xref: /openbmc/linux/include/linux/netdevice.h (revision b24413180f5600bcb3bb70fbed5cf186b60864bd)
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/workqueue.h>
39 #include <linux/dynamic_queue_limits.h>
40 
41 #include <linux/ethtool.h>
42 #include <net/net_namespace.h>
43 #ifdef CONFIG_DCB
44 #include <net/dcbnl.h>
45 #endif
46 #include <net/netprio_cgroup.h>
47 
48 #include <linux/netdev_features.h>
49 #include <linux/neighbour.h>
50 #include <uapi/linux/netdevice.h>
51 #include <uapi/linux/if_bonding.h>
52 #include <uapi/linux/pkt_cls.h>
53 #include <linux/hashtable.h>
54 
55 struct netpoll_info;
56 struct device;
57 struct phy_device;
58 struct dsa_switch_tree;
59 
60 /* 802.11 specific */
61 struct wireless_dev;
62 /* 802.15.4 specific */
63 struct wpan_dev;
64 struct mpls_dev;
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
67 struct bpf_prog;
68 struct xdp_buff;
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 	unsigned int	hh_len;
241 	seqlock_t	hh_lock;
242 
243 	/* cached hardware header; allow for machine alignment needs.        */
244 #define HH_DATA_MOD	16
245 #define HH_DATA_OFF(__len) \
246 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250 };
251 
252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
253  * Alternative is:
254  *   dev->hard_header_len ? (dev->hard_header_len +
255  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256  *
257  * We could use other alignment values, but we must maintain the
258  * relationship HH alignment <= LL alignment.
259  */
260 #define LL_RESERVED_SPACE(dev) \
261 	((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 	((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264 
265 struct header_ops {
266 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
267 			   unsigned short type, const void *daddr,
268 			   const void *saddr, unsigned int len);
269 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
271 	void	(*cache_update)(struct hh_cache *hh,
272 				const struct net_device *dev,
273 				const unsigned char *haddr);
274 	bool	(*validate)(const char *ll_header, unsigned int len);
275 };
276 
277 /* These flag bits are private to the generic network queueing
278  * layer; they may not be explicitly referenced by any other
279  * code.
280  */
281 
282 enum netdev_state_t {
283 	__LINK_STATE_START,
284 	__LINK_STATE_PRESENT,
285 	__LINK_STATE_NOCARRIER,
286 	__LINK_STATE_LINKWATCH_PENDING,
287 	__LINK_STATE_DORMANT,
288 };
289 
290 
291 /*
292  * This structure holds boot-time configured netdevice settings. They
293  * are then used in the device probing.
294  */
295 struct netdev_boot_setup {
296 	char name[IFNAMSIZ];
297 	struct ifmap map;
298 };
299 #define NETDEV_BOOT_SETUP_MAX 8
300 
301 int __init netdev_boot_setup(char *str);
302 
303 /*
304  * Structure for NAPI scheduling similar to tasklet but with weighting
305  */
306 struct napi_struct {
307 	/* The poll_list must only be managed by the entity which
308 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
309 	 * whoever atomically sets that bit can add this napi_struct
310 	 * to the per-CPU poll_list, and whoever clears that bit
311 	 * can remove from the list right before clearing the bit.
312 	 */
313 	struct list_head	poll_list;
314 
315 	unsigned long		state;
316 	int			weight;
317 	unsigned int		gro_count;
318 	int			(*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
320 	int			poll_owner;
321 #endif
322 	struct net_device	*dev;
323 	struct sk_buff		*gro_list;
324 	struct sk_buff		*skb;
325 	struct hrtimer		timer;
326 	struct list_head	dev_list;
327 	struct hlist_node	napi_hash_node;
328 	unsigned int		napi_id;
329 };
330 
331 enum {
332 	NAPI_STATE_SCHED,	/* Poll is scheduled */
333 	NAPI_STATE_MISSED,	/* reschedule a napi */
334 	NAPI_STATE_DISABLE,	/* Disable pending */
335 	NAPI_STATE_NPSVC,	/* Netpoll - don't dequeue from poll_list */
336 	NAPI_STATE_HASHED,	/* In NAPI hash (busy polling possible) */
337 	NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
338 	NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
339 };
340 
341 enum {
342 	NAPIF_STATE_SCHED	 = BIT(NAPI_STATE_SCHED),
343 	NAPIF_STATE_MISSED	 = BIT(NAPI_STATE_MISSED),
344 	NAPIF_STATE_DISABLE	 = BIT(NAPI_STATE_DISABLE),
345 	NAPIF_STATE_NPSVC	 = BIT(NAPI_STATE_NPSVC),
346 	NAPIF_STATE_HASHED	 = BIT(NAPI_STATE_HASHED),
347 	NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
348 	NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
349 };
350 
351 enum gro_result {
352 	GRO_MERGED,
353 	GRO_MERGED_FREE,
354 	GRO_HELD,
355 	GRO_NORMAL,
356 	GRO_DROP,
357 	GRO_CONSUMED,
358 };
359 typedef enum gro_result gro_result_t;
360 
361 /*
362  * enum rx_handler_result - Possible return values for rx_handlers.
363  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
364  * further.
365  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366  * case skb->dev was changed by rx_handler.
367  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
368  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
369  *
370  * rx_handlers are functions called from inside __netif_receive_skb(), to do
371  * special processing of the skb, prior to delivery to protocol handlers.
372  *
373  * Currently, a net_device can only have a single rx_handler registered. Trying
374  * to register a second rx_handler will return -EBUSY.
375  *
376  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377  * To unregister a rx_handler on a net_device, use
378  * netdev_rx_handler_unregister().
379  *
380  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
381  * do with the skb.
382  *
383  * If the rx_handler consumed the skb in some way, it should return
384  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
385  * the skb to be delivered in some other way.
386  *
387  * If the rx_handler changed skb->dev, to divert the skb to another
388  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389  * new device will be called if it exists.
390  *
391  * If the rx_handler decides the skb should be ignored, it should return
392  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
393  * are registered on exact device (ptype->dev == skb->dev).
394  *
395  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
396  * delivered, it should return RX_HANDLER_PASS.
397  *
398  * A device without a registered rx_handler will behave as if rx_handler
399  * returned RX_HANDLER_PASS.
400  */
401 
402 enum rx_handler_result {
403 	RX_HANDLER_CONSUMED,
404 	RX_HANDLER_ANOTHER,
405 	RX_HANDLER_EXACT,
406 	RX_HANDLER_PASS,
407 };
408 typedef enum rx_handler_result rx_handler_result_t;
409 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
410 
411 void __napi_schedule(struct napi_struct *n);
412 void __napi_schedule_irqoff(struct napi_struct *n);
413 
414 static inline bool napi_disable_pending(struct napi_struct *n)
415 {
416 	return test_bit(NAPI_STATE_DISABLE, &n->state);
417 }
418 
419 bool napi_schedule_prep(struct napi_struct *n);
420 
421 /**
422  *	napi_schedule - schedule NAPI poll
423  *	@n: NAPI context
424  *
425  * Schedule NAPI poll routine to be called if it is not already
426  * running.
427  */
428 static inline void napi_schedule(struct napi_struct *n)
429 {
430 	if (napi_schedule_prep(n))
431 		__napi_schedule(n);
432 }
433 
434 /**
435  *	napi_schedule_irqoff - schedule NAPI poll
436  *	@n: NAPI context
437  *
438  * Variant of napi_schedule(), assuming hard irqs are masked.
439  */
440 static inline void napi_schedule_irqoff(struct napi_struct *n)
441 {
442 	if (napi_schedule_prep(n))
443 		__napi_schedule_irqoff(n);
444 }
445 
446 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
447 static inline bool napi_reschedule(struct napi_struct *napi)
448 {
449 	if (napi_schedule_prep(napi)) {
450 		__napi_schedule(napi);
451 		return true;
452 	}
453 	return false;
454 }
455 
456 bool napi_complete_done(struct napi_struct *n, int work_done);
457 /**
458  *	napi_complete - NAPI processing complete
459  *	@n: NAPI context
460  *
461  * Mark NAPI processing as complete.
462  * Consider using napi_complete_done() instead.
463  * Return false if device should avoid rearming interrupts.
464  */
465 static inline bool napi_complete(struct napi_struct *n)
466 {
467 	return napi_complete_done(n, 0);
468 }
469 
470 /**
471  *	napi_hash_del - remove a NAPI from global table
472  *	@napi: NAPI context
473  *
474  * Warning: caller must observe RCU grace period
475  * before freeing memory containing @napi, if
476  * this function returns true.
477  * Note: core networking stack automatically calls it
478  * from netif_napi_del().
479  * Drivers might want to call this helper to combine all
480  * the needed RCU grace periods into a single one.
481  */
482 bool napi_hash_del(struct napi_struct *napi);
483 
484 /**
485  *	napi_disable - prevent NAPI from scheduling
486  *	@n: NAPI context
487  *
488  * Stop NAPI from being scheduled on this context.
489  * Waits till any outstanding processing completes.
490  */
491 void napi_disable(struct napi_struct *n);
492 
493 /**
494  *	napi_enable - enable NAPI scheduling
495  *	@n: NAPI context
496  *
497  * Resume NAPI from being scheduled on this context.
498  * Must be paired with napi_disable.
499  */
500 static inline void napi_enable(struct napi_struct *n)
501 {
502 	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
503 	smp_mb__before_atomic();
504 	clear_bit(NAPI_STATE_SCHED, &n->state);
505 	clear_bit(NAPI_STATE_NPSVC, &n->state);
506 }
507 
508 /**
509  *	napi_synchronize - wait until NAPI is not running
510  *	@n: NAPI context
511  *
512  * Wait until NAPI is done being scheduled on this context.
513  * Waits till any outstanding processing completes but
514  * does not disable future activations.
515  */
516 static inline void napi_synchronize(const struct napi_struct *n)
517 {
518 	if (IS_ENABLED(CONFIG_SMP))
519 		while (test_bit(NAPI_STATE_SCHED, &n->state))
520 			msleep(1);
521 	else
522 		barrier();
523 }
524 
525 enum netdev_queue_state_t {
526 	__QUEUE_STATE_DRV_XOFF,
527 	__QUEUE_STATE_STACK_XOFF,
528 	__QUEUE_STATE_FROZEN,
529 };
530 
531 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
532 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
533 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
534 
535 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
537 					QUEUE_STATE_FROZEN)
538 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
539 					QUEUE_STATE_FROZEN)
540 
541 /*
542  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
543  * netif_tx_* functions below are used to manipulate this flag.  The
544  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545  * queue independently.  The netif_xmit_*stopped functions below are called
546  * to check if the queue has been stopped by the driver or stack (either
547  * of the XOFF bits are set in the state).  Drivers should not need to call
548  * netif_xmit*stopped functions, they should only be using netif_tx_*.
549  */
550 
551 struct netdev_queue {
552 /*
553  * read-mostly part
554  */
555 	struct net_device	*dev;
556 	struct Qdisc __rcu	*qdisc;
557 	struct Qdisc		*qdisc_sleeping;
558 #ifdef CONFIG_SYSFS
559 	struct kobject		kobj;
560 #endif
561 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 	int			numa_node;
563 #endif
564 	unsigned long		tx_maxrate;
565 	/*
566 	 * Number of TX timeouts for this queue
567 	 * (/sys/class/net/DEV/Q/trans_timeout)
568 	 */
569 	unsigned long		trans_timeout;
570 /*
571  * write-mostly part
572  */
573 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
574 	int			xmit_lock_owner;
575 	/*
576 	 * Time (in jiffies) of last Tx
577 	 */
578 	unsigned long		trans_start;
579 
580 	unsigned long		state;
581 
582 #ifdef CONFIG_BQL
583 	struct dql		dql;
584 #endif
585 } ____cacheline_aligned_in_smp;
586 
587 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
588 {
589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 	return q->numa_node;
591 #else
592 	return NUMA_NO_NODE;
593 #endif
594 }
595 
596 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
597 {
598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 	q->numa_node = node;
600 #endif
601 }
602 
603 #ifdef CONFIG_RPS
604 /*
605  * This structure holds an RPS map which can be of variable length.  The
606  * map is an array of CPUs.
607  */
608 struct rps_map {
609 	unsigned int len;
610 	struct rcu_head rcu;
611 	u16 cpus[0];
612 };
613 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
614 
615 /*
616  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617  * tail pointer for that CPU's input queue at the time of last enqueue, and
618  * a hardware filter index.
619  */
620 struct rps_dev_flow {
621 	u16 cpu;
622 	u16 filter;
623 	unsigned int last_qtail;
624 };
625 #define RPS_NO_FILTER 0xffff
626 
627 /*
628  * The rps_dev_flow_table structure contains a table of flow mappings.
629  */
630 struct rps_dev_flow_table {
631 	unsigned int mask;
632 	struct rcu_head rcu;
633 	struct rps_dev_flow flows[0];
634 };
635 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
636     ((_num) * sizeof(struct rps_dev_flow)))
637 
638 /*
639  * The rps_sock_flow_table contains mappings of flows to the last CPU
640  * on which they were processed by the application (set in recvmsg).
641  * Each entry is a 32bit value. Upper part is the high-order bits
642  * of flow hash, lower part is CPU number.
643  * rps_cpu_mask is used to partition the space, depending on number of
644  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
646  * meaning we use 32-6=26 bits for the hash.
647  */
648 struct rps_sock_flow_table {
649 	u32	mask;
650 
651 	u32	ents[0] ____cacheline_aligned_in_smp;
652 };
653 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
654 
655 #define RPS_NO_CPU 0xffff
656 
657 extern u32 rps_cpu_mask;
658 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
659 
660 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
661 					u32 hash)
662 {
663 	if (table && hash) {
664 		unsigned int index = hash & table->mask;
665 		u32 val = hash & ~rps_cpu_mask;
666 
667 		/* We only give a hint, preemption can change CPU under us */
668 		val |= raw_smp_processor_id();
669 
670 		if (table->ents[index] != val)
671 			table->ents[index] = val;
672 	}
673 }
674 
675 #ifdef CONFIG_RFS_ACCEL
676 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
677 			 u16 filter_id);
678 #endif
679 #endif /* CONFIG_RPS */
680 
681 /* This structure contains an instance of an RX queue. */
682 struct netdev_rx_queue {
683 #ifdef CONFIG_RPS
684 	struct rps_map __rcu		*rps_map;
685 	struct rps_dev_flow_table __rcu	*rps_flow_table;
686 #endif
687 	struct kobject			kobj;
688 	struct net_device		*dev;
689 } ____cacheline_aligned_in_smp;
690 
691 /*
692  * RX queue sysfs structures and functions.
693  */
694 struct rx_queue_attribute {
695 	struct attribute attr;
696 	ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
697 	ssize_t (*store)(struct netdev_rx_queue *queue,
698 			 const char *buf, size_t len);
699 };
700 
701 #ifdef CONFIG_XPS
702 /*
703  * This structure holds an XPS map which can be of variable length.  The
704  * map is an array of queues.
705  */
706 struct xps_map {
707 	unsigned int len;
708 	unsigned int alloc_len;
709 	struct rcu_head rcu;
710 	u16 queues[0];
711 };
712 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
713 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
714        - sizeof(struct xps_map)) / sizeof(u16))
715 
716 /*
717  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
718  */
719 struct xps_dev_maps {
720 	struct rcu_head rcu;
721 	struct xps_map __rcu *cpu_map[0];
722 };
723 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +		\
724 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
725 #endif /* CONFIG_XPS */
726 
727 #define TC_MAX_QUEUE	16
728 #define TC_BITMASK	15
729 /* HW offloaded queuing disciplines txq count and offset maps */
730 struct netdev_tc_txq {
731 	u16 count;
732 	u16 offset;
733 };
734 
735 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
736 /*
737  * This structure is to hold information about the device
738  * configured to run FCoE protocol stack.
739  */
740 struct netdev_fcoe_hbainfo {
741 	char	manufacturer[64];
742 	char	serial_number[64];
743 	char	hardware_version[64];
744 	char	driver_version[64];
745 	char	optionrom_version[64];
746 	char	firmware_version[64];
747 	char	model[256];
748 	char	model_description[256];
749 };
750 #endif
751 
752 #define MAX_PHYS_ITEM_ID_LEN 32
753 
754 /* This structure holds a unique identifier to identify some
755  * physical item (port for example) used by a netdevice.
756  */
757 struct netdev_phys_item_id {
758 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
759 	unsigned char id_len;
760 };
761 
762 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
763 					    struct netdev_phys_item_id *b)
764 {
765 	return a->id_len == b->id_len &&
766 	       memcmp(a->id, b->id, a->id_len) == 0;
767 }
768 
769 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 				       struct sk_buff *skb);
771 
772 enum tc_setup_type {
773 	TC_SETUP_MQPRIO,
774 	TC_SETUP_CLSU32,
775 	TC_SETUP_CLSFLOWER,
776 	TC_SETUP_CLSMATCHALL,
777 	TC_SETUP_CLSBPF,
778 };
779 
780 /* These structures hold the attributes of xdp state that are being passed
781  * to the netdevice through the xdp op.
782  */
783 enum xdp_netdev_command {
784 	/* Set or clear a bpf program used in the earliest stages of packet
785 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
786 	 * is responsible for calling bpf_prog_put on any old progs that are
787 	 * stored. In case of error, the callee need not release the new prog
788 	 * reference, but on success it takes ownership and must bpf_prog_put
789 	 * when it is no longer used.
790 	 */
791 	XDP_SETUP_PROG,
792 	XDP_SETUP_PROG_HW,
793 	/* Check if a bpf program is set on the device.  The callee should
794 	 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
795 	 * is equivalent to XDP_ATTACHED_DRV.
796 	 */
797 	XDP_QUERY_PROG,
798 };
799 
800 struct netlink_ext_ack;
801 
802 struct netdev_xdp {
803 	enum xdp_netdev_command command;
804 	union {
805 		/* XDP_SETUP_PROG */
806 		struct {
807 			u32 flags;
808 			struct bpf_prog *prog;
809 			struct netlink_ext_ack *extack;
810 		};
811 		/* XDP_QUERY_PROG */
812 		struct {
813 			u8 prog_attached;
814 			u32 prog_id;
815 		};
816 	};
817 };
818 
819 #ifdef CONFIG_XFRM_OFFLOAD
820 struct xfrmdev_ops {
821 	int	(*xdo_dev_state_add) (struct xfrm_state *x);
822 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
823 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
824 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
825 				       struct xfrm_state *x);
826 };
827 #endif
828 
829 /*
830  * This structure defines the management hooks for network devices.
831  * The following hooks can be defined; unless noted otherwise, they are
832  * optional and can be filled with a null pointer.
833  *
834  * int (*ndo_init)(struct net_device *dev);
835  *     This function is called once when a network device is registered.
836  *     The network device can use this for any late stage initialization
837  *     or semantic validation. It can fail with an error code which will
838  *     be propagated back to register_netdev.
839  *
840  * void (*ndo_uninit)(struct net_device *dev);
841  *     This function is called when device is unregistered or when registration
842  *     fails. It is not called if init fails.
843  *
844  * int (*ndo_open)(struct net_device *dev);
845  *     This function is called when a network device transitions to the up
846  *     state.
847  *
848  * int (*ndo_stop)(struct net_device *dev);
849  *     This function is called when a network device transitions to the down
850  *     state.
851  *
852  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
853  *                               struct net_device *dev);
854  *	Called when a packet needs to be transmitted.
855  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
856  *	the queue before that can happen; it's for obsolete devices and weird
857  *	corner cases, but the stack really does a non-trivial amount
858  *	of useless work if you return NETDEV_TX_BUSY.
859  *	Required; cannot be NULL.
860  *
861  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
862  *					   struct net_device *dev
863  *					   netdev_features_t features);
864  *	Called by core transmit path to determine if device is capable of
865  *	performing offload operations on a given packet. This is to give
866  *	the device an opportunity to implement any restrictions that cannot
867  *	be otherwise expressed by feature flags. The check is called with
868  *	the set of features that the stack has calculated and it returns
869  *	those the driver believes to be appropriate.
870  *
871  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
872  *                         void *accel_priv, select_queue_fallback_t fallback);
873  *	Called to decide which queue to use when device supports multiple
874  *	transmit queues.
875  *
876  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
877  *	This function is called to allow device receiver to make
878  *	changes to configuration when multicast or promiscuous is enabled.
879  *
880  * void (*ndo_set_rx_mode)(struct net_device *dev);
881  *	This function is called device changes address list filtering.
882  *	If driver handles unicast address filtering, it should set
883  *	IFF_UNICAST_FLT in its priv_flags.
884  *
885  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
886  *	This function  is called when the Media Access Control address
887  *	needs to be changed. If this interface is not defined, the
888  *	MAC address can not be changed.
889  *
890  * int (*ndo_validate_addr)(struct net_device *dev);
891  *	Test if Media Access Control address is valid for the device.
892  *
893  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
894  *	Called when a user requests an ioctl which can't be handled by
895  *	the generic interface code. If not defined ioctls return
896  *	not supported error code.
897  *
898  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
899  *	Used to set network devices bus interface parameters. This interface
900  *	is retained for legacy reasons; new devices should use the bus
901  *	interface (PCI) for low level management.
902  *
903  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
904  *	Called when a user wants to change the Maximum Transfer Unit
905  *	of a device.
906  *
907  * void (*ndo_tx_timeout)(struct net_device *dev);
908  *	Callback used when the transmitter has not made any progress
909  *	for dev->watchdog ticks.
910  *
911  * void (*ndo_get_stats64)(struct net_device *dev,
912  *                         struct rtnl_link_stats64 *storage);
913  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
914  *	Called when a user wants to get the network device usage
915  *	statistics. Drivers must do one of the following:
916  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
917  *	   rtnl_link_stats64 structure passed by the caller.
918  *	2. Define @ndo_get_stats to update a net_device_stats structure
919  *	   (which should normally be dev->stats) and return a pointer to
920  *	   it. The structure may be changed asynchronously only if each
921  *	   field is written atomically.
922  *	3. Update dev->stats asynchronously and atomically, and define
923  *	   neither operation.
924  *
925  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
926  *	Return true if this device supports offload stats of this attr_id.
927  *
928  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
929  *	void *attr_data)
930  *	Get statistics for offload operations by attr_id. Write it into the
931  *	attr_data pointer.
932  *
933  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
934  *	If device supports VLAN filtering this function is called when a
935  *	VLAN id is registered.
936  *
937  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
938  *	If device supports VLAN filtering this function is called when a
939  *	VLAN id is unregistered.
940  *
941  * void (*ndo_poll_controller)(struct net_device *dev);
942  *
943  *	SR-IOV management functions.
944  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
945  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
946  *			  u8 qos, __be16 proto);
947  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
948  *			  int max_tx_rate);
949  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
950  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
951  * int (*ndo_get_vf_config)(struct net_device *dev,
952  *			    int vf, struct ifla_vf_info *ivf);
953  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
954  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
955  *			  struct nlattr *port[]);
956  *
957  *      Enable or disable the VF ability to query its RSS Redirection Table and
958  *      Hash Key. This is needed since on some devices VF share this information
959  *      with PF and querying it may introduce a theoretical security risk.
960  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
961  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
962  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
963  *		       void *type_data);
964  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
965  *	This is always called from the stack with the rtnl lock held and netif
966  *	tx queues stopped. This allows the netdevice to perform queue
967  *	management safely.
968  *
969  *	Fiber Channel over Ethernet (FCoE) offload functions.
970  * int (*ndo_fcoe_enable)(struct net_device *dev);
971  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
972  *	so the underlying device can perform whatever needed configuration or
973  *	initialization to support acceleration of FCoE traffic.
974  *
975  * int (*ndo_fcoe_disable)(struct net_device *dev);
976  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
977  *	so the underlying device can perform whatever needed clean-ups to
978  *	stop supporting acceleration of FCoE traffic.
979  *
980  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
981  *			     struct scatterlist *sgl, unsigned int sgc);
982  *	Called when the FCoE Initiator wants to initialize an I/O that
983  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
984  *	perform necessary setup and returns 1 to indicate the device is set up
985  *	successfully to perform DDP on this I/O, otherwise this returns 0.
986  *
987  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
988  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
989  *	indicated by the FC exchange id 'xid', so the underlying device can
990  *	clean up and reuse resources for later DDP requests.
991  *
992  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
993  *			      struct scatterlist *sgl, unsigned int sgc);
994  *	Called when the FCoE Target wants to initialize an I/O that
995  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
996  *	perform necessary setup and returns 1 to indicate the device is set up
997  *	successfully to perform DDP on this I/O, otherwise this returns 0.
998  *
999  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1000  *			       struct netdev_fcoe_hbainfo *hbainfo);
1001  *	Called when the FCoE Protocol stack wants information on the underlying
1002  *	device. This information is utilized by the FCoE protocol stack to
1003  *	register attributes with Fiber Channel management service as per the
1004  *	FC-GS Fabric Device Management Information(FDMI) specification.
1005  *
1006  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1007  *	Called when the underlying device wants to override default World Wide
1008  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1009  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1010  *	protocol stack to use.
1011  *
1012  *	RFS acceleration.
1013  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1014  *			    u16 rxq_index, u32 flow_id);
1015  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1016  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1017  *	Return the filter ID on success, or a negative error code.
1018  *
1019  *	Slave management functions (for bridge, bonding, etc).
1020  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1021  *	Called to make another netdev an underling.
1022  *
1023  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1024  *	Called to release previously enslaved netdev.
1025  *
1026  *      Feature/offload setting functions.
1027  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1028  *		netdev_features_t features);
1029  *	Adjusts the requested feature flags according to device-specific
1030  *	constraints, and returns the resulting flags. Must not modify
1031  *	the device state.
1032  *
1033  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1034  *	Called to update device configuration to new features. Passed
1035  *	feature set might be less than what was returned by ndo_fix_features()).
1036  *	Must return >0 or -errno if it changed dev->features itself.
1037  *
1038  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1039  *		      struct net_device *dev,
1040  *		      const unsigned char *addr, u16 vid, u16 flags)
1041  *	Adds an FDB entry to dev for addr.
1042  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1043  *		      struct net_device *dev,
1044  *		      const unsigned char *addr, u16 vid)
1045  *	Deletes the FDB entry from dev coresponding to addr.
1046  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1047  *		       struct net_device *dev, struct net_device *filter_dev,
1048  *		       int *idx)
1049  *	Used to add FDB entries to dump requests. Implementers should add
1050  *	entries to skb and update idx with the number of entries.
1051  *
1052  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1053  *			     u16 flags)
1054  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1055  *			     struct net_device *dev, u32 filter_mask,
1056  *			     int nlflags)
1057  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1058  *			     u16 flags);
1059  *
1060  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1061  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1062  *	which do not represent real hardware may define this to allow their
1063  *	userspace components to manage their virtual carrier state. Devices
1064  *	that determine carrier state from physical hardware properties (eg
1065  *	network cables) or protocol-dependent mechanisms (eg
1066  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1067  *
1068  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1069  *			       struct netdev_phys_item_id *ppid);
1070  *	Called to get ID of physical port of this device. If driver does
1071  *	not implement this, it is assumed that the hw is not able to have
1072  *	multiple net devices on single physical port.
1073  *
1074  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1075  *			      struct udp_tunnel_info *ti);
1076  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1077  *	address family that a UDP tunnel is listnening to. It is called only
1078  *	when a new port starts listening. The operation is protected by the
1079  *	RTNL.
1080  *
1081  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1082  *			      struct udp_tunnel_info *ti);
1083  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1084  *	address family that the UDP tunnel is not listening to anymore. The
1085  *	operation is protected by the RTNL.
1086  *
1087  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1088  *				 struct net_device *dev)
1089  *	Called by upper layer devices to accelerate switching or other
1090  *	station functionality into hardware. 'pdev is the lowerdev
1091  *	to use for the offload and 'dev' is the net device that will
1092  *	back the offload. Returns a pointer to the private structure
1093  *	the upper layer will maintain.
1094  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1095  *	Called by upper layer device to delete the station created
1096  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1097  *	the station and priv is the structure returned by the add
1098  *	operation.
1099  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1100  *			     int queue_index, u32 maxrate);
1101  *	Called when a user wants to set a max-rate limitation of specific
1102  *	TX queue.
1103  * int (*ndo_get_iflink)(const struct net_device *dev);
1104  *	Called to get the iflink value of this device.
1105  * void (*ndo_change_proto_down)(struct net_device *dev,
1106  *				 bool proto_down);
1107  *	This function is used to pass protocol port error state information
1108  *	to the switch driver. The switch driver can react to the proto_down
1109  *      by doing a phys down on the associated switch port.
1110  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1111  *	This function is used to get egress tunnel information for given skb.
1112  *	This is useful for retrieving outer tunnel header parameters while
1113  *	sampling packet.
1114  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1115  *	This function is used to specify the headroom that the skb must
1116  *	consider when allocation skb during packet reception. Setting
1117  *	appropriate rx headroom value allows avoiding skb head copy on
1118  *	forward. Setting a negative value resets the rx headroom to the
1119  *	default value.
1120  * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1121  *	This function is used to set or query state related to XDP on the
1122  *	netdevice. See definition of enum xdp_netdev_command for details.
1123  * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1124  *	This function is used to submit a XDP packet for transmit on a
1125  *	netdevice.
1126  * void (*ndo_xdp_flush)(struct net_device *dev);
1127  *	This function is used to inform the driver to flush a particular
1128  *	xdp tx queue. Must be called on same CPU as xdp_xmit.
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 						enum tc_setup_type type,
1212 						void *type_data);
1213 #if IS_ENABLED(CONFIG_FCOE)
1214 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1215 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1216 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1217 						      u16 xid,
1218 						      struct scatterlist *sgl,
1219 						      unsigned int sgc);
1220 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1221 						     u16 xid);
1222 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1223 						       u16 xid,
1224 						       struct scatterlist *sgl,
1225 						       unsigned int sgc);
1226 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1227 							struct netdev_fcoe_hbainfo *hbainfo);
1228 #endif
1229 
1230 #if IS_ENABLED(CONFIG_LIBFCOE)
1231 #define NETDEV_FCOE_WWNN 0
1232 #define NETDEV_FCOE_WWPN 1
1233 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1234 						    u64 *wwn, int type);
1235 #endif
1236 
1237 #ifdef CONFIG_RFS_ACCEL
1238 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1239 						     const struct sk_buff *skb,
1240 						     u16 rxq_index,
1241 						     u32 flow_id);
1242 #endif
1243 	int			(*ndo_add_slave)(struct net_device *dev,
1244 						 struct net_device *slave_dev);
1245 	int			(*ndo_del_slave)(struct net_device *dev,
1246 						 struct net_device *slave_dev);
1247 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1248 						    netdev_features_t features);
1249 	int			(*ndo_set_features)(struct net_device *dev,
1250 						    netdev_features_t features);
1251 	int			(*ndo_neigh_construct)(struct net_device *dev,
1252 						       struct neighbour *n);
1253 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1254 						     struct neighbour *n);
1255 
1256 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1257 					       struct nlattr *tb[],
1258 					       struct net_device *dev,
1259 					       const unsigned char *addr,
1260 					       u16 vid,
1261 					       u16 flags);
1262 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1263 					       struct nlattr *tb[],
1264 					       struct net_device *dev,
1265 					       const unsigned char *addr,
1266 					       u16 vid);
1267 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1268 						struct netlink_callback *cb,
1269 						struct net_device *dev,
1270 						struct net_device *filter_dev,
1271 						int *idx);
1272 
1273 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1274 						      struct nlmsghdr *nlh,
1275 						      u16 flags);
1276 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1277 						      u32 pid, u32 seq,
1278 						      struct net_device *dev,
1279 						      u32 filter_mask,
1280 						      int nlflags);
1281 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1282 						      struct nlmsghdr *nlh,
1283 						      u16 flags);
1284 	int			(*ndo_change_carrier)(struct net_device *dev,
1285 						      bool new_carrier);
1286 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1287 							struct netdev_phys_item_id *ppid);
1288 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1289 							  char *name, size_t len);
1290 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1291 						      struct udp_tunnel_info *ti);
1292 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1293 						      struct udp_tunnel_info *ti);
1294 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1295 							struct net_device *dev);
1296 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1297 							void *priv);
1298 
1299 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1300 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1301 						      int queue_index,
1302 						      u32 maxrate);
1303 	int			(*ndo_get_iflink)(const struct net_device *dev);
1304 	int			(*ndo_change_proto_down)(struct net_device *dev,
1305 							 bool proto_down);
1306 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1307 						       struct sk_buff *skb);
1308 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1309 						       int needed_headroom);
1310 	int			(*ndo_xdp)(struct net_device *dev,
1311 					   struct netdev_xdp *xdp);
1312 	int			(*ndo_xdp_xmit)(struct net_device *dev,
1313 						struct xdp_buff *xdp);
1314 	void			(*ndo_xdp_flush)(struct net_device *dev);
1315 };
1316 
1317 /**
1318  * enum net_device_priv_flags - &struct net_device priv_flags
1319  *
1320  * These are the &struct net_device, they are only set internally
1321  * by drivers and used in the kernel. These flags are invisible to
1322  * userspace; this means that the order of these flags can change
1323  * during any kernel release.
1324  *
1325  * You should have a pretty good reason to be extending these flags.
1326  *
1327  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1328  * @IFF_EBRIDGE: Ethernet bridging device
1329  * @IFF_BONDING: bonding master or slave
1330  * @IFF_ISATAP: ISATAP interface (RFC4214)
1331  * @IFF_WAN_HDLC: WAN HDLC device
1332  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1333  *	release skb->dst
1334  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1335  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1336  * @IFF_MACVLAN_PORT: device used as macvlan port
1337  * @IFF_BRIDGE_PORT: device used as bridge port
1338  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1339  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1340  * @IFF_UNICAST_FLT: Supports unicast filtering
1341  * @IFF_TEAM_PORT: device used as team port
1342  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1343  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1344  *	change when it's running
1345  * @IFF_MACVLAN: Macvlan device
1346  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1347  *	underlying stacked devices
1348  * @IFF_IPVLAN_MASTER: IPvlan master device
1349  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1350  * @IFF_L3MDEV_MASTER: device is an L3 master device
1351  * @IFF_NO_QUEUE: device can run without qdisc attached
1352  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1353  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1354  * @IFF_TEAM: device is a team device
1355  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1356  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1357  *	entity (i.e. the master device for bridged veth)
1358  * @IFF_MACSEC: device is a MACsec device
1359  */
1360 enum netdev_priv_flags {
1361 	IFF_802_1Q_VLAN			= 1<<0,
1362 	IFF_EBRIDGE			= 1<<1,
1363 	IFF_BONDING			= 1<<2,
1364 	IFF_ISATAP			= 1<<3,
1365 	IFF_WAN_HDLC			= 1<<4,
1366 	IFF_XMIT_DST_RELEASE		= 1<<5,
1367 	IFF_DONT_BRIDGE			= 1<<6,
1368 	IFF_DISABLE_NETPOLL		= 1<<7,
1369 	IFF_MACVLAN_PORT		= 1<<8,
1370 	IFF_BRIDGE_PORT			= 1<<9,
1371 	IFF_OVS_DATAPATH		= 1<<10,
1372 	IFF_TX_SKB_SHARING		= 1<<11,
1373 	IFF_UNICAST_FLT			= 1<<12,
1374 	IFF_TEAM_PORT			= 1<<13,
1375 	IFF_SUPP_NOFCS			= 1<<14,
1376 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1377 	IFF_MACVLAN			= 1<<16,
1378 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1379 	IFF_IPVLAN_MASTER		= 1<<18,
1380 	IFF_IPVLAN_SLAVE		= 1<<19,
1381 	IFF_L3MDEV_MASTER		= 1<<20,
1382 	IFF_NO_QUEUE			= 1<<21,
1383 	IFF_OPENVSWITCH			= 1<<22,
1384 	IFF_L3MDEV_SLAVE		= 1<<23,
1385 	IFF_TEAM			= 1<<24,
1386 	IFF_RXFH_CONFIGURED		= 1<<25,
1387 	IFF_PHONY_HEADROOM		= 1<<26,
1388 	IFF_MACSEC			= 1<<27,
1389 };
1390 
1391 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1392 #define IFF_EBRIDGE			IFF_EBRIDGE
1393 #define IFF_BONDING			IFF_BONDING
1394 #define IFF_ISATAP			IFF_ISATAP
1395 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1396 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1397 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1398 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1399 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1400 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1401 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1402 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1403 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1404 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1405 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1406 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1407 #define IFF_MACVLAN			IFF_MACVLAN
1408 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1409 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1410 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1411 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1412 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1413 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1414 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1415 #define IFF_TEAM			IFF_TEAM
1416 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1417 #define IFF_MACSEC			IFF_MACSEC
1418 
1419 /**
1420  *	struct net_device - The DEVICE structure.
1421  *
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  *	@needs_free_netdev:	Should unregister perform free_netdev?
1586  *	@priv_destructor:	Called from unregister
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 #ifdef CONFIG_XFRM
1701 	const struct xfrmdev_ops *xfrmdev_ops;
1702 #endif
1703 
1704 	const struct header_ops *header_ops;
1705 
1706 	unsigned int		flags;
1707 	unsigned int		priv_flags;
1708 
1709 	unsigned short		gflags;
1710 	unsigned short		padded;
1711 
1712 	unsigned char		operstate;
1713 	unsigned char		link_mode;
1714 
1715 	unsigned char		if_port;
1716 	unsigned char		dma;
1717 
1718 	unsigned int		mtu;
1719 	unsigned int		min_mtu;
1720 	unsigned int		max_mtu;
1721 	unsigned short		type;
1722 	unsigned short		hard_header_len;
1723 	unsigned char		min_header_len;
1724 
1725 	unsigned short		needed_headroom;
1726 	unsigned short		needed_tailroom;
1727 
1728 	/* Interface address info. */
1729 	unsigned char		perm_addr[MAX_ADDR_LEN];
1730 	unsigned char		addr_assign_type;
1731 	unsigned char		addr_len;
1732 	unsigned short		neigh_priv_len;
1733 	unsigned short          dev_id;
1734 	unsigned short          dev_port;
1735 	spinlock_t		addr_list_lock;
1736 	unsigned char		name_assign_type;
1737 	bool			uc_promisc;
1738 	struct netdev_hw_addr_list	uc;
1739 	struct netdev_hw_addr_list	mc;
1740 	struct netdev_hw_addr_list	dev_addrs;
1741 
1742 #ifdef CONFIG_SYSFS
1743 	struct kset		*queues_kset;
1744 #endif
1745 	unsigned int		promiscuity;
1746 	unsigned int		allmulti;
1747 
1748 
1749 	/* Protocol-specific pointers */
1750 
1751 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1752 	struct vlan_info __rcu	*vlan_info;
1753 #endif
1754 #if IS_ENABLED(CONFIG_NET_DSA)
1755 	struct dsa_switch_tree	*dsa_ptr;
1756 #endif
1757 #if IS_ENABLED(CONFIG_TIPC)
1758 	struct tipc_bearer __rcu *tipc_ptr;
1759 #endif
1760 	void 			*atalk_ptr;
1761 	struct in_device __rcu	*ip_ptr;
1762 	struct dn_dev __rcu     *dn_ptr;
1763 	struct inet6_dev __rcu	*ip6_ptr;
1764 	void			*ax25_ptr;
1765 	struct wireless_dev	*ieee80211_ptr;
1766 	struct wpan_dev		*ieee802154_ptr;
1767 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1768 	struct mpls_dev __rcu	*mpls_ptr;
1769 #endif
1770 
1771 /*
1772  * Cache lines mostly used on receive path (including eth_type_trans())
1773  */
1774 	/* Interface address info used in eth_type_trans() */
1775 	unsigned char		*dev_addr;
1776 
1777 #ifdef CONFIG_SYSFS
1778 	struct netdev_rx_queue	*_rx;
1779 
1780 	unsigned int		num_rx_queues;
1781 	unsigned int		real_num_rx_queues;
1782 #endif
1783 
1784 	struct bpf_prog __rcu	*xdp_prog;
1785 	unsigned long		gro_flush_timeout;
1786 	rx_handler_func_t __rcu	*rx_handler;
1787 	void __rcu		*rx_handler_data;
1788 
1789 #ifdef CONFIG_NET_CLS_ACT
1790 	struct tcf_proto __rcu  *ingress_cl_list;
1791 #endif
1792 	struct netdev_queue __rcu *ingress_queue;
1793 #ifdef CONFIG_NETFILTER_INGRESS
1794 	struct nf_hook_entries __rcu *nf_hooks_ingress;
1795 #endif
1796 
1797 	unsigned char		broadcast[MAX_ADDR_LEN];
1798 #ifdef CONFIG_RFS_ACCEL
1799 	struct cpu_rmap		*rx_cpu_rmap;
1800 #endif
1801 	struct hlist_node	index_hlist;
1802 
1803 /*
1804  * Cache lines mostly used on transmit path
1805  */
1806 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1807 	unsigned int		num_tx_queues;
1808 	unsigned int		real_num_tx_queues;
1809 	struct Qdisc		*qdisc;
1810 #ifdef CONFIG_NET_SCHED
1811 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1812 #endif
1813 	unsigned int		tx_queue_len;
1814 	spinlock_t		tx_global_lock;
1815 	int			watchdog_timeo;
1816 
1817 #ifdef CONFIG_XPS
1818 	struct xps_dev_maps __rcu *xps_maps;
1819 #endif
1820 #ifdef CONFIG_NET_CLS_ACT
1821 	struct tcf_proto __rcu  *egress_cl_list;
1822 #endif
1823 
1824 	/* These may be needed for future network-power-down code. */
1825 	struct timer_list	watchdog_timer;
1826 
1827 	int __percpu		*pcpu_refcnt;
1828 	struct list_head	todo_list;
1829 
1830 	struct list_head	link_watch_list;
1831 
1832 	enum { NETREG_UNINITIALIZED=0,
1833 	       NETREG_REGISTERED,	/* completed register_netdevice */
1834 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1835 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1836 	       NETREG_RELEASED,		/* called free_netdev */
1837 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1838 	} reg_state:8;
1839 
1840 	bool dismantle;
1841 
1842 	enum {
1843 		RTNL_LINK_INITIALIZED,
1844 		RTNL_LINK_INITIALIZING,
1845 	} rtnl_link_state:16;
1846 
1847 	bool needs_free_netdev;
1848 	void (*priv_destructor)(struct net_device *dev);
1849 
1850 #ifdef CONFIG_NETPOLL
1851 	struct netpoll_info __rcu	*npinfo;
1852 #endif
1853 
1854 	possible_net_t			nd_net;
1855 
1856 	/* mid-layer private */
1857 	union {
1858 		void					*ml_priv;
1859 		struct pcpu_lstats __percpu		*lstats;
1860 		struct pcpu_sw_netstats __percpu	*tstats;
1861 		struct pcpu_dstats __percpu		*dstats;
1862 		struct pcpu_vstats __percpu		*vstats;
1863 	};
1864 
1865 #if IS_ENABLED(CONFIG_GARP)
1866 	struct garp_port __rcu	*garp_port;
1867 #endif
1868 #if IS_ENABLED(CONFIG_MRP)
1869 	struct mrp_port __rcu	*mrp_port;
1870 #endif
1871 
1872 	struct device		dev;
1873 	const struct attribute_group *sysfs_groups[4];
1874 	const struct attribute_group *sysfs_rx_queue_group;
1875 
1876 	const struct rtnl_link_ops *rtnl_link_ops;
1877 
1878 	/* for setting kernel sock attribute on TCP connection setup */
1879 #define GSO_MAX_SIZE		65536
1880 	unsigned int		gso_max_size;
1881 #define GSO_MAX_SEGS		65535
1882 	u16			gso_max_segs;
1883 
1884 #ifdef CONFIG_DCB
1885 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1886 #endif
1887 	u8			num_tc;
1888 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1889 	u8			prio_tc_map[TC_BITMASK + 1];
1890 
1891 #if IS_ENABLED(CONFIG_FCOE)
1892 	unsigned int		fcoe_ddp_xid;
1893 #endif
1894 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1895 	struct netprio_map __rcu *priomap;
1896 #endif
1897 	struct phy_device	*phydev;
1898 	struct lock_class_key	*qdisc_tx_busylock;
1899 	struct lock_class_key	*qdisc_running_key;
1900 	bool			proto_down;
1901 };
1902 #define to_net_dev(d) container_of(d, struct net_device, dev)
1903 
1904 static inline bool netif_elide_gro(const struct net_device *dev)
1905 {
1906 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1907 		return true;
1908 	return false;
1909 }
1910 
1911 #define	NETDEV_ALIGN		32
1912 
1913 static inline
1914 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1915 {
1916 	return dev->prio_tc_map[prio & TC_BITMASK];
1917 }
1918 
1919 static inline
1920 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1921 {
1922 	if (tc >= dev->num_tc)
1923 		return -EINVAL;
1924 
1925 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1926 	return 0;
1927 }
1928 
1929 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1930 void netdev_reset_tc(struct net_device *dev);
1931 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1932 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1933 
1934 static inline
1935 int netdev_get_num_tc(struct net_device *dev)
1936 {
1937 	return dev->num_tc;
1938 }
1939 
1940 static inline
1941 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1942 					 unsigned int index)
1943 {
1944 	return &dev->_tx[index];
1945 }
1946 
1947 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1948 						    const struct sk_buff *skb)
1949 {
1950 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1951 }
1952 
1953 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1954 					    void (*f)(struct net_device *,
1955 						      struct netdev_queue *,
1956 						      void *),
1957 					    void *arg)
1958 {
1959 	unsigned int i;
1960 
1961 	for (i = 0; i < dev->num_tx_queues; i++)
1962 		f(dev, &dev->_tx[i], arg);
1963 }
1964 
1965 #define netdev_lockdep_set_classes(dev)				\
1966 {								\
1967 	static struct lock_class_key qdisc_tx_busylock_key;	\
1968 	static struct lock_class_key qdisc_running_key;		\
1969 	static struct lock_class_key qdisc_xmit_lock_key;	\
1970 	static struct lock_class_key dev_addr_list_lock_key;	\
1971 	unsigned int i;						\
1972 								\
1973 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
1974 	(dev)->qdisc_running_key = &qdisc_running_key;		\
1975 	lockdep_set_class(&(dev)->addr_list_lock,		\
1976 			  &dev_addr_list_lock_key); 		\
1977 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
1978 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
1979 				  &qdisc_xmit_lock_key);	\
1980 }
1981 
1982 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1983 				    struct sk_buff *skb,
1984 				    void *accel_priv);
1985 
1986 /* returns the headroom that the master device needs to take in account
1987  * when forwarding to this dev
1988  */
1989 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1990 {
1991 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1992 }
1993 
1994 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1995 {
1996 	if (dev->netdev_ops->ndo_set_rx_headroom)
1997 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1998 }
1999 
2000 /* set the device rx headroom to the dev's default */
2001 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2002 {
2003 	netdev_set_rx_headroom(dev, -1);
2004 }
2005 
2006 /*
2007  * Net namespace inlines
2008  */
2009 static inline
2010 struct net *dev_net(const struct net_device *dev)
2011 {
2012 	return read_pnet(&dev->nd_net);
2013 }
2014 
2015 static inline
2016 void dev_net_set(struct net_device *dev, struct net *net)
2017 {
2018 	write_pnet(&dev->nd_net, net);
2019 }
2020 
2021 /**
2022  *	netdev_priv - access network device private data
2023  *	@dev: network device
2024  *
2025  * Get network device private data
2026  */
2027 static inline void *netdev_priv(const struct net_device *dev)
2028 {
2029 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2030 }
2031 
2032 /* Set the sysfs physical device reference for the network logical device
2033  * if set prior to registration will cause a symlink during initialization.
2034  */
2035 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2036 
2037 /* Set the sysfs device type for the network logical device to allow
2038  * fine-grained identification of different network device types. For
2039  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2040  */
2041 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2042 
2043 /* Default NAPI poll() weight
2044  * Device drivers are strongly advised to not use bigger value
2045  */
2046 #define NAPI_POLL_WEIGHT 64
2047 
2048 /**
2049  *	netif_napi_add - initialize a NAPI context
2050  *	@dev:  network device
2051  *	@napi: NAPI context
2052  *	@poll: polling function
2053  *	@weight: default weight
2054  *
2055  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2056  * *any* of the other NAPI-related functions.
2057  */
2058 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2059 		    int (*poll)(struct napi_struct *, int), int weight);
2060 
2061 /**
2062  *	netif_tx_napi_add - initialize a NAPI context
2063  *	@dev:  network device
2064  *	@napi: NAPI context
2065  *	@poll: polling function
2066  *	@weight: default weight
2067  *
2068  * This variant of netif_napi_add() should be used from drivers using NAPI
2069  * to exclusively poll a TX queue.
2070  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2071  */
2072 static inline void netif_tx_napi_add(struct net_device *dev,
2073 				     struct napi_struct *napi,
2074 				     int (*poll)(struct napi_struct *, int),
2075 				     int weight)
2076 {
2077 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2078 	netif_napi_add(dev, napi, poll, weight);
2079 }
2080 
2081 /**
2082  *  netif_napi_del - remove a NAPI context
2083  *  @napi: NAPI context
2084  *
2085  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2086  */
2087 void netif_napi_del(struct napi_struct *napi);
2088 
2089 struct napi_gro_cb {
2090 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2091 	void	*frag0;
2092 
2093 	/* Length of frag0. */
2094 	unsigned int frag0_len;
2095 
2096 	/* This indicates where we are processing relative to skb->data. */
2097 	int	data_offset;
2098 
2099 	/* This is non-zero if the packet cannot be merged with the new skb. */
2100 	u16	flush;
2101 
2102 	/* Save the IP ID here and check when we get to the transport layer */
2103 	u16	flush_id;
2104 
2105 	/* Number of segments aggregated. */
2106 	u16	count;
2107 
2108 	/* Start offset for remote checksum offload */
2109 	u16	gro_remcsum_start;
2110 
2111 	/* jiffies when first packet was created/queued */
2112 	unsigned long age;
2113 
2114 	/* Used in ipv6_gro_receive() and foo-over-udp */
2115 	u16	proto;
2116 
2117 	/* This is non-zero if the packet may be of the same flow. */
2118 	u8	same_flow:1;
2119 
2120 	/* Used in tunnel GRO receive */
2121 	u8	encap_mark:1;
2122 
2123 	/* GRO checksum is valid */
2124 	u8	csum_valid:1;
2125 
2126 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2127 	u8	csum_cnt:3;
2128 
2129 	/* Free the skb? */
2130 	u8	free:2;
2131 #define NAPI_GRO_FREE		  1
2132 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2133 
2134 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2135 	u8	is_ipv6:1;
2136 
2137 	/* Used in GRE, set in fou/gue_gro_receive */
2138 	u8	is_fou:1;
2139 
2140 	/* Used to determine if flush_id can be ignored */
2141 	u8	is_atomic:1;
2142 
2143 	/* Number of gro_receive callbacks this packet already went through */
2144 	u8 recursion_counter:4;
2145 
2146 	/* 1 bit hole */
2147 
2148 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2149 	__wsum	csum;
2150 
2151 	/* used in skb_gro_receive() slow path */
2152 	struct sk_buff *last;
2153 };
2154 
2155 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2156 
2157 #define GRO_RECURSION_LIMIT 15
2158 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2159 {
2160 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2161 }
2162 
2163 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2164 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2165 						struct sk_buff **head,
2166 						struct sk_buff *skb)
2167 {
2168 	if (unlikely(gro_recursion_inc_test(skb))) {
2169 		NAPI_GRO_CB(skb)->flush |= 1;
2170 		return NULL;
2171 	}
2172 
2173 	return cb(head, skb);
2174 }
2175 
2176 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2177 					     struct sk_buff *);
2178 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2179 						   struct sock *sk,
2180 						   struct sk_buff **head,
2181 						   struct sk_buff *skb)
2182 {
2183 	if (unlikely(gro_recursion_inc_test(skb))) {
2184 		NAPI_GRO_CB(skb)->flush |= 1;
2185 		return NULL;
2186 	}
2187 
2188 	return cb(sk, head, skb);
2189 }
2190 
2191 struct packet_type {
2192 	__be16			type;	/* This is really htons(ether_type). */
2193 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2194 	int			(*func) (struct sk_buff *,
2195 					 struct net_device *,
2196 					 struct packet_type *,
2197 					 struct net_device *);
2198 	bool			(*id_match)(struct packet_type *ptype,
2199 					    struct sock *sk);
2200 	void			*af_packet_priv;
2201 	struct list_head	list;
2202 };
2203 
2204 struct offload_callbacks {
2205 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2206 						netdev_features_t features);
2207 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2208 						 struct sk_buff *skb);
2209 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2210 };
2211 
2212 struct packet_offload {
2213 	__be16			 type;	/* This is really htons(ether_type). */
2214 	u16			 priority;
2215 	struct offload_callbacks callbacks;
2216 	struct list_head	 list;
2217 };
2218 
2219 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2220 struct pcpu_sw_netstats {
2221 	u64     rx_packets;
2222 	u64     rx_bytes;
2223 	u64     tx_packets;
2224 	u64     tx_bytes;
2225 	struct u64_stats_sync   syncp;
2226 };
2227 
2228 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2229 ({									\
2230 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2231 	if (pcpu_stats)	{						\
2232 		int __cpu;						\
2233 		for_each_possible_cpu(__cpu) {				\
2234 			typeof(type) *stat;				\
2235 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2236 			u64_stats_init(&stat->syncp);			\
2237 		}							\
2238 	}								\
2239 	pcpu_stats;							\
2240 })
2241 
2242 #define netdev_alloc_pcpu_stats(type)					\
2243 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2244 
2245 enum netdev_lag_tx_type {
2246 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2247 	NETDEV_LAG_TX_TYPE_RANDOM,
2248 	NETDEV_LAG_TX_TYPE_BROADCAST,
2249 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2250 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2251 	NETDEV_LAG_TX_TYPE_HASH,
2252 };
2253 
2254 struct netdev_lag_upper_info {
2255 	enum netdev_lag_tx_type tx_type;
2256 };
2257 
2258 struct netdev_lag_lower_state_info {
2259 	u8 link_up : 1,
2260 	   tx_enabled : 1;
2261 };
2262 
2263 #include <linux/notifier.h>
2264 
2265 /* netdevice notifier chain. Please remember to update the rtnetlink
2266  * notification exclusion list in rtnetlink_event() when adding new
2267  * types.
2268  */
2269 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2270 #define NETDEV_DOWN	0x0002
2271 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2272 				   detected a hardware crash and restarted
2273 				   - we can use this eg to kick tcp sessions
2274 				   once done */
2275 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2276 #define NETDEV_REGISTER 0x0005
2277 #define NETDEV_UNREGISTER	0x0006
2278 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2279 #define NETDEV_CHANGEADDR	0x0008
2280 #define NETDEV_GOING_DOWN	0x0009
2281 #define NETDEV_CHANGENAME	0x000A
2282 #define NETDEV_FEAT_CHANGE	0x000B
2283 #define NETDEV_BONDING_FAILOVER 0x000C
2284 #define NETDEV_PRE_UP		0x000D
2285 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2286 #define NETDEV_POST_TYPE_CHANGE	0x000F
2287 #define NETDEV_POST_INIT	0x0010
2288 #define NETDEV_UNREGISTER_FINAL 0x0011
2289 #define NETDEV_RELEASE		0x0012
2290 #define NETDEV_NOTIFY_PEERS	0x0013
2291 #define NETDEV_JOIN		0x0014
2292 #define NETDEV_CHANGEUPPER	0x0015
2293 #define NETDEV_RESEND_IGMP	0x0016
2294 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2295 #define NETDEV_CHANGEINFODATA	0x0018
2296 #define NETDEV_BONDING_INFO	0x0019
2297 #define NETDEV_PRECHANGEUPPER	0x001A
2298 #define NETDEV_CHANGELOWERSTATE	0x001B
2299 #define NETDEV_UDP_TUNNEL_PUSH_INFO	0x001C
2300 #define NETDEV_UDP_TUNNEL_DROP_INFO	0x001D
2301 #define NETDEV_CHANGE_TX_QUEUE_LEN	0x001E
2302 
2303 int register_netdevice_notifier(struct notifier_block *nb);
2304 int unregister_netdevice_notifier(struct notifier_block *nb);
2305 
2306 struct netdev_notifier_info {
2307 	struct net_device *dev;
2308 };
2309 
2310 struct netdev_notifier_change_info {
2311 	struct netdev_notifier_info info; /* must be first */
2312 	unsigned int flags_changed;
2313 };
2314 
2315 struct netdev_notifier_changeupper_info {
2316 	struct netdev_notifier_info info; /* must be first */
2317 	struct net_device *upper_dev; /* new upper dev */
2318 	bool master; /* is upper dev master */
2319 	bool linking; /* is the notification for link or unlink */
2320 	void *upper_info; /* upper dev info */
2321 };
2322 
2323 struct netdev_notifier_changelowerstate_info {
2324 	struct netdev_notifier_info info; /* must be first */
2325 	void *lower_state_info; /* is lower dev state */
2326 };
2327 
2328 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2329 					     struct net_device *dev)
2330 {
2331 	info->dev = dev;
2332 }
2333 
2334 static inline struct net_device *
2335 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2336 {
2337 	return info->dev;
2338 }
2339 
2340 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2341 
2342 
2343 extern rwlock_t				dev_base_lock;		/* Device list lock */
2344 
2345 #define for_each_netdev(net, d)		\
2346 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2347 #define for_each_netdev_reverse(net, d)	\
2348 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2349 #define for_each_netdev_rcu(net, d)		\
2350 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2351 #define for_each_netdev_safe(net, d, n)	\
2352 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2353 #define for_each_netdev_continue(net, d)		\
2354 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2355 #define for_each_netdev_continue_rcu(net, d)		\
2356 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2357 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2358 		for_each_netdev_rcu(&init_net, slave)	\
2359 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2360 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2361 
2362 static inline struct net_device *next_net_device(struct net_device *dev)
2363 {
2364 	struct list_head *lh;
2365 	struct net *net;
2366 
2367 	net = dev_net(dev);
2368 	lh = dev->dev_list.next;
2369 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2370 }
2371 
2372 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2373 {
2374 	struct list_head *lh;
2375 	struct net *net;
2376 
2377 	net = dev_net(dev);
2378 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2379 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2380 }
2381 
2382 static inline struct net_device *first_net_device(struct net *net)
2383 {
2384 	return list_empty(&net->dev_base_head) ? NULL :
2385 		net_device_entry(net->dev_base_head.next);
2386 }
2387 
2388 static inline struct net_device *first_net_device_rcu(struct net *net)
2389 {
2390 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2391 
2392 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2393 }
2394 
2395 int netdev_boot_setup_check(struct net_device *dev);
2396 unsigned long netdev_boot_base(const char *prefix, int unit);
2397 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2398 				       const char *hwaddr);
2399 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2400 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2401 void dev_add_pack(struct packet_type *pt);
2402 void dev_remove_pack(struct packet_type *pt);
2403 void __dev_remove_pack(struct packet_type *pt);
2404 void dev_add_offload(struct packet_offload *po);
2405 void dev_remove_offload(struct packet_offload *po);
2406 
2407 int dev_get_iflink(const struct net_device *dev);
2408 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2409 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2410 				      unsigned short mask);
2411 struct net_device *dev_get_by_name(struct net *net, const char *name);
2412 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2413 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2414 int dev_alloc_name(struct net_device *dev, const char *name);
2415 int dev_open(struct net_device *dev);
2416 void dev_close(struct net_device *dev);
2417 void dev_close_many(struct list_head *head, bool unlink);
2418 void dev_disable_lro(struct net_device *dev);
2419 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2420 int dev_queue_xmit(struct sk_buff *skb);
2421 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2422 int register_netdevice(struct net_device *dev);
2423 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2424 void unregister_netdevice_many(struct list_head *head);
2425 static inline void unregister_netdevice(struct net_device *dev)
2426 {
2427 	unregister_netdevice_queue(dev, NULL);
2428 }
2429 
2430 int netdev_refcnt_read(const struct net_device *dev);
2431 void free_netdev(struct net_device *dev);
2432 void netdev_freemem(struct net_device *dev);
2433 void synchronize_net(void);
2434 int init_dummy_netdev(struct net_device *dev);
2435 
2436 DECLARE_PER_CPU(int, xmit_recursion);
2437 #define XMIT_RECURSION_LIMIT	10
2438 
2439 static inline int dev_recursion_level(void)
2440 {
2441 	return this_cpu_read(xmit_recursion);
2442 }
2443 
2444 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2445 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2446 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2447 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2448 int netdev_get_name(struct net *net, char *name, int ifindex);
2449 int dev_restart(struct net_device *dev);
2450 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2451 
2452 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2453 {
2454 	return NAPI_GRO_CB(skb)->data_offset;
2455 }
2456 
2457 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2458 {
2459 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2460 }
2461 
2462 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2463 {
2464 	NAPI_GRO_CB(skb)->data_offset += len;
2465 }
2466 
2467 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2468 					unsigned int offset)
2469 {
2470 	return NAPI_GRO_CB(skb)->frag0 + offset;
2471 }
2472 
2473 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2474 {
2475 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2476 }
2477 
2478 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2479 {
2480 	NAPI_GRO_CB(skb)->frag0 = NULL;
2481 	NAPI_GRO_CB(skb)->frag0_len = 0;
2482 }
2483 
2484 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2485 					unsigned int offset)
2486 {
2487 	if (!pskb_may_pull(skb, hlen))
2488 		return NULL;
2489 
2490 	skb_gro_frag0_invalidate(skb);
2491 	return skb->data + offset;
2492 }
2493 
2494 static inline void *skb_gro_network_header(struct sk_buff *skb)
2495 {
2496 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2497 	       skb_network_offset(skb);
2498 }
2499 
2500 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2501 					const void *start, unsigned int len)
2502 {
2503 	if (NAPI_GRO_CB(skb)->csum_valid)
2504 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2505 						  csum_partial(start, len, 0));
2506 }
2507 
2508 /* GRO checksum functions. These are logical equivalents of the normal
2509  * checksum functions (in skbuff.h) except that they operate on the GRO
2510  * offsets and fields in sk_buff.
2511  */
2512 
2513 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2514 
2515 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2516 {
2517 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2518 }
2519 
2520 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2521 						      bool zero_okay,
2522 						      __sum16 check)
2523 {
2524 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2525 		skb_checksum_start_offset(skb) <
2526 		 skb_gro_offset(skb)) &&
2527 		!skb_at_gro_remcsum_start(skb) &&
2528 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2529 		(!zero_okay || check));
2530 }
2531 
2532 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2533 							   __wsum psum)
2534 {
2535 	if (NAPI_GRO_CB(skb)->csum_valid &&
2536 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2537 		return 0;
2538 
2539 	NAPI_GRO_CB(skb)->csum = psum;
2540 
2541 	return __skb_gro_checksum_complete(skb);
2542 }
2543 
2544 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2545 {
2546 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2547 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2548 		NAPI_GRO_CB(skb)->csum_cnt--;
2549 	} else {
2550 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2551 		 * verified a new top level checksum or an encapsulated one
2552 		 * during GRO. This saves work if we fallback to normal path.
2553 		 */
2554 		__skb_incr_checksum_unnecessary(skb);
2555 	}
2556 }
2557 
2558 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2559 				    compute_pseudo)			\
2560 ({									\
2561 	__sum16 __ret = 0;						\
2562 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2563 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2564 				compute_pseudo(skb, proto));		\
2565 	if (!__ret)							\
2566 		skb_gro_incr_csum_unnecessary(skb);			\
2567 	__ret;								\
2568 })
2569 
2570 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2571 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2572 
2573 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2574 					     compute_pseudo)		\
2575 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2576 
2577 #define skb_gro_checksum_simple_validate(skb)				\
2578 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2579 
2580 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2581 {
2582 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2583 		!NAPI_GRO_CB(skb)->csum_valid);
2584 }
2585 
2586 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2587 					      __sum16 check, __wsum pseudo)
2588 {
2589 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2590 	NAPI_GRO_CB(skb)->csum_valid = 1;
2591 }
2592 
2593 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2594 do {									\
2595 	if (__skb_gro_checksum_convert_check(skb))			\
2596 		__skb_gro_checksum_convert(skb, check,			\
2597 					   compute_pseudo(skb, proto));	\
2598 } while (0)
2599 
2600 struct gro_remcsum {
2601 	int offset;
2602 	__wsum delta;
2603 };
2604 
2605 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2606 {
2607 	grc->offset = 0;
2608 	grc->delta = 0;
2609 }
2610 
2611 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2612 					    unsigned int off, size_t hdrlen,
2613 					    int start, int offset,
2614 					    struct gro_remcsum *grc,
2615 					    bool nopartial)
2616 {
2617 	__wsum delta;
2618 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2619 
2620 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2621 
2622 	if (!nopartial) {
2623 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2624 		return ptr;
2625 	}
2626 
2627 	ptr = skb_gro_header_fast(skb, off);
2628 	if (skb_gro_header_hard(skb, off + plen)) {
2629 		ptr = skb_gro_header_slow(skb, off + plen, off);
2630 		if (!ptr)
2631 			return NULL;
2632 	}
2633 
2634 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2635 			       start, offset);
2636 
2637 	/* Adjust skb->csum since we changed the packet */
2638 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2639 
2640 	grc->offset = off + hdrlen + offset;
2641 	grc->delta = delta;
2642 
2643 	return ptr;
2644 }
2645 
2646 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2647 					   struct gro_remcsum *grc)
2648 {
2649 	void *ptr;
2650 	size_t plen = grc->offset + sizeof(u16);
2651 
2652 	if (!grc->delta)
2653 		return;
2654 
2655 	ptr = skb_gro_header_fast(skb, grc->offset);
2656 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2657 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2658 		if (!ptr)
2659 			return;
2660 	}
2661 
2662 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2663 }
2664 
2665 #ifdef CONFIG_XFRM_OFFLOAD
2666 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2667 {
2668 	if (PTR_ERR(pp) != -EINPROGRESS)
2669 		NAPI_GRO_CB(skb)->flush |= flush;
2670 }
2671 #else
2672 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2673 {
2674 	NAPI_GRO_CB(skb)->flush |= flush;
2675 }
2676 #endif
2677 
2678 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2679 				  unsigned short type,
2680 				  const void *daddr, const void *saddr,
2681 				  unsigned int len)
2682 {
2683 	if (!dev->header_ops || !dev->header_ops->create)
2684 		return 0;
2685 
2686 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2687 }
2688 
2689 static inline int dev_parse_header(const struct sk_buff *skb,
2690 				   unsigned char *haddr)
2691 {
2692 	const struct net_device *dev = skb->dev;
2693 
2694 	if (!dev->header_ops || !dev->header_ops->parse)
2695 		return 0;
2696 	return dev->header_ops->parse(skb, haddr);
2697 }
2698 
2699 /* ll_header must have at least hard_header_len allocated */
2700 static inline bool dev_validate_header(const struct net_device *dev,
2701 				       char *ll_header, int len)
2702 {
2703 	if (likely(len >= dev->hard_header_len))
2704 		return true;
2705 	if (len < dev->min_header_len)
2706 		return false;
2707 
2708 	if (capable(CAP_SYS_RAWIO)) {
2709 		memset(ll_header + len, 0, dev->hard_header_len - len);
2710 		return true;
2711 	}
2712 
2713 	if (dev->header_ops && dev->header_ops->validate)
2714 		return dev->header_ops->validate(ll_header, len);
2715 
2716 	return false;
2717 }
2718 
2719 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2720 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2721 static inline int unregister_gifconf(unsigned int family)
2722 {
2723 	return register_gifconf(family, NULL);
2724 }
2725 
2726 #ifdef CONFIG_NET_FLOW_LIMIT
2727 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2728 struct sd_flow_limit {
2729 	u64			count;
2730 	unsigned int		num_buckets;
2731 	unsigned int		history_head;
2732 	u16			history[FLOW_LIMIT_HISTORY];
2733 	u8			buckets[];
2734 };
2735 
2736 extern int netdev_flow_limit_table_len;
2737 #endif /* CONFIG_NET_FLOW_LIMIT */
2738 
2739 /*
2740  * Incoming packets are placed on per-CPU queues
2741  */
2742 struct softnet_data {
2743 	struct list_head	poll_list;
2744 	struct sk_buff_head	process_queue;
2745 
2746 	/* stats */
2747 	unsigned int		processed;
2748 	unsigned int		time_squeeze;
2749 	unsigned int		received_rps;
2750 #ifdef CONFIG_RPS
2751 	struct softnet_data	*rps_ipi_list;
2752 #endif
2753 #ifdef CONFIG_NET_FLOW_LIMIT
2754 	struct sd_flow_limit __rcu *flow_limit;
2755 #endif
2756 	struct Qdisc		*output_queue;
2757 	struct Qdisc		**output_queue_tailp;
2758 	struct sk_buff		*completion_queue;
2759 
2760 #ifdef CONFIG_RPS
2761 	/* input_queue_head should be written by cpu owning this struct,
2762 	 * and only read by other cpus. Worth using a cache line.
2763 	 */
2764 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2765 
2766 	/* Elements below can be accessed between CPUs for RPS/RFS */
2767 	call_single_data_t	csd ____cacheline_aligned_in_smp;
2768 	struct softnet_data	*rps_ipi_next;
2769 	unsigned int		cpu;
2770 	unsigned int		input_queue_tail;
2771 #endif
2772 	unsigned int		dropped;
2773 	struct sk_buff_head	input_pkt_queue;
2774 	struct napi_struct	backlog;
2775 
2776 };
2777 
2778 static inline void input_queue_head_incr(struct softnet_data *sd)
2779 {
2780 #ifdef CONFIG_RPS
2781 	sd->input_queue_head++;
2782 #endif
2783 }
2784 
2785 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2786 					      unsigned int *qtail)
2787 {
2788 #ifdef CONFIG_RPS
2789 	*qtail = ++sd->input_queue_tail;
2790 #endif
2791 }
2792 
2793 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2794 
2795 void __netif_schedule(struct Qdisc *q);
2796 void netif_schedule_queue(struct netdev_queue *txq);
2797 
2798 static inline void netif_tx_schedule_all(struct net_device *dev)
2799 {
2800 	unsigned int i;
2801 
2802 	for (i = 0; i < dev->num_tx_queues; i++)
2803 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2804 }
2805 
2806 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2807 {
2808 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2809 }
2810 
2811 /**
2812  *	netif_start_queue - allow transmit
2813  *	@dev: network device
2814  *
2815  *	Allow upper layers to call the device hard_start_xmit routine.
2816  */
2817 static inline void netif_start_queue(struct net_device *dev)
2818 {
2819 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2820 }
2821 
2822 static inline void netif_tx_start_all_queues(struct net_device *dev)
2823 {
2824 	unsigned int i;
2825 
2826 	for (i = 0; i < dev->num_tx_queues; i++) {
2827 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2828 		netif_tx_start_queue(txq);
2829 	}
2830 }
2831 
2832 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2833 
2834 /**
2835  *	netif_wake_queue - restart transmit
2836  *	@dev: network device
2837  *
2838  *	Allow upper layers to call the device hard_start_xmit routine.
2839  *	Used for flow control when transmit resources are available.
2840  */
2841 static inline void netif_wake_queue(struct net_device *dev)
2842 {
2843 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2844 }
2845 
2846 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2847 {
2848 	unsigned int i;
2849 
2850 	for (i = 0; i < dev->num_tx_queues; i++) {
2851 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2852 		netif_tx_wake_queue(txq);
2853 	}
2854 }
2855 
2856 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2857 {
2858 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2859 }
2860 
2861 /**
2862  *	netif_stop_queue - stop transmitted packets
2863  *	@dev: network device
2864  *
2865  *	Stop upper layers calling the device hard_start_xmit routine.
2866  *	Used for flow control when transmit resources are unavailable.
2867  */
2868 static inline void netif_stop_queue(struct net_device *dev)
2869 {
2870 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2871 }
2872 
2873 void netif_tx_stop_all_queues(struct net_device *dev);
2874 
2875 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2876 {
2877 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2878 }
2879 
2880 /**
2881  *	netif_queue_stopped - test if transmit queue is flowblocked
2882  *	@dev: network device
2883  *
2884  *	Test if transmit queue on device is currently unable to send.
2885  */
2886 static inline bool netif_queue_stopped(const struct net_device *dev)
2887 {
2888 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2889 }
2890 
2891 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2892 {
2893 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2894 }
2895 
2896 static inline bool
2897 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2898 {
2899 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2900 }
2901 
2902 static inline bool
2903 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2904 {
2905 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2906 }
2907 
2908 /**
2909  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2910  *	@dev_queue: pointer to transmit queue
2911  *
2912  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2913  * to give appropriate hint to the CPU.
2914  */
2915 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2916 {
2917 #ifdef CONFIG_BQL
2918 	prefetchw(&dev_queue->dql.num_queued);
2919 #endif
2920 }
2921 
2922 /**
2923  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2924  *	@dev_queue: pointer to transmit queue
2925  *
2926  * BQL enabled drivers might use this helper in their TX completion path,
2927  * to give appropriate hint to the CPU.
2928  */
2929 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2930 {
2931 #ifdef CONFIG_BQL
2932 	prefetchw(&dev_queue->dql.limit);
2933 #endif
2934 }
2935 
2936 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2937 					unsigned int bytes)
2938 {
2939 #ifdef CONFIG_BQL
2940 	dql_queued(&dev_queue->dql, bytes);
2941 
2942 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2943 		return;
2944 
2945 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2946 
2947 	/*
2948 	 * The XOFF flag must be set before checking the dql_avail below,
2949 	 * because in netdev_tx_completed_queue we update the dql_completed
2950 	 * before checking the XOFF flag.
2951 	 */
2952 	smp_mb();
2953 
2954 	/* check again in case another CPU has just made room avail */
2955 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2956 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2957 #endif
2958 }
2959 
2960 /**
2961  * 	netdev_sent_queue - report the number of bytes queued to hardware
2962  * 	@dev: network device
2963  * 	@bytes: number of bytes queued to the hardware device queue
2964  *
2965  * 	Report the number of bytes queued for sending/completion to the network
2966  * 	device hardware queue. @bytes should be a good approximation and should
2967  * 	exactly match netdev_completed_queue() @bytes
2968  */
2969 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2970 {
2971 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2972 }
2973 
2974 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2975 					     unsigned int pkts, unsigned int bytes)
2976 {
2977 #ifdef CONFIG_BQL
2978 	if (unlikely(!bytes))
2979 		return;
2980 
2981 	dql_completed(&dev_queue->dql, bytes);
2982 
2983 	/*
2984 	 * Without the memory barrier there is a small possiblity that
2985 	 * netdev_tx_sent_queue will miss the update and cause the queue to
2986 	 * be stopped forever
2987 	 */
2988 	smp_mb();
2989 
2990 	if (dql_avail(&dev_queue->dql) < 0)
2991 		return;
2992 
2993 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2994 		netif_schedule_queue(dev_queue);
2995 #endif
2996 }
2997 
2998 /**
2999  * 	netdev_completed_queue - report bytes and packets completed by device
3000  * 	@dev: network device
3001  * 	@pkts: actual number of packets sent over the medium
3002  * 	@bytes: actual number of bytes sent over the medium
3003  *
3004  * 	Report the number of bytes and packets transmitted by the network device
3005  * 	hardware queue over the physical medium, @bytes must exactly match the
3006  * 	@bytes amount passed to netdev_sent_queue()
3007  */
3008 static inline void netdev_completed_queue(struct net_device *dev,
3009 					  unsigned int pkts, unsigned int bytes)
3010 {
3011 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3012 }
3013 
3014 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3015 {
3016 #ifdef CONFIG_BQL
3017 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3018 	dql_reset(&q->dql);
3019 #endif
3020 }
3021 
3022 /**
3023  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3024  * 	@dev_queue: network device
3025  *
3026  * 	Reset the bytes and packet count of a network device and clear the
3027  * 	software flow control OFF bit for this network device
3028  */
3029 static inline void netdev_reset_queue(struct net_device *dev_queue)
3030 {
3031 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3032 }
3033 
3034 /**
3035  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3036  * 	@dev: network device
3037  * 	@queue_index: given tx queue index
3038  *
3039  * 	Returns 0 if given tx queue index >= number of device tx queues,
3040  * 	otherwise returns the originally passed tx queue index.
3041  */
3042 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3043 {
3044 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3045 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3046 				     dev->name, queue_index,
3047 				     dev->real_num_tx_queues);
3048 		return 0;
3049 	}
3050 
3051 	return queue_index;
3052 }
3053 
3054 /**
3055  *	netif_running - test if up
3056  *	@dev: network device
3057  *
3058  *	Test if the device has been brought up.
3059  */
3060 static inline bool netif_running(const struct net_device *dev)
3061 {
3062 	return test_bit(__LINK_STATE_START, &dev->state);
3063 }
3064 
3065 /*
3066  * Routines to manage the subqueues on a device.  We only need start,
3067  * stop, and a check if it's stopped.  All other device management is
3068  * done at the overall netdevice level.
3069  * Also test the device if we're multiqueue.
3070  */
3071 
3072 /**
3073  *	netif_start_subqueue - allow sending packets on subqueue
3074  *	@dev: network device
3075  *	@queue_index: sub queue index
3076  *
3077  * Start individual transmit queue of a device with multiple transmit queues.
3078  */
3079 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3080 {
3081 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3082 
3083 	netif_tx_start_queue(txq);
3084 }
3085 
3086 /**
3087  *	netif_stop_subqueue - stop sending packets on subqueue
3088  *	@dev: network device
3089  *	@queue_index: sub queue index
3090  *
3091  * Stop individual transmit queue of a device with multiple transmit queues.
3092  */
3093 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3094 {
3095 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3096 	netif_tx_stop_queue(txq);
3097 }
3098 
3099 /**
3100  *	netif_subqueue_stopped - test status of subqueue
3101  *	@dev: network device
3102  *	@queue_index: sub queue index
3103  *
3104  * Check individual transmit queue of a device with multiple transmit queues.
3105  */
3106 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3107 					    u16 queue_index)
3108 {
3109 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3110 
3111 	return netif_tx_queue_stopped(txq);
3112 }
3113 
3114 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3115 					  struct sk_buff *skb)
3116 {
3117 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3118 }
3119 
3120 /**
3121  *	netif_wake_subqueue - allow sending packets on subqueue
3122  *	@dev: network device
3123  *	@queue_index: sub queue index
3124  *
3125  * Resume individual transmit queue of a device with multiple transmit queues.
3126  */
3127 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3128 {
3129 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3130 
3131 	netif_tx_wake_queue(txq);
3132 }
3133 
3134 #ifdef CONFIG_XPS
3135 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3136 			u16 index);
3137 #else
3138 static inline int netif_set_xps_queue(struct net_device *dev,
3139 				      const struct cpumask *mask,
3140 				      u16 index)
3141 {
3142 	return 0;
3143 }
3144 #endif
3145 
3146 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3147 		  unsigned int num_tx_queues);
3148 
3149 /*
3150  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3151  * as a distribution range limit for the returned value.
3152  */
3153 static inline u16 skb_tx_hash(const struct net_device *dev,
3154 			      struct sk_buff *skb)
3155 {
3156 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3157 }
3158 
3159 /**
3160  *	netif_is_multiqueue - test if device has multiple transmit queues
3161  *	@dev: network device
3162  *
3163  * Check if device has multiple transmit queues
3164  */
3165 static inline bool netif_is_multiqueue(const struct net_device *dev)
3166 {
3167 	return dev->num_tx_queues > 1;
3168 }
3169 
3170 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3171 
3172 #ifdef CONFIG_SYSFS
3173 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3174 #else
3175 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3176 						unsigned int rxq)
3177 {
3178 	return 0;
3179 }
3180 #endif
3181 
3182 #ifdef CONFIG_SYSFS
3183 static inline unsigned int get_netdev_rx_queue_index(
3184 		struct netdev_rx_queue *queue)
3185 {
3186 	struct net_device *dev = queue->dev;
3187 	int index = queue - dev->_rx;
3188 
3189 	BUG_ON(index >= dev->num_rx_queues);
3190 	return index;
3191 }
3192 #endif
3193 
3194 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3195 int netif_get_num_default_rss_queues(void);
3196 
3197 enum skb_free_reason {
3198 	SKB_REASON_CONSUMED,
3199 	SKB_REASON_DROPPED,
3200 };
3201 
3202 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3203 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3204 
3205 /*
3206  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3207  * interrupt context or with hardware interrupts being disabled.
3208  * (in_irq() || irqs_disabled())
3209  *
3210  * We provide four helpers that can be used in following contexts :
3211  *
3212  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3213  *  replacing kfree_skb(skb)
3214  *
3215  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3216  *  Typically used in place of consume_skb(skb) in TX completion path
3217  *
3218  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3219  *  replacing kfree_skb(skb)
3220  *
3221  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3222  *  and consumed a packet. Used in place of consume_skb(skb)
3223  */
3224 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3225 {
3226 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3227 }
3228 
3229 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3230 {
3231 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3232 }
3233 
3234 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3235 {
3236 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3237 }
3238 
3239 static inline void dev_consume_skb_any(struct sk_buff *skb)
3240 {
3241 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3242 }
3243 
3244 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3245 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3246 int netif_rx(struct sk_buff *skb);
3247 int netif_rx_ni(struct sk_buff *skb);
3248 int netif_receive_skb(struct sk_buff *skb);
3249 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3250 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3251 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3252 gro_result_t napi_gro_frags(struct napi_struct *napi);
3253 struct packet_offload *gro_find_receive_by_type(__be16 type);
3254 struct packet_offload *gro_find_complete_by_type(__be16 type);
3255 
3256 static inline void napi_free_frags(struct napi_struct *napi)
3257 {
3258 	kfree_skb(napi->skb);
3259 	napi->skb = NULL;
3260 }
3261 
3262 bool netdev_is_rx_handler_busy(struct net_device *dev);
3263 int netdev_rx_handler_register(struct net_device *dev,
3264 			       rx_handler_func_t *rx_handler,
3265 			       void *rx_handler_data);
3266 void netdev_rx_handler_unregister(struct net_device *dev);
3267 
3268 bool dev_valid_name(const char *name);
3269 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3270 int dev_ethtool(struct net *net, struct ifreq *);
3271 unsigned int dev_get_flags(const struct net_device *);
3272 int __dev_change_flags(struct net_device *, unsigned int flags);
3273 int dev_change_flags(struct net_device *, unsigned int);
3274 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3275 			unsigned int gchanges);
3276 int dev_change_name(struct net_device *, const char *);
3277 int dev_set_alias(struct net_device *, const char *, size_t);
3278 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3279 int __dev_set_mtu(struct net_device *, int);
3280 int dev_set_mtu(struct net_device *, int);
3281 void dev_set_group(struct net_device *, int);
3282 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3283 int dev_change_carrier(struct net_device *, bool new_carrier);
3284 int dev_get_phys_port_id(struct net_device *dev,
3285 			 struct netdev_phys_item_id *ppid);
3286 int dev_get_phys_port_name(struct net_device *dev,
3287 			   char *name, size_t len);
3288 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3289 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3290 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3291 				    struct netdev_queue *txq, int *ret);
3292 
3293 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3294 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3295 		      int fd, u32 flags);
3296 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
3297 
3298 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3299 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3300 bool is_skb_forwardable(const struct net_device *dev,
3301 			const struct sk_buff *skb);
3302 
3303 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3304 					       struct sk_buff *skb)
3305 {
3306 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3307 	    unlikely(!is_skb_forwardable(dev, skb))) {
3308 		atomic_long_inc(&dev->rx_dropped);
3309 		kfree_skb(skb);
3310 		return NET_RX_DROP;
3311 	}
3312 
3313 	skb_scrub_packet(skb, true);
3314 	skb->priority = 0;
3315 	return 0;
3316 }
3317 
3318 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3319 
3320 extern int		netdev_budget;
3321 extern unsigned int	netdev_budget_usecs;
3322 
3323 /* Called by rtnetlink.c:rtnl_unlock() */
3324 void netdev_run_todo(void);
3325 
3326 /**
3327  *	dev_put - release reference to device
3328  *	@dev: network device
3329  *
3330  * Release reference to device to allow it to be freed.
3331  */
3332 static inline void dev_put(struct net_device *dev)
3333 {
3334 	this_cpu_dec(*dev->pcpu_refcnt);
3335 }
3336 
3337 /**
3338  *	dev_hold - get reference to device
3339  *	@dev: network device
3340  *
3341  * Hold reference to device to keep it from being freed.
3342  */
3343 static inline void dev_hold(struct net_device *dev)
3344 {
3345 	this_cpu_inc(*dev->pcpu_refcnt);
3346 }
3347 
3348 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3349  * and _off may be called from IRQ context, but it is caller
3350  * who is responsible for serialization of these calls.
3351  *
3352  * The name carrier is inappropriate, these functions should really be
3353  * called netif_lowerlayer_*() because they represent the state of any
3354  * kind of lower layer not just hardware media.
3355  */
3356 
3357 void linkwatch_init_dev(struct net_device *dev);
3358 void linkwatch_fire_event(struct net_device *dev);
3359 void linkwatch_forget_dev(struct net_device *dev);
3360 
3361 /**
3362  *	netif_carrier_ok - test if carrier present
3363  *	@dev: network device
3364  *
3365  * Check if carrier is present on device
3366  */
3367 static inline bool netif_carrier_ok(const struct net_device *dev)
3368 {
3369 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3370 }
3371 
3372 unsigned long dev_trans_start(struct net_device *dev);
3373 
3374 void __netdev_watchdog_up(struct net_device *dev);
3375 
3376 void netif_carrier_on(struct net_device *dev);
3377 
3378 void netif_carrier_off(struct net_device *dev);
3379 
3380 /**
3381  *	netif_dormant_on - mark device as dormant.
3382  *	@dev: network device
3383  *
3384  * Mark device as dormant (as per RFC2863).
3385  *
3386  * The dormant state indicates that the relevant interface is not
3387  * actually in a condition to pass packets (i.e., it is not 'up') but is
3388  * in a "pending" state, waiting for some external event.  For "on-
3389  * demand" interfaces, this new state identifies the situation where the
3390  * interface is waiting for events to place it in the up state.
3391  */
3392 static inline void netif_dormant_on(struct net_device *dev)
3393 {
3394 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3395 		linkwatch_fire_event(dev);
3396 }
3397 
3398 /**
3399  *	netif_dormant_off - set device as not dormant.
3400  *	@dev: network device
3401  *
3402  * Device is not in dormant state.
3403  */
3404 static inline void netif_dormant_off(struct net_device *dev)
3405 {
3406 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3407 		linkwatch_fire_event(dev);
3408 }
3409 
3410 /**
3411  *	netif_dormant - test if device is dormant
3412  *	@dev: network device
3413  *
3414  * Check if device is dormant.
3415  */
3416 static inline bool netif_dormant(const struct net_device *dev)
3417 {
3418 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3419 }
3420 
3421 
3422 /**
3423  *	netif_oper_up - test if device is operational
3424  *	@dev: network device
3425  *
3426  * Check if carrier is operational
3427  */
3428 static inline bool netif_oper_up(const struct net_device *dev)
3429 {
3430 	return (dev->operstate == IF_OPER_UP ||
3431 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3432 }
3433 
3434 /**
3435  *	netif_device_present - is device available or removed
3436  *	@dev: network device
3437  *
3438  * Check if device has not been removed from system.
3439  */
3440 static inline bool netif_device_present(struct net_device *dev)
3441 {
3442 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3443 }
3444 
3445 void netif_device_detach(struct net_device *dev);
3446 
3447 void netif_device_attach(struct net_device *dev);
3448 
3449 /*
3450  * Network interface message level settings
3451  */
3452 
3453 enum {
3454 	NETIF_MSG_DRV		= 0x0001,
3455 	NETIF_MSG_PROBE		= 0x0002,
3456 	NETIF_MSG_LINK		= 0x0004,
3457 	NETIF_MSG_TIMER		= 0x0008,
3458 	NETIF_MSG_IFDOWN	= 0x0010,
3459 	NETIF_MSG_IFUP		= 0x0020,
3460 	NETIF_MSG_RX_ERR	= 0x0040,
3461 	NETIF_MSG_TX_ERR	= 0x0080,
3462 	NETIF_MSG_TX_QUEUED	= 0x0100,
3463 	NETIF_MSG_INTR		= 0x0200,
3464 	NETIF_MSG_TX_DONE	= 0x0400,
3465 	NETIF_MSG_RX_STATUS	= 0x0800,
3466 	NETIF_MSG_PKTDATA	= 0x1000,
3467 	NETIF_MSG_HW		= 0x2000,
3468 	NETIF_MSG_WOL		= 0x4000,
3469 };
3470 
3471 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3472 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3473 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3474 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3475 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3476 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3477 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3478 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3479 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3480 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3481 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3482 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3483 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3484 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3485 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3486 
3487 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3488 {
3489 	/* use default */
3490 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3491 		return default_msg_enable_bits;
3492 	if (debug_value == 0)	/* no output */
3493 		return 0;
3494 	/* set low N bits */
3495 	return (1 << debug_value) - 1;
3496 }
3497 
3498 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3499 {
3500 	spin_lock(&txq->_xmit_lock);
3501 	txq->xmit_lock_owner = cpu;
3502 }
3503 
3504 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3505 {
3506 	__acquire(&txq->_xmit_lock);
3507 	return true;
3508 }
3509 
3510 static inline void __netif_tx_release(struct netdev_queue *txq)
3511 {
3512 	__release(&txq->_xmit_lock);
3513 }
3514 
3515 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3516 {
3517 	spin_lock_bh(&txq->_xmit_lock);
3518 	txq->xmit_lock_owner = smp_processor_id();
3519 }
3520 
3521 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3522 {
3523 	bool ok = spin_trylock(&txq->_xmit_lock);
3524 	if (likely(ok))
3525 		txq->xmit_lock_owner = smp_processor_id();
3526 	return ok;
3527 }
3528 
3529 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3530 {
3531 	txq->xmit_lock_owner = -1;
3532 	spin_unlock(&txq->_xmit_lock);
3533 }
3534 
3535 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3536 {
3537 	txq->xmit_lock_owner = -1;
3538 	spin_unlock_bh(&txq->_xmit_lock);
3539 }
3540 
3541 static inline void txq_trans_update(struct netdev_queue *txq)
3542 {
3543 	if (txq->xmit_lock_owner != -1)
3544 		txq->trans_start = jiffies;
3545 }
3546 
3547 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3548 static inline void netif_trans_update(struct net_device *dev)
3549 {
3550 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3551 
3552 	if (txq->trans_start != jiffies)
3553 		txq->trans_start = jiffies;
3554 }
3555 
3556 /**
3557  *	netif_tx_lock - grab network device transmit lock
3558  *	@dev: network device
3559  *
3560  * Get network device transmit lock
3561  */
3562 static inline void netif_tx_lock(struct net_device *dev)
3563 {
3564 	unsigned int i;
3565 	int cpu;
3566 
3567 	spin_lock(&dev->tx_global_lock);
3568 	cpu = smp_processor_id();
3569 	for (i = 0; i < dev->num_tx_queues; i++) {
3570 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3571 
3572 		/* We are the only thread of execution doing a
3573 		 * freeze, but we have to grab the _xmit_lock in
3574 		 * order to synchronize with threads which are in
3575 		 * the ->hard_start_xmit() handler and already
3576 		 * checked the frozen bit.
3577 		 */
3578 		__netif_tx_lock(txq, cpu);
3579 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3580 		__netif_tx_unlock(txq);
3581 	}
3582 }
3583 
3584 static inline void netif_tx_lock_bh(struct net_device *dev)
3585 {
3586 	local_bh_disable();
3587 	netif_tx_lock(dev);
3588 }
3589 
3590 static inline void netif_tx_unlock(struct net_device *dev)
3591 {
3592 	unsigned int i;
3593 
3594 	for (i = 0; i < dev->num_tx_queues; i++) {
3595 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3596 
3597 		/* No need to grab the _xmit_lock here.  If the
3598 		 * queue is not stopped for another reason, we
3599 		 * force a schedule.
3600 		 */
3601 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3602 		netif_schedule_queue(txq);
3603 	}
3604 	spin_unlock(&dev->tx_global_lock);
3605 }
3606 
3607 static inline void netif_tx_unlock_bh(struct net_device *dev)
3608 {
3609 	netif_tx_unlock(dev);
3610 	local_bh_enable();
3611 }
3612 
3613 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3614 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3615 		__netif_tx_lock(txq, cpu);		\
3616 	} else {					\
3617 		__netif_tx_acquire(txq);		\
3618 	}						\
3619 }
3620 
3621 #define HARD_TX_TRYLOCK(dev, txq)			\
3622 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3623 		__netif_tx_trylock(txq) :		\
3624 		__netif_tx_acquire(txq))
3625 
3626 #define HARD_TX_UNLOCK(dev, txq) {			\
3627 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3628 		__netif_tx_unlock(txq);			\
3629 	} else {					\
3630 		__netif_tx_release(txq);		\
3631 	}						\
3632 }
3633 
3634 static inline void netif_tx_disable(struct net_device *dev)
3635 {
3636 	unsigned int i;
3637 	int cpu;
3638 
3639 	local_bh_disable();
3640 	cpu = smp_processor_id();
3641 	for (i = 0; i < dev->num_tx_queues; i++) {
3642 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3643 
3644 		__netif_tx_lock(txq, cpu);
3645 		netif_tx_stop_queue(txq);
3646 		__netif_tx_unlock(txq);
3647 	}
3648 	local_bh_enable();
3649 }
3650 
3651 static inline void netif_addr_lock(struct net_device *dev)
3652 {
3653 	spin_lock(&dev->addr_list_lock);
3654 }
3655 
3656 static inline void netif_addr_lock_nested(struct net_device *dev)
3657 {
3658 	int subclass = SINGLE_DEPTH_NESTING;
3659 
3660 	if (dev->netdev_ops->ndo_get_lock_subclass)
3661 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3662 
3663 	spin_lock_nested(&dev->addr_list_lock, subclass);
3664 }
3665 
3666 static inline void netif_addr_lock_bh(struct net_device *dev)
3667 {
3668 	spin_lock_bh(&dev->addr_list_lock);
3669 }
3670 
3671 static inline void netif_addr_unlock(struct net_device *dev)
3672 {
3673 	spin_unlock(&dev->addr_list_lock);
3674 }
3675 
3676 static inline void netif_addr_unlock_bh(struct net_device *dev)
3677 {
3678 	spin_unlock_bh(&dev->addr_list_lock);
3679 }
3680 
3681 /*
3682  * dev_addrs walker. Should be used only for read access. Call with
3683  * rcu_read_lock held.
3684  */
3685 #define for_each_dev_addr(dev, ha) \
3686 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3687 
3688 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3689 
3690 void ether_setup(struct net_device *dev);
3691 
3692 /* Support for loadable net-drivers */
3693 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3694 				    unsigned char name_assign_type,
3695 				    void (*setup)(struct net_device *),
3696 				    unsigned int txqs, unsigned int rxqs);
3697 int dev_get_valid_name(struct net *net, struct net_device *dev,
3698 		       const char *name);
3699 
3700 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3701 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3702 
3703 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3704 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3705 			 count)
3706 
3707 int register_netdev(struct net_device *dev);
3708 void unregister_netdev(struct net_device *dev);
3709 
3710 /* General hardware address lists handling functions */
3711 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3712 		   struct netdev_hw_addr_list *from_list, int addr_len);
3713 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3714 		      struct netdev_hw_addr_list *from_list, int addr_len);
3715 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3716 		       struct net_device *dev,
3717 		       int (*sync)(struct net_device *, const unsigned char *),
3718 		       int (*unsync)(struct net_device *,
3719 				     const unsigned char *));
3720 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3721 			  struct net_device *dev,
3722 			  int (*unsync)(struct net_device *,
3723 					const unsigned char *));
3724 void __hw_addr_init(struct netdev_hw_addr_list *list);
3725 
3726 /* Functions used for device addresses handling */
3727 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3728 		 unsigned char addr_type);
3729 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3730 		 unsigned char addr_type);
3731 void dev_addr_flush(struct net_device *dev);
3732 int dev_addr_init(struct net_device *dev);
3733 
3734 /* Functions used for unicast addresses handling */
3735 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3736 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3737 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3738 int dev_uc_sync(struct net_device *to, struct net_device *from);
3739 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3740 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3741 void dev_uc_flush(struct net_device *dev);
3742 void dev_uc_init(struct net_device *dev);
3743 
3744 /**
3745  *  __dev_uc_sync - Synchonize device's unicast list
3746  *  @dev:  device to sync
3747  *  @sync: function to call if address should be added
3748  *  @unsync: function to call if address should be removed
3749  *
3750  *  Add newly added addresses to the interface, and release
3751  *  addresses that have been deleted.
3752  */
3753 static inline int __dev_uc_sync(struct net_device *dev,
3754 				int (*sync)(struct net_device *,
3755 					    const unsigned char *),
3756 				int (*unsync)(struct net_device *,
3757 					      const unsigned char *))
3758 {
3759 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3760 }
3761 
3762 /**
3763  *  __dev_uc_unsync - Remove synchronized addresses from device
3764  *  @dev:  device to sync
3765  *  @unsync: function to call if address should be removed
3766  *
3767  *  Remove all addresses that were added to the device by dev_uc_sync().
3768  */
3769 static inline void __dev_uc_unsync(struct net_device *dev,
3770 				   int (*unsync)(struct net_device *,
3771 						 const unsigned char *))
3772 {
3773 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3774 }
3775 
3776 /* Functions used for multicast addresses handling */
3777 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3778 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3779 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3780 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3781 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3782 int dev_mc_sync(struct net_device *to, struct net_device *from);
3783 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3784 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3785 void dev_mc_flush(struct net_device *dev);
3786 void dev_mc_init(struct net_device *dev);
3787 
3788 /**
3789  *  __dev_mc_sync - Synchonize device's multicast list
3790  *  @dev:  device to sync
3791  *  @sync: function to call if address should be added
3792  *  @unsync: function to call if address should be removed
3793  *
3794  *  Add newly added addresses to the interface, and release
3795  *  addresses that have been deleted.
3796  */
3797 static inline int __dev_mc_sync(struct net_device *dev,
3798 				int (*sync)(struct net_device *,
3799 					    const unsigned char *),
3800 				int (*unsync)(struct net_device *,
3801 					      const unsigned char *))
3802 {
3803 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3804 }
3805 
3806 /**
3807  *  __dev_mc_unsync - Remove synchronized addresses from device
3808  *  @dev:  device to sync
3809  *  @unsync: function to call if address should be removed
3810  *
3811  *  Remove all addresses that were added to the device by dev_mc_sync().
3812  */
3813 static inline void __dev_mc_unsync(struct net_device *dev,
3814 				   int (*unsync)(struct net_device *,
3815 						 const unsigned char *))
3816 {
3817 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3818 }
3819 
3820 /* Functions used for secondary unicast and multicast support */
3821 void dev_set_rx_mode(struct net_device *dev);
3822 void __dev_set_rx_mode(struct net_device *dev);
3823 int dev_set_promiscuity(struct net_device *dev, int inc);
3824 int dev_set_allmulti(struct net_device *dev, int inc);
3825 void netdev_state_change(struct net_device *dev);
3826 void netdev_notify_peers(struct net_device *dev);
3827 void netdev_features_change(struct net_device *dev);
3828 /* Load a device via the kmod */
3829 void dev_load(struct net *net, const char *name);
3830 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3831 					struct rtnl_link_stats64 *storage);
3832 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3833 			     const struct net_device_stats *netdev_stats);
3834 
3835 extern int		netdev_max_backlog;
3836 extern int		netdev_tstamp_prequeue;
3837 extern int		weight_p;
3838 extern int		dev_weight_rx_bias;
3839 extern int		dev_weight_tx_bias;
3840 extern int		dev_rx_weight;
3841 extern int		dev_tx_weight;
3842 
3843 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3844 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3845 						     struct list_head **iter);
3846 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3847 						     struct list_head **iter);
3848 
3849 /* iterate through upper list, must be called under RCU read lock */
3850 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3851 	for (iter = &(dev)->adj_list.upper, \
3852 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3853 	     updev; \
3854 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3855 
3856 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3857 				  int (*fn)(struct net_device *upper_dev,
3858 					    void *data),
3859 				  void *data);
3860 
3861 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3862 				  struct net_device *upper_dev);
3863 
3864 bool netdev_has_any_upper_dev(struct net_device *dev);
3865 
3866 void *netdev_lower_get_next_private(struct net_device *dev,
3867 				    struct list_head **iter);
3868 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3869 					struct list_head **iter);
3870 
3871 #define netdev_for_each_lower_private(dev, priv, iter) \
3872 	for (iter = (dev)->adj_list.lower.next, \
3873 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3874 	     priv; \
3875 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3876 
3877 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3878 	for (iter = &(dev)->adj_list.lower, \
3879 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3880 	     priv; \
3881 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3882 
3883 void *netdev_lower_get_next(struct net_device *dev,
3884 				struct list_head **iter);
3885 
3886 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3887 	for (iter = (dev)->adj_list.lower.next, \
3888 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3889 	     ldev; \
3890 	     ldev = netdev_lower_get_next(dev, &(iter)))
3891 
3892 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3893 					     struct list_head **iter);
3894 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3895 						 struct list_head **iter);
3896 
3897 int netdev_walk_all_lower_dev(struct net_device *dev,
3898 			      int (*fn)(struct net_device *lower_dev,
3899 					void *data),
3900 			      void *data);
3901 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3902 				  int (*fn)(struct net_device *lower_dev,
3903 					    void *data),
3904 				  void *data);
3905 
3906 void *netdev_adjacent_get_private(struct list_head *adj_list);
3907 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3908 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3909 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3910 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3911 int netdev_master_upper_dev_link(struct net_device *dev,
3912 				 struct net_device *upper_dev,
3913 				 void *upper_priv, void *upper_info);
3914 void netdev_upper_dev_unlink(struct net_device *dev,
3915 			     struct net_device *upper_dev);
3916 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3917 void *netdev_lower_dev_get_private(struct net_device *dev,
3918 				   struct net_device *lower_dev);
3919 void netdev_lower_state_changed(struct net_device *lower_dev,
3920 				void *lower_state_info);
3921 
3922 /* RSS keys are 40 or 52 bytes long */
3923 #define NETDEV_RSS_KEY_LEN 52
3924 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3925 void netdev_rss_key_fill(void *buffer, size_t len);
3926 
3927 int dev_get_nest_level(struct net_device *dev);
3928 int skb_checksum_help(struct sk_buff *skb);
3929 int skb_crc32c_csum_help(struct sk_buff *skb);
3930 int skb_csum_hwoffload_help(struct sk_buff *skb,
3931 			    const netdev_features_t features);
3932 
3933 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3934 				  netdev_features_t features, bool tx_path);
3935 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3936 				    netdev_features_t features);
3937 
3938 struct netdev_bonding_info {
3939 	ifslave	slave;
3940 	ifbond	master;
3941 };
3942 
3943 struct netdev_notifier_bonding_info {
3944 	struct netdev_notifier_info info; /* must be first */
3945 	struct netdev_bonding_info  bonding_info;
3946 };
3947 
3948 void netdev_bonding_info_change(struct net_device *dev,
3949 				struct netdev_bonding_info *bonding_info);
3950 
3951 static inline
3952 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3953 {
3954 	return __skb_gso_segment(skb, features, true);
3955 }
3956 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3957 
3958 static inline bool can_checksum_protocol(netdev_features_t features,
3959 					 __be16 protocol)
3960 {
3961 	if (protocol == htons(ETH_P_FCOE))
3962 		return !!(features & NETIF_F_FCOE_CRC);
3963 
3964 	/* Assume this is an IP checksum (not SCTP CRC) */
3965 
3966 	if (features & NETIF_F_HW_CSUM) {
3967 		/* Can checksum everything */
3968 		return true;
3969 	}
3970 
3971 	switch (protocol) {
3972 	case htons(ETH_P_IP):
3973 		return !!(features & NETIF_F_IP_CSUM);
3974 	case htons(ETH_P_IPV6):
3975 		return !!(features & NETIF_F_IPV6_CSUM);
3976 	default:
3977 		return false;
3978 	}
3979 }
3980 
3981 #ifdef CONFIG_BUG
3982 void netdev_rx_csum_fault(struct net_device *dev);
3983 #else
3984 static inline void netdev_rx_csum_fault(struct net_device *dev)
3985 {
3986 }
3987 #endif
3988 /* rx skb timestamps */
3989 void net_enable_timestamp(void);
3990 void net_disable_timestamp(void);
3991 
3992 #ifdef CONFIG_PROC_FS
3993 int __init dev_proc_init(void);
3994 #else
3995 #define dev_proc_init() 0
3996 #endif
3997 
3998 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3999 					      struct sk_buff *skb, struct net_device *dev,
4000 					      bool more)
4001 {
4002 	skb->xmit_more = more ? 1 : 0;
4003 	return ops->ndo_start_xmit(skb, dev);
4004 }
4005 
4006 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4007 					    struct netdev_queue *txq, bool more)
4008 {
4009 	const struct net_device_ops *ops = dev->netdev_ops;
4010 	int rc;
4011 
4012 	rc = __netdev_start_xmit(ops, skb, dev, more);
4013 	if (rc == NETDEV_TX_OK)
4014 		txq_trans_update(txq);
4015 
4016 	return rc;
4017 }
4018 
4019 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4020 				const void *ns);
4021 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4022 				 const void *ns);
4023 
4024 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4025 {
4026 	return netdev_class_create_file_ns(class_attr, NULL);
4027 }
4028 
4029 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4030 {
4031 	netdev_class_remove_file_ns(class_attr, NULL);
4032 }
4033 
4034 extern const struct kobj_ns_type_operations net_ns_type_operations;
4035 
4036 const char *netdev_drivername(const struct net_device *dev);
4037 
4038 void linkwatch_run_queue(void);
4039 
4040 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4041 							  netdev_features_t f2)
4042 {
4043 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4044 		if (f1 & NETIF_F_HW_CSUM)
4045 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4046 		else
4047 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4048 	}
4049 
4050 	return f1 & f2;
4051 }
4052 
4053 static inline netdev_features_t netdev_get_wanted_features(
4054 	struct net_device *dev)
4055 {
4056 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4057 }
4058 netdev_features_t netdev_increment_features(netdev_features_t all,
4059 	netdev_features_t one, netdev_features_t mask);
4060 
4061 /* Allow TSO being used on stacked device :
4062  * Performing the GSO segmentation before last device
4063  * is a performance improvement.
4064  */
4065 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4066 							netdev_features_t mask)
4067 {
4068 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4069 }
4070 
4071 int __netdev_update_features(struct net_device *dev);
4072 void netdev_update_features(struct net_device *dev);
4073 void netdev_change_features(struct net_device *dev);
4074 
4075 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4076 					struct net_device *dev);
4077 
4078 netdev_features_t passthru_features_check(struct sk_buff *skb,
4079 					  struct net_device *dev,
4080 					  netdev_features_t features);
4081 netdev_features_t netif_skb_features(struct sk_buff *skb);
4082 
4083 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4084 {
4085 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4086 
4087 	/* check flags correspondence */
4088 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4089 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4090 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4091 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4092 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4093 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4094 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4095 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4096 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4097 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4098 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4099 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4100 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4101 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4102 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4103 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4104 
4105 	return (features & feature) == feature;
4106 }
4107 
4108 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4109 {
4110 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4111 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4112 }
4113 
4114 static inline bool netif_needs_gso(struct sk_buff *skb,
4115 				   netdev_features_t features)
4116 {
4117 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4118 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4119 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4120 }
4121 
4122 static inline void netif_set_gso_max_size(struct net_device *dev,
4123 					  unsigned int size)
4124 {
4125 	dev->gso_max_size = size;
4126 }
4127 
4128 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4129 					int pulled_hlen, u16 mac_offset,
4130 					int mac_len)
4131 {
4132 	skb->protocol = protocol;
4133 	skb->encapsulation = 1;
4134 	skb_push(skb, pulled_hlen);
4135 	skb_reset_transport_header(skb);
4136 	skb->mac_header = mac_offset;
4137 	skb->network_header = skb->mac_header + mac_len;
4138 	skb->mac_len = mac_len;
4139 }
4140 
4141 static inline bool netif_is_macsec(const struct net_device *dev)
4142 {
4143 	return dev->priv_flags & IFF_MACSEC;
4144 }
4145 
4146 static inline bool netif_is_macvlan(const struct net_device *dev)
4147 {
4148 	return dev->priv_flags & IFF_MACVLAN;
4149 }
4150 
4151 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4152 {
4153 	return dev->priv_flags & IFF_MACVLAN_PORT;
4154 }
4155 
4156 static inline bool netif_is_ipvlan(const struct net_device *dev)
4157 {
4158 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4159 }
4160 
4161 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4162 {
4163 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4164 }
4165 
4166 static inline bool netif_is_bond_master(const struct net_device *dev)
4167 {
4168 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4169 }
4170 
4171 static inline bool netif_is_bond_slave(const struct net_device *dev)
4172 {
4173 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4174 }
4175 
4176 static inline bool netif_supports_nofcs(struct net_device *dev)
4177 {
4178 	return dev->priv_flags & IFF_SUPP_NOFCS;
4179 }
4180 
4181 static inline bool netif_is_l3_master(const struct net_device *dev)
4182 {
4183 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4184 }
4185 
4186 static inline bool netif_is_l3_slave(const struct net_device *dev)
4187 {
4188 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4189 }
4190 
4191 static inline bool netif_is_bridge_master(const struct net_device *dev)
4192 {
4193 	return dev->priv_flags & IFF_EBRIDGE;
4194 }
4195 
4196 static inline bool netif_is_bridge_port(const struct net_device *dev)
4197 {
4198 	return dev->priv_flags & IFF_BRIDGE_PORT;
4199 }
4200 
4201 static inline bool netif_is_ovs_master(const struct net_device *dev)
4202 {
4203 	return dev->priv_flags & IFF_OPENVSWITCH;
4204 }
4205 
4206 static inline bool netif_is_ovs_port(const struct net_device *dev)
4207 {
4208 	return dev->priv_flags & IFF_OVS_DATAPATH;
4209 }
4210 
4211 static inline bool netif_is_team_master(const struct net_device *dev)
4212 {
4213 	return dev->priv_flags & IFF_TEAM;
4214 }
4215 
4216 static inline bool netif_is_team_port(const struct net_device *dev)
4217 {
4218 	return dev->priv_flags & IFF_TEAM_PORT;
4219 }
4220 
4221 static inline bool netif_is_lag_master(const struct net_device *dev)
4222 {
4223 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4224 }
4225 
4226 static inline bool netif_is_lag_port(const struct net_device *dev)
4227 {
4228 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4229 }
4230 
4231 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4232 {
4233 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4234 }
4235 
4236 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4237 static inline void netif_keep_dst(struct net_device *dev)
4238 {
4239 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4240 }
4241 
4242 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4243 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4244 {
4245 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4246 	return dev->priv_flags & IFF_MACSEC;
4247 }
4248 
4249 extern struct pernet_operations __net_initdata loopback_net_ops;
4250 
4251 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4252 
4253 /* netdev_printk helpers, similar to dev_printk */
4254 
4255 static inline const char *netdev_name(const struct net_device *dev)
4256 {
4257 	if (!dev->name[0] || strchr(dev->name, '%'))
4258 		return "(unnamed net_device)";
4259 	return dev->name;
4260 }
4261 
4262 static inline bool netdev_unregistering(const struct net_device *dev)
4263 {
4264 	return dev->reg_state == NETREG_UNREGISTERING;
4265 }
4266 
4267 static inline const char *netdev_reg_state(const struct net_device *dev)
4268 {
4269 	switch (dev->reg_state) {
4270 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4271 	case NETREG_REGISTERED: return "";
4272 	case NETREG_UNREGISTERING: return " (unregistering)";
4273 	case NETREG_UNREGISTERED: return " (unregistered)";
4274 	case NETREG_RELEASED: return " (released)";
4275 	case NETREG_DUMMY: return " (dummy)";
4276 	}
4277 
4278 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4279 	return " (unknown)";
4280 }
4281 
4282 __printf(3, 4)
4283 void netdev_printk(const char *level, const struct net_device *dev,
4284 		   const char *format, ...);
4285 __printf(2, 3)
4286 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4287 __printf(2, 3)
4288 void netdev_alert(const struct net_device *dev, const char *format, ...);
4289 __printf(2, 3)
4290 void netdev_crit(const struct net_device *dev, const char *format, ...);
4291 __printf(2, 3)
4292 void netdev_err(const struct net_device *dev, const char *format, ...);
4293 __printf(2, 3)
4294 void netdev_warn(const struct net_device *dev, const char *format, ...);
4295 __printf(2, 3)
4296 void netdev_notice(const struct net_device *dev, const char *format, ...);
4297 __printf(2, 3)
4298 void netdev_info(const struct net_device *dev, const char *format, ...);
4299 
4300 #define MODULE_ALIAS_NETDEV(device) \
4301 	MODULE_ALIAS("netdev-" device)
4302 
4303 #if defined(CONFIG_DYNAMIC_DEBUG)
4304 #define netdev_dbg(__dev, format, args...)			\
4305 do {								\
4306 	dynamic_netdev_dbg(__dev, format, ##args);		\
4307 } while (0)
4308 #elif defined(DEBUG)
4309 #define netdev_dbg(__dev, format, args...)			\
4310 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4311 #else
4312 #define netdev_dbg(__dev, format, args...)			\
4313 ({								\
4314 	if (0)							\
4315 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4316 })
4317 #endif
4318 
4319 #if defined(VERBOSE_DEBUG)
4320 #define netdev_vdbg	netdev_dbg
4321 #else
4322 
4323 #define netdev_vdbg(dev, format, args...)			\
4324 ({								\
4325 	if (0)							\
4326 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4327 	0;							\
4328 })
4329 #endif
4330 
4331 /*
4332  * netdev_WARN() acts like dev_printk(), but with the key difference
4333  * of using a WARN/WARN_ON to get the message out, including the
4334  * file/line information and a backtrace.
4335  */
4336 #define netdev_WARN(dev, format, args...)			\
4337 	WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),	\
4338 	     netdev_reg_state(dev), ##args)
4339 
4340 /* netif printk helpers, similar to netdev_printk */
4341 
4342 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4343 do {					  			\
4344 	if (netif_msg_##type(priv))				\
4345 		netdev_printk(level, (dev), fmt, ##args);	\
4346 } while (0)
4347 
4348 #define netif_level(level, priv, type, dev, fmt, args...)	\
4349 do {								\
4350 	if (netif_msg_##type(priv))				\
4351 		netdev_##level(dev, fmt, ##args);		\
4352 } while (0)
4353 
4354 #define netif_emerg(priv, type, dev, fmt, args...)		\
4355 	netif_level(emerg, priv, type, dev, fmt, ##args)
4356 #define netif_alert(priv, type, dev, fmt, args...)		\
4357 	netif_level(alert, priv, type, dev, fmt, ##args)
4358 #define netif_crit(priv, type, dev, fmt, args...)		\
4359 	netif_level(crit, priv, type, dev, fmt, ##args)
4360 #define netif_err(priv, type, dev, fmt, args...)		\
4361 	netif_level(err, priv, type, dev, fmt, ##args)
4362 #define netif_warn(priv, type, dev, fmt, args...)		\
4363 	netif_level(warn, priv, type, dev, fmt, ##args)
4364 #define netif_notice(priv, type, dev, fmt, args...)		\
4365 	netif_level(notice, priv, type, dev, fmt, ##args)
4366 #define netif_info(priv, type, dev, fmt, args...)		\
4367 	netif_level(info, priv, type, dev, fmt, ##args)
4368 
4369 #if defined(CONFIG_DYNAMIC_DEBUG)
4370 #define netif_dbg(priv, type, netdev, format, args...)		\
4371 do {								\
4372 	if (netif_msg_##type(priv))				\
4373 		dynamic_netdev_dbg(netdev, format, ##args);	\
4374 } while (0)
4375 #elif defined(DEBUG)
4376 #define netif_dbg(priv, type, dev, format, args...)		\
4377 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4378 #else
4379 #define netif_dbg(priv, type, dev, format, args...)			\
4380 ({									\
4381 	if (0)								\
4382 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4383 	0;								\
4384 })
4385 #endif
4386 
4387 /* if @cond then downgrade to debug, else print at @level */
4388 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4389 	do {                                                              \
4390 		if (cond)                                                 \
4391 			netif_dbg(priv, type, netdev, fmt, ##args);       \
4392 		else                                                      \
4393 			netif_ ## level(priv, type, netdev, fmt, ##args); \
4394 	} while (0)
4395 
4396 #if defined(VERBOSE_DEBUG)
4397 #define netif_vdbg	netif_dbg
4398 #else
4399 #define netif_vdbg(priv, type, dev, format, args...)		\
4400 ({								\
4401 	if (0)							\
4402 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4403 	0;							\
4404 })
4405 #endif
4406 
4407 /*
4408  *	The list of packet types we will receive (as opposed to discard)
4409  *	and the routines to invoke.
4410  *
4411  *	Why 16. Because with 16 the only overlap we get on a hash of the
4412  *	low nibble of the protocol value is RARP/SNAP/X.25.
4413  *
4414  *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4415  *             sure which should go first, but I bet it won't make much
4416  *             difference if we are running VLANs.  The good news is that
4417  *             this protocol won't be in the list unless compiled in, so
4418  *             the average user (w/out VLANs) will not be adversely affected.
4419  *             --BLG
4420  *
4421  *		0800	IP
4422  *		8100    802.1Q VLAN
4423  *		0001	802.3
4424  *		0002	AX.25
4425  *		0004	802.2
4426  *		8035	RARP
4427  *		0005	SNAP
4428  *		0805	X.25
4429  *		0806	ARP
4430  *		8137	IPX
4431  *		0009	Localtalk
4432  *		86DD	IPv6
4433  */
4434 #define PTYPE_HASH_SIZE	(16)
4435 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4436 
4437 #endif	/* _LINUX_NETDEVICE_H */
4438