xref: /openbmc/linux/include/linux/netdevice.h (revision 4f139972b489f8bc2c821aa25ac65018d92af3f7)
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, <waltje@uWalt.NL.Mugnet.ORG>
12  *		Corey Minyard <wf-rch!minyard@relay.EU.net>
13  *		Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14  *		Alan Cox, <alan@lxorguk.ukuu.org.uk>
15  *		Bjorn Ekwall. <bj0rn@blox.se>
16  *              Pekka Riikonen <priikone@poseidon.pspt.fi>
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 
70 void netdev_set_default_ethtool_ops(struct net_device *dev,
71 				    const struct ethtool_ops *ops);
72 
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
75 #define NET_RX_DROP		1	/* packet dropped */
76 
77 /*
78  * Transmit return codes: transmit return codes originate from three different
79  * namespaces:
80  *
81  * - qdisc return codes
82  * - driver transmit return codes
83  * - errno values
84  *
85  * Drivers are allowed to return any one of those in their hard_start_xmit()
86  * function. Real network devices commonly used with qdiscs should only return
87  * the driver transmit return codes though - when qdiscs are used, the actual
88  * transmission happens asynchronously, so the value is not propagated to
89  * higher layers. Virtual network devices transmit synchronously; in this case
90  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
91  * others are propagated to higher layers.
92  */
93 
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS	0x00
96 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
97 #define NET_XMIT_CN		0x02	/* congestion notification	*/
98 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
99 
100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101  * indicates that the device will soon be dropping packets, or already drops
102  * some packets of the same priority; prompting us to send less aggressively. */
103 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
104 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105 
106 /* Driver transmit return codes */
107 #define NETDEV_TX_MASK		0xf0
108 
109 enum netdev_tx {
110 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
111 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
112 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
113 };
114 typedef enum netdev_tx netdev_tx_t;
115 
116 /*
117  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119  */
120 static inline bool dev_xmit_complete(int rc)
121 {
122 	/*
123 	 * Positive cases with an skb consumed by a driver:
124 	 * - successful transmission (rc == NETDEV_TX_OK)
125 	 * - error while transmitting (rc < 0)
126 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 	 */
128 	if (likely(rc < NET_XMIT_MASK))
129 		return true;
130 
131 	return false;
132 }
133 
134 /*
135  *	Compute the worst-case header length according to the protocols
136  *	used.
137  */
138 
139 #if defined(CONFIG_HYPERV_NET)
140 # define LL_MAX_HEADER 128
141 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 #  define LL_MAX_HEADER 128
144 # else
145 #  define LL_MAX_HEADER 96
146 # endif
147 #else
148 # define LL_MAX_HEADER 32
149 #endif
150 
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
154 #else
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
156 #endif
157 
158 /*
159  *	Old network device statistics. Fields are native words
160  *	(unsigned long) so they can be read and written atomically.
161  */
162 
163 struct net_device_stats {
164 	unsigned long	rx_packets;
165 	unsigned long	tx_packets;
166 	unsigned long	rx_bytes;
167 	unsigned long	tx_bytes;
168 	unsigned long	rx_errors;
169 	unsigned long	tx_errors;
170 	unsigned long	rx_dropped;
171 	unsigned long	tx_dropped;
172 	unsigned long	multicast;
173 	unsigned long	collisions;
174 	unsigned long	rx_length_errors;
175 	unsigned long	rx_over_errors;
176 	unsigned long	rx_crc_errors;
177 	unsigned long	rx_frame_errors;
178 	unsigned long	rx_fifo_errors;
179 	unsigned long	rx_missed_errors;
180 	unsigned long	tx_aborted_errors;
181 	unsigned long	tx_carrier_errors;
182 	unsigned long	tx_fifo_errors;
183 	unsigned long	tx_heartbeat_errors;
184 	unsigned long	tx_window_errors;
185 	unsigned long	rx_compressed;
186 	unsigned long	tx_compressed;
187 };
188 
189 
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
192 
193 #ifdef CONFIG_RPS
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed;
196 extern struct static_key rfs_needed;
197 #endif
198 
199 struct neighbour;
200 struct neigh_parms;
201 struct sk_buff;
202 
203 struct netdev_hw_addr {
204 	struct list_head	list;
205 	unsigned char		addr[MAX_ADDR_LEN];
206 	unsigned char		type;
207 #define NETDEV_HW_ADDR_T_LAN		1
208 #define NETDEV_HW_ADDR_T_SAN		2
209 #define NETDEV_HW_ADDR_T_SLAVE		3
210 #define NETDEV_HW_ADDR_T_UNICAST	4
211 #define NETDEV_HW_ADDR_T_MULTICAST	5
212 	bool			global_use;
213 	int			sync_cnt;
214 	int			refcount;
215 	int			synced;
216 	struct rcu_head		rcu_head;
217 };
218 
219 struct netdev_hw_addr_list {
220 	struct list_head	list;
221 	int			count;
222 };
223 
224 #define netdev_hw_addr_list_count(l) ((l)->count)
225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226 #define netdev_hw_addr_list_for_each(ha, l) \
227 	list_for_each_entry(ha, &(l)->list, list)
228 
229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231 #define netdev_for_each_uc_addr(ha, dev) \
232 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
233 
234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
236 #define netdev_for_each_mc_addr(ha, dev) \
237 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
238 
239 struct hh_cache {
240 	u16		hh_len;
241 	u16		__pad;
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 	/* Check if a bpf program is set on the device.  The callee should
812 	 * return true if a program is currently attached and running.
813 	 */
814 	XDP_QUERY_PROG,
815 };
816 
817 struct netdev_xdp {
818 	enum xdp_netdev_command command;
819 	union {
820 		/* XDP_SETUP_PROG */
821 		struct bpf_prog *prog;
822 		/* XDP_QUERY_PROG */
823 		bool prog_attached;
824 	};
825 };
826 
827 /*
828  * This structure defines the management hooks for network devices.
829  * The following hooks can be defined; unless noted otherwise, they are
830  * optional and can be filled with a null pointer.
831  *
832  * int (*ndo_init)(struct net_device *dev);
833  *     This function is called once when a network device is registered.
834  *     The network device can use this for any late stage initialization
835  *     or semantic validation. It can fail with an error code which will
836  *     be propagated back to register_netdev.
837  *
838  * void (*ndo_uninit)(struct net_device *dev);
839  *     This function is called when device is unregistered or when registration
840  *     fails. It is not called if init fails.
841  *
842  * int (*ndo_open)(struct net_device *dev);
843  *     This function is called when a network device transitions to the up
844  *     state.
845  *
846  * int (*ndo_stop)(struct net_device *dev);
847  *     This function is called when a network device transitions to the down
848  *     state.
849  *
850  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
851  *                               struct net_device *dev);
852  *	Called when a packet needs to be transmitted.
853  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
854  *	the queue before that can happen; it's for obsolete devices and weird
855  *	corner cases, but the stack really does a non-trivial amount
856  *	of useless work if you return NETDEV_TX_BUSY.
857  *	Required; cannot be NULL.
858  *
859  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
860  *					   struct net_device *dev
861  *					   netdev_features_t features);
862  *	Called by core transmit path to determine if device is capable of
863  *	performing offload operations on a given packet. This is to give
864  *	the device an opportunity to implement any restrictions that cannot
865  *	be otherwise expressed by feature flags. The check is called with
866  *	the set of features that the stack has calculated and it returns
867  *	those the driver believes to be appropriate.
868  *
869  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
870  *                         void *accel_priv, select_queue_fallback_t fallback);
871  *	Called to decide which queue to use when device supports multiple
872  *	transmit queues.
873  *
874  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
875  *	This function is called to allow device receiver to make
876  *	changes to configuration when multicast or promiscuous is enabled.
877  *
878  * void (*ndo_set_rx_mode)(struct net_device *dev);
879  *	This function is called device changes address list filtering.
880  *	If driver handles unicast address filtering, it should set
881  *	IFF_UNICAST_FLT in its priv_flags.
882  *
883  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
884  *	This function  is called when the Media Access Control address
885  *	needs to be changed. If this interface is not defined, the
886  *	MAC address can not be changed.
887  *
888  * int (*ndo_validate_addr)(struct net_device *dev);
889  *	Test if Media Access Control address is valid for the device.
890  *
891  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
892  *	Called when a user requests an ioctl which can't be handled by
893  *	the generic interface code. If not defined ioctls return
894  *	not supported error code.
895  *
896  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
897  *	Used to set network devices bus interface parameters. This interface
898  *	is retained for legacy reasons; new devices should use the bus
899  *	interface (PCI) for low level management.
900  *
901  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
902  *	Called when a user wants to change the Maximum Transfer Unit
903  *	of a device. If not defined, any request to change MTU will
904  *	will return an error.
905  *
906  * void (*ndo_tx_timeout)(struct net_device *dev);
907  *	Callback used when the transmitter has not made any progress
908  *	for dev->watchdog ticks.
909  *
910  * void (*ndo_get_stats64)(struct net_device *dev,
911  *                         struct rtnl_link_stats64 *storage);
912  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
913  *	Called when a user wants to get the network device usage
914  *	statistics. Drivers must do one of the following:
915  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
916  *	   rtnl_link_stats64 structure passed by the caller.
917  *	2. Define @ndo_get_stats to update a net_device_stats structure
918  *	   (which should normally be dev->stats) and return a pointer to
919  *	   it. The structure may be changed asynchronously only if each
920  *	   field is written atomically.
921  *	3. Update dev->stats asynchronously and atomically, and define
922  *	   neither operation.
923  *
924  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
925  *	Return true if this device supports offload stats of this attr_id.
926  *
927  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
928  *	void *attr_data)
929  *	Get statistics for offload operations by attr_id. Write it into the
930  *	attr_data pointer.
931  *
932  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
933  *	If device supports VLAN filtering this function is called when a
934  *	VLAN id is registered.
935  *
936  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
937  *	If device supports VLAN filtering this function is called when a
938  *	VLAN id is unregistered.
939  *
940  * void (*ndo_poll_controller)(struct net_device *dev);
941  *
942  *	SR-IOV management functions.
943  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
944  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
945  *			  u8 qos, __be16 proto);
946  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
947  *			  int max_tx_rate);
948  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
949  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
950  * int (*ndo_get_vf_config)(struct net_device *dev,
951  *			    int vf, struct ifla_vf_info *ivf);
952  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
953  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
954  *			  struct nlattr *port[]);
955  *
956  *      Enable or disable the VF ability to query its RSS Redirection Table and
957  *      Hash Key. This is needed since on some devices VF share this information
958  *      with PF and querying it may introduce a theoretical security risk.
959  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
960  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
961  * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
962  *		       __be16 protocol, struct tc_to_netdev *tc);
963  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
964  *	This is always called from the stack with the rtnl lock held and netif
965  *	tx queues stopped. This allows the netdevice to perform queue
966  *	management safely.
967  *
968  *	Fiber Channel over Ethernet (FCoE) offload functions.
969  * int (*ndo_fcoe_enable)(struct net_device *dev);
970  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
971  *	so the underlying device can perform whatever needed configuration or
972  *	initialization to support acceleration of FCoE traffic.
973  *
974  * int (*ndo_fcoe_disable)(struct net_device *dev);
975  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
976  *	so the underlying device can perform whatever needed clean-ups to
977  *	stop supporting acceleration of FCoE traffic.
978  *
979  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
980  *			     struct scatterlist *sgl, unsigned int sgc);
981  *	Called when the FCoE Initiator wants to initialize an I/O that
982  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
983  *	perform necessary setup and returns 1 to indicate the device is set up
984  *	successfully to perform DDP on this I/O, otherwise this returns 0.
985  *
986  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
987  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
988  *	indicated by the FC exchange id 'xid', so the underlying device can
989  *	clean up and reuse resources for later DDP requests.
990  *
991  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
992  *			      struct scatterlist *sgl, unsigned int sgc);
993  *	Called when the FCoE Target wants to initialize an I/O that
994  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
995  *	perform necessary setup and returns 1 to indicate the device is set up
996  *	successfully to perform DDP on this I/O, otherwise this returns 0.
997  *
998  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
999  *			       struct netdev_fcoe_hbainfo *hbainfo);
1000  *	Called when the FCoE Protocol stack wants information on the underlying
1001  *	device. This information is utilized by the FCoE protocol stack to
1002  *	register attributes with Fiber Channel management service as per the
1003  *	FC-GS Fabric Device Management Information(FDMI) specification.
1004  *
1005  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1006  *	Called when the underlying device wants to override default World Wide
1007  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1008  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1009  *	protocol stack to use.
1010  *
1011  *	RFS acceleration.
1012  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1013  *			    u16 rxq_index, u32 flow_id);
1014  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1015  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1016  *	Return the filter ID on success, or a negative error code.
1017  *
1018  *	Slave management functions (for bridge, bonding, etc).
1019  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1020  *	Called to make another netdev an underling.
1021  *
1022  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1023  *	Called to release previously enslaved netdev.
1024  *
1025  *      Feature/offload setting functions.
1026  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1027  *		netdev_features_t features);
1028  *	Adjusts the requested feature flags according to device-specific
1029  *	constraints, and returns the resulting flags. Must not modify
1030  *	the device state.
1031  *
1032  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1033  *	Called to update device configuration to new features. Passed
1034  *	feature set might be less than what was returned by ndo_fix_features()).
1035  *	Must return >0 or -errno if it changed dev->features itself.
1036  *
1037  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1038  *		      struct net_device *dev,
1039  *		      const unsigned char *addr, u16 vid, u16 flags)
1040  *	Adds an FDB entry to dev for addr.
1041  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1042  *		      struct net_device *dev,
1043  *		      const unsigned char *addr, u16 vid)
1044  *	Deletes the FDB entry from dev coresponding to addr.
1045  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1046  *		       struct net_device *dev, struct net_device *filter_dev,
1047  *		       int *idx)
1048  *	Used to add FDB entries to dump requests. Implementers should add
1049  *	entries to skb and update idx with the number of entries.
1050  *
1051  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1052  *			     u16 flags)
1053  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1054  *			     struct net_device *dev, u32 filter_mask,
1055  *			     int nlflags)
1056  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1057  *			     u16 flags);
1058  *
1059  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1060  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1061  *	which do not represent real hardware may define this to allow their
1062  *	userspace components to manage their virtual carrier state. Devices
1063  *	that determine carrier state from physical hardware properties (eg
1064  *	network cables) or protocol-dependent mechanisms (eg
1065  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1066  *
1067  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1068  *			       struct netdev_phys_item_id *ppid);
1069  *	Called to get ID of physical port of this device. If driver does
1070  *	not implement this, it is assumed that the hw is not able to have
1071  *	multiple net devices on single physical port.
1072  *
1073  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1074  *			      struct udp_tunnel_info *ti);
1075  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1076  *	address family that a UDP tunnel is listnening to. It is called only
1077  *	when a new port starts listening. The operation is protected by the
1078  *	RTNL.
1079  *
1080  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1081  *			      struct udp_tunnel_info *ti);
1082  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1083  *	address family that the UDP tunnel is not listening to anymore. The
1084  *	operation is protected by the RTNL.
1085  *
1086  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1087  *				 struct net_device *dev)
1088  *	Called by upper layer devices to accelerate switching or other
1089  *	station functionality into hardware. 'pdev is the lowerdev
1090  *	to use for the offload and 'dev' is the net device that will
1091  *	back the offload. Returns a pointer to the private structure
1092  *	the upper layer will maintain.
1093  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1094  *	Called by upper layer device to delete the station created
1095  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1096  *	the station and priv is the structure returned by the add
1097  *	operation.
1098  * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1099  *				      struct net_device *dev,
1100  *				      void *priv);
1101  *	Callback to use for xmit over the accelerated station. This
1102  *	is used in place of ndo_start_xmit on accelerated net
1103  *	devices.
1104  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1105  *			     int queue_index, u32 maxrate);
1106  *	Called when a user wants to set a max-rate limitation of specific
1107  *	TX queue.
1108  * int (*ndo_get_iflink)(const struct net_device *dev);
1109  *	Called to get the iflink value of this device.
1110  * void (*ndo_change_proto_down)(struct net_device *dev,
1111  *				 bool proto_down);
1112  *	This function is used to pass protocol port error state information
1113  *	to the switch driver. The switch driver can react to the proto_down
1114  *      by doing a phys down on the associated switch port.
1115  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1116  *	This function is used to get egress tunnel information for given skb.
1117  *	This is useful for retrieving outer tunnel header parameters while
1118  *	sampling packet.
1119  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1120  *	This function is used to specify the headroom that the skb must
1121  *	consider when allocation skb during packet reception. Setting
1122  *	appropriate rx headroom value allows avoiding skb head copy on
1123  *	forward. Setting a negative value resets the rx headroom to the
1124  *	default value.
1125  * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1126  *	This function is used to set or query state related to XDP on the
1127  *	netdevice. See definition of enum xdp_netdev_command for details.
1128  *
1129  */
1130 struct net_device_ops {
1131 	int			(*ndo_init)(struct net_device *dev);
1132 	void			(*ndo_uninit)(struct net_device *dev);
1133 	int			(*ndo_open)(struct net_device *dev);
1134 	int			(*ndo_stop)(struct net_device *dev);
1135 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1136 						  struct net_device *dev);
1137 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1138 						      struct net_device *dev,
1139 						      netdev_features_t features);
1140 	u16			(*ndo_select_queue)(struct net_device *dev,
1141 						    struct sk_buff *skb,
1142 						    void *accel_priv,
1143 						    select_queue_fallback_t fallback);
1144 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1145 						       int flags);
1146 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1147 	int			(*ndo_set_mac_address)(struct net_device *dev,
1148 						       void *addr);
1149 	int			(*ndo_validate_addr)(struct net_device *dev);
1150 	int			(*ndo_do_ioctl)(struct net_device *dev,
1151 					        struct ifreq *ifr, int cmd);
1152 	int			(*ndo_set_config)(struct net_device *dev,
1153 					          struct ifmap *map);
1154 	int			(*ndo_change_mtu)(struct net_device *dev,
1155 						  int new_mtu);
1156 	int			(*ndo_neigh_setup)(struct net_device *dev,
1157 						   struct neigh_parms *);
1158 	void			(*ndo_tx_timeout) (struct net_device *dev);
1159 
1160 	void			(*ndo_get_stats64)(struct net_device *dev,
1161 						   struct rtnl_link_stats64 *storage);
1162 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1163 	int			(*ndo_get_offload_stats)(int attr_id,
1164 							 const struct net_device *dev,
1165 							 void *attr_data);
1166 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1167 
1168 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1169 						       __be16 proto, u16 vid);
1170 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1171 						        __be16 proto, u16 vid);
1172 #ifdef CONFIG_NET_POLL_CONTROLLER
1173 	void                    (*ndo_poll_controller)(struct net_device *dev);
1174 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1175 						     struct netpoll_info *info);
1176 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1177 #endif
1178 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1179 						  int queue, u8 *mac);
1180 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1181 						   int queue, u16 vlan,
1182 						   u8 qos, __be16 proto);
1183 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1184 						   int vf, int min_tx_rate,
1185 						   int max_tx_rate);
1186 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1187 						       int vf, bool setting);
1188 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1189 						    int vf, bool setting);
1190 	int			(*ndo_get_vf_config)(struct net_device *dev,
1191 						     int vf,
1192 						     struct ifla_vf_info *ivf);
1193 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1194 							 int vf, int link_state);
1195 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1196 						    int vf,
1197 						    struct ifla_vf_stats
1198 						    *vf_stats);
1199 	int			(*ndo_set_vf_port)(struct net_device *dev,
1200 						   int vf,
1201 						   struct nlattr *port[]);
1202 	int			(*ndo_get_vf_port)(struct net_device *dev,
1203 						   int vf, struct sk_buff *skb);
1204 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1205 						   int vf, u64 guid,
1206 						   int guid_type);
1207 	int			(*ndo_set_vf_rss_query_en)(
1208 						   struct net_device *dev,
1209 						   int vf, bool setting);
1210 	int			(*ndo_setup_tc)(struct net_device *dev,
1211 						u32 handle,
1212 						__be16 protocol,
1213 						struct tc_to_netdev *tc);
1214 #if IS_ENABLED(CONFIG_FCOE)
1215 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1216 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1217 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1218 						      u16 xid,
1219 						      struct scatterlist *sgl,
1220 						      unsigned int sgc);
1221 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1222 						     u16 xid);
1223 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1224 						       u16 xid,
1225 						       struct scatterlist *sgl,
1226 						       unsigned int sgc);
1227 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1228 							struct netdev_fcoe_hbainfo *hbainfo);
1229 #endif
1230 
1231 #if IS_ENABLED(CONFIG_LIBFCOE)
1232 #define NETDEV_FCOE_WWNN 0
1233 #define NETDEV_FCOE_WWPN 1
1234 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1235 						    u64 *wwn, int type);
1236 #endif
1237 
1238 #ifdef CONFIG_RFS_ACCEL
1239 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1240 						     const struct sk_buff *skb,
1241 						     u16 rxq_index,
1242 						     u32 flow_id);
1243 #endif
1244 	int			(*ndo_add_slave)(struct net_device *dev,
1245 						 struct net_device *slave_dev);
1246 	int			(*ndo_del_slave)(struct net_device *dev,
1247 						 struct net_device *slave_dev);
1248 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1249 						    netdev_features_t features);
1250 	int			(*ndo_set_features)(struct net_device *dev,
1251 						    netdev_features_t features);
1252 	int			(*ndo_neigh_construct)(struct net_device *dev,
1253 						       struct neighbour *n);
1254 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1255 						     struct neighbour *n);
1256 
1257 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1258 					       struct nlattr *tb[],
1259 					       struct net_device *dev,
1260 					       const unsigned char *addr,
1261 					       u16 vid,
1262 					       u16 flags);
1263 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1264 					       struct nlattr *tb[],
1265 					       struct net_device *dev,
1266 					       const unsigned char *addr,
1267 					       u16 vid);
1268 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1269 						struct netlink_callback *cb,
1270 						struct net_device *dev,
1271 						struct net_device *filter_dev,
1272 						int *idx);
1273 
1274 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1275 						      struct nlmsghdr *nlh,
1276 						      u16 flags);
1277 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1278 						      u32 pid, u32 seq,
1279 						      struct net_device *dev,
1280 						      u32 filter_mask,
1281 						      int nlflags);
1282 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1283 						      struct nlmsghdr *nlh,
1284 						      u16 flags);
1285 	int			(*ndo_change_carrier)(struct net_device *dev,
1286 						      bool new_carrier);
1287 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1288 							struct netdev_phys_item_id *ppid);
1289 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1290 							  char *name, size_t len);
1291 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1292 						      struct udp_tunnel_info *ti);
1293 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1294 						      struct udp_tunnel_info *ti);
1295 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1296 							struct net_device *dev);
1297 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1298 							void *priv);
1299 
1300 	netdev_tx_t		(*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1301 							struct net_device *dev,
1302 							void *priv);
1303 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1304 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1305 						      int queue_index,
1306 						      u32 maxrate);
1307 	int			(*ndo_get_iflink)(const struct net_device *dev);
1308 	int			(*ndo_change_proto_down)(struct net_device *dev,
1309 							 bool proto_down);
1310 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1311 						       struct sk_buff *skb);
1312 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1313 						       int needed_headroom);
1314 	int			(*ndo_xdp)(struct net_device *dev,
1315 					   struct netdev_xdp *xdp);
1316 };
1317 
1318 /**
1319  * enum net_device_priv_flags - &struct net_device priv_flags
1320  *
1321  * These are the &struct net_device, they are only set internally
1322  * by drivers and used in the kernel. These flags are invisible to
1323  * userspace; this means that the order of these flags can change
1324  * during any kernel release.
1325  *
1326  * You should have a pretty good reason to be extending these flags.
1327  *
1328  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1329  * @IFF_EBRIDGE: Ethernet bridging device
1330  * @IFF_BONDING: bonding master or slave
1331  * @IFF_ISATAP: ISATAP interface (RFC4214)
1332  * @IFF_WAN_HDLC: WAN HDLC device
1333  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1334  *	release skb->dst
1335  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1336  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1337  * @IFF_MACVLAN_PORT: device used as macvlan port
1338  * @IFF_BRIDGE_PORT: device used as bridge port
1339  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1340  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1341  * @IFF_UNICAST_FLT: Supports unicast filtering
1342  * @IFF_TEAM_PORT: device used as team port
1343  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1344  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1345  *	change when it's running
1346  * @IFF_MACVLAN: Macvlan device
1347  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1348  *	underlying stacked devices
1349  * @IFF_IPVLAN_MASTER: IPvlan master device
1350  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1351  * @IFF_L3MDEV_MASTER: device is an L3 master device
1352  * @IFF_NO_QUEUE: device can run without qdisc attached
1353  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1354  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1355  * @IFF_TEAM: device is a team device
1356  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1357  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1358  *	entity (i.e. the master device for bridged veth)
1359  * @IFF_MACSEC: device is a MACsec device
1360  */
1361 enum netdev_priv_flags {
1362 	IFF_802_1Q_VLAN			= 1<<0,
1363 	IFF_EBRIDGE			= 1<<1,
1364 	IFF_BONDING			= 1<<2,
1365 	IFF_ISATAP			= 1<<3,
1366 	IFF_WAN_HDLC			= 1<<4,
1367 	IFF_XMIT_DST_RELEASE		= 1<<5,
1368 	IFF_DONT_BRIDGE			= 1<<6,
1369 	IFF_DISABLE_NETPOLL		= 1<<7,
1370 	IFF_MACVLAN_PORT		= 1<<8,
1371 	IFF_BRIDGE_PORT			= 1<<9,
1372 	IFF_OVS_DATAPATH		= 1<<10,
1373 	IFF_TX_SKB_SHARING		= 1<<11,
1374 	IFF_UNICAST_FLT			= 1<<12,
1375 	IFF_TEAM_PORT			= 1<<13,
1376 	IFF_SUPP_NOFCS			= 1<<14,
1377 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1378 	IFF_MACVLAN			= 1<<16,
1379 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1380 	IFF_IPVLAN_MASTER		= 1<<18,
1381 	IFF_IPVLAN_SLAVE		= 1<<19,
1382 	IFF_L3MDEV_MASTER		= 1<<20,
1383 	IFF_NO_QUEUE			= 1<<21,
1384 	IFF_OPENVSWITCH			= 1<<22,
1385 	IFF_L3MDEV_SLAVE		= 1<<23,
1386 	IFF_TEAM			= 1<<24,
1387 	IFF_RXFH_CONFIGURED		= 1<<25,
1388 	IFF_PHONY_HEADROOM		= 1<<26,
1389 	IFF_MACSEC			= 1<<27,
1390 };
1391 
1392 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1393 #define IFF_EBRIDGE			IFF_EBRIDGE
1394 #define IFF_BONDING			IFF_BONDING
1395 #define IFF_ISATAP			IFF_ISATAP
1396 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1397 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1398 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1399 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1400 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1401 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1402 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1403 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1404 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1405 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1406 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1407 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1408 #define IFF_MACVLAN			IFF_MACVLAN
1409 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1410 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1411 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1412 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1413 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1414 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1415 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1416 #define IFF_TEAM			IFF_TEAM
1417 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1418 #define IFF_MACSEC			IFF_MACSEC
1419 
1420 /**
1421  *	struct net_device - The DEVICE structure.
1422  *		Actually, this whole structure is a big mistake.  It mixes I/O
1423  *		data with strictly "high-level" data, and it has to know about
1424  *		almost every data structure used in the INET module.
1425  *
1426  *	@name:	This is the first field of the "visible" part of this structure
1427  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1428  *	 	of the interface.
1429  *
1430  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1431  *	@ifalias:	SNMP alias
1432  *	@mem_end:	Shared memory end
1433  *	@mem_start:	Shared memory start
1434  *	@base_addr:	Device I/O address
1435  *	@irq:		Device IRQ number
1436  *
1437  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1438  *
1439  *	@state:		Generic network queuing layer state, see netdev_state_t
1440  *	@dev_list:	The global list of network devices
1441  *	@napi_list:	List entry used for polling NAPI devices
1442  *	@unreg_list:	List entry  when we are unregistering the
1443  *			device; see the function unregister_netdev
1444  *	@close_list:	List entry used when we are closing the device
1445  *	@ptype_all:     Device-specific packet handlers for all protocols
1446  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1447  *
1448  *	@adj_list:	Directly linked devices, like slaves for bonding
1449  *	@features:	Currently active device features
1450  *	@hw_features:	User-changeable features
1451  *
1452  *	@wanted_features:	User-requested features
1453  *	@vlan_features:		Mask of features inheritable by VLAN devices
1454  *
1455  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1456  *				This field indicates what encapsulation
1457  *				offloads the hardware is capable of doing,
1458  *				and drivers will need to set them appropriately.
1459  *
1460  *	@mpls_features:	Mask of features inheritable by MPLS
1461  *
1462  *	@ifindex:	interface index
1463  *	@group:		The group the device belongs to
1464  *
1465  *	@stats:		Statistics struct, which was left as a legacy, use
1466  *			rtnl_link_stats64 instead
1467  *
1468  *	@rx_dropped:	Dropped packets by core network,
1469  *			do not use this in drivers
1470  *	@tx_dropped:	Dropped packets by core network,
1471  *			do not use this in drivers
1472  *	@rx_nohandler:	nohandler dropped packets by core network on
1473  *			inactive devices, do not use this in drivers
1474  *
1475  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1476  *				instead of ioctl,
1477  *				see <net/iw_handler.h> for details.
1478  *	@wireless_data:	Instance data managed by the core of wireless extensions
1479  *
1480  *	@netdev_ops:	Includes several pointers to callbacks,
1481  *			if one wants to override the ndo_*() functions
1482  *	@ethtool_ops:	Management operations
1483  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1484  *			discovery handling. Necessary for e.g. 6LoWPAN.
1485  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1486  *			of Layer 2 headers.
1487  *
1488  *	@flags:		Interface flags (a la BSD)
1489  *	@priv_flags:	Like 'flags' but invisible to userspace,
1490  *			see if.h for the definitions
1491  *	@gflags:	Global flags ( kept as legacy )
1492  *	@padded:	How much padding added by alloc_netdev()
1493  *	@operstate:	RFC2863 operstate
1494  *	@link_mode:	Mapping policy to operstate
1495  *	@if_port:	Selectable AUI, TP, ...
1496  *	@dma:		DMA channel
1497  *	@mtu:		Interface MTU value
1498  *	@min_mtu:	Interface Minimum MTU value
1499  *	@max_mtu:	Interface Maximum MTU value
1500  *	@type:		Interface hardware type
1501  *	@hard_header_len: Maximum hardware header length.
1502  *	@min_header_len:  Minimum hardware header length
1503  *
1504  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1505  *			  cases can this be guaranteed
1506  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1507  *			  cases can this be guaranteed. Some cases also use
1508  *			  LL_MAX_HEADER instead to allocate the skb
1509  *
1510  *	interface address info:
1511  *
1512  * 	@perm_addr:		Permanent hw address
1513  * 	@addr_assign_type:	Hw address assignment type
1514  * 	@addr_len:		Hardware address length
1515  *	@neigh_priv_len:	Used in neigh_alloc()
1516  * 	@dev_id:		Used to differentiate devices that share
1517  * 				the same link layer address
1518  * 	@dev_port:		Used to differentiate devices that share
1519  * 				the same function
1520  *	@addr_list_lock:	XXX: need comments on this one
1521  *	@uc_promisc:		Counter that indicates promiscuous mode
1522  *				has been enabled due to the need to listen to
1523  *				additional unicast addresses in a device that
1524  *				does not implement ndo_set_rx_mode()
1525  *	@uc:			unicast mac addresses
1526  *	@mc:			multicast mac addresses
1527  *	@dev_addrs:		list of device hw addresses
1528  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1529  *	@promiscuity:		Number of times the NIC is told to work in
1530  *				promiscuous mode; if it becomes 0 the NIC will
1531  *				exit promiscuous mode
1532  *	@allmulti:		Counter, enables or disables allmulticast mode
1533  *
1534  *	@vlan_info:	VLAN info
1535  *	@dsa_ptr:	dsa specific data
1536  *	@tipc_ptr:	TIPC specific data
1537  *	@atalk_ptr:	AppleTalk link
1538  *	@ip_ptr:	IPv4 specific data
1539  *	@dn_ptr:	DECnet specific data
1540  *	@ip6_ptr:	IPv6 specific data
1541  *	@ax25_ptr:	AX.25 specific data
1542  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1543  *
1544  *	@dev_addr:	Hw address (before bcast,
1545  *			because most packets are unicast)
1546  *
1547  *	@_rx:			Array of RX queues
1548  *	@num_rx_queues:		Number of RX queues
1549  *				allocated at register_netdev() time
1550  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1551  *
1552  *	@rx_handler:		handler for received packets
1553  *	@rx_handler_data: 	XXX: need comments on this one
1554  *	@ingress_queue:		XXX: need comments on this one
1555  *	@broadcast:		hw bcast address
1556  *
1557  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1558  *			indexed by RX queue number. Assigned by driver.
1559  *			This must only be set if the ndo_rx_flow_steer
1560  *			operation is defined
1561  *	@index_hlist:		Device index hash chain
1562  *
1563  *	@_tx:			Array of TX queues
1564  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1565  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1566  *	@qdisc:			Root qdisc from userspace point of view
1567  *	@tx_queue_len:		Max frames per queue allowed
1568  *	@tx_global_lock: 	XXX: need comments on this one
1569  *
1570  *	@xps_maps:	XXX: need comments on this one
1571  *
1572  *	@watchdog_timeo:	Represents the timeout that is used by
1573  *				the watchdog (see dev_watchdog())
1574  *	@watchdog_timer:	List of timers
1575  *
1576  *	@pcpu_refcnt:		Number of references to this device
1577  *	@todo_list:		Delayed register/unregister
1578  *	@link_watch_list:	XXX: need comments on this one
1579  *
1580  *	@reg_state:		Register/unregister state machine
1581  *	@dismantle:		Device is going to be freed
1582  *	@rtnl_link_state:	This enum represents the phases of creating
1583  *				a new link
1584  *
1585  *	@destructor:		Called from unregister,
1586  *				can be used to call free_netdev
1587  *	@npinfo:		XXX: need comments on this one
1588  * 	@nd_net:		Network namespace this network device is inside
1589  *
1590  * 	@ml_priv:	Mid-layer private
1591  * 	@lstats:	Loopback statistics
1592  * 	@tstats:	Tunnel statistics
1593  * 	@dstats:	Dummy statistics
1594  * 	@vstats:	Virtual ethernet statistics
1595  *
1596  *	@garp_port:	GARP
1597  *	@mrp_port:	MRP
1598  *
1599  *	@dev:		Class/net/name entry
1600  *	@sysfs_groups:	Space for optional device, statistics and wireless
1601  *			sysfs groups
1602  *
1603  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1604  *	@rtnl_link_ops:	Rtnl_link_ops
1605  *
1606  *	@gso_max_size:	Maximum size of generic segmentation offload
1607  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1608  *			NIC for GSO
1609  *
1610  *	@dcbnl_ops:	Data Center Bridging netlink ops
1611  *	@num_tc:	Number of traffic classes in the net device
1612  *	@tc_to_txq:	XXX: need comments on this one
1613  *	@prio_tc_map:	XXX: need comments on this one
1614  *
1615  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1616  *
1617  *	@priomap:	XXX: need comments on this one
1618  *	@phydev:	Physical device may attach itself
1619  *			for hardware timestamping
1620  *
1621  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1622  *	@qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1623  *
1624  *	@proto_down:	protocol port state information can be sent to the
1625  *			switch driver and used to set the phys state of the
1626  *			switch port.
1627  *
1628  *	FIXME: cleanup struct net_device such that network protocol info
1629  *	moves out.
1630  */
1631 
1632 struct net_device {
1633 	char			name[IFNAMSIZ];
1634 	struct hlist_node	name_hlist;
1635 	char 			*ifalias;
1636 	/*
1637 	 *	I/O specific fields
1638 	 *	FIXME: Merge these and struct ifmap into one
1639 	 */
1640 	unsigned long		mem_end;
1641 	unsigned long		mem_start;
1642 	unsigned long		base_addr;
1643 	int			irq;
1644 
1645 	atomic_t		carrier_changes;
1646 
1647 	/*
1648 	 *	Some hardware also needs these fields (state,dev_list,
1649 	 *	napi_list,unreg_list,close_list) but they are not
1650 	 *	part of the usual set specified in Space.c.
1651 	 */
1652 
1653 	unsigned long		state;
1654 
1655 	struct list_head	dev_list;
1656 	struct list_head	napi_list;
1657 	struct list_head	unreg_list;
1658 	struct list_head	close_list;
1659 	struct list_head	ptype_all;
1660 	struct list_head	ptype_specific;
1661 
1662 	struct {
1663 		struct list_head upper;
1664 		struct list_head lower;
1665 	} adj_list;
1666 
1667 	netdev_features_t	features;
1668 	netdev_features_t	hw_features;
1669 	netdev_features_t	wanted_features;
1670 	netdev_features_t	vlan_features;
1671 	netdev_features_t	hw_enc_features;
1672 	netdev_features_t	mpls_features;
1673 	netdev_features_t	gso_partial_features;
1674 
1675 	int			ifindex;
1676 	int			group;
1677 
1678 	struct net_device_stats	stats;
1679 
1680 	atomic_long_t		rx_dropped;
1681 	atomic_long_t		tx_dropped;
1682 	atomic_long_t		rx_nohandler;
1683 
1684 #ifdef CONFIG_WIRELESS_EXT
1685 	const struct iw_handler_def *wireless_handlers;
1686 	struct iw_public_data	*wireless_data;
1687 #endif
1688 	const struct net_device_ops *netdev_ops;
1689 	const struct ethtool_ops *ethtool_ops;
1690 #ifdef CONFIG_NET_SWITCHDEV
1691 	const struct switchdev_ops *switchdev_ops;
1692 #endif
1693 #ifdef CONFIG_NET_L3_MASTER_DEV
1694 	const struct l3mdev_ops	*l3mdev_ops;
1695 #endif
1696 #if IS_ENABLED(CONFIG_IPV6)
1697 	const struct ndisc_ops *ndisc_ops;
1698 #endif
1699 
1700 	const struct header_ops *header_ops;
1701 
1702 	unsigned int		flags;
1703 	unsigned int		priv_flags;
1704 
1705 	unsigned short		gflags;
1706 	unsigned short		padded;
1707 
1708 	unsigned char		operstate;
1709 	unsigned char		link_mode;
1710 
1711 	unsigned char		if_port;
1712 	unsigned char		dma;
1713 
1714 	unsigned int		mtu;
1715 	unsigned int		min_mtu;
1716 	unsigned int		max_mtu;
1717 	unsigned short		type;
1718 	unsigned short		hard_header_len;
1719 	unsigned short		min_header_len;
1720 
1721 	unsigned short		needed_headroom;
1722 	unsigned short		needed_tailroom;
1723 
1724 	/* Interface address info. */
1725 	unsigned char		perm_addr[MAX_ADDR_LEN];
1726 	unsigned char		addr_assign_type;
1727 	unsigned char		addr_len;
1728 	unsigned short		neigh_priv_len;
1729 	unsigned short          dev_id;
1730 	unsigned short          dev_port;
1731 	spinlock_t		addr_list_lock;
1732 	unsigned char		name_assign_type;
1733 	bool			uc_promisc;
1734 	struct netdev_hw_addr_list	uc;
1735 	struct netdev_hw_addr_list	mc;
1736 	struct netdev_hw_addr_list	dev_addrs;
1737 
1738 #ifdef CONFIG_SYSFS
1739 	struct kset		*queues_kset;
1740 #endif
1741 	unsigned int		promiscuity;
1742 	unsigned int		allmulti;
1743 
1744 
1745 	/* Protocol-specific pointers */
1746 
1747 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1748 	struct vlan_info __rcu	*vlan_info;
1749 #endif
1750 #if IS_ENABLED(CONFIG_NET_DSA)
1751 	struct dsa_switch_tree	*dsa_ptr;
1752 #endif
1753 #if IS_ENABLED(CONFIG_TIPC)
1754 	struct tipc_bearer __rcu *tipc_ptr;
1755 #endif
1756 	void 			*atalk_ptr;
1757 	struct in_device __rcu	*ip_ptr;
1758 	struct dn_dev __rcu     *dn_ptr;
1759 	struct inet6_dev __rcu	*ip6_ptr;
1760 	void			*ax25_ptr;
1761 	struct wireless_dev	*ieee80211_ptr;
1762 	struct wpan_dev		*ieee802154_ptr;
1763 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1764 	struct mpls_dev __rcu	*mpls_ptr;
1765 #endif
1766 
1767 /*
1768  * Cache lines mostly used on receive path (including eth_type_trans())
1769  */
1770 	/* Interface address info used in eth_type_trans() */
1771 	unsigned char		*dev_addr;
1772 
1773 #ifdef CONFIG_SYSFS
1774 	struct netdev_rx_queue	*_rx;
1775 
1776 	unsigned int		num_rx_queues;
1777 	unsigned int		real_num_rx_queues;
1778 #endif
1779 
1780 	unsigned long		gro_flush_timeout;
1781 	rx_handler_func_t __rcu	*rx_handler;
1782 	void __rcu		*rx_handler_data;
1783 
1784 #ifdef CONFIG_NET_CLS_ACT
1785 	struct tcf_proto __rcu  *ingress_cl_list;
1786 #endif
1787 	struct netdev_queue __rcu *ingress_queue;
1788 #ifdef CONFIG_NETFILTER_INGRESS
1789 	struct nf_hook_entry __rcu *nf_hooks_ingress;
1790 #endif
1791 
1792 	unsigned char		broadcast[MAX_ADDR_LEN];
1793 #ifdef CONFIG_RFS_ACCEL
1794 	struct cpu_rmap		*rx_cpu_rmap;
1795 #endif
1796 	struct hlist_node	index_hlist;
1797 
1798 /*
1799  * Cache lines mostly used on transmit path
1800  */
1801 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1802 	unsigned int		num_tx_queues;
1803 	unsigned int		real_num_tx_queues;
1804 	struct Qdisc		*qdisc;
1805 #ifdef CONFIG_NET_SCHED
1806 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1807 #endif
1808 	unsigned long		tx_queue_len;
1809 	spinlock_t		tx_global_lock;
1810 	int			watchdog_timeo;
1811 
1812 #ifdef CONFIG_XPS
1813 	struct xps_dev_maps __rcu *xps_maps;
1814 #endif
1815 #ifdef CONFIG_NET_CLS_ACT
1816 	struct tcf_proto __rcu  *egress_cl_list;
1817 #endif
1818 
1819 	/* These may be needed for future network-power-down code. */
1820 	struct timer_list	watchdog_timer;
1821 
1822 	int __percpu		*pcpu_refcnt;
1823 	struct list_head	todo_list;
1824 
1825 	struct list_head	link_watch_list;
1826 
1827 	enum { NETREG_UNINITIALIZED=0,
1828 	       NETREG_REGISTERED,	/* completed register_netdevice */
1829 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1830 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1831 	       NETREG_RELEASED,		/* called free_netdev */
1832 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1833 	} reg_state:8;
1834 
1835 	bool dismantle;
1836 
1837 	enum {
1838 		RTNL_LINK_INITIALIZED,
1839 		RTNL_LINK_INITIALIZING,
1840 	} rtnl_link_state:16;
1841 
1842 	void (*destructor)(struct net_device *dev);
1843 
1844 #ifdef CONFIG_NETPOLL
1845 	struct netpoll_info __rcu	*npinfo;
1846 #endif
1847 
1848 	possible_net_t			nd_net;
1849 
1850 	/* mid-layer private */
1851 	union {
1852 		void					*ml_priv;
1853 		struct pcpu_lstats __percpu		*lstats;
1854 		struct pcpu_sw_netstats __percpu	*tstats;
1855 		struct pcpu_dstats __percpu		*dstats;
1856 		struct pcpu_vstats __percpu		*vstats;
1857 	};
1858 
1859 #if IS_ENABLED(CONFIG_GARP)
1860 	struct garp_port __rcu	*garp_port;
1861 #endif
1862 #if IS_ENABLED(CONFIG_MRP)
1863 	struct mrp_port __rcu	*mrp_port;
1864 #endif
1865 
1866 	struct device		dev;
1867 	const struct attribute_group *sysfs_groups[4];
1868 	const struct attribute_group *sysfs_rx_queue_group;
1869 
1870 	const struct rtnl_link_ops *rtnl_link_ops;
1871 
1872 	/* for setting kernel sock attribute on TCP connection setup */
1873 #define GSO_MAX_SIZE		65536
1874 	unsigned int		gso_max_size;
1875 #define GSO_MAX_SEGS		65535
1876 	u16			gso_max_segs;
1877 
1878 #ifdef CONFIG_DCB
1879 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1880 #endif
1881 	u8			num_tc;
1882 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1883 	u8			prio_tc_map[TC_BITMASK + 1];
1884 
1885 #if IS_ENABLED(CONFIG_FCOE)
1886 	unsigned int		fcoe_ddp_xid;
1887 #endif
1888 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1889 	struct netprio_map __rcu *priomap;
1890 #endif
1891 	struct phy_device	*phydev;
1892 	struct lock_class_key	*qdisc_tx_busylock;
1893 	struct lock_class_key	*qdisc_running_key;
1894 	bool			proto_down;
1895 };
1896 #define to_net_dev(d) container_of(d, struct net_device, dev)
1897 
1898 #define	NETDEV_ALIGN		32
1899 
1900 static inline
1901 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1902 {
1903 	return dev->prio_tc_map[prio & TC_BITMASK];
1904 }
1905 
1906 static inline
1907 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1908 {
1909 	if (tc >= dev->num_tc)
1910 		return -EINVAL;
1911 
1912 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1913 	return 0;
1914 }
1915 
1916 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1917 void netdev_reset_tc(struct net_device *dev);
1918 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1919 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1920 
1921 static inline
1922 int netdev_get_num_tc(struct net_device *dev)
1923 {
1924 	return dev->num_tc;
1925 }
1926 
1927 static inline
1928 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1929 					 unsigned int index)
1930 {
1931 	return &dev->_tx[index];
1932 }
1933 
1934 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1935 						    const struct sk_buff *skb)
1936 {
1937 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1938 }
1939 
1940 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1941 					    void (*f)(struct net_device *,
1942 						      struct netdev_queue *,
1943 						      void *),
1944 					    void *arg)
1945 {
1946 	unsigned int i;
1947 
1948 	for (i = 0; i < dev->num_tx_queues; i++)
1949 		f(dev, &dev->_tx[i], arg);
1950 }
1951 
1952 #define netdev_lockdep_set_classes(dev)				\
1953 {								\
1954 	static struct lock_class_key qdisc_tx_busylock_key;	\
1955 	static struct lock_class_key qdisc_running_key;		\
1956 	static struct lock_class_key qdisc_xmit_lock_key;	\
1957 	static struct lock_class_key dev_addr_list_lock_key;	\
1958 	unsigned int i;						\
1959 								\
1960 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
1961 	(dev)->qdisc_running_key = &qdisc_running_key;		\
1962 	lockdep_set_class(&(dev)->addr_list_lock,		\
1963 			  &dev_addr_list_lock_key); 		\
1964 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
1965 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
1966 				  &qdisc_xmit_lock_key);	\
1967 }
1968 
1969 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1970 				    struct sk_buff *skb,
1971 				    void *accel_priv);
1972 
1973 /* returns the headroom that the master device needs to take in account
1974  * when forwarding to this dev
1975  */
1976 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1977 {
1978 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1979 }
1980 
1981 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1982 {
1983 	if (dev->netdev_ops->ndo_set_rx_headroom)
1984 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1985 }
1986 
1987 /* set the device rx headroom to the dev's default */
1988 static inline void netdev_reset_rx_headroom(struct net_device *dev)
1989 {
1990 	netdev_set_rx_headroom(dev, -1);
1991 }
1992 
1993 /*
1994  * Net namespace inlines
1995  */
1996 static inline
1997 struct net *dev_net(const struct net_device *dev)
1998 {
1999 	return read_pnet(&dev->nd_net);
2000 }
2001 
2002 static inline
2003 void dev_net_set(struct net_device *dev, struct net *net)
2004 {
2005 	write_pnet(&dev->nd_net, net);
2006 }
2007 
2008 /**
2009  *	netdev_priv - access network device private data
2010  *	@dev: network device
2011  *
2012  * Get network device private data
2013  */
2014 static inline void *netdev_priv(const struct net_device *dev)
2015 {
2016 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2017 }
2018 
2019 /* Set the sysfs physical device reference for the network logical device
2020  * if set prior to registration will cause a symlink during initialization.
2021  */
2022 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2023 
2024 /* Set the sysfs device type for the network logical device to allow
2025  * fine-grained identification of different network device types. For
2026  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2027  */
2028 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2029 
2030 /* Default NAPI poll() weight
2031  * Device drivers are strongly advised to not use bigger value
2032  */
2033 #define NAPI_POLL_WEIGHT 64
2034 
2035 /**
2036  *	netif_napi_add - initialize a NAPI context
2037  *	@dev:  network device
2038  *	@napi: NAPI context
2039  *	@poll: polling function
2040  *	@weight: default weight
2041  *
2042  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2043  * *any* of the other NAPI-related functions.
2044  */
2045 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2046 		    int (*poll)(struct napi_struct *, int), int weight);
2047 
2048 /**
2049  *	netif_tx_napi_add - initialize a NAPI context
2050  *	@dev:  network device
2051  *	@napi: NAPI context
2052  *	@poll: polling function
2053  *	@weight: default weight
2054  *
2055  * This variant of netif_napi_add() should be used from drivers using NAPI
2056  * to exclusively poll a TX queue.
2057  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2058  */
2059 static inline void netif_tx_napi_add(struct net_device *dev,
2060 				     struct napi_struct *napi,
2061 				     int (*poll)(struct napi_struct *, int),
2062 				     int weight)
2063 {
2064 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2065 	netif_napi_add(dev, napi, poll, weight);
2066 }
2067 
2068 /**
2069  *  netif_napi_del - remove a NAPI context
2070  *  @napi: NAPI context
2071  *
2072  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2073  */
2074 void netif_napi_del(struct napi_struct *napi);
2075 
2076 struct napi_gro_cb {
2077 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2078 	void	*frag0;
2079 
2080 	/* Length of frag0. */
2081 	unsigned int frag0_len;
2082 
2083 	/* This indicates where we are processing relative to skb->data. */
2084 	int	data_offset;
2085 
2086 	/* This is non-zero if the packet cannot be merged with the new skb. */
2087 	u16	flush;
2088 
2089 	/* Save the IP ID here and check when we get to the transport layer */
2090 	u16	flush_id;
2091 
2092 	/* Number of segments aggregated. */
2093 	u16	count;
2094 
2095 	/* Start offset for remote checksum offload */
2096 	u16	gro_remcsum_start;
2097 
2098 	/* jiffies when first packet was created/queued */
2099 	unsigned long age;
2100 
2101 	/* Used in ipv6_gro_receive() and foo-over-udp */
2102 	u16	proto;
2103 
2104 	/* This is non-zero if the packet may be of the same flow. */
2105 	u8	same_flow:1;
2106 
2107 	/* Used in tunnel GRO receive */
2108 	u8	encap_mark:1;
2109 
2110 	/* GRO checksum is valid */
2111 	u8	csum_valid:1;
2112 
2113 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2114 	u8	csum_cnt:3;
2115 
2116 	/* Free the skb? */
2117 	u8	free:2;
2118 #define NAPI_GRO_FREE		  1
2119 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2120 
2121 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2122 	u8	is_ipv6:1;
2123 
2124 	/* Used in GRE, set in fou/gue_gro_receive */
2125 	u8	is_fou:1;
2126 
2127 	/* Used to determine if flush_id can be ignored */
2128 	u8	is_atomic:1;
2129 
2130 	/* Number of gro_receive callbacks this packet already went through */
2131 	u8 recursion_counter:4;
2132 
2133 	/* 1 bit hole */
2134 
2135 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2136 	__wsum	csum;
2137 
2138 	/* used in skb_gro_receive() slow path */
2139 	struct sk_buff *last;
2140 };
2141 
2142 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2143 
2144 #define GRO_RECURSION_LIMIT 15
2145 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2146 {
2147 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2148 }
2149 
2150 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2151 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2152 						struct sk_buff **head,
2153 						struct sk_buff *skb)
2154 {
2155 	if (unlikely(gro_recursion_inc_test(skb))) {
2156 		NAPI_GRO_CB(skb)->flush |= 1;
2157 		return NULL;
2158 	}
2159 
2160 	return cb(head, skb);
2161 }
2162 
2163 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2164 					     struct sk_buff *);
2165 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2166 						   struct sock *sk,
2167 						   struct sk_buff **head,
2168 						   struct sk_buff *skb)
2169 {
2170 	if (unlikely(gro_recursion_inc_test(skb))) {
2171 		NAPI_GRO_CB(skb)->flush |= 1;
2172 		return NULL;
2173 	}
2174 
2175 	return cb(sk, head, skb);
2176 }
2177 
2178 struct packet_type {
2179 	__be16			type;	/* This is really htons(ether_type). */
2180 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2181 	int			(*func) (struct sk_buff *,
2182 					 struct net_device *,
2183 					 struct packet_type *,
2184 					 struct net_device *);
2185 	bool			(*id_match)(struct packet_type *ptype,
2186 					    struct sock *sk);
2187 	void			*af_packet_priv;
2188 	struct list_head	list;
2189 };
2190 
2191 struct offload_callbacks {
2192 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2193 						netdev_features_t features);
2194 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2195 						 struct sk_buff *skb);
2196 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2197 };
2198 
2199 struct packet_offload {
2200 	__be16			 type;	/* This is really htons(ether_type). */
2201 	u16			 priority;
2202 	struct offload_callbacks callbacks;
2203 	struct list_head	 list;
2204 };
2205 
2206 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2207 struct pcpu_sw_netstats {
2208 	u64     rx_packets;
2209 	u64     rx_bytes;
2210 	u64     tx_packets;
2211 	u64     tx_bytes;
2212 	struct u64_stats_sync   syncp;
2213 };
2214 
2215 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2216 ({									\
2217 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2218 	if (pcpu_stats)	{						\
2219 		int __cpu;						\
2220 		for_each_possible_cpu(__cpu) {				\
2221 			typeof(type) *stat;				\
2222 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2223 			u64_stats_init(&stat->syncp);			\
2224 		}							\
2225 	}								\
2226 	pcpu_stats;							\
2227 })
2228 
2229 #define netdev_alloc_pcpu_stats(type)					\
2230 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2231 
2232 enum netdev_lag_tx_type {
2233 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2234 	NETDEV_LAG_TX_TYPE_RANDOM,
2235 	NETDEV_LAG_TX_TYPE_BROADCAST,
2236 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2237 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2238 	NETDEV_LAG_TX_TYPE_HASH,
2239 };
2240 
2241 struct netdev_lag_upper_info {
2242 	enum netdev_lag_tx_type tx_type;
2243 };
2244 
2245 struct netdev_lag_lower_state_info {
2246 	u8 link_up : 1,
2247 	   tx_enabled : 1;
2248 };
2249 
2250 #include <linux/notifier.h>
2251 
2252 /* netdevice notifier chain. Please remember to update the rtnetlink
2253  * notification exclusion list in rtnetlink_event() when adding new
2254  * types.
2255  */
2256 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2257 #define NETDEV_DOWN	0x0002
2258 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2259 				   detected a hardware crash and restarted
2260 				   - we can use this eg to kick tcp sessions
2261 				   once done */
2262 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2263 #define NETDEV_REGISTER 0x0005
2264 #define NETDEV_UNREGISTER	0x0006
2265 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2266 #define NETDEV_CHANGEADDR	0x0008
2267 #define NETDEV_GOING_DOWN	0x0009
2268 #define NETDEV_CHANGENAME	0x000A
2269 #define NETDEV_FEAT_CHANGE	0x000B
2270 #define NETDEV_BONDING_FAILOVER 0x000C
2271 #define NETDEV_PRE_UP		0x000D
2272 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2273 #define NETDEV_POST_TYPE_CHANGE	0x000F
2274 #define NETDEV_POST_INIT	0x0010
2275 #define NETDEV_UNREGISTER_FINAL 0x0011
2276 #define NETDEV_RELEASE		0x0012
2277 #define NETDEV_NOTIFY_PEERS	0x0013
2278 #define NETDEV_JOIN		0x0014
2279 #define NETDEV_CHANGEUPPER	0x0015
2280 #define NETDEV_RESEND_IGMP	0x0016
2281 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2282 #define NETDEV_CHANGEINFODATA	0x0018
2283 #define NETDEV_BONDING_INFO	0x0019
2284 #define NETDEV_PRECHANGEUPPER	0x001A
2285 #define NETDEV_CHANGELOWERSTATE	0x001B
2286 #define NETDEV_UDP_TUNNEL_PUSH_INFO	0x001C
2287 #define NETDEV_CHANGE_TX_QUEUE_LEN	0x001E
2288 
2289 int register_netdevice_notifier(struct notifier_block *nb);
2290 int unregister_netdevice_notifier(struct notifier_block *nb);
2291 
2292 struct netdev_notifier_info {
2293 	struct net_device *dev;
2294 };
2295 
2296 struct netdev_notifier_change_info {
2297 	struct netdev_notifier_info info; /* must be first */
2298 	unsigned int flags_changed;
2299 };
2300 
2301 struct netdev_notifier_changeupper_info {
2302 	struct netdev_notifier_info info; /* must be first */
2303 	struct net_device *upper_dev; /* new upper dev */
2304 	bool master; /* is upper dev master */
2305 	bool linking; /* is the notification for link or unlink */
2306 	void *upper_info; /* upper dev info */
2307 };
2308 
2309 struct netdev_notifier_changelowerstate_info {
2310 	struct netdev_notifier_info info; /* must be first */
2311 	void *lower_state_info; /* is lower dev state */
2312 };
2313 
2314 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2315 					     struct net_device *dev)
2316 {
2317 	info->dev = dev;
2318 }
2319 
2320 static inline struct net_device *
2321 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2322 {
2323 	return info->dev;
2324 }
2325 
2326 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2327 
2328 
2329 extern rwlock_t				dev_base_lock;		/* Device list lock */
2330 
2331 #define for_each_netdev(net, d)		\
2332 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2333 #define for_each_netdev_reverse(net, d)	\
2334 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2335 #define for_each_netdev_rcu(net, d)		\
2336 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2337 #define for_each_netdev_safe(net, d, n)	\
2338 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2339 #define for_each_netdev_continue(net, d)		\
2340 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2341 #define for_each_netdev_continue_rcu(net, d)		\
2342 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2343 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2344 		for_each_netdev_rcu(&init_net, slave)	\
2345 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2346 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2347 
2348 static inline struct net_device *next_net_device(struct net_device *dev)
2349 {
2350 	struct list_head *lh;
2351 	struct net *net;
2352 
2353 	net = dev_net(dev);
2354 	lh = dev->dev_list.next;
2355 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2356 }
2357 
2358 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2359 {
2360 	struct list_head *lh;
2361 	struct net *net;
2362 
2363 	net = dev_net(dev);
2364 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2365 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2366 }
2367 
2368 static inline struct net_device *first_net_device(struct net *net)
2369 {
2370 	return list_empty(&net->dev_base_head) ? NULL :
2371 		net_device_entry(net->dev_base_head.next);
2372 }
2373 
2374 static inline struct net_device *first_net_device_rcu(struct net *net)
2375 {
2376 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2377 
2378 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2379 }
2380 
2381 int netdev_boot_setup_check(struct net_device *dev);
2382 unsigned long netdev_boot_base(const char *prefix, int unit);
2383 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2384 				       const char *hwaddr);
2385 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2386 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2387 void dev_add_pack(struct packet_type *pt);
2388 void dev_remove_pack(struct packet_type *pt);
2389 void __dev_remove_pack(struct packet_type *pt);
2390 void dev_add_offload(struct packet_offload *po);
2391 void dev_remove_offload(struct packet_offload *po);
2392 
2393 int dev_get_iflink(const struct net_device *dev);
2394 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2395 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2396 				      unsigned short mask);
2397 struct net_device *dev_get_by_name(struct net *net, const char *name);
2398 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2399 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2400 int dev_alloc_name(struct net_device *dev, const char *name);
2401 int dev_open(struct net_device *dev);
2402 int dev_close(struct net_device *dev);
2403 int dev_close_many(struct list_head *head, bool unlink);
2404 void dev_disable_lro(struct net_device *dev);
2405 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2406 int dev_queue_xmit(struct sk_buff *skb);
2407 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2408 int register_netdevice(struct net_device *dev);
2409 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2410 void unregister_netdevice_many(struct list_head *head);
2411 static inline void unregister_netdevice(struct net_device *dev)
2412 {
2413 	unregister_netdevice_queue(dev, NULL);
2414 }
2415 
2416 int netdev_refcnt_read(const struct net_device *dev);
2417 void free_netdev(struct net_device *dev);
2418 void netdev_freemem(struct net_device *dev);
2419 void synchronize_net(void);
2420 int init_dummy_netdev(struct net_device *dev);
2421 
2422 DECLARE_PER_CPU(int, xmit_recursion);
2423 #define XMIT_RECURSION_LIMIT	10
2424 
2425 static inline int dev_recursion_level(void)
2426 {
2427 	return this_cpu_read(xmit_recursion);
2428 }
2429 
2430 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2431 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2432 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2433 int netdev_get_name(struct net *net, char *name, int ifindex);
2434 int dev_restart(struct net_device *dev);
2435 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2436 
2437 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2438 {
2439 	return NAPI_GRO_CB(skb)->data_offset;
2440 }
2441 
2442 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2443 {
2444 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2445 }
2446 
2447 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2448 {
2449 	NAPI_GRO_CB(skb)->data_offset += len;
2450 }
2451 
2452 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2453 					unsigned int offset)
2454 {
2455 	return NAPI_GRO_CB(skb)->frag0 + offset;
2456 }
2457 
2458 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2459 {
2460 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2461 }
2462 
2463 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2464 {
2465 	NAPI_GRO_CB(skb)->frag0 = NULL;
2466 	NAPI_GRO_CB(skb)->frag0_len = 0;
2467 }
2468 
2469 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2470 					unsigned int offset)
2471 {
2472 	if (!pskb_may_pull(skb, hlen))
2473 		return NULL;
2474 
2475 	skb_gro_frag0_invalidate(skb);
2476 	return skb->data + offset;
2477 }
2478 
2479 static inline void *skb_gro_network_header(struct sk_buff *skb)
2480 {
2481 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2482 	       skb_network_offset(skb);
2483 }
2484 
2485 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2486 					const void *start, unsigned int len)
2487 {
2488 	if (NAPI_GRO_CB(skb)->csum_valid)
2489 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2490 						  csum_partial(start, len, 0));
2491 }
2492 
2493 /* GRO checksum functions. These are logical equivalents of the normal
2494  * checksum functions (in skbuff.h) except that they operate on the GRO
2495  * offsets and fields in sk_buff.
2496  */
2497 
2498 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2499 
2500 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2501 {
2502 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2503 }
2504 
2505 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2506 						      bool zero_okay,
2507 						      __sum16 check)
2508 {
2509 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2510 		skb_checksum_start_offset(skb) <
2511 		 skb_gro_offset(skb)) &&
2512 		!skb_at_gro_remcsum_start(skb) &&
2513 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2514 		(!zero_okay || check));
2515 }
2516 
2517 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2518 							   __wsum psum)
2519 {
2520 	if (NAPI_GRO_CB(skb)->csum_valid &&
2521 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2522 		return 0;
2523 
2524 	NAPI_GRO_CB(skb)->csum = psum;
2525 
2526 	return __skb_gro_checksum_complete(skb);
2527 }
2528 
2529 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2530 {
2531 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2532 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2533 		NAPI_GRO_CB(skb)->csum_cnt--;
2534 	} else {
2535 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2536 		 * verified a new top level checksum or an encapsulated one
2537 		 * during GRO. This saves work if we fallback to normal path.
2538 		 */
2539 		__skb_incr_checksum_unnecessary(skb);
2540 	}
2541 }
2542 
2543 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2544 				    compute_pseudo)			\
2545 ({									\
2546 	__sum16 __ret = 0;						\
2547 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2548 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2549 				compute_pseudo(skb, proto));		\
2550 	if (__ret)							\
2551 		__skb_mark_checksum_bad(skb);				\
2552 	else								\
2553 		skb_gro_incr_csum_unnecessary(skb);			\
2554 	__ret;								\
2555 })
2556 
2557 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2558 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2559 
2560 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2561 					     compute_pseudo)		\
2562 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2563 
2564 #define skb_gro_checksum_simple_validate(skb)				\
2565 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2566 
2567 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2568 {
2569 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2570 		!NAPI_GRO_CB(skb)->csum_valid);
2571 }
2572 
2573 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2574 					      __sum16 check, __wsum pseudo)
2575 {
2576 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2577 	NAPI_GRO_CB(skb)->csum_valid = 1;
2578 }
2579 
2580 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2581 do {									\
2582 	if (__skb_gro_checksum_convert_check(skb))			\
2583 		__skb_gro_checksum_convert(skb, check,			\
2584 					   compute_pseudo(skb, proto));	\
2585 } while (0)
2586 
2587 struct gro_remcsum {
2588 	int offset;
2589 	__wsum delta;
2590 };
2591 
2592 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2593 {
2594 	grc->offset = 0;
2595 	grc->delta = 0;
2596 }
2597 
2598 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2599 					    unsigned int off, size_t hdrlen,
2600 					    int start, int offset,
2601 					    struct gro_remcsum *grc,
2602 					    bool nopartial)
2603 {
2604 	__wsum delta;
2605 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2606 
2607 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2608 
2609 	if (!nopartial) {
2610 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2611 		return ptr;
2612 	}
2613 
2614 	ptr = skb_gro_header_fast(skb, off);
2615 	if (skb_gro_header_hard(skb, off + plen)) {
2616 		ptr = skb_gro_header_slow(skb, off + plen, off);
2617 		if (!ptr)
2618 			return NULL;
2619 	}
2620 
2621 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2622 			       start, offset);
2623 
2624 	/* Adjust skb->csum since we changed the packet */
2625 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2626 
2627 	grc->offset = off + hdrlen + offset;
2628 	grc->delta = delta;
2629 
2630 	return ptr;
2631 }
2632 
2633 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2634 					   struct gro_remcsum *grc)
2635 {
2636 	void *ptr;
2637 	size_t plen = grc->offset + sizeof(u16);
2638 
2639 	if (!grc->delta)
2640 		return;
2641 
2642 	ptr = skb_gro_header_fast(skb, grc->offset);
2643 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2644 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2645 		if (!ptr)
2646 			return;
2647 	}
2648 
2649 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2650 }
2651 
2652 #ifdef CONFIG_XFRM_OFFLOAD
2653 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2654 {
2655 	if (PTR_ERR(pp) != -EINPROGRESS)
2656 		NAPI_GRO_CB(skb)->flush |= flush;
2657 }
2658 #else
2659 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2660 {
2661 	NAPI_GRO_CB(skb)->flush |= flush;
2662 }
2663 #endif
2664 
2665 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2666 				  unsigned short type,
2667 				  const void *daddr, const void *saddr,
2668 				  unsigned int len)
2669 {
2670 	if (!dev->header_ops || !dev->header_ops->create)
2671 		return 0;
2672 
2673 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2674 }
2675 
2676 static inline int dev_parse_header(const struct sk_buff *skb,
2677 				   unsigned char *haddr)
2678 {
2679 	const struct net_device *dev = skb->dev;
2680 
2681 	if (!dev->header_ops || !dev->header_ops->parse)
2682 		return 0;
2683 	return dev->header_ops->parse(skb, haddr);
2684 }
2685 
2686 /* ll_header must have at least hard_header_len allocated */
2687 static inline bool dev_validate_header(const struct net_device *dev,
2688 				       char *ll_header, int len)
2689 {
2690 	if (likely(len >= dev->hard_header_len))
2691 		return true;
2692 	if (len < dev->min_header_len)
2693 		return false;
2694 
2695 	if (capable(CAP_SYS_RAWIO)) {
2696 		memset(ll_header + len, 0, dev->hard_header_len - len);
2697 		return true;
2698 	}
2699 
2700 	if (dev->header_ops && dev->header_ops->validate)
2701 		return dev->header_ops->validate(ll_header, len);
2702 
2703 	return false;
2704 }
2705 
2706 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2707 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2708 static inline int unregister_gifconf(unsigned int family)
2709 {
2710 	return register_gifconf(family, NULL);
2711 }
2712 
2713 #ifdef CONFIG_NET_FLOW_LIMIT
2714 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2715 struct sd_flow_limit {
2716 	u64			count;
2717 	unsigned int		num_buckets;
2718 	unsigned int		history_head;
2719 	u16			history[FLOW_LIMIT_HISTORY];
2720 	u8			buckets[];
2721 };
2722 
2723 extern int netdev_flow_limit_table_len;
2724 #endif /* CONFIG_NET_FLOW_LIMIT */
2725 
2726 /*
2727  * Incoming packets are placed on per-CPU queues
2728  */
2729 struct softnet_data {
2730 	struct list_head	poll_list;
2731 	struct sk_buff_head	process_queue;
2732 
2733 	/* stats */
2734 	unsigned int		processed;
2735 	unsigned int		time_squeeze;
2736 	unsigned int		received_rps;
2737 #ifdef CONFIG_RPS
2738 	struct softnet_data	*rps_ipi_list;
2739 #endif
2740 #ifdef CONFIG_NET_FLOW_LIMIT
2741 	struct sd_flow_limit __rcu *flow_limit;
2742 #endif
2743 	struct Qdisc		*output_queue;
2744 	struct Qdisc		**output_queue_tailp;
2745 	struct sk_buff		*completion_queue;
2746 
2747 #ifdef CONFIG_RPS
2748 	/* input_queue_head should be written by cpu owning this struct,
2749 	 * and only read by other cpus. Worth using a cache line.
2750 	 */
2751 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2752 
2753 	/* Elements below can be accessed between CPUs for RPS/RFS */
2754 	struct call_single_data	csd ____cacheline_aligned_in_smp;
2755 	struct softnet_data	*rps_ipi_next;
2756 	unsigned int		cpu;
2757 	unsigned int		input_queue_tail;
2758 #endif
2759 	unsigned int		dropped;
2760 	struct sk_buff_head	input_pkt_queue;
2761 	struct napi_struct	backlog;
2762 
2763 };
2764 
2765 static inline void input_queue_head_incr(struct softnet_data *sd)
2766 {
2767 #ifdef CONFIG_RPS
2768 	sd->input_queue_head++;
2769 #endif
2770 }
2771 
2772 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2773 					      unsigned int *qtail)
2774 {
2775 #ifdef CONFIG_RPS
2776 	*qtail = ++sd->input_queue_tail;
2777 #endif
2778 }
2779 
2780 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2781 
2782 void __netif_schedule(struct Qdisc *q);
2783 void netif_schedule_queue(struct netdev_queue *txq);
2784 
2785 static inline void netif_tx_schedule_all(struct net_device *dev)
2786 {
2787 	unsigned int i;
2788 
2789 	for (i = 0; i < dev->num_tx_queues; i++)
2790 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2791 }
2792 
2793 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2794 {
2795 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2796 }
2797 
2798 /**
2799  *	netif_start_queue - allow transmit
2800  *	@dev: network device
2801  *
2802  *	Allow upper layers to call the device hard_start_xmit routine.
2803  */
2804 static inline void netif_start_queue(struct net_device *dev)
2805 {
2806 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2807 }
2808 
2809 static inline void netif_tx_start_all_queues(struct net_device *dev)
2810 {
2811 	unsigned int i;
2812 
2813 	for (i = 0; i < dev->num_tx_queues; i++) {
2814 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2815 		netif_tx_start_queue(txq);
2816 	}
2817 }
2818 
2819 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2820 
2821 /**
2822  *	netif_wake_queue - restart transmit
2823  *	@dev: network device
2824  *
2825  *	Allow upper layers to call the device hard_start_xmit routine.
2826  *	Used for flow control when transmit resources are available.
2827  */
2828 static inline void netif_wake_queue(struct net_device *dev)
2829 {
2830 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2831 }
2832 
2833 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2834 {
2835 	unsigned int i;
2836 
2837 	for (i = 0; i < dev->num_tx_queues; i++) {
2838 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2839 		netif_tx_wake_queue(txq);
2840 	}
2841 }
2842 
2843 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2844 {
2845 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2846 }
2847 
2848 /**
2849  *	netif_stop_queue - stop transmitted packets
2850  *	@dev: network device
2851  *
2852  *	Stop upper layers calling the device hard_start_xmit routine.
2853  *	Used for flow control when transmit resources are unavailable.
2854  */
2855 static inline void netif_stop_queue(struct net_device *dev)
2856 {
2857 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2858 }
2859 
2860 void netif_tx_stop_all_queues(struct net_device *dev);
2861 
2862 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2863 {
2864 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2865 }
2866 
2867 /**
2868  *	netif_queue_stopped - test if transmit queue is flowblocked
2869  *	@dev: network device
2870  *
2871  *	Test if transmit queue on device is currently unable to send.
2872  */
2873 static inline bool netif_queue_stopped(const struct net_device *dev)
2874 {
2875 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2876 }
2877 
2878 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2879 {
2880 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2881 }
2882 
2883 static inline bool
2884 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2885 {
2886 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2887 }
2888 
2889 static inline bool
2890 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2891 {
2892 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2893 }
2894 
2895 /**
2896  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2897  *	@dev_queue: pointer to transmit queue
2898  *
2899  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2900  * to give appropriate hint to the CPU.
2901  */
2902 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2903 {
2904 #ifdef CONFIG_BQL
2905 	prefetchw(&dev_queue->dql.num_queued);
2906 #endif
2907 }
2908 
2909 /**
2910  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2911  *	@dev_queue: pointer to transmit queue
2912  *
2913  * BQL enabled drivers might use this helper in their TX completion path,
2914  * to give appropriate hint to the CPU.
2915  */
2916 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2917 {
2918 #ifdef CONFIG_BQL
2919 	prefetchw(&dev_queue->dql.limit);
2920 #endif
2921 }
2922 
2923 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2924 					unsigned int bytes)
2925 {
2926 #ifdef CONFIG_BQL
2927 	dql_queued(&dev_queue->dql, bytes);
2928 
2929 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2930 		return;
2931 
2932 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2933 
2934 	/*
2935 	 * The XOFF flag must be set before checking the dql_avail below,
2936 	 * because in netdev_tx_completed_queue we update the dql_completed
2937 	 * before checking the XOFF flag.
2938 	 */
2939 	smp_mb();
2940 
2941 	/* check again in case another CPU has just made room avail */
2942 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2943 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2944 #endif
2945 }
2946 
2947 /**
2948  * 	netdev_sent_queue - report the number of bytes queued to hardware
2949  * 	@dev: network device
2950  * 	@bytes: number of bytes queued to the hardware device queue
2951  *
2952  * 	Report the number of bytes queued for sending/completion to the network
2953  * 	device hardware queue. @bytes should be a good approximation and should
2954  * 	exactly match netdev_completed_queue() @bytes
2955  */
2956 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2957 {
2958 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2959 }
2960 
2961 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2962 					     unsigned int pkts, unsigned int bytes)
2963 {
2964 #ifdef CONFIG_BQL
2965 	if (unlikely(!bytes))
2966 		return;
2967 
2968 	dql_completed(&dev_queue->dql, bytes);
2969 
2970 	/*
2971 	 * Without the memory barrier there is a small possiblity that
2972 	 * netdev_tx_sent_queue will miss the update and cause the queue to
2973 	 * be stopped forever
2974 	 */
2975 	smp_mb();
2976 
2977 	if (dql_avail(&dev_queue->dql) < 0)
2978 		return;
2979 
2980 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2981 		netif_schedule_queue(dev_queue);
2982 #endif
2983 }
2984 
2985 /**
2986  * 	netdev_completed_queue - report bytes and packets completed by device
2987  * 	@dev: network device
2988  * 	@pkts: actual number of packets sent over the medium
2989  * 	@bytes: actual number of bytes sent over the medium
2990  *
2991  * 	Report the number of bytes and packets transmitted by the network device
2992  * 	hardware queue over the physical medium, @bytes must exactly match the
2993  * 	@bytes amount passed to netdev_sent_queue()
2994  */
2995 static inline void netdev_completed_queue(struct net_device *dev,
2996 					  unsigned int pkts, unsigned int bytes)
2997 {
2998 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2999 }
3000 
3001 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3002 {
3003 #ifdef CONFIG_BQL
3004 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3005 	dql_reset(&q->dql);
3006 #endif
3007 }
3008 
3009 /**
3010  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3011  * 	@dev_queue: network device
3012  *
3013  * 	Reset the bytes and packet count of a network device and clear the
3014  * 	software flow control OFF bit for this network device
3015  */
3016 static inline void netdev_reset_queue(struct net_device *dev_queue)
3017 {
3018 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3019 }
3020 
3021 /**
3022  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3023  * 	@dev: network device
3024  * 	@queue_index: given tx queue index
3025  *
3026  * 	Returns 0 if given tx queue index >= number of device tx queues,
3027  * 	otherwise returns the originally passed tx queue index.
3028  */
3029 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3030 {
3031 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3032 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3033 				     dev->name, queue_index,
3034 				     dev->real_num_tx_queues);
3035 		return 0;
3036 	}
3037 
3038 	return queue_index;
3039 }
3040 
3041 /**
3042  *	netif_running - test if up
3043  *	@dev: network device
3044  *
3045  *	Test if the device has been brought up.
3046  */
3047 static inline bool netif_running(const struct net_device *dev)
3048 {
3049 	return test_bit(__LINK_STATE_START, &dev->state);
3050 }
3051 
3052 /*
3053  * Routines to manage the subqueues on a device.  We only need start,
3054  * stop, and a check if it's stopped.  All other device management is
3055  * done at the overall netdevice level.
3056  * Also test the device if we're multiqueue.
3057  */
3058 
3059 /**
3060  *	netif_start_subqueue - allow sending packets on subqueue
3061  *	@dev: network device
3062  *	@queue_index: sub queue index
3063  *
3064  * Start individual transmit queue of a device with multiple transmit queues.
3065  */
3066 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3067 {
3068 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3069 
3070 	netif_tx_start_queue(txq);
3071 }
3072 
3073 /**
3074  *	netif_stop_subqueue - stop sending packets on subqueue
3075  *	@dev: network device
3076  *	@queue_index: sub queue index
3077  *
3078  * Stop individual transmit queue of a device with multiple transmit queues.
3079  */
3080 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3081 {
3082 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3083 	netif_tx_stop_queue(txq);
3084 }
3085 
3086 /**
3087  *	netif_subqueue_stopped - test status of subqueue
3088  *	@dev: network device
3089  *	@queue_index: sub queue index
3090  *
3091  * Check individual transmit queue of a device with multiple transmit queues.
3092  */
3093 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3094 					    u16 queue_index)
3095 {
3096 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3097 
3098 	return netif_tx_queue_stopped(txq);
3099 }
3100 
3101 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3102 					  struct sk_buff *skb)
3103 {
3104 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3105 }
3106 
3107 /**
3108  *	netif_wake_subqueue - allow sending packets on subqueue
3109  *	@dev: network device
3110  *	@queue_index: sub queue index
3111  *
3112  * Resume individual transmit queue of a device with multiple transmit queues.
3113  */
3114 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3115 {
3116 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3117 
3118 	netif_tx_wake_queue(txq);
3119 }
3120 
3121 #ifdef CONFIG_XPS
3122 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3123 			u16 index);
3124 #else
3125 static inline int netif_set_xps_queue(struct net_device *dev,
3126 				      const struct cpumask *mask,
3127 				      u16 index)
3128 {
3129 	return 0;
3130 }
3131 #endif
3132 
3133 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3134 		  unsigned int num_tx_queues);
3135 
3136 /*
3137  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3138  * as a distribution range limit for the returned value.
3139  */
3140 static inline u16 skb_tx_hash(const struct net_device *dev,
3141 			      struct sk_buff *skb)
3142 {
3143 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3144 }
3145 
3146 /**
3147  *	netif_is_multiqueue - test if device has multiple transmit queues
3148  *	@dev: network device
3149  *
3150  * Check if device has multiple transmit queues
3151  */
3152 static inline bool netif_is_multiqueue(const struct net_device *dev)
3153 {
3154 	return dev->num_tx_queues > 1;
3155 }
3156 
3157 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3158 
3159 #ifdef CONFIG_SYSFS
3160 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3161 #else
3162 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3163 						unsigned int rxq)
3164 {
3165 	return 0;
3166 }
3167 #endif
3168 
3169 #ifdef CONFIG_SYSFS
3170 static inline unsigned int get_netdev_rx_queue_index(
3171 		struct netdev_rx_queue *queue)
3172 {
3173 	struct net_device *dev = queue->dev;
3174 	int index = queue - dev->_rx;
3175 
3176 	BUG_ON(index >= dev->num_rx_queues);
3177 	return index;
3178 }
3179 #endif
3180 
3181 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3182 int netif_get_num_default_rss_queues(void);
3183 
3184 enum skb_free_reason {
3185 	SKB_REASON_CONSUMED,
3186 	SKB_REASON_DROPPED,
3187 };
3188 
3189 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3190 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3191 
3192 /*
3193  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3194  * interrupt context or with hardware interrupts being disabled.
3195  * (in_irq() || irqs_disabled())
3196  *
3197  * We provide four helpers that can be used in following contexts :
3198  *
3199  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3200  *  replacing kfree_skb(skb)
3201  *
3202  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3203  *  Typically used in place of consume_skb(skb) in TX completion path
3204  *
3205  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3206  *  replacing kfree_skb(skb)
3207  *
3208  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3209  *  and consumed a packet. Used in place of consume_skb(skb)
3210  */
3211 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3212 {
3213 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3214 }
3215 
3216 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3217 {
3218 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3219 }
3220 
3221 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3222 {
3223 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3224 }
3225 
3226 static inline void dev_consume_skb_any(struct sk_buff *skb)
3227 {
3228 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3229 }
3230 
3231 int netif_rx(struct sk_buff *skb);
3232 int netif_rx_ni(struct sk_buff *skb);
3233 int netif_receive_skb(struct sk_buff *skb);
3234 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3235 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3236 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3237 gro_result_t napi_gro_frags(struct napi_struct *napi);
3238 struct packet_offload *gro_find_receive_by_type(__be16 type);
3239 struct packet_offload *gro_find_complete_by_type(__be16 type);
3240 
3241 static inline void napi_free_frags(struct napi_struct *napi)
3242 {
3243 	kfree_skb(napi->skb);
3244 	napi->skb = NULL;
3245 }
3246 
3247 bool netdev_is_rx_handler_busy(struct net_device *dev);
3248 int netdev_rx_handler_register(struct net_device *dev,
3249 			       rx_handler_func_t *rx_handler,
3250 			       void *rx_handler_data);
3251 void netdev_rx_handler_unregister(struct net_device *dev);
3252 
3253 bool dev_valid_name(const char *name);
3254 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3255 int dev_ethtool(struct net *net, struct ifreq *);
3256 unsigned int dev_get_flags(const struct net_device *);
3257 int __dev_change_flags(struct net_device *, unsigned int flags);
3258 int dev_change_flags(struct net_device *, unsigned int);
3259 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3260 			unsigned int gchanges);
3261 int dev_change_name(struct net_device *, const char *);
3262 int dev_set_alias(struct net_device *, const char *, size_t);
3263 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3264 int dev_set_mtu(struct net_device *, int);
3265 void dev_set_group(struct net_device *, int);
3266 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3267 int dev_change_carrier(struct net_device *, bool new_carrier);
3268 int dev_get_phys_port_id(struct net_device *dev,
3269 			 struct netdev_phys_item_id *ppid);
3270 int dev_get_phys_port_name(struct net_device *dev,
3271 			   char *name, size_t len);
3272 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3273 int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
3274 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3275 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3276 				    struct netdev_queue *txq, int *ret);
3277 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3278 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3279 bool is_skb_forwardable(const struct net_device *dev,
3280 			const struct sk_buff *skb);
3281 
3282 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3283 					       struct sk_buff *skb)
3284 {
3285 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3286 	    unlikely(!is_skb_forwardable(dev, skb))) {
3287 		atomic_long_inc(&dev->rx_dropped);
3288 		kfree_skb(skb);
3289 		return NET_RX_DROP;
3290 	}
3291 
3292 	skb_scrub_packet(skb, true);
3293 	skb->priority = 0;
3294 	return 0;
3295 }
3296 
3297 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3298 
3299 extern int		netdev_budget;
3300 
3301 /* Called by rtnetlink.c:rtnl_unlock() */
3302 void netdev_run_todo(void);
3303 
3304 /**
3305  *	dev_put - release reference to device
3306  *	@dev: network device
3307  *
3308  * Release reference to device to allow it to be freed.
3309  */
3310 static inline void dev_put(struct net_device *dev)
3311 {
3312 	this_cpu_dec(*dev->pcpu_refcnt);
3313 }
3314 
3315 /**
3316  *	dev_hold - get reference to device
3317  *	@dev: network device
3318  *
3319  * Hold reference to device to keep it from being freed.
3320  */
3321 static inline void dev_hold(struct net_device *dev)
3322 {
3323 	this_cpu_inc(*dev->pcpu_refcnt);
3324 }
3325 
3326 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3327  * and _off may be called from IRQ context, but it is caller
3328  * who is responsible for serialization of these calls.
3329  *
3330  * The name carrier is inappropriate, these functions should really be
3331  * called netif_lowerlayer_*() because they represent the state of any
3332  * kind of lower layer not just hardware media.
3333  */
3334 
3335 void linkwatch_init_dev(struct net_device *dev);
3336 void linkwatch_fire_event(struct net_device *dev);
3337 void linkwatch_forget_dev(struct net_device *dev);
3338 
3339 /**
3340  *	netif_carrier_ok - test if carrier present
3341  *	@dev: network device
3342  *
3343  * Check if carrier is present on device
3344  */
3345 static inline bool netif_carrier_ok(const struct net_device *dev)
3346 {
3347 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3348 }
3349 
3350 unsigned long dev_trans_start(struct net_device *dev);
3351 
3352 void __netdev_watchdog_up(struct net_device *dev);
3353 
3354 void netif_carrier_on(struct net_device *dev);
3355 
3356 void netif_carrier_off(struct net_device *dev);
3357 
3358 /**
3359  *	netif_dormant_on - mark device as dormant.
3360  *	@dev: network device
3361  *
3362  * Mark device as dormant (as per RFC2863).
3363  *
3364  * The dormant state indicates that the relevant interface is not
3365  * actually in a condition to pass packets (i.e., it is not 'up') but is
3366  * in a "pending" state, waiting for some external event.  For "on-
3367  * demand" interfaces, this new state identifies the situation where the
3368  * interface is waiting for events to place it in the up state.
3369  */
3370 static inline void netif_dormant_on(struct net_device *dev)
3371 {
3372 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3373 		linkwatch_fire_event(dev);
3374 }
3375 
3376 /**
3377  *	netif_dormant_off - set device as not dormant.
3378  *	@dev: network device
3379  *
3380  * Device is not in dormant state.
3381  */
3382 static inline void netif_dormant_off(struct net_device *dev)
3383 {
3384 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3385 		linkwatch_fire_event(dev);
3386 }
3387 
3388 /**
3389  *	netif_dormant - test if carrier present
3390  *	@dev: network device
3391  *
3392  * Check if carrier is present on device
3393  */
3394 static inline bool netif_dormant(const struct net_device *dev)
3395 {
3396 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3397 }
3398 
3399 
3400 /**
3401  *	netif_oper_up - test if device is operational
3402  *	@dev: network device
3403  *
3404  * Check if carrier is operational
3405  */
3406 static inline bool netif_oper_up(const struct net_device *dev)
3407 {
3408 	return (dev->operstate == IF_OPER_UP ||
3409 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3410 }
3411 
3412 /**
3413  *	netif_device_present - is device available or removed
3414  *	@dev: network device
3415  *
3416  * Check if device has not been removed from system.
3417  */
3418 static inline bool netif_device_present(struct net_device *dev)
3419 {
3420 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3421 }
3422 
3423 void netif_device_detach(struct net_device *dev);
3424 
3425 void netif_device_attach(struct net_device *dev);
3426 
3427 /*
3428  * Network interface message level settings
3429  */
3430 
3431 enum {
3432 	NETIF_MSG_DRV		= 0x0001,
3433 	NETIF_MSG_PROBE		= 0x0002,
3434 	NETIF_MSG_LINK		= 0x0004,
3435 	NETIF_MSG_TIMER		= 0x0008,
3436 	NETIF_MSG_IFDOWN	= 0x0010,
3437 	NETIF_MSG_IFUP		= 0x0020,
3438 	NETIF_MSG_RX_ERR	= 0x0040,
3439 	NETIF_MSG_TX_ERR	= 0x0080,
3440 	NETIF_MSG_TX_QUEUED	= 0x0100,
3441 	NETIF_MSG_INTR		= 0x0200,
3442 	NETIF_MSG_TX_DONE	= 0x0400,
3443 	NETIF_MSG_RX_STATUS	= 0x0800,
3444 	NETIF_MSG_PKTDATA	= 0x1000,
3445 	NETIF_MSG_HW		= 0x2000,
3446 	NETIF_MSG_WOL		= 0x4000,
3447 };
3448 
3449 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3450 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3451 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3452 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3453 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3454 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3455 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3456 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3457 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3458 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3459 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3460 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3461 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3462 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3463 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3464 
3465 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3466 {
3467 	/* use default */
3468 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3469 		return default_msg_enable_bits;
3470 	if (debug_value == 0)	/* no output */
3471 		return 0;
3472 	/* set low N bits */
3473 	return (1 << debug_value) - 1;
3474 }
3475 
3476 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3477 {
3478 	spin_lock(&txq->_xmit_lock);
3479 	txq->xmit_lock_owner = cpu;
3480 }
3481 
3482 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3483 {
3484 	__acquire(&txq->_xmit_lock);
3485 	return true;
3486 }
3487 
3488 static inline void __netif_tx_release(struct netdev_queue *txq)
3489 {
3490 	__release(&txq->_xmit_lock);
3491 }
3492 
3493 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3494 {
3495 	spin_lock_bh(&txq->_xmit_lock);
3496 	txq->xmit_lock_owner = smp_processor_id();
3497 }
3498 
3499 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3500 {
3501 	bool ok = spin_trylock(&txq->_xmit_lock);
3502 	if (likely(ok))
3503 		txq->xmit_lock_owner = smp_processor_id();
3504 	return ok;
3505 }
3506 
3507 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3508 {
3509 	txq->xmit_lock_owner = -1;
3510 	spin_unlock(&txq->_xmit_lock);
3511 }
3512 
3513 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3514 {
3515 	txq->xmit_lock_owner = -1;
3516 	spin_unlock_bh(&txq->_xmit_lock);
3517 }
3518 
3519 static inline void txq_trans_update(struct netdev_queue *txq)
3520 {
3521 	if (txq->xmit_lock_owner != -1)
3522 		txq->trans_start = jiffies;
3523 }
3524 
3525 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3526 static inline void netif_trans_update(struct net_device *dev)
3527 {
3528 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3529 
3530 	if (txq->trans_start != jiffies)
3531 		txq->trans_start = jiffies;
3532 }
3533 
3534 /**
3535  *	netif_tx_lock - grab network device transmit lock
3536  *	@dev: network device
3537  *
3538  * Get network device transmit lock
3539  */
3540 static inline void netif_tx_lock(struct net_device *dev)
3541 {
3542 	unsigned int i;
3543 	int cpu;
3544 
3545 	spin_lock(&dev->tx_global_lock);
3546 	cpu = smp_processor_id();
3547 	for (i = 0; i < dev->num_tx_queues; i++) {
3548 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3549 
3550 		/* We are the only thread of execution doing a
3551 		 * freeze, but we have to grab the _xmit_lock in
3552 		 * order to synchronize with threads which are in
3553 		 * the ->hard_start_xmit() handler and already
3554 		 * checked the frozen bit.
3555 		 */
3556 		__netif_tx_lock(txq, cpu);
3557 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3558 		__netif_tx_unlock(txq);
3559 	}
3560 }
3561 
3562 static inline void netif_tx_lock_bh(struct net_device *dev)
3563 {
3564 	local_bh_disable();
3565 	netif_tx_lock(dev);
3566 }
3567 
3568 static inline void netif_tx_unlock(struct net_device *dev)
3569 {
3570 	unsigned int i;
3571 
3572 	for (i = 0; i < dev->num_tx_queues; i++) {
3573 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3574 
3575 		/* No need to grab the _xmit_lock here.  If the
3576 		 * queue is not stopped for another reason, we
3577 		 * force a schedule.
3578 		 */
3579 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3580 		netif_schedule_queue(txq);
3581 	}
3582 	spin_unlock(&dev->tx_global_lock);
3583 }
3584 
3585 static inline void netif_tx_unlock_bh(struct net_device *dev)
3586 {
3587 	netif_tx_unlock(dev);
3588 	local_bh_enable();
3589 }
3590 
3591 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3592 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3593 		__netif_tx_lock(txq, cpu);		\
3594 	} else {					\
3595 		__netif_tx_acquire(txq);		\
3596 	}						\
3597 }
3598 
3599 #define HARD_TX_TRYLOCK(dev, txq)			\
3600 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3601 		__netif_tx_trylock(txq) :		\
3602 		__netif_tx_acquire(txq))
3603 
3604 #define HARD_TX_UNLOCK(dev, txq) {			\
3605 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3606 		__netif_tx_unlock(txq);			\
3607 	} else {					\
3608 		__netif_tx_release(txq);		\
3609 	}						\
3610 }
3611 
3612 static inline void netif_tx_disable(struct net_device *dev)
3613 {
3614 	unsigned int i;
3615 	int cpu;
3616 
3617 	local_bh_disable();
3618 	cpu = smp_processor_id();
3619 	for (i = 0; i < dev->num_tx_queues; i++) {
3620 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3621 
3622 		__netif_tx_lock(txq, cpu);
3623 		netif_tx_stop_queue(txq);
3624 		__netif_tx_unlock(txq);
3625 	}
3626 	local_bh_enable();
3627 }
3628 
3629 static inline void netif_addr_lock(struct net_device *dev)
3630 {
3631 	spin_lock(&dev->addr_list_lock);
3632 }
3633 
3634 static inline void netif_addr_lock_nested(struct net_device *dev)
3635 {
3636 	int subclass = SINGLE_DEPTH_NESTING;
3637 
3638 	if (dev->netdev_ops->ndo_get_lock_subclass)
3639 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3640 
3641 	spin_lock_nested(&dev->addr_list_lock, subclass);
3642 }
3643 
3644 static inline void netif_addr_lock_bh(struct net_device *dev)
3645 {
3646 	spin_lock_bh(&dev->addr_list_lock);
3647 }
3648 
3649 static inline void netif_addr_unlock(struct net_device *dev)
3650 {
3651 	spin_unlock(&dev->addr_list_lock);
3652 }
3653 
3654 static inline void netif_addr_unlock_bh(struct net_device *dev)
3655 {
3656 	spin_unlock_bh(&dev->addr_list_lock);
3657 }
3658 
3659 /*
3660  * dev_addrs walker. Should be used only for read access. Call with
3661  * rcu_read_lock held.
3662  */
3663 #define for_each_dev_addr(dev, ha) \
3664 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3665 
3666 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3667 
3668 void ether_setup(struct net_device *dev);
3669 
3670 /* Support for loadable net-drivers */
3671 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3672 				    unsigned char name_assign_type,
3673 				    void (*setup)(struct net_device *),
3674 				    unsigned int txqs, unsigned int rxqs);
3675 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3676 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3677 
3678 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3679 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3680 			 count)
3681 
3682 int register_netdev(struct net_device *dev);
3683 void unregister_netdev(struct net_device *dev);
3684 
3685 /* General hardware address lists handling functions */
3686 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3687 		   struct netdev_hw_addr_list *from_list, int addr_len);
3688 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3689 		      struct netdev_hw_addr_list *from_list, int addr_len);
3690 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3691 		       struct net_device *dev,
3692 		       int (*sync)(struct net_device *, const unsigned char *),
3693 		       int (*unsync)(struct net_device *,
3694 				     const unsigned char *));
3695 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3696 			  struct net_device *dev,
3697 			  int (*unsync)(struct net_device *,
3698 					const unsigned char *));
3699 void __hw_addr_init(struct netdev_hw_addr_list *list);
3700 
3701 /* Functions used for device addresses handling */
3702 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3703 		 unsigned char addr_type);
3704 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3705 		 unsigned char addr_type);
3706 void dev_addr_flush(struct net_device *dev);
3707 int dev_addr_init(struct net_device *dev);
3708 
3709 /* Functions used for unicast addresses handling */
3710 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3711 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3712 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3713 int dev_uc_sync(struct net_device *to, struct net_device *from);
3714 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3715 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3716 void dev_uc_flush(struct net_device *dev);
3717 void dev_uc_init(struct net_device *dev);
3718 
3719 /**
3720  *  __dev_uc_sync - Synchonize device's unicast list
3721  *  @dev:  device to sync
3722  *  @sync: function to call if address should be added
3723  *  @unsync: function to call if address should be removed
3724  *
3725  *  Add newly added addresses to the interface, and release
3726  *  addresses that have been deleted.
3727  */
3728 static inline int __dev_uc_sync(struct net_device *dev,
3729 				int (*sync)(struct net_device *,
3730 					    const unsigned char *),
3731 				int (*unsync)(struct net_device *,
3732 					      const unsigned char *))
3733 {
3734 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3735 }
3736 
3737 /**
3738  *  __dev_uc_unsync - Remove synchronized addresses from device
3739  *  @dev:  device to sync
3740  *  @unsync: function to call if address should be removed
3741  *
3742  *  Remove all addresses that were added to the device by dev_uc_sync().
3743  */
3744 static inline void __dev_uc_unsync(struct net_device *dev,
3745 				   int (*unsync)(struct net_device *,
3746 						 const unsigned char *))
3747 {
3748 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3749 }
3750 
3751 /* Functions used for multicast addresses handling */
3752 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3753 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3754 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3755 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3756 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3757 int dev_mc_sync(struct net_device *to, struct net_device *from);
3758 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3759 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3760 void dev_mc_flush(struct net_device *dev);
3761 void dev_mc_init(struct net_device *dev);
3762 
3763 /**
3764  *  __dev_mc_sync - Synchonize device's multicast list
3765  *  @dev:  device to sync
3766  *  @sync: function to call if address should be added
3767  *  @unsync: function to call if address should be removed
3768  *
3769  *  Add newly added addresses to the interface, and release
3770  *  addresses that have been deleted.
3771  */
3772 static inline int __dev_mc_sync(struct net_device *dev,
3773 				int (*sync)(struct net_device *,
3774 					    const unsigned char *),
3775 				int (*unsync)(struct net_device *,
3776 					      const unsigned char *))
3777 {
3778 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3779 }
3780 
3781 /**
3782  *  __dev_mc_unsync - Remove synchronized addresses from device
3783  *  @dev:  device to sync
3784  *  @unsync: function to call if address should be removed
3785  *
3786  *  Remove all addresses that were added to the device by dev_mc_sync().
3787  */
3788 static inline void __dev_mc_unsync(struct net_device *dev,
3789 				   int (*unsync)(struct net_device *,
3790 						 const unsigned char *))
3791 {
3792 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3793 }
3794 
3795 /* Functions used for secondary unicast and multicast support */
3796 void dev_set_rx_mode(struct net_device *dev);
3797 void __dev_set_rx_mode(struct net_device *dev);
3798 int dev_set_promiscuity(struct net_device *dev, int inc);
3799 int dev_set_allmulti(struct net_device *dev, int inc);
3800 void netdev_state_change(struct net_device *dev);
3801 void netdev_notify_peers(struct net_device *dev);
3802 void netdev_features_change(struct net_device *dev);
3803 /* Load a device via the kmod */
3804 void dev_load(struct net *net, const char *name);
3805 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3806 					struct rtnl_link_stats64 *storage);
3807 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3808 			     const struct net_device_stats *netdev_stats);
3809 
3810 extern int		netdev_max_backlog;
3811 extern int		netdev_tstamp_prequeue;
3812 extern int		weight_p;
3813 extern int		dev_weight_rx_bias;
3814 extern int		dev_weight_tx_bias;
3815 extern int		dev_rx_weight;
3816 extern int		dev_tx_weight;
3817 
3818 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3819 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3820 						     struct list_head **iter);
3821 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3822 						     struct list_head **iter);
3823 
3824 /* iterate through upper list, must be called under RCU read lock */
3825 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3826 	for (iter = &(dev)->adj_list.upper, \
3827 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3828 	     updev; \
3829 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3830 
3831 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3832 				  int (*fn)(struct net_device *upper_dev,
3833 					    void *data),
3834 				  void *data);
3835 
3836 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3837 				  struct net_device *upper_dev);
3838 
3839 void *netdev_lower_get_next_private(struct net_device *dev,
3840 				    struct list_head **iter);
3841 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3842 					struct list_head **iter);
3843 
3844 #define netdev_for_each_lower_private(dev, priv, iter) \
3845 	for (iter = (dev)->adj_list.lower.next, \
3846 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3847 	     priv; \
3848 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3849 
3850 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3851 	for (iter = &(dev)->adj_list.lower, \
3852 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3853 	     priv; \
3854 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3855 
3856 void *netdev_lower_get_next(struct net_device *dev,
3857 				struct list_head **iter);
3858 
3859 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3860 	for (iter = (dev)->adj_list.lower.next, \
3861 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3862 	     ldev; \
3863 	     ldev = netdev_lower_get_next(dev, &(iter)))
3864 
3865 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3866 					     struct list_head **iter);
3867 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3868 						 struct list_head **iter);
3869 
3870 int netdev_walk_all_lower_dev(struct net_device *dev,
3871 			      int (*fn)(struct net_device *lower_dev,
3872 					void *data),
3873 			      void *data);
3874 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3875 				  int (*fn)(struct net_device *lower_dev,
3876 					    void *data),
3877 				  void *data);
3878 
3879 void *netdev_adjacent_get_private(struct list_head *adj_list);
3880 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3881 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3882 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3883 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3884 int netdev_master_upper_dev_link(struct net_device *dev,
3885 				 struct net_device *upper_dev,
3886 				 void *upper_priv, void *upper_info);
3887 void netdev_upper_dev_unlink(struct net_device *dev,
3888 			     struct net_device *upper_dev);
3889 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3890 void *netdev_lower_dev_get_private(struct net_device *dev,
3891 				   struct net_device *lower_dev);
3892 void netdev_lower_state_changed(struct net_device *lower_dev,
3893 				void *lower_state_info);
3894 
3895 /* RSS keys are 40 or 52 bytes long */
3896 #define NETDEV_RSS_KEY_LEN 52
3897 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3898 void netdev_rss_key_fill(void *buffer, size_t len);
3899 
3900 int dev_get_nest_level(struct net_device *dev);
3901 int skb_checksum_help(struct sk_buff *skb);
3902 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3903 				  netdev_features_t features, bool tx_path);
3904 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3905 				    netdev_features_t features);
3906 
3907 struct netdev_bonding_info {
3908 	ifslave	slave;
3909 	ifbond	master;
3910 };
3911 
3912 struct netdev_notifier_bonding_info {
3913 	struct netdev_notifier_info info; /* must be first */
3914 	struct netdev_bonding_info  bonding_info;
3915 };
3916 
3917 void netdev_bonding_info_change(struct net_device *dev,
3918 				struct netdev_bonding_info *bonding_info);
3919 
3920 static inline
3921 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3922 {
3923 	return __skb_gso_segment(skb, features, true);
3924 }
3925 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3926 
3927 static inline bool can_checksum_protocol(netdev_features_t features,
3928 					 __be16 protocol)
3929 {
3930 	if (protocol == htons(ETH_P_FCOE))
3931 		return !!(features & NETIF_F_FCOE_CRC);
3932 
3933 	/* Assume this is an IP checksum (not SCTP CRC) */
3934 
3935 	if (features & NETIF_F_HW_CSUM) {
3936 		/* Can checksum everything */
3937 		return true;
3938 	}
3939 
3940 	switch (protocol) {
3941 	case htons(ETH_P_IP):
3942 		return !!(features & NETIF_F_IP_CSUM);
3943 	case htons(ETH_P_IPV6):
3944 		return !!(features & NETIF_F_IPV6_CSUM);
3945 	default:
3946 		return false;
3947 	}
3948 }
3949 
3950 #ifdef CONFIG_BUG
3951 void netdev_rx_csum_fault(struct net_device *dev);
3952 #else
3953 static inline void netdev_rx_csum_fault(struct net_device *dev)
3954 {
3955 }
3956 #endif
3957 /* rx skb timestamps */
3958 void net_enable_timestamp(void);
3959 void net_disable_timestamp(void);
3960 
3961 #ifdef CONFIG_PROC_FS
3962 int __init dev_proc_init(void);
3963 #else
3964 #define dev_proc_init() 0
3965 #endif
3966 
3967 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3968 					      struct sk_buff *skb, struct net_device *dev,
3969 					      bool more)
3970 {
3971 	skb->xmit_more = more ? 1 : 0;
3972 	return ops->ndo_start_xmit(skb, dev);
3973 }
3974 
3975 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3976 					    struct netdev_queue *txq, bool more)
3977 {
3978 	const struct net_device_ops *ops = dev->netdev_ops;
3979 	int rc;
3980 
3981 	rc = __netdev_start_xmit(ops, skb, dev, more);
3982 	if (rc == NETDEV_TX_OK)
3983 		txq_trans_update(txq);
3984 
3985 	return rc;
3986 }
3987 
3988 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3989 				const void *ns);
3990 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3991 				 const void *ns);
3992 
3993 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3994 {
3995 	return netdev_class_create_file_ns(class_attr, NULL);
3996 }
3997 
3998 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3999 {
4000 	netdev_class_remove_file_ns(class_attr, NULL);
4001 }
4002 
4003 extern struct kobj_ns_type_operations net_ns_type_operations;
4004 
4005 const char *netdev_drivername(const struct net_device *dev);
4006 
4007 void linkwatch_run_queue(void);
4008 
4009 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4010 							  netdev_features_t f2)
4011 {
4012 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4013 		if (f1 & NETIF_F_HW_CSUM)
4014 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4015 		else
4016 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4017 	}
4018 
4019 	return f1 & f2;
4020 }
4021 
4022 static inline netdev_features_t netdev_get_wanted_features(
4023 	struct net_device *dev)
4024 {
4025 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4026 }
4027 netdev_features_t netdev_increment_features(netdev_features_t all,
4028 	netdev_features_t one, netdev_features_t mask);
4029 
4030 /* Allow TSO being used on stacked device :
4031  * Performing the GSO segmentation before last device
4032  * is a performance improvement.
4033  */
4034 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4035 							netdev_features_t mask)
4036 {
4037 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4038 }
4039 
4040 int __netdev_update_features(struct net_device *dev);
4041 void netdev_update_features(struct net_device *dev);
4042 void netdev_change_features(struct net_device *dev);
4043 
4044 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4045 					struct net_device *dev);
4046 
4047 netdev_features_t passthru_features_check(struct sk_buff *skb,
4048 					  struct net_device *dev,
4049 					  netdev_features_t features);
4050 netdev_features_t netif_skb_features(struct sk_buff *skb);
4051 
4052 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4053 {
4054 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4055 
4056 	/* check flags correspondence */
4057 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4058 	BUILD_BUG_ON(SKB_GSO_UDP     != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4059 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4060 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4061 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4062 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4063 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4064 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4065 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4066 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4067 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4068 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4069 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4070 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4071 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4072 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4073 
4074 	return (features & feature) == feature;
4075 }
4076 
4077 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4078 {
4079 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4080 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4081 }
4082 
4083 static inline bool netif_needs_gso(struct sk_buff *skb,
4084 				   netdev_features_t features)
4085 {
4086 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4087 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4088 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4089 }
4090 
4091 static inline void netif_set_gso_max_size(struct net_device *dev,
4092 					  unsigned int size)
4093 {
4094 	dev->gso_max_size = size;
4095 }
4096 
4097 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4098 					int pulled_hlen, u16 mac_offset,
4099 					int mac_len)
4100 {
4101 	skb->protocol = protocol;
4102 	skb->encapsulation = 1;
4103 	skb_push(skb, pulled_hlen);
4104 	skb_reset_transport_header(skb);
4105 	skb->mac_header = mac_offset;
4106 	skb->network_header = skb->mac_header + mac_len;
4107 	skb->mac_len = mac_len;
4108 }
4109 
4110 static inline bool netif_is_macsec(const struct net_device *dev)
4111 {
4112 	return dev->priv_flags & IFF_MACSEC;
4113 }
4114 
4115 static inline bool netif_is_macvlan(const struct net_device *dev)
4116 {
4117 	return dev->priv_flags & IFF_MACVLAN;
4118 }
4119 
4120 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4121 {
4122 	return dev->priv_flags & IFF_MACVLAN_PORT;
4123 }
4124 
4125 static inline bool netif_is_ipvlan(const struct net_device *dev)
4126 {
4127 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4128 }
4129 
4130 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4131 {
4132 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4133 }
4134 
4135 static inline bool netif_is_bond_master(const struct net_device *dev)
4136 {
4137 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4138 }
4139 
4140 static inline bool netif_is_bond_slave(const struct net_device *dev)
4141 {
4142 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4143 }
4144 
4145 static inline bool netif_supports_nofcs(struct net_device *dev)
4146 {
4147 	return dev->priv_flags & IFF_SUPP_NOFCS;
4148 }
4149 
4150 static inline bool netif_is_l3_master(const struct net_device *dev)
4151 {
4152 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4153 }
4154 
4155 static inline bool netif_is_l3_slave(const struct net_device *dev)
4156 {
4157 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4158 }
4159 
4160 static inline bool netif_is_bridge_master(const struct net_device *dev)
4161 {
4162 	return dev->priv_flags & IFF_EBRIDGE;
4163 }
4164 
4165 static inline bool netif_is_bridge_port(const struct net_device *dev)
4166 {
4167 	return dev->priv_flags & IFF_BRIDGE_PORT;
4168 }
4169 
4170 static inline bool netif_is_ovs_master(const struct net_device *dev)
4171 {
4172 	return dev->priv_flags & IFF_OPENVSWITCH;
4173 }
4174 
4175 static inline bool netif_is_team_master(const struct net_device *dev)
4176 {
4177 	return dev->priv_flags & IFF_TEAM;
4178 }
4179 
4180 static inline bool netif_is_team_port(const struct net_device *dev)
4181 {
4182 	return dev->priv_flags & IFF_TEAM_PORT;
4183 }
4184 
4185 static inline bool netif_is_lag_master(const struct net_device *dev)
4186 {
4187 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4188 }
4189 
4190 static inline bool netif_is_lag_port(const struct net_device *dev)
4191 {
4192 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4193 }
4194 
4195 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4196 {
4197 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4198 }
4199 
4200 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4201 static inline void netif_keep_dst(struct net_device *dev)
4202 {
4203 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4204 }
4205 
4206 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4207 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4208 {
4209 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4210 	return dev->priv_flags & IFF_MACSEC;
4211 }
4212 
4213 extern struct pernet_operations __net_initdata loopback_net_ops;
4214 
4215 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4216 
4217 /* netdev_printk helpers, similar to dev_printk */
4218 
4219 static inline const char *netdev_name(const struct net_device *dev)
4220 {
4221 	if (!dev->name[0] || strchr(dev->name, '%'))
4222 		return "(unnamed net_device)";
4223 	return dev->name;
4224 }
4225 
4226 static inline const char *netdev_reg_state(const struct net_device *dev)
4227 {
4228 	switch (dev->reg_state) {
4229 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4230 	case NETREG_REGISTERED: return "";
4231 	case NETREG_UNREGISTERING: return " (unregistering)";
4232 	case NETREG_UNREGISTERED: return " (unregistered)";
4233 	case NETREG_RELEASED: return " (released)";
4234 	case NETREG_DUMMY: return " (dummy)";
4235 	}
4236 
4237 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4238 	return " (unknown)";
4239 }
4240 
4241 __printf(3, 4)
4242 void netdev_printk(const char *level, const struct net_device *dev,
4243 		   const char *format, ...);
4244 __printf(2, 3)
4245 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4246 __printf(2, 3)
4247 void netdev_alert(const struct net_device *dev, const char *format, ...);
4248 __printf(2, 3)
4249 void netdev_crit(const struct net_device *dev, const char *format, ...);
4250 __printf(2, 3)
4251 void netdev_err(const struct net_device *dev, const char *format, ...);
4252 __printf(2, 3)
4253 void netdev_warn(const struct net_device *dev, const char *format, ...);
4254 __printf(2, 3)
4255 void netdev_notice(const struct net_device *dev, const char *format, ...);
4256 __printf(2, 3)
4257 void netdev_info(const struct net_device *dev, const char *format, ...);
4258 
4259 #define MODULE_ALIAS_NETDEV(device) \
4260 	MODULE_ALIAS("netdev-" device)
4261 
4262 #if defined(CONFIG_DYNAMIC_DEBUG)
4263 #define netdev_dbg(__dev, format, args...)			\
4264 do {								\
4265 	dynamic_netdev_dbg(__dev, format, ##args);		\
4266 } while (0)
4267 #elif defined(DEBUG)
4268 #define netdev_dbg(__dev, format, args...)			\
4269 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4270 #else
4271 #define netdev_dbg(__dev, format, args...)			\
4272 ({								\
4273 	if (0)							\
4274 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4275 })
4276 #endif
4277 
4278 #if defined(VERBOSE_DEBUG)
4279 #define netdev_vdbg	netdev_dbg
4280 #else
4281 
4282 #define netdev_vdbg(dev, format, args...)			\
4283 ({								\
4284 	if (0)							\
4285 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4286 	0;							\
4287 })
4288 #endif
4289 
4290 /*
4291  * netdev_WARN() acts like dev_printk(), but with the key difference
4292  * of using a WARN/WARN_ON to get the message out, including the
4293  * file/line information and a backtrace.
4294  */
4295 #define netdev_WARN(dev, format, args...)			\
4296 	WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),	\
4297 	     netdev_reg_state(dev), ##args)
4298 
4299 /* netif printk helpers, similar to netdev_printk */
4300 
4301 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4302 do {					  			\
4303 	if (netif_msg_##type(priv))				\
4304 		netdev_printk(level, (dev), fmt, ##args);	\
4305 } while (0)
4306 
4307 #define netif_level(level, priv, type, dev, fmt, args...)	\
4308 do {								\
4309 	if (netif_msg_##type(priv))				\
4310 		netdev_##level(dev, fmt, ##args);		\
4311 } while (0)
4312 
4313 #define netif_emerg(priv, type, dev, fmt, args...)		\
4314 	netif_level(emerg, priv, type, dev, fmt, ##args)
4315 #define netif_alert(priv, type, dev, fmt, args...)		\
4316 	netif_level(alert, priv, type, dev, fmt, ##args)
4317 #define netif_crit(priv, type, dev, fmt, args...)		\
4318 	netif_level(crit, priv, type, dev, fmt, ##args)
4319 #define netif_err(priv, type, dev, fmt, args...)		\
4320 	netif_level(err, priv, type, dev, fmt, ##args)
4321 #define netif_warn(priv, type, dev, fmt, args...)		\
4322 	netif_level(warn, priv, type, dev, fmt, ##args)
4323 #define netif_notice(priv, type, dev, fmt, args...)		\
4324 	netif_level(notice, priv, type, dev, fmt, ##args)
4325 #define netif_info(priv, type, dev, fmt, args...)		\
4326 	netif_level(info, priv, type, dev, fmt, ##args)
4327 
4328 #if defined(CONFIG_DYNAMIC_DEBUG)
4329 #define netif_dbg(priv, type, netdev, format, args...)		\
4330 do {								\
4331 	if (netif_msg_##type(priv))				\
4332 		dynamic_netdev_dbg(netdev, format, ##args);	\
4333 } while (0)
4334 #elif defined(DEBUG)
4335 #define netif_dbg(priv, type, dev, format, args...)		\
4336 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4337 #else
4338 #define netif_dbg(priv, type, dev, format, args...)			\
4339 ({									\
4340 	if (0)								\
4341 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4342 	0;								\
4343 })
4344 #endif
4345 
4346 /* if @cond then downgrade to debug, else print at @level */
4347 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4348 	do {                                                              \
4349 		if (cond)                                                 \
4350 			netif_dbg(priv, type, netdev, fmt, ##args);       \
4351 		else                                                      \
4352 			netif_ ## level(priv, type, netdev, fmt, ##args); \
4353 	} while (0)
4354 
4355 #if defined(VERBOSE_DEBUG)
4356 #define netif_vdbg	netif_dbg
4357 #else
4358 #define netif_vdbg(priv, type, dev, format, args...)		\
4359 ({								\
4360 	if (0)							\
4361 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4362 	0;							\
4363 })
4364 #endif
4365 
4366 /*
4367  *	The list of packet types we will receive (as opposed to discard)
4368  *	and the routines to invoke.
4369  *
4370  *	Why 16. Because with 16 the only overlap we get on a hash of the
4371  *	low nibble of the protocol value is RARP/SNAP/X.25.
4372  *
4373  *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4374  *             sure which should go first, but I bet it won't make much
4375  *             difference if we are running VLANs.  The good news is that
4376  *             this protocol won't be in the list unless compiled in, so
4377  *             the average user (w/out VLANs) will not be adversely affected.
4378  *             --BLG
4379  *
4380  *		0800	IP
4381  *		8100    802.1Q VLAN
4382  *		0001	802.3
4383  *		0002	AX.25
4384  *		0004	802.2
4385  *		8035	RARP
4386  *		0005	SNAP
4387  *		0805	X.25
4388  *		0806	ARP
4389  *		8137	IPX
4390  *		0009	Localtalk
4391  *		86DD	IPv6
4392  */
4393 #define PTYPE_HASH_SIZE	(16)
4394 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4395 
4396 #endif	/* _LINUX_NETDEVICE_H */
4397