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