xref: /linux-6.15/include/linux/netdevice.h (revision bbb03029)
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/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41 
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #ifdef CONFIG_DCB
45 #include <net/dcbnl.h>
46 #endif
47 #include <net/netprio_cgroup.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_switch_tree;
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 } ____cacheline_aligned_in_smp;
691 
692 /*
693  * RX queue sysfs structures and functions.
694  */
695 struct rx_queue_attribute {
696 	struct attribute attr;
697 	ssize_t (*show)(struct netdev_rx_queue *queue,
698 	    struct rx_queue_attribute *attr, char *buf);
699 	ssize_t (*store)(struct netdev_rx_queue *queue,
700 	    struct rx_queue_attribute *attr, 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 /* These structures hold the attributes of qdisc and classifiers
775  * that are being passed to the netdevice through the setup_tc op.
776  */
777 enum {
778 	TC_SETUP_MQPRIO,
779 	TC_SETUP_CLSU32,
780 	TC_SETUP_CLSFLOWER,
781 	TC_SETUP_MATCHALL,
782 	TC_SETUP_CLSBPF,
783 };
784 
785 struct tc_cls_u32_offload;
786 
787 struct tc_to_netdev {
788 	unsigned int type;
789 	union {
790 		struct tc_cls_u32_offload *cls_u32;
791 		struct tc_cls_flower_offload *cls_flower;
792 		struct tc_cls_matchall_offload *cls_mall;
793 		struct tc_cls_bpf_offload *cls_bpf;
794 		struct tc_mqprio_qopt *mqprio;
795 	};
796 	bool egress_dev;
797 };
798 
799 /* These structures hold the attributes of xdp state that are being passed
800  * to the netdevice through the xdp op.
801  */
802 enum xdp_netdev_command {
803 	/* Set or clear a bpf program used in the earliest stages of packet
804 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
805 	 * is responsible for calling bpf_prog_put on any old progs that are
806 	 * stored. In case of error, the callee need not release the new prog
807 	 * reference, but on success it takes ownership and must bpf_prog_put
808 	 * when it is no longer used.
809 	 */
810 	XDP_SETUP_PROG,
811 	XDP_SETUP_PROG_HW,
812 	/* Check if a bpf program is set on the device.  The callee should
813 	 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
814 	 * is equivalent to XDP_ATTACHED_DRV.
815 	 */
816 	XDP_QUERY_PROG,
817 };
818 
819 struct netlink_ext_ack;
820 
821 struct netdev_xdp {
822 	enum xdp_netdev_command command;
823 	union {
824 		/* XDP_SETUP_PROG */
825 		struct {
826 			u32 flags;
827 			struct bpf_prog *prog;
828 			struct netlink_ext_ack *extack;
829 		};
830 		/* XDP_QUERY_PROG */
831 		struct {
832 			u8 prog_attached;
833 			u32 prog_id;
834 		};
835 	};
836 };
837 
838 #ifdef CONFIG_XFRM_OFFLOAD
839 struct xfrmdev_ops {
840 	int	(*xdo_dev_state_add) (struct xfrm_state *x);
841 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
842 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
843 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
844 				       struct xfrm_state *x);
845 };
846 #endif
847 
848 /*
849  * This structure defines the management hooks for network devices.
850  * The following hooks can be defined; unless noted otherwise, they are
851  * optional and can be filled with a null pointer.
852  *
853  * int (*ndo_init)(struct net_device *dev);
854  *     This function is called once when a network device is registered.
855  *     The network device can use this for any late stage initialization
856  *     or semantic validation. It can fail with an error code which will
857  *     be propagated back to register_netdev.
858  *
859  * void (*ndo_uninit)(struct net_device *dev);
860  *     This function is called when device is unregistered or when registration
861  *     fails. It is not called if init fails.
862  *
863  * int (*ndo_open)(struct net_device *dev);
864  *     This function is called when a network device transitions to the up
865  *     state.
866  *
867  * int (*ndo_stop)(struct net_device *dev);
868  *     This function is called when a network device transitions to the down
869  *     state.
870  *
871  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
872  *                               struct net_device *dev);
873  *	Called when a packet needs to be transmitted.
874  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
875  *	the queue before that can happen; it's for obsolete devices and weird
876  *	corner cases, but the stack really does a non-trivial amount
877  *	of useless work if you return NETDEV_TX_BUSY.
878  *	Required; cannot be NULL.
879  *
880  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
881  *					   struct net_device *dev
882  *					   netdev_features_t features);
883  *	Called by core transmit path to determine if device is capable of
884  *	performing offload operations on a given packet. This is to give
885  *	the device an opportunity to implement any restrictions that cannot
886  *	be otherwise expressed by feature flags. The check is called with
887  *	the set of features that the stack has calculated and it returns
888  *	those the driver believes to be appropriate.
889  *
890  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
891  *                         void *accel_priv, select_queue_fallback_t fallback);
892  *	Called to decide which queue to use when device supports multiple
893  *	transmit queues.
894  *
895  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
896  *	This function is called to allow device receiver to make
897  *	changes to configuration when multicast or promiscuous is enabled.
898  *
899  * void (*ndo_set_rx_mode)(struct net_device *dev);
900  *	This function is called device changes address list filtering.
901  *	If driver handles unicast address filtering, it should set
902  *	IFF_UNICAST_FLT in its priv_flags.
903  *
904  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
905  *	This function  is called when the Media Access Control address
906  *	needs to be changed. If this interface is not defined, the
907  *	MAC address can not be changed.
908  *
909  * int (*ndo_validate_addr)(struct net_device *dev);
910  *	Test if Media Access Control address is valid for the device.
911  *
912  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
913  *	Called when a user requests an ioctl which can't be handled by
914  *	the generic interface code. If not defined ioctls return
915  *	not supported error code.
916  *
917  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
918  *	Used to set network devices bus interface parameters. This interface
919  *	is retained for legacy reasons; new devices should use the bus
920  *	interface (PCI) for low level management.
921  *
922  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
923  *	Called when a user wants to change the Maximum Transfer Unit
924  *	of a device.
925  *
926  * void (*ndo_tx_timeout)(struct net_device *dev);
927  *	Callback used when the transmitter has not made any progress
928  *	for dev->watchdog ticks.
929  *
930  * void (*ndo_get_stats64)(struct net_device *dev,
931  *                         struct rtnl_link_stats64 *storage);
932  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
933  *	Called when a user wants to get the network device usage
934  *	statistics. Drivers must do one of the following:
935  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
936  *	   rtnl_link_stats64 structure passed by the caller.
937  *	2. Define @ndo_get_stats to update a net_device_stats structure
938  *	   (which should normally be dev->stats) and return a pointer to
939  *	   it. The structure may be changed asynchronously only if each
940  *	   field is written atomically.
941  *	3. Update dev->stats asynchronously and atomically, and define
942  *	   neither operation.
943  *
944  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
945  *	Return true if this device supports offload stats of this attr_id.
946  *
947  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
948  *	void *attr_data)
949  *	Get statistics for offload operations by attr_id. Write it into the
950  *	attr_data pointer.
951  *
952  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
953  *	If device supports VLAN filtering this function is called when a
954  *	VLAN id is registered.
955  *
956  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
957  *	If device supports VLAN filtering this function is called when a
958  *	VLAN id is unregistered.
959  *
960  * void (*ndo_poll_controller)(struct net_device *dev);
961  *
962  *	SR-IOV management functions.
963  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
964  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
965  *			  u8 qos, __be16 proto);
966  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
967  *			  int max_tx_rate);
968  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
969  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
970  * int (*ndo_get_vf_config)(struct net_device *dev,
971  *			    int vf, struct ifla_vf_info *ivf);
972  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
973  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
974  *			  struct nlattr *port[]);
975  *
976  *      Enable or disable the VF ability to query its RSS Redirection Table and
977  *      Hash Key. This is needed since on some devices VF share this information
978  *      with PF and querying it may introduce a theoretical security risk.
979  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
980  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
981  * int (*ndo_setup_tc)(struct net_device *dev, u32 handle, u32 chain_index,
982  *		       __be16 protocol, struct tc_to_netdev *tc);
983  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
984  *	This is always called from the stack with the rtnl lock held and netif
985  *	tx queues stopped. This allows the netdevice to perform queue
986  *	management safely.
987  *
988  *	Fiber Channel over Ethernet (FCoE) offload functions.
989  * int (*ndo_fcoe_enable)(struct net_device *dev);
990  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
991  *	so the underlying device can perform whatever needed configuration or
992  *	initialization to support acceleration of FCoE traffic.
993  *
994  * int (*ndo_fcoe_disable)(struct net_device *dev);
995  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
996  *	so the underlying device can perform whatever needed clean-ups to
997  *	stop supporting acceleration of FCoE traffic.
998  *
999  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1000  *			     struct scatterlist *sgl, unsigned int sgc);
1001  *	Called when the FCoE Initiator wants to initialize an I/O that
1002  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1003  *	perform necessary setup and returns 1 to indicate the device is set up
1004  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1005  *
1006  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1007  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1008  *	indicated by the FC exchange id 'xid', so the underlying device can
1009  *	clean up and reuse resources for later DDP requests.
1010  *
1011  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1012  *			      struct scatterlist *sgl, unsigned int sgc);
1013  *	Called when the FCoE Target wants to initialize an I/O that
1014  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1015  *	perform necessary setup and returns 1 to indicate the device is set up
1016  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1017  *
1018  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1019  *			       struct netdev_fcoe_hbainfo *hbainfo);
1020  *	Called when the FCoE Protocol stack wants information on the underlying
1021  *	device. This information is utilized by the FCoE protocol stack to
1022  *	register attributes with Fiber Channel management service as per the
1023  *	FC-GS Fabric Device Management Information(FDMI) specification.
1024  *
1025  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1026  *	Called when the underlying device wants to override default World Wide
1027  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1028  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1029  *	protocol stack to use.
1030  *
1031  *	RFS acceleration.
1032  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1033  *			    u16 rxq_index, u32 flow_id);
1034  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1035  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1036  *	Return the filter ID on success, or a negative error code.
1037  *
1038  *	Slave management functions (for bridge, bonding, etc).
1039  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1040  *	Called to make another netdev an underling.
1041  *
1042  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1043  *	Called to release previously enslaved netdev.
1044  *
1045  *      Feature/offload setting functions.
1046  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1047  *		netdev_features_t features);
1048  *	Adjusts the requested feature flags according to device-specific
1049  *	constraints, and returns the resulting flags. Must not modify
1050  *	the device state.
1051  *
1052  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1053  *	Called to update device configuration to new features. Passed
1054  *	feature set might be less than what was returned by ndo_fix_features()).
1055  *	Must return >0 or -errno if it changed dev->features itself.
1056  *
1057  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1058  *		      struct net_device *dev,
1059  *		      const unsigned char *addr, u16 vid, u16 flags)
1060  *	Adds an FDB entry to dev for addr.
1061  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1062  *		      struct net_device *dev,
1063  *		      const unsigned char *addr, u16 vid)
1064  *	Deletes the FDB entry from dev coresponding to addr.
1065  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1066  *		       struct net_device *dev, struct net_device *filter_dev,
1067  *		       int *idx)
1068  *	Used to add FDB entries to dump requests. Implementers should add
1069  *	entries to skb and update idx with the number of entries.
1070  *
1071  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1072  *			     u16 flags)
1073  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1074  *			     struct net_device *dev, u32 filter_mask,
1075  *			     int nlflags)
1076  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1077  *			     u16 flags);
1078  *
1079  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1080  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1081  *	which do not represent real hardware may define this to allow their
1082  *	userspace components to manage their virtual carrier state. Devices
1083  *	that determine carrier state from physical hardware properties (eg
1084  *	network cables) or protocol-dependent mechanisms (eg
1085  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1086  *
1087  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1088  *			       struct netdev_phys_item_id *ppid);
1089  *	Called to get ID of physical port of this device. If driver does
1090  *	not implement this, it is assumed that the hw is not able to have
1091  *	multiple net devices on single physical port.
1092  *
1093  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1094  *			      struct udp_tunnel_info *ti);
1095  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1096  *	address family that a UDP tunnel is listnening to. It is called only
1097  *	when a new port starts listening. The operation is protected by the
1098  *	RTNL.
1099  *
1100  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1101  *			      struct udp_tunnel_info *ti);
1102  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1103  *	address family that the UDP tunnel is not listening to anymore. The
1104  *	operation is protected by the RTNL.
1105  *
1106  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1107  *				 struct net_device *dev)
1108  *	Called by upper layer devices to accelerate switching or other
1109  *	station functionality into hardware. 'pdev is the lowerdev
1110  *	to use for the offload and 'dev' is the net device that will
1111  *	back the offload. Returns a pointer to the private structure
1112  *	the upper layer will maintain.
1113  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1114  *	Called by upper layer device to delete the station created
1115  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1116  *	the station and priv is the structure returned by the add
1117  *	operation.
1118  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1119  *			     int queue_index, u32 maxrate);
1120  *	Called when a user wants to set a max-rate limitation of specific
1121  *	TX queue.
1122  * int (*ndo_get_iflink)(const struct net_device *dev);
1123  *	Called to get the iflink value of this device.
1124  * void (*ndo_change_proto_down)(struct net_device *dev,
1125  *				 bool proto_down);
1126  *	This function is used to pass protocol port error state information
1127  *	to the switch driver. The switch driver can react to the proto_down
1128  *      by doing a phys down on the associated switch port.
1129  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1130  *	This function is used to get egress tunnel information for given skb.
1131  *	This is useful for retrieving outer tunnel header parameters while
1132  *	sampling packet.
1133  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1134  *	This function is used to specify the headroom that the skb must
1135  *	consider when allocation skb during packet reception. Setting
1136  *	appropriate rx headroom value allows avoiding skb head copy on
1137  *	forward. Setting a negative value resets the rx headroom to the
1138  *	default value.
1139  * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1140  *	This function is used to set or query state related to XDP on the
1141  *	netdevice. See definition of enum xdp_netdev_command for details.
1142  * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1143  *	This function is used to submit a XDP packet for transmit on a
1144  *	netdevice.
1145  * void (*ndo_xdp_flush)(struct net_device *dev);
1146  *	This function is used to inform the driver to flush a paticular
1147  *	xpd tx queue. Must be called on same CPU as xdp_xmit.
1148  */
1149 struct net_device_ops {
1150 	int			(*ndo_init)(struct net_device *dev);
1151 	void			(*ndo_uninit)(struct net_device *dev);
1152 	int			(*ndo_open)(struct net_device *dev);
1153 	int			(*ndo_stop)(struct net_device *dev);
1154 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1155 						  struct net_device *dev);
1156 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1157 						      struct net_device *dev,
1158 						      netdev_features_t features);
1159 	u16			(*ndo_select_queue)(struct net_device *dev,
1160 						    struct sk_buff *skb,
1161 						    void *accel_priv,
1162 						    select_queue_fallback_t fallback);
1163 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1164 						       int flags);
1165 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1166 	int			(*ndo_set_mac_address)(struct net_device *dev,
1167 						       void *addr);
1168 	int			(*ndo_validate_addr)(struct net_device *dev);
1169 	int			(*ndo_do_ioctl)(struct net_device *dev,
1170 					        struct ifreq *ifr, int cmd);
1171 	int			(*ndo_set_config)(struct net_device *dev,
1172 					          struct ifmap *map);
1173 	int			(*ndo_change_mtu)(struct net_device *dev,
1174 						  int new_mtu);
1175 	int			(*ndo_neigh_setup)(struct net_device *dev,
1176 						   struct neigh_parms *);
1177 	void			(*ndo_tx_timeout) (struct net_device *dev);
1178 
1179 	void			(*ndo_get_stats64)(struct net_device *dev,
1180 						   struct rtnl_link_stats64 *storage);
1181 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1182 	int			(*ndo_get_offload_stats)(int attr_id,
1183 							 const struct net_device *dev,
1184 							 void *attr_data);
1185 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1186 
1187 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1188 						       __be16 proto, u16 vid);
1189 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1190 						        __be16 proto, u16 vid);
1191 #ifdef CONFIG_NET_POLL_CONTROLLER
1192 	void                    (*ndo_poll_controller)(struct net_device *dev);
1193 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1194 						     struct netpoll_info *info);
1195 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1196 #endif
1197 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1198 						  int queue, u8 *mac);
1199 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1200 						   int queue, u16 vlan,
1201 						   u8 qos, __be16 proto);
1202 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1203 						   int vf, int min_tx_rate,
1204 						   int max_tx_rate);
1205 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1206 						       int vf, bool setting);
1207 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1208 						    int vf, bool setting);
1209 	int			(*ndo_get_vf_config)(struct net_device *dev,
1210 						     int vf,
1211 						     struct ifla_vf_info *ivf);
1212 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1213 							 int vf, int link_state);
1214 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1215 						    int vf,
1216 						    struct ifla_vf_stats
1217 						    *vf_stats);
1218 	int			(*ndo_set_vf_port)(struct net_device *dev,
1219 						   int vf,
1220 						   struct nlattr *port[]);
1221 	int			(*ndo_get_vf_port)(struct net_device *dev,
1222 						   int vf, struct sk_buff *skb);
1223 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1224 						   int vf, u64 guid,
1225 						   int guid_type);
1226 	int			(*ndo_set_vf_rss_query_en)(
1227 						   struct net_device *dev,
1228 						   int vf, bool setting);
1229 	int			(*ndo_setup_tc)(struct net_device *dev,
1230 						u32 handle, u32 chain_index,
1231 						__be16 protocol,
1232 						struct tc_to_netdev *tc);
1233 #if IS_ENABLED(CONFIG_FCOE)
1234 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1235 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1236 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1237 						      u16 xid,
1238 						      struct scatterlist *sgl,
1239 						      unsigned int sgc);
1240 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1241 						     u16 xid);
1242 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1243 						       u16 xid,
1244 						       struct scatterlist *sgl,
1245 						       unsigned int sgc);
1246 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1247 							struct netdev_fcoe_hbainfo *hbainfo);
1248 #endif
1249 
1250 #if IS_ENABLED(CONFIG_LIBFCOE)
1251 #define NETDEV_FCOE_WWNN 0
1252 #define NETDEV_FCOE_WWPN 1
1253 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1254 						    u64 *wwn, int type);
1255 #endif
1256 
1257 #ifdef CONFIG_RFS_ACCEL
1258 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1259 						     const struct sk_buff *skb,
1260 						     u16 rxq_index,
1261 						     u32 flow_id);
1262 #endif
1263 	int			(*ndo_add_slave)(struct net_device *dev,
1264 						 struct net_device *slave_dev);
1265 	int			(*ndo_del_slave)(struct net_device *dev,
1266 						 struct net_device *slave_dev);
1267 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1268 						    netdev_features_t features);
1269 	int			(*ndo_set_features)(struct net_device *dev,
1270 						    netdev_features_t features);
1271 	int			(*ndo_neigh_construct)(struct net_device *dev,
1272 						       struct neighbour *n);
1273 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1274 						     struct neighbour *n);
1275 
1276 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1277 					       struct nlattr *tb[],
1278 					       struct net_device *dev,
1279 					       const unsigned char *addr,
1280 					       u16 vid,
1281 					       u16 flags);
1282 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1283 					       struct nlattr *tb[],
1284 					       struct net_device *dev,
1285 					       const unsigned char *addr,
1286 					       u16 vid);
1287 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1288 						struct netlink_callback *cb,
1289 						struct net_device *dev,
1290 						struct net_device *filter_dev,
1291 						int *idx);
1292 
1293 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1294 						      struct nlmsghdr *nlh,
1295 						      u16 flags);
1296 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1297 						      u32 pid, u32 seq,
1298 						      struct net_device *dev,
1299 						      u32 filter_mask,
1300 						      int nlflags);
1301 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1302 						      struct nlmsghdr *nlh,
1303 						      u16 flags);
1304 	int			(*ndo_change_carrier)(struct net_device *dev,
1305 						      bool new_carrier);
1306 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1307 							struct netdev_phys_item_id *ppid);
1308 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1309 							  char *name, size_t len);
1310 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1311 						      struct udp_tunnel_info *ti);
1312 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1313 						      struct udp_tunnel_info *ti);
1314 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1315 							struct net_device *dev);
1316 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1317 							void *priv);
1318 
1319 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1320 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1321 						      int queue_index,
1322 						      u32 maxrate);
1323 	int			(*ndo_get_iflink)(const struct net_device *dev);
1324 	int			(*ndo_change_proto_down)(struct net_device *dev,
1325 							 bool proto_down);
1326 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1327 						       struct sk_buff *skb);
1328 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1329 						       int needed_headroom);
1330 	int			(*ndo_xdp)(struct net_device *dev,
1331 					   struct netdev_xdp *xdp);
1332 	int			(*ndo_xdp_xmit)(struct net_device *dev,
1333 						struct xdp_buff *xdp);
1334 	void			(*ndo_xdp_flush)(struct net_device *dev);
1335 };
1336 
1337 /**
1338  * enum net_device_priv_flags - &struct net_device priv_flags
1339  *
1340  * These are the &struct net_device, they are only set internally
1341  * by drivers and used in the kernel. These flags are invisible to
1342  * userspace; this means that the order of these flags can change
1343  * during any kernel release.
1344  *
1345  * You should have a pretty good reason to be extending these flags.
1346  *
1347  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1348  * @IFF_EBRIDGE: Ethernet bridging device
1349  * @IFF_BONDING: bonding master or slave
1350  * @IFF_ISATAP: ISATAP interface (RFC4214)
1351  * @IFF_WAN_HDLC: WAN HDLC device
1352  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1353  *	release skb->dst
1354  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1355  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1356  * @IFF_MACVLAN_PORT: device used as macvlan port
1357  * @IFF_BRIDGE_PORT: device used as bridge port
1358  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1359  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1360  * @IFF_UNICAST_FLT: Supports unicast filtering
1361  * @IFF_TEAM_PORT: device used as team port
1362  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1363  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1364  *	change when it's running
1365  * @IFF_MACVLAN: Macvlan device
1366  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1367  *	underlying stacked devices
1368  * @IFF_IPVLAN_MASTER: IPvlan master device
1369  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1370  * @IFF_L3MDEV_MASTER: device is an L3 master device
1371  * @IFF_NO_QUEUE: device can run without qdisc attached
1372  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1373  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1374  * @IFF_TEAM: device is a team device
1375  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1376  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1377  *	entity (i.e. the master device for bridged veth)
1378  * @IFF_MACSEC: device is a MACsec device
1379  */
1380 enum netdev_priv_flags {
1381 	IFF_802_1Q_VLAN			= 1<<0,
1382 	IFF_EBRIDGE			= 1<<1,
1383 	IFF_BONDING			= 1<<2,
1384 	IFF_ISATAP			= 1<<3,
1385 	IFF_WAN_HDLC			= 1<<4,
1386 	IFF_XMIT_DST_RELEASE		= 1<<5,
1387 	IFF_DONT_BRIDGE			= 1<<6,
1388 	IFF_DISABLE_NETPOLL		= 1<<7,
1389 	IFF_MACVLAN_PORT		= 1<<8,
1390 	IFF_BRIDGE_PORT			= 1<<9,
1391 	IFF_OVS_DATAPATH		= 1<<10,
1392 	IFF_TX_SKB_SHARING		= 1<<11,
1393 	IFF_UNICAST_FLT			= 1<<12,
1394 	IFF_TEAM_PORT			= 1<<13,
1395 	IFF_SUPP_NOFCS			= 1<<14,
1396 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1397 	IFF_MACVLAN			= 1<<16,
1398 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1399 	IFF_IPVLAN_MASTER		= 1<<18,
1400 	IFF_IPVLAN_SLAVE		= 1<<19,
1401 	IFF_L3MDEV_MASTER		= 1<<20,
1402 	IFF_NO_QUEUE			= 1<<21,
1403 	IFF_OPENVSWITCH			= 1<<22,
1404 	IFF_L3MDEV_SLAVE		= 1<<23,
1405 	IFF_TEAM			= 1<<24,
1406 	IFF_RXFH_CONFIGURED		= 1<<25,
1407 	IFF_PHONY_HEADROOM		= 1<<26,
1408 	IFF_MACSEC			= 1<<27,
1409 };
1410 
1411 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1412 #define IFF_EBRIDGE			IFF_EBRIDGE
1413 #define IFF_BONDING			IFF_BONDING
1414 #define IFF_ISATAP			IFF_ISATAP
1415 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1416 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1417 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1418 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1419 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1420 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1421 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1422 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1423 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1424 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1425 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1426 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1427 #define IFF_MACVLAN			IFF_MACVLAN
1428 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1429 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1430 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1431 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1432 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1433 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1434 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1435 #define IFF_TEAM			IFF_TEAM
1436 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1437 #define IFF_MACSEC			IFF_MACSEC
1438 
1439 /**
1440  *	struct net_device - The DEVICE structure.
1441  *
1442  *	Actually, this whole structure is a big mistake.  It mixes I/O
1443  *	data with strictly "high-level" data, and it has to know about
1444  *	almost every data structure used in the INET module.
1445  *
1446  *	@name:	This is the first field of the "visible" part of this structure
1447  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1448  *		of the interface.
1449  *
1450  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1451  *	@ifalias:	SNMP alias
1452  *	@mem_end:	Shared memory end
1453  *	@mem_start:	Shared memory start
1454  *	@base_addr:	Device I/O address
1455  *	@irq:		Device IRQ number
1456  *
1457  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1458  *
1459  *	@state:		Generic network queuing layer state, see netdev_state_t
1460  *	@dev_list:	The global list of network devices
1461  *	@napi_list:	List entry used for polling NAPI devices
1462  *	@unreg_list:	List entry  when we are unregistering the
1463  *			device; see the function unregister_netdev
1464  *	@close_list:	List entry used when we are closing the device
1465  *	@ptype_all:     Device-specific packet handlers for all protocols
1466  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1467  *
1468  *	@adj_list:	Directly linked devices, like slaves for bonding
1469  *	@features:	Currently active device features
1470  *	@hw_features:	User-changeable features
1471  *
1472  *	@wanted_features:	User-requested features
1473  *	@vlan_features:		Mask of features inheritable by VLAN devices
1474  *
1475  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1476  *				This field indicates what encapsulation
1477  *				offloads the hardware is capable of doing,
1478  *				and drivers will need to set them appropriately.
1479  *
1480  *	@mpls_features:	Mask of features inheritable by MPLS
1481  *
1482  *	@ifindex:	interface index
1483  *	@group:		The group the device belongs to
1484  *
1485  *	@stats:		Statistics struct, which was left as a legacy, use
1486  *			rtnl_link_stats64 instead
1487  *
1488  *	@rx_dropped:	Dropped packets by core network,
1489  *			do not use this in drivers
1490  *	@tx_dropped:	Dropped packets by core network,
1491  *			do not use this in drivers
1492  *	@rx_nohandler:	nohandler dropped packets by core network on
1493  *			inactive devices, do not use this in drivers
1494  *
1495  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1496  *				instead of ioctl,
1497  *				see <net/iw_handler.h> for details.
1498  *	@wireless_data:	Instance data managed by the core of wireless extensions
1499  *
1500  *	@netdev_ops:	Includes several pointers to callbacks,
1501  *			if one wants to override the ndo_*() functions
1502  *	@ethtool_ops:	Management operations
1503  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1504  *			discovery handling. Necessary for e.g. 6LoWPAN.
1505  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1506  *			of Layer 2 headers.
1507  *
1508  *	@flags:		Interface flags (a la BSD)
1509  *	@priv_flags:	Like 'flags' but invisible to userspace,
1510  *			see if.h for the definitions
1511  *	@gflags:	Global flags ( kept as legacy )
1512  *	@padded:	How much padding added by alloc_netdev()
1513  *	@operstate:	RFC2863 operstate
1514  *	@link_mode:	Mapping policy to operstate
1515  *	@if_port:	Selectable AUI, TP, ...
1516  *	@dma:		DMA channel
1517  *	@mtu:		Interface MTU value
1518  *	@min_mtu:	Interface Minimum MTU value
1519  *	@max_mtu:	Interface Maximum MTU value
1520  *	@type:		Interface hardware type
1521  *	@hard_header_len: Maximum hardware header length.
1522  *	@min_header_len:  Minimum hardware header length
1523  *
1524  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1525  *			  cases can this be guaranteed
1526  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1527  *			  cases can this be guaranteed. Some cases also use
1528  *			  LL_MAX_HEADER instead to allocate the skb
1529  *
1530  *	interface address info:
1531  *
1532  * 	@perm_addr:		Permanent hw address
1533  * 	@addr_assign_type:	Hw address assignment type
1534  * 	@addr_len:		Hardware address length
1535  *	@neigh_priv_len:	Used in neigh_alloc()
1536  * 	@dev_id:		Used to differentiate devices that share
1537  * 				the same link layer address
1538  * 	@dev_port:		Used to differentiate devices that share
1539  * 				the same function
1540  *	@addr_list_lock:	XXX: need comments on this one
1541  *	@uc_promisc:		Counter that indicates promiscuous mode
1542  *				has been enabled due to the need to listen to
1543  *				additional unicast addresses in a device that
1544  *				does not implement ndo_set_rx_mode()
1545  *	@uc:			unicast mac addresses
1546  *	@mc:			multicast mac addresses
1547  *	@dev_addrs:		list of device hw addresses
1548  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1549  *	@promiscuity:		Number of times the NIC is told to work in
1550  *				promiscuous mode; if it becomes 0 the NIC will
1551  *				exit promiscuous mode
1552  *	@allmulti:		Counter, enables or disables allmulticast mode
1553  *
1554  *	@vlan_info:	VLAN info
1555  *	@dsa_ptr:	dsa specific data
1556  *	@tipc_ptr:	TIPC specific data
1557  *	@atalk_ptr:	AppleTalk link
1558  *	@ip_ptr:	IPv4 specific data
1559  *	@dn_ptr:	DECnet specific data
1560  *	@ip6_ptr:	IPv6 specific data
1561  *	@ax25_ptr:	AX.25 specific data
1562  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1563  *
1564  *	@dev_addr:	Hw address (before bcast,
1565  *			because most packets are unicast)
1566  *
1567  *	@_rx:			Array of RX queues
1568  *	@num_rx_queues:		Number of RX queues
1569  *				allocated at register_netdev() time
1570  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1571  *
1572  *	@rx_handler:		handler for received packets
1573  *	@rx_handler_data: 	XXX: need comments on this one
1574  *	@ingress_queue:		XXX: need comments on this one
1575  *	@broadcast:		hw bcast address
1576  *
1577  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1578  *			indexed by RX queue number. Assigned by driver.
1579  *			This must only be set if the ndo_rx_flow_steer
1580  *			operation is defined
1581  *	@index_hlist:		Device index hash chain
1582  *
1583  *	@_tx:			Array of TX queues
1584  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1585  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1586  *	@qdisc:			Root qdisc from userspace point of view
1587  *	@tx_queue_len:		Max frames per queue allowed
1588  *	@tx_global_lock: 	XXX: need comments on this one
1589  *
1590  *	@xps_maps:	XXX: need comments on this one
1591  *
1592  *	@watchdog_timeo:	Represents the timeout that is used by
1593  *				the watchdog (see dev_watchdog())
1594  *	@watchdog_timer:	List of timers
1595  *
1596  *	@pcpu_refcnt:		Number of references to this device
1597  *	@todo_list:		Delayed register/unregister
1598  *	@link_watch_list:	XXX: need comments on this one
1599  *
1600  *	@reg_state:		Register/unregister state machine
1601  *	@dismantle:		Device is going to be freed
1602  *	@rtnl_link_state:	This enum represents the phases of creating
1603  *				a new link
1604  *
1605  *	@needs_free_netdev:	Should unregister perform free_netdev?
1606  *	@priv_destructor:	Called from unregister
1607  *	@npinfo:		XXX: need comments on this one
1608  * 	@nd_net:		Network namespace this network device is inside
1609  *
1610  * 	@ml_priv:	Mid-layer private
1611  * 	@lstats:	Loopback statistics
1612  * 	@tstats:	Tunnel statistics
1613  * 	@dstats:	Dummy statistics
1614  * 	@vstats:	Virtual ethernet statistics
1615  *
1616  *	@garp_port:	GARP
1617  *	@mrp_port:	MRP
1618  *
1619  *	@dev:		Class/net/name entry
1620  *	@sysfs_groups:	Space for optional device, statistics and wireless
1621  *			sysfs groups
1622  *
1623  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1624  *	@rtnl_link_ops:	Rtnl_link_ops
1625  *
1626  *	@gso_max_size:	Maximum size of generic segmentation offload
1627  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1628  *			NIC for GSO
1629  *
1630  *	@dcbnl_ops:	Data Center Bridging netlink ops
1631  *	@num_tc:	Number of traffic classes in the net device
1632  *	@tc_to_txq:	XXX: need comments on this one
1633  *	@prio_tc_map:	XXX: need comments on this one
1634  *
1635  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1636  *
1637  *	@priomap:	XXX: need comments on this one
1638  *	@phydev:	Physical device may attach itself
1639  *			for hardware timestamping
1640  *
1641  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1642  *	@qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1643  *
1644  *	@proto_down:	protocol port state information can be sent to the
1645  *			switch driver and used to set the phys state of the
1646  *			switch port.
1647  *
1648  *	FIXME: cleanup struct net_device such that network protocol info
1649  *	moves out.
1650  */
1651 
1652 struct net_device {
1653 	char			name[IFNAMSIZ];
1654 	struct hlist_node	name_hlist;
1655 	char 			*ifalias;
1656 	/*
1657 	 *	I/O specific fields
1658 	 *	FIXME: Merge these and struct ifmap into one
1659 	 */
1660 	unsigned long		mem_end;
1661 	unsigned long		mem_start;
1662 	unsigned long		base_addr;
1663 	int			irq;
1664 
1665 	atomic_t		carrier_changes;
1666 
1667 	/*
1668 	 *	Some hardware also needs these fields (state,dev_list,
1669 	 *	napi_list,unreg_list,close_list) but they are not
1670 	 *	part of the usual set specified in Space.c.
1671 	 */
1672 
1673 	unsigned long		state;
1674 
1675 	struct list_head	dev_list;
1676 	struct list_head	napi_list;
1677 	struct list_head	unreg_list;
1678 	struct list_head	close_list;
1679 	struct list_head	ptype_all;
1680 	struct list_head	ptype_specific;
1681 
1682 	struct {
1683 		struct list_head upper;
1684 		struct list_head lower;
1685 	} adj_list;
1686 
1687 	netdev_features_t	features;
1688 	netdev_features_t	hw_features;
1689 	netdev_features_t	wanted_features;
1690 	netdev_features_t	vlan_features;
1691 	netdev_features_t	hw_enc_features;
1692 	netdev_features_t	mpls_features;
1693 	netdev_features_t	gso_partial_features;
1694 
1695 	int			ifindex;
1696 	int			group;
1697 
1698 	struct net_device_stats	stats;
1699 
1700 	atomic_long_t		rx_dropped;
1701 	atomic_long_t		tx_dropped;
1702 	atomic_long_t		rx_nohandler;
1703 
1704 #ifdef CONFIG_WIRELESS_EXT
1705 	const struct iw_handler_def *wireless_handlers;
1706 	struct iw_public_data	*wireless_data;
1707 #endif
1708 	const struct net_device_ops *netdev_ops;
1709 	const struct ethtool_ops *ethtool_ops;
1710 #ifdef CONFIG_NET_SWITCHDEV
1711 	const struct switchdev_ops *switchdev_ops;
1712 #endif
1713 #ifdef CONFIG_NET_L3_MASTER_DEV
1714 	const struct l3mdev_ops	*l3mdev_ops;
1715 #endif
1716 #if IS_ENABLED(CONFIG_IPV6)
1717 	const struct ndisc_ops *ndisc_ops;
1718 #endif
1719 
1720 #ifdef CONFIG_XFRM
1721 	const struct xfrmdev_ops *xfrmdev_ops;
1722 #endif
1723 
1724 	const struct header_ops *header_ops;
1725 
1726 	unsigned int		flags;
1727 	unsigned int		priv_flags;
1728 
1729 	unsigned short		gflags;
1730 	unsigned short		padded;
1731 
1732 	unsigned char		operstate;
1733 	unsigned char		link_mode;
1734 
1735 	unsigned char		if_port;
1736 	unsigned char		dma;
1737 
1738 	unsigned int		mtu;
1739 	unsigned int		min_mtu;
1740 	unsigned int		max_mtu;
1741 	unsigned short		type;
1742 	unsigned short		hard_header_len;
1743 	unsigned char		min_header_len;
1744 
1745 	unsigned short		needed_headroom;
1746 	unsigned short		needed_tailroom;
1747 
1748 	/* Interface address info. */
1749 	unsigned char		perm_addr[MAX_ADDR_LEN];
1750 	unsigned char		addr_assign_type;
1751 	unsigned char		addr_len;
1752 	unsigned short		neigh_priv_len;
1753 	unsigned short          dev_id;
1754 	unsigned short          dev_port;
1755 	spinlock_t		addr_list_lock;
1756 	unsigned char		name_assign_type;
1757 	bool			uc_promisc;
1758 	struct netdev_hw_addr_list	uc;
1759 	struct netdev_hw_addr_list	mc;
1760 	struct netdev_hw_addr_list	dev_addrs;
1761 
1762 #ifdef CONFIG_SYSFS
1763 	struct kset		*queues_kset;
1764 #endif
1765 	unsigned int		promiscuity;
1766 	unsigned int		allmulti;
1767 
1768 
1769 	/* Protocol-specific pointers */
1770 
1771 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1772 	struct vlan_info __rcu	*vlan_info;
1773 #endif
1774 #if IS_ENABLED(CONFIG_NET_DSA)
1775 	struct dsa_switch_tree	*dsa_ptr;
1776 #endif
1777 #if IS_ENABLED(CONFIG_TIPC)
1778 	struct tipc_bearer __rcu *tipc_ptr;
1779 #endif
1780 	void 			*atalk_ptr;
1781 	struct in_device __rcu	*ip_ptr;
1782 	struct dn_dev __rcu     *dn_ptr;
1783 	struct inet6_dev __rcu	*ip6_ptr;
1784 	void			*ax25_ptr;
1785 	struct wireless_dev	*ieee80211_ptr;
1786 	struct wpan_dev		*ieee802154_ptr;
1787 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1788 	struct mpls_dev __rcu	*mpls_ptr;
1789 #endif
1790 
1791 /*
1792  * Cache lines mostly used on receive path (including eth_type_trans())
1793  */
1794 	/* Interface address info used in eth_type_trans() */
1795 	unsigned char		*dev_addr;
1796 
1797 #ifdef CONFIG_SYSFS
1798 	struct netdev_rx_queue	*_rx;
1799 
1800 	unsigned int		num_rx_queues;
1801 	unsigned int		real_num_rx_queues;
1802 #endif
1803 
1804 	struct bpf_prog __rcu	*xdp_prog;
1805 	unsigned long		gro_flush_timeout;
1806 	rx_handler_func_t __rcu	*rx_handler;
1807 	void __rcu		*rx_handler_data;
1808 
1809 #ifdef CONFIG_NET_CLS_ACT
1810 	struct tcf_proto __rcu  *ingress_cl_list;
1811 #endif
1812 	struct netdev_queue __rcu *ingress_queue;
1813 #ifdef CONFIG_NETFILTER_INGRESS
1814 	struct nf_hook_entry __rcu *nf_hooks_ingress;
1815 #endif
1816 
1817 	unsigned char		broadcast[MAX_ADDR_LEN];
1818 #ifdef CONFIG_RFS_ACCEL
1819 	struct cpu_rmap		*rx_cpu_rmap;
1820 #endif
1821 	struct hlist_node	index_hlist;
1822 
1823 /*
1824  * Cache lines mostly used on transmit path
1825  */
1826 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1827 	unsigned int		num_tx_queues;
1828 	unsigned int		real_num_tx_queues;
1829 	struct Qdisc		*qdisc;
1830 #ifdef CONFIG_NET_SCHED
1831 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1832 #endif
1833 	unsigned int		tx_queue_len;
1834 	spinlock_t		tx_global_lock;
1835 	int			watchdog_timeo;
1836 
1837 #ifdef CONFIG_XPS
1838 	struct xps_dev_maps __rcu *xps_maps;
1839 #endif
1840 #ifdef CONFIG_NET_CLS_ACT
1841 	struct tcf_proto __rcu  *egress_cl_list;
1842 #endif
1843 
1844 	/* These may be needed for future network-power-down code. */
1845 	struct timer_list	watchdog_timer;
1846 
1847 	int __percpu		*pcpu_refcnt;
1848 	struct list_head	todo_list;
1849 
1850 	struct list_head	link_watch_list;
1851 
1852 	enum { NETREG_UNINITIALIZED=0,
1853 	       NETREG_REGISTERED,	/* completed register_netdevice */
1854 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1855 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1856 	       NETREG_RELEASED,		/* called free_netdev */
1857 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1858 	} reg_state:8;
1859 
1860 	bool dismantle;
1861 
1862 	enum {
1863 		RTNL_LINK_INITIALIZED,
1864 		RTNL_LINK_INITIALIZING,
1865 	} rtnl_link_state:16;
1866 
1867 	bool needs_free_netdev;
1868 	void (*priv_destructor)(struct net_device *dev);
1869 
1870 #ifdef CONFIG_NETPOLL
1871 	struct netpoll_info __rcu	*npinfo;
1872 #endif
1873 
1874 	possible_net_t			nd_net;
1875 
1876 	/* mid-layer private */
1877 	union {
1878 		void					*ml_priv;
1879 		struct pcpu_lstats __percpu		*lstats;
1880 		struct pcpu_sw_netstats __percpu	*tstats;
1881 		struct pcpu_dstats __percpu		*dstats;
1882 		struct pcpu_vstats __percpu		*vstats;
1883 	};
1884 
1885 #if IS_ENABLED(CONFIG_GARP)
1886 	struct garp_port __rcu	*garp_port;
1887 #endif
1888 #if IS_ENABLED(CONFIG_MRP)
1889 	struct mrp_port __rcu	*mrp_port;
1890 #endif
1891 
1892 	struct device		dev;
1893 	const struct attribute_group *sysfs_groups[4];
1894 	const struct attribute_group *sysfs_rx_queue_group;
1895 
1896 	const struct rtnl_link_ops *rtnl_link_ops;
1897 
1898 	/* for setting kernel sock attribute on TCP connection setup */
1899 #define GSO_MAX_SIZE		65536
1900 	unsigned int		gso_max_size;
1901 #define GSO_MAX_SEGS		65535
1902 	u16			gso_max_segs;
1903 
1904 #ifdef CONFIG_DCB
1905 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1906 #endif
1907 	u8			num_tc;
1908 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1909 	u8			prio_tc_map[TC_BITMASK + 1];
1910 
1911 #if IS_ENABLED(CONFIG_FCOE)
1912 	unsigned int		fcoe_ddp_xid;
1913 #endif
1914 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1915 	struct netprio_map __rcu *priomap;
1916 #endif
1917 	struct phy_device	*phydev;
1918 	struct lock_class_key	*qdisc_tx_busylock;
1919 	struct lock_class_key	*qdisc_running_key;
1920 	bool			proto_down;
1921 };
1922 #define to_net_dev(d) container_of(d, struct net_device, dev)
1923 
1924 static inline bool netif_elide_gro(const struct net_device *dev)
1925 {
1926 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1927 		return true;
1928 	return false;
1929 }
1930 
1931 #define	NETDEV_ALIGN		32
1932 
1933 static inline
1934 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1935 {
1936 	return dev->prio_tc_map[prio & TC_BITMASK];
1937 }
1938 
1939 static inline
1940 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1941 {
1942 	if (tc >= dev->num_tc)
1943 		return -EINVAL;
1944 
1945 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1946 	return 0;
1947 }
1948 
1949 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1950 void netdev_reset_tc(struct net_device *dev);
1951 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1952 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1953 
1954 static inline
1955 int netdev_get_num_tc(struct net_device *dev)
1956 {
1957 	return dev->num_tc;
1958 }
1959 
1960 static inline
1961 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1962 					 unsigned int index)
1963 {
1964 	return &dev->_tx[index];
1965 }
1966 
1967 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1968 						    const struct sk_buff *skb)
1969 {
1970 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1971 }
1972 
1973 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1974 					    void (*f)(struct net_device *,
1975 						      struct netdev_queue *,
1976 						      void *),
1977 					    void *arg)
1978 {
1979 	unsigned int i;
1980 
1981 	for (i = 0; i < dev->num_tx_queues; i++)
1982 		f(dev, &dev->_tx[i], arg);
1983 }
1984 
1985 #define netdev_lockdep_set_classes(dev)				\
1986 {								\
1987 	static struct lock_class_key qdisc_tx_busylock_key;	\
1988 	static struct lock_class_key qdisc_running_key;		\
1989 	static struct lock_class_key qdisc_xmit_lock_key;	\
1990 	static struct lock_class_key dev_addr_list_lock_key;	\
1991 	unsigned int i;						\
1992 								\
1993 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
1994 	(dev)->qdisc_running_key = &qdisc_running_key;		\
1995 	lockdep_set_class(&(dev)->addr_list_lock,		\
1996 			  &dev_addr_list_lock_key); 		\
1997 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
1998 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
1999 				  &qdisc_xmit_lock_key);	\
2000 }
2001 
2002 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2003 				    struct sk_buff *skb,
2004 				    void *accel_priv);
2005 
2006 /* returns the headroom that the master device needs to take in account
2007  * when forwarding to this dev
2008  */
2009 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2010 {
2011 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2012 }
2013 
2014 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2015 {
2016 	if (dev->netdev_ops->ndo_set_rx_headroom)
2017 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2018 }
2019 
2020 /* set the device rx headroom to the dev's default */
2021 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2022 {
2023 	netdev_set_rx_headroom(dev, -1);
2024 }
2025 
2026 /*
2027  * Net namespace inlines
2028  */
2029 static inline
2030 struct net *dev_net(const struct net_device *dev)
2031 {
2032 	return read_pnet(&dev->nd_net);
2033 }
2034 
2035 static inline
2036 void dev_net_set(struct net_device *dev, struct net *net)
2037 {
2038 	write_pnet(&dev->nd_net, net);
2039 }
2040 
2041 /**
2042  *	netdev_priv - access network device private data
2043  *	@dev: network device
2044  *
2045  * Get network device private data
2046  */
2047 static inline void *netdev_priv(const struct net_device *dev)
2048 {
2049 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2050 }
2051 
2052 /* Set the sysfs physical device reference for the network logical device
2053  * if set prior to registration will cause a symlink during initialization.
2054  */
2055 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2056 
2057 /* Set the sysfs device type for the network logical device to allow
2058  * fine-grained identification of different network device types. For
2059  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2060  */
2061 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2062 
2063 /* Default NAPI poll() weight
2064  * Device drivers are strongly advised to not use bigger value
2065  */
2066 #define NAPI_POLL_WEIGHT 64
2067 
2068 /**
2069  *	netif_napi_add - initialize a NAPI context
2070  *	@dev:  network device
2071  *	@napi: NAPI context
2072  *	@poll: polling function
2073  *	@weight: default weight
2074  *
2075  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2076  * *any* of the other NAPI-related functions.
2077  */
2078 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2079 		    int (*poll)(struct napi_struct *, int), int weight);
2080 
2081 /**
2082  *	netif_tx_napi_add - initialize a NAPI context
2083  *	@dev:  network device
2084  *	@napi: NAPI context
2085  *	@poll: polling function
2086  *	@weight: default weight
2087  *
2088  * This variant of netif_napi_add() should be used from drivers using NAPI
2089  * to exclusively poll a TX queue.
2090  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2091  */
2092 static inline void netif_tx_napi_add(struct net_device *dev,
2093 				     struct napi_struct *napi,
2094 				     int (*poll)(struct napi_struct *, int),
2095 				     int weight)
2096 {
2097 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2098 	netif_napi_add(dev, napi, poll, weight);
2099 }
2100 
2101 /**
2102  *  netif_napi_del - remove a NAPI context
2103  *  @napi: NAPI context
2104  *
2105  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2106  */
2107 void netif_napi_del(struct napi_struct *napi);
2108 
2109 struct napi_gro_cb {
2110 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2111 	void	*frag0;
2112 
2113 	/* Length of frag0. */
2114 	unsigned int frag0_len;
2115 
2116 	/* This indicates where we are processing relative to skb->data. */
2117 	int	data_offset;
2118 
2119 	/* This is non-zero if the packet cannot be merged with the new skb. */
2120 	u16	flush;
2121 
2122 	/* Save the IP ID here and check when we get to the transport layer */
2123 	u16	flush_id;
2124 
2125 	/* Number of segments aggregated. */
2126 	u16	count;
2127 
2128 	/* Start offset for remote checksum offload */
2129 	u16	gro_remcsum_start;
2130 
2131 	/* jiffies when first packet was created/queued */
2132 	unsigned long age;
2133 
2134 	/* Used in ipv6_gro_receive() and foo-over-udp */
2135 	u16	proto;
2136 
2137 	/* This is non-zero if the packet may be of the same flow. */
2138 	u8	same_flow:1;
2139 
2140 	/* Used in tunnel GRO receive */
2141 	u8	encap_mark:1;
2142 
2143 	/* GRO checksum is valid */
2144 	u8	csum_valid:1;
2145 
2146 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2147 	u8	csum_cnt:3;
2148 
2149 	/* Free the skb? */
2150 	u8	free:2;
2151 #define NAPI_GRO_FREE		  1
2152 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2153 
2154 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2155 	u8	is_ipv6:1;
2156 
2157 	/* Used in GRE, set in fou/gue_gro_receive */
2158 	u8	is_fou:1;
2159 
2160 	/* Used to determine if flush_id can be ignored */
2161 	u8	is_atomic:1;
2162 
2163 	/* Number of gro_receive callbacks this packet already went through */
2164 	u8 recursion_counter:4;
2165 
2166 	/* 1 bit hole */
2167 
2168 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2169 	__wsum	csum;
2170 
2171 	/* used in skb_gro_receive() slow path */
2172 	struct sk_buff *last;
2173 };
2174 
2175 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2176 
2177 #define GRO_RECURSION_LIMIT 15
2178 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2179 {
2180 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2181 }
2182 
2183 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2184 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2185 						struct sk_buff **head,
2186 						struct sk_buff *skb)
2187 {
2188 	if (unlikely(gro_recursion_inc_test(skb))) {
2189 		NAPI_GRO_CB(skb)->flush |= 1;
2190 		return NULL;
2191 	}
2192 
2193 	return cb(head, skb);
2194 }
2195 
2196 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2197 					     struct sk_buff *);
2198 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2199 						   struct sock *sk,
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(sk, head, skb);
2209 }
2210 
2211 struct packet_type {
2212 	__be16			type;	/* This is really htons(ether_type). */
2213 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2214 	int			(*func) (struct sk_buff *,
2215 					 struct net_device *,
2216 					 struct packet_type *,
2217 					 struct net_device *);
2218 	bool			(*id_match)(struct packet_type *ptype,
2219 					    struct sock *sk);
2220 	void			*af_packet_priv;
2221 	struct list_head	list;
2222 };
2223 
2224 struct offload_callbacks {
2225 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2226 						netdev_features_t features);
2227 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2228 						 struct sk_buff *skb);
2229 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2230 };
2231 
2232 struct packet_offload {
2233 	__be16			 type;	/* This is really htons(ether_type). */
2234 	u16			 priority;
2235 	struct offload_callbacks callbacks;
2236 	struct list_head	 list;
2237 };
2238 
2239 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2240 struct pcpu_sw_netstats {
2241 	u64     rx_packets;
2242 	u64     rx_bytes;
2243 	u64     tx_packets;
2244 	u64     tx_bytes;
2245 	struct u64_stats_sync   syncp;
2246 };
2247 
2248 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2249 ({									\
2250 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2251 	if (pcpu_stats)	{						\
2252 		int __cpu;						\
2253 		for_each_possible_cpu(__cpu) {				\
2254 			typeof(type) *stat;				\
2255 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2256 			u64_stats_init(&stat->syncp);			\
2257 		}							\
2258 	}								\
2259 	pcpu_stats;							\
2260 })
2261 
2262 #define netdev_alloc_pcpu_stats(type)					\
2263 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2264 
2265 enum netdev_lag_tx_type {
2266 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2267 	NETDEV_LAG_TX_TYPE_RANDOM,
2268 	NETDEV_LAG_TX_TYPE_BROADCAST,
2269 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2270 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2271 	NETDEV_LAG_TX_TYPE_HASH,
2272 };
2273 
2274 struct netdev_lag_upper_info {
2275 	enum netdev_lag_tx_type tx_type;
2276 };
2277 
2278 struct netdev_lag_lower_state_info {
2279 	u8 link_up : 1,
2280 	   tx_enabled : 1;
2281 };
2282 
2283 #include <linux/notifier.h>
2284 
2285 /* netdevice notifier chain. Please remember to update the rtnetlink
2286  * notification exclusion list in rtnetlink_event() when adding new
2287  * types.
2288  */
2289 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2290 #define NETDEV_DOWN	0x0002
2291 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2292 				   detected a hardware crash and restarted
2293 				   - we can use this eg to kick tcp sessions
2294 				   once done */
2295 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2296 #define NETDEV_REGISTER 0x0005
2297 #define NETDEV_UNREGISTER	0x0006
2298 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2299 #define NETDEV_CHANGEADDR	0x0008
2300 #define NETDEV_GOING_DOWN	0x0009
2301 #define NETDEV_CHANGENAME	0x000A
2302 #define NETDEV_FEAT_CHANGE	0x000B
2303 #define NETDEV_BONDING_FAILOVER 0x000C
2304 #define NETDEV_PRE_UP		0x000D
2305 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2306 #define NETDEV_POST_TYPE_CHANGE	0x000F
2307 #define NETDEV_POST_INIT	0x0010
2308 #define NETDEV_UNREGISTER_FINAL 0x0011
2309 #define NETDEV_RELEASE		0x0012
2310 #define NETDEV_NOTIFY_PEERS	0x0013
2311 #define NETDEV_JOIN		0x0014
2312 #define NETDEV_CHANGEUPPER	0x0015
2313 #define NETDEV_RESEND_IGMP	0x0016
2314 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2315 #define NETDEV_CHANGEINFODATA	0x0018
2316 #define NETDEV_BONDING_INFO	0x0019
2317 #define NETDEV_PRECHANGEUPPER	0x001A
2318 #define NETDEV_CHANGELOWERSTATE	0x001B
2319 #define NETDEV_UDP_TUNNEL_PUSH_INFO	0x001C
2320 #define NETDEV_UDP_TUNNEL_DROP_INFO	0x001D
2321 #define NETDEV_CHANGE_TX_QUEUE_LEN	0x001E
2322 
2323 int register_netdevice_notifier(struct notifier_block *nb);
2324 int unregister_netdevice_notifier(struct notifier_block *nb);
2325 
2326 struct netdev_notifier_info {
2327 	struct net_device *dev;
2328 };
2329 
2330 struct netdev_notifier_change_info {
2331 	struct netdev_notifier_info info; /* must be first */
2332 	unsigned int flags_changed;
2333 };
2334 
2335 struct netdev_notifier_changeupper_info {
2336 	struct netdev_notifier_info info; /* must be first */
2337 	struct net_device *upper_dev; /* new upper dev */
2338 	bool master; /* is upper dev master */
2339 	bool linking; /* is the notification for link or unlink */
2340 	void *upper_info; /* upper dev info */
2341 };
2342 
2343 struct netdev_notifier_changelowerstate_info {
2344 	struct netdev_notifier_info info; /* must be first */
2345 	void *lower_state_info; /* is lower dev state */
2346 };
2347 
2348 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2349 					     struct net_device *dev)
2350 {
2351 	info->dev = dev;
2352 }
2353 
2354 static inline struct net_device *
2355 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2356 {
2357 	return info->dev;
2358 }
2359 
2360 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2361 
2362 
2363 extern rwlock_t				dev_base_lock;		/* Device list lock */
2364 
2365 #define for_each_netdev(net, d)		\
2366 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2367 #define for_each_netdev_reverse(net, d)	\
2368 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2369 #define for_each_netdev_rcu(net, d)		\
2370 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2371 #define for_each_netdev_safe(net, d, n)	\
2372 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2373 #define for_each_netdev_continue(net, d)		\
2374 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2375 #define for_each_netdev_continue_rcu(net, d)		\
2376 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2377 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2378 		for_each_netdev_rcu(&init_net, slave)	\
2379 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2380 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2381 
2382 static inline struct net_device *next_net_device(struct net_device *dev)
2383 {
2384 	struct list_head *lh;
2385 	struct net *net;
2386 
2387 	net = dev_net(dev);
2388 	lh = dev->dev_list.next;
2389 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2390 }
2391 
2392 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2393 {
2394 	struct list_head *lh;
2395 	struct net *net;
2396 
2397 	net = dev_net(dev);
2398 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2399 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2400 }
2401 
2402 static inline struct net_device *first_net_device(struct net *net)
2403 {
2404 	return list_empty(&net->dev_base_head) ? NULL :
2405 		net_device_entry(net->dev_base_head.next);
2406 }
2407 
2408 static inline struct net_device *first_net_device_rcu(struct net *net)
2409 {
2410 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2411 
2412 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2413 }
2414 
2415 int netdev_boot_setup_check(struct net_device *dev);
2416 unsigned long netdev_boot_base(const char *prefix, int unit);
2417 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2418 				       const char *hwaddr);
2419 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2420 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2421 void dev_add_pack(struct packet_type *pt);
2422 void dev_remove_pack(struct packet_type *pt);
2423 void __dev_remove_pack(struct packet_type *pt);
2424 void dev_add_offload(struct packet_offload *po);
2425 void dev_remove_offload(struct packet_offload *po);
2426 
2427 int dev_get_iflink(const struct net_device *dev);
2428 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2429 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2430 				      unsigned short mask);
2431 struct net_device *dev_get_by_name(struct net *net, const char *name);
2432 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2433 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2434 int dev_alloc_name(struct net_device *dev, const char *name);
2435 int dev_open(struct net_device *dev);
2436 void dev_close(struct net_device *dev);
2437 void dev_close_many(struct list_head *head, bool unlink);
2438 void dev_disable_lro(struct net_device *dev);
2439 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2440 int dev_queue_xmit(struct sk_buff *skb);
2441 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2442 int register_netdevice(struct net_device *dev);
2443 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2444 void unregister_netdevice_many(struct list_head *head);
2445 static inline void unregister_netdevice(struct net_device *dev)
2446 {
2447 	unregister_netdevice_queue(dev, NULL);
2448 }
2449 
2450 int netdev_refcnt_read(const struct net_device *dev);
2451 void free_netdev(struct net_device *dev);
2452 void netdev_freemem(struct net_device *dev);
2453 void synchronize_net(void);
2454 int init_dummy_netdev(struct net_device *dev);
2455 
2456 DECLARE_PER_CPU(int, xmit_recursion);
2457 #define XMIT_RECURSION_LIMIT	10
2458 
2459 static inline int dev_recursion_level(void)
2460 {
2461 	return this_cpu_read(xmit_recursion);
2462 }
2463 
2464 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2465 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2466 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2467 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2468 int netdev_get_name(struct net *net, char *name, int ifindex);
2469 int dev_restart(struct net_device *dev);
2470 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2471 
2472 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2473 {
2474 	return NAPI_GRO_CB(skb)->data_offset;
2475 }
2476 
2477 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2478 {
2479 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2480 }
2481 
2482 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2483 {
2484 	NAPI_GRO_CB(skb)->data_offset += len;
2485 }
2486 
2487 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2488 					unsigned int offset)
2489 {
2490 	return NAPI_GRO_CB(skb)->frag0 + offset;
2491 }
2492 
2493 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2494 {
2495 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2496 }
2497 
2498 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2499 {
2500 	NAPI_GRO_CB(skb)->frag0 = NULL;
2501 	NAPI_GRO_CB(skb)->frag0_len = 0;
2502 }
2503 
2504 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2505 					unsigned int offset)
2506 {
2507 	if (!pskb_may_pull(skb, hlen))
2508 		return NULL;
2509 
2510 	skb_gro_frag0_invalidate(skb);
2511 	return skb->data + offset;
2512 }
2513 
2514 static inline void *skb_gro_network_header(struct sk_buff *skb)
2515 {
2516 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2517 	       skb_network_offset(skb);
2518 }
2519 
2520 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2521 					const void *start, unsigned int len)
2522 {
2523 	if (NAPI_GRO_CB(skb)->csum_valid)
2524 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2525 						  csum_partial(start, len, 0));
2526 }
2527 
2528 /* GRO checksum functions. These are logical equivalents of the normal
2529  * checksum functions (in skbuff.h) except that they operate on the GRO
2530  * offsets and fields in sk_buff.
2531  */
2532 
2533 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2534 
2535 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2536 {
2537 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2538 }
2539 
2540 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2541 						      bool zero_okay,
2542 						      __sum16 check)
2543 {
2544 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2545 		skb_checksum_start_offset(skb) <
2546 		 skb_gro_offset(skb)) &&
2547 		!skb_at_gro_remcsum_start(skb) &&
2548 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2549 		(!zero_okay || check));
2550 }
2551 
2552 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2553 							   __wsum psum)
2554 {
2555 	if (NAPI_GRO_CB(skb)->csum_valid &&
2556 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2557 		return 0;
2558 
2559 	NAPI_GRO_CB(skb)->csum = psum;
2560 
2561 	return __skb_gro_checksum_complete(skb);
2562 }
2563 
2564 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2565 {
2566 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2567 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2568 		NAPI_GRO_CB(skb)->csum_cnt--;
2569 	} else {
2570 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2571 		 * verified a new top level checksum or an encapsulated one
2572 		 * during GRO. This saves work if we fallback to normal path.
2573 		 */
2574 		__skb_incr_checksum_unnecessary(skb);
2575 	}
2576 }
2577 
2578 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2579 				    compute_pseudo)			\
2580 ({									\
2581 	__sum16 __ret = 0;						\
2582 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2583 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2584 				compute_pseudo(skb, proto));		\
2585 	if (!__ret)							\
2586 		skb_gro_incr_csum_unnecessary(skb);			\
2587 	__ret;								\
2588 })
2589 
2590 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2591 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2592 
2593 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2594 					     compute_pseudo)		\
2595 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2596 
2597 #define skb_gro_checksum_simple_validate(skb)				\
2598 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2599 
2600 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2601 {
2602 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2603 		!NAPI_GRO_CB(skb)->csum_valid);
2604 }
2605 
2606 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2607 					      __sum16 check, __wsum pseudo)
2608 {
2609 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2610 	NAPI_GRO_CB(skb)->csum_valid = 1;
2611 }
2612 
2613 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2614 do {									\
2615 	if (__skb_gro_checksum_convert_check(skb))			\
2616 		__skb_gro_checksum_convert(skb, check,			\
2617 					   compute_pseudo(skb, proto));	\
2618 } while (0)
2619 
2620 struct gro_remcsum {
2621 	int offset;
2622 	__wsum delta;
2623 };
2624 
2625 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2626 {
2627 	grc->offset = 0;
2628 	grc->delta = 0;
2629 }
2630 
2631 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2632 					    unsigned int off, size_t hdrlen,
2633 					    int start, int offset,
2634 					    struct gro_remcsum *grc,
2635 					    bool nopartial)
2636 {
2637 	__wsum delta;
2638 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2639 
2640 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2641 
2642 	if (!nopartial) {
2643 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2644 		return ptr;
2645 	}
2646 
2647 	ptr = skb_gro_header_fast(skb, off);
2648 	if (skb_gro_header_hard(skb, off + plen)) {
2649 		ptr = skb_gro_header_slow(skb, off + plen, off);
2650 		if (!ptr)
2651 			return NULL;
2652 	}
2653 
2654 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2655 			       start, offset);
2656 
2657 	/* Adjust skb->csum since we changed the packet */
2658 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2659 
2660 	grc->offset = off + hdrlen + offset;
2661 	grc->delta = delta;
2662 
2663 	return ptr;
2664 }
2665 
2666 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2667 					   struct gro_remcsum *grc)
2668 {
2669 	void *ptr;
2670 	size_t plen = grc->offset + sizeof(u16);
2671 
2672 	if (!grc->delta)
2673 		return;
2674 
2675 	ptr = skb_gro_header_fast(skb, grc->offset);
2676 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2677 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2678 		if (!ptr)
2679 			return;
2680 	}
2681 
2682 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2683 }
2684 
2685 #ifdef CONFIG_XFRM_OFFLOAD
2686 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2687 {
2688 	if (PTR_ERR(pp) != -EINPROGRESS)
2689 		NAPI_GRO_CB(skb)->flush |= flush;
2690 }
2691 #else
2692 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2693 {
2694 	NAPI_GRO_CB(skb)->flush |= flush;
2695 }
2696 #endif
2697 
2698 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2699 				  unsigned short type,
2700 				  const void *daddr, const void *saddr,
2701 				  unsigned int len)
2702 {
2703 	if (!dev->header_ops || !dev->header_ops->create)
2704 		return 0;
2705 
2706 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2707 }
2708 
2709 static inline int dev_parse_header(const struct sk_buff *skb,
2710 				   unsigned char *haddr)
2711 {
2712 	const struct net_device *dev = skb->dev;
2713 
2714 	if (!dev->header_ops || !dev->header_ops->parse)
2715 		return 0;
2716 	return dev->header_ops->parse(skb, haddr);
2717 }
2718 
2719 /* ll_header must have at least hard_header_len allocated */
2720 static inline bool dev_validate_header(const struct net_device *dev,
2721 				       char *ll_header, int len)
2722 {
2723 	if (likely(len >= dev->hard_header_len))
2724 		return true;
2725 	if (len < dev->min_header_len)
2726 		return false;
2727 
2728 	if (capable(CAP_SYS_RAWIO)) {
2729 		memset(ll_header + len, 0, dev->hard_header_len - len);
2730 		return true;
2731 	}
2732 
2733 	if (dev->header_ops && dev->header_ops->validate)
2734 		return dev->header_ops->validate(ll_header, len);
2735 
2736 	return false;
2737 }
2738 
2739 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2740 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2741 static inline int unregister_gifconf(unsigned int family)
2742 {
2743 	return register_gifconf(family, NULL);
2744 }
2745 
2746 #ifdef CONFIG_NET_FLOW_LIMIT
2747 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2748 struct sd_flow_limit {
2749 	u64			count;
2750 	unsigned int		num_buckets;
2751 	unsigned int		history_head;
2752 	u16			history[FLOW_LIMIT_HISTORY];
2753 	u8			buckets[];
2754 };
2755 
2756 extern int netdev_flow_limit_table_len;
2757 #endif /* CONFIG_NET_FLOW_LIMIT */
2758 
2759 /*
2760  * Incoming packets are placed on per-CPU queues
2761  */
2762 struct softnet_data {
2763 	struct list_head	poll_list;
2764 	struct sk_buff_head	process_queue;
2765 
2766 	/* stats */
2767 	unsigned int		processed;
2768 	unsigned int		time_squeeze;
2769 	unsigned int		received_rps;
2770 #ifdef CONFIG_RPS
2771 	struct softnet_data	*rps_ipi_list;
2772 #endif
2773 #ifdef CONFIG_NET_FLOW_LIMIT
2774 	struct sd_flow_limit __rcu *flow_limit;
2775 #endif
2776 	struct Qdisc		*output_queue;
2777 	struct Qdisc		**output_queue_tailp;
2778 	struct sk_buff		*completion_queue;
2779 
2780 #ifdef CONFIG_RPS
2781 	/* input_queue_head should be written by cpu owning this struct,
2782 	 * and only read by other cpus. Worth using a cache line.
2783 	 */
2784 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2785 
2786 	/* Elements below can be accessed between CPUs for RPS/RFS */
2787 	struct call_single_data	csd ____cacheline_aligned_in_smp;
2788 	struct softnet_data	*rps_ipi_next;
2789 	unsigned int		cpu;
2790 	unsigned int		input_queue_tail;
2791 #endif
2792 	unsigned int		dropped;
2793 	struct sk_buff_head	input_pkt_queue;
2794 	struct napi_struct	backlog;
2795 
2796 };
2797 
2798 static inline void input_queue_head_incr(struct softnet_data *sd)
2799 {
2800 #ifdef CONFIG_RPS
2801 	sd->input_queue_head++;
2802 #endif
2803 }
2804 
2805 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2806 					      unsigned int *qtail)
2807 {
2808 #ifdef CONFIG_RPS
2809 	*qtail = ++sd->input_queue_tail;
2810 #endif
2811 }
2812 
2813 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2814 
2815 void __netif_schedule(struct Qdisc *q);
2816 void netif_schedule_queue(struct netdev_queue *txq);
2817 
2818 static inline void netif_tx_schedule_all(struct net_device *dev)
2819 {
2820 	unsigned int i;
2821 
2822 	for (i = 0; i < dev->num_tx_queues; i++)
2823 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2824 }
2825 
2826 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2827 {
2828 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2829 }
2830 
2831 /**
2832  *	netif_start_queue - allow transmit
2833  *	@dev: network device
2834  *
2835  *	Allow upper layers to call the device hard_start_xmit routine.
2836  */
2837 static inline void netif_start_queue(struct net_device *dev)
2838 {
2839 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2840 }
2841 
2842 static inline void netif_tx_start_all_queues(struct net_device *dev)
2843 {
2844 	unsigned int i;
2845 
2846 	for (i = 0; i < dev->num_tx_queues; i++) {
2847 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2848 		netif_tx_start_queue(txq);
2849 	}
2850 }
2851 
2852 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2853 
2854 /**
2855  *	netif_wake_queue - restart transmit
2856  *	@dev: network device
2857  *
2858  *	Allow upper layers to call the device hard_start_xmit routine.
2859  *	Used for flow control when transmit resources are available.
2860  */
2861 static inline void netif_wake_queue(struct net_device *dev)
2862 {
2863 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2864 }
2865 
2866 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2867 {
2868 	unsigned int i;
2869 
2870 	for (i = 0; i < dev->num_tx_queues; i++) {
2871 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2872 		netif_tx_wake_queue(txq);
2873 	}
2874 }
2875 
2876 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2877 {
2878 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2879 }
2880 
2881 /**
2882  *	netif_stop_queue - stop transmitted packets
2883  *	@dev: network device
2884  *
2885  *	Stop upper layers calling the device hard_start_xmit routine.
2886  *	Used for flow control when transmit resources are unavailable.
2887  */
2888 static inline void netif_stop_queue(struct net_device *dev)
2889 {
2890 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2891 }
2892 
2893 void netif_tx_stop_all_queues(struct net_device *dev);
2894 
2895 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2896 {
2897 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2898 }
2899 
2900 /**
2901  *	netif_queue_stopped - test if transmit queue is flowblocked
2902  *	@dev: network device
2903  *
2904  *	Test if transmit queue on device is currently unable to send.
2905  */
2906 static inline bool netif_queue_stopped(const struct net_device *dev)
2907 {
2908 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2909 }
2910 
2911 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2912 {
2913 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2914 }
2915 
2916 static inline bool
2917 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2918 {
2919 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2920 }
2921 
2922 static inline bool
2923 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2924 {
2925 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2926 }
2927 
2928 /**
2929  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2930  *	@dev_queue: pointer to transmit queue
2931  *
2932  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2933  * to give appropriate hint to the CPU.
2934  */
2935 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2936 {
2937 #ifdef CONFIG_BQL
2938 	prefetchw(&dev_queue->dql.num_queued);
2939 #endif
2940 }
2941 
2942 /**
2943  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2944  *	@dev_queue: pointer to transmit queue
2945  *
2946  * BQL enabled drivers might use this helper in their TX completion path,
2947  * to give appropriate hint to the CPU.
2948  */
2949 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2950 {
2951 #ifdef CONFIG_BQL
2952 	prefetchw(&dev_queue->dql.limit);
2953 #endif
2954 }
2955 
2956 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2957 					unsigned int bytes)
2958 {
2959 #ifdef CONFIG_BQL
2960 	dql_queued(&dev_queue->dql, bytes);
2961 
2962 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2963 		return;
2964 
2965 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2966 
2967 	/*
2968 	 * The XOFF flag must be set before checking the dql_avail below,
2969 	 * because in netdev_tx_completed_queue we update the dql_completed
2970 	 * before checking the XOFF flag.
2971 	 */
2972 	smp_mb();
2973 
2974 	/* check again in case another CPU has just made room avail */
2975 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2976 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2977 #endif
2978 }
2979 
2980 /**
2981  * 	netdev_sent_queue - report the number of bytes queued to hardware
2982  * 	@dev: network device
2983  * 	@bytes: number of bytes queued to the hardware device queue
2984  *
2985  * 	Report the number of bytes queued for sending/completion to the network
2986  * 	device hardware queue. @bytes should be a good approximation and should
2987  * 	exactly match netdev_completed_queue() @bytes
2988  */
2989 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2990 {
2991 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2992 }
2993 
2994 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2995 					     unsigned int pkts, unsigned int bytes)
2996 {
2997 #ifdef CONFIG_BQL
2998 	if (unlikely(!bytes))
2999 		return;
3000 
3001 	dql_completed(&dev_queue->dql, bytes);
3002 
3003 	/*
3004 	 * Without the memory barrier there is a small possiblity that
3005 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3006 	 * be stopped forever
3007 	 */
3008 	smp_mb();
3009 
3010 	if (dql_avail(&dev_queue->dql) < 0)
3011 		return;
3012 
3013 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3014 		netif_schedule_queue(dev_queue);
3015 #endif
3016 }
3017 
3018 /**
3019  * 	netdev_completed_queue - report bytes and packets completed by device
3020  * 	@dev: network device
3021  * 	@pkts: actual number of packets sent over the medium
3022  * 	@bytes: actual number of bytes sent over the medium
3023  *
3024  * 	Report the number of bytes and packets transmitted by the network device
3025  * 	hardware queue over the physical medium, @bytes must exactly match the
3026  * 	@bytes amount passed to netdev_sent_queue()
3027  */
3028 static inline void netdev_completed_queue(struct net_device *dev,
3029 					  unsigned int pkts, unsigned int bytes)
3030 {
3031 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3032 }
3033 
3034 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3035 {
3036 #ifdef CONFIG_BQL
3037 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3038 	dql_reset(&q->dql);
3039 #endif
3040 }
3041 
3042 /**
3043  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3044  * 	@dev_queue: network device
3045  *
3046  * 	Reset the bytes and packet count of a network device and clear the
3047  * 	software flow control OFF bit for this network device
3048  */
3049 static inline void netdev_reset_queue(struct net_device *dev_queue)
3050 {
3051 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3052 }
3053 
3054 /**
3055  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3056  * 	@dev: network device
3057  * 	@queue_index: given tx queue index
3058  *
3059  * 	Returns 0 if given tx queue index >= number of device tx queues,
3060  * 	otherwise returns the originally passed tx queue index.
3061  */
3062 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3063 {
3064 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3065 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3066 				     dev->name, queue_index,
3067 				     dev->real_num_tx_queues);
3068 		return 0;
3069 	}
3070 
3071 	return queue_index;
3072 }
3073 
3074 /**
3075  *	netif_running - test if up
3076  *	@dev: network device
3077  *
3078  *	Test if the device has been brought up.
3079  */
3080 static inline bool netif_running(const struct net_device *dev)
3081 {
3082 	return test_bit(__LINK_STATE_START, &dev->state);
3083 }
3084 
3085 /*
3086  * Routines to manage the subqueues on a device.  We only need start,
3087  * stop, and a check if it's stopped.  All other device management is
3088  * done at the overall netdevice level.
3089  * Also test the device if we're multiqueue.
3090  */
3091 
3092 /**
3093  *	netif_start_subqueue - allow sending packets on subqueue
3094  *	@dev: network device
3095  *	@queue_index: sub queue index
3096  *
3097  * Start individual transmit queue of a device with multiple transmit queues.
3098  */
3099 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3100 {
3101 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3102 
3103 	netif_tx_start_queue(txq);
3104 }
3105 
3106 /**
3107  *	netif_stop_subqueue - stop sending packets on subqueue
3108  *	@dev: network device
3109  *	@queue_index: sub queue index
3110  *
3111  * Stop individual transmit queue of a device with multiple transmit queues.
3112  */
3113 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3114 {
3115 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3116 	netif_tx_stop_queue(txq);
3117 }
3118 
3119 /**
3120  *	netif_subqueue_stopped - test status of subqueue
3121  *	@dev: network device
3122  *	@queue_index: sub queue index
3123  *
3124  * Check individual transmit queue of a device with multiple transmit queues.
3125  */
3126 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3127 					    u16 queue_index)
3128 {
3129 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3130 
3131 	return netif_tx_queue_stopped(txq);
3132 }
3133 
3134 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3135 					  struct sk_buff *skb)
3136 {
3137 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3138 }
3139 
3140 /**
3141  *	netif_wake_subqueue - allow sending packets on subqueue
3142  *	@dev: network device
3143  *	@queue_index: sub queue index
3144  *
3145  * Resume individual transmit queue of a device with multiple transmit queues.
3146  */
3147 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3148 {
3149 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3150 
3151 	netif_tx_wake_queue(txq);
3152 }
3153 
3154 #ifdef CONFIG_XPS
3155 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3156 			u16 index);
3157 #else
3158 static inline int netif_set_xps_queue(struct net_device *dev,
3159 				      const struct cpumask *mask,
3160 				      u16 index)
3161 {
3162 	return 0;
3163 }
3164 #endif
3165 
3166 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3167 		  unsigned int num_tx_queues);
3168 
3169 /*
3170  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3171  * as a distribution range limit for the returned value.
3172  */
3173 static inline u16 skb_tx_hash(const struct net_device *dev,
3174 			      struct sk_buff *skb)
3175 {
3176 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3177 }
3178 
3179 /**
3180  *	netif_is_multiqueue - test if device has multiple transmit queues
3181  *	@dev: network device
3182  *
3183  * Check if device has multiple transmit queues
3184  */
3185 static inline bool netif_is_multiqueue(const struct net_device *dev)
3186 {
3187 	return dev->num_tx_queues > 1;
3188 }
3189 
3190 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3191 
3192 #ifdef CONFIG_SYSFS
3193 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3194 #else
3195 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3196 						unsigned int rxq)
3197 {
3198 	return 0;
3199 }
3200 #endif
3201 
3202 #ifdef CONFIG_SYSFS
3203 static inline unsigned int get_netdev_rx_queue_index(
3204 		struct netdev_rx_queue *queue)
3205 {
3206 	struct net_device *dev = queue->dev;
3207 	int index = queue - dev->_rx;
3208 
3209 	BUG_ON(index >= dev->num_rx_queues);
3210 	return index;
3211 }
3212 #endif
3213 
3214 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3215 int netif_get_num_default_rss_queues(void);
3216 
3217 enum skb_free_reason {
3218 	SKB_REASON_CONSUMED,
3219 	SKB_REASON_DROPPED,
3220 };
3221 
3222 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3223 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3224 
3225 /*
3226  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3227  * interrupt context or with hardware interrupts being disabled.
3228  * (in_irq() || irqs_disabled())
3229  *
3230  * We provide four helpers that can be used in following contexts :
3231  *
3232  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3233  *  replacing kfree_skb(skb)
3234  *
3235  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3236  *  Typically used in place of consume_skb(skb) in TX completion path
3237  *
3238  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3239  *  replacing kfree_skb(skb)
3240  *
3241  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3242  *  and consumed a packet. Used in place of consume_skb(skb)
3243  */
3244 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3245 {
3246 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3247 }
3248 
3249 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3250 {
3251 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3252 }
3253 
3254 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3255 {
3256 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3257 }
3258 
3259 static inline void dev_consume_skb_any(struct sk_buff *skb)
3260 {
3261 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3262 }
3263 
3264 int netif_rx(struct sk_buff *skb);
3265 int netif_rx_ni(struct sk_buff *skb);
3266 int netif_receive_skb(struct sk_buff *skb);
3267 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3268 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3269 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3270 gro_result_t napi_gro_frags(struct napi_struct *napi);
3271 struct packet_offload *gro_find_receive_by_type(__be16 type);
3272 struct packet_offload *gro_find_complete_by_type(__be16 type);
3273 
3274 static inline void napi_free_frags(struct napi_struct *napi)
3275 {
3276 	kfree_skb(napi->skb);
3277 	napi->skb = NULL;
3278 }
3279 
3280 bool netdev_is_rx_handler_busy(struct net_device *dev);
3281 int netdev_rx_handler_register(struct net_device *dev,
3282 			       rx_handler_func_t *rx_handler,
3283 			       void *rx_handler_data);
3284 void netdev_rx_handler_unregister(struct net_device *dev);
3285 
3286 bool dev_valid_name(const char *name);
3287 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3288 int dev_ethtool(struct net *net, struct ifreq *);
3289 unsigned int dev_get_flags(const struct net_device *);
3290 int __dev_change_flags(struct net_device *, unsigned int flags);
3291 int dev_change_flags(struct net_device *, unsigned int);
3292 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3293 			unsigned int gchanges);
3294 int dev_change_name(struct net_device *, const char *);
3295 int dev_set_alias(struct net_device *, const char *, size_t);
3296 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3297 int __dev_set_mtu(struct net_device *, int);
3298 int dev_set_mtu(struct net_device *, int);
3299 void dev_set_group(struct net_device *, int);
3300 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3301 int dev_change_carrier(struct net_device *, bool new_carrier);
3302 int dev_get_phys_port_id(struct net_device *dev,
3303 			 struct netdev_phys_item_id *ppid);
3304 int dev_get_phys_port_name(struct net_device *dev,
3305 			   char *name, size_t len);
3306 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3307 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3308 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3309 				    struct netdev_queue *txq, int *ret);
3310 
3311 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3312 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3313 		      int fd, u32 flags);
3314 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
3315 
3316 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3317 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3318 bool is_skb_forwardable(const struct net_device *dev,
3319 			const struct sk_buff *skb);
3320 
3321 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3322 					       struct sk_buff *skb)
3323 {
3324 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3325 	    unlikely(!is_skb_forwardable(dev, skb))) {
3326 		atomic_long_inc(&dev->rx_dropped);
3327 		kfree_skb(skb);
3328 		return NET_RX_DROP;
3329 	}
3330 
3331 	skb_scrub_packet(skb, true);
3332 	skb->priority = 0;
3333 	return 0;
3334 }
3335 
3336 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3337 
3338 extern int		netdev_budget;
3339 extern unsigned int	netdev_budget_usecs;
3340 
3341 /* Called by rtnetlink.c:rtnl_unlock() */
3342 void netdev_run_todo(void);
3343 
3344 /**
3345  *	dev_put - release reference to device
3346  *	@dev: network device
3347  *
3348  * Release reference to device to allow it to be freed.
3349  */
3350 static inline void dev_put(struct net_device *dev)
3351 {
3352 	this_cpu_dec(*dev->pcpu_refcnt);
3353 }
3354 
3355 /**
3356  *	dev_hold - get reference to device
3357  *	@dev: network device
3358  *
3359  * Hold reference to device to keep it from being freed.
3360  */
3361 static inline void dev_hold(struct net_device *dev)
3362 {
3363 	this_cpu_inc(*dev->pcpu_refcnt);
3364 }
3365 
3366 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3367  * and _off may be called from IRQ context, but it is caller
3368  * who is responsible for serialization of these calls.
3369  *
3370  * The name carrier is inappropriate, these functions should really be
3371  * called netif_lowerlayer_*() because they represent the state of any
3372  * kind of lower layer not just hardware media.
3373  */
3374 
3375 void linkwatch_init_dev(struct net_device *dev);
3376 void linkwatch_fire_event(struct net_device *dev);
3377 void linkwatch_forget_dev(struct net_device *dev);
3378 
3379 /**
3380  *	netif_carrier_ok - test if carrier present
3381  *	@dev: network device
3382  *
3383  * Check if carrier is present on device
3384  */
3385 static inline bool netif_carrier_ok(const struct net_device *dev)
3386 {
3387 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3388 }
3389 
3390 unsigned long dev_trans_start(struct net_device *dev);
3391 
3392 void __netdev_watchdog_up(struct net_device *dev);
3393 
3394 void netif_carrier_on(struct net_device *dev);
3395 
3396 void netif_carrier_off(struct net_device *dev);
3397 
3398 /**
3399  *	netif_dormant_on - mark device as dormant.
3400  *	@dev: network device
3401  *
3402  * Mark device as dormant (as per RFC2863).
3403  *
3404  * The dormant state indicates that the relevant interface is not
3405  * actually in a condition to pass packets (i.e., it is not 'up') but is
3406  * in a "pending" state, waiting for some external event.  For "on-
3407  * demand" interfaces, this new state identifies the situation where the
3408  * interface is waiting for events to place it in the up state.
3409  */
3410 static inline void netif_dormant_on(struct net_device *dev)
3411 {
3412 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3413 		linkwatch_fire_event(dev);
3414 }
3415 
3416 /**
3417  *	netif_dormant_off - set device as not dormant.
3418  *	@dev: network device
3419  *
3420  * Device is not in dormant state.
3421  */
3422 static inline void netif_dormant_off(struct net_device *dev)
3423 {
3424 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3425 		linkwatch_fire_event(dev);
3426 }
3427 
3428 /**
3429  *	netif_dormant - test if device is dormant
3430  *	@dev: network device
3431  *
3432  * Check if device is dormant.
3433  */
3434 static inline bool netif_dormant(const struct net_device *dev)
3435 {
3436 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3437 }
3438 
3439 
3440 /**
3441  *	netif_oper_up - test if device is operational
3442  *	@dev: network device
3443  *
3444  * Check if carrier is operational
3445  */
3446 static inline bool netif_oper_up(const struct net_device *dev)
3447 {
3448 	return (dev->operstate == IF_OPER_UP ||
3449 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3450 }
3451 
3452 /**
3453  *	netif_device_present - is device available or removed
3454  *	@dev: network device
3455  *
3456  * Check if device has not been removed from system.
3457  */
3458 static inline bool netif_device_present(struct net_device *dev)
3459 {
3460 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3461 }
3462 
3463 void netif_device_detach(struct net_device *dev);
3464 
3465 void netif_device_attach(struct net_device *dev);
3466 
3467 /*
3468  * Network interface message level settings
3469  */
3470 
3471 enum {
3472 	NETIF_MSG_DRV		= 0x0001,
3473 	NETIF_MSG_PROBE		= 0x0002,
3474 	NETIF_MSG_LINK		= 0x0004,
3475 	NETIF_MSG_TIMER		= 0x0008,
3476 	NETIF_MSG_IFDOWN	= 0x0010,
3477 	NETIF_MSG_IFUP		= 0x0020,
3478 	NETIF_MSG_RX_ERR	= 0x0040,
3479 	NETIF_MSG_TX_ERR	= 0x0080,
3480 	NETIF_MSG_TX_QUEUED	= 0x0100,
3481 	NETIF_MSG_INTR		= 0x0200,
3482 	NETIF_MSG_TX_DONE	= 0x0400,
3483 	NETIF_MSG_RX_STATUS	= 0x0800,
3484 	NETIF_MSG_PKTDATA	= 0x1000,
3485 	NETIF_MSG_HW		= 0x2000,
3486 	NETIF_MSG_WOL		= 0x4000,
3487 };
3488 
3489 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3490 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3491 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3492 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3493 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3494 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3495 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3496 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3497 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3498 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3499 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3500 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3501 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3502 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3503 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3504 
3505 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3506 {
3507 	/* use default */
3508 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3509 		return default_msg_enable_bits;
3510 	if (debug_value == 0)	/* no output */
3511 		return 0;
3512 	/* set low N bits */
3513 	return (1 << debug_value) - 1;
3514 }
3515 
3516 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3517 {
3518 	spin_lock(&txq->_xmit_lock);
3519 	txq->xmit_lock_owner = cpu;
3520 }
3521 
3522 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3523 {
3524 	__acquire(&txq->_xmit_lock);
3525 	return true;
3526 }
3527 
3528 static inline void __netif_tx_release(struct netdev_queue *txq)
3529 {
3530 	__release(&txq->_xmit_lock);
3531 }
3532 
3533 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3534 {
3535 	spin_lock_bh(&txq->_xmit_lock);
3536 	txq->xmit_lock_owner = smp_processor_id();
3537 }
3538 
3539 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3540 {
3541 	bool ok = spin_trylock(&txq->_xmit_lock);
3542 	if (likely(ok))
3543 		txq->xmit_lock_owner = smp_processor_id();
3544 	return ok;
3545 }
3546 
3547 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3548 {
3549 	txq->xmit_lock_owner = -1;
3550 	spin_unlock(&txq->_xmit_lock);
3551 }
3552 
3553 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3554 {
3555 	txq->xmit_lock_owner = -1;
3556 	spin_unlock_bh(&txq->_xmit_lock);
3557 }
3558 
3559 static inline void txq_trans_update(struct netdev_queue *txq)
3560 {
3561 	if (txq->xmit_lock_owner != -1)
3562 		txq->trans_start = jiffies;
3563 }
3564 
3565 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3566 static inline void netif_trans_update(struct net_device *dev)
3567 {
3568 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3569 
3570 	if (txq->trans_start != jiffies)
3571 		txq->trans_start = jiffies;
3572 }
3573 
3574 /**
3575  *	netif_tx_lock - grab network device transmit lock
3576  *	@dev: network device
3577  *
3578  * Get network device transmit lock
3579  */
3580 static inline void netif_tx_lock(struct net_device *dev)
3581 {
3582 	unsigned int i;
3583 	int cpu;
3584 
3585 	spin_lock(&dev->tx_global_lock);
3586 	cpu = smp_processor_id();
3587 	for (i = 0; i < dev->num_tx_queues; i++) {
3588 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3589 
3590 		/* We are the only thread of execution doing a
3591 		 * freeze, but we have to grab the _xmit_lock in
3592 		 * order to synchronize with threads which are in
3593 		 * the ->hard_start_xmit() handler and already
3594 		 * checked the frozen bit.
3595 		 */
3596 		__netif_tx_lock(txq, cpu);
3597 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3598 		__netif_tx_unlock(txq);
3599 	}
3600 }
3601 
3602 static inline void netif_tx_lock_bh(struct net_device *dev)
3603 {
3604 	local_bh_disable();
3605 	netif_tx_lock(dev);
3606 }
3607 
3608 static inline void netif_tx_unlock(struct net_device *dev)
3609 {
3610 	unsigned int i;
3611 
3612 	for (i = 0; i < dev->num_tx_queues; i++) {
3613 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3614 
3615 		/* No need to grab the _xmit_lock here.  If the
3616 		 * queue is not stopped for another reason, we
3617 		 * force a schedule.
3618 		 */
3619 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3620 		netif_schedule_queue(txq);
3621 	}
3622 	spin_unlock(&dev->tx_global_lock);
3623 }
3624 
3625 static inline void netif_tx_unlock_bh(struct net_device *dev)
3626 {
3627 	netif_tx_unlock(dev);
3628 	local_bh_enable();
3629 }
3630 
3631 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3632 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3633 		__netif_tx_lock(txq, cpu);		\
3634 	} else {					\
3635 		__netif_tx_acquire(txq);		\
3636 	}						\
3637 }
3638 
3639 #define HARD_TX_TRYLOCK(dev, txq)			\
3640 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3641 		__netif_tx_trylock(txq) :		\
3642 		__netif_tx_acquire(txq))
3643 
3644 #define HARD_TX_UNLOCK(dev, txq) {			\
3645 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3646 		__netif_tx_unlock(txq);			\
3647 	} else {					\
3648 		__netif_tx_release(txq);		\
3649 	}						\
3650 }
3651 
3652 static inline void netif_tx_disable(struct net_device *dev)
3653 {
3654 	unsigned int i;
3655 	int cpu;
3656 
3657 	local_bh_disable();
3658 	cpu = smp_processor_id();
3659 	for (i = 0; i < dev->num_tx_queues; i++) {
3660 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3661 
3662 		__netif_tx_lock(txq, cpu);
3663 		netif_tx_stop_queue(txq);
3664 		__netif_tx_unlock(txq);
3665 	}
3666 	local_bh_enable();
3667 }
3668 
3669 static inline void netif_addr_lock(struct net_device *dev)
3670 {
3671 	spin_lock(&dev->addr_list_lock);
3672 }
3673 
3674 static inline void netif_addr_lock_nested(struct net_device *dev)
3675 {
3676 	int subclass = SINGLE_DEPTH_NESTING;
3677 
3678 	if (dev->netdev_ops->ndo_get_lock_subclass)
3679 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3680 
3681 	spin_lock_nested(&dev->addr_list_lock, subclass);
3682 }
3683 
3684 static inline void netif_addr_lock_bh(struct net_device *dev)
3685 {
3686 	spin_lock_bh(&dev->addr_list_lock);
3687 }
3688 
3689 static inline void netif_addr_unlock(struct net_device *dev)
3690 {
3691 	spin_unlock(&dev->addr_list_lock);
3692 }
3693 
3694 static inline void netif_addr_unlock_bh(struct net_device *dev)
3695 {
3696 	spin_unlock_bh(&dev->addr_list_lock);
3697 }
3698 
3699 /*
3700  * dev_addrs walker. Should be used only for read access. Call with
3701  * rcu_read_lock held.
3702  */
3703 #define for_each_dev_addr(dev, ha) \
3704 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3705 
3706 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3707 
3708 void ether_setup(struct net_device *dev);
3709 
3710 /* Support for loadable net-drivers */
3711 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3712 				    unsigned char name_assign_type,
3713 				    void (*setup)(struct net_device *),
3714 				    unsigned int txqs, unsigned int rxqs);
3715 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3716 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3717 
3718 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3719 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3720 			 count)
3721 
3722 int register_netdev(struct net_device *dev);
3723 void unregister_netdev(struct net_device *dev);
3724 
3725 /* General hardware address lists handling functions */
3726 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3727 		   struct netdev_hw_addr_list *from_list, int addr_len);
3728 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3729 		      struct netdev_hw_addr_list *from_list, int addr_len);
3730 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3731 		       struct net_device *dev,
3732 		       int (*sync)(struct net_device *, const unsigned char *),
3733 		       int (*unsync)(struct net_device *,
3734 				     const unsigned char *));
3735 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3736 			  struct net_device *dev,
3737 			  int (*unsync)(struct net_device *,
3738 					const unsigned char *));
3739 void __hw_addr_init(struct netdev_hw_addr_list *list);
3740 
3741 /* Functions used for device addresses handling */
3742 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3743 		 unsigned char addr_type);
3744 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3745 		 unsigned char addr_type);
3746 void dev_addr_flush(struct net_device *dev);
3747 int dev_addr_init(struct net_device *dev);
3748 
3749 /* Functions used for unicast addresses handling */
3750 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3751 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3752 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3753 int dev_uc_sync(struct net_device *to, struct net_device *from);
3754 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3755 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3756 void dev_uc_flush(struct net_device *dev);
3757 void dev_uc_init(struct net_device *dev);
3758 
3759 /**
3760  *  __dev_uc_sync - Synchonize device's unicast list
3761  *  @dev:  device to sync
3762  *  @sync: function to call if address should be added
3763  *  @unsync: function to call if address should be removed
3764  *
3765  *  Add newly added addresses to the interface, and release
3766  *  addresses that have been deleted.
3767  */
3768 static inline int __dev_uc_sync(struct net_device *dev,
3769 				int (*sync)(struct net_device *,
3770 					    const unsigned char *),
3771 				int (*unsync)(struct net_device *,
3772 					      const unsigned char *))
3773 {
3774 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3775 }
3776 
3777 /**
3778  *  __dev_uc_unsync - Remove synchronized addresses from device
3779  *  @dev:  device to sync
3780  *  @unsync: function to call if address should be removed
3781  *
3782  *  Remove all addresses that were added to the device by dev_uc_sync().
3783  */
3784 static inline void __dev_uc_unsync(struct net_device *dev,
3785 				   int (*unsync)(struct net_device *,
3786 						 const unsigned char *))
3787 {
3788 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3789 }
3790 
3791 /* Functions used for multicast addresses handling */
3792 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3793 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3794 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3795 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3796 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3797 int dev_mc_sync(struct net_device *to, struct net_device *from);
3798 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3799 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3800 void dev_mc_flush(struct net_device *dev);
3801 void dev_mc_init(struct net_device *dev);
3802 
3803 /**
3804  *  __dev_mc_sync - Synchonize device's multicast list
3805  *  @dev:  device to sync
3806  *  @sync: function to call if address should be added
3807  *  @unsync: function to call if address should be removed
3808  *
3809  *  Add newly added addresses to the interface, and release
3810  *  addresses that have been deleted.
3811  */
3812 static inline int __dev_mc_sync(struct net_device *dev,
3813 				int (*sync)(struct net_device *,
3814 					    const unsigned char *),
3815 				int (*unsync)(struct net_device *,
3816 					      const unsigned char *))
3817 {
3818 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3819 }
3820 
3821 /**
3822  *  __dev_mc_unsync - Remove synchronized addresses from device
3823  *  @dev:  device to sync
3824  *  @unsync: function to call if address should be removed
3825  *
3826  *  Remove all addresses that were added to the device by dev_mc_sync().
3827  */
3828 static inline void __dev_mc_unsync(struct net_device *dev,
3829 				   int (*unsync)(struct net_device *,
3830 						 const unsigned char *))
3831 {
3832 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3833 }
3834 
3835 /* Functions used for secondary unicast and multicast support */
3836 void dev_set_rx_mode(struct net_device *dev);
3837 void __dev_set_rx_mode(struct net_device *dev);
3838 int dev_set_promiscuity(struct net_device *dev, int inc);
3839 int dev_set_allmulti(struct net_device *dev, int inc);
3840 void netdev_state_change(struct net_device *dev);
3841 void netdev_notify_peers(struct net_device *dev);
3842 void netdev_features_change(struct net_device *dev);
3843 /* Load a device via the kmod */
3844 void dev_load(struct net *net, const char *name);
3845 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3846 					struct rtnl_link_stats64 *storage);
3847 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3848 			     const struct net_device_stats *netdev_stats);
3849 
3850 extern int		netdev_max_backlog;
3851 extern int		netdev_tstamp_prequeue;
3852 extern int		weight_p;
3853 extern int		dev_weight_rx_bias;
3854 extern int		dev_weight_tx_bias;
3855 extern int		dev_rx_weight;
3856 extern int		dev_tx_weight;
3857 
3858 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3859 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3860 						     struct list_head **iter);
3861 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3862 						     struct list_head **iter);
3863 
3864 /* iterate through upper list, must be called under RCU read lock */
3865 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3866 	for (iter = &(dev)->adj_list.upper, \
3867 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3868 	     updev; \
3869 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3870 
3871 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3872 				  int (*fn)(struct net_device *upper_dev,
3873 					    void *data),
3874 				  void *data);
3875 
3876 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3877 				  struct net_device *upper_dev);
3878 
3879 void *netdev_lower_get_next_private(struct net_device *dev,
3880 				    struct list_head **iter);
3881 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3882 					struct list_head **iter);
3883 
3884 #define netdev_for_each_lower_private(dev, priv, iter) \
3885 	for (iter = (dev)->adj_list.lower.next, \
3886 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3887 	     priv; \
3888 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3889 
3890 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3891 	for (iter = &(dev)->adj_list.lower, \
3892 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3893 	     priv; \
3894 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3895 
3896 void *netdev_lower_get_next(struct net_device *dev,
3897 				struct list_head **iter);
3898 
3899 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3900 	for (iter = (dev)->adj_list.lower.next, \
3901 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3902 	     ldev; \
3903 	     ldev = netdev_lower_get_next(dev, &(iter)))
3904 
3905 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3906 					     struct list_head **iter);
3907 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3908 						 struct list_head **iter);
3909 
3910 int netdev_walk_all_lower_dev(struct net_device *dev,
3911 			      int (*fn)(struct net_device *lower_dev,
3912 					void *data),
3913 			      void *data);
3914 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3915 				  int (*fn)(struct net_device *lower_dev,
3916 					    void *data),
3917 				  void *data);
3918 
3919 void *netdev_adjacent_get_private(struct list_head *adj_list);
3920 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3921 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3922 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3923 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3924 int netdev_master_upper_dev_link(struct net_device *dev,
3925 				 struct net_device *upper_dev,
3926 				 void *upper_priv, void *upper_info);
3927 void netdev_upper_dev_unlink(struct net_device *dev,
3928 			     struct net_device *upper_dev);
3929 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3930 void *netdev_lower_dev_get_private(struct net_device *dev,
3931 				   struct net_device *lower_dev);
3932 void netdev_lower_state_changed(struct net_device *lower_dev,
3933 				void *lower_state_info);
3934 
3935 /* RSS keys are 40 or 52 bytes long */
3936 #define NETDEV_RSS_KEY_LEN 52
3937 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3938 void netdev_rss_key_fill(void *buffer, size_t len);
3939 
3940 int dev_get_nest_level(struct net_device *dev);
3941 int skb_checksum_help(struct sk_buff *skb);
3942 int skb_crc32c_csum_help(struct sk_buff *skb);
3943 int skb_csum_hwoffload_help(struct sk_buff *skb,
3944 			    const netdev_features_t features);
3945 
3946 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3947 				  netdev_features_t features, bool tx_path);
3948 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3949 				    netdev_features_t features);
3950 
3951 struct netdev_bonding_info {
3952 	ifslave	slave;
3953 	ifbond	master;
3954 };
3955 
3956 struct netdev_notifier_bonding_info {
3957 	struct netdev_notifier_info info; /* must be first */
3958 	struct netdev_bonding_info  bonding_info;
3959 };
3960 
3961 void netdev_bonding_info_change(struct net_device *dev,
3962 				struct netdev_bonding_info *bonding_info);
3963 
3964 static inline
3965 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3966 {
3967 	return __skb_gso_segment(skb, features, true);
3968 }
3969 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3970 
3971 static inline bool can_checksum_protocol(netdev_features_t features,
3972 					 __be16 protocol)
3973 {
3974 	if (protocol == htons(ETH_P_FCOE))
3975 		return !!(features & NETIF_F_FCOE_CRC);
3976 
3977 	/* Assume this is an IP checksum (not SCTP CRC) */
3978 
3979 	if (features & NETIF_F_HW_CSUM) {
3980 		/* Can checksum everything */
3981 		return true;
3982 	}
3983 
3984 	switch (protocol) {
3985 	case htons(ETH_P_IP):
3986 		return !!(features & NETIF_F_IP_CSUM);
3987 	case htons(ETH_P_IPV6):
3988 		return !!(features & NETIF_F_IPV6_CSUM);
3989 	default:
3990 		return false;
3991 	}
3992 }
3993 
3994 #ifdef CONFIG_BUG
3995 void netdev_rx_csum_fault(struct net_device *dev);
3996 #else
3997 static inline void netdev_rx_csum_fault(struct net_device *dev)
3998 {
3999 }
4000 #endif
4001 /* rx skb timestamps */
4002 void net_enable_timestamp(void);
4003 void net_disable_timestamp(void);
4004 
4005 #ifdef CONFIG_PROC_FS
4006 int __init dev_proc_init(void);
4007 #else
4008 #define dev_proc_init() 0
4009 #endif
4010 
4011 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4012 					      struct sk_buff *skb, struct net_device *dev,
4013 					      bool more)
4014 {
4015 	skb->xmit_more = more ? 1 : 0;
4016 	return ops->ndo_start_xmit(skb, dev);
4017 }
4018 
4019 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4020 					    struct netdev_queue *txq, bool more)
4021 {
4022 	const struct net_device_ops *ops = dev->netdev_ops;
4023 	int rc;
4024 
4025 	rc = __netdev_start_xmit(ops, skb, dev, more);
4026 	if (rc == NETDEV_TX_OK)
4027 		txq_trans_update(txq);
4028 
4029 	return rc;
4030 }
4031 
4032 int netdev_class_create_file_ns(struct class_attribute *class_attr,
4033 				const void *ns);
4034 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4035 				 const void *ns);
4036 
4037 static inline int netdev_class_create_file(struct class_attribute *class_attr)
4038 {
4039 	return netdev_class_create_file_ns(class_attr, NULL);
4040 }
4041 
4042 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4043 {
4044 	netdev_class_remove_file_ns(class_attr, NULL);
4045 }
4046 
4047 extern struct kobj_ns_type_operations net_ns_type_operations;
4048 
4049 const char *netdev_drivername(const struct net_device *dev);
4050 
4051 void linkwatch_run_queue(void);
4052 
4053 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4054 							  netdev_features_t f2)
4055 {
4056 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4057 		if (f1 & NETIF_F_HW_CSUM)
4058 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4059 		else
4060 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4061 	}
4062 
4063 	return f1 & f2;
4064 }
4065 
4066 static inline netdev_features_t netdev_get_wanted_features(
4067 	struct net_device *dev)
4068 {
4069 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4070 }
4071 netdev_features_t netdev_increment_features(netdev_features_t all,
4072 	netdev_features_t one, netdev_features_t mask);
4073 
4074 /* Allow TSO being used on stacked device :
4075  * Performing the GSO segmentation before last device
4076  * is a performance improvement.
4077  */
4078 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4079 							netdev_features_t mask)
4080 {
4081 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4082 }
4083 
4084 int __netdev_update_features(struct net_device *dev);
4085 void netdev_update_features(struct net_device *dev);
4086 void netdev_change_features(struct net_device *dev);
4087 
4088 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4089 					struct net_device *dev);
4090 
4091 netdev_features_t passthru_features_check(struct sk_buff *skb,
4092 					  struct net_device *dev,
4093 					  netdev_features_t features);
4094 netdev_features_t netif_skb_features(struct sk_buff *skb);
4095 
4096 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4097 {
4098 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4099 
4100 	/* check flags correspondence */
4101 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4102 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4103 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4104 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4105 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4106 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4107 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4108 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4109 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4110 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4111 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4112 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4113 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4114 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4115 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4116 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4117 
4118 	return (features & feature) == feature;
4119 }
4120 
4121 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4122 {
4123 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4124 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4125 }
4126 
4127 static inline bool netif_needs_gso(struct sk_buff *skb,
4128 				   netdev_features_t features)
4129 {
4130 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4131 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4132 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4133 }
4134 
4135 static inline void netif_set_gso_max_size(struct net_device *dev,
4136 					  unsigned int size)
4137 {
4138 	dev->gso_max_size = size;
4139 }
4140 
4141 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4142 					int pulled_hlen, u16 mac_offset,
4143 					int mac_len)
4144 {
4145 	skb->protocol = protocol;
4146 	skb->encapsulation = 1;
4147 	skb_push(skb, pulled_hlen);
4148 	skb_reset_transport_header(skb);
4149 	skb->mac_header = mac_offset;
4150 	skb->network_header = skb->mac_header + mac_len;
4151 	skb->mac_len = mac_len;
4152 }
4153 
4154 static inline bool netif_is_macsec(const struct net_device *dev)
4155 {
4156 	return dev->priv_flags & IFF_MACSEC;
4157 }
4158 
4159 static inline bool netif_is_macvlan(const struct net_device *dev)
4160 {
4161 	return dev->priv_flags & IFF_MACVLAN;
4162 }
4163 
4164 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4165 {
4166 	return dev->priv_flags & IFF_MACVLAN_PORT;
4167 }
4168 
4169 static inline bool netif_is_ipvlan(const struct net_device *dev)
4170 {
4171 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4172 }
4173 
4174 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4175 {
4176 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4177 }
4178 
4179 static inline bool netif_is_bond_master(const struct net_device *dev)
4180 {
4181 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4182 }
4183 
4184 static inline bool netif_is_bond_slave(const struct net_device *dev)
4185 {
4186 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4187 }
4188 
4189 static inline bool netif_supports_nofcs(struct net_device *dev)
4190 {
4191 	return dev->priv_flags & IFF_SUPP_NOFCS;
4192 }
4193 
4194 static inline bool netif_is_l3_master(const struct net_device *dev)
4195 {
4196 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4197 }
4198 
4199 static inline bool netif_is_l3_slave(const struct net_device *dev)
4200 {
4201 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4202 }
4203 
4204 static inline bool netif_is_bridge_master(const struct net_device *dev)
4205 {
4206 	return dev->priv_flags & IFF_EBRIDGE;
4207 }
4208 
4209 static inline bool netif_is_bridge_port(const struct net_device *dev)
4210 {
4211 	return dev->priv_flags & IFF_BRIDGE_PORT;
4212 }
4213 
4214 static inline bool netif_is_ovs_master(const struct net_device *dev)
4215 {
4216 	return dev->priv_flags & IFF_OPENVSWITCH;
4217 }
4218 
4219 static inline bool netif_is_ovs_port(const struct net_device *dev)
4220 {
4221 	return dev->priv_flags & IFF_OVS_DATAPATH;
4222 }
4223 
4224 static inline bool netif_is_team_master(const struct net_device *dev)
4225 {
4226 	return dev->priv_flags & IFF_TEAM;
4227 }
4228 
4229 static inline bool netif_is_team_port(const struct net_device *dev)
4230 {
4231 	return dev->priv_flags & IFF_TEAM_PORT;
4232 }
4233 
4234 static inline bool netif_is_lag_master(const struct net_device *dev)
4235 {
4236 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4237 }
4238 
4239 static inline bool netif_is_lag_port(const struct net_device *dev)
4240 {
4241 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4242 }
4243 
4244 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4245 {
4246 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4247 }
4248 
4249 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4250 static inline void netif_keep_dst(struct net_device *dev)
4251 {
4252 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4253 }
4254 
4255 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4256 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4257 {
4258 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4259 	return dev->priv_flags & IFF_MACSEC;
4260 }
4261 
4262 extern struct pernet_operations __net_initdata loopback_net_ops;
4263 
4264 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4265 
4266 /* netdev_printk helpers, similar to dev_printk */
4267 
4268 static inline const char *netdev_name(const struct net_device *dev)
4269 {
4270 	if (!dev->name[0] || strchr(dev->name, '%'))
4271 		return "(unnamed net_device)";
4272 	return dev->name;
4273 }
4274 
4275 static inline bool netdev_unregistering(const struct net_device *dev)
4276 {
4277 	return dev->reg_state == NETREG_UNREGISTERING;
4278 }
4279 
4280 static inline const char *netdev_reg_state(const struct net_device *dev)
4281 {
4282 	switch (dev->reg_state) {
4283 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4284 	case NETREG_REGISTERED: return "";
4285 	case NETREG_UNREGISTERING: return " (unregistering)";
4286 	case NETREG_UNREGISTERED: return " (unregistered)";
4287 	case NETREG_RELEASED: return " (released)";
4288 	case NETREG_DUMMY: return " (dummy)";
4289 	}
4290 
4291 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4292 	return " (unknown)";
4293 }
4294 
4295 __printf(3, 4)
4296 void netdev_printk(const char *level, const struct net_device *dev,
4297 		   const char *format, ...);
4298 __printf(2, 3)
4299 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4300 __printf(2, 3)
4301 void netdev_alert(const struct net_device *dev, const char *format, ...);
4302 __printf(2, 3)
4303 void netdev_crit(const struct net_device *dev, const char *format, ...);
4304 __printf(2, 3)
4305 void netdev_err(const struct net_device *dev, const char *format, ...);
4306 __printf(2, 3)
4307 void netdev_warn(const struct net_device *dev, const char *format, ...);
4308 __printf(2, 3)
4309 void netdev_notice(const struct net_device *dev, const char *format, ...);
4310 __printf(2, 3)
4311 void netdev_info(const struct net_device *dev, const char *format, ...);
4312 
4313 #define MODULE_ALIAS_NETDEV(device) \
4314 	MODULE_ALIAS("netdev-" device)
4315 
4316 #if defined(CONFIG_DYNAMIC_DEBUG)
4317 #define netdev_dbg(__dev, format, args...)			\
4318 do {								\
4319 	dynamic_netdev_dbg(__dev, format, ##args);		\
4320 } while (0)
4321 #elif defined(DEBUG)
4322 #define netdev_dbg(__dev, format, args...)			\
4323 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4324 #else
4325 #define netdev_dbg(__dev, format, args...)			\
4326 ({								\
4327 	if (0)							\
4328 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4329 })
4330 #endif
4331 
4332 #if defined(VERBOSE_DEBUG)
4333 #define netdev_vdbg	netdev_dbg
4334 #else
4335 
4336 #define netdev_vdbg(dev, format, args...)			\
4337 ({								\
4338 	if (0)							\
4339 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4340 	0;							\
4341 })
4342 #endif
4343 
4344 /*
4345  * netdev_WARN() acts like dev_printk(), but with the key difference
4346  * of using a WARN/WARN_ON to get the message out, including the
4347  * file/line information and a backtrace.
4348  */
4349 #define netdev_WARN(dev, format, args...)			\
4350 	WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),	\
4351 	     netdev_reg_state(dev), ##args)
4352 
4353 /* netif printk helpers, similar to netdev_printk */
4354 
4355 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4356 do {					  			\
4357 	if (netif_msg_##type(priv))				\
4358 		netdev_printk(level, (dev), fmt, ##args);	\
4359 } while (0)
4360 
4361 #define netif_level(level, priv, type, dev, fmt, args...)	\
4362 do {								\
4363 	if (netif_msg_##type(priv))				\
4364 		netdev_##level(dev, fmt, ##args);		\
4365 } while (0)
4366 
4367 #define netif_emerg(priv, type, dev, fmt, args...)		\
4368 	netif_level(emerg, priv, type, dev, fmt, ##args)
4369 #define netif_alert(priv, type, dev, fmt, args...)		\
4370 	netif_level(alert, priv, type, dev, fmt, ##args)
4371 #define netif_crit(priv, type, dev, fmt, args...)		\
4372 	netif_level(crit, priv, type, dev, fmt, ##args)
4373 #define netif_err(priv, type, dev, fmt, args...)		\
4374 	netif_level(err, priv, type, dev, fmt, ##args)
4375 #define netif_warn(priv, type, dev, fmt, args...)		\
4376 	netif_level(warn, priv, type, dev, fmt, ##args)
4377 #define netif_notice(priv, type, dev, fmt, args...)		\
4378 	netif_level(notice, priv, type, dev, fmt, ##args)
4379 #define netif_info(priv, type, dev, fmt, args...)		\
4380 	netif_level(info, priv, type, dev, fmt, ##args)
4381 
4382 #if defined(CONFIG_DYNAMIC_DEBUG)
4383 #define netif_dbg(priv, type, netdev, format, args...)		\
4384 do {								\
4385 	if (netif_msg_##type(priv))				\
4386 		dynamic_netdev_dbg(netdev, format, ##args);	\
4387 } while (0)
4388 #elif defined(DEBUG)
4389 #define netif_dbg(priv, type, dev, format, args...)		\
4390 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4391 #else
4392 #define netif_dbg(priv, type, dev, format, args...)			\
4393 ({									\
4394 	if (0)								\
4395 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4396 	0;								\
4397 })
4398 #endif
4399 
4400 /* if @cond then downgrade to debug, else print at @level */
4401 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4402 	do {                                                              \
4403 		if (cond)                                                 \
4404 			netif_dbg(priv, type, netdev, fmt, ##args);       \
4405 		else                                                      \
4406 			netif_ ## level(priv, type, netdev, fmt, ##args); \
4407 	} while (0)
4408 
4409 #if defined(VERBOSE_DEBUG)
4410 #define netif_vdbg	netif_dbg
4411 #else
4412 #define netif_vdbg(priv, type, dev, format, args...)		\
4413 ({								\
4414 	if (0)							\
4415 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4416 	0;							\
4417 })
4418 #endif
4419 
4420 /*
4421  *	The list of packet types we will receive (as opposed to discard)
4422  *	and the routines to invoke.
4423  *
4424  *	Why 16. Because with 16 the only overlap we get on a hash of the
4425  *	low nibble of the protocol value is RARP/SNAP/X.25.
4426  *
4427  *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4428  *             sure which should go first, but I bet it won't make much
4429  *             difference if we are running VLANs.  The good news is that
4430  *             this protocol won't be in the list unless compiled in, so
4431  *             the average user (w/out VLANs) will not be adversely affected.
4432  *             --BLG
4433  *
4434  *		0800	IP
4435  *		8100    802.1Q VLAN
4436  *		0001	802.3
4437  *		0002	AX.25
4438  *		0004	802.2
4439  *		8035	RARP
4440  *		0005	SNAP
4441  *		0805	X.25
4442  *		0806	ARP
4443  *		8137	IPX
4444  *		0009	Localtalk
4445  *		86DD	IPv6
4446  */
4447 #define PTYPE_HASH_SIZE	(16)
4448 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4449 
4450 #endif	/* _LINUX_NETDEVICE_H */
4451