device *to, struct net_device *from);
4142 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4143 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4144 void dev_uc_flush(struct net_device *dev);
4145 void dev_uc_init(struct net_device *dev);
4146 
4147 /**
4148  *  __dev_uc_sync - Synchonize device's unicast list
4149  *  @dev:  device to sync
4150  *  @sync: function to call if address should be added
4151  *  @unsync: function to call if address should be removed
4152  *
4153  *  Add newly added addresses to the interface, and release
4154  *  addresses that have been deleted.
4155  */
4156 static inline int __dev_uc_sync(struct net_device *dev,
4157 				int (*sync)(struct net_device *,
4158 					    const unsigned char *),
4159 				int (*unsync)(struct net_device *,
4160 					      const unsigned char *))
4161 {
4162 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4163 }
4164 
4165 /**
4166  *  __dev_uc_unsync - Remove synchronized addresses from device
4167  *  @dev:  device to sync
4168  *  @unsync: function to call if address should be removed
4169  *
4170  *  Remove all addresses that were added to the device by dev_uc_sync().
4171  */
4172 static inline void __dev_uc_unsync(struct net_device *dev,
4173 				   int (*unsync)(struct net_device *,
4174 						 const unsigned char *))
4175 {
4176 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
4177 }
4178 
4179 /* Functions used for multicast addresses handling */
4180 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4181 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4182 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4183 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4184 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4185 int dev_mc_sync(struct net_device *to, struct net_device *from);
4186 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4187 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4188 void dev_mc_flush(struct net_device *dev);
4189 void dev_mc_init(struct net_device *dev);
4190 
4191 /**
4192  *  __dev_mc_sync - Synchonize device's multicast list
4193  *  @dev:  device to sync
4194  *  @sync: function to call if address should be added
4195  *  @unsync: function to call if address should be removed
4196  *
4197  *  Add newly added addresses to the interface, and release
4198  *  addresses that have been deleted.
4199  */
4200 static inline int __dev_mc_sync(struct net_device *dev,
4201 				int (*sync)(struct net_device *,
4202 					    const unsigned char *),
4203 				int (*unsync)(struct net_device *,
4204 					      const unsigned char *))
4205 {
4206 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4207 }
4208 
4209 /**
4210  *  __dev_mc_unsync - Remove synchronized addresses from device
4211  *  @dev:  device to sync
4212  *  @unsync: function to call if address should be removed
4213  *
4214  *  Remove all addresses that were added to the device by dev_mc_sync().
4215  */
4216 static inline void __dev_mc_unsync(struct net_device *dev,
4217 				   int (*unsync)(struct net_device *,
4218 						 const unsigned char *))
4219 {
4220 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
4221 }
4222 
4223 /* Functions used for secondary unicast and multicast support */
4224 void dev_set_rx_mode(struct net_device *dev);
4225 void __dev_set_rx_mode(struct net_device *dev);
4226 int dev_set_promiscuity(struct net_device *dev, int inc);
4227 int dev_set_allmulti(struct net_device *dev, int inc);
4228 void netdev_state_change(struct net_device *dev);
4229 void netdev_notify_peers(struct net_device *dev);
4230 void netdev_features_change(struct net_device *dev);
4231 /* Load a device via the kmod */
4232 void dev_load(struct net *net, const char *name);
4233 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4234 					struct rtnl_link_stats64 *storage);
4235 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4236 			     const struct net_device_stats *netdev_stats);
4237 
4238 extern int		netdev_max_backlog;
4239 extern int		netdev_tstamp_prequeue;
4240 extern int		weight_p;
4241 extern int		dev_weight_rx_bias;
4242 extern int		dev_weight_tx_bias;
4243 extern int		dev_rx_weight;
4244 extern int		dev_tx_weight;
4245 
4246 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4247 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4248 						     struct list_head **iter);
4249 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4250 						     struct list_head **iter);
4251 
4252 /* iterate through upper list, must be called under RCU read lock */
4253 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4254 	for (iter = &(dev)->adj_list.upper, \
4255 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4256 	     updev; \
4257 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4258 
4259 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4260 				  int (*fn)(struct net_device *upper_dev,
4261 					    void *data),
4262 				  void *data);
4263 
4264 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4265 				  struct net_device *upper_dev);
4266 
4267 bool netdev_has_any_upper_dev(struct net_device *dev);
4268 
4269 void *netdev_lower_get_next_private(struct net_device *dev,
4270 				    struct list_head **iter);
4271 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4272 					struct list_head **iter);
4273 
4274 #define netdev_for_each_lower_private(dev, priv, iter) \
4275 	for (iter = (dev)->adj_list.lower.next, \
4276 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
4277 	     priv; \
4278 	     priv = netdev_lower_get_next_private(dev, &(iter)))
4279 
4280 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4281 	for (iter = &(dev)->adj_list.lower, \
4282 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4283 	     priv; \
4284 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4285 
4286 void *netdev_lower_get_next(struct net_device *dev,
4287 				struct list_head **iter);
4288 
4289 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4290 	for (iter = (dev)->adj_list.lower.next, \
4291 	     ldev = netdev_lower_get_next(dev, &(iter)); \
4292 	     ldev; \
4293 	     ldev = netdev_lower_get_next(dev, &(iter)))
4294 
4295 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4296 					     struct list_head **iter);
4297 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4298 						 struct list_head **iter);
4299 
4300 int netdev_walk_all_lower_dev(struct net_device *dev,
4301 			      int (*fn)(struct net_device *lower_dev,
4302 					void *data),
4303 			      void *data);
4304 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4305 				  int (*fn)(struct net_device *lower_dev,
4306 					    void *data),
4307 				  void *data);
4308 
4309 void *netdev_adjacent_get_private(struct list_head *adj_list);
4310 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4311 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4312 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4313 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4314 			  struct netlink_ext_ack *extack);
4315 int netdev_master_upper_dev_link(struct net_device *dev,
4316 				 struct net_device *upper_dev,
4317 				 void *upper_priv, void *upper_info,
4318 				 struct netlink_ext_ack *extack);
4319 void netdev_upper_dev_unlink(struct net_device *dev,
4320 			     struct net_device *upper_dev);
4321 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4322 void *netdev_lower_dev_get_private(struct net_device *dev,
4323 				   struct net_device *lower_dev);
4324 void netdev_lower_state_changed(struct net_device *lower_dev,
4325 				void *lower_state_info);
4326 
4327 /* RSS keys are 40 or 52 bytes long */
4328 #define NETDEV_RSS_KEY_LEN 52
4329 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4330 void netdev_rss_key_fill(void *buffer, size_t len);
4331 
4332 int dev_get_nest_level(struct net_device *dev);
4333 int skb_checksum_help(struct sk_buff *skb);
4334 int skb_crc32c_csum_help(struct sk_buff *skb);
4335 int skb_csum_hwoffload_help(struct sk_buff *skb,
4336 			    const netdev_features_t features);
4337 
4338 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4339 				  netdev_features_t features, bool tx_path);
4340 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4341 				    netdev_features_t features);
4342 
4343 struct netdev_bonding_info {
4344 	ifslave	slave;
4345 	ifbond	master;
4346 };
4347 
4348 struct netdev_notifier_bonding_info {
4349 	struct netdev_notifier_info info; /* must be first */
4350 	struct netdev_bonding_info  bonding_info;
4351 };
4352 
4353 void netdev_bonding_info_change(struct net_device *dev,
4354 				struct netdev_bonding_info *bonding_info);
4355 
4356 static inline
4357 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4358 {
4359 	return __skb_gso_segment(skb, features, true);
4360 }
4361 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4362 
4363 static inline bool can_checksum_protocol(netdev_features_t features,
4364 					 __be16 protocol)
4365 {
4366 	if (protocol == htons(ETH_P_FCOE))
4367 		return !!(features & NETIF_F_FCOE_CRC);
4368 
4369 	/* Assume this is an IP checksum (not SCTP CRC) */
4370 
4371 	if (features & NETIF_F_HW_CSUM) {
4372 		/* Can checksum everything */
4373 		return true;
4374 	}
4375 
4376 	switch (protocol) {
4377 	case htons(ETH_P_IP):
4378 		return !!(features & NETIF_F_IP_CSUM);
4379 	case htons(ETH_P_IPV6):
4380 		return !!(features & NETIF_F_IPV6_CSUM);
4381 	default:
4382 		return false;
4383 	}
4384 }
4385 
4386 #ifdef CONFIG_BUG
4387 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4388 #else
4389 static inline void netdev_rx_csum_fault(struct net_device *dev,
4390 					struct sk_buff *skb)
4391 {
4392 }
4393 #endif
4394 /* rx skb timestamps */
4395 void net_enable_timestamp(void);
4396 void net_disable_timestamp(void);
4397 
4398 #ifdef CONFIG_PROC_FS
4399 int __init dev_proc_init(void);
4400 #else
4401 #define dev_proc_init() 0
4402 #endif
4403 
4404 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4405 					      struct sk_buff *skb, struct net_device *dev,
4406 					      bool more)
4407 {
4408 	skb->xmit_more = more ? 1 : 0;
4409 	return ops->ndo_start_xmit(skb, dev);
4410 }
4411 
4412 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4413 					    struct netdev_queue *txq, bool more)
4414 {
4415 	const struct net_device_ops *ops = dev->netdev_ops;
4416 	netdev_tx_t rc;
4417 
4418 	rc = __netdev_start_xmit(ops, skb, dev, more);
4419 	if (rc == NETDEV_TX_OK)
4420 		txq_trans_update(txq);
4421 
4422 	return rc;
4423 }
4424 
4425 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4426 				const void *ns);
4427 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4428 				 const void *ns);
4429 
4430 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4431 {
4432 	return netdev_class_create_file_ns(class_attr, NULL);
4433 }
4434 
4435 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4436 {
4437 	netdev_class_remove_file_ns(class_attr, NULL);
4438 }
4439 
4440 extern const struct kobj_ns_type_operations net_ns_type_operations;
4441 
4442 const char *netdev_drivername(const struct net_device *dev);
4443 
4444 void linkwatch_run_queue(void);
4445 
4446 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4447 							  netdev_features_t f2)
4448 {
4449 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4450 		if (f1 & NETIF_F_HW_CSUM)
4451 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4452 		else
4453 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4454 	}
4455 
4456 	return f1 & f2;
4457 }
4458 
4459 static inline netdev_features_t netdev_get_wanted_features(
4460 	struct net_device *dev)
4461 {
4462 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4463 }
4464 netdev_features_t netdev_increment_features(netdev_features_t all,
4465 	netdev_features_t one, netdev_features_t mask);
4466 
4467 /* Allow TSO being used on stacked device :
4468  * Performing the GSO segmentation before last device
4469  * is a performance improvement.
4470  */
4471 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4472 							netdev_features_t mask)
4473 {
4474 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4475 }
4476 
4477 int __netdev_update_features(struct net_device *dev);
4478 void netdev_update_features(struct net_device *dev);
4479 void netdev_change_features(struct net_device *dev);
4480 
4481 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4482 					struct net_device *dev);
4483 
4484 netdev_features_t passthru_features_check(struct sk_buff *skb,
4485 					  struct net_device *dev,
4486 					  netdev_features_t features);
4487 netdev_features_t netif_skb_features(struct sk_buff *skb);
4488 
4489 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4490 {
4491 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4492 
4493 	/* check flags correspondence */
4494 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4495 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4496 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4497 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4498 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4499 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4500 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4501 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4502 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4503 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4504 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4505 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4506 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4507 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4508 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4509 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4510 	BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4511 	BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4512 
4513 	return (features & feature) == feature;
4514 }
4515 
4516 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4517 {
4518 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4519 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4520 }
4521 
4522 static inline bool netif_needs_gso(struct sk_buff *skb,
4523 				   netdev_features_t features)
4524 {
4525 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4526 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4527 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4528 }
4529 
4530 static inline void netif_set_gso_max_size(struct net_device *dev,
4531 					  unsigned int size)
4532 {
4533 	dev->gso_max_size = size;
4534 }
4535 
4536 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4537 					int pulled_hlen, u16 mac_offset,
4538 					int mac_len)
4539 {
4540 	skb->protocol = protocol;
4541 	skb->encapsulation = 1;
4542 	skb_push(skb, pulled_hlen);
4543 	skb_reset_transport_header(skb);
4544 	skb->mac_header = mac_offset;
4545 	skb->network_header = skb->mac_header + mac_len;
4546 	skb->mac_len = mac_len;
4547 }
4548 
4549 static inline bool netif_is_macsec(const struct net_device *dev)
4550 {
4551 	return dev->priv_flags & IFF_MACSEC;
4552 }
4553 
4554 static inline bool netif_is_macvlan(const struct net_device *dev)
4555 {
4556 	return dev->priv_flags & IFF_MACVLAN;
4557 }
4558 
4559 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4560 {
4561 	return dev->priv_flags & IFF_MACVLAN_PORT;
4562 }
4563 
4564 static inline bool netif_is_bond_master(const struct net_device *dev)
4565 {
4566 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4567 }
4568 
4569 static inline bool netif_is_bond_slave(const struct net_device *dev)
4570 {
4571 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4572 }
4573 
4574 static inline bool netif_supports_nofcs(struct net_device *dev)
4575 {
4576 	return dev->priv_flags & IFF_SUPP_NOFCS;
4577 }
4578 
4579 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4580 {
4581 	return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4582 }
4583 
4584 static inline bool netif_is_l3_master(const struct net_device *dev)
4585 {
4586 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4587 }
4588 
4589 static inline bool netif_is_l3_slave(const struct net_device *dev)
4590 {
4591 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4592 }
4593 
4594 static inline bool netif_is_bridge_master(const struct net_device *dev)
4595 {
4596 	return dev->priv_flags & IFF_EBRIDGE;
4597 }
4598 
4599 static inline bool netif_is_bridge_port(const struct net_device *dev)
4600 {
4601 	return dev->priv_flags & IFF_BRIDGE_PORT;
4602 }
4603 
4604 static inline bool netif_is_ovs_master(const struct net_device *dev)
4605 {
4606 	return dev->priv_flags & IFF_OPENVSWITCH;
4607 }
4608 
4609 static inline bool netif_is_ovs_port(const struct net_device *dev)
4610 {
4611 	return dev->priv_flags & IFF_OVS_DATAPATH;
4612 }
4613 
4614 static inline bool netif_is_team_master(const struct net_device *dev)
4615 {
4616 	return dev->priv_flags & IFF_TEAM;
4617 }
4618 
4619 static inline bool netif_is_team_port(const struct net_device *dev)
4620 {
4621 	return dev->priv_flags & IFF_TEAM_PORT;
4622 }
4623 
4624 static inline bool netif_is_lag_master(const struct net_device *dev)
4625 {
4626 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4627 }
4628 
4629 static inline bool netif_is_lag_port(const struct net_device *dev)
4630 {
4631 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4632 }
4633 
4634 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4635 {
4636 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4637 }
4638 
4639 static inline bool netif_is_failover(const struct net_device *dev)
4640 {
4641 	return dev->priv_flags & IFF_FAILOVER;
4642 }
4643 
4644 static inline bool netif_is_failover_slave(const struct net_device *dev)
4645 {
4646 	return dev->priv_flags & IFF_FAILOVER_SLAVE;
4647 }
4648 
4649 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4650 static inline void netif_keep_dst(struct net_device *dev)
4651 {
4652 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4653 }
4654 
4655 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4656 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4657 {
4658 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4659 	return dev->priv_flags & IFF_MACSEC;
4660 }
4661 
4662 extern struct pernet_operations __net_initdata loopback_net_ops;
4663 
4664 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4665 
4666 /* netdev_printk helpers, similar to dev_printk */
4667 
4668 static inline const char *netdev_name(const struct net_device *dev)
4669 {
4670 	if (!dev->name[0] || strchr(dev->name, '%'))
4671 		return "(unnamed net_device)";
4672 	return dev->name;
4673 }
4674 
4675 static inline bool netdev_unregistering(const struct net_device *dev)
4676 {
4677 	return dev->reg_state == NETREG_UNREGISTERING;
4678 }
4679 
4680 static inline const char *netdev_reg_state(const struct net_device *dev)
4681 {
4682 	switch (dev->reg_state) {
4683 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4684 	case NETREG_REGISTERED: return "";
4685 	case NETREG_UNREGISTERING: return " (unregistering)";
4686 	case NETREG_UNREGISTERED: return " (unregistered)";
4687 	case NETREG_RELEASED: return " (released)";
4688 	case NETREG_DUMMY: return " (dummy)";
4689 	}
4690 
4691 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4692 	return " (unknown)";
4693 }
4694 
4695 __printf(3, 4) __cold
4696 void netdev_printk(const char *level, const struct net_device *dev,
4697 		   const char *format, ...);
4698 __printf(2, 3) __cold
4699 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4700 __printf(2, 3) __cold
4701 void netdev_alert(const struct net_device *dev, const char *format, ...);
4702 __printf(2, 3) __cold
4703 void netdev_crit(const struct net_device *dev, const char *format, ...);
4704 __printf(2, 3) __cold
4705 void netdev_err(const struct net_device *dev, const char *format, ...);
4706 __printf(2, 3) __cold
4707 void netdev_warn(const struct net_device *dev, const char *format, ...);
4708 __printf(2, 3) __cold
4709 void netdev_notice(const struct net_device *dev, const char *format, ...);
4710 __printf(2, 3) __cold
4711 void netdev_info(const struct net_device *dev, const char *format, ...);
4712 
4713 #define netdev_level_once(level, dev, fmt, ...)			\
4714 do {								\
4715 	static bool __print_once __read_mostly;			\
4716 								\
4717 	if (!__print_once) {					\
4718 		__print_once = true;				\
4719 		netdev_printk(level, dev, fmt, ##__VA_ARGS__);	\
4720 	}							\
4721 } while (0)
4722 
4723 #define netdev_emerg_once(dev, fmt, ...) \
4724 	netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4725 #define netdev_alert_once(dev, fmt, ...) \
4726 	netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4727 #define netdev_crit_once(dev, fmt, ...) \
4728 	netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4729 #define netdev_err_once(dev, fmt, ...) \
4730 	netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4731 #define netdev_warn_once(dev, fmt, ...) \
4732 	netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4733 #define netdev_notice_once(dev, fmt, ...) \
4734 	netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4735 #define netdev_info_once(dev, fmt, ...) \
4736 	netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4737 
4738 #define MODULE_ALIAS_NETDEV(device) \
4739 	MODULE_ALIAS("netdev-" device)
4740 
4741 #if defined(CONFIG_DYNAMIC_DEBUG)
4742 #define netdev_dbg(__dev, format, args...)			\
4743 do {								\
4744 	dynamic_netdev_dbg(__dev, format, ##args);		\
4745 } while (0)
4746 #elif defined(DEBUG)
4747 #define netdev_dbg(__dev, format, args...)			\
4748 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4749 #else
4750 #define netdev_dbg(__dev, format, args...)			\
4751 ({								\
4752 	if (0)							\
4753 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4754 })
4755 #endif
4756 
4757 #if defined(VERBOSE_DEBUG)
4758 #define netdev_vdbg	netdev_dbg
4759 #else
4760 
4761 #define netdev_vdbg(dev, format, args...)			\
4762 ({								\
4763 	if (0)							\
4764 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4765 	0;							\
4766 })
4767 #endif
4768 
4769 /*
4770  * netdev_WARN() acts like dev_printk(), but with the key difference
4771  * of using a WARN/WARN_ON to get the message out, including the
4772  * file/line information and a backtrace.
4773  */
4774 #define netdev_WARN(dev, format, args...)			\
4775 	WARN(1, "netdevice: %s%s: " format, netdev_name(dev),	\
4776 	     netdev_reg_state(dev), ##args)
4777 
4778 #define netdev_WARN_ONCE(dev, format, args...)				\
4779 	WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),	\
4780 		  netdev_reg_state(dev), ##args)
4781 
4782 /* netif printk helpers, similar to netdev_printk */
4783 
4784 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4785 do {					  			\
4786 	if (netif_msg_##type(priv))				\
4787 		netdev_printk(level, (dev), fmt, ##args);	\
4788 } while (0)
4789 
4790 #define netif_level(level, priv, type, dev, fmt, args...)	\
4791 do {								\
4792 	if (netif_msg_##type(priv))				\
4793 		netdev_##level(dev, fmt, ##args);		\
4794 } while (0)
4795 
4796 #define netif_emerg(priv, type, dev, fmt, args...)		\
4797 	netif_level(emerg, priv, type, dev, fmt, ##args)
4798 #define netif_alert(priv, type, dev, fmt, args...)		\
4799 	netif_level(alert, priv, type, dev, fmt, ##args)
4800 #define netif_crit(priv, type, dev, fmt, args...)		\
4801 	netif_level(crit, priv, type, dev, fmt, ##args)
4802 #define netif_err(priv, type, dev, fmt, args...)		\
4803 	netif_level(err, priv, type, dev, fmt, ##args)
4804 #define netif_warn(priv, type, dev, fmt, args...)		\
4805 	netif_level(warn, priv, type, dev, fmt, ##args)
4806 #define netif_notice(priv, type, dev, fmt, args...)		\
4807 	netif_level(notice, priv, type, dev, fmt, ##args)
4808 #define netif_info(priv, type, dev, fmt, args...)		\
4809 	netif_level(info, priv, type, dev, fmt, ##args)
4810 
4811 #if defined(CONFIG_DYNAMIC_DEBUG)
4812 #define netif_dbg(priv, type, netdev, format, args...)		\
4813 do {								\
4814 	if (netif_msg_##type(priv))				\
4815 		dynamic_netdev_dbg(netdev, format, ##args);	\
4816 } while (0)
4817 #elif defined(DEBUG)
4818 #define netif_dbg(priv, type, dev, format, args...)		\
4819 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4820 #else
4821 #define netif_dbg(priv, type, dev, format, args...)			\
4822 ({									\
4823 	if (0)								\
4824 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4825 	0;								\
4826 })
4827 #endif
4828 
4829 /* if @cond then downgrade to debug, else print at @level */
4830 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4831 	do {                                                              \
4832 		if (cond)                                                 \
4833 			netif_dbg(priv, type, netdev, fmt, ##args);       \
4834 		else                                                      \
4835 			netif_ ## level(priv, type, netdev, fmt, ##args); \
4836 	} while (0)
4837 
4838 #if defined(VERBOSE_DEBUG)
4839 #define netif_vdbg	netif_dbg
4840 #else
4841 #define netif_vdbg(priv, type, dev, format, args...)		\
4842 ({								\
4843 	if (0)							\
4844 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4845 	0;							\
4846 })
4847 #endif
4848 
4849 /*
4850  *	The list of packet types we will receive (as opposed to discard)
4851  *	and the routines to invoke.
4852  *
4853  *	Why 16. Because with 16 the only overlap we get on a hash of the
4854  *	low nibble of the protocol value is RARP/SNAP/X.25.
4855  *
4856  *		0800	IP
4857  *		0001	802.3
4858  *		0002	AX.25
4859  *		0004	802.2
4860  *		8035	RARP
4861  *		0005	SNAP
4862  *		0805	X.25
4863  *		0806	ARP
4864  *		8137	IPX
4865  *		0009	Localtalk
4866  *		86DD	IPv6
4867  */
4868 #define PTYPE_HASH_SIZE	(16)
4869 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4870 
4871 #endif	/* _LINUX_NETDEVICE_H */
4872