xref: /linux-6.15/include/linux/netdevice.h (revision ce628966)
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 /* 802.11 specific */
62 struct wireless_dev;
63 /* 802.15.4 specific */
64 struct wpan_dev;
65 struct mpls_dev;
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info;
68 struct bpf_prog;
69 struct xdp_buff;
70 
71 void netdev_set_default_ethtool_ops(struct net_device *dev,
72 				    const struct ethtool_ops *ops);
73 
74 /* Backlog congestion levels */
75 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
76 #define NET_RX_DROP		1	/* packet dropped */
77 
78 /*
79  * Transmit return codes: transmit return codes originate from three different
80  * namespaces:
81  *
82  * - qdisc return codes
83  * - driver transmit return codes
84  * - errno values
85  *
86  * Drivers are allowed to return any one of those in their hard_start_xmit()
87  * function. Real network devices commonly used with qdiscs should only return
88  * the driver transmit return codes though - when qdiscs are used, the actual
89  * transmission happens asynchronously, so the value is not propagated to
90  * higher layers. Virtual network devices transmit synchronously; in this case
91  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
92  * others are propagated to higher layers.
93  */
94 
95 /* qdisc ->enqueue() return codes. */
96 #define NET_XMIT_SUCCESS	0x00
97 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
98 #define NET_XMIT_CN		0x02	/* congestion notification	*/
99 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
100 
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102  * indicates that the device will soon be dropping packets, or already drops
103  * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106 
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK		0xf0
109 
110 enum netdev_tx {
111 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
112 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
113 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
114 };
115 typedef enum netdev_tx netdev_tx_t;
116 
117 /*
118  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
119  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120  */
121 static inline bool dev_xmit_complete(int rc)
122 {
123 	/*
124 	 * Positive cases with an skb consumed by a driver:
125 	 * - successful transmission (rc == NETDEV_TX_OK)
126 	 * - error while transmitting (rc < 0)
127 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 	 */
129 	if (likely(rc < NET_XMIT_MASK))
130 		return true;
131 
132 	return false;
133 }
134 
135 /*
136  *	Compute the worst-case header length according to the protocols
137  *	used.
138  */
139 
140 #if defined(CONFIG_HYPERV_NET)
141 # define LL_MAX_HEADER 128
142 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
143 # if defined(CONFIG_MAC80211_MESH)
144 #  define LL_MAX_HEADER 128
145 # else
146 #  define LL_MAX_HEADER 96
147 # endif
148 #else
149 # define LL_MAX_HEADER 32
150 #endif
151 
152 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
153     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
154 #define MAX_HEADER LL_MAX_HEADER
155 #else
156 #define MAX_HEADER (LL_MAX_HEADER + 48)
157 #endif
158 
159 /*
160  *	Old network device statistics. Fields are native words
161  *	(unsigned long) so they can be read and written atomically.
162  */
163 
164 struct net_device_stats {
165 	unsigned long	rx_packets;
166 	unsigned long	tx_packets;
167 	unsigned long	rx_bytes;
168 	unsigned long	tx_bytes;
169 	unsigned long	rx_errors;
170 	unsigned long	tx_errors;
171 	unsigned long	rx_dropped;
172 	unsigned long	tx_dropped;
173 	unsigned long	multicast;
174 	unsigned long	collisions;
175 	unsigned long	rx_length_errors;
176 	unsigned long	rx_over_errors;
177 	unsigned long	rx_crc_errors;
178 	unsigned long	rx_frame_errors;
179 	unsigned long	rx_fifo_errors;
180 	unsigned long	rx_missed_errors;
181 	unsigned long	tx_aborted_errors;
182 	unsigned long	tx_carrier_errors;
183 	unsigned long	tx_fifo_errors;
184 	unsigned long	tx_heartbeat_errors;
185 	unsigned long	tx_window_errors;
186 	unsigned long	rx_compressed;
187 	unsigned long	tx_compressed;
188 };
189 
190 
191 #include <linux/cache.h>
192 #include <linux/skbuff.h>
193 
194 #ifdef CONFIG_RPS
195 #include <linux/static_key.h>
196 extern struct static_key rps_needed;
197 extern struct static_key rfs_needed;
198 #endif
199 
200 struct neighbour;
201 struct neigh_parms;
202 struct sk_buff;
203 
204 struct netdev_hw_addr {
205 	struct list_head	list;
206 	unsigned char		addr[MAX_ADDR_LEN];
207 	unsigned char		type;
208 #define NETDEV_HW_ADDR_T_LAN		1
209 #define NETDEV_HW_ADDR_T_SAN		2
210 #define NETDEV_HW_ADDR_T_SLAVE		3
211 #define NETDEV_HW_ADDR_T_UNICAST	4
212 #define NETDEV_HW_ADDR_T_MULTICAST	5
213 	bool			global_use;
214 	int			sync_cnt;
215 	int			refcount;
216 	int			synced;
217 	struct rcu_head		rcu_head;
218 };
219 
220 struct netdev_hw_addr_list {
221 	struct list_head	list;
222 	int			count;
223 };
224 
225 #define netdev_hw_addr_list_count(l) ((l)->count)
226 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
227 #define netdev_hw_addr_list_for_each(ha, l) \
228 	list_for_each_entry(ha, &(l)->list, list)
229 
230 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
231 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
232 #define netdev_for_each_uc_addr(ha, dev) \
233 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
234 
235 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
236 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
237 #define netdev_for_each_mc_addr(ha, dev) \
238 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
239 
240 struct hh_cache {
241 	unsigned int	hh_len;
242 	seqlock_t	hh_lock;
243 
244 	/* cached hardware header; allow for machine alignment needs.        */
245 #define HH_DATA_MOD	16
246 #define HH_DATA_OFF(__len) \
247 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
248 #define HH_DATA_ALIGN(__len) \
249 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
250 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251 };
252 
253 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
254  * Alternative is:
255  *   dev->hard_header_len ? (dev->hard_header_len +
256  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257  *
258  * We could use other alignment values, but we must maintain the
259  * relationship HH alignment <= LL alignment.
260  */
261 #define LL_RESERVED_SPACE(dev) \
262 	((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
263 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
264 	((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
265 
266 struct header_ops {
267 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
268 			   unsigned short type, const void *daddr,
269 			   const void *saddr, unsigned int len);
270 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
271 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
272 	void	(*cache_update)(struct hh_cache *hh,
273 				const struct net_device *dev,
274 				const unsigned char *haddr);
275 	bool	(*validate)(const char *ll_header, unsigned int len);
276 };
277 
278 /* These flag bits are private to the generic network queueing
279  * layer; they may not be explicitly referenced by any other
280  * code.
281  */
282 
283 enum netdev_state_t {
284 	__LINK_STATE_START,
285 	__LINK_STATE_PRESENT,
286 	__LINK_STATE_NOCARRIER,
287 	__LINK_STATE_LINKWATCH_PENDING,
288 	__LINK_STATE_DORMANT,
289 };
290 
291 
292 /*
293  * This structure holds boot-time configured netdevice settings. They
294  * are then used in the device probing.
295  */
296 struct netdev_boot_setup {
297 	char name[IFNAMSIZ];
298 	struct ifmap map;
299 };
300 #define NETDEV_BOOT_SETUP_MAX 8
301 
302 int __init netdev_boot_setup(char *str);
303 
304 /*
305  * Structure for NAPI scheduling similar to tasklet but with weighting
306  */
307 struct napi_struct {
308 	/* The poll_list must only be managed by the entity which
309 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
310 	 * whoever atomically sets that bit can add this napi_struct
311 	 * to the per-CPU poll_list, and whoever clears that bit
312 	 * can remove from the list right before clearing the bit.
313 	 */
314 	struct list_head	poll_list;
315 
316 	unsigned long		state;
317 	int			weight;
318 	unsigned int		gro_count;
319 	int			(*poll)(struct napi_struct *, int);
320 #ifdef CONFIG_NETPOLL
321 	int			poll_owner;
322 #endif
323 	struct net_device	*dev;
324 	struct sk_buff		*gro_list;
325 	struct sk_buff		*skb;
326 	struct hrtimer		timer;
327 	struct list_head	dev_list;
328 	struct hlist_node	napi_hash_node;
329 	unsigned int		napi_id;
330 };
331 
332 enum {
333 	NAPI_STATE_SCHED,	/* Poll is scheduled */
334 	NAPI_STATE_MISSED,	/* reschedule a napi */
335 	NAPI_STATE_DISABLE,	/* Disable pending */
336 	NAPI_STATE_NPSVC,	/* Netpoll - don't dequeue from poll_list */
337 	NAPI_STATE_HASHED,	/* In NAPI hash (busy polling possible) */
338 	NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
339 	NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
340 };
341 
342 enum {
343 	NAPIF_STATE_SCHED	 = BIT(NAPI_STATE_SCHED),
344 	NAPIF_STATE_MISSED	 = BIT(NAPI_STATE_MISSED),
345 	NAPIF_STATE_DISABLE	 = BIT(NAPI_STATE_DISABLE),
346 	NAPIF_STATE_NPSVC	 = BIT(NAPI_STATE_NPSVC),
347 	NAPIF_STATE_HASHED	 = BIT(NAPI_STATE_HASHED),
348 	NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
349 	NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
350 };
351 
352 enum gro_result {
353 	GRO_MERGED,
354 	GRO_MERGED_FREE,
355 	GRO_HELD,
356 	GRO_NORMAL,
357 	GRO_DROP,
358 	GRO_CONSUMED,
359 };
360 typedef enum gro_result gro_result_t;
361 
362 /*
363  * enum rx_handler_result - Possible return values for rx_handlers.
364  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365  * further.
366  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
367  * case skb->dev was changed by rx_handler.
368  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
369  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
370  *
371  * rx_handlers are functions called from inside __netif_receive_skb(), to do
372  * special processing of the skb, prior to delivery to protocol handlers.
373  *
374  * Currently, a net_device can only have a single rx_handler registered. Trying
375  * to register a second rx_handler will return -EBUSY.
376  *
377  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
378  * To unregister a rx_handler on a net_device, use
379  * netdev_rx_handler_unregister().
380  *
381  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
382  * do with the skb.
383  *
384  * If the rx_handler consumed the skb in some way, it should return
385  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
386  * the skb to be delivered in some other way.
387  *
388  * If the rx_handler changed skb->dev, to divert the skb to another
389  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
390  * new device will be called if it exists.
391  *
392  * If the rx_handler decides the skb should be ignored, it should return
393  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
394  * are registered on exact device (ptype->dev == skb->dev).
395  *
396  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
397  * delivered, it should return RX_HANDLER_PASS.
398  *
399  * A device without a registered rx_handler will behave as if rx_handler
400  * returned RX_HANDLER_PASS.
401  */
402 
403 enum rx_handler_result {
404 	RX_HANDLER_CONSUMED,
405 	RX_HANDLER_ANOTHER,
406 	RX_HANDLER_EXACT,
407 	RX_HANDLER_PASS,
408 };
409 typedef enum rx_handler_result rx_handler_result_t;
410 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
411 
412 void __napi_schedule(struct napi_struct *n);
413 void __napi_schedule_irqoff(struct napi_struct *n);
414 
415 static inline bool napi_disable_pending(struct napi_struct *n)
416 {
417 	return test_bit(NAPI_STATE_DISABLE, &n->state);
418 }
419 
420 bool napi_schedule_prep(struct napi_struct *n);
421 
422 /**
423  *	napi_schedule - schedule NAPI poll
424  *	@n: NAPI context
425  *
426  * Schedule NAPI poll routine to be called if it is not already
427  * running.
428  */
429 static inline void napi_schedule(struct napi_struct *n)
430 {
431 	if (napi_schedule_prep(n))
432 		__napi_schedule(n);
433 }
434 
435 /**
436  *	napi_schedule_irqoff - schedule NAPI poll
437  *	@n: NAPI context
438  *
439  * Variant of napi_schedule(), assuming hard irqs are masked.
440  */
441 static inline void napi_schedule_irqoff(struct napi_struct *n)
442 {
443 	if (napi_schedule_prep(n))
444 		__napi_schedule_irqoff(n);
445 }
446 
447 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
448 static inline bool napi_reschedule(struct napi_struct *napi)
449 {
450 	if (napi_schedule_prep(napi)) {
451 		__napi_schedule(napi);
452 		return true;
453 	}
454 	return false;
455 }
456 
457 bool napi_complete_done(struct napi_struct *n, int work_done);
458 /**
459  *	napi_complete - NAPI processing complete
460  *	@n: NAPI context
461  *
462  * Mark NAPI processing as complete.
463  * Consider using napi_complete_done() instead.
464  * Return false if device should avoid rearming interrupts.
465  */
466 static inline bool napi_complete(struct napi_struct *n)
467 {
468 	return napi_complete_done(n, 0);
469 }
470 
471 /**
472  *	napi_hash_del - remove a NAPI from global table
473  *	@napi: NAPI context
474  *
475  * Warning: caller must observe RCU grace period
476  * before freeing memory containing @napi, if
477  * this function returns true.
478  * Note: core networking stack automatically calls it
479  * from netif_napi_del().
480  * Drivers might want to call this helper to combine all
481  * the needed RCU grace periods into a single one.
482  */
483 bool napi_hash_del(struct napi_struct *napi);
484 
485 /**
486  *	napi_disable - prevent NAPI from scheduling
487  *	@n: NAPI context
488  *
489  * Stop NAPI from being scheduled on this context.
490  * Waits till any outstanding processing completes.
491  */
492 void napi_disable(struct napi_struct *n);
493 
494 /**
495  *	napi_enable - enable NAPI scheduling
496  *	@n: NAPI context
497  *
498  * Resume NAPI from being scheduled on this context.
499  * Must be paired with napi_disable.
500  */
501 static inline void napi_enable(struct napi_struct *n)
502 {
503 	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
504 	smp_mb__before_atomic();
505 	clear_bit(NAPI_STATE_SCHED, &n->state);
506 	clear_bit(NAPI_STATE_NPSVC, &n->state);
507 }
508 
509 /**
510  *	napi_synchronize - wait until NAPI is not running
511  *	@n: NAPI context
512  *
513  * Wait until NAPI is done being scheduled on this context.
514  * Waits till any outstanding processing completes but
515  * does not disable future activations.
516  */
517 static inline void napi_synchronize(const struct napi_struct *n)
518 {
519 	if (IS_ENABLED(CONFIG_SMP))
520 		while (test_bit(NAPI_STATE_SCHED, &n->state))
521 			msleep(1);
522 	else
523 		barrier();
524 }
525 
526 enum netdev_queue_state_t {
527 	__QUEUE_STATE_DRV_XOFF,
528 	__QUEUE_STATE_STACK_XOFF,
529 	__QUEUE_STATE_FROZEN,
530 };
531 
532 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
533 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
534 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
535 
536 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
537 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 					QUEUE_STATE_FROZEN)
539 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
540 					QUEUE_STATE_FROZEN)
541 
542 /*
543  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
544  * netif_tx_* functions below are used to manipulate this flag.  The
545  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
546  * queue independently.  The netif_xmit_*stopped functions below are called
547  * to check if the queue has been stopped by the driver or stack (either
548  * of the XOFF bits are set in the state).  Drivers should not need to call
549  * netif_xmit*stopped functions, they should only be using netif_tx_*.
550  */
551 
552 struct netdev_queue {
553 /*
554  * read-mostly part
555  */
556 	struct net_device	*dev;
557 	struct Qdisc __rcu	*qdisc;
558 	struct Qdisc		*qdisc_sleeping;
559 #ifdef CONFIG_SYSFS
560 	struct kobject		kobj;
561 #endif
562 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
563 	int			numa_node;
564 #endif
565 	unsigned long		tx_maxrate;
566 	/*
567 	 * Number of TX timeouts for this queue
568 	 * (/sys/class/net/DEV/Q/trans_timeout)
569 	 */
570 	unsigned long		trans_timeout;
571 /*
572  * write-mostly part
573  */
574 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
575 	int			xmit_lock_owner;
576 	/*
577 	 * Time (in jiffies) of last Tx
578 	 */
579 	unsigned long		trans_start;
580 
581 	unsigned long		state;
582 
583 #ifdef CONFIG_BQL
584 	struct dql		dql;
585 #endif
586 } ____cacheline_aligned_in_smp;
587 
588 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
589 {
590 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
591 	return q->numa_node;
592 #else
593 	return NUMA_NO_NODE;
594 #endif
595 }
596 
597 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
598 {
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 	q->numa_node = node;
601 #endif
602 }
603 
604 #ifdef CONFIG_RPS
605 /*
606  * This structure holds an RPS map which can be of variable length.  The
607  * map is an array of CPUs.
608  */
609 struct rps_map {
610 	unsigned int len;
611 	struct rcu_head rcu;
612 	u16 cpus[0];
613 };
614 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
615 
616 /*
617  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
618  * tail pointer for that CPU's input queue at the time of last enqueue, and
619  * a hardware filter index.
620  */
621 struct rps_dev_flow {
622 	u16 cpu;
623 	u16 filter;
624 	unsigned int last_qtail;
625 };
626 #define RPS_NO_FILTER 0xffff
627 
628 /*
629  * The rps_dev_flow_table structure contains a table of flow mappings.
630  */
631 struct rps_dev_flow_table {
632 	unsigned int mask;
633 	struct rcu_head rcu;
634 	struct rps_dev_flow flows[0];
635 };
636 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
637     ((_num) * sizeof(struct rps_dev_flow)))
638 
639 /*
640  * The rps_sock_flow_table contains mappings of flows to the last CPU
641  * on which they were processed by the application (set in recvmsg).
642  * Each entry is a 32bit value. Upper part is the high-order bits
643  * of flow hash, lower part is CPU number.
644  * rps_cpu_mask is used to partition the space, depending on number of
645  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
646  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
647  * meaning we use 32-6=26 bits for the hash.
648  */
649 struct rps_sock_flow_table {
650 	u32	mask;
651 
652 	u32	ents[0] ____cacheline_aligned_in_smp;
653 };
654 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
655 
656 #define RPS_NO_CPU 0xffff
657 
658 extern u32 rps_cpu_mask;
659 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
660 
661 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
662 					u32 hash)
663 {
664 	if (table && hash) {
665 		unsigned int index = hash & table->mask;
666 		u32 val = hash & ~rps_cpu_mask;
667 
668 		/* We only give a hint, preemption can change CPU under us */
669 		val |= raw_smp_processor_id();
670 
671 		if (table->ents[index] != val)
672 			table->ents[index] = val;
673 	}
674 }
675 
676 #ifdef CONFIG_RFS_ACCEL
677 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
678 			 u16 filter_id);
679 #endif
680 #endif /* CONFIG_RPS */
681 
682 /* This structure contains an instance of an RX queue. */
683 struct netdev_rx_queue {
684 #ifdef CONFIG_RPS
685 	struct rps_map __rcu		*rps_map;
686 	struct rps_dev_flow_table __rcu	*rps_flow_table;
687 #endif
688 	struct kobject			kobj;
689 	struct net_device		*dev;
690 	struct xdp_rxq_info		xdp_rxq;
691 } ____cacheline_aligned_in_smp;
692 
693 /*
694  * RX queue sysfs structures and functions.
695  */
696 struct rx_queue_attribute {
697 	struct attribute attr;
698 	ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
699 	ssize_t (*store)(struct netdev_rx_queue *queue,
700 			 const char *buf, size_t len);
701 };
702 
703 #ifdef CONFIG_XPS
704 /*
705  * This structure holds an XPS map which can be of variable length.  The
706  * map is an array of queues.
707  */
708 struct xps_map {
709 	unsigned int len;
710 	unsigned int alloc_len;
711 	struct rcu_head rcu;
712 	u16 queues[0];
713 };
714 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
715 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
716        - sizeof(struct xps_map)) / sizeof(u16))
717 
718 /*
719  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
720  */
721 struct xps_dev_maps {
722 	struct rcu_head rcu;
723 	struct xps_map __rcu *cpu_map[0];
724 };
725 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +		\
726 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
727 #endif /* CONFIG_XPS */
728 
729 #define TC_MAX_QUEUE	16
730 #define TC_BITMASK	15
731 /* HW offloaded queuing disciplines txq count and offset maps */
732 struct netdev_tc_txq {
733 	u16 count;
734 	u16 offset;
735 };
736 
737 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
738 /*
739  * This structure is to hold information about the device
740  * configured to run FCoE protocol stack.
741  */
742 struct netdev_fcoe_hbainfo {
743 	char	manufacturer[64];
744 	char	serial_number[64];
745 	char	hardware_version[64];
746 	char	driver_version[64];
747 	char	optionrom_version[64];
748 	char	firmware_version[64];
749 	char	model[256];
750 	char	model_description[256];
751 };
752 #endif
753 
754 #define MAX_PHYS_ITEM_ID_LEN 32
755 
756 /* This structure holds a unique identifier to identify some
757  * physical item (port for example) used by a netdevice.
758  */
759 struct netdev_phys_item_id {
760 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
761 	unsigned char id_len;
762 };
763 
764 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
765 					    struct netdev_phys_item_id *b)
766 {
767 	return a->id_len == b->id_len &&
768 	       memcmp(a->id, b->id, a->id_len) == 0;
769 }
770 
771 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
772 				       struct sk_buff *skb);
773 
774 enum tc_setup_type {
775 	TC_SETUP_QDISC_MQPRIO,
776 	TC_SETUP_CLSU32,
777 	TC_SETUP_CLSFLOWER,
778 	TC_SETUP_CLSMATCHALL,
779 	TC_SETUP_CLSBPF,
780 	TC_SETUP_BLOCK,
781 	TC_SETUP_QDISC_CBS,
782 	TC_SETUP_QDISC_RED,
783 	TC_SETUP_QDISC_PRIO,
784 };
785 
786 /* These structures hold the attributes of bpf state that are being passed
787  * to the netdevice through the bpf op.
788  */
789 enum bpf_netdev_command {
790 	/* Set or clear a bpf program used in the earliest stages of packet
791 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
792 	 * is responsible for calling bpf_prog_put on any old progs that are
793 	 * stored. In case of error, the callee need not release the new prog
794 	 * reference, but on success it takes ownership and must bpf_prog_put
795 	 * when it is no longer used.
796 	 */
797 	XDP_SETUP_PROG,
798 	XDP_SETUP_PROG_HW,
799 	/* Check if a bpf program is set on the device.  The callee should
800 	 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
801 	 * is equivalent to XDP_ATTACHED_DRV.
802 	 */
803 	XDP_QUERY_PROG,
804 	/* BPF program for offload callbacks, invoked at program load time. */
805 	BPF_OFFLOAD_VERIFIER_PREP,
806 	BPF_OFFLOAD_TRANSLATE,
807 	BPF_OFFLOAD_DESTROY,
808 	BPF_OFFLOAD_MAP_ALLOC,
809 	BPF_OFFLOAD_MAP_FREE,
810 };
811 
812 struct bpf_prog_offload_ops;
813 struct netlink_ext_ack;
814 
815 struct netdev_bpf {
816 	enum bpf_netdev_command command;
817 	union {
818 		/* XDP_SETUP_PROG */
819 		struct {
820 			u32 flags;
821 			struct bpf_prog *prog;
822 			struct netlink_ext_ack *extack;
823 		};
824 		/* XDP_QUERY_PROG */
825 		struct {
826 			u8 prog_attached;
827 			u32 prog_id;
828 			/* flags with which program was installed */
829 			u32 prog_flags;
830 		};
831 		/* BPF_OFFLOAD_VERIFIER_PREP */
832 		struct {
833 			struct bpf_prog *prog;
834 			const struct bpf_prog_offload_ops *ops; /* callee set */
835 		} verifier;
836 		/* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
837 		struct {
838 			struct bpf_prog *prog;
839 		} offload;
840 		/* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
841 		struct {
842 			struct bpf_offloaded_map *offmap;
843 		};
844 	};
845 };
846 
847 #ifdef CONFIG_XFRM_OFFLOAD
848 struct xfrmdev_ops {
849 	int	(*xdo_dev_state_add) (struct xfrm_state *x);
850 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
851 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
852 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
853 				       struct xfrm_state *x);
854 };
855 #endif
856 
857 struct dev_ifalias {
858 	struct rcu_head rcuhead;
859 	char ifalias[];
860 };
861 
862 /*
863  * This structure defines the management hooks for network devices.
864  * The following hooks can be defined; unless noted otherwise, they are
865  * optional and can be filled with a null pointer.
866  *
867  * int (*ndo_init)(struct net_device *dev);
868  *     This function is called once when a network device is registered.
869  *     The network device can use this for any late stage initialization
870  *     or semantic validation. It can fail with an error code which will
871  *     be propagated back to register_netdev.
872  *
873  * void (*ndo_uninit)(struct net_device *dev);
874  *     This function is called when device is unregistered or when registration
875  *     fails. It is not called if init fails.
876  *
877  * int (*ndo_open)(struct net_device *dev);
878  *     This function is called when a network device transitions to the up
879  *     state.
880  *
881  * int (*ndo_stop)(struct net_device *dev);
882  *     This function is called when a network device transitions to the down
883  *     state.
884  *
885  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
886  *                               struct net_device *dev);
887  *	Called when a packet needs to be transmitted.
888  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
889  *	the queue before that can happen; it's for obsolete devices and weird
890  *	corner cases, but the stack really does a non-trivial amount
891  *	of useless work if you return NETDEV_TX_BUSY.
892  *	Required; cannot be NULL.
893  *
894  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
895  *					   struct net_device *dev
896  *					   netdev_features_t features);
897  *	Called by core transmit path to determine if device is capable of
898  *	performing offload operations on a given packet. This is to give
899  *	the device an opportunity to implement any restrictions that cannot
900  *	be otherwise expressed by feature flags. The check is called with
901  *	the set of features that the stack has calculated and it returns
902  *	those the driver believes to be appropriate.
903  *
904  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
905  *                         void *accel_priv, select_queue_fallback_t fallback);
906  *	Called to decide which queue to use when device supports multiple
907  *	transmit queues.
908  *
909  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
910  *	This function is called to allow device receiver to make
911  *	changes to configuration when multicast or promiscuous is enabled.
912  *
913  * void (*ndo_set_rx_mode)(struct net_device *dev);
914  *	This function is called device changes address list filtering.
915  *	If driver handles unicast address filtering, it should set
916  *	IFF_UNICAST_FLT in its priv_flags.
917  *
918  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
919  *	This function  is called when the Media Access Control address
920  *	needs to be changed. If this interface is not defined, the
921  *	MAC address can not be changed.
922  *
923  * int (*ndo_validate_addr)(struct net_device *dev);
924  *	Test if Media Access Control address is valid for the device.
925  *
926  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
927  *	Called when a user requests an ioctl which can't be handled by
928  *	the generic interface code. If not defined ioctls return
929  *	not supported error code.
930  *
931  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
932  *	Used to set network devices bus interface parameters. This interface
933  *	is retained for legacy reasons; new devices should use the bus
934  *	interface (PCI) for low level management.
935  *
936  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
937  *	Called when a user wants to change the Maximum Transfer Unit
938  *	of a device.
939  *
940  * void (*ndo_tx_timeout)(struct net_device *dev);
941  *	Callback used when the transmitter has not made any progress
942  *	for dev->watchdog ticks.
943  *
944  * void (*ndo_get_stats64)(struct net_device *dev,
945  *                         struct rtnl_link_stats64 *storage);
946  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
947  *	Called when a user wants to get the network device usage
948  *	statistics. Drivers must do one of the following:
949  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
950  *	   rtnl_link_stats64 structure passed by the caller.
951  *	2. Define @ndo_get_stats to update a net_device_stats structure
952  *	   (which should normally be dev->stats) and return a pointer to
953  *	   it. The structure may be changed asynchronously only if each
954  *	   field is written atomically.
955  *	3. Update dev->stats asynchronously and atomically, and define
956  *	   neither operation.
957  *
958  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
959  *	Return true if this device supports offload stats of this attr_id.
960  *
961  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
962  *	void *attr_data)
963  *	Get statistics for offload operations by attr_id. Write it into the
964  *	attr_data pointer.
965  *
966  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
967  *	If device supports VLAN filtering this function is called when a
968  *	VLAN id is registered.
969  *
970  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
971  *	If device supports VLAN filtering this function is called when a
972  *	VLAN id is unregistered.
973  *
974  * void (*ndo_poll_controller)(struct net_device *dev);
975  *
976  *	SR-IOV management functions.
977  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
978  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
979  *			  u8 qos, __be16 proto);
980  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
981  *			  int max_tx_rate);
982  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
983  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
984  * int (*ndo_get_vf_config)(struct net_device *dev,
985  *			    int vf, struct ifla_vf_info *ivf);
986  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
987  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
988  *			  struct nlattr *port[]);
989  *
990  *      Enable or disable the VF ability to query its RSS Redirection Table and
991  *      Hash Key. This is needed since on some devices VF share this information
992  *      with PF and querying it may introduce a theoretical security risk.
993  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
994  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
995  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
996  *		       void *type_data);
997  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
998  *	This is always called from the stack with the rtnl lock held and netif
999  *	tx queues stopped. This allows the netdevice to perform queue
1000  *	management safely.
1001  *
1002  *	Fiber Channel over Ethernet (FCoE) offload functions.
1003  * int (*ndo_fcoe_enable)(struct net_device *dev);
1004  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
1005  *	so the underlying device can perform whatever needed configuration or
1006  *	initialization to support acceleration of FCoE traffic.
1007  *
1008  * int (*ndo_fcoe_disable)(struct net_device *dev);
1009  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
1010  *	so the underlying device can perform whatever needed clean-ups to
1011  *	stop supporting acceleration of FCoE traffic.
1012  *
1013  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1014  *			     struct scatterlist *sgl, unsigned int sgc);
1015  *	Called when the FCoE Initiator wants to initialize an I/O that
1016  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1017  *	perform necessary setup and returns 1 to indicate the device is set up
1018  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1019  *
1020  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1021  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1022  *	indicated by the FC exchange id 'xid', so the underlying device can
1023  *	clean up and reuse resources for later DDP requests.
1024  *
1025  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1026  *			      struct scatterlist *sgl, unsigned int sgc);
1027  *	Called when the FCoE Target wants to initialize an I/O that
1028  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1029  *	perform necessary setup and returns 1 to indicate the device is set up
1030  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1031  *
1032  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1033  *			       struct netdev_fcoe_hbainfo *hbainfo);
1034  *	Called when the FCoE Protocol stack wants information on the underlying
1035  *	device. This information is utilized by the FCoE protocol stack to
1036  *	register attributes with Fiber Channel management service as per the
1037  *	FC-GS Fabric Device Management Information(FDMI) specification.
1038  *
1039  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1040  *	Called when the underlying device wants to override default World Wide
1041  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1042  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1043  *	protocol stack to use.
1044  *
1045  *	RFS acceleration.
1046  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1047  *			    u16 rxq_index, u32 flow_id);
1048  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1049  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1050  *	Return the filter ID on success, or a negative error code.
1051  *
1052  *	Slave management functions (for bridge, bonding, etc).
1053  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1054  *	Called to make another netdev an underling.
1055  *
1056  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1057  *	Called to release previously enslaved netdev.
1058  *
1059  *      Feature/offload setting functions.
1060  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1061  *		netdev_features_t features);
1062  *	Adjusts the requested feature flags according to device-specific
1063  *	constraints, and returns the resulting flags. Must not modify
1064  *	the device state.
1065  *
1066  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1067  *	Called to update device configuration to new features. Passed
1068  *	feature set might be less than what was returned by ndo_fix_features()).
1069  *	Must return >0 or -errno if it changed dev->features itself.
1070  *
1071  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1072  *		      struct net_device *dev,
1073  *		      const unsigned char *addr, u16 vid, u16 flags)
1074  *	Adds an FDB entry to dev for addr.
1075  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1076  *		      struct net_device *dev,
1077  *		      const unsigned char *addr, u16 vid)
1078  *	Deletes the FDB entry from dev coresponding to addr.
1079  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1080  *		       struct net_device *dev, struct net_device *filter_dev,
1081  *		       int *idx)
1082  *	Used to add FDB entries to dump requests. Implementers should add
1083  *	entries to skb and update idx with the number of entries.
1084  *
1085  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1086  *			     u16 flags)
1087  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1088  *			     struct net_device *dev, u32 filter_mask,
1089  *			     int nlflags)
1090  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1091  *			     u16 flags);
1092  *
1093  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1094  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1095  *	which do not represent real hardware may define this to allow their
1096  *	userspace components to manage their virtual carrier state. Devices
1097  *	that determine carrier state from physical hardware properties (eg
1098  *	network cables) or protocol-dependent mechanisms (eg
1099  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1100  *
1101  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1102  *			       struct netdev_phys_item_id *ppid);
1103  *	Called to get ID of physical port of this device. If driver does
1104  *	not implement this, it is assumed that the hw is not able to have
1105  *	multiple net devices on single physical port.
1106  *
1107  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1108  *			      struct udp_tunnel_info *ti);
1109  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1110  *	address family that a UDP tunnel is listnening to. It is called only
1111  *	when a new port starts listening. The operation is protected by the
1112  *	RTNL.
1113  *
1114  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1115  *			      struct udp_tunnel_info *ti);
1116  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1117  *	address family that the UDP tunnel is not listening to anymore. The
1118  *	operation is protected by the RTNL.
1119  *
1120  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1121  *				 struct net_device *dev)
1122  *	Called by upper layer devices to accelerate switching or other
1123  *	station functionality into hardware. 'pdev is the lowerdev
1124  *	to use for the offload and 'dev' is the net device that will
1125  *	back the offload. Returns a pointer to the private structure
1126  *	the upper layer will maintain.
1127  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1128  *	Called by upper layer device to delete the station created
1129  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1130  *	the station and priv is the structure returned by the add
1131  *	operation.
1132  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1133  *			     int queue_index, u32 maxrate);
1134  *	Called when a user wants to set a max-rate limitation of specific
1135  *	TX queue.
1136  * int (*ndo_get_iflink)(const struct net_device *dev);
1137  *	Called to get the iflink value of this device.
1138  * void (*ndo_change_proto_down)(struct net_device *dev,
1139  *				 bool proto_down);
1140  *	This function is used to pass protocol port error state information
1141  *	to the switch driver. The switch driver can react to the proto_down
1142  *      by doing a phys down on the associated switch port.
1143  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1144  *	This function is used to get egress tunnel information for given skb.
1145  *	This is useful for retrieving outer tunnel header parameters while
1146  *	sampling packet.
1147  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1148  *	This function is used to specify the headroom that the skb must
1149  *	consider when allocation skb during packet reception. Setting
1150  *	appropriate rx headroom value allows avoiding skb head copy on
1151  *	forward. Setting a negative value resets the rx headroom to the
1152  *	default value.
1153  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1154  *	This function is used to set or query state related to XDP on the
1155  *	netdevice and manage BPF offload. See definition of
1156  *	enum bpf_netdev_command for details.
1157  * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1158  *	This function is used to submit a XDP packet for transmit on a
1159  *	netdevice.
1160  * void (*ndo_xdp_flush)(struct net_device *dev);
1161  *	This function is used to inform the driver to flush a particular
1162  *	xdp tx queue. Must be called on same CPU as xdp_xmit.
1163  */
1164 struct net_device_ops {
1165 	int			(*ndo_init)(struct net_device *dev);
1166 	void			(*ndo_uninit)(struct net_device *dev);
1167 	int			(*ndo_open)(struct net_device *dev);
1168 	int			(*ndo_stop)(struct net_device *dev);
1169 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1170 						  struct net_device *dev);
1171 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1172 						      struct net_device *dev,
1173 						      netdev_features_t features);
1174 	u16			(*ndo_select_queue)(struct net_device *dev,
1175 						    struct sk_buff *skb,
1176 						    void *accel_priv,
1177 						    select_queue_fallback_t fallback);
1178 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1179 						       int flags);
1180 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1181 	int			(*ndo_set_mac_address)(struct net_device *dev,
1182 						       void *addr);
1183 	int			(*ndo_validate_addr)(struct net_device *dev);
1184 	int			(*ndo_do_ioctl)(struct net_device *dev,
1185 					        struct ifreq *ifr, int cmd);
1186 	int			(*ndo_set_config)(struct net_device *dev,
1187 					          struct ifmap *map);
1188 	int			(*ndo_change_mtu)(struct net_device *dev,
1189 						  int new_mtu);
1190 	int			(*ndo_neigh_setup)(struct net_device *dev,
1191 						   struct neigh_parms *);
1192 	void			(*ndo_tx_timeout) (struct net_device *dev);
1193 
1194 	void			(*ndo_get_stats64)(struct net_device *dev,
1195 						   struct rtnl_link_stats64 *storage);
1196 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1197 	int			(*ndo_get_offload_stats)(int attr_id,
1198 							 const struct net_device *dev,
1199 							 void *attr_data);
1200 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1201 
1202 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1203 						       __be16 proto, u16 vid);
1204 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1205 						        __be16 proto, u16 vid);
1206 #ifdef CONFIG_NET_POLL_CONTROLLER
1207 	void                    (*ndo_poll_controller)(struct net_device *dev);
1208 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1209 						     struct netpoll_info *info);
1210 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1211 #endif
1212 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1213 						  int queue, u8 *mac);
1214 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1215 						   int queue, u16 vlan,
1216 						   u8 qos, __be16 proto);
1217 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1218 						   int vf, int min_tx_rate,
1219 						   int max_tx_rate);
1220 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1221 						       int vf, bool setting);
1222 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1223 						    int vf, bool setting);
1224 	int			(*ndo_get_vf_config)(struct net_device *dev,
1225 						     int vf,
1226 						     struct ifla_vf_info *ivf);
1227 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1228 							 int vf, int link_state);
1229 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1230 						    int vf,
1231 						    struct ifla_vf_stats
1232 						    *vf_stats);
1233 	int			(*ndo_set_vf_port)(struct net_device *dev,
1234 						   int vf,
1235 						   struct nlattr *port[]);
1236 	int			(*ndo_get_vf_port)(struct net_device *dev,
1237 						   int vf, struct sk_buff *skb);
1238 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1239 						   int vf, u64 guid,
1240 						   int guid_type);
1241 	int			(*ndo_set_vf_rss_query_en)(
1242 						   struct net_device *dev,
1243 						   int vf, bool setting);
1244 	int			(*ndo_setup_tc)(struct net_device *dev,
1245 						enum tc_setup_type type,
1246 						void *type_data);
1247 #if IS_ENABLED(CONFIG_FCOE)
1248 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1249 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1250 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1251 						      u16 xid,
1252 						      struct scatterlist *sgl,
1253 						      unsigned int sgc);
1254 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1255 						     u16 xid);
1256 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1257 						       u16 xid,
1258 						       struct scatterlist *sgl,
1259 						       unsigned int sgc);
1260 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1261 							struct netdev_fcoe_hbainfo *hbainfo);
1262 #endif
1263 
1264 #if IS_ENABLED(CONFIG_LIBFCOE)
1265 #define NETDEV_FCOE_WWNN 0
1266 #define NETDEV_FCOE_WWPN 1
1267 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1268 						    u64 *wwn, int type);
1269 #endif
1270 
1271 #ifdef CONFIG_RFS_ACCEL
1272 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1273 						     const struct sk_buff *skb,
1274 						     u16 rxq_index,
1275 						     u32 flow_id);
1276 #endif
1277 	int			(*ndo_add_slave)(struct net_device *dev,
1278 						 struct net_device *slave_dev,
1279 						 struct netlink_ext_ack *extack);
1280 	int			(*ndo_del_slave)(struct net_device *dev,
1281 						 struct net_device *slave_dev);
1282 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1283 						    netdev_features_t features);
1284 	int			(*ndo_set_features)(struct net_device *dev,
1285 						    netdev_features_t features);
1286 	int			(*ndo_neigh_construct)(struct net_device *dev,
1287 						       struct neighbour *n);
1288 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1289 						     struct neighbour *n);
1290 
1291 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1292 					       struct nlattr *tb[],
1293 					       struct net_device *dev,
1294 					       const unsigned char *addr,
1295 					       u16 vid,
1296 					       u16 flags);
1297 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1298 					       struct nlattr *tb[],
1299 					       struct net_device *dev,
1300 					       const unsigned char *addr,
1301 					       u16 vid);
1302 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1303 						struct netlink_callback *cb,
1304 						struct net_device *dev,
1305 						struct net_device *filter_dev,
1306 						int *idx);
1307 
1308 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1309 						      struct nlmsghdr *nlh,
1310 						      u16 flags);
1311 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1312 						      u32 pid, u32 seq,
1313 						      struct net_device *dev,
1314 						      u32 filter_mask,
1315 						      int nlflags);
1316 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1317 						      struct nlmsghdr *nlh,
1318 						      u16 flags);
1319 	int			(*ndo_change_carrier)(struct net_device *dev,
1320 						      bool new_carrier);
1321 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1322 							struct netdev_phys_item_id *ppid);
1323 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1324 							  char *name, size_t len);
1325 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1326 						      struct udp_tunnel_info *ti);
1327 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1328 						      struct udp_tunnel_info *ti);
1329 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1330 							struct net_device *dev);
1331 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1332 							void *priv);
1333 
1334 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1335 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1336 						      int queue_index,
1337 						      u32 maxrate);
1338 	int			(*ndo_get_iflink)(const struct net_device *dev);
1339 	int			(*ndo_change_proto_down)(struct net_device *dev,
1340 							 bool proto_down);
1341 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1342 						       struct sk_buff *skb);
1343 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1344 						       int needed_headroom);
1345 	int			(*ndo_bpf)(struct net_device *dev,
1346 					   struct netdev_bpf *bpf);
1347 	int			(*ndo_xdp_xmit)(struct net_device *dev,
1348 						struct xdp_buff *xdp);
1349 	void			(*ndo_xdp_flush)(struct net_device *dev);
1350 };
1351 
1352 /**
1353  * enum net_device_priv_flags - &struct net_device priv_flags
1354  *
1355  * These are the &struct net_device, they are only set internally
1356  * by drivers and used in the kernel. These flags are invisible to
1357  * userspace; this means that the order of these flags can change
1358  * during any kernel release.
1359  *
1360  * You should have a pretty good reason to be extending these flags.
1361  *
1362  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1363  * @IFF_EBRIDGE: Ethernet bridging device
1364  * @IFF_BONDING: bonding master or slave
1365  * @IFF_ISATAP: ISATAP interface (RFC4214)
1366  * @IFF_WAN_HDLC: WAN HDLC device
1367  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1368  *	release skb->dst
1369  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1370  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1371  * @IFF_MACVLAN_PORT: device used as macvlan port
1372  * @IFF_BRIDGE_PORT: device used as bridge port
1373  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1374  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1375  * @IFF_UNICAST_FLT: Supports unicast filtering
1376  * @IFF_TEAM_PORT: device used as team port
1377  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1378  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1379  *	change when it's running
1380  * @IFF_MACVLAN: Macvlan device
1381  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1382  *	underlying stacked devices
1383  * @IFF_IPVLAN_MASTER: IPvlan master device
1384  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1385  * @IFF_L3MDEV_MASTER: device is an L3 master device
1386  * @IFF_NO_QUEUE: device can run without qdisc attached
1387  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1388  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1389  * @IFF_TEAM: device is a team device
1390  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1391  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1392  *	entity (i.e. the master device for bridged veth)
1393  * @IFF_MACSEC: device is a MACsec device
1394  */
1395 enum netdev_priv_flags {
1396 	IFF_802_1Q_VLAN			= 1<<0,
1397 	IFF_EBRIDGE			= 1<<1,
1398 	IFF_BONDING			= 1<<2,
1399 	IFF_ISATAP			= 1<<3,
1400 	IFF_WAN_HDLC			= 1<<4,
1401 	IFF_XMIT_DST_RELEASE		= 1<<5,
1402 	IFF_DONT_BRIDGE			= 1<<6,
1403 	IFF_DISABLE_NETPOLL		= 1<<7,
1404 	IFF_MACVLAN_PORT		= 1<<8,
1405 	IFF_BRIDGE_PORT			= 1<<9,
1406 	IFF_OVS_DATAPATH		= 1<<10,
1407 	IFF_TX_SKB_SHARING		= 1<<11,
1408 	IFF_UNICAST_FLT			= 1<<12,
1409 	IFF_TEAM_PORT			= 1<<13,
1410 	IFF_SUPP_NOFCS			= 1<<14,
1411 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1412 	IFF_MACVLAN			= 1<<16,
1413 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1414 	IFF_IPVLAN_MASTER		= 1<<18,
1415 	IFF_IPVLAN_SLAVE		= 1<<19,
1416 	IFF_L3MDEV_MASTER		= 1<<20,
1417 	IFF_NO_QUEUE			= 1<<21,
1418 	IFF_OPENVSWITCH			= 1<<22,
1419 	IFF_L3MDEV_SLAVE		= 1<<23,
1420 	IFF_TEAM			= 1<<24,
1421 	IFF_RXFH_CONFIGURED		= 1<<25,
1422 	IFF_PHONY_HEADROOM		= 1<<26,
1423 	IFF_MACSEC			= 1<<27,
1424 };
1425 
1426 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1427 #define IFF_EBRIDGE			IFF_EBRIDGE
1428 #define IFF_BONDING			IFF_BONDING
1429 #define IFF_ISATAP			IFF_ISATAP
1430 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1431 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1432 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1433 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1434 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1435 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1436 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1437 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1438 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1439 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1440 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1441 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1442 #define IFF_MACVLAN			IFF_MACVLAN
1443 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1444 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1445 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1446 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1447 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1448 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1449 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1450 #define IFF_TEAM			IFF_TEAM
1451 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1452 #define IFF_MACSEC			IFF_MACSEC
1453 
1454 /**
1455  *	struct net_device - The DEVICE structure.
1456  *
1457  *	Actually, this whole structure is a big mistake.  It mixes I/O
1458  *	data with strictly "high-level" data, and it has to know about
1459  *	almost every data structure used in the INET module.
1460  *
1461  *	@name:	This is the first field of the "visible" part of this structure
1462  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1463  *		of the interface.
1464  *
1465  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1466  *	@ifalias:	SNMP alias
1467  *	@mem_end:	Shared memory end
1468  *	@mem_start:	Shared memory start
1469  *	@base_addr:	Device I/O address
1470  *	@irq:		Device IRQ number
1471  *
1472  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1473  *
1474  *	@state:		Generic network queuing layer state, see netdev_state_t
1475  *	@dev_list:	The global list of network devices
1476  *	@napi_list:	List entry used for polling NAPI devices
1477  *	@unreg_list:	List entry  when we are unregistering the
1478  *			device; see the function unregister_netdev
1479  *	@close_list:	List entry used when we are closing the device
1480  *	@ptype_all:     Device-specific packet handlers for all protocols
1481  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1482  *
1483  *	@adj_list:	Directly linked devices, like slaves for bonding
1484  *	@features:	Currently active device features
1485  *	@hw_features:	User-changeable features
1486  *
1487  *	@wanted_features:	User-requested features
1488  *	@vlan_features:		Mask of features inheritable by VLAN devices
1489  *
1490  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1491  *				This field indicates what encapsulation
1492  *				offloads the hardware is capable of doing,
1493  *				and drivers will need to set them appropriately.
1494  *
1495  *	@mpls_features:	Mask of features inheritable by MPLS
1496  *
1497  *	@ifindex:	interface index
1498  *	@group:		The group the device belongs to
1499  *
1500  *	@stats:		Statistics struct, which was left as a legacy, use
1501  *			rtnl_link_stats64 instead
1502  *
1503  *	@rx_dropped:	Dropped packets by core network,
1504  *			do not use this in drivers
1505  *	@tx_dropped:	Dropped packets by core network,
1506  *			do not use this in drivers
1507  *	@rx_nohandler:	nohandler dropped packets by core network on
1508  *			inactive devices, do not use this in drivers
1509  *
1510  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1511  *				instead of ioctl,
1512  *				see <net/iw_handler.h> for details.
1513  *	@wireless_data:	Instance data managed by the core of wireless extensions
1514  *
1515  *	@netdev_ops:	Includes several pointers to callbacks,
1516  *			if one wants to override the ndo_*() functions
1517  *	@ethtool_ops:	Management operations
1518  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1519  *			discovery handling. Necessary for e.g. 6LoWPAN.
1520  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1521  *			of Layer 2 headers.
1522  *
1523  *	@flags:		Interface flags (a la BSD)
1524  *	@priv_flags:	Like 'flags' but invisible to userspace,
1525  *			see if.h for the definitions
1526  *	@gflags:	Global flags ( kept as legacy )
1527  *	@padded:	How much padding added by alloc_netdev()
1528  *	@operstate:	RFC2863 operstate
1529  *	@link_mode:	Mapping policy to operstate
1530  *	@if_port:	Selectable AUI, TP, ...
1531  *	@dma:		DMA channel
1532  *	@mtu:		Interface MTU value
1533  *	@min_mtu:	Interface Minimum MTU value
1534  *	@max_mtu:	Interface Maximum MTU value
1535  *	@type:		Interface hardware type
1536  *	@hard_header_len: Maximum hardware header length.
1537  *	@min_header_len:  Minimum hardware header length
1538  *
1539  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1540  *			  cases can this be guaranteed
1541  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1542  *			  cases can this be guaranteed. Some cases also use
1543  *			  LL_MAX_HEADER instead to allocate the skb
1544  *
1545  *	interface address info:
1546  *
1547  * 	@perm_addr:		Permanent hw address
1548  * 	@addr_assign_type:	Hw address assignment type
1549  * 	@addr_len:		Hardware address length
1550  *	@neigh_priv_len:	Used in neigh_alloc()
1551  * 	@dev_id:		Used to differentiate devices that share
1552  * 				the same link layer address
1553  * 	@dev_port:		Used to differentiate devices that share
1554  * 				the same function
1555  *	@addr_list_lock:	XXX: need comments on this one
1556  *	@uc_promisc:		Counter that indicates promiscuous mode
1557  *				has been enabled due to the need to listen to
1558  *				additional unicast addresses in a device that
1559  *				does not implement ndo_set_rx_mode()
1560  *	@uc:			unicast mac addresses
1561  *	@mc:			multicast mac addresses
1562  *	@dev_addrs:		list of device hw addresses
1563  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1564  *	@promiscuity:		Number of times the NIC is told to work in
1565  *				promiscuous mode; if it becomes 0 the NIC will
1566  *				exit promiscuous mode
1567  *	@allmulti:		Counter, enables or disables allmulticast mode
1568  *
1569  *	@vlan_info:	VLAN info
1570  *	@dsa_ptr:	dsa specific data
1571  *	@tipc_ptr:	TIPC specific data
1572  *	@atalk_ptr:	AppleTalk link
1573  *	@ip_ptr:	IPv4 specific data
1574  *	@dn_ptr:	DECnet specific data
1575  *	@ip6_ptr:	IPv6 specific data
1576  *	@ax25_ptr:	AX.25 specific data
1577  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1578  *
1579  *	@dev_addr:	Hw address (before bcast,
1580  *			because most packets are unicast)
1581  *
1582  *	@_rx:			Array of RX queues
1583  *	@num_rx_queues:		Number of RX queues
1584  *				allocated at register_netdev() time
1585  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1586  *
1587  *	@rx_handler:		handler for received packets
1588  *	@rx_handler_data: 	XXX: need comments on this one
1589  *	@miniq_ingress:		ingress/clsact qdisc specific data for
1590  *				ingress processing
1591  *	@ingress_queue:		XXX: need comments on this one
1592  *	@broadcast:		hw bcast address
1593  *
1594  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1595  *			indexed by RX queue number. Assigned by driver.
1596  *			This must only be set if the ndo_rx_flow_steer
1597  *			operation is defined
1598  *	@index_hlist:		Device index hash chain
1599  *
1600  *	@_tx:			Array of TX queues
1601  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1602  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1603  *	@qdisc:			Root qdisc from userspace point of view
1604  *	@tx_queue_len:		Max frames per queue allowed
1605  *	@tx_global_lock: 	XXX: need comments on this one
1606  *
1607  *	@xps_maps:	XXX: need comments on this one
1608  *	@miniq_egress:		clsact qdisc specific data for
1609  *				egress processing
1610  *	@watchdog_timeo:	Represents the timeout that is used by
1611  *				the watchdog (see dev_watchdog())
1612  *	@watchdog_timer:	List of timers
1613  *
1614  *	@pcpu_refcnt:		Number of references to this device
1615  *	@todo_list:		Delayed register/unregister
1616  *	@link_watch_list:	XXX: need comments on this one
1617  *
1618  *	@reg_state:		Register/unregister state machine
1619  *	@dismantle:		Device is going to be freed
1620  *	@rtnl_link_state:	This enum represents the phases of creating
1621  *				a new link
1622  *
1623  *	@needs_free_netdev:	Should unregister perform free_netdev?
1624  *	@priv_destructor:	Called from unregister
1625  *	@npinfo:		XXX: need comments on this one
1626  * 	@nd_net:		Network namespace this network device is inside
1627  *
1628  * 	@ml_priv:	Mid-layer private
1629  * 	@lstats:	Loopback statistics
1630  * 	@tstats:	Tunnel statistics
1631  * 	@dstats:	Dummy statistics
1632  * 	@vstats:	Virtual ethernet statistics
1633  *
1634  *	@garp_port:	GARP
1635  *	@mrp_port:	MRP
1636  *
1637  *	@dev:		Class/net/name entry
1638  *	@sysfs_groups:	Space for optional device, statistics and wireless
1639  *			sysfs groups
1640  *
1641  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1642  *	@rtnl_link_ops:	Rtnl_link_ops
1643  *
1644  *	@gso_max_size:	Maximum size of generic segmentation offload
1645  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1646  *			NIC for GSO
1647  *
1648  *	@dcbnl_ops:	Data Center Bridging netlink ops
1649  *	@num_tc:	Number of traffic classes in the net device
1650  *	@tc_to_txq:	XXX: need comments on this one
1651  *	@prio_tc_map:	XXX: need comments on this one
1652  *
1653  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1654  *
1655  *	@priomap:	XXX: need comments on this one
1656  *	@phydev:	Physical device may attach itself
1657  *			for hardware timestamping
1658  *
1659  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1660  *	@qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1661  *
1662  *	@proto_down:	protocol port state information can be sent to the
1663  *			switch driver and used to set the phys state of the
1664  *			switch port.
1665  *
1666  *	FIXME: cleanup struct net_device such that network protocol info
1667  *	moves out.
1668  */
1669 
1670 struct net_device {
1671 	char			name[IFNAMSIZ];
1672 	struct hlist_node	name_hlist;
1673 	struct dev_ifalias	__rcu *ifalias;
1674 	/*
1675 	 *	I/O specific fields
1676 	 *	FIXME: Merge these and struct ifmap into one
1677 	 */
1678 	unsigned long		mem_end;
1679 	unsigned long		mem_start;
1680 	unsigned long		base_addr;
1681 	int			irq;
1682 
1683 	atomic_t		carrier_changes;
1684 
1685 	/*
1686 	 *	Some hardware also needs these fields (state,dev_list,
1687 	 *	napi_list,unreg_list,close_list) but they are not
1688 	 *	part of the usual set specified in Space.c.
1689 	 */
1690 
1691 	unsigned long		state;
1692 
1693 	struct list_head	dev_list;
1694 	struct list_head	napi_list;
1695 	struct list_head	unreg_list;
1696 	struct list_head	close_list;
1697 	struct list_head	ptype_all;
1698 	struct list_head	ptype_specific;
1699 
1700 	struct {
1701 		struct list_head upper;
1702 		struct list_head lower;
1703 	} adj_list;
1704 
1705 	netdev_features_t	features;
1706 	netdev_features_t	hw_features;
1707 	netdev_features_t	wanted_features;
1708 	netdev_features_t	vlan_features;
1709 	netdev_features_t	hw_enc_features;
1710 	netdev_features_t	mpls_features;
1711 	netdev_features_t	gso_partial_features;
1712 
1713 	int			ifindex;
1714 	int			group;
1715 
1716 	struct net_device_stats	stats;
1717 
1718 	atomic_long_t		rx_dropped;
1719 	atomic_long_t		tx_dropped;
1720 	atomic_long_t		rx_nohandler;
1721 
1722 #ifdef CONFIG_WIRELESS_EXT
1723 	const struct iw_handler_def *wireless_handlers;
1724 	struct iw_public_data	*wireless_data;
1725 #endif
1726 	const struct net_device_ops *netdev_ops;
1727 	const struct ethtool_ops *ethtool_ops;
1728 #ifdef CONFIG_NET_SWITCHDEV
1729 	const struct switchdev_ops *switchdev_ops;
1730 #endif
1731 #ifdef CONFIG_NET_L3_MASTER_DEV
1732 	const struct l3mdev_ops	*l3mdev_ops;
1733 #endif
1734 #if IS_ENABLED(CONFIG_IPV6)
1735 	const struct ndisc_ops *ndisc_ops;
1736 #endif
1737 
1738 #ifdef CONFIG_XFRM_OFFLOAD
1739 	const struct xfrmdev_ops *xfrmdev_ops;
1740 #endif
1741 
1742 	const struct header_ops *header_ops;
1743 
1744 	unsigned int		flags;
1745 	unsigned int		priv_flags;
1746 
1747 	unsigned short		gflags;
1748 	unsigned short		padded;
1749 
1750 	unsigned char		operstate;
1751 	unsigned char		link_mode;
1752 
1753 	unsigned char		if_port;
1754 	unsigned char		dma;
1755 
1756 	unsigned int		mtu;
1757 	unsigned int		min_mtu;
1758 	unsigned int		max_mtu;
1759 	unsigned short		type;
1760 	unsigned short		hard_header_len;
1761 	unsigned char		min_header_len;
1762 
1763 	unsigned short		needed_headroom;
1764 	unsigned short		needed_tailroom;
1765 
1766 	/* Interface address info. */
1767 	unsigned char		perm_addr[MAX_ADDR_LEN];
1768 	unsigned char		addr_assign_type;
1769 	unsigned char		addr_len;
1770 	unsigned short		neigh_priv_len;
1771 	unsigned short          dev_id;
1772 	unsigned short          dev_port;
1773 	spinlock_t		addr_list_lock;
1774 	unsigned char		name_assign_type;
1775 	bool			uc_promisc;
1776 	struct netdev_hw_addr_list	uc;
1777 	struct netdev_hw_addr_list	mc;
1778 	struct netdev_hw_addr_list	dev_addrs;
1779 
1780 #ifdef CONFIG_SYSFS
1781 	struct kset		*queues_kset;
1782 #endif
1783 	unsigned int		promiscuity;
1784 	unsigned int		allmulti;
1785 
1786 
1787 	/* Protocol-specific pointers */
1788 
1789 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1790 	struct vlan_info __rcu	*vlan_info;
1791 #endif
1792 #if IS_ENABLED(CONFIG_NET_DSA)
1793 	struct dsa_port		*dsa_ptr;
1794 #endif
1795 #if IS_ENABLED(CONFIG_TIPC)
1796 	struct tipc_bearer __rcu *tipc_ptr;
1797 #endif
1798 	void 			*atalk_ptr;
1799 	struct in_device __rcu	*ip_ptr;
1800 	struct dn_dev __rcu     *dn_ptr;
1801 	struct inet6_dev __rcu	*ip6_ptr;
1802 	void			*ax25_ptr;
1803 	struct wireless_dev	*ieee80211_ptr;
1804 	struct wpan_dev		*ieee802154_ptr;
1805 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1806 	struct mpls_dev __rcu	*mpls_ptr;
1807 #endif
1808 
1809 /*
1810  * Cache lines mostly used on receive path (including eth_type_trans())
1811  */
1812 	/* Interface address info used in eth_type_trans() */
1813 	unsigned char		*dev_addr;
1814 
1815 	struct netdev_rx_queue	*_rx;
1816 	unsigned int		num_rx_queues;
1817 	unsigned int		real_num_rx_queues;
1818 
1819 	struct bpf_prog __rcu	*xdp_prog;
1820 	unsigned long		gro_flush_timeout;
1821 	rx_handler_func_t __rcu	*rx_handler;
1822 	void __rcu		*rx_handler_data;
1823 
1824 #ifdef CONFIG_NET_CLS_ACT
1825 	struct mini_Qdisc __rcu	*miniq_ingress;
1826 #endif
1827 	struct netdev_queue __rcu *ingress_queue;
1828 #ifdef CONFIG_NETFILTER_INGRESS
1829 	struct nf_hook_entries __rcu *nf_hooks_ingress;
1830 #endif
1831 
1832 	unsigned char		broadcast[MAX_ADDR_LEN];
1833 #ifdef CONFIG_RFS_ACCEL
1834 	struct cpu_rmap		*rx_cpu_rmap;
1835 #endif
1836 	struct hlist_node	index_hlist;
1837 
1838 /*
1839  * Cache lines mostly used on transmit path
1840  */
1841 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1842 	unsigned int		num_tx_queues;
1843 	unsigned int		real_num_tx_queues;
1844 	struct Qdisc		*qdisc;
1845 #ifdef CONFIG_NET_SCHED
1846 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1847 #endif
1848 	unsigned int		tx_queue_len;
1849 	spinlock_t		tx_global_lock;
1850 	int			watchdog_timeo;
1851 
1852 #ifdef CONFIG_XPS
1853 	struct xps_dev_maps __rcu *xps_maps;
1854 #endif
1855 #ifdef CONFIG_NET_CLS_ACT
1856 	struct mini_Qdisc __rcu	*miniq_egress;
1857 #endif
1858 
1859 	/* These may be needed for future network-power-down code. */
1860 	struct timer_list	watchdog_timer;
1861 
1862 	int __percpu		*pcpu_refcnt;
1863 	struct list_head	todo_list;
1864 
1865 	struct list_head	link_watch_list;
1866 
1867 	enum { NETREG_UNINITIALIZED=0,
1868 	       NETREG_REGISTERED,	/* completed register_netdevice */
1869 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1870 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1871 	       NETREG_RELEASED,		/* called free_netdev */
1872 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1873 	} reg_state:8;
1874 
1875 	bool dismantle;
1876 
1877 	enum {
1878 		RTNL_LINK_INITIALIZED,
1879 		RTNL_LINK_INITIALIZING,
1880 	} rtnl_link_state:16;
1881 
1882 	bool needs_free_netdev;
1883 	void (*priv_destructor)(struct net_device *dev);
1884 
1885 #ifdef CONFIG_NETPOLL
1886 	struct netpoll_info __rcu	*npinfo;
1887 #endif
1888 
1889 	possible_net_t			nd_net;
1890 
1891 	/* mid-layer private */
1892 	union {
1893 		void					*ml_priv;
1894 		struct pcpu_lstats __percpu		*lstats;
1895 		struct pcpu_sw_netstats __percpu	*tstats;
1896 		struct pcpu_dstats __percpu		*dstats;
1897 		struct pcpu_vstats __percpu		*vstats;
1898 	};
1899 
1900 #if IS_ENABLED(CONFIG_GARP)
1901 	struct garp_port __rcu	*garp_port;
1902 #endif
1903 #if IS_ENABLED(CONFIG_MRP)
1904 	struct mrp_port __rcu	*mrp_port;
1905 #endif
1906 
1907 	struct device		dev;
1908 	const struct attribute_group *sysfs_groups[4];
1909 	const struct attribute_group *sysfs_rx_queue_group;
1910 
1911 	const struct rtnl_link_ops *rtnl_link_ops;
1912 
1913 	/* for setting kernel sock attribute on TCP connection setup */
1914 #define GSO_MAX_SIZE		65536
1915 	unsigned int		gso_max_size;
1916 #define GSO_MAX_SEGS		65535
1917 	u16			gso_max_segs;
1918 
1919 #ifdef CONFIG_DCB
1920 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1921 #endif
1922 	u8			num_tc;
1923 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1924 	u8			prio_tc_map[TC_BITMASK + 1];
1925 
1926 #if IS_ENABLED(CONFIG_FCOE)
1927 	unsigned int		fcoe_ddp_xid;
1928 #endif
1929 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1930 	struct netprio_map __rcu *priomap;
1931 #endif
1932 	struct phy_device	*phydev;
1933 	struct lock_class_key	*qdisc_tx_busylock;
1934 	struct lock_class_key	*qdisc_running_key;
1935 	bool			proto_down;
1936 };
1937 #define to_net_dev(d) container_of(d, struct net_device, dev)
1938 
1939 static inline bool netif_elide_gro(const struct net_device *dev)
1940 {
1941 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1942 		return true;
1943 	return false;
1944 }
1945 
1946 #define	NETDEV_ALIGN		32
1947 
1948 static inline
1949 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1950 {
1951 	return dev->prio_tc_map[prio & TC_BITMASK];
1952 }
1953 
1954 static inline
1955 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1956 {
1957 	if (tc >= dev->num_tc)
1958 		return -EINVAL;
1959 
1960 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1961 	return 0;
1962 }
1963 
1964 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1965 void netdev_reset_tc(struct net_device *dev);
1966 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1967 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1968 
1969 static inline
1970 int netdev_get_num_tc(struct net_device *dev)
1971 {
1972 	return dev->num_tc;
1973 }
1974 
1975 static inline
1976 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1977 					 unsigned int index)
1978 {
1979 	return &dev->_tx[index];
1980 }
1981 
1982 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1983 						    const struct sk_buff *skb)
1984 {
1985 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1986 }
1987 
1988 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1989 					    void (*f)(struct net_device *,
1990 						      struct netdev_queue *,
1991 						      void *),
1992 					    void *arg)
1993 {
1994 	unsigned int i;
1995 
1996 	for (i = 0; i < dev->num_tx_queues; i++)
1997 		f(dev, &dev->_tx[i], arg);
1998 }
1999 
2000 #define netdev_lockdep_set_classes(dev)				\
2001 {								\
2002 	static struct lock_class_key qdisc_tx_busylock_key;	\
2003 	static struct lock_class_key qdisc_running_key;		\
2004 	static struct lock_class_key qdisc_xmit_lock_key;	\
2005 	static struct lock_class_key dev_addr_list_lock_key;	\
2006 	unsigned int i;						\
2007 								\
2008 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
2009 	(dev)->qdisc_running_key = &qdisc_running_key;		\
2010 	lockdep_set_class(&(dev)->addr_list_lock,		\
2011 			  &dev_addr_list_lock_key); 		\
2012 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
2013 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
2014 				  &qdisc_xmit_lock_key);	\
2015 }
2016 
2017 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2018 				    struct sk_buff *skb,
2019 				    void *accel_priv);
2020 
2021 /* returns the headroom that the master device needs to take in account
2022  * when forwarding to this dev
2023  */
2024 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2025 {
2026 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2027 }
2028 
2029 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2030 {
2031 	if (dev->netdev_ops->ndo_set_rx_headroom)
2032 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2033 }
2034 
2035 /* set the device rx headroom to the dev's default */
2036 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2037 {
2038 	netdev_set_rx_headroom(dev, -1);
2039 }
2040 
2041 /*
2042  * Net namespace inlines
2043  */
2044 static inline
2045 struct net *dev_net(const struct net_device *dev)
2046 {
2047 	return read_pnet(&dev->nd_net);
2048 }
2049 
2050 static inline
2051 void dev_net_set(struct net_device *dev, struct net *net)
2052 {
2053 	write_pnet(&dev->nd_net, net);
2054 }
2055 
2056 /**
2057  *	netdev_priv - access network device private data
2058  *	@dev: network device
2059  *
2060  * Get network device private data
2061  */
2062 static inline void *netdev_priv(const struct net_device *dev)
2063 {
2064 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2065 }
2066 
2067 /* Set the sysfs physical device reference for the network logical device
2068  * if set prior to registration will cause a symlink during initialization.
2069  */
2070 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2071 
2072 /* Set the sysfs device type for the network logical device to allow
2073  * fine-grained identification of different network device types. For
2074  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2075  */
2076 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2077 
2078 /* Default NAPI poll() weight
2079  * Device drivers are strongly advised to not use bigger value
2080  */
2081 #define NAPI_POLL_WEIGHT 64
2082 
2083 /**
2084  *	netif_napi_add - initialize a NAPI context
2085  *	@dev:  network device
2086  *	@napi: NAPI context
2087  *	@poll: polling function
2088  *	@weight: default weight
2089  *
2090  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2091  * *any* of the other NAPI-related functions.
2092  */
2093 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2094 		    int (*poll)(struct napi_struct *, int), int weight);
2095 
2096 /**
2097  *	netif_tx_napi_add - initialize a NAPI context
2098  *	@dev:  network device
2099  *	@napi: NAPI context
2100  *	@poll: polling function
2101  *	@weight: default weight
2102  *
2103  * This variant of netif_napi_add() should be used from drivers using NAPI
2104  * to exclusively poll a TX queue.
2105  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2106  */
2107 static inline void netif_tx_napi_add(struct net_device *dev,
2108 				     struct napi_struct *napi,
2109 				     int (*poll)(struct napi_struct *, int),
2110 				     int weight)
2111 {
2112 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2113 	netif_napi_add(dev, napi, poll, weight);
2114 }
2115 
2116 /**
2117  *  netif_napi_del - remove a NAPI context
2118  *  @napi: NAPI context
2119  *
2120  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2121  */
2122 void netif_napi_del(struct napi_struct *napi);
2123 
2124 struct napi_gro_cb {
2125 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2126 	void	*frag0;
2127 
2128 	/* Length of frag0. */
2129 	unsigned int frag0_len;
2130 
2131 	/* This indicates where we are processing relative to skb->data. */
2132 	int	data_offset;
2133 
2134 	/* This is non-zero if the packet cannot be merged with the new skb. */
2135 	u16	flush;
2136 
2137 	/* Save the IP ID here and check when we get to the transport layer */
2138 	u16	flush_id;
2139 
2140 	/* Number of segments aggregated. */
2141 	u16	count;
2142 
2143 	/* Start offset for remote checksum offload */
2144 	u16	gro_remcsum_start;
2145 
2146 	/* jiffies when first packet was created/queued */
2147 	unsigned long age;
2148 
2149 	/* Used in ipv6_gro_receive() and foo-over-udp */
2150 	u16	proto;
2151 
2152 	/* This is non-zero if the packet may be of the same flow. */
2153 	u8	same_flow:1;
2154 
2155 	/* Used in tunnel GRO receive */
2156 	u8	encap_mark:1;
2157 
2158 	/* GRO checksum is valid */
2159 	u8	csum_valid:1;
2160 
2161 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2162 	u8	csum_cnt:3;
2163 
2164 	/* Free the skb? */
2165 	u8	free:2;
2166 #define NAPI_GRO_FREE		  1
2167 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2168 
2169 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2170 	u8	is_ipv6:1;
2171 
2172 	/* Used in GRE, set in fou/gue_gro_receive */
2173 	u8	is_fou:1;
2174 
2175 	/* Used to determine if flush_id can be ignored */
2176 	u8	is_atomic:1;
2177 
2178 	/* Number of gro_receive callbacks this packet already went through */
2179 	u8 recursion_counter:4;
2180 
2181 	/* 1 bit hole */
2182 
2183 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2184 	__wsum	csum;
2185 
2186 	/* used in skb_gro_receive() slow path */
2187 	struct sk_buff *last;
2188 };
2189 
2190 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2191 
2192 #define GRO_RECURSION_LIMIT 15
2193 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2194 {
2195 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2196 }
2197 
2198 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2199 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2200 						struct sk_buff **head,
2201 						struct sk_buff *skb)
2202 {
2203 	if (unlikely(gro_recursion_inc_test(skb))) {
2204 		NAPI_GRO_CB(skb)->flush |= 1;
2205 		return NULL;
2206 	}
2207 
2208 	return cb(head, skb);
2209 }
2210 
2211 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2212 					     struct sk_buff *);
2213 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2214 						   struct sock *sk,
2215 						   struct sk_buff **head,
2216 						   struct sk_buff *skb)
2217 {
2218 	if (unlikely(gro_recursion_inc_test(skb))) {
2219 		NAPI_GRO_CB(skb)->flush |= 1;
2220 		return NULL;
2221 	}
2222 
2223 	return cb(sk, head, skb);
2224 }
2225 
2226 struct packet_type {
2227 	__be16			type;	/* This is really htons(ether_type). */
2228 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2229 	int			(*func) (struct sk_buff *,
2230 					 struct net_device *,
2231 					 struct packet_type *,
2232 					 struct net_device *);
2233 	bool			(*id_match)(struct packet_type *ptype,
2234 					    struct sock *sk);
2235 	void			*af_packet_priv;
2236 	struct list_head	list;
2237 };
2238 
2239 struct offload_callbacks {
2240 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2241 						netdev_features_t features);
2242 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2243 						 struct sk_buff *skb);
2244 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2245 };
2246 
2247 struct packet_offload {
2248 	__be16			 type;	/* This is really htons(ether_type). */
2249 	u16			 priority;
2250 	struct offload_callbacks callbacks;
2251 	struct list_head	 list;
2252 };
2253 
2254 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2255 struct pcpu_sw_netstats {
2256 	u64     rx_packets;
2257 	u64     rx_bytes;
2258 	u64     tx_packets;
2259 	u64     tx_bytes;
2260 	struct u64_stats_sync   syncp;
2261 };
2262 
2263 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2264 ({									\
2265 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2266 	if (pcpu_stats)	{						\
2267 		int __cpu;						\
2268 		for_each_possible_cpu(__cpu) {				\
2269 			typeof(type) *stat;				\
2270 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2271 			u64_stats_init(&stat->syncp);			\
2272 		}							\
2273 	}								\
2274 	pcpu_stats;							\
2275 })
2276 
2277 #define netdev_alloc_pcpu_stats(type)					\
2278 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2279 
2280 enum netdev_lag_tx_type {
2281 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2282 	NETDEV_LAG_TX_TYPE_RANDOM,
2283 	NETDEV_LAG_TX_TYPE_BROADCAST,
2284 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2285 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2286 	NETDEV_LAG_TX_TYPE_HASH,
2287 };
2288 
2289 struct netdev_lag_upper_info {
2290 	enum netdev_lag_tx_type tx_type;
2291 };
2292 
2293 struct netdev_lag_lower_state_info {
2294 	u8 link_up : 1,
2295 	   tx_enabled : 1;
2296 };
2297 
2298 #include <linux/notifier.h>
2299 
2300 /* netdevice notifier chain. Please remember to update the rtnetlink
2301  * notification exclusion list in rtnetlink_event() when adding new
2302  * types.
2303  */
2304 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2305 #define NETDEV_DOWN	0x0002
2306 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2307 				   detected a hardware crash and restarted
2308 				   - we can use this eg to kick tcp sessions
2309 				   once done */
2310 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2311 #define NETDEV_REGISTER 0x0005
2312 #define NETDEV_UNREGISTER	0x0006
2313 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2314 #define NETDEV_CHANGEADDR	0x0008
2315 #define NETDEV_GOING_DOWN	0x0009
2316 #define NETDEV_CHANGENAME	0x000A
2317 #define NETDEV_FEAT_CHANGE	0x000B
2318 #define NETDEV_BONDING_FAILOVER 0x000C
2319 #define NETDEV_PRE_UP		0x000D
2320 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2321 #define NETDEV_POST_TYPE_CHANGE	0x000F
2322 #define NETDEV_POST_INIT	0x0010
2323 #define NETDEV_UNREGISTER_FINAL 0x0011
2324 #define NETDEV_RELEASE		0x0012
2325 #define NETDEV_NOTIFY_PEERS	0x0013
2326 #define NETDEV_JOIN		0x0014
2327 #define NETDEV_CHANGEUPPER	0x0015
2328 #define NETDEV_RESEND_IGMP	0x0016
2329 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2330 #define NETDEV_CHANGEINFODATA	0x0018
2331 #define NETDEV_BONDING_INFO	0x0019
2332 #define NETDEV_PRECHANGEUPPER	0x001A
2333 #define NETDEV_CHANGELOWERSTATE	0x001B
2334 #define NETDEV_UDP_TUNNEL_PUSH_INFO	0x001C
2335 #define NETDEV_UDP_TUNNEL_DROP_INFO	0x001D
2336 #define NETDEV_CHANGE_TX_QUEUE_LEN	0x001E
2337 
2338 int register_netdevice_notifier(struct notifier_block *nb);
2339 int unregister_netdevice_notifier(struct notifier_block *nb);
2340 
2341 struct netdev_notifier_info {
2342 	struct net_device	*dev;
2343 	struct netlink_ext_ack	*extack;
2344 };
2345 
2346 struct netdev_notifier_change_info {
2347 	struct netdev_notifier_info info; /* must be first */
2348 	unsigned int flags_changed;
2349 };
2350 
2351 struct netdev_notifier_changeupper_info {
2352 	struct netdev_notifier_info info; /* must be first */
2353 	struct net_device *upper_dev; /* new upper dev */
2354 	bool master; /* is upper dev master */
2355 	bool linking; /* is the notification for link or unlink */
2356 	void *upper_info; /* upper dev info */
2357 };
2358 
2359 struct netdev_notifier_changelowerstate_info {
2360 	struct netdev_notifier_info info; /* must be first */
2361 	void *lower_state_info; /* is lower dev state */
2362 };
2363 
2364 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2365 					     struct net_device *dev)
2366 {
2367 	info->dev = dev;
2368 	info->extack = NULL;
2369 }
2370 
2371 static inline struct net_device *
2372 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2373 {
2374 	return info->dev;
2375 }
2376 
2377 static inline struct netlink_ext_ack *
2378 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2379 {
2380 	return info->extack;
2381 }
2382 
2383 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2384 
2385 
2386 extern rwlock_t				dev_base_lock;		/* Device list lock */
2387 
2388 #define for_each_netdev(net, d)		\
2389 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2390 #define for_each_netdev_reverse(net, d)	\
2391 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2392 #define for_each_netdev_rcu(net, d)		\
2393 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2394 #define for_each_netdev_safe(net, d, n)	\
2395 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2396 #define for_each_netdev_continue(net, d)		\
2397 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2398 #define for_each_netdev_continue_rcu(net, d)		\
2399 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2400 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2401 		for_each_netdev_rcu(&init_net, slave)	\
2402 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2403 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2404 
2405 static inline struct net_device *next_net_device(struct net_device *dev)
2406 {
2407 	struct list_head *lh;
2408 	struct net *net;
2409 
2410 	net = dev_net(dev);
2411 	lh = dev->dev_list.next;
2412 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2413 }
2414 
2415 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2416 {
2417 	struct list_head *lh;
2418 	struct net *net;
2419 
2420 	net = dev_net(dev);
2421 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2422 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2423 }
2424 
2425 static inline struct net_device *first_net_device(struct net *net)
2426 {
2427 	return list_empty(&net->dev_base_head) ? NULL :
2428 		net_device_entry(net->dev_base_head.next);
2429 }
2430 
2431 static inline struct net_device *first_net_device_rcu(struct net *net)
2432 {
2433 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2434 
2435 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2436 }
2437 
2438 int netdev_boot_setup_check(struct net_device *dev);
2439 unsigned long netdev_boot_base(const char *prefix, int unit);
2440 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2441 				       const char *hwaddr);
2442 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2443 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2444 void dev_add_pack(struct packet_type *pt);
2445 void dev_remove_pack(struct packet_type *pt);
2446 void __dev_remove_pack(struct packet_type *pt);
2447 void dev_add_offload(struct packet_offload *po);
2448 void dev_remove_offload(struct packet_offload *po);
2449 
2450 int dev_get_iflink(const struct net_device *dev);
2451 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2452 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2453 				      unsigned short mask);
2454 struct net_device *dev_get_by_name(struct net *net, const char *name);
2455 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2456 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2457 int dev_alloc_name(struct net_device *dev, const char *name);
2458 int dev_open(struct net_device *dev);
2459 void dev_close(struct net_device *dev);
2460 void dev_close_many(struct list_head *head, bool unlink);
2461 void dev_disable_lro(struct net_device *dev);
2462 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2463 int dev_queue_xmit(struct sk_buff *skb);
2464 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2465 int register_netdevice(struct net_device *dev);
2466 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2467 void unregister_netdevice_many(struct list_head *head);
2468 static inline void unregister_netdevice(struct net_device *dev)
2469 {
2470 	unregister_netdevice_queue(dev, NULL);
2471 }
2472 
2473 int netdev_refcnt_read(const struct net_device *dev);
2474 void free_netdev(struct net_device *dev);
2475 void netdev_freemem(struct net_device *dev);
2476 void synchronize_net(void);
2477 int init_dummy_netdev(struct net_device *dev);
2478 
2479 DECLARE_PER_CPU(int, xmit_recursion);
2480 #define XMIT_RECURSION_LIMIT	10
2481 
2482 static inline int dev_recursion_level(void)
2483 {
2484 	return this_cpu_read(xmit_recursion);
2485 }
2486 
2487 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2488 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2489 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2490 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2491 int netdev_get_name(struct net *net, char *name, int ifindex);
2492 int dev_restart(struct net_device *dev);
2493 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2494 
2495 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2496 {
2497 	return NAPI_GRO_CB(skb)->data_offset;
2498 }
2499 
2500 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2501 {
2502 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2503 }
2504 
2505 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2506 {
2507 	NAPI_GRO_CB(skb)->data_offset += len;
2508 }
2509 
2510 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2511 					unsigned int offset)
2512 {
2513 	return NAPI_GRO_CB(skb)->frag0 + offset;
2514 }
2515 
2516 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2517 {
2518 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2519 }
2520 
2521 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2522 {
2523 	NAPI_GRO_CB(skb)->frag0 = NULL;
2524 	NAPI_GRO_CB(skb)->frag0_len = 0;
2525 }
2526 
2527 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2528 					unsigned int offset)
2529 {
2530 	if (!pskb_may_pull(skb, hlen))
2531 		return NULL;
2532 
2533 	skb_gro_frag0_invalidate(skb);
2534 	return skb->data + offset;
2535 }
2536 
2537 static inline void *skb_gro_network_header(struct sk_buff *skb)
2538 {
2539 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2540 	       skb_network_offset(skb);
2541 }
2542 
2543 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2544 					const void *start, unsigned int len)
2545 {
2546 	if (NAPI_GRO_CB(skb)->csum_valid)
2547 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2548 						  csum_partial(start, len, 0));
2549 }
2550 
2551 /* GRO checksum functions. These are logical equivalents of the normal
2552  * checksum functions (in skbuff.h) except that they operate on the GRO
2553  * offsets and fields in sk_buff.
2554  */
2555 
2556 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2557 
2558 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2559 {
2560 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2561 }
2562 
2563 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2564 						      bool zero_okay,
2565 						      __sum16 check)
2566 {
2567 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2568 		skb_checksum_start_offset(skb) <
2569 		 skb_gro_offset(skb)) &&
2570 		!skb_at_gro_remcsum_start(skb) &&
2571 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2572 		(!zero_okay || check));
2573 }
2574 
2575 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2576 							   __wsum psum)
2577 {
2578 	if (NAPI_GRO_CB(skb)->csum_valid &&
2579 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2580 		return 0;
2581 
2582 	NAPI_GRO_CB(skb)->csum = psum;
2583 
2584 	return __skb_gro_checksum_complete(skb);
2585 }
2586 
2587 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2588 {
2589 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2590 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2591 		NAPI_GRO_CB(skb)->csum_cnt--;
2592 	} else {
2593 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2594 		 * verified a new top level checksum or an encapsulated one
2595 		 * during GRO. This saves work if we fallback to normal path.
2596 		 */
2597 		__skb_incr_checksum_unnecessary(skb);
2598 	}
2599 }
2600 
2601 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2602 				    compute_pseudo)			\
2603 ({									\
2604 	__sum16 __ret = 0;						\
2605 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2606 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2607 				compute_pseudo(skb, proto));		\
2608 	if (!__ret)							\
2609 		skb_gro_incr_csum_unnecessary(skb);			\
2610 	__ret;								\
2611 })
2612 
2613 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2614 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2615 
2616 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2617 					     compute_pseudo)		\
2618 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2619 
2620 #define skb_gro_checksum_simple_validate(skb)				\
2621 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2622 
2623 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2624 {
2625 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2626 		!NAPI_GRO_CB(skb)->csum_valid);
2627 }
2628 
2629 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2630 					      __sum16 check, __wsum pseudo)
2631 {
2632 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2633 	NAPI_GRO_CB(skb)->csum_valid = 1;
2634 }
2635 
2636 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2637 do {									\
2638 	if (__skb_gro_checksum_convert_check(skb))			\
2639 		__skb_gro_checksum_convert(skb, check,			\
2640 					   compute_pseudo(skb, proto));	\
2641 } while (0)
2642 
2643 struct gro_remcsum {
2644 	int offset;
2645 	__wsum delta;
2646 };
2647 
2648 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2649 {
2650 	grc->offset = 0;
2651 	grc->delta = 0;
2652 }
2653 
2654 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2655 					    unsigned int off, size_t hdrlen,
2656 					    int start, int offset,
2657 					    struct gro_remcsum *grc,
2658 					    bool nopartial)
2659 {
2660 	__wsum delta;
2661 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2662 
2663 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2664 
2665 	if (!nopartial) {
2666 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2667 		return ptr;
2668 	}
2669 
2670 	ptr = skb_gro_header_fast(skb, off);
2671 	if (skb_gro_header_hard(skb, off + plen)) {
2672 		ptr = skb_gro_header_slow(skb, off + plen, off);
2673 		if (!ptr)
2674 			return NULL;
2675 	}
2676 
2677 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2678 			       start, offset);
2679 
2680 	/* Adjust skb->csum since we changed the packet */
2681 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2682 
2683 	grc->offset = off + hdrlen + offset;
2684 	grc->delta = delta;
2685 
2686 	return ptr;
2687 }
2688 
2689 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2690 					   struct gro_remcsum *grc)
2691 {
2692 	void *ptr;
2693 	size_t plen = grc->offset + sizeof(u16);
2694 
2695 	if (!grc->delta)
2696 		return;
2697 
2698 	ptr = skb_gro_header_fast(skb, grc->offset);
2699 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2700 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2701 		if (!ptr)
2702 			return;
2703 	}
2704 
2705 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2706 }
2707 
2708 #ifdef CONFIG_XFRM_OFFLOAD
2709 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2710 {
2711 	if (PTR_ERR(pp) != -EINPROGRESS)
2712 		NAPI_GRO_CB(skb)->flush |= flush;
2713 }
2714 #else
2715 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2716 {
2717 	NAPI_GRO_CB(skb)->flush |= flush;
2718 }
2719 #endif
2720 
2721 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2722 				  unsigned short type,
2723 				  const void *daddr, const void *saddr,
2724 				  unsigned int len)
2725 {
2726 	if (!dev->header_ops || !dev->header_ops->create)
2727 		return 0;
2728 
2729 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2730 }
2731 
2732 static inline int dev_parse_header(const struct sk_buff *skb,
2733 				   unsigned char *haddr)
2734 {
2735 	const struct net_device *dev = skb->dev;
2736 
2737 	if (!dev->header_ops || !dev->header_ops->parse)
2738 		return 0;
2739 	return dev->header_ops->parse(skb, haddr);
2740 }
2741 
2742 /* ll_header must have at least hard_header_len allocated */
2743 static inline bool dev_validate_header(const struct net_device *dev,
2744 				       char *ll_header, int len)
2745 {
2746 	if (likely(len >= dev->hard_header_len))
2747 		return true;
2748 	if (len < dev->min_header_len)
2749 		return false;
2750 
2751 	if (capable(CAP_SYS_RAWIO)) {
2752 		memset(ll_header + len, 0, dev->hard_header_len - len);
2753 		return true;
2754 	}
2755 
2756 	if (dev->header_ops && dev->header_ops->validate)
2757 		return dev->header_ops->validate(ll_header, len);
2758 
2759 	return false;
2760 }
2761 
2762 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2763 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2764 static inline int unregister_gifconf(unsigned int family)
2765 {
2766 	return register_gifconf(family, NULL);
2767 }
2768 
2769 #ifdef CONFIG_NET_FLOW_LIMIT
2770 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2771 struct sd_flow_limit {
2772 	u64			count;
2773 	unsigned int		num_buckets;
2774 	unsigned int		history_head;
2775 	u16			history[FLOW_LIMIT_HISTORY];
2776 	u8			buckets[];
2777 };
2778 
2779 extern int netdev_flow_limit_table_len;
2780 #endif /* CONFIG_NET_FLOW_LIMIT */
2781 
2782 /*
2783  * Incoming packets are placed on per-CPU queues
2784  */
2785 struct softnet_data {
2786 	struct list_head	poll_list;
2787 	struct sk_buff_head	process_queue;
2788 
2789 	/* stats */
2790 	unsigned int		processed;
2791 	unsigned int		time_squeeze;
2792 	unsigned int		received_rps;
2793 #ifdef CONFIG_RPS
2794 	struct softnet_data	*rps_ipi_list;
2795 #endif
2796 #ifdef CONFIG_NET_FLOW_LIMIT
2797 	struct sd_flow_limit __rcu *flow_limit;
2798 #endif
2799 	struct Qdisc		*output_queue;
2800 	struct Qdisc		**output_queue_tailp;
2801 	struct sk_buff		*completion_queue;
2802 #ifdef CONFIG_XFRM_OFFLOAD
2803 	struct sk_buff_head	xfrm_backlog;
2804 #endif
2805 #ifdef CONFIG_RPS
2806 	/* input_queue_head should be written by cpu owning this struct,
2807 	 * and only read by other cpus. Worth using a cache line.
2808 	 */
2809 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2810 
2811 	/* Elements below can be accessed between CPUs for RPS/RFS */
2812 	call_single_data_t	csd ____cacheline_aligned_in_smp;
2813 	struct softnet_data	*rps_ipi_next;
2814 	unsigned int		cpu;
2815 	unsigned int		input_queue_tail;
2816 #endif
2817 	unsigned int		dropped;
2818 	struct sk_buff_head	input_pkt_queue;
2819 	struct napi_struct	backlog;
2820 
2821 };
2822 
2823 static inline void input_queue_head_incr(struct softnet_data *sd)
2824 {
2825 #ifdef CONFIG_RPS
2826 	sd->input_queue_head++;
2827 #endif
2828 }
2829 
2830 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2831 					      unsigned int *qtail)
2832 {
2833 #ifdef CONFIG_RPS
2834 	*qtail = ++sd->input_queue_tail;
2835 #endif
2836 }
2837 
2838 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2839 
2840 void __netif_schedule(struct Qdisc *q);
2841 void netif_schedule_queue(struct netdev_queue *txq);
2842 
2843 static inline void netif_tx_schedule_all(struct net_device *dev)
2844 {
2845 	unsigned int i;
2846 
2847 	for (i = 0; i < dev->num_tx_queues; i++)
2848 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2849 }
2850 
2851 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2852 {
2853 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2854 }
2855 
2856 /**
2857  *	netif_start_queue - allow transmit
2858  *	@dev: network device
2859  *
2860  *	Allow upper layers to call the device hard_start_xmit routine.
2861  */
2862 static inline void netif_start_queue(struct net_device *dev)
2863 {
2864 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2865 }
2866 
2867 static inline void netif_tx_start_all_queues(struct net_device *dev)
2868 {
2869 	unsigned int i;
2870 
2871 	for (i = 0; i < dev->num_tx_queues; i++) {
2872 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2873 		netif_tx_start_queue(txq);
2874 	}
2875 }
2876 
2877 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2878 
2879 /**
2880  *	netif_wake_queue - restart transmit
2881  *	@dev: network device
2882  *
2883  *	Allow upper layers to call the device hard_start_xmit routine.
2884  *	Used for flow control when transmit resources are available.
2885  */
2886 static inline void netif_wake_queue(struct net_device *dev)
2887 {
2888 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2889 }
2890 
2891 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2892 {
2893 	unsigned int i;
2894 
2895 	for (i = 0; i < dev->num_tx_queues; i++) {
2896 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2897 		netif_tx_wake_queue(txq);
2898 	}
2899 }
2900 
2901 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2902 {
2903 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2904 }
2905 
2906 /**
2907  *	netif_stop_queue - stop transmitted packets
2908  *	@dev: network device
2909  *
2910  *	Stop upper layers calling the device hard_start_xmit routine.
2911  *	Used for flow control when transmit resources are unavailable.
2912  */
2913 static inline void netif_stop_queue(struct net_device *dev)
2914 {
2915 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2916 }
2917 
2918 void netif_tx_stop_all_queues(struct net_device *dev);
2919 
2920 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2921 {
2922 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2923 }
2924 
2925 /**
2926  *	netif_queue_stopped - test if transmit queue is flowblocked
2927  *	@dev: network device
2928  *
2929  *	Test if transmit queue on device is currently unable to send.
2930  */
2931 static inline bool netif_queue_stopped(const struct net_device *dev)
2932 {
2933 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2934 }
2935 
2936 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2937 {
2938 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2939 }
2940 
2941 static inline bool
2942 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2943 {
2944 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2945 }
2946 
2947 static inline bool
2948 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2949 {
2950 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2951 }
2952 
2953 /**
2954  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2955  *	@dev_queue: pointer to transmit queue
2956  *
2957  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2958  * to give appropriate hint to the CPU.
2959  */
2960 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2961 {
2962 #ifdef CONFIG_BQL
2963 	prefetchw(&dev_queue->dql.num_queued);
2964 #endif
2965 }
2966 
2967 /**
2968  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2969  *	@dev_queue: pointer to transmit queue
2970  *
2971  * BQL enabled drivers might use this helper in their TX completion path,
2972  * to give appropriate hint to the CPU.
2973  */
2974 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2975 {
2976 #ifdef CONFIG_BQL
2977 	prefetchw(&dev_queue->dql.limit);
2978 #endif
2979 }
2980 
2981 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2982 					unsigned int bytes)
2983 {
2984 #ifdef CONFIG_BQL
2985 	dql_queued(&dev_queue->dql, bytes);
2986 
2987 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2988 		return;
2989 
2990 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2991 
2992 	/*
2993 	 * The XOFF flag must be set before checking the dql_avail below,
2994 	 * because in netdev_tx_completed_queue we update the dql_completed
2995 	 * before checking the XOFF flag.
2996 	 */
2997 	smp_mb();
2998 
2999 	/* check again in case another CPU has just made room avail */
3000 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3001 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3002 #endif
3003 }
3004 
3005 /**
3006  * 	netdev_sent_queue - report the number of bytes queued to hardware
3007  * 	@dev: network device
3008  * 	@bytes: number of bytes queued to the hardware device queue
3009  *
3010  * 	Report the number of bytes queued for sending/completion to the network
3011  * 	device hardware queue. @bytes should be a good approximation and should
3012  * 	exactly match netdev_completed_queue() @bytes
3013  */
3014 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3015 {
3016 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3017 }
3018 
3019 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3020 					     unsigned int pkts, unsigned int bytes)
3021 {
3022 #ifdef CONFIG_BQL
3023 	if (unlikely(!bytes))
3024 		return;
3025 
3026 	dql_completed(&dev_queue->dql, bytes);
3027 
3028 	/*
3029 	 * Without the memory barrier there is a small possiblity that
3030 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3031 	 * be stopped forever
3032 	 */
3033 	smp_mb();
3034 
3035 	if (dql_avail(&dev_queue->dql) < 0)
3036 		return;
3037 
3038 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3039 		netif_schedule_queue(dev_queue);
3040 #endif
3041 }
3042 
3043 /**
3044  * 	netdev_completed_queue - report bytes and packets completed by device
3045  * 	@dev: network device
3046  * 	@pkts: actual number of packets sent over the medium
3047  * 	@bytes: actual number of bytes sent over the medium
3048  *
3049  * 	Report the number of bytes and packets transmitted by the network device
3050  * 	hardware queue over the physical medium, @bytes must exactly match the
3051  * 	@bytes amount passed to netdev_sent_queue()
3052  */
3053 static inline void netdev_completed_queue(struct net_device *dev,
3054 					  unsigned int pkts, unsigned int bytes)
3055 {
3056 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3057 }
3058 
3059 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3060 {
3061 #ifdef CONFIG_BQL
3062 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3063 	dql_reset(&q->dql);
3064 #endif
3065 }
3066 
3067 /**
3068  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3069  * 	@dev_queue: network device
3070  *
3071  * 	Reset the bytes and packet count of a network device and clear the
3072  * 	software flow control OFF bit for this network device
3073  */
3074 static inline void netdev_reset_queue(struct net_device *dev_queue)
3075 {
3076 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3077 }
3078 
3079 /**
3080  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3081  * 	@dev: network device
3082  * 	@queue_index: given tx queue index
3083  *
3084  * 	Returns 0 if given tx queue index >= number of device tx queues,
3085  * 	otherwise returns the originally passed tx queue index.
3086  */
3087 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3088 {
3089 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3090 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3091 				     dev->name, queue_index,
3092 				     dev->real_num_tx_queues);
3093 		return 0;
3094 	}
3095 
3096 	return queue_index;
3097 }
3098 
3099 /**
3100  *	netif_running - test if up
3101  *	@dev: network device
3102  *
3103  *	Test if the device has been brought up.
3104  */
3105 static inline bool netif_running(const struct net_device *dev)
3106 {
3107 	return test_bit(__LINK_STATE_START, &dev->state);
3108 }
3109 
3110 /*
3111  * Routines to manage the subqueues on a device.  We only need start,
3112  * stop, and a check if it's stopped.  All other device management is
3113  * done at the overall netdevice level.
3114  * Also test the device if we're multiqueue.
3115  */
3116 
3117 /**
3118  *	netif_start_subqueue - allow sending packets on subqueue
3119  *	@dev: network device
3120  *	@queue_index: sub queue index
3121  *
3122  * Start individual transmit queue of a device with multiple transmit queues.
3123  */
3124 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3125 {
3126 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3127 
3128 	netif_tx_start_queue(txq);
3129 }
3130 
3131 /**
3132  *	netif_stop_subqueue - stop sending packets on subqueue
3133  *	@dev: network device
3134  *	@queue_index: sub queue index
3135  *
3136  * Stop individual transmit queue of a device with multiple transmit queues.
3137  */
3138 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3139 {
3140 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3141 	netif_tx_stop_queue(txq);
3142 }
3143 
3144 /**
3145  *	netif_subqueue_stopped - test status of subqueue
3146  *	@dev: network device
3147  *	@queue_index: sub queue index
3148  *
3149  * Check individual transmit queue of a device with multiple transmit queues.
3150  */
3151 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3152 					    u16 queue_index)
3153 {
3154 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3155 
3156 	return netif_tx_queue_stopped(txq);
3157 }
3158 
3159 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3160 					  struct sk_buff *skb)
3161 {
3162 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3163 }
3164 
3165 /**
3166  *	netif_wake_subqueue - allow sending packets on subqueue
3167  *	@dev: network device
3168  *	@queue_index: sub queue index
3169  *
3170  * Resume individual transmit queue of a device with multiple transmit queues.
3171  */
3172 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3173 {
3174 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3175 
3176 	netif_tx_wake_queue(txq);
3177 }
3178 
3179 #ifdef CONFIG_XPS
3180 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3181 			u16 index);
3182 #else
3183 static inline int netif_set_xps_queue(struct net_device *dev,
3184 				      const struct cpumask *mask,
3185 				      u16 index)
3186 {
3187 	return 0;
3188 }
3189 #endif
3190 
3191 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3192 		  unsigned int num_tx_queues);
3193 
3194 /*
3195  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3196  * as a distribution range limit for the returned value.
3197  */
3198 static inline u16 skb_tx_hash(const struct net_device *dev,
3199 			      struct sk_buff *skb)
3200 {
3201 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3202 }
3203 
3204 /**
3205  *	netif_is_multiqueue - test if device has multiple transmit queues
3206  *	@dev: network device
3207  *
3208  * Check if device has multiple transmit queues
3209  */
3210 static inline bool netif_is_multiqueue(const struct net_device *dev)
3211 {
3212 	return dev->num_tx_queues > 1;
3213 }
3214 
3215 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3216 
3217 #ifdef CONFIG_SYSFS
3218 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3219 #else
3220 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3221 						unsigned int rxq)
3222 {
3223 	return 0;
3224 }
3225 #endif
3226 
3227 #ifdef CONFIG_SYSFS
3228 static inline unsigned int get_netdev_rx_queue_index(
3229 		struct netdev_rx_queue *queue)
3230 {
3231 	struct net_device *dev = queue->dev;
3232 	int index = queue - dev->_rx;
3233 
3234 	BUG_ON(index >= dev->num_rx_queues);
3235 	return index;
3236 }
3237 #endif
3238 
3239 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3240 int netif_get_num_default_rss_queues(void);
3241 
3242 enum skb_free_reason {
3243 	SKB_REASON_CONSUMED,
3244 	SKB_REASON_DROPPED,
3245 };
3246 
3247 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3248 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3249 
3250 /*
3251  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3252  * interrupt context or with hardware interrupts being disabled.
3253  * (in_irq() || irqs_disabled())
3254  *
3255  * We provide four helpers that can be used in following contexts :
3256  *
3257  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3258  *  replacing kfree_skb(skb)
3259  *
3260  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3261  *  Typically used in place of consume_skb(skb) in TX completion path
3262  *
3263  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3264  *  replacing kfree_skb(skb)
3265  *
3266  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3267  *  and consumed a packet. Used in place of consume_skb(skb)
3268  */
3269 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3270 {
3271 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3272 }
3273 
3274 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3275 {
3276 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3277 }
3278 
3279 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3280 {
3281 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3282 }
3283 
3284 static inline void dev_consume_skb_any(struct sk_buff *skb)
3285 {
3286 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3287 }
3288 
3289 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3290 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3291 int netif_rx(struct sk_buff *skb);
3292 int netif_rx_ni(struct sk_buff *skb);
3293 int netif_receive_skb(struct sk_buff *skb);
3294 int netif_receive_skb_core(struct sk_buff *skb);
3295 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3296 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3297 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3298 gro_result_t napi_gro_frags(struct napi_struct *napi);
3299 struct packet_offload *gro_find_receive_by_type(__be16 type);
3300 struct packet_offload *gro_find_complete_by_type(__be16 type);
3301 
3302 static inline void napi_free_frags(struct napi_struct *napi)
3303 {
3304 	kfree_skb(napi->skb);
3305 	napi->skb = NULL;
3306 }
3307 
3308 bool netdev_is_rx_handler_busy(struct net_device *dev);
3309 int netdev_rx_handler_register(struct net_device *dev,
3310 			       rx_handler_func_t *rx_handler,
3311 			       void *rx_handler_data);
3312 void netdev_rx_handler_unregister(struct net_device *dev);
3313 
3314 bool dev_valid_name(const char *name);
3315 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3316 int dev_ethtool(struct net *net, struct ifreq *);
3317 unsigned int dev_get_flags(const struct net_device *);
3318 int __dev_change_flags(struct net_device *, unsigned int flags);
3319 int dev_change_flags(struct net_device *, unsigned int);
3320 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3321 			unsigned int gchanges);
3322 int dev_change_name(struct net_device *, const char *);
3323 int dev_set_alias(struct net_device *, const char *, size_t);
3324 int dev_get_alias(const struct net_device *, char *, size_t);
3325 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3326 int __dev_set_mtu(struct net_device *, int);
3327 int dev_set_mtu(struct net_device *, int);
3328 void dev_set_group(struct net_device *, int);
3329 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3330 int dev_change_carrier(struct net_device *, bool new_carrier);
3331 int dev_get_phys_port_id(struct net_device *dev,
3332 			 struct netdev_phys_item_id *ppid);
3333 int dev_get_phys_port_name(struct net_device *dev,
3334 			   char *name, size_t len);
3335 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3336 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3337 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3338 				    struct netdev_queue *txq, int *ret);
3339 
3340 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3341 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3342 		      int fd, u32 flags);
3343 void __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3344 		     struct netdev_bpf *xdp);
3345 
3346 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3347 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3348 bool is_skb_forwardable(const struct net_device *dev,
3349 			const struct sk_buff *skb);
3350 
3351 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3352 					       struct sk_buff *skb)
3353 {
3354 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3355 	    unlikely(!is_skb_forwardable(dev, skb))) {
3356 		atomic_long_inc(&dev->rx_dropped);
3357 		kfree_skb(skb);
3358 		return NET_RX_DROP;
3359 	}
3360 
3361 	skb_scrub_packet(skb, true);
3362 	skb->priority = 0;
3363 	return 0;
3364 }
3365 
3366 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3367 
3368 extern int		netdev_budget;
3369 extern unsigned int	netdev_budget_usecs;
3370 
3371 /* Called by rtnetlink.c:rtnl_unlock() */
3372 void netdev_run_todo(void);
3373 
3374 /**
3375  *	dev_put - release reference to device
3376  *	@dev: network device
3377  *
3378  * Release reference to device to allow it to be freed.
3379  */
3380 static inline void dev_put(struct net_device *dev)
3381 {
3382 	this_cpu_dec(*dev->pcpu_refcnt);
3383 }
3384 
3385 /**
3386  *	dev_hold - get reference to device
3387  *	@dev: network device
3388  *
3389  * Hold reference to device to keep it from being freed.
3390  */
3391 static inline void dev_hold(struct net_device *dev)
3392 {
3393 	this_cpu_inc(*dev->pcpu_refcnt);
3394 }
3395 
3396 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3397  * and _off may be called from IRQ context, but it is caller
3398  * who is responsible for serialization of these calls.
3399  *
3400  * The name carrier is inappropriate, these functions should really be
3401  * called netif_lowerlayer_*() because they represent the state of any
3402  * kind of lower layer not just hardware media.
3403  */
3404 
3405 void linkwatch_init_dev(struct net_device *dev);
3406 void linkwatch_fire_event(struct net_device *dev);
3407 void linkwatch_forget_dev(struct net_device *dev);
3408 
3409 /**
3410  *	netif_carrier_ok - test if carrier present
3411  *	@dev: network device
3412  *
3413  * Check if carrier is present on device
3414  */
3415 static inline bool netif_carrier_ok(const struct net_device *dev)
3416 {
3417 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3418 }
3419 
3420 unsigned long dev_trans_start(struct net_device *dev);
3421 
3422 void __netdev_watchdog_up(struct net_device *dev);
3423 
3424 void netif_carrier_on(struct net_device *dev);
3425 
3426 void netif_carrier_off(struct net_device *dev);
3427 
3428 /**
3429  *	netif_dormant_on - mark device as dormant.
3430  *	@dev: network device
3431  *
3432  * Mark device as dormant (as per RFC2863).
3433  *
3434  * The dormant state indicates that the relevant interface is not
3435  * actually in a condition to pass packets (i.e., it is not 'up') but is
3436  * in a "pending" state, waiting for some external event.  For "on-
3437  * demand" interfaces, this new state identifies the situation where the
3438  * interface is waiting for events to place it in the up state.
3439  */
3440 static inline void netif_dormant_on(struct net_device *dev)
3441 {
3442 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3443 		linkwatch_fire_event(dev);
3444 }
3445 
3446 /**
3447  *	netif_dormant_off - set device as not dormant.
3448  *	@dev: network device
3449  *
3450  * Device is not in dormant state.
3451  */
3452 static inline void netif_dormant_off(struct net_device *dev)
3453 {
3454 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3455 		linkwatch_fire_event(dev);
3456 }
3457 
3458 /**
3459  *	netif_dormant - test if device is dormant
3460  *	@dev: network device
3461  *
3462  * Check if device is dormant.
3463  */
3464 static inline bool netif_dormant(const struct net_device *dev)
3465 {
3466 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3467 }
3468 
3469 
3470 /**
3471  *	netif_oper_up - test if device is operational
3472  *	@dev: network device
3473  *
3474  * Check if carrier is operational
3475  */
3476 static inline bool netif_oper_up(const struct net_device *dev)
3477 {
3478 	return (dev->operstate == IF_OPER_UP ||
3479 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3480 }
3481 
3482 /**
3483  *	netif_device_present - is device available or removed
3484  *	@dev: network device
3485  *
3486  * Check if device has not been removed from system.
3487  */
3488 static inline bool netif_device_present(struct net_device *dev)
3489 {
3490 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3491 }
3492 
3493 void netif_device_detach(struct net_device *dev);
3494 
3495 void netif_device_attach(struct net_device *dev);
3496 
3497 /*
3498  * Network interface message level settings
3499  */
3500 
3501 enum {
3502 	NETIF_MSG_DRV		= 0x0001,
3503 	NETIF_MSG_PROBE		= 0x0002,
3504 	NETIF_MSG_LINK		= 0x0004,
3505 	NETIF_MSG_TIMER		= 0x0008,
3506 	NETIF_MSG_IFDOWN	= 0x0010,
3507 	NETIF_MSG_IFUP		= 0x0020,
3508 	NETIF_MSG_RX_ERR	= 0x0040,
3509 	NETIF_MSG_TX_ERR	= 0x0080,
3510 	NETIF_MSG_TX_QUEUED	= 0x0100,
3511 	NETIF_MSG_INTR		= 0x0200,
3512 	NETIF_MSG_TX_DONE	= 0x0400,
3513 	NETIF_MSG_RX_STATUS	= 0x0800,
3514 	NETIF_MSG_PKTDATA	= 0x1000,
3515 	NETIF_MSG_HW		= 0x2000,
3516 	NETIF_MSG_WOL		= 0x4000,
3517 };
3518 
3519 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3520 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3521 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3522 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3523 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3524 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3525 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3526 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3527 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3528 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3529 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3530 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3531 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3532 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3533 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3534 
3535 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3536 {
3537 	/* use default */
3538 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3539 		return default_msg_enable_bits;
3540 	if (debug_value == 0)	/* no output */
3541 		return 0;
3542 	/* set low N bits */
3543 	return (1 << debug_value) - 1;
3544 }
3545 
3546 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3547 {
3548 	spin_lock(&txq->_xmit_lock);
3549 	txq->xmit_lock_owner = cpu;
3550 }
3551 
3552 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3553 {
3554 	__acquire(&txq->_xmit_lock);
3555 	return true;
3556 }
3557 
3558 static inline void __netif_tx_release(struct netdev_queue *txq)
3559 {
3560 	__release(&txq->_xmit_lock);
3561 }
3562 
3563 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3564 {
3565 	spin_lock_bh(&txq->_xmit_lock);
3566 	txq->xmit_lock_owner = smp_processor_id();
3567 }
3568 
3569 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3570 {
3571 	bool ok = spin_trylock(&txq->_xmit_lock);
3572 	if (likely(ok))
3573 		txq->xmit_lock_owner = smp_processor_id();
3574 	return ok;
3575 }
3576 
3577 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3578 {
3579 	txq->xmit_lock_owner = -1;
3580 	spin_unlock(&txq->_xmit_lock);
3581 }
3582 
3583 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3584 {
3585 	txq->xmit_lock_owner = -1;
3586 	spin_unlock_bh(&txq->_xmit_lock);
3587 }
3588 
3589 static inline void txq_trans_update(struct netdev_queue *txq)
3590 {
3591 	if (txq->xmit_lock_owner != -1)
3592 		txq->trans_start = jiffies;
3593 }
3594 
3595 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3596 static inline void netif_trans_update(struct net_device *dev)
3597 {
3598 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3599 
3600 	if (txq->trans_start != jiffies)
3601 		txq->trans_start = jiffies;
3602 }
3603 
3604 /**
3605  *	netif_tx_lock - grab network device transmit lock
3606  *	@dev: network device
3607  *
3608  * Get network device transmit lock
3609  */
3610 static inline void netif_tx_lock(struct net_device *dev)
3611 {
3612 	unsigned int i;
3613 	int cpu;
3614 
3615 	spin_lock(&dev->tx_global_lock);
3616 	cpu = smp_processor_id();
3617 	for (i = 0; i < dev->num_tx_queues; i++) {
3618 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3619 
3620 		/* We are the only thread of execution doing a
3621 		 * freeze, but we have to grab the _xmit_lock in
3622 		 * order to synchronize with threads which are in
3623 		 * the ->hard_start_xmit() handler and already
3624 		 * checked the frozen bit.
3625 		 */
3626 		__netif_tx_lock(txq, cpu);
3627 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3628 		__netif_tx_unlock(txq);
3629 	}
3630 }
3631 
3632 static inline void netif_tx_lock_bh(struct net_device *dev)
3633 {
3634 	local_bh_disable();
3635 	netif_tx_lock(dev);
3636 }
3637 
3638 static inline void netif_tx_unlock(struct net_device *dev)
3639 {
3640 	unsigned int i;
3641 
3642 	for (i = 0; i < dev->num_tx_queues; i++) {
3643 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3644 
3645 		/* No need to grab the _xmit_lock here.  If the
3646 		 * queue is not stopped for another reason, we
3647 		 * force a schedule.
3648 		 */
3649 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3650 		netif_schedule_queue(txq);
3651 	}
3652 	spin_unlock(&dev->tx_global_lock);
3653 }
3654 
3655 static inline void netif_tx_unlock_bh(struct net_device *dev)
3656 {
3657 	netif_tx_unlock(dev);
3658 	local_bh_enable();
3659 }
3660 
3661 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3662 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3663 		__netif_tx_lock(txq, cpu);		\
3664 	} else {					\
3665 		__netif_tx_acquire(txq);		\
3666 	}						\
3667 }
3668 
3669 #define HARD_TX_TRYLOCK(dev, txq)			\
3670 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3671 		__netif_tx_trylock(txq) :		\
3672 		__netif_tx_acquire(txq))
3673 
3674 #define HARD_TX_UNLOCK(dev, txq) {			\
3675 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3676 		__netif_tx_unlock(txq);			\
3677 	} else {					\
3678 		__netif_tx_release(txq);		\
3679 	}						\
3680 }
3681 
3682 static inline void netif_tx_disable(struct net_device *dev)
3683 {
3684 	unsigned int i;
3685 	int cpu;
3686 
3687 	local_bh_disable();
3688 	cpu = smp_processor_id();
3689 	for (i = 0; i < dev->num_tx_queues; i++) {
3690 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3691 
3692 		__netif_tx_lock(txq, cpu);
3693 		netif_tx_stop_queue(txq);
3694 		__netif_tx_unlock(txq);
3695 	}
3696 	local_bh_enable();
3697 }
3698 
3699 static inline void netif_addr_lock(struct net_device *dev)
3700 {
3701 	spin_lock(&dev->addr_list_lock);
3702 }
3703 
3704 static inline void netif_addr_lock_nested(struct net_device *dev)
3705 {
3706 	int subclass = SINGLE_DEPTH_NESTING;
3707 
3708 	if (dev->netdev_ops->ndo_get_lock_subclass)
3709 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3710 
3711 	spin_lock_nested(&dev->addr_list_lock, subclass);
3712 }
3713 
3714 static inline void netif_addr_lock_bh(struct net_device *dev)
3715 {
3716 	spin_lock_bh(&dev->addr_list_lock);
3717 }
3718 
3719 static inline void netif_addr_unlock(struct net_device *dev)
3720 {
3721 	spin_unlock(&dev->addr_list_lock);
3722 }
3723 
3724 static inline void netif_addr_unlock_bh(struct net_device *dev)
3725 {
3726 	spin_unlock_bh(&dev->addr_list_lock);
3727 }
3728 
3729 /*
3730  * dev_addrs walker. Should be used only for read access. Call with
3731  * rcu_read_lock held.
3732  */
3733 #define for_each_dev_addr(dev, ha) \
3734 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3735 
3736 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3737 
3738 void ether_setup(struct net_device *dev);
3739 
3740 /* Support for loadable net-drivers */
3741 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3742 				    unsigned char name_assign_type,
3743 				    void (*setup)(struct net_device *),
3744 				    unsigned int txqs, unsigned int rxqs);
3745 int dev_get_valid_name(struct net *net, struct net_device *dev,
3746 		       const char *name);
3747 
3748 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3749 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3750 
3751 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3752 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3753 			 count)
3754 
3755 int register_netdev(struct net_device *dev);
3756 void unregister_netdev(struct net_device *dev);
3757 
3758 /* General hardware address lists handling functions */
3759 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3760 		   struct netdev_hw_addr_list *from_list, int addr_len);
3761 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3762 		      struct netdev_hw_addr_list *from_list, int addr_len);
3763 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3764 		       struct net_device *dev,
3765 		       int (*sync)(struct net_device *, const unsigned char *),
3766 		       int (*unsync)(struct net_device *,
3767 				     const unsigned char *));
3768 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3769 			  struct net_device *dev,
3770 			  int (*unsync)(struct net_device *,
3771 					const unsigned char *));
3772 void __hw_addr_init(struct netdev_hw_addr_list *list);
3773 
3774 /* Functions used for device addresses handling */
3775 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3776 		 unsigned char addr_type);
3777 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3778 		 unsigned char addr_type);
3779 void dev_addr_flush(struct net_device *dev);
3780 int dev_addr_init(struct net_device *dev);
3781 
3782 /* Functions used for unicast addresses handling */
3783 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3784 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3785 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3786 int dev_uc_sync(struct net_device *to, struct net_device *from);
3787 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3788 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3789 void dev_uc_flush(struct net_device *dev);
3790 void dev_uc_init(struct net_device *dev);
3791 
3792 /**
3793  *  __dev_uc_sync - Synchonize device's unicast list
3794  *  @dev:  device to sync
3795  *  @sync: function to call if address should be added
3796  *  @unsync: function to call if address should be removed
3797  *
3798  *  Add newly added addresses to the interface, and release
3799  *  addresses that have been deleted.
3800  */
3801 static inline int __dev_uc_sync(struct net_device *dev,
3802 				int (*sync)(struct net_device *,
3803 					    const unsigned char *),
3804 				int (*unsync)(struct net_device *,
3805 					      const unsigned char *))
3806 {
3807 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3808 }
3809 
3810 /**
3811  *  __dev_uc_unsync - Remove synchronized addresses from device
3812  *  @dev:  device to sync
3813  *  @unsync: function to call if address should be removed
3814  *
3815  *  Remove all addresses that were added to the device by dev_uc_sync().
3816  */
3817 static inline void __dev_uc_unsync(struct net_device *dev,
3818 				   int (*unsync)(struct net_device *,
3819 						 const unsigned char *))
3820 {
3821 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3822 }
3823 
3824 /* Functions used for multicast addresses handling */
3825 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3826 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3827 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3828 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3829 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3830 int dev_mc_sync(struct net_device *to, struct net_device *from);
3831 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3832 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3833 void dev_mc_flush(struct net_device *dev);
3834 void dev_mc_init(struct net_device *dev);
3835 
3836 /**
3837  *  __dev_mc_sync - Synchonize device's multicast list
3838  *  @dev:  device to sync
3839  *  @sync: function to call if address should be added
3840  *  @unsync: function to call if address should be removed
3841  *
3842  *  Add newly added addresses to the interface, and release
3843  *  addresses that have been deleted.
3844  */
3845 static inline int __dev_mc_sync(struct net_device *dev,
3846 				int (*sync)(struct net_device *,
3847 					    const unsigned char *),
3848 				int (*unsync)(struct net_device *,
3849 					      const unsigned char *))
3850 {
3851 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3852 }
3853 
3854 /**
3855  *  __dev_mc_unsync - Remove synchronized addresses from device
3856  *  @dev:  device to sync
3857  *  @unsync: function to call if address should be removed
3858  *
3859  *  Remove all addresses that were added to the device by dev_mc_sync().
3860  */
3861 static inline void __dev_mc_unsync(struct net_device *dev,
3862 				   int (*unsync)(struct net_device *,
3863 						 const unsigned char *))
3864 {
3865 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3866 }
3867 
3868 /* Functions used for secondary unicast and multicast support */
3869 void dev_set_rx_mode(struct net_device *dev);
3870 void __dev_set_rx_mode(struct net_device *dev);
3871 int dev_set_promiscuity(struct net_device *dev, int inc);
3872 int dev_set_allmulti(struct net_device *dev, int inc);
3873 void netdev_state_change(struct net_device *dev);
3874 void netdev_notify_peers(struct net_device *dev);
3875 void netdev_features_change(struct net_device *dev);
3876 /* Load a device via the kmod */
3877 void dev_load(struct net *net, const char *name);
3878 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3879 					struct rtnl_link_stats64 *storage);
3880 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3881 			     const struct net_device_stats *netdev_stats);
3882 
3883 extern int		netdev_max_backlog;
3884 extern int		netdev_tstamp_prequeue;
3885 extern int		weight_p;
3886 extern int		dev_weight_rx_bias;
3887 extern int		dev_weight_tx_bias;
3888 extern int		dev_rx_weight;
3889 extern int		dev_tx_weight;
3890 
3891 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3892 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3893 						     struct list_head **iter);
3894 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3895 						     struct list_head **iter);
3896 
3897 /* iterate through upper list, must be called under RCU read lock */
3898 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3899 	for (iter = &(dev)->adj_list.upper, \
3900 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3901 	     updev; \
3902 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3903 
3904 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3905 				  int (*fn)(struct net_device *upper_dev,
3906 					    void *data),
3907 				  void *data);
3908 
3909 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3910 				  struct net_device *upper_dev);
3911 
3912 bool netdev_has_any_upper_dev(struct net_device *dev);
3913 
3914 void *netdev_lower_get_next_private(struct net_device *dev,
3915 				    struct list_head **iter);
3916 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3917 					struct list_head **iter);
3918 
3919 #define netdev_for_each_lower_private(dev, priv, iter) \
3920 	for (iter = (dev)->adj_list.lower.next, \
3921 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3922 	     priv; \
3923 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3924 
3925 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3926 	for (iter = &(dev)->adj_list.lower, \
3927 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3928 	     priv; \
3929 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3930 
3931 void *netdev_lower_get_next(struct net_device *dev,
3932 				struct list_head **iter);
3933 
3934 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3935 	for (iter = (dev)->adj_list.lower.next, \
3936 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3937 	     ldev; \
3938 	     ldev = netdev_lower_get_next(dev, &(iter)))
3939 
3940 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3941 					     struct list_head **iter);
3942 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3943 						 struct list_head **iter);
3944 
3945 int netdev_walk_all_lower_dev(struct net_device *dev,
3946 			      int (*fn)(struct net_device *lower_dev,
3947 					void *data),
3948 			      void *data);
3949 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3950 				  int (*fn)(struct net_device *lower_dev,
3951 					    void *data),
3952 				  void *data);
3953 
3954 void *netdev_adjacent_get_private(struct list_head *adj_list);
3955 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3956 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3957 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3958 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
3959 			  struct netlink_ext_ack *extack);
3960 int netdev_master_upper_dev_link(struct net_device *dev,
3961 				 struct net_device *upper_dev,
3962 				 void *upper_priv, void *upper_info,
3963 				 struct netlink_ext_ack *extack);
3964 void netdev_upper_dev_unlink(struct net_device *dev,
3965 			     struct net_device *upper_dev);
3966 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3967 void *netdev_lower_dev_get_private(struct net_device *dev,
3968 				   struct net_device *lower_dev);
3969 void netdev_lower_state_changed(struct net_device *lower_dev,
3970 				void *lower_state_info);
3971 
3972 /* RSS keys are 40 or 52 bytes long */
3973 #define NETDEV_RSS_KEY_LEN 52
3974 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3975 void netdev_rss_key_fill(void *buffer, size_t len);
3976 
3977 int dev_get_nest_level(struct net_device *dev);
3978 int skb_checksum_help(struct sk_buff *skb);
3979 int skb_crc32c_csum_help(struct sk_buff *skb);
3980 int skb_csum_hwoffload_help(struct sk_buff *skb,
3981 			    const netdev_features_t features);
3982 
3983 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3984 				  netdev_features_t features, bool tx_path);
3985 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3986 				    netdev_features_t features);
3987 
3988 struct netdev_bonding_info {
3989 	ifslave	slave;
3990 	ifbond	master;
3991 };
3992 
3993 struct netdev_notifier_bonding_info {
3994 	struct netdev_notifier_info info; /* must be first */
3995 	struct netdev_bonding_info  bonding_info;
3996 };
3997 
3998 void netdev_bonding_info_change(struct net_device *dev,
3999 				struct netdev_bonding_info *bonding_info);
4000 
4001 static inline
4002 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4003 {
4004 	return __skb_gso_segment(skb, features, true);
4005 }
4006 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4007 
4008 static inline bool can_checksum_protocol(netdev_features_t features,
4009 					 __be16 protocol)
4010 {
4011 	if (protocol == htons(ETH_P_FCOE))
4012 		return !!(features & NETIF_F_FCOE_CRC);
4013 
4014 	/* Assume this is an IP checksum (not SCTP CRC) */
4015 
4016 	if (features & NETIF_F_HW_CSUM) {
4017 		/* Can checksum everything */
4018 		return true;
4019 	}
4020 
4021 	switch (protocol) {
4022 	case htons(ETH_P_IP):
4023 		return !!(features & NETIF_F_IP_CSUM);
4024 	case htons(ETH_P_IPV6):
4025 		return !!(features & NETIF_F_IPV6_CSUM);
4026 	default:
4027 		return false;
4028 	}
4029 }
4030 
4031 #ifdef CONFIG_BUG
4032 void netdev_rx_csum_fault(struct net_device *dev);
4033 #else
4034 static inline void netdev_rx_csum_fault(struct net_device *dev)
4035 {
4036 }
4037 #endif
4038 /* rx skb timestamps */
4039 void net_enable_timestamp(void);
4040 void net_disable_timestamp(void);
4041 
4042 #ifdef CONFIG_PROC_FS
4043 int __init dev_proc_init(void);
4044 #else
4045 #define dev_proc_init() 0
4046 #endif
4047 
4048 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4049 					      struct sk_buff *skb, struct net_device *dev,
4050 					      bool more)
4051 {
4052 	skb->xmit_more = more ? 1 : 0;
4053 	return ops->ndo_start_xmit(skb, dev);
4054 }
4055 
4056 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4057 					    struct netdev_queue *txq, bool more)
4058 {
4059 	const struct net_device_ops *ops = dev->netdev_ops;
4060 	int rc;
4061 
4062 	rc = __netdev_start_xmit(ops, skb, dev, more);
4063 	if (rc == NETDEV_TX_OK)
4064 		txq_trans_update(txq);
4065 
4066 	return rc;
4067 }
4068 
4069 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4070 				const void *ns);
4071 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4072 				 const void *ns);
4073 
4074 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4075 {
4076 	return netdev_class_create_file_ns(class_attr, NULL);
4077 }
4078 
4079 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4080 {
4081 	netdev_class_remove_file_ns(class_attr, NULL);
4082 }
4083 
4084 extern const struct kobj_ns_type_operations net_ns_type_operations;
4085 
4086 const char *netdev_drivername(const struct net_device *dev);
4087 
4088 void linkwatch_run_queue(void);
4089 
4090 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4091 							  netdev_features_t f2)
4092 {
4093 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4094 		if (f1 & NETIF_F_HW_CSUM)
4095 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4096 		else
4097 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4098 	}
4099 
4100 	return f1 & f2;
4101 }
4102 
4103 static inline netdev_features_t netdev_get_wanted_features(
4104 	struct net_device *dev)
4105 {
4106 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4107 }
4108 netdev_features_t netdev_increment_features(netdev_features_t all,
4109 	netdev_features_t one, netdev_features_t mask);
4110 
4111 /* Allow TSO being used on stacked device :
4112  * Performing the GSO segmentation before last device
4113  * is a performance improvement.
4114  */
4115 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4116 							netdev_features_t mask)
4117 {
4118 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4119 }
4120 
4121 int __netdev_update_features(struct net_device *dev);
4122 void netdev_update_features(struct net_device *dev);
4123 void netdev_change_features(struct net_device *dev);
4124 
4125 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4126 					struct net_device *dev);
4127 
4128 netdev_features_t passthru_features_check(struct sk_buff *skb,
4129 					  struct net_device *dev,
4130 					  netdev_features_t features);
4131 netdev_features_t netif_skb_features(struct sk_buff *skb);
4132 
4133 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4134 {
4135 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4136 
4137 	/* check flags correspondence */
4138 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4139 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4140 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4141 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4142 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4143 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4144 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4145 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4146 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4147 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4148 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4149 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4150 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4151 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4152 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4153 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4154 	BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4155 
4156 	return (features & feature) == feature;
4157 }
4158 
4159 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4160 {
4161 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4162 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4163 }
4164 
4165 static inline bool netif_needs_gso(struct sk_buff *skb,
4166 				   netdev_features_t features)
4167 {
4168 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4169 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4170 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4171 }
4172 
4173 static inline void netif_set_gso_max_size(struct net_device *dev,
4174 					  unsigned int size)
4175 {
4176 	dev->gso_max_size = size;
4177 }
4178 
4179 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4180 					int pulled_hlen, u16 mac_offset,
4181 					int mac_len)
4182 {
4183 	skb->protocol = protocol;
4184 	skb->encapsulation = 1;
4185 	skb_push(skb, pulled_hlen);
4186 	skb_reset_transport_header(skb);
4187 	skb->mac_header = mac_offset;
4188 	skb->network_header = skb->mac_header + mac_len;
4189 	skb->mac_len = mac_len;
4190 }
4191 
4192 static inline bool netif_is_macsec(const struct net_device *dev)
4193 {
4194 	return dev->priv_flags & IFF_MACSEC;
4195 }
4196 
4197 static inline bool netif_is_macvlan(const struct net_device *dev)
4198 {
4199 	return dev->priv_flags & IFF_MACVLAN;
4200 }
4201 
4202 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4203 {
4204 	return dev->priv_flags & IFF_MACVLAN_PORT;
4205 }
4206 
4207 static inline bool netif_is_ipvlan(const struct net_device *dev)
4208 {
4209 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4210 }
4211 
4212 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4213 {
4214 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4215 }
4216 
4217 static inline bool netif_is_bond_master(const struct net_device *dev)
4218 {
4219 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4220 }
4221 
4222 static inline bool netif_is_bond_slave(const struct net_device *dev)
4223 {
4224 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4225 }
4226 
4227 static inline bool netif_supports_nofcs(struct net_device *dev)
4228 {
4229 	return dev->priv_flags & IFF_SUPP_NOFCS;
4230 }
4231 
4232 static inline bool netif_is_l3_master(const struct net_device *dev)
4233 {
4234 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4235 }
4236 
4237 static inline bool netif_is_l3_slave(const struct net_device *dev)
4238 {
4239 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4240 }
4241 
4242 static inline bool netif_is_bridge_master(const struct net_device *dev)
4243 {
4244 	return dev->priv_flags & IFF_EBRIDGE;
4245 }
4246 
4247 static inline bool netif_is_bridge_port(const struct net_device *dev)
4248 {
4249 	return dev->priv_flags & IFF_BRIDGE_PORT;
4250 }
4251 
4252 static inline bool netif_is_ovs_master(const struct net_device *dev)
4253 {
4254 	return dev->priv_flags & IFF_OPENVSWITCH;
4255 }
4256 
4257 static inline bool netif_is_ovs_port(const struct net_device *dev)
4258 {
4259 	return dev->priv_flags & IFF_OVS_DATAPATH;
4260 }
4261 
4262 static inline bool netif_is_team_master(const struct net_device *dev)
4263 {
4264 	return dev->priv_flags & IFF_TEAM;
4265 }
4266 
4267 static inline bool netif_is_team_port(const struct net_device *dev)
4268 {
4269 	return dev->priv_flags & IFF_TEAM_PORT;
4270 }
4271 
4272 static inline bool netif_is_lag_master(const struct net_device *dev)
4273 {
4274 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4275 }
4276 
4277 static inline bool netif_is_lag_port(const struct net_device *dev)
4278 {
4279 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4280 }
4281 
4282 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4283 {
4284 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4285 }
4286 
4287 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4288 static inline void netif_keep_dst(struct net_device *dev)
4289 {
4290 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4291 }
4292 
4293 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4294 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4295 {
4296 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4297 	return dev->priv_flags & IFF_MACSEC;
4298 }
4299 
4300 extern struct pernet_operations __net_initdata loopback_net_ops;
4301 
4302 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4303 
4304 /* netdev_printk helpers, similar to dev_printk */
4305 
4306 static inline const char *netdev_name(const struct net_device *dev)
4307 {
4308 	if (!dev->name[0] || strchr(dev->name, '%'))
4309 		return "(unnamed net_device)";
4310 	return dev->name;
4311 }
4312 
4313 static inline bool netdev_unregistering(const struct net_device *dev)
4314 {
4315 	return dev->reg_state == NETREG_UNREGISTERING;
4316 }
4317 
4318 static inline const char *netdev_reg_state(const struct net_device *dev)
4319 {
4320 	switch (dev->reg_state) {
4321 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4322 	case NETREG_REGISTERED: return "";
4323 	case NETREG_UNREGISTERING: return " (unregistering)";
4324 	case NETREG_UNREGISTERED: return " (unregistered)";
4325 	case NETREG_RELEASED: return " (released)";
4326 	case NETREG_DUMMY: return " (dummy)";
4327 	}
4328 
4329 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4330 	return " (unknown)";
4331 }
4332 
4333 __printf(3, 4)
4334 void netdev_printk(const char *level, const struct net_device *dev,
4335 		   const char *format, ...);
4336 __printf(2, 3)
4337 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4338 __printf(2, 3)
4339 void netdev_alert(const struct net_device *dev, const char *format, ...);
4340 __printf(2, 3)
4341 void netdev_crit(const struct net_device *dev, const char *format, ...);
4342 __printf(2, 3)
4343 void netdev_err(const struct net_device *dev, const char *format, ...);
4344 __printf(2, 3)
4345 void netdev_warn(const struct net_device *dev, const char *format, ...);
4346 __printf(2, 3)
4347 void netdev_notice(const struct net_device *dev, const char *format, ...);
4348 __printf(2, 3)
4349 void netdev_info(const struct net_device *dev, const char *format, ...);
4350 
4351 #define netdev_level_once(level, dev, fmt, ...)			\
4352 do {								\
4353 	static bool __print_once __read_mostly;			\
4354 								\
4355 	if (!__print_once) {					\
4356 		__print_once = true;				\
4357 		netdev_printk(level, dev, fmt, ##__VA_ARGS__);	\
4358 	}							\
4359 } while (0)
4360 
4361 #define netdev_emerg_once(dev, fmt, ...) \
4362 	netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4363 #define netdev_alert_once(dev, fmt, ...) \
4364 	netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4365 #define netdev_crit_once(dev, fmt, ...) \
4366 	netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4367 #define netdev_err_once(dev, fmt, ...) \
4368 	netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4369 #define netdev_warn_once(dev, fmt, ...) \
4370 	netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4371 #define netdev_notice_once(dev, fmt, ...) \
4372 	netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4373 #define netdev_info_once(dev, fmt, ...) \
4374 	netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4375 
4376 #define MODULE_ALIAS_NETDEV(device) \
4377 	MODULE_ALIAS("netdev-" device)
4378 
4379 #if defined(CONFIG_DYNAMIC_DEBUG)
4380 #define netdev_dbg(__dev, format, args...)			\
4381 do {								\
4382 	dynamic_netdev_dbg(__dev, format, ##args);		\
4383 } while (0)
4384 #elif defined(DEBUG)
4385 #define netdev_dbg(__dev, format, args...)			\
4386 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4387 #else
4388 #define netdev_dbg(__dev, format, args...)			\
4389 ({								\
4390 	if (0)							\
4391 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4392 })
4393 #endif
4394 
4395 #if defined(VERBOSE_DEBUG)
4396 #define netdev_vdbg	netdev_dbg
4397 #else
4398 
4399 #define netdev_vdbg(dev, format, args...)			\
4400 ({								\
4401 	if (0)							\
4402 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4403 	0;							\
4404 })
4405 #endif
4406 
4407 /*
4408  * netdev_WARN() acts like dev_printk(), but with the key difference
4409  * of using a WARN/WARN_ON to get the message out, including the
4410  * file/line information and a backtrace.
4411  */
4412 #define netdev_WARN(dev, format, args...)			\
4413 	WARN(1, "netdevice: %s%s: " format, netdev_name(dev),	\
4414 	     netdev_reg_state(dev), ##args)
4415 
4416 #define netdev_WARN_ONCE(dev, format, args...)				\
4417 	WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),	\
4418 		  netdev_reg_state(dev), ##args)
4419 
4420 /* netif printk helpers, similar to netdev_printk */
4421 
4422 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4423 do {					  			\
4424 	if (netif_msg_##type(priv))				\
4425 		netdev_printk(level, (dev), fmt, ##args);	\
4426 } while (0)
4427 
4428 #define netif_level(level, priv, type, dev, fmt, args...)	\
4429 do {								\
4430 	if (netif_msg_##type(priv))				\
4431 		netdev_##level(dev, fmt, ##args);		\
4432 } while (0)
4433 
4434 #define netif_emerg(priv, type, dev, fmt, args...)		\
4435 	netif_level(emerg, priv, type, dev, fmt, ##args)
4436 #define netif_alert(priv, type, dev, fmt, args...)		\
4437 	netif_level(alert, priv, type, dev, fmt, ##args)
4438 #define netif_crit(priv, type, dev, fmt, args...)		\
4439 	netif_level(crit, priv, type, dev, fmt, ##args)
4440 #define netif_err(priv, type, dev, fmt, args...)		\
4441 	netif_level(err, priv, type, dev, fmt, ##args)
4442 #define netif_warn(priv, type, dev, fmt, args...)		\
4443 	netif_level(warn, priv, type, dev, fmt, ##args)
4444 #define netif_notice(priv, type, dev, fmt, args...)		\
4445 	netif_level(notice, priv, type, dev, fmt, ##args)
4446 #define netif_info(priv, type, dev, fmt, args...)		\
4447 	netif_level(info, priv, type, dev, fmt, ##args)
4448 
4449 #if defined(CONFIG_DYNAMIC_DEBUG)
4450 #define netif_dbg(priv, type, netdev, format, args...)		\
4451 do {								\
4452 	if (netif_msg_##type(priv))				\
4453 		dynamic_netdev_dbg(netdev, format, ##args);	\
4454 } while (0)
4455 #elif defined(DEBUG)
4456 #define netif_dbg(priv, type, dev, format, args...)		\
4457 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4458 #else
4459 #define netif_dbg(priv, type, dev, format, args...)			\
4460 ({									\
4461 	if (0)								\
4462 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4463 	0;								\
4464 })
4465 #endif
4466 
4467 /* if @cond then downgrade to debug, else print at @level */
4468 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4469 	do {                                                              \
4470 		if (cond)                                                 \
4471 			netif_dbg(priv, type, netdev, fmt, ##args);       \
4472 		else                                                      \
4473 			netif_ ## level(priv, type, netdev, fmt, ##args); \
4474 	} while (0)
4475 
4476 #if defined(VERBOSE_DEBUG)
4477 #define netif_vdbg	netif_dbg
4478 #else
4479 #define netif_vdbg(priv, type, dev, format, args...)		\
4480 ({								\
4481 	if (0)							\
4482 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4483 	0;							\
4484 })
4485 #endif
4486 
4487 /*
4488  *	The list of packet types we will receive (as opposed to discard)
4489  *	and the routines to invoke.
4490  *
4491  *	Why 16. Because with 16 the only overlap we get on a hash of the
4492  *	low nibble of the protocol value is RARP/SNAP/X.25.
4493  *
4494  *		0800	IP
4495  *		0001	802.3
4496  *		0002	AX.25
4497  *		0004	802.2
4498  *		8035	RARP
4499  *		0005	SNAP
4500  *		0805	X.25
4501  *		0806	ARP
4502  *		8137	IPX
4503  *		0009	Localtalk
4504  *		86DD	IPv6
4505  */
4506 #define PTYPE_HASH_SIZE	(16)
4507 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4508 
4509 #endif	/* _LINUX_NETDEVICE_H */
4510