]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blob - include/linux/netdevice.h
netfilter: nf_flow_table: hardware offload support
[mirror_ubuntu-hirsute-kernel.git] / include / linux / netdevice.h
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
23
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
31
32 #include <linux/percpu.h>
33 #include <linux/rculist.h>
34 #include <linux/workqueue.h>
35 #include <linux/dynamic_queue_limits.h>
36
37 #include <linux/ethtool.h>
38 #include <net/net_namespace.h>
39 #ifdef CONFIG_DCB
40 #include <net/dcbnl.h>
41 #endif
42 #include <net/netprio_cgroup.h>
43 #include <net/xdp.h>
44
45 #include <linux/netdev_features.h>
46 #include <linux/neighbour.h>
47 #include <uapi/linux/netdevice.h>
48 #include <uapi/linux/if_bonding.h>
49 #include <uapi/linux/pkt_cls.h>
50 #include <linux/hashtable.h>
51
52 struct netpoll_info;
53 struct device;
54 struct phy_device;
55 struct dsa_port;
56
57 struct sfp_bus;
58 /* 802.11 specific */
59 struct wireless_dev;
60 /* 802.15.4 specific */
61 struct wpan_dev;
62 struct mpls_dev;
63 /* UDP Tunnel offloads */
64 struct udp_tunnel_info;
65 struct bpf_prog;
66 struct xdp_buff;
67
68 void netdev_set_default_ethtool_ops(struct net_device *dev,
69 const struct ethtool_ops *ops);
70
71 /* Backlog congestion levels */
72 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
73 #define NET_RX_DROP 1 /* packet dropped */
74
75 /*
76 * Transmit return codes: transmit return codes originate from three different
77 * namespaces:
78 *
79 * - qdisc return codes
80 * - driver transmit return codes
81 * - errno values
82 *
83 * Drivers are allowed to return any one of those in their hard_start_xmit()
84 * function. Real network devices commonly used with qdiscs should only return
85 * the driver transmit return codes though - when qdiscs are used, the actual
86 * transmission happens asynchronously, so the value is not propagated to
87 * higher layers. Virtual network devices transmit synchronously; in this case
88 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
89 * others are propagated to higher layers.
90 */
91
92 /* qdisc ->enqueue() return codes. */
93 #define NET_XMIT_SUCCESS 0x00
94 #define NET_XMIT_DROP 0x01 /* skb dropped */
95 #define NET_XMIT_CN 0x02 /* congestion notification */
96 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
97
98 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
99 * indicates that the device will soon be dropping packets, or already drops
100 * some packets of the same priority; prompting us to send less aggressively. */
101 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
102 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
103
104 /* Driver transmit return codes */
105 #define NETDEV_TX_MASK 0xf0
106
107 enum netdev_tx {
108 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
109 NETDEV_TX_OK = 0x00, /* driver took care of packet */
110 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
111 };
112 typedef enum netdev_tx netdev_tx_t;
113
114 /*
115 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
116 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
117 */
118 static inline bool dev_xmit_complete(int rc)
119 {
120 /*
121 * Positive cases with an skb consumed by a driver:
122 * - successful transmission (rc == NETDEV_TX_OK)
123 * - error while transmitting (rc < 0)
124 * - error while queueing to a different device (rc & NET_XMIT_MASK)
125 */
126 if (likely(rc < NET_XMIT_MASK))
127 return true;
128
129 return false;
130 }
131
132 /*
133 * Compute the worst-case header length according to the protocols
134 * used.
135 */
136
137 #if defined(CONFIG_HYPERV_NET)
138 # define LL_MAX_HEADER 128
139 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
140 # if defined(CONFIG_MAC80211_MESH)
141 # define LL_MAX_HEADER 128
142 # else
143 # define LL_MAX_HEADER 96
144 # endif
145 #else
146 # define LL_MAX_HEADER 32
147 #endif
148
149 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
150 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
151 #define MAX_HEADER LL_MAX_HEADER
152 #else
153 #define MAX_HEADER (LL_MAX_HEADER + 48)
154 #endif
155
156 /*
157 * Old network device statistics. Fields are native words
158 * (unsigned long) so they can be read and written atomically.
159 */
160
161 struct net_device_stats {
162 unsigned long rx_packets;
163 unsigned long tx_packets;
164 unsigned long rx_bytes;
165 unsigned long tx_bytes;
166 unsigned long rx_errors;
167 unsigned long tx_errors;
168 unsigned long rx_dropped;
169 unsigned long tx_dropped;
170 unsigned long multicast;
171 unsigned long collisions;
172 unsigned long rx_length_errors;
173 unsigned long rx_over_errors;
174 unsigned long rx_crc_errors;
175 unsigned long rx_frame_errors;
176 unsigned long rx_fifo_errors;
177 unsigned long rx_missed_errors;
178 unsigned long tx_aborted_errors;
179 unsigned long tx_carrier_errors;
180 unsigned long tx_fifo_errors;
181 unsigned long tx_heartbeat_errors;
182 unsigned long tx_window_errors;
183 unsigned long rx_compressed;
184 unsigned long tx_compressed;
185 };
186
187
188 #include <linux/cache.h>
189 #include <linux/skbuff.h>
190
191 #ifdef CONFIG_RPS
192 #include <linux/static_key.h>
193 extern struct static_key_false rps_needed;
194 extern struct static_key_false rfs_needed;
195 #endif
196
197 struct neighbour;
198 struct neigh_parms;
199 struct sk_buff;
200
201 struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
205 #define NETDEV_HW_ADDR_T_LAN 1
206 #define NETDEV_HW_ADDR_T_SAN 2
207 #define NETDEV_HW_ADDR_T_SLAVE 3
208 #define NETDEV_HW_ADDR_T_UNICAST 4
209 #define NETDEV_HW_ADDR_T_MULTICAST 5
210 bool global_use;
211 int sync_cnt;
212 int refcount;
213 int synced;
214 struct rcu_head rcu_head;
215 };
216
217 struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220 };
221
222 #define netdev_hw_addr_list_count(l) ((l)->count)
223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224 #define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
226
227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229 #define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
231
232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234 #define netdev_for_each_mc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
236
237 struct hh_cache {
238 unsigned int hh_len;
239 seqlock_t hh_lock;
240
241 /* cached hardware header; allow for machine alignment needs. */
242 #define HH_DATA_MOD 16
243 #define HH_DATA_OFF(__len) \
244 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
245 #define HH_DATA_ALIGN(__len) \
246 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
247 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
248 };
249
250 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
251 * Alternative is:
252 * dev->hard_header_len ? (dev->hard_header_len +
253 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
254 *
255 * We could use other alignment values, but we must maintain the
256 * relationship HH alignment <= LL alignment.
257 */
258 #define LL_RESERVED_SPACE(dev) \
259 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262
263 struct header_ops {
264 int (*create) (struct sk_buff *skb, struct net_device *dev,
265 unsigned short type, const void *daddr,
266 const void *saddr, unsigned int len);
267 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
268 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
269 void (*cache_update)(struct hh_cache *hh,
270 const struct net_device *dev,
271 const unsigned char *haddr);
272 bool (*validate)(const char *ll_header, unsigned int len);
273 __be16 (*parse_protocol)(const struct sk_buff *skb);
274 };
275
276 /* These flag bits are private to the generic network queueing
277 * layer; they may not be explicitly referenced by any other
278 * code.
279 */
280
281 enum netdev_state_t {
282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
284 __LINK_STATE_NOCARRIER,
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
287 };
288
289
290 /*
291 * This structure holds boot-time configured netdevice settings. They
292 * are then used in the device probing.
293 */
294 struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297 };
298 #define NETDEV_BOOT_SETUP_MAX 8
299
300 int __init netdev_boot_setup(char *str);
301
302 struct gro_list {
303 struct list_head list;
304 int count;
305 };
306
307 /*
308 * size of gro hash buckets, must less than bit number of
309 * napi_struct::gro_bitmask
310 */
311 #define GRO_HASH_BUCKETS 8
312
313 /*
314 * Structure for NAPI scheduling similar to tasklet but with weighting
315 */
316 struct napi_struct {
317 /* The poll_list must only be managed by the entity which
318 * changes the state of the NAPI_STATE_SCHED bit. This means
319 * whoever atomically sets that bit can add this napi_struct
320 * to the per-CPU poll_list, and whoever clears that bit
321 * can remove from the list right before clearing the bit.
322 */
323 struct list_head poll_list;
324
325 unsigned long state;
326 int weight;
327 unsigned long gro_bitmask;
328 int (*poll)(struct napi_struct *, int);
329 #ifdef CONFIG_NETPOLL
330 int poll_owner;
331 #endif
332 struct net_device *dev;
333 struct gro_list gro_hash[GRO_HASH_BUCKETS];
334 struct sk_buff *skb;
335 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
336 int rx_count; /* length of rx_list */
337 struct hrtimer timer;
338 struct list_head dev_list;
339 struct hlist_node napi_hash_node;
340 unsigned int napi_id;
341 };
342
343 enum {
344 NAPI_STATE_SCHED, /* Poll is scheduled */
345 NAPI_STATE_MISSED, /* reschedule a napi */
346 NAPI_STATE_DISABLE, /* Disable pending */
347 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
348 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
349 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
350 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
351 };
352
353 enum {
354 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
355 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
356 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
357 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
358 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
359 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
360 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
361 };
362
363 enum gro_result {
364 GRO_MERGED,
365 GRO_MERGED_FREE,
366 GRO_HELD,
367 GRO_NORMAL,
368 GRO_DROP,
369 GRO_CONSUMED,
370 };
371 typedef enum gro_result gro_result_t;
372
373 /*
374 * enum rx_handler_result - Possible return values for rx_handlers.
375 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
376 * further.
377 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
378 * case skb->dev was changed by rx_handler.
379 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
380 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
381 *
382 * rx_handlers are functions called from inside __netif_receive_skb(), to do
383 * special processing of the skb, prior to delivery to protocol handlers.
384 *
385 * Currently, a net_device can only have a single rx_handler registered. Trying
386 * to register a second rx_handler will return -EBUSY.
387 *
388 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
389 * To unregister a rx_handler on a net_device, use
390 * netdev_rx_handler_unregister().
391 *
392 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
393 * do with the skb.
394 *
395 * If the rx_handler consumed the skb in some way, it should return
396 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
397 * the skb to be delivered in some other way.
398 *
399 * If the rx_handler changed skb->dev, to divert the skb to another
400 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
401 * new device will be called if it exists.
402 *
403 * If the rx_handler decides the skb should be ignored, it should return
404 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
405 * are registered on exact device (ptype->dev == skb->dev).
406 *
407 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
408 * delivered, it should return RX_HANDLER_PASS.
409 *
410 * A device without a registered rx_handler will behave as if rx_handler
411 * returned RX_HANDLER_PASS.
412 */
413
414 enum rx_handler_result {
415 RX_HANDLER_CONSUMED,
416 RX_HANDLER_ANOTHER,
417 RX_HANDLER_EXACT,
418 RX_HANDLER_PASS,
419 };
420 typedef enum rx_handler_result rx_handler_result_t;
421 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
422
423 void __napi_schedule(struct napi_struct *n);
424 void __napi_schedule_irqoff(struct napi_struct *n);
425
426 static inline bool napi_disable_pending(struct napi_struct *n)
427 {
428 return test_bit(NAPI_STATE_DISABLE, &n->state);
429 }
430
431 bool napi_schedule_prep(struct napi_struct *n);
432
433 /**
434 * napi_schedule - schedule NAPI poll
435 * @n: NAPI context
436 *
437 * Schedule NAPI poll routine to be called if it is not already
438 * running.
439 */
440 static inline void napi_schedule(struct napi_struct *n)
441 {
442 if (napi_schedule_prep(n))
443 __napi_schedule(n);
444 }
445
446 /**
447 * napi_schedule_irqoff - schedule NAPI poll
448 * @n: NAPI context
449 *
450 * Variant of napi_schedule(), assuming hard irqs are masked.
451 */
452 static inline void napi_schedule_irqoff(struct napi_struct *n)
453 {
454 if (napi_schedule_prep(n))
455 __napi_schedule_irqoff(n);
456 }
457
458 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
459 static inline bool napi_reschedule(struct napi_struct *napi)
460 {
461 if (napi_schedule_prep(napi)) {
462 __napi_schedule(napi);
463 return true;
464 }
465 return false;
466 }
467
468 bool napi_complete_done(struct napi_struct *n, int work_done);
469 /**
470 * napi_complete - NAPI processing complete
471 * @n: NAPI context
472 *
473 * Mark NAPI processing as complete.
474 * Consider using napi_complete_done() instead.
475 * Return false if device should avoid rearming interrupts.
476 */
477 static inline bool napi_complete(struct napi_struct *n)
478 {
479 return napi_complete_done(n, 0);
480 }
481
482 /**
483 * napi_hash_del - remove a NAPI from global table
484 * @napi: NAPI context
485 *
486 * Warning: caller must observe RCU grace period
487 * before freeing memory containing @napi, if
488 * this function returns true.
489 * Note: core networking stack automatically calls it
490 * from netif_napi_del().
491 * Drivers might want to call this helper to combine all
492 * the needed RCU grace periods into a single one.
493 */
494 bool napi_hash_del(struct napi_struct *napi);
495
496 /**
497 * napi_disable - prevent NAPI from scheduling
498 * @n: NAPI context
499 *
500 * Stop NAPI from being scheduled on this context.
501 * Waits till any outstanding processing completes.
502 */
503 void napi_disable(struct napi_struct *n);
504
505 /**
506 * napi_enable - enable NAPI scheduling
507 * @n: NAPI context
508 *
509 * Resume NAPI from being scheduled on this context.
510 * Must be paired with napi_disable.
511 */
512 static inline void napi_enable(struct napi_struct *n)
513 {
514 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
515 smp_mb__before_atomic();
516 clear_bit(NAPI_STATE_SCHED, &n->state);
517 clear_bit(NAPI_STATE_NPSVC, &n->state);
518 }
519
520 /**
521 * napi_synchronize - wait until NAPI is not running
522 * @n: NAPI context
523 *
524 * Wait until NAPI is done being scheduled on this context.
525 * Waits till any outstanding processing completes but
526 * does not disable future activations.
527 */
528 static inline void napi_synchronize(const struct napi_struct *n)
529 {
530 if (IS_ENABLED(CONFIG_SMP))
531 while (test_bit(NAPI_STATE_SCHED, &n->state))
532 msleep(1);
533 else
534 barrier();
535 }
536
537 /**
538 * napi_if_scheduled_mark_missed - if napi is running, set the
539 * NAPIF_STATE_MISSED
540 * @n: NAPI context
541 *
542 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
543 * NAPI is scheduled.
544 **/
545 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
546 {
547 unsigned long val, new;
548
549 do {
550 val = READ_ONCE(n->state);
551 if (val & NAPIF_STATE_DISABLE)
552 return true;
553
554 if (!(val & NAPIF_STATE_SCHED))
555 return false;
556
557 new = val | NAPIF_STATE_MISSED;
558 } while (cmpxchg(&n->state, val, new) != val);
559
560 return true;
561 }
562
563 enum netdev_queue_state_t {
564 __QUEUE_STATE_DRV_XOFF,
565 __QUEUE_STATE_STACK_XOFF,
566 __QUEUE_STATE_FROZEN,
567 };
568
569 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
570 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
571 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
572
573 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
574 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
575 QUEUE_STATE_FROZEN)
576 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
577 QUEUE_STATE_FROZEN)
578
579 /*
580 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
581 * netif_tx_* functions below are used to manipulate this flag. The
582 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
583 * queue independently. The netif_xmit_*stopped functions below are called
584 * to check if the queue has been stopped by the driver or stack (either
585 * of the XOFF bits are set in the state). Drivers should not need to call
586 * netif_xmit*stopped functions, they should only be using netif_tx_*.
587 */
588
589 struct netdev_queue {
590 /*
591 * read-mostly part
592 */
593 struct net_device *dev;
594 struct Qdisc __rcu *qdisc;
595 struct Qdisc *qdisc_sleeping;
596 #ifdef CONFIG_SYSFS
597 struct kobject kobj;
598 #endif
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 int numa_node;
601 #endif
602 unsigned long tx_maxrate;
603 /*
604 * Number of TX timeouts for this queue
605 * (/sys/class/net/DEV/Q/trans_timeout)
606 */
607 unsigned long trans_timeout;
608
609 /* Subordinate device that the queue has been assigned to */
610 struct net_device *sb_dev;
611 #ifdef CONFIG_XDP_SOCKETS
612 struct xdp_umem *umem;
613 #endif
614 /*
615 * write-mostly part
616 */
617 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
618 int xmit_lock_owner;
619 /*
620 * Time (in jiffies) of last Tx
621 */
622 unsigned long trans_start;
623
624 unsigned long state;
625
626 #ifdef CONFIG_BQL
627 struct dql dql;
628 #endif
629 } ____cacheline_aligned_in_smp;
630
631 extern int sysctl_fb_tunnels_only_for_init_net;
632 extern int sysctl_devconf_inherit_init_net;
633
634 static inline bool net_has_fallback_tunnels(const struct net *net)
635 {
636 return net == &init_net ||
637 !IS_ENABLED(CONFIG_SYSCTL) ||
638 !sysctl_fb_tunnels_only_for_init_net;
639 }
640
641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
642 {
643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
644 return q->numa_node;
645 #else
646 return NUMA_NO_NODE;
647 #endif
648 }
649
650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
651 {
652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
653 q->numa_node = node;
654 #endif
655 }
656
657 #ifdef CONFIG_RPS
658 /*
659 * This structure holds an RPS map which can be of variable length. The
660 * map is an array of CPUs.
661 */
662 struct rps_map {
663 unsigned int len;
664 struct rcu_head rcu;
665 u16 cpus[0];
666 };
667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
668
669 /*
670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
671 * tail pointer for that CPU's input queue at the time of last enqueue, and
672 * a hardware filter index.
673 */
674 struct rps_dev_flow {
675 u16 cpu;
676 u16 filter;
677 unsigned int last_qtail;
678 };
679 #define RPS_NO_FILTER 0xffff
680
681 /*
682 * The rps_dev_flow_table structure contains a table of flow mappings.
683 */
684 struct rps_dev_flow_table {
685 unsigned int mask;
686 struct rcu_head rcu;
687 struct rps_dev_flow flows[0];
688 };
689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
690 ((_num) * sizeof(struct rps_dev_flow)))
691
692 /*
693 * The rps_sock_flow_table contains mappings of flows to the last CPU
694 * on which they were processed by the application (set in recvmsg).
695 * Each entry is a 32bit value. Upper part is the high-order bits
696 * of flow hash, lower part is CPU number.
697 * rps_cpu_mask is used to partition the space, depending on number of
698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
700 * meaning we use 32-6=26 bits for the hash.
701 */
702 struct rps_sock_flow_table {
703 u32 mask;
704
705 u32 ents[0] ____cacheline_aligned_in_smp;
706 };
707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
708
709 #define RPS_NO_CPU 0xffff
710
711 extern u32 rps_cpu_mask;
712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
713
714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
715 u32 hash)
716 {
717 if (table && hash) {
718 unsigned int index = hash & table->mask;
719 u32 val = hash & ~rps_cpu_mask;
720
721 /* We only give a hint, preemption can change CPU under us */
722 val |= raw_smp_processor_id();
723
724 if (table->ents[index] != val)
725 table->ents[index] = val;
726 }
727 }
728
729 #ifdef CONFIG_RFS_ACCEL
730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
731 u16 filter_id);
732 #endif
733 #endif /* CONFIG_RPS */
734
735 /* This structure contains an instance of an RX queue. */
736 struct netdev_rx_queue {
737 #ifdef CONFIG_RPS
738 struct rps_map __rcu *rps_map;
739 struct rps_dev_flow_table __rcu *rps_flow_table;
740 #endif
741 struct kobject kobj;
742 struct net_device *dev;
743 struct xdp_rxq_info xdp_rxq;
744 #ifdef CONFIG_XDP_SOCKETS
745 struct xdp_umem *umem;
746 #endif
747 } ____cacheline_aligned_in_smp;
748
749 /*
750 * RX queue sysfs structures and functions.
751 */
752 struct rx_queue_attribute {
753 struct attribute attr;
754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
755 ssize_t (*store)(struct netdev_rx_queue *queue,
756 const char *buf, size_t len);
757 };
758
759 #ifdef CONFIG_XPS
760 /*
761 * This structure holds an XPS map which can be of variable length. The
762 * map is an array of queues.
763 */
764 struct xps_map {
765 unsigned int len;
766 unsigned int alloc_len;
767 struct rcu_head rcu;
768 u16 queues[0];
769 };
770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
772 - sizeof(struct xps_map)) / sizeof(u16))
773
774 /*
775 * This structure holds all XPS maps for device. Maps are indexed by CPU.
776 */
777 struct xps_dev_maps {
778 struct rcu_head rcu;
779 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
780 };
781
782 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
784
785 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
786 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
787
788 #endif /* CONFIG_XPS */
789
790 #define TC_MAX_QUEUE 16
791 #define TC_BITMASK 15
792 /* HW offloaded queuing disciplines txq count and offset maps */
793 struct netdev_tc_txq {
794 u16 count;
795 u16 offset;
796 };
797
798 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
799 /*
800 * This structure is to hold information about the device
801 * configured to run FCoE protocol stack.
802 */
803 struct netdev_fcoe_hbainfo {
804 char manufacturer[64];
805 char serial_number[64];
806 char hardware_version[64];
807 char driver_version[64];
808 char optionrom_version[64];
809 char firmware_version[64];
810 char model[256];
811 char model_description[256];
812 };
813 #endif
814
815 #define MAX_PHYS_ITEM_ID_LEN 32
816
817 /* This structure holds a unique identifier to identify some
818 * physical item (port for example) used by a netdevice.
819 */
820 struct netdev_phys_item_id {
821 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
822 unsigned char id_len;
823 };
824
825 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
826 struct netdev_phys_item_id *b)
827 {
828 return a->id_len == b->id_len &&
829 memcmp(a->id, b->id, a->id_len) == 0;
830 }
831
832 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
833 struct sk_buff *skb,
834 struct net_device *sb_dev);
835
836 enum tc_setup_type {
837 TC_SETUP_QDISC_MQPRIO,
838 TC_SETUP_CLSU32,
839 TC_SETUP_CLSFLOWER,
840 TC_SETUP_CLSMATCHALL,
841 TC_SETUP_CLSBPF,
842 TC_SETUP_BLOCK,
843 TC_SETUP_QDISC_CBS,
844 TC_SETUP_QDISC_RED,
845 TC_SETUP_QDISC_PRIO,
846 TC_SETUP_QDISC_MQ,
847 TC_SETUP_QDISC_ETF,
848 TC_SETUP_ROOT_QDISC,
849 TC_SETUP_QDISC_GRED,
850 TC_SETUP_QDISC_TAPRIO,
851 TC_SETUP_FT,
852 };
853
854 /* These structures hold the attributes of bpf state that are being passed
855 * to the netdevice through the bpf op.
856 */
857 enum bpf_netdev_command {
858 /* Set or clear a bpf program used in the earliest stages of packet
859 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
860 * is responsible for calling bpf_prog_put on any old progs that are
861 * stored. In case of error, the callee need not release the new prog
862 * reference, but on success it takes ownership and must bpf_prog_put
863 * when it is no longer used.
864 */
865 XDP_SETUP_PROG,
866 XDP_SETUP_PROG_HW,
867 XDP_QUERY_PROG,
868 XDP_QUERY_PROG_HW,
869 /* BPF program for offload callbacks, invoked at program load time. */
870 BPF_OFFLOAD_MAP_ALLOC,
871 BPF_OFFLOAD_MAP_FREE,
872 XDP_SETUP_XSK_UMEM,
873 };
874
875 struct bpf_prog_offload_ops;
876 struct netlink_ext_ack;
877 struct xdp_umem;
878
879 struct netdev_bpf {
880 enum bpf_netdev_command command;
881 union {
882 /* XDP_SETUP_PROG */
883 struct {
884 u32 flags;
885 struct bpf_prog *prog;
886 struct netlink_ext_ack *extack;
887 };
888 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
889 struct {
890 u32 prog_id;
891 /* flags with which program was installed */
892 u32 prog_flags;
893 };
894 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
895 struct {
896 struct bpf_offloaded_map *offmap;
897 };
898 /* XDP_SETUP_XSK_UMEM */
899 struct {
900 struct xdp_umem *umem;
901 u16 queue_id;
902 } xsk;
903 };
904 };
905
906 /* Flags for ndo_xsk_wakeup. */
907 #define XDP_WAKEUP_RX (1 << 0)
908 #define XDP_WAKEUP_TX (1 << 1)
909
910 #ifdef CONFIG_XFRM_OFFLOAD
911 struct xfrmdev_ops {
912 int (*xdo_dev_state_add) (struct xfrm_state *x);
913 void (*xdo_dev_state_delete) (struct xfrm_state *x);
914 void (*xdo_dev_state_free) (struct xfrm_state *x);
915 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
916 struct xfrm_state *x);
917 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
918 };
919 #endif
920
921 struct dev_ifalias {
922 struct rcu_head rcuhead;
923 char ifalias[];
924 };
925
926 struct devlink;
927 struct tlsdev_ops;
928
929 struct netdev_name_node {
930 struct hlist_node hlist;
931 struct list_head list;
932 struct net_device *dev;
933 const char *name;
934 };
935
936 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
937 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
938
939 /*
940 * This structure defines the management hooks for network devices.
941 * The following hooks can be defined; unless noted otherwise, they are
942 * optional and can be filled with a null pointer.
943 *
944 * int (*ndo_init)(struct net_device *dev);
945 * This function is called once when a network device is registered.
946 * The network device can use this for any late stage initialization
947 * or semantic validation. It can fail with an error code which will
948 * be propagated back to register_netdev.
949 *
950 * void (*ndo_uninit)(struct net_device *dev);
951 * This function is called when device is unregistered or when registration
952 * fails. It is not called if init fails.
953 *
954 * int (*ndo_open)(struct net_device *dev);
955 * This function is called when a network device transitions to the up
956 * state.
957 *
958 * int (*ndo_stop)(struct net_device *dev);
959 * This function is called when a network device transitions to the down
960 * state.
961 *
962 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
963 * struct net_device *dev);
964 * Called when a packet needs to be transmitted.
965 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
966 * the queue before that can happen; it's for obsolete devices and weird
967 * corner cases, but the stack really does a non-trivial amount
968 * of useless work if you return NETDEV_TX_BUSY.
969 * Required; cannot be NULL.
970 *
971 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
972 * struct net_device *dev
973 * netdev_features_t features);
974 * Called by core transmit path to determine if device is capable of
975 * performing offload operations on a given packet. This is to give
976 * the device an opportunity to implement any restrictions that cannot
977 * be otherwise expressed by feature flags. The check is called with
978 * the set of features that the stack has calculated and it returns
979 * those the driver believes to be appropriate.
980 *
981 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
982 * struct net_device *sb_dev);
983 * Called to decide which queue to use when device supports multiple
984 * transmit queues.
985 *
986 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
987 * This function is called to allow device receiver to make
988 * changes to configuration when multicast or promiscuous is enabled.
989 *
990 * void (*ndo_set_rx_mode)(struct net_device *dev);
991 * This function is called device changes address list filtering.
992 * If driver handles unicast address filtering, it should set
993 * IFF_UNICAST_FLT in its priv_flags.
994 *
995 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
996 * This function is called when the Media Access Control address
997 * needs to be changed. If this interface is not defined, the
998 * MAC address can not be changed.
999 *
1000 * int (*ndo_validate_addr)(struct net_device *dev);
1001 * Test if Media Access Control address is valid for the device.
1002 *
1003 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1004 * Called when a user requests an ioctl which can't be handled by
1005 * the generic interface code. If not defined ioctls return
1006 * not supported error code.
1007 *
1008 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1009 * Used to set network devices bus interface parameters. This interface
1010 * is retained for legacy reasons; new devices should use the bus
1011 * interface (PCI) for low level management.
1012 *
1013 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1014 * Called when a user wants to change the Maximum Transfer Unit
1015 * of a device.
1016 *
1017 * void (*ndo_tx_timeout)(struct net_device *dev);
1018 * Callback used when the transmitter has not made any progress
1019 * for dev->watchdog ticks.
1020 *
1021 * void (*ndo_get_stats64)(struct net_device *dev,
1022 * struct rtnl_link_stats64 *storage);
1023 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1024 * Called when a user wants to get the network device usage
1025 * statistics. Drivers must do one of the following:
1026 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1027 * rtnl_link_stats64 structure passed by the caller.
1028 * 2. Define @ndo_get_stats to update a net_device_stats structure
1029 * (which should normally be dev->stats) and return a pointer to
1030 * it. The structure may be changed asynchronously only if each
1031 * field is written atomically.
1032 * 3. Update dev->stats asynchronously and atomically, and define
1033 * neither operation.
1034 *
1035 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1036 * Return true if this device supports offload stats of this attr_id.
1037 *
1038 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1039 * void *attr_data)
1040 * Get statistics for offload operations by attr_id. Write it into the
1041 * attr_data pointer.
1042 *
1043 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1044 * If device supports VLAN filtering this function is called when a
1045 * VLAN id is registered.
1046 *
1047 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1048 * If device supports VLAN filtering this function is called when a
1049 * VLAN id is unregistered.
1050 *
1051 * void (*ndo_poll_controller)(struct net_device *dev);
1052 *
1053 * SR-IOV management functions.
1054 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1055 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1056 * u8 qos, __be16 proto);
1057 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1058 * int max_tx_rate);
1059 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1060 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1061 * int (*ndo_get_vf_config)(struct net_device *dev,
1062 * int vf, struct ifla_vf_info *ivf);
1063 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1064 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1065 * struct nlattr *port[]);
1066 *
1067 * Enable or disable the VF ability to query its RSS Redirection Table and
1068 * Hash Key. This is needed since on some devices VF share this information
1069 * with PF and querying it may introduce a theoretical security risk.
1070 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1071 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1072 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1073 * void *type_data);
1074 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1075 * This is always called from the stack with the rtnl lock held and netif
1076 * tx queues stopped. This allows the netdevice to perform queue
1077 * management safely.
1078 *
1079 * Fiber Channel over Ethernet (FCoE) offload functions.
1080 * int (*ndo_fcoe_enable)(struct net_device *dev);
1081 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1082 * so the underlying device can perform whatever needed configuration or
1083 * initialization to support acceleration of FCoE traffic.
1084 *
1085 * int (*ndo_fcoe_disable)(struct net_device *dev);
1086 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1087 * so the underlying device can perform whatever needed clean-ups to
1088 * stop supporting acceleration of FCoE traffic.
1089 *
1090 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1091 * struct scatterlist *sgl, unsigned int sgc);
1092 * Called when the FCoE Initiator wants to initialize an I/O that
1093 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1094 * perform necessary setup and returns 1 to indicate the device is set up
1095 * successfully to perform DDP on this I/O, otherwise this returns 0.
1096 *
1097 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1098 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1099 * indicated by the FC exchange id 'xid', so the underlying device can
1100 * clean up and reuse resources for later DDP requests.
1101 *
1102 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1103 * struct scatterlist *sgl, unsigned int sgc);
1104 * Called when the FCoE Target wants to initialize an I/O that
1105 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1106 * perform necessary setup and returns 1 to indicate the device is set up
1107 * successfully to perform DDP on this I/O, otherwise this returns 0.
1108 *
1109 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1110 * struct netdev_fcoe_hbainfo *hbainfo);
1111 * Called when the FCoE Protocol stack wants information on the underlying
1112 * device. This information is utilized by the FCoE protocol stack to
1113 * register attributes with Fiber Channel management service as per the
1114 * FC-GS Fabric Device Management Information(FDMI) specification.
1115 *
1116 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1117 * Called when the underlying device wants to override default World Wide
1118 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1119 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1120 * protocol stack to use.
1121 *
1122 * RFS acceleration.
1123 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1124 * u16 rxq_index, u32 flow_id);
1125 * Set hardware filter for RFS. rxq_index is the target queue index;
1126 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1127 * Return the filter ID on success, or a negative error code.
1128 *
1129 * Slave management functions (for bridge, bonding, etc).
1130 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1131 * Called to make another netdev an underling.
1132 *
1133 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1134 * Called to release previously enslaved netdev.
1135 *
1136 * Feature/offload setting functions.
1137 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1138 * netdev_features_t features);
1139 * Adjusts the requested feature flags according to device-specific
1140 * constraints, and returns the resulting flags. Must not modify
1141 * the device state.
1142 *
1143 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1144 * Called to update device configuration to new features. Passed
1145 * feature set might be less than what was returned by ndo_fix_features()).
1146 * Must return >0 or -errno if it changed dev->features itself.
1147 *
1148 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1149 * struct net_device *dev,
1150 * const unsigned char *addr, u16 vid, u16 flags,
1151 * struct netlink_ext_ack *extack);
1152 * Adds an FDB entry to dev for addr.
1153 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1154 * struct net_device *dev,
1155 * const unsigned char *addr, u16 vid)
1156 * Deletes the FDB entry from dev coresponding to addr.
1157 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1158 * struct net_device *dev, struct net_device *filter_dev,
1159 * int *idx)
1160 * Used to add FDB entries to dump requests. Implementers should add
1161 * entries to skb and update idx with the number of entries.
1162 *
1163 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1164 * u16 flags, struct netlink_ext_ack *extack)
1165 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1166 * struct net_device *dev, u32 filter_mask,
1167 * int nlflags)
1168 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1169 * u16 flags);
1170 *
1171 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1172 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1173 * which do not represent real hardware may define this to allow their
1174 * userspace components to manage their virtual carrier state. Devices
1175 * that determine carrier state from physical hardware properties (eg
1176 * network cables) or protocol-dependent mechanisms (eg
1177 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1178 *
1179 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1180 * struct netdev_phys_item_id *ppid);
1181 * Called to get ID of physical port of this device. If driver does
1182 * not implement this, it is assumed that the hw is not able to have
1183 * multiple net devices on single physical port.
1184 *
1185 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1186 * struct netdev_phys_item_id *ppid)
1187 * Called to get the parent ID of the physical port of this device.
1188 *
1189 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1190 * struct udp_tunnel_info *ti);
1191 * Called by UDP tunnel to notify a driver about the UDP port and socket
1192 * address family that a UDP tunnel is listnening to. It is called only
1193 * when a new port starts listening. The operation is protected by the
1194 * RTNL.
1195 *
1196 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1197 * struct udp_tunnel_info *ti);
1198 * Called by UDP tunnel to notify the driver about a UDP port and socket
1199 * address family that the UDP tunnel is not listening to anymore. The
1200 * operation is protected by the RTNL.
1201 *
1202 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1203 * struct net_device *dev)
1204 * Called by upper layer devices to accelerate switching or other
1205 * station functionality into hardware. 'pdev is the lowerdev
1206 * to use for the offload and 'dev' is the net device that will
1207 * back the offload. Returns a pointer to the private structure
1208 * the upper layer will maintain.
1209 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1210 * Called by upper layer device to delete the station created
1211 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1212 * the station and priv is the structure returned by the add
1213 * operation.
1214 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1215 * int queue_index, u32 maxrate);
1216 * Called when a user wants to set a max-rate limitation of specific
1217 * TX queue.
1218 * int (*ndo_get_iflink)(const struct net_device *dev);
1219 * Called to get the iflink value of this device.
1220 * void (*ndo_change_proto_down)(struct net_device *dev,
1221 * bool proto_down);
1222 * This function is used to pass protocol port error state information
1223 * to the switch driver. The switch driver can react to the proto_down
1224 * by doing a phys down on the associated switch port.
1225 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1226 * This function is used to get egress tunnel information for given skb.
1227 * This is useful for retrieving outer tunnel header parameters while
1228 * sampling packet.
1229 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1230 * This function is used to specify the headroom that the skb must
1231 * consider when allocation skb during packet reception. Setting
1232 * appropriate rx headroom value allows avoiding skb head copy on
1233 * forward. Setting a negative value resets the rx headroom to the
1234 * default value.
1235 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1236 * This function is used to set or query state related to XDP on the
1237 * netdevice and manage BPF offload. See definition of
1238 * enum bpf_netdev_command for details.
1239 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1240 * u32 flags);
1241 * This function is used to submit @n XDP packets for transmit on a
1242 * netdevice. Returns number of frames successfully transmitted, frames
1243 * that got dropped are freed/returned via xdp_return_frame().
1244 * Returns negative number, means general error invoking ndo, meaning
1245 * no frames were xmit'ed and core-caller will free all frames.
1246 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1247 * This function is used to wake up the softirq, ksoftirqd or kthread
1248 * responsible for sending and/or receiving packets on a specific
1249 * queue id bound to an AF_XDP socket. The flags field specifies if
1250 * only RX, only Tx, or both should be woken up using the flags
1251 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1252 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1253 * Get devlink port instance associated with a given netdev.
1254 * Called with a reference on the netdevice and devlink locks only,
1255 * rtnl_lock is not held.
1256 */
1257 struct net_device_ops {
1258 int (*ndo_init)(struct net_device *dev);
1259 void (*ndo_uninit)(struct net_device *dev);
1260 int (*ndo_open)(struct net_device *dev);
1261 int (*ndo_stop)(struct net_device *dev);
1262 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1263 struct net_device *dev);
1264 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1265 struct net_device *dev,
1266 netdev_features_t features);
1267 u16 (*ndo_select_queue)(struct net_device *dev,
1268 struct sk_buff *skb,
1269 struct net_device *sb_dev);
1270 void (*ndo_change_rx_flags)(struct net_device *dev,
1271 int flags);
1272 void (*ndo_set_rx_mode)(struct net_device *dev);
1273 int (*ndo_set_mac_address)(struct net_device *dev,
1274 void *addr);
1275 int (*ndo_validate_addr)(struct net_device *dev);
1276 int (*ndo_do_ioctl)(struct net_device *dev,
1277 struct ifreq *ifr, int cmd);
1278 int (*ndo_set_config)(struct net_device *dev,
1279 struct ifmap *map);
1280 int (*ndo_change_mtu)(struct net_device *dev,
1281 int new_mtu);
1282 int (*ndo_neigh_setup)(struct net_device *dev,
1283 struct neigh_parms *);
1284 void (*ndo_tx_timeout) (struct net_device *dev);
1285
1286 void (*ndo_get_stats64)(struct net_device *dev,
1287 struct rtnl_link_stats64 *storage);
1288 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1289 int (*ndo_get_offload_stats)(int attr_id,
1290 const struct net_device *dev,
1291 void *attr_data);
1292 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1293
1294 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1295 __be16 proto, u16 vid);
1296 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1297 __be16 proto, u16 vid);
1298 #ifdef CONFIG_NET_POLL_CONTROLLER
1299 void (*ndo_poll_controller)(struct net_device *dev);
1300 int (*ndo_netpoll_setup)(struct net_device *dev,
1301 struct netpoll_info *info);
1302 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1303 #endif
1304 int (*ndo_set_vf_mac)(struct net_device *dev,
1305 int queue, u8 *mac);
1306 int (*ndo_set_vf_vlan)(struct net_device *dev,
1307 int queue, u16 vlan,
1308 u8 qos, __be16 proto);
1309 int (*ndo_set_vf_rate)(struct net_device *dev,
1310 int vf, int min_tx_rate,
1311 int max_tx_rate);
1312 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1313 int vf, bool setting);
1314 int (*ndo_set_vf_trust)(struct net_device *dev,
1315 int vf, bool setting);
1316 int (*ndo_get_vf_config)(struct net_device *dev,
1317 int vf,
1318 struct ifla_vf_info *ivf);
1319 int (*ndo_set_vf_link_state)(struct net_device *dev,
1320 int vf, int link_state);
1321 int (*ndo_get_vf_stats)(struct net_device *dev,
1322 int vf,
1323 struct ifla_vf_stats
1324 *vf_stats);
1325 int (*ndo_set_vf_port)(struct net_device *dev,
1326 int vf,
1327 struct nlattr *port[]);
1328 int (*ndo_get_vf_port)(struct net_device *dev,
1329 int vf, struct sk_buff *skb);
1330 int (*ndo_set_vf_guid)(struct net_device *dev,
1331 int vf, u64 guid,
1332 int guid_type);
1333 int (*ndo_set_vf_rss_query_en)(
1334 struct net_device *dev,
1335 int vf, bool setting);
1336 int (*ndo_setup_tc)(struct net_device *dev,
1337 enum tc_setup_type type,
1338 void *type_data);
1339 #if IS_ENABLED(CONFIG_FCOE)
1340 int (*ndo_fcoe_enable)(struct net_device *dev);
1341 int (*ndo_fcoe_disable)(struct net_device *dev);
1342 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1343 u16 xid,
1344 struct scatterlist *sgl,
1345 unsigned int sgc);
1346 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1347 u16 xid);
1348 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1349 u16 xid,
1350 struct scatterlist *sgl,
1351 unsigned int sgc);
1352 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1353 struct netdev_fcoe_hbainfo *hbainfo);
1354 #endif
1355
1356 #if IS_ENABLED(CONFIG_LIBFCOE)
1357 #define NETDEV_FCOE_WWNN 0
1358 #define NETDEV_FCOE_WWPN 1
1359 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1360 u64 *wwn, int type);
1361 #endif
1362
1363 #ifdef CONFIG_RFS_ACCEL
1364 int (*ndo_rx_flow_steer)(struct net_device *dev,
1365 const struct sk_buff *skb,
1366 u16 rxq_index,
1367 u32 flow_id);
1368 #endif
1369 int (*ndo_add_slave)(struct net_device *dev,
1370 struct net_device *slave_dev,
1371 struct netlink_ext_ack *extack);
1372 int (*ndo_del_slave)(struct net_device *dev,
1373 struct net_device *slave_dev);
1374 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1375 netdev_features_t features);
1376 int (*ndo_set_features)(struct net_device *dev,
1377 netdev_features_t features);
1378 int (*ndo_neigh_construct)(struct net_device *dev,
1379 struct neighbour *n);
1380 void (*ndo_neigh_destroy)(struct net_device *dev,
1381 struct neighbour *n);
1382
1383 int (*ndo_fdb_add)(struct ndmsg *ndm,
1384 struct nlattr *tb[],
1385 struct net_device *dev,
1386 const unsigned char *addr,
1387 u16 vid,
1388 u16 flags,
1389 struct netlink_ext_ack *extack);
1390 int (*ndo_fdb_del)(struct ndmsg *ndm,
1391 struct nlattr *tb[],
1392 struct net_device *dev,
1393 const unsigned char *addr,
1394 u16 vid);
1395 int (*ndo_fdb_dump)(struct sk_buff *skb,
1396 struct netlink_callback *cb,
1397 struct net_device *dev,
1398 struct net_device *filter_dev,
1399 int *idx);
1400 int (*ndo_fdb_get)(struct sk_buff *skb,
1401 struct nlattr *tb[],
1402 struct net_device *dev,
1403 const unsigned char *addr,
1404 u16 vid, u32 portid, u32 seq,
1405 struct netlink_ext_ack *extack);
1406 int (*ndo_bridge_setlink)(struct net_device *dev,
1407 struct nlmsghdr *nlh,
1408 u16 flags,
1409 struct netlink_ext_ack *extack);
1410 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1411 u32 pid, u32 seq,
1412 struct net_device *dev,
1413 u32 filter_mask,
1414 int nlflags);
1415 int (*ndo_bridge_dellink)(struct net_device *dev,
1416 struct nlmsghdr *nlh,
1417 u16 flags);
1418 int (*ndo_change_carrier)(struct net_device *dev,
1419 bool new_carrier);
1420 int (*ndo_get_phys_port_id)(struct net_device *dev,
1421 struct netdev_phys_item_id *ppid);
1422 int (*ndo_get_port_parent_id)(struct net_device *dev,
1423 struct netdev_phys_item_id *ppid);
1424 int (*ndo_get_phys_port_name)(struct net_device *dev,
1425 char *name, size_t len);
1426 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1427 struct udp_tunnel_info *ti);
1428 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1429 struct udp_tunnel_info *ti);
1430 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1431 struct net_device *dev);
1432 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1433 void *priv);
1434
1435 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1436 int queue_index,
1437 u32 maxrate);
1438 int (*ndo_get_iflink)(const struct net_device *dev);
1439 int (*ndo_change_proto_down)(struct net_device *dev,
1440 bool proto_down);
1441 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1442 struct sk_buff *skb);
1443 void (*ndo_set_rx_headroom)(struct net_device *dev,
1444 int needed_headroom);
1445 int (*ndo_bpf)(struct net_device *dev,
1446 struct netdev_bpf *bpf);
1447 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1448 struct xdp_frame **xdp,
1449 u32 flags);
1450 int (*ndo_xsk_wakeup)(struct net_device *dev,
1451 u32 queue_id, u32 flags);
1452 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1453 };
1454
1455 /**
1456 * enum net_device_priv_flags - &struct net_device priv_flags
1457 *
1458 * These are the &struct net_device, they are only set internally
1459 * by drivers and used in the kernel. These flags are invisible to
1460 * userspace; this means that the order of these flags can change
1461 * during any kernel release.
1462 *
1463 * You should have a pretty good reason to be extending these flags.
1464 *
1465 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1466 * @IFF_EBRIDGE: Ethernet bridging device
1467 * @IFF_BONDING: bonding master or slave
1468 * @IFF_ISATAP: ISATAP interface (RFC4214)
1469 * @IFF_WAN_HDLC: WAN HDLC device
1470 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1471 * release skb->dst
1472 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1473 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1474 * @IFF_MACVLAN_PORT: device used as macvlan port
1475 * @IFF_BRIDGE_PORT: device used as bridge port
1476 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1477 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1478 * @IFF_UNICAST_FLT: Supports unicast filtering
1479 * @IFF_TEAM_PORT: device used as team port
1480 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1481 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1482 * change when it's running
1483 * @IFF_MACVLAN: Macvlan device
1484 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1485 * underlying stacked devices
1486 * @IFF_L3MDEV_MASTER: device is an L3 master device
1487 * @IFF_NO_QUEUE: device can run without qdisc attached
1488 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1489 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1490 * @IFF_TEAM: device is a team device
1491 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1492 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1493 * entity (i.e. the master device for bridged veth)
1494 * @IFF_MACSEC: device is a MACsec device
1495 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1496 * @IFF_FAILOVER: device is a failover master device
1497 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1498 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1499 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1500 */
1501 enum netdev_priv_flags {
1502 IFF_802_1Q_VLAN = 1<<0,
1503 IFF_EBRIDGE = 1<<1,
1504 IFF_BONDING = 1<<2,
1505 IFF_ISATAP = 1<<3,
1506 IFF_WAN_HDLC = 1<<4,
1507 IFF_XMIT_DST_RELEASE = 1<<5,
1508 IFF_DONT_BRIDGE = 1<<6,
1509 IFF_DISABLE_NETPOLL = 1<<7,
1510 IFF_MACVLAN_PORT = 1<<8,
1511 IFF_BRIDGE_PORT = 1<<9,
1512 IFF_OVS_DATAPATH = 1<<10,
1513 IFF_TX_SKB_SHARING = 1<<11,
1514 IFF_UNICAST_FLT = 1<<12,
1515 IFF_TEAM_PORT = 1<<13,
1516 IFF_SUPP_NOFCS = 1<<14,
1517 IFF_LIVE_ADDR_CHANGE = 1<<15,
1518 IFF_MACVLAN = 1<<16,
1519 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1520 IFF_L3MDEV_MASTER = 1<<18,
1521 IFF_NO_QUEUE = 1<<19,
1522 IFF_OPENVSWITCH = 1<<20,
1523 IFF_L3MDEV_SLAVE = 1<<21,
1524 IFF_TEAM = 1<<22,
1525 IFF_RXFH_CONFIGURED = 1<<23,
1526 IFF_PHONY_HEADROOM = 1<<24,
1527 IFF_MACSEC = 1<<25,
1528 IFF_NO_RX_HANDLER = 1<<26,
1529 IFF_FAILOVER = 1<<27,
1530 IFF_FAILOVER_SLAVE = 1<<28,
1531 IFF_L3MDEV_RX_HANDLER = 1<<29,
1532 IFF_LIVE_RENAME_OK = 1<<30,
1533 };
1534
1535 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1536 #define IFF_EBRIDGE IFF_EBRIDGE
1537 #define IFF_BONDING IFF_BONDING
1538 #define IFF_ISATAP IFF_ISATAP
1539 #define IFF_WAN_HDLC IFF_WAN_HDLC
1540 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1541 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1542 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1543 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1544 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1545 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1546 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1547 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1548 #define IFF_TEAM_PORT IFF_TEAM_PORT
1549 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1550 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1551 #define IFF_MACVLAN IFF_MACVLAN
1552 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1553 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1554 #define IFF_NO_QUEUE IFF_NO_QUEUE
1555 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1556 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1557 #define IFF_TEAM IFF_TEAM
1558 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1559 #define IFF_MACSEC IFF_MACSEC
1560 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1561 #define IFF_FAILOVER IFF_FAILOVER
1562 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1563 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1564 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1565
1566 /**
1567 * struct net_device - The DEVICE structure.
1568 *
1569 * Actually, this whole structure is a big mistake. It mixes I/O
1570 * data with strictly "high-level" data, and it has to know about
1571 * almost every data structure used in the INET module.
1572 *
1573 * @name: This is the first field of the "visible" part of this structure
1574 * (i.e. as seen by users in the "Space.c" file). It is the name
1575 * of the interface.
1576 *
1577 * @name_node: Name hashlist node
1578 * @ifalias: SNMP alias
1579 * @mem_end: Shared memory end
1580 * @mem_start: Shared memory start
1581 * @base_addr: Device I/O address
1582 * @irq: Device IRQ number
1583 *
1584 * @state: Generic network queuing layer state, see netdev_state_t
1585 * @dev_list: The global list of network devices
1586 * @napi_list: List entry used for polling NAPI devices
1587 * @unreg_list: List entry when we are unregistering the
1588 * device; see the function unregister_netdev
1589 * @close_list: List entry used when we are closing the device
1590 * @ptype_all: Device-specific packet handlers for all protocols
1591 * @ptype_specific: Device-specific, protocol-specific packet handlers
1592 *
1593 * @adj_list: Directly linked devices, like slaves for bonding
1594 * @features: Currently active device features
1595 * @hw_features: User-changeable features
1596 *
1597 * @wanted_features: User-requested features
1598 * @vlan_features: Mask of features inheritable by VLAN devices
1599 *
1600 * @hw_enc_features: Mask of features inherited by encapsulating devices
1601 * This field indicates what encapsulation
1602 * offloads the hardware is capable of doing,
1603 * and drivers will need to set them appropriately.
1604 *
1605 * @mpls_features: Mask of features inheritable by MPLS
1606 *
1607 * @ifindex: interface index
1608 * @group: The group the device belongs to
1609 *
1610 * @stats: Statistics struct, which was left as a legacy, use
1611 * rtnl_link_stats64 instead
1612 *
1613 * @rx_dropped: Dropped packets by core network,
1614 * do not use this in drivers
1615 * @tx_dropped: Dropped packets by core network,
1616 * do not use this in drivers
1617 * @rx_nohandler: nohandler dropped packets by core network on
1618 * inactive devices, do not use this in drivers
1619 * @carrier_up_count: Number of times the carrier has been up
1620 * @carrier_down_count: Number of times the carrier has been down
1621 *
1622 * @wireless_handlers: List of functions to handle Wireless Extensions,
1623 * instead of ioctl,
1624 * see <net/iw_handler.h> for details.
1625 * @wireless_data: Instance data managed by the core of wireless extensions
1626 *
1627 * @netdev_ops: Includes several pointers to callbacks,
1628 * if one wants to override the ndo_*() functions
1629 * @ethtool_ops: Management operations
1630 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1631 * discovery handling. Necessary for e.g. 6LoWPAN.
1632 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1633 * of Layer 2 headers.
1634 *
1635 * @flags: Interface flags (a la BSD)
1636 * @priv_flags: Like 'flags' but invisible to userspace,
1637 * see if.h for the definitions
1638 * @gflags: Global flags ( kept as legacy )
1639 * @padded: How much padding added by alloc_netdev()
1640 * @operstate: RFC2863 operstate
1641 * @link_mode: Mapping policy to operstate
1642 * @if_port: Selectable AUI, TP, ...
1643 * @dma: DMA channel
1644 * @mtu: Interface MTU value
1645 * @min_mtu: Interface Minimum MTU value
1646 * @max_mtu: Interface Maximum MTU value
1647 * @type: Interface hardware type
1648 * @hard_header_len: Maximum hardware header length.
1649 * @min_header_len: Minimum hardware header length
1650 *
1651 * @needed_headroom: Extra headroom the hardware may need, but not in all
1652 * cases can this be guaranteed
1653 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1654 * cases can this be guaranteed. Some cases also use
1655 * LL_MAX_HEADER instead to allocate the skb
1656 *
1657 * interface address info:
1658 *
1659 * @perm_addr: Permanent hw address
1660 * @addr_assign_type: Hw address assignment type
1661 * @addr_len: Hardware address length
1662 * @upper_level: Maximum depth level of upper devices.
1663 * @lower_level: Maximum depth level of lower devices.
1664 * @neigh_priv_len: Used in neigh_alloc()
1665 * @dev_id: Used to differentiate devices that share
1666 * the same link layer address
1667 * @dev_port: Used to differentiate devices that share
1668 * the same function
1669 * @addr_list_lock: XXX: need comments on this one
1670 * @uc_promisc: Counter that indicates promiscuous mode
1671 * has been enabled due to the need to listen to
1672 * additional unicast addresses in a device that
1673 * does not implement ndo_set_rx_mode()
1674 * @uc: unicast mac addresses
1675 * @mc: multicast mac addresses
1676 * @dev_addrs: list of device hw addresses
1677 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1678 * @promiscuity: Number of times the NIC is told to work in
1679 * promiscuous mode; if it becomes 0 the NIC will
1680 * exit promiscuous mode
1681 * @allmulti: Counter, enables or disables allmulticast mode
1682 *
1683 * @vlan_info: VLAN info
1684 * @dsa_ptr: dsa specific data
1685 * @tipc_ptr: TIPC specific data
1686 * @atalk_ptr: AppleTalk link
1687 * @ip_ptr: IPv4 specific data
1688 * @dn_ptr: DECnet specific data
1689 * @ip6_ptr: IPv6 specific data
1690 * @ax25_ptr: AX.25 specific data
1691 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1692 *
1693 * @dev_addr: Hw address (before bcast,
1694 * because most packets are unicast)
1695 *
1696 * @_rx: Array of RX queues
1697 * @num_rx_queues: Number of RX queues
1698 * allocated at register_netdev() time
1699 * @real_num_rx_queues: Number of RX queues currently active in device
1700 *
1701 * @rx_handler: handler for received packets
1702 * @rx_handler_data: XXX: need comments on this one
1703 * @miniq_ingress: ingress/clsact qdisc specific data for
1704 * ingress processing
1705 * @ingress_queue: XXX: need comments on this one
1706 * @broadcast: hw bcast address
1707 *
1708 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1709 * indexed by RX queue number. Assigned by driver.
1710 * This must only be set if the ndo_rx_flow_steer
1711 * operation is defined
1712 * @index_hlist: Device index hash chain
1713 *
1714 * @_tx: Array of TX queues
1715 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1716 * @real_num_tx_queues: Number of TX queues currently active in device
1717 * @qdisc: Root qdisc from userspace point of view
1718 * @tx_queue_len: Max frames per queue allowed
1719 * @tx_global_lock: XXX: need comments on this one
1720 *
1721 * @xps_maps: XXX: need comments on this one
1722 * @miniq_egress: clsact qdisc specific data for
1723 * egress processing
1724 * @watchdog_timeo: Represents the timeout that is used by
1725 * the watchdog (see dev_watchdog())
1726 * @watchdog_timer: List of timers
1727 *
1728 * @pcpu_refcnt: Number of references to this device
1729 * @todo_list: Delayed register/unregister
1730 * @link_watch_list: XXX: need comments on this one
1731 *
1732 * @reg_state: Register/unregister state machine
1733 * @dismantle: Device is going to be freed
1734 * @rtnl_link_state: This enum represents the phases of creating
1735 * a new link
1736 *
1737 * @needs_free_netdev: Should unregister perform free_netdev?
1738 * @priv_destructor: Called from unregister
1739 * @npinfo: XXX: need comments on this one
1740 * @nd_net: Network namespace this network device is inside
1741 *
1742 * @ml_priv: Mid-layer private
1743 * @lstats: Loopback statistics
1744 * @tstats: Tunnel statistics
1745 * @dstats: Dummy statistics
1746 * @vstats: Virtual ethernet statistics
1747 *
1748 * @garp_port: GARP
1749 * @mrp_port: MRP
1750 *
1751 * @dev: Class/net/name entry
1752 * @sysfs_groups: Space for optional device, statistics and wireless
1753 * sysfs groups
1754 *
1755 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1756 * @rtnl_link_ops: Rtnl_link_ops
1757 *
1758 * @gso_max_size: Maximum size of generic segmentation offload
1759 * @gso_max_segs: Maximum number of segments that can be passed to the
1760 * NIC for GSO
1761 *
1762 * @dcbnl_ops: Data Center Bridging netlink ops
1763 * @num_tc: Number of traffic classes in the net device
1764 * @tc_to_txq: XXX: need comments on this one
1765 * @prio_tc_map: XXX: need comments on this one
1766 *
1767 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1768 *
1769 * @priomap: XXX: need comments on this one
1770 * @phydev: Physical device may attach itself
1771 * for hardware timestamping
1772 * @sfp_bus: attached &struct sfp_bus structure.
1773 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock
1774 spinlock
1775 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1776 * @qdisc_xmit_lock_key: lockdep class annotating
1777 * netdev_queue->_xmit_lock spinlock
1778 * @addr_list_lock_key: lockdep class annotating
1779 * net_device->addr_list_lock spinlock
1780 *
1781 * @proto_down: protocol port state information can be sent to the
1782 * switch driver and used to set the phys state of the
1783 * switch port.
1784 *
1785 * @wol_enabled: Wake-on-LAN is enabled
1786 *
1787 * FIXME: cleanup struct net_device such that network protocol info
1788 * moves out.
1789 */
1790
1791 struct net_device {
1792 char name[IFNAMSIZ];
1793 struct netdev_name_node *name_node;
1794 struct dev_ifalias __rcu *ifalias;
1795 /*
1796 * I/O specific fields
1797 * FIXME: Merge these and struct ifmap into one
1798 */
1799 unsigned long mem_end;
1800 unsigned long mem_start;
1801 unsigned long base_addr;
1802 int irq;
1803
1804 /*
1805 * Some hardware also needs these fields (state,dev_list,
1806 * napi_list,unreg_list,close_list) but they are not
1807 * part of the usual set specified in Space.c.
1808 */
1809
1810 unsigned long state;
1811
1812 struct list_head dev_list;
1813 struct list_head napi_list;
1814 struct list_head unreg_list;
1815 struct list_head close_list;
1816 struct list_head ptype_all;
1817 struct list_head ptype_specific;
1818
1819 struct {
1820 struct list_head upper;
1821 struct list_head lower;
1822 } adj_list;
1823
1824 netdev_features_t features;
1825 netdev_features_t hw_features;
1826 netdev_features_t wanted_features;
1827 netdev_features_t vlan_features;
1828 netdev_features_t hw_enc_features;
1829 netdev_features_t mpls_features;
1830 netdev_features_t gso_partial_features;
1831
1832 int ifindex;
1833 int group;
1834
1835 struct net_device_stats stats;
1836
1837 atomic_long_t rx_dropped;
1838 atomic_long_t tx_dropped;
1839 atomic_long_t rx_nohandler;
1840
1841 /* Stats to monitor link on/off, flapping */
1842 atomic_t carrier_up_count;
1843 atomic_t carrier_down_count;
1844
1845 #ifdef CONFIG_WIRELESS_EXT
1846 const struct iw_handler_def *wireless_handlers;
1847 struct iw_public_data *wireless_data;
1848 #endif
1849 const struct net_device_ops *netdev_ops;
1850 const struct ethtool_ops *ethtool_ops;
1851 #ifdef CONFIG_NET_L3_MASTER_DEV
1852 const struct l3mdev_ops *l3mdev_ops;
1853 #endif
1854 #if IS_ENABLED(CONFIG_IPV6)
1855 const struct ndisc_ops *ndisc_ops;
1856 #endif
1857
1858 #ifdef CONFIG_XFRM_OFFLOAD
1859 const struct xfrmdev_ops *xfrmdev_ops;
1860 #endif
1861
1862 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1863 const struct tlsdev_ops *tlsdev_ops;
1864 #endif
1865
1866 const struct header_ops *header_ops;
1867
1868 unsigned int flags;
1869 unsigned int priv_flags;
1870
1871 unsigned short gflags;
1872 unsigned short padded;
1873
1874 unsigned char operstate;
1875 unsigned char link_mode;
1876
1877 unsigned char if_port;
1878 unsigned char dma;
1879
1880 unsigned int mtu;
1881 unsigned int min_mtu;
1882 unsigned int max_mtu;
1883 unsigned short type;
1884 unsigned short hard_header_len;
1885 unsigned char min_header_len;
1886
1887 unsigned short needed_headroom;
1888 unsigned short needed_tailroom;
1889
1890 /* Interface address info. */
1891 unsigned char perm_addr[MAX_ADDR_LEN];
1892 unsigned char addr_assign_type;
1893 unsigned char addr_len;
1894 unsigned char upper_level;
1895 unsigned char lower_level;
1896 unsigned short neigh_priv_len;
1897 unsigned short dev_id;
1898 unsigned short dev_port;
1899 spinlock_t addr_list_lock;
1900 unsigned char name_assign_type;
1901 bool uc_promisc;
1902 struct netdev_hw_addr_list uc;
1903 struct netdev_hw_addr_list mc;
1904 struct netdev_hw_addr_list dev_addrs;
1905
1906 #ifdef CONFIG_SYSFS
1907 struct kset *queues_kset;
1908 #endif
1909 unsigned int promiscuity;
1910 unsigned int allmulti;
1911
1912
1913 /* Protocol-specific pointers */
1914
1915 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1916 struct vlan_info __rcu *vlan_info;
1917 #endif
1918 #if IS_ENABLED(CONFIG_NET_DSA)
1919 struct dsa_port *dsa_ptr;
1920 #endif
1921 #if IS_ENABLED(CONFIG_TIPC)
1922 struct tipc_bearer __rcu *tipc_ptr;
1923 #endif
1924 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1925 void *atalk_ptr;
1926 #endif
1927 struct in_device __rcu *ip_ptr;
1928 #if IS_ENABLED(CONFIG_DECNET)
1929 struct dn_dev __rcu *dn_ptr;
1930 #endif
1931 struct inet6_dev __rcu *ip6_ptr;
1932 #if IS_ENABLED(CONFIG_AX25)
1933 void *ax25_ptr;
1934 #endif
1935 struct wireless_dev *ieee80211_ptr;
1936 struct wpan_dev *ieee802154_ptr;
1937 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1938 struct mpls_dev __rcu *mpls_ptr;
1939 #endif
1940
1941 /*
1942 * Cache lines mostly used on receive path (including eth_type_trans())
1943 */
1944 /* Interface address info used in eth_type_trans() */
1945 unsigned char *dev_addr;
1946
1947 struct netdev_rx_queue *_rx;
1948 unsigned int num_rx_queues;
1949 unsigned int real_num_rx_queues;
1950
1951 struct bpf_prog __rcu *xdp_prog;
1952 unsigned long gro_flush_timeout;
1953 rx_handler_func_t __rcu *rx_handler;
1954 void __rcu *rx_handler_data;
1955
1956 #ifdef CONFIG_NET_CLS_ACT
1957 struct mini_Qdisc __rcu *miniq_ingress;
1958 #endif
1959 struct netdev_queue __rcu *ingress_queue;
1960 #ifdef CONFIG_NETFILTER_INGRESS
1961 struct nf_hook_entries __rcu *nf_hooks_ingress;
1962 #endif
1963
1964 unsigned char broadcast[MAX_ADDR_LEN];
1965 #ifdef CONFIG_RFS_ACCEL
1966 struct cpu_rmap *rx_cpu_rmap;
1967 #endif
1968 struct hlist_node index_hlist;
1969
1970 /*
1971 * Cache lines mostly used on transmit path
1972 */
1973 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1974 unsigned int num_tx_queues;
1975 unsigned int real_num_tx_queues;
1976 struct Qdisc *qdisc;
1977 #ifdef CONFIG_NET_SCHED
1978 DECLARE_HASHTABLE (qdisc_hash, 4);
1979 #endif
1980 unsigned int tx_queue_len;
1981 spinlock_t tx_global_lock;
1982 int watchdog_timeo;
1983
1984 #ifdef CONFIG_XPS
1985 struct xps_dev_maps __rcu *xps_cpus_map;
1986 struct xps_dev_maps __rcu *xps_rxqs_map;
1987 #endif
1988 #ifdef CONFIG_NET_CLS_ACT
1989 struct mini_Qdisc __rcu *miniq_egress;
1990 #endif
1991
1992 /* These may be needed for future network-power-down code. */
1993 struct timer_list watchdog_timer;
1994
1995 int __percpu *pcpu_refcnt;
1996 struct list_head todo_list;
1997
1998 struct list_head link_watch_list;
1999
2000 enum { NETREG_UNINITIALIZED=0,
2001 NETREG_REGISTERED, /* completed register_netdevice */
2002 NETREG_UNREGISTERING, /* called unregister_netdevice */
2003 NETREG_UNREGISTERED, /* completed unregister todo */
2004 NETREG_RELEASED, /* called free_netdev */
2005 NETREG_DUMMY, /* dummy device for NAPI poll */
2006 } reg_state:8;
2007
2008 bool dismantle;
2009
2010 enum {
2011 RTNL_LINK_INITIALIZED,
2012 RTNL_LINK_INITIALIZING,
2013 } rtnl_link_state:16;
2014
2015 bool needs_free_netdev;
2016 void (*priv_destructor)(struct net_device *dev);
2017
2018 #ifdef CONFIG_NETPOLL
2019 struct netpoll_info __rcu *npinfo;
2020 #endif
2021
2022 possible_net_t nd_net;
2023
2024 /* mid-layer private */
2025 union {
2026 void *ml_priv;
2027 struct pcpu_lstats __percpu *lstats;
2028 struct pcpu_sw_netstats __percpu *tstats;
2029 struct pcpu_dstats __percpu *dstats;
2030 };
2031
2032 #if IS_ENABLED(CONFIG_GARP)
2033 struct garp_port __rcu *garp_port;
2034 #endif
2035 #if IS_ENABLED(CONFIG_MRP)
2036 struct mrp_port __rcu *mrp_port;
2037 #endif
2038
2039 struct device dev;
2040 const struct attribute_group *sysfs_groups[4];
2041 const struct attribute_group *sysfs_rx_queue_group;
2042
2043 const struct rtnl_link_ops *rtnl_link_ops;
2044
2045 /* for setting kernel sock attribute on TCP connection setup */
2046 #define GSO_MAX_SIZE 65536
2047 unsigned int gso_max_size;
2048 #define GSO_MAX_SEGS 65535
2049 u16 gso_max_segs;
2050
2051 #ifdef CONFIG_DCB
2052 const struct dcbnl_rtnl_ops *dcbnl_ops;
2053 #endif
2054 s16 num_tc;
2055 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2056 u8 prio_tc_map[TC_BITMASK + 1];
2057
2058 #if IS_ENABLED(CONFIG_FCOE)
2059 unsigned int fcoe_ddp_xid;
2060 #endif
2061 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2062 struct netprio_map __rcu *priomap;
2063 #endif
2064 struct phy_device *phydev;
2065 struct sfp_bus *sfp_bus;
2066 struct lock_class_key qdisc_tx_busylock_key;
2067 struct lock_class_key qdisc_running_key;
2068 struct lock_class_key qdisc_xmit_lock_key;
2069 struct lock_class_key addr_list_lock_key;
2070 bool proto_down;
2071 unsigned wol_enabled:1;
2072 };
2073 #define to_net_dev(d) container_of(d, struct net_device, dev)
2074
2075 static inline bool netif_elide_gro(const struct net_device *dev)
2076 {
2077 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2078 return true;
2079 return false;
2080 }
2081
2082 #define NETDEV_ALIGN 32
2083
2084 static inline
2085 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2086 {
2087 return dev->prio_tc_map[prio & TC_BITMASK];
2088 }
2089
2090 static inline
2091 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2092 {
2093 if (tc >= dev->num_tc)
2094 return -EINVAL;
2095
2096 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2097 return 0;
2098 }
2099
2100 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2101 void netdev_reset_tc(struct net_device *dev);
2102 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2103 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2104
2105 static inline
2106 int netdev_get_num_tc(struct net_device *dev)
2107 {
2108 return dev->num_tc;
2109 }
2110
2111 void netdev_unbind_sb_channel(struct net_device *dev,
2112 struct net_device *sb_dev);
2113 int netdev_bind_sb_channel_queue(struct net_device *dev,
2114 struct net_device *sb_dev,
2115 u8 tc, u16 count, u16 offset);
2116 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2117 static inline int netdev_get_sb_channel(struct net_device *dev)
2118 {
2119 return max_t(int, -dev->num_tc, 0);
2120 }
2121
2122 static inline
2123 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2124 unsigned int index)
2125 {
2126 return &dev->_tx[index];
2127 }
2128
2129 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2130 const struct sk_buff *skb)
2131 {
2132 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2133 }
2134
2135 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2136 void (*f)(struct net_device *,
2137 struct netdev_queue *,
2138 void *),
2139 void *arg)
2140 {
2141 unsigned int i;
2142
2143 for (i = 0; i < dev->num_tx_queues; i++)
2144 f(dev, &dev->_tx[i], arg);
2145 }
2146
2147 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2148 struct net_device *sb_dev);
2149 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2150 struct sk_buff *skb,
2151 struct net_device *sb_dev);
2152
2153 /* returns the headroom that the master device needs to take in account
2154 * when forwarding to this dev
2155 */
2156 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2157 {
2158 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2159 }
2160
2161 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2162 {
2163 if (dev->netdev_ops->ndo_set_rx_headroom)
2164 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2165 }
2166
2167 /* set the device rx headroom to the dev's default */
2168 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2169 {
2170 netdev_set_rx_headroom(dev, -1);
2171 }
2172
2173 /*
2174 * Net namespace inlines
2175 */
2176 static inline
2177 struct net *dev_net(const struct net_device *dev)
2178 {
2179 return read_pnet(&dev->nd_net);
2180 }
2181
2182 static inline
2183 void dev_net_set(struct net_device *dev, struct net *net)
2184 {
2185 write_pnet(&dev->nd_net, net);
2186 }
2187
2188 /**
2189 * netdev_priv - access network device private data
2190 * @dev: network device
2191 *
2192 * Get network device private data
2193 */
2194 static inline void *netdev_priv(const struct net_device *dev)
2195 {
2196 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2197 }
2198
2199 /* Set the sysfs physical device reference for the network logical device
2200 * if set prior to registration will cause a symlink during initialization.
2201 */
2202 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2203
2204 /* Set the sysfs device type for the network logical device to allow
2205 * fine-grained identification of different network device types. For
2206 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2207 */
2208 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2209
2210 /* Default NAPI poll() weight
2211 * Device drivers are strongly advised to not use bigger value
2212 */
2213 #define NAPI_POLL_WEIGHT 64
2214
2215 /**
2216 * netif_napi_add - initialize a NAPI context
2217 * @dev: network device
2218 * @napi: NAPI context
2219 * @poll: polling function
2220 * @weight: default weight
2221 *
2222 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2223 * *any* of the other NAPI-related functions.
2224 */
2225 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2226 int (*poll)(struct napi_struct *, int), int weight);
2227
2228 /**
2229 * netif_tx_napi_add - initialize a NAPI context
2230 * @dev: network device
2231 * @napi: NAPI context
2232 * @poll: polling function
2233 * @weight: default weight
2234 *
2235 * This variant of netif_napi_add() should be used from drivers using NAPI
2236 * to exclusively poll a TX queue.
2237 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2238 */
2239 static inline void netif_tx_napi_add(struct net_device *dev,
2240 struct napi_struct *napi,
2241 int (*poll)(struct napi_struct *, int),
2242 int weight)
2243 {
2244 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2245 netif_napi_add(dev, napi, poll, weight);
2246 }
2247
2248 /**
2249 * netif_napi_del - remove a NAPI context
2250 * @napi: NAPI context
2251 *
2252 * netif_napi_del() removes a NAPI context from the network device NAPI list
2253 */
2254 void netif_napi_del(struct napi_struct *napi);
2255
2256 struct napi_gro_cb {
2257 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2258 void *frag0;
2259
2260 /* Length of frag0. */
2261 unsigned int frag0_len;
2262
2263 /* This indicates where we are processing relative to skb->data. */
2264 int data_offset;
2265
2266 /* This is non-zero if the packet cannot be merged with the new skb. */
2267 u16 flush;
2268
2269 /* Save the IP ID here and check when we get to the transport layer */
2270 u16 flush_id;
2271
2272 /* Number of segments aggregated. */
2273 u16 count;
2274
2275 /* Start offset for remote checksum offload */
2276 u16 gro_remcsum_start;
2277
2278 /* jiffies when first packet was created/queued */
2279 unsigned long age;
2280
2281 /* Used in ipv6_gro_receive() and foo-over-udp */
2282 u16 proto;
2283
2284 /* This is non-zero if the packet may be of the same flow. */
2285 u8 same_flow:1;
2286
2287 /* Used in tunnel GRO receive */
2288 u8 encap_mark:1;
2289
2290 /* GRO checksum is valid */
2291 u8 csum_valid:1;
2292
2293 /* Number of checksums via CHECKSUM_UNNECESSARY */
2294 u8 csum_cnt:3;
2295
2296 /* Free the skb? */
2297 u8 free:2;
2298 #define NAPI_GRO_FREE 1
2299 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2300
2301 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2302 u8 is_ipv6:1;
2303
2304 /* Used in GRE, set in fou/gue_gro_receive */
2305 u8 is_fou:1;
2306
2307 /* Used to determine if flush_id can be ignored */
2308 u8 is_atomic:1;
2309
2310 /* Number of gro_receive callbacks this packet already went through */
2311 u8 recursion_counter:4;
2312
2313 /* 1 bit hole */
2314
2315 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2316 __wsum csum;
2317
2318 /* used in skb_gro_receive() slow path */
2319 struct sk_buff *last;
2320 };
2321
2322 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2323
2324 #define GRO_RECURSION_LIMIT 15
2325 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2326 {
2327 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2328 }
2329
2330 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2331 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2332 struct list_head *head,
2333 struct sk_buff *skb)
2334 {
2335 if (unlikely(gro_recursion_inc_test(skb))) {
2336 NAPI_GRO_CB(skb)->flush |= 1;
2337 return NULL;
2338 }
2339
2340 return cb(head, skb);
2341 }
2342
2343 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2344 struct sk_buff *);
2345 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2346 struct sock *sk,
2347 struct list_head *head,
2348 struct sk_buff *skb)
2349 {
2350 if (unlikely(gro_recursion_inc_test(skb))) {
2351 NAPI_GRO_CB(skb)->flush |= 1;
2352 return NULL;
2353 }
2354
2355 return cb(sk, head, skb);
2356 }
2357
2358 struct packet_type {
2359 __be16 type; /* This is really htons(ether_type). */
2360 bool ignore_outgoing;
2361 struct net_device *dev; /* NULL is wildcarded here */
2362 int (*func) (struct sk_buff *,
2363 struct net_device *,
2364 struct packet_type *,
2365 struct net_device *);
2366 void (*list_func) (struct list_head *,
2367 struct packet_type *,
2368 struct net_device *);
2369 bool (*id_match)(struct packet_type *ptype,
2370 struct sock *sk);
2371 void *af_packet_priv;
2372 struct list_head list;
2373 };
2374
2375 struct offload_callbacks {
2376 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2377 netdev_features_t features);
2378 struct sk_buff *(*gro_receive)(struct list_head *head,
2379 struct sk_buff *skb);
2380 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2381 };
2382
2383 struct packet_offload {
2384 __be16 type; /* This is really htons(ether_type). */
2385 u16 priority;
2386 struct offload_callbacks callbacks;
2387 struct list_head list;
2388 };
2389
2390 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2391 struct pcpu_sw_netstats {
2392 u64 rx_packets;
2393 u64 rx_bytes;
2394 u64 tx_packets;
2395 u64 tx_bytes;
2396 struct u64_stats_sync syncp;
2397 } __aligned(4 * sizeof(u64));
2398
2399 struct pcpu_lstats {
2400 u64_stats_t packets;
2401 u64_stats_t bytes;
2402 struct u64_stats_sync syncp;
2403 } __aligned(2 * sizeof(u64));
2404
2405 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2406
2407 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2408 {
2409 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2410
2411 u64_stats_update_begin(&lstats->syncp);
2412 u64_stats_add(&lstats->bytes, len);
2413 u64_stats_inc(&lstats->packets);
2414 u64_stats_update_end(&lstats->syncp);
2415 }
2416
2417 #define __netdev_alloc_pcpu_stats(type, gfp) \
2418 ({ \
2419 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2420 if (pcpu_stats) { \
2421 int __cpu; \
2422 for_each_possible_cpu(__cpu) { \
2423 typeof(type) *stat; \
2424 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2425 u64_stats_init(&stat->syncp); \
2426 } \
2427 } \
2428 pcpu_stats; \
2429 })
2430
2431 #define netdev_alloc_pcpu_stats(type) \
2432 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2433
2434 enum netdev_lag_tx_type {
2435 NETDEV_LAG_TX_TYPE_UNKNOWN,
2436 NETDEV_LAG_TX_TYPE_RANDOM,
2437 NETDEV_LAG_TX_TYPE_BROADCAST,
2438 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2439 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2440 NETDEV_LAG_TX_TYPE_HASH,
2441 };
2442
2443 enum netdev_lag_hash {
2444 NETDEV_LAG_HASH_NONE,
2445 NETDEV_LAG_HASH_L2,
2446 NETDEV_LAG_HASH_L34,
2447 NETDEV_LAG_HASH_L23,
2448 NETDEV_LAG_HASH_E23,
2449 NETDEV_LAG_HASH_E34,
2450 NETDEV_LAG_HASH_UNKNOWN,
2451 };
2452
2453 struct netdev_lag_upper_info {
2454 enum netdev_lag_tx_type tx_type;
2455 enum netdev_lag_hash hash_type;
2456 };
2457
2458 struct netdev_lag_lower_state_info {
2459 u8 link_up : 1,
2460 tx_enabled : 1;
2461 };
2462
2463 #include <linux/notifier.h>
2464
2465 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2466 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2467 * adding new types.
2468 */
2469 enum netdev_cmd {
2470 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2471 NETDEV_DOWN,
2472 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2473 detected a hardware crash and restarted
2474 - we can use this eg to kick tcp sessions
2475 once done */
2476 NETDEV_CHANGE, /* Notify device state change */
2477 NETDEV_REGISTER,
2478 NETDEV_UNREGISTER,
2479 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2480 NETDEV_CHANGEADDR, /* notify after the address change */
2481 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2482 NETDEV_GOING_DOWN,
2483 NETDEV_CHANGENAME,
2484 NETDEV_FEAT_CHANGE,
2485 NETDEV_BONDING_FAILOVER,
2486 NETDEV_PRE_UP,
2487 NETDEV_PRE_TYPE_CHANGE,
2488 NETDEV_POST_TYPE_CHANGE,
2489 NETDEV_POST_INIT,
2490 NETDEV_RELEASE,
2491 NETDEV_NOTIFY_PEERS,
2492 NETDEV_JOIN,
2493 NETDEV_CHANGEUPPER,
2494 NETDEV_RESEND_IGMP,
2495 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2496 NETDEV_CHANGEINFODATA,
2497 NETDEV_BONDING_INFO,
2498 NETDEV_PRECHANGEUPPER,
2499 NETDEV_CHANGELOWERSTATE,
2500 NETDEV_UDP_TUNNEL_PUSH_INFO,
2501 NETDEV_UDP_TUNNEL_DROP_INFO,
2502 NETDEV_CHANGE_TX_QUEUE_LEN,
2503 NETDEV_CVLAN_FILTER_PUSH_INFO,
2504 NETDEV_CVLAN_FILTER_DROP_INFO,
2505 NETDEV_SVLAN_FILTER_PUSH_INFO,
2506 NETDEV_SVLAN_FILTER_DROP_INFO,
2507 };
2508 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2509
2510 int register_netdevice_notifier(struct notifier_block *nb);
2511 int unregister_netdevice_notifier(struct notifier_block *nb);
2512 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2513 int unregister_netdevice_notifier_net(struct net *net,
2514 struct notifier_block *nb);
2515
2516 struct netdev_notifier_info {
2517 struct net_device *dev;
2518 struct netlink_ext_ack *extack;
2519 };
2520
2521 struct netdev_notifier_info_ext {
2522 struct netdev_notifier_info info; /* must be first */
2523 union {
2524 u32 mtu;
2525 } ext;
2526 };
2527
2528 struct netdev_notifier_change_info {
2529 struct netdev_notifier_info info; /* must be first */
2530 unsigned int flags_changed;
2531 };
2532
2533 struct netdev_notifier_changeupper_info {
2534 struct netdev_notifier_info info; /* must be first */
2535 struct net_device *upper_dev; /* new upper dev */
2536 bool master; /* is upper dev master */
2537 bool linking; /* is the notification for link or unlink */
2538 void *upper_info; /* upper dev info */
2539 };
2540
2541 struct netdev_notifier_changelowerstate_info {
2542 struct netdev_notifier_info info; /* must be first */
2543 void *lower_state_info; /* is lower dev state */
2544 };
2545
2546 struct netdev_notifier_pre_changeaddr_info {
2547 struct netdev_notifier_info info; /* must be first */
2548 const unsigned char *dev_addr;
2549 };
2550
2551 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2552 struct net_device *dev)
2553 {
2554 info->dev = dev;
2555 info->extack = NULL;
2556 }
2557
2558 static inline struct net_device *
2559 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2560 {
2561 return info->dev;
2562 }
2563
2564 static inline struct netlink_ext_ack *
2565 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2566 {
2567 return info->extack;
2568 }
2569
2570 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2571
2572
2573 extern rwlock_t dev_base_lock; /* Device list lock */
2574
2575 #define for_each_netdev(net, d) \
2576 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2577 #define for_each_netdev_reverse(net, d) \
2578 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2579 #define for_each_netdev_rcu(net, d) \
2580 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2581 #define for_each_netdev_safe(net, d, n) \
2582 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2583 #define for_each_netdev_continue(net, d) \
2584 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2585 #define for_each_netdev_continue_reverse(net, d) \
2586 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2587 dev_list)
2588 #define for_each_netdev_continue_rcu(net, d) \
2589 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2590 #define for_each_netdev_in_bond_rcu(bond, slave) \
2591 for_each_netdev_rcu(&init_net, slave) \
2592 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2593 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2594
2595 static inline struct net_device *next_net_device(struct net_device *dev)
2596 {
2597 struct list_head *lh;
2598 struct net *net;
2599
2600 net = dev_net(dev);
2601 lh = dev->dev_list.next;
2602 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2603 }
2604
2605 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2606 {
2607 struct list_head *lh;
2608 struct net *net;
2609
2610 net = dev_net(dev);
2611 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2612 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2613 }
2614
2615 static inline struct net_device *first_net_device(struct net *net)
2616 {
2617 return list_empty(&net->dev_base_head) ? NULL :
2618 net_device_entry(net->dev_base_head.next);
2619 }
2620
2621 static inline struct net_device *first_net_device_rcu(struct net *net)
2622 {
2623 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2624
2625 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2626 }
2627
2628 int netdev_boot_setup_check(struct net_device *dev);
2629 unsigned long netdev_boot_base(const char *prefix, int unit);
2630 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2631 const char *hwaddr);
2632 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2633 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2634 void dev_add_pack(struct packet_type *pt);
2635 void dev_remove_pack(struct packet_type *pt);
2636 void __dev_remove_pack(struct packet_type *pt);
2637 void dev_add_offload(struct packet_offload *po);
2638 void dev_remove_offload(struct packet_offload *po);
2639
2640 int dev_get_iflink(const struct net_device *dev);
2641 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2642 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2643 unsigned short mask);
2644 struct net_device *dev_get_by_name(struct net *net, const char *name);
2645 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2646 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2647 int dev_alloc_name(struct net_device *dev, const char *name);
2648 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2649 void dev_close(struct net_device *dev);
2650 void dev_close_many(struct list_head *head, bool unlink);
2651 void dev_disable_lro(struct net_device *dev);
2652 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2653 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2654 struct net_device *sb_dev);
2655 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2656 struct net_device *sb_dev);
2657 int dev_queue_xmit(struct sk_buff *skb);
2658 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2659 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2660 int register_netdevice(struct net_device *dev);
2661 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2662 void unregister_netdevice_many(struct list_head *head);
2663 static inline void unregister_netdevice(struct net_device *dev)
2664 {
2665 unregister_netdevice_queue(dev, NULL);
2666 }
2667
2668 int netdev_refcnt_read(const struct net_device *dev);
2669 void free_netdev(struct net_device *dev);
2670 void netdev_freemem(struct net_device *dev);
2671 void synchronize_net(void);
2672 int init_dummy_netdev(struct net_device *dev);
2673
2674 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2675 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2676 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2677 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2678 int netdev_get_name(struct net *net, char *name, int ifindex);
2679 int dev_restart(struct net_device *dev);
2680 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2681
2682 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2683 {
2684 return NAPI_GRO_CB(skb)->data_offset;
2685 }
2686
2687 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2688 {
2689 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2690 }
2691
2692 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2693 {
2694 NAPI_GRO_CB(skb)->data_offset += len;
2695 }
2696
2697 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2698 unsigned int offset)
2699 {
2700 return NAPI_GRO_CB(skb)->frag0 + offset;
2701 }
2702
2703 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2704 {
2705 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2706 }
2707
2708 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2709 {
2710 NAPI_GRO_CB(skb)->frag0 = NULL;
2711 NAPI_GRO_CB(skb)->frag0_len = 0;
2712 }
2713
2714 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2715 unsigned int offset)
2716 {
2717 if (!pskb_may_pull(skb, hlen))
2718 return NULL;
2719
2720 skb_gro_frag0_invalidate(skb);
2721 return skb->data + offset;
2722 }
2723
2724 static inline void *skb_gro_network_header(struct sk_buff *skb)
2725 {
2726 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2727 skb_network_offset(skb);
2728 }
2729
2730 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2731 const void *start, unsigned int len)
2732 {
2733 if (NAPI_GRO_CB(skb)->csum_valid)
2734 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2735 csum_partial(start, len, 0));
2736 }
2737
2738 /* GRO checksum functions. These are logical equivalents of the normal
2739 * checksum functions (in skbuff.h) except that they operate on the GRO
2740 * offsets and fields in sk_buff.
2741 */
2742
2743 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2744
2745 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2746 {
2747 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2748 }
2749
2750 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2751 bool zero_okay,
2752 __sum16 check)
2753 {
2754 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2755 skb_checksum_start_offset(skb) <
2756 skb_gro_offset(skb)) &&
2757 !skb_at_gro_remcsum_start(skb) &&
2758 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2759 (!zero_okay || check));
2760 }
2761
2762 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2763 __wsum psum)
2764 {
2765 if (NAPI_GRO_CB(skb)->csum_valid &&
2766 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2767 return 0;
2768
2769 NAPI_GRO_CB(skb)->csum = psum;
2770
2771 return __skb_gro_checksum_complete(skb);
2772 }
2773
2774 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2775 {
2776 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2777 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2778 NAPI_GRO_CB(skb)->csum_cnt--;
2779 } else {
2780 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2781 * verified a new top level checksum or an encapsulated one
2782 * during GRO. This saves work if we fallback to normal path.
2783 */
2784 __skb_incr_checksum_unnecessary(skb);
2785 }
2786 }
2787
2788 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2789 compute_pseudo) \
2790 ({ \
2791 __sum16 __ret = 0; \
2792 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2793 __ret = __skb_gro_checksum_validate_complete(skb, \
2794 compute_pseudo(skb, proto)); \
2795 if (!__ret) \
2796 skb_gro_incr_csum_unnecessary(skb); \
2797 __ret; \
2798 })
2799
2800 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2801 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2802
2803 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2804 compute_pseudo) \
2805 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2806
2807 #define skb_gro_checksum_simple_validate(skb) \
2808 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2809
2810 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2811 {
2812 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2813 !NAPI_GRO_CB(skb)->csum_valid);
2814 }
2815
2816 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2817 __sum16 check, __wsum pseudo)
2818 {
2819 NAPI_GRO_CB(skb)->csum = ~pseudo;
2820 NAPI_GRO_CB(skb)->csum_valid = 1;
2821 }
2822
2823 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2824 do { \
2825 if (__skb_gro_checksum_convert_check(skb)) \
2826 __skb_gro_checksum_convert(skb, check, \
2827 compute_pseudo(skb, proto)); \
2828 } while (0)
2829
2830 struct gro_remcsum {
2831 int offset;
2832 __wsum delta;
2833 };
2834
2835 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2836 {
2837 grc->offset = 0;
2838 grc->delta = 0;
2839 }
2840
2841 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2842 unsigned int off, size_t hdrlen,
2843 int start, int offset,
2844 struct gro_remcsum *grc,
2845 bool nopartial)
2846 {
2847 __wsum delta;
2848 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2849
2850 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2851
2852 if (!nopartial) {
2853 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2854 return ptr;
2855 }
2856
2857 ptr = skb_gro_header_fast(skb, off);
2858 if (skb_gro_header_hard(skb, off + plen)) {
2859 ptr = skb_gro_header_slow(skb, off + plen, off);
2860 if (!ptr)
2861 return NULL;
2862 }
2863
2864 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2865 start, offset);
2866
2867 /* Adjust skb->csum since we changed the packet */
2868 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2869
2870 grc->offset = off + hdrlen + offset;
2871 grc->delta = delta;
2872
2873 return ptr;
2874 }
2875
2876 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2877 struct gro_remcsum *grc)
2878 {
2879 void *ptr;
2880 size_t plen = grc->offset + sizeof(u16);
2881
2882 if (!grc->delta)
2883 return;
2884
2885 ptr = skb_gro_header_fast(skb, grc->offset);
2886 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2887 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2888 if (!ptr)
2889 return;
2890 }
2891
2892 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2893 }
2894
2895 #ifdef CONFIG_XFRM_OFFLOAD
2896 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2897 {
2898 if (PTR_ERR(pp) != -EINPROGRESS)
2899 NAPI_GRO_CB(skb)->flush |= flush;
2900 }
2901 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2902 struct sk_buff *pp,
2903 int flush,
2904 struct gro_remcsum *grc)
2905 {
2906 if (PTR_ERR(pp) != -EINPROGRESS) {
2907 NAPI_GRO_CB(skb)->flush |= flush;
2908 skb_gro_remcsum_cleanup(skb, grc);
2909 skb->remcsum_offload = 0;
2910 }
2911 }
2912 #else
2913 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2914 {
2915 NAPI_GRO_CB(skb)->flush |= flush;
2916 }
2917 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2918 struct sk_buff *pp,
2919 int flush,
2920 struct gro_remcsum *grc)
2921 {
2922 NAPI_GRO_CB(skb)->flush |= flush;
2923 skb_gro_remcsum_cleanup(skb, grc);
2924 skb->remcsum_offload = 0;
2925 }
2926 #endif
2927
2928 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2929 unsigned short type,
2930 const void *daddr, const void *saddr,
2931 unsigned int len)
2932 {
2933 if (!dev->header_ops || !dev->header_ops->create)
2934 return 0;
2935
2936 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2937 }
2938
2939 static inline int dev_parse_header(const struct sk_buff *skb,
2940 unsigned char *haddr)
2941 {
2942 const struct net_device *dev = skb->dev;
2943
2944 if (!dev->header_ops || !dev->header_ops->parse)
2945 return 0;
2946 return dev->header_ops->parse(skb, haddr);
2947 }
2948
2949 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2950 {
2951 const struct net_device *dev = skb->dev;
2952
2953 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2954 return 0;
2955 return dev->header_ops->parse_protocol(skb);
2956 }
2957
2958 /* ll_header must have at least hard_header_len allocated */
2959 static inline bool dev_validate_header(const struct net_device *dev,
2960 char *ll_header, int len)
2961 {
2962 if (likely(len >= dev->hard_header_len))
2963 return true;
2964 if (len < dev->min_header_len)
2965 return false;
2966
2967 if (capable(CAP_SYS_RAWIO)) {
2968 memset(ll_header + len, 0, dev->hard_header_len - len);
2969 return true;
2970 }
2971
2972 if (dev->header_ops && dev->header_ops->validate)
2973 return dev->header_ops->validate(ll_header, len);
2974
2975 return false;
2976 }
2977
2978 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2979 int len, int size);
2980 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2981 static inline int unregister_gifconf(unsigned int family)
2982 {
2983 return register_gifconf(family, NULL);
2984 }
2985
2986 #ifdef CONFIG_NET_FLOW_LIMIT
2987 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2988 struct sd_flow_limit {
2989 u64 count;
2990 unsigned int num_buckets;
2991 unsigned int history_head;
2992 u16 history[FLOW_LIMIT_HISTORY];
2993 u8 buckets[];
2994 };
2995
2996 extern int netdev_flow_limit_table_len;
2997 #endif /* CONFIG_NET_FLOW_LIMIT */
2998
2999 /*
3000 * Incoming packets are placed on per-CPU queues
3001 */
3002 struct softnet_data {
3003 struct list_head poll_list;
3004 struct sk_buff_head process_queue;
3005
3006 /* stats */
3007 unsigned int processed;
3008 unsigned int time_squeeze;
3009 unsigned int received_rps;
3010 #ifdef CONFIG_RPS
3011 struct softnet_data *rps_ipi_list;
3012 #endif
3013 #ifdef CONFIG_NET_FLOW_LIMIT
3014 struct sd_flow_limit __rcu *flow_limit;
3015 #endif
3016 struct Qdisc *output_queue;
3017 struct Qdisc **output_queue_tailp;
3018 struct sk_buff *completion_queue;
3019 #ifdef CONFIG_XFRM_OFFLOAD
3020 struct sk_buff_head xfrm_backlog;
3021 #endif
3022 /* written and read only by owning cpu: */
3023 struct {
3024 u16 recursion;
3025 u8 more;
3026 } xmit;
3027 #ifdef CONFIG_RPS
3028 /* input_queue_head should be written by cpu owning this struct,
3029 * and only read by other cpus. Worth using a cache line.
3030 */
3031 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3032
3033 /* Elements below can be accessed between CPUs for RPS/RFS */
3034 call_single_data_t csd ____cacheline_aligned_in_smp;
3035 struct softnet_data *rps_ipi_next;
3036 unsigned int cpu;
3037 unsigned int input_queue_tail;
3038 #endif
3039 unsigned int dropped;
3040 struct sk_buff_head input_pkt_queue;
3041 struct napi_struct backlog;
3042
3043 };
3044
3045 static inline void input_queue_head_incr(struct softnet_data *sd)
3046 {
3047 #ifdef CONFIG_RPS
3048 sd->input_queue_head++;
3049 #endif
3050 }
3051
3052 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3053 unsigned int *qtail)
3054 {
3055 #ifdef CONFIG_RPS
3056 *qtail = ++sd->input_queue_tail;
3057 #endif
3058 }
3059
3060 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3061
3062 static inline int dev_recursion_level(void)
3063 {
3064 return this_cpu_read(softnet_data.xmit.recursion);
3065 }
3066
3067 #define XMIT_RECURSION_LIMIT 10
3068 static inline bool dev_xmit_recursion(void)
3069 {
3070 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3071 XMIT_RECURSION_LIMIT);
3072 }
3073
3074 static inline void dev_xmit_recursion_inc(void)
3075 {
3076 __this_cpu_inc(softnet_data.xmit.recursion);
3077 }
3078
3079 static inline void dev_xmit_recursion_dec(void)
3080 {
3081 __this_cpu_dec(softnet_data.xmit.recursion);
3082 }
3083
3084 void __netif_schedule(struct Qdisc *q);
3085 void netif_schedule_queue(struct netdev_queue *txq);
3086
3087 static inline void netif_tx_schedule_all(struct net_device *dev)
3088 {
3089 unsigned int i;
3090
3091 for (i = 0; i < dev->num_tx_queues; i++)
3092 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3093 }
3094
3095 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3096 {
3097 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3098 }
3099
3100 /**
3101 * netif_start_queue - allow transmit
3102 * @dev: network device
3103 *
3104 * Allow upper layers to call the device hard_start_xmit routine.
3105 */
3106 static inline void netif_start_queue(struct net_device *dev)
3107 {
3108 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3109 }
3110
3111 static inline void netif_tx_start_all_queues(struct net_device *dev)
3112 {
3113 unsigned int i;
3114
3115 for (i = 0; i < dev->num_tx_queues; i++) {
3116 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3117 netif_tx_start_queue(txq);
3118 }
3119 }
3120
3121 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3122
3123 /**
3124 * netif_wake_queue - restart transmit
3125 * @dev: network device
3126 *
3127 * Allow upper layers to call the device hard_start_xmit routine.
3128 * Used for flow control when transmit resources are available.
3129 */
3130 static inline void netif_wake_queue(struct net_device *dev)
3131 {
3132 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3133 }
3134
3135 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3136 {
3137 unsigned int i;
3138
3139 for (i = 0; i < dev->num_tx_queues; i++) {
3140 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3141 netif_tx_wake_queue(txq);
3142 }
3143 }
3144
3145 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3146 {
3147 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3148 }
3149
3150 /**
3151 * netif_stop_queue - stop transmitted packets
3152 * @dev: network device
3153 *
3154 * Stop upper layers calling the device hard_start_xmit routine.
3155 * Used for flow control when transmit resources are unavailable.
3156 */
3157 static inline void netif_stop_queue(struct net_device *dev)
3158 {
3159 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3160 }
3161
3162 void netif_tx_stop_all_queues(struct net_device *dev);
3163 void netdev_update_lockdep_key(struct net_device *dev);
3164
3165 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3166 {
3167 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3168 }
3169
3170 /**
3171 * netif_queue_stopped - test if transmit queue is flowblocked
3172 * @dev: network device
3173 *
3174 * Test if transmit queue on device is currently unable to send.
3175 */
3176 static inline bool netif_queue_stopped(const struct net_device *dev)
3177 {
3178 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3179 }
3180
3181 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3182 {
3183 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3184 }
3185
3186 static inline bool
3187 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3188 {
3189 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3190 }
3191
3192 static inline bool
3193 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3194 {
3195 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3196 }
3197
3198 /**
3199 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3200 * @dev_queue: pointer to transmit queue
3201 *
3202 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3203 * to give appropriate hint to the CPU.
3204 */
3205 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3206 {
3207 #ifdef CONFIG_BQL
3208 prefetchw(&dev_queue->dql.num_queued);
3209 #endif
3210 }
3211
3212 /**
3213 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3214 * @dev_queue: pointer to transmit queue
3215 *
3216 * BQL enabled drivers might use this helper in their TX completion path,
3217 * to give appropriate hint to the CPU.
3218 */
3219 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3220 {
3221 #ifdef CONFIG_BQL
3222 prefetchw(&dev_queue->dql.limit);
3223 #endif
3224 }
3225
3226 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3227 unsigned int bytes)
3228 {
3229 #ifdef CONFIG_BQL
3230 dql_queued(&dev_queue->dql, bytes);
3231
3232 if (likely(dql_avail(&dev_queue->dql) >= 0))
3233 return;
3234
3235 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3236
3237 /*
3238 * The XOFF flag must be set before checking the dql_avail below,
3239 * because in netdev_tx_completed_queue we update the dql_completed
3240 * before checking the XOFF flag.
3241 */
3242 smp_mb();
3243
3244 /* check again in case another CPU has just made room avail */
3245 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3246 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3247 #endif
3248 }
3249
3250 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3251 * that they should not test BQL status themselves.
3252 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3253 * skb of a batch.
3254 * Returns true if the doorbell must be used to kick the NIC.
3255 */
3256 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3257 unsigned int bytes,
3258 bool xmit_more)
3259 {
3260 if (xmit_more) {
3261 #ifdef CONFIG_BQL
3262 dql_queued(&dev_queue->dql, bytes);
3263 #endif
3264 return netif_tx_queue_stopped(dev_queue);
3265 }
3266 netdev_tx_sent_queue(dev_queue, bytes);
3267 return true;
3268 }
3269
3270 /**
3271 * netdev_sent_queue - report the number of bytes queued to hardware
3272 * @dev: network device
3273 * @bytes: number of bytes queued to the hardware device queue
3274 *
3275 * Report the number of bytes queued for sending/completion to the network
3276 * device hardware queue. @bytes should be a good approximation and should
3277 * exactly match netdev_completed_queue() @bytes
3278 */
3279 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3280 {
3281 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3282 }
3283
3284 static inline bool __netdev_sent_queue(struct net_device *dev,
3285 unsigned int bytes,
3286 bool xmit_more)
3287 {
3288 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3289 xmit_more);
3290 }
3291
3292 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3293 unsigned int pkts, unsigned int bytes)
3294 {
3295 #ifdef CONFIG_BQL
3296 if (unlikely(!bytes))
3297 return;
3298
3299 dql_completed(&dev_queue->dql, bytes);
3300
3301 /*
3302 * Without the memory barrier there is a small possiblity that
3303 * netdev_tx_sent_queue will miss the update and cause the queue to
3304 * be stopped forever
3305 */
3306 smp_mb();
3307
3308 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3309 return;
3310
3311 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3312 netif_schedule_queue(dev_queue);
3313 #endif
3314 }
3315
3316 /**
3317 * netdev_completed_queue - report bytes and packets completed by device
3318 * @dev: network device
3319 * @pkts: actual number of packets sent over the medium
3320 * @bytes: actual number of bytes sent over the medium
3321 *
3322 * Report the number of bytes and packets transmitted by the network device
3323 * hardware queue over the physical medium, @bytes must exactly match the
3324 * @bytes amount passed to netdev_sent_queue()
3325 */
3326 static inline void netdev_completed_queue(struct net_device *dev,
3327 unsigned int pkts, unsigned int bytes)
3328 {
3329 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3330 }
3331
3332 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3333 {
3334 #ifdef CONFIG_BQL
3335 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3336 dql_reset(&q->dql);
3337 #endif
3338 }
3339
3340 /**
3341 * netdev_reset_queue - reset the packets and bytes count of a network device
3342 * @dev_queue: network device
3343 *
3344 * Reset the bytes and packet count of a network device and clear the
3345 * software flow control OFF bit for this network device
3346 */
3347 static inline void netdev_reset_queue(struct net_device *dev_queue)
3348 {
3349 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3350 }
3351
3352 /**
3353 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3354 * @dev: network device
3355 * @queue_index: given tx queue index
3356 *
3357 * Returns 0 if given tx queue index >= number of device tx queues,
3358 * otherwise returns the originally passed tx queue index.
3359 */
3360 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3361 {
3362 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3363 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3364 dev->name, queue_index,
3365 dev->real_num_tx_queues);
3366 return 0;
3367 }
3368
3369 return queue_index;
3370 }
3371
3372 /**
3373 * netif_running - test if up
3374 * @dev: network device
3375 *
3376 * Test if the device has been brought up.
3377 */
3378 static inline bool netif_running(const struct net_device *dev)
3379 {
3380 return test_bit(__LINK_STATE_START, &dev->state);
3381 }
3382
3383 /*
3384 * Routines to manage the subqueues on a device. We only need start,
3385 * stop, and a check if it's stopped. All other device management is
3386 * done at the overall netdevice level.
3387 * Also test the device if we're multiqueue.
3388 */
3389
3390 /**
3391 * netif_start_subqueue - allow sending packets on subqueue
3392 * @dev: network device
3393 * @queue_index: sub queue index
3394 *
3395 * Start individual transmit queue of a device with multiple transmit queues.
3396 */
3397 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3398 {
3399 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3400
3401 netif_tx_start_queue(txq);
3402 }
3403
3404 /**
3405 * netif_stop_subqueue - stop sending packets on subqueue
3406 * @dev: network device
3407 * @queue_index: sub queue index
3408 *
3409 * Stop individual transmit queue of a device with multiple transmit queues.
3410 */
3411 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3412 {
3413 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3414 netif_tx_stop_queue(txq);
3415 }
3416
3417 /**
3418 * netif_subqueue_stopped - test status of subqueue
3419 * @dev: network device
3420 * @queue_index: sub queue index
3421 *
3422 * Check individual transmit queue of a device with multiple transmit queues.
3423 */
3424 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3425 u16 queue_index)
3426 {
3427 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3428
3429 return netif_tx_queue_stopped(txq);
3430 }
3431
3432 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3433 struct sk_buff *skb)
3434 {
3435 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3436 }
3437
3438 /**
3439 * netif_wake_subqueue - allow sending packets on subqueue
3440 * @dev: network device
3441 * @queue_index: sub queue index
3442 *
3443 * Resume individual transmit queue of a device with multiple transmit queues.
3444 */
3445 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3446 {
3447 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3448
3449 netif_tx_wake_queue(txq);
3450 }
3451
3452 #ifdef CONFIG_XPS
3453 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3454 u16 index);
3455 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3456 u16 index, bool is_rxqs_map);
3457
3458 /**
3459 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3460 * @j: CPU/Rx queue index
3461 * @mask: bitmask of all cpus/rx queues
3462 * @nr_bits: number of bits in the bitmask
3463 *
3464 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3465 */
3466 static inline bool netif_attr_test_mask(unsigned long j,
3467 const unsigned long *mask,
3468 unsigned int nr_bits)
3469 {
3470 cpu_max_bits_warn(j, nr_bits);
3471 return test_bit(j, mask);
3472 }
3473
3474 /**
3475 * netif_attr_test_online - Test for online CPU/Rx queue
3476 * @j: CPU/Rx queue index
3477 * @online_mask: bitmask for CPUs/Rx queues that are online
3478 * @nr_bits: number of bits in the bitmask
3479 *
3480 * Returns true if a CPU/Rx queue is online.
3481 */
3482 static inline bool netif_attr_test_online(unsigned long j,
3483 const unsigned long *online_mask,
3484 unsigned int nr_bits)
3485 {
3486 cpu_max_bits_warn(j, nr_bits);
3487
3488 if (online_mask)
3489 return test_bit(j, online_mask);
3490
3491 return (j < nr_bits);
3492 }
3493
3494 /**
3495 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3496 * @n: CPU/Rx queue index
3497 * @srcp: the cpumask/Rx queue mask pointer
3498 * @nr_bits: number of bits in the bitmask
3499 *
3500 * Returns >= nr_bits if no further CPUs/Rx queues set.
3501 */
3502 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3503 unsigned int nr_bits)
3504 {
3505 /* -1 is a legal arg here. */
3506 if (n != -1)
3507 cpu_max_bits_warn(n, nr_bits);
3508
3509 if (srcp)
3510 return find_next_bit(srcp, nr_bits, n + 1);
3511
3512 return n + 1;
3513 }
3514
3515 /**
3516 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3517 * @n: CPU/Rx queue index
3518 * @src1p: the first CPUs/Rx queues mask pointer
3519 * @src2p: the second CPUs/Rx queues mask pointer
3520 * @nr_bits: number of bits in the bitmask
3521 *
3522 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3523 */
3524 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3525 const unsigned long *src2p,
3526 unsigned int nr_bits)
3527 {
3528 /* -1 is a legal arg here. */
3529 if (n != -1)
3530 cpu_max_bits_warn(n, nr_bits);
3531
3532 if (src1p && src2p)
3533 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3534 else if (src1p)
3535 return find_next_bit(src1p, nr_bits, n + 1);
3536 else if (src2p)
3537 return find_next_bit(src2p, nr_bits, n + 1);
3538
3539 return n + 1;
3540 }
3541 #else
3542 static inline int netif_set_xps_queue(struct net_device *dev,
3543 const struct cpumask *mask,
3544 u16 index)
3545 {
3546 return 0;
3547 }
3548
3549 static inline int __netif_set_xps_queue(struct net_device *dev,
3550 const unsigned long *mask,
3551 u16 index, bool is_rxqs_map)
3552 {
3553 return 0;
3554 }
3555 #endif
3556
3557 /**
3558 * netif_is_multiqueue - test if device has multiple transmit queues
3559 * @dev: network device
3560 *
3561 * Check if device has multiple transmit queues
3562 */
3563 static inline bool netif_is_multiqueue(const struct net_device *dev)
3564 {
3565 return dev->num_tx_queues > 1;
3566 }
3567
3568 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3569
3570 #ifdef CONFIG_SYSFS
3571 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3572 #else
3573 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3574 unsigned int rxqs)
3575 {
3576 dev->real_num_rx_queues = rxqs;
3577 return 0;
3578 }
3579 #endif
3580
3581 static inline struct netdev_rx_queue *
3582 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3583 {
3584 return dev->_rx + rxq;
3585 }
3586
3587 #ifdef CONFIG_SYSFS
3588 static inline unsigned int get_netdev_rx_queue_index(
3589 struct netdev_rx_queue *queue)
3590 {
3591 struct net_device *dev = queue->dev;
3592 int index = queue - dev->_rx;
3593
3594 BUG_ON(index >= dev->num_rx_queues);
3595 return index;
3596 }
3597 #endif
3598
3599 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3600 int netif_get_num_default_rss_queues(void);
3601
3602 enum skb_free_reason {
3603 SKB_REASON_CONSUMED,
3604 SKB_REASON_DROPPED,
3605 };
3606
3607 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3608 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3609
3610 /*
3611 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3612 * interrupt context or with hardware interrupts being disabled.
3613 * (in_irq() || irqs_disabled())
3614 *
3615 * We provide four helpers that can be used in following contexts :
3616 *
3617 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3618 * replacing kfree_skb(skb)
3619 *
3620 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3621 * Typically used in place of consume_skb(skb) in TX completion path
3622 *
3623 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3624 * replacing kfree_skb(skb)
3625 *
3626 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3627 * and consumed a packet. Used in place of consume_skb(skb)
3628 */
3629 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3630 {
3631 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3632 }
3633
3634 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3635 {
3636 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3637 }
3638
3639 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3640 {
3641 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3642 }
3643
3644 static inline void dev_consume_skb_any(struct sk_buff *skb)
3645 {
3646 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3647 }
3648
3649 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3650 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3651 int netif_rx(struct sk_buff *skb);
3652 int netif_rx_ni(struct sk_buff *skb);
3653 int netif_receive_skb(struct sk_buff *skb);
3654 int netif_receive_skb_core(struct sk_buff *skb);
3655 void netif_receive_skb_list(struct list_head *head);
3656 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3657 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3658 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3659 gro_result_t napi_gro_frags(struct napi_struct *napi);
3660 struct packet_offload *gro_find_receive_by_type(__be16 type);
3661 struct packet_offload *gro_find_complete_by_type(__be16 type);
3662
3663 static inline void napi_free_frags(struct napi_struct *napi)
3664 {
3665 kfree_skb(napi->skb);
3666 napi->skb = NULL;
3667 }
3668
3669 bool netdev_is_rx_handler_busy(struct net_device *dev);
3670 int netdev_rx_handler_register(struct net_device *dev,
3671 rx_handler_func_t *rx_handler,
3672 void *rx_handler_data);
3673 void netdev_rx_handler_unregister(struct net_device *dev);
3674
3675 bool dev_valid_name(const char *name);
3676 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3677 bool *need_copyout);
3678 int dev_ifconf(struct net *net, struct ifconf *, int);
3679 int dev_ethtool(struct net *net, struct ifreq *);
3680 unsigned int dev_get_flags(const struct net_device *);
3681 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3682 struct netlink_ext_ack *extack);
3683 int dev_change_flags(struct net_device *dev, unsigned int flags,
3684 struct netlink_ext_ack *extack);
3685 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3686 unsigned int gchanges);
3687 int dev_change_name(struct net_device *, const char *);
3688 int dev_set_alias(struct net_device *, const char *, size_t);
3689 int dev_get_alias(const struct net_device *, char *, size_t);
3690 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3691 int __dev_set_mtu(struct net_device *, int);
3692 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3693 struct netlink_ext_ack *extack);
3694 int dev_set_mtu(struct net_device *, int);
3695 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3696 void dev_set_group(struct net_device *, int);
3697 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3698 struct netlink_ext_ack *extack);
3699 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3700 struct netlink_ext_ack *extack);
3701 int dev_change_carrier(struct net_device *, bool new_carrier);
3702 int dev_get_phys_port_id(struct net_device *dev,
3703 struct netdev_phys_item_id *ppid);
3704 int dev_get_phys_port_name(struct net_device *dev,
3705 char *name, size_t len);
3706 int dev_get_port_parent_id(struct net_device *dev,
3707 struct netdev_phys_item_id *ppid, bool recurse);
3708 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3709 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3710 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3711 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3712 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3713 struct netdev_queue *txq, int *ret);
3714
3715 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3716 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3717 int fd, u32 flags);
3718 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3719 enum bpf_netdev_command cmd);
3720 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3721
3722 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3723 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3724 bool is_skb_forwardable(const struct net_device *dev,
3725 const struct sk_buff *skb);
3726
3727 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3728 struct sk_buff *skb)
3729 {
3730 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3731 unlikely(!is_skb_forwardable(dev, skb))) {
3732 atomic_long_inc(&dev->rx_dropped);
3733 kfree_skb(skb);
3734 return NET_RX_DROP;
3735 }
3736
3737 skb_scrub_packet(skb, true);
3738 skb->priority = 0;
3739 return 0;
3740 }
3741
3742 bool dev_nit_active(struct net_device *dev);
3743 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3744
3745 extern int netdev_budget;
3746 extern unsigned int netdev_budget_usecs;
3747
3748 /* Called by rtnetlink.c:rtnl_unlock() */
3749 void netdev_run_todo(void);
3750
3751 /**
3752 * dev_put - release reference to device
3753 * @dev: network device
3754 *
3755 * Release reference to device to allow it to be freed.
3756 */
3757 static inline void dev_put(struct net_device *dev)
3758 {
3759 this_cpu_dec(*dev->pcpu_refcnt);
3760 }
3761
3762 /**
3763 * dev_hold - get reference to device
3764 * @dev: network device
3765 *
3766 * Hold reference to device to keep it from being freed.
3767 */
3768 static inline void dev_hold(struct net_device *dev)
3769 {
3770 this_cpu_inc(*dev->pcpu_refcnt);
3771 }
3772
3773 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3774 * and _off may be called from IRQ context, but it is caller
3775 * who is responsible for serialization of these calls.
3776 *
3777 * The name carrier is inappropriate, these functions should really be
3778 * called netif_lowerlayer_*() because they represent the state of any
3779 * kind of lower layer not just hardware media.
3780 */
3781
3782 void linkwatch_init_dev(struct net_device *dev);
3783 void linkwatch_fire_event(struct net_device *dev);
3784 void linkwatch_forget_dev(struct net_device *dev);
3785
3786 /**
3787 * netif_carrier_ok - test if carrier present
3788 * @dev: network device
3789 *
3790 * Check if carrier is present on device
3791 */
3792 static inline bool netif_carrier_ok(const struct net_device *dev)
3793 {
3794 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3795 }
3796
3797 unsigned long dev_trans_start(struct net_device *dev);
3798
3799 void __netdev_watchdog_up(struct net_device *dev);
3800
3801 void netif_carrier_on(struct net_device *dev);
3802
3803 void netif_carrier_off(struct net_device *dev);
3804
3805 /**
3806 * netif_dormant_on - mark device as dormant.
3807 * @dev: network device
3808 *
3809 * Mark device as dormant (as per RFC2863).
3810 *
3811 * The dormant state indicates that the relevant interface is not
3812 * actually in a condition to pass packets (i.e., it is not 'up') but is
3813 * in a "pending" state, waiting for some external event. For "on-
3814 * demand" interfaces, this new state identifies the situation where the
3815 * interface is waiting for events to place it in the up state.
3816 */
3817 static inline void netif_dormant_on(struct net_device *dev)
3818 {
3819 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3820 linkwatch_fire_event(dev);
3821 }
3822
3823 /**
3824 * netif_dormant_off - set device as not dormant.
3825 * @dev: network device
3826 *
3827 * Device is not in dormant state.
3828 */
3829 static inline void netif_dormant_off(struct net_device *dev)
3830 {
3831 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3832 linkwatch_fire_event(dev);
3833 }
3834
3835 /**
3836 * netif_dormant - test if device is dormant
3837 * @dev: network device
3838 *
3839 * Check if device is dormant.
3840 */
3841 static inline bool netif_dormant(const struct net_device *dev)
3842 {
3843 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3844 }
3845
3846
3847 /**
3848 * netif_oper_up - test if device is operational
3849 * @dev: network device
3850 *
3851 * Check if carrier is operational
3852 */
3853 static inline bool netif_oper_up(const struct net_device *dev)
3854 {
3855 return (dev->operstate == IF_OPER_UP ||
3856 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3857 }
3858
3859 /**
3860 * netif_device_present - is device available or removed
3861 * @dev: network device
3862 *
3863 * Check if device has not been removed from system.
3864 */
3865 static inline bool netif_device_present(struct net_device *dev)
3866 {
3867 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3868 }
3869
3870 void netif_device_detach(struct net_device *dev);
3871
3872 void netif_device_attach(struct net_device *dev);
3873
3874 /*
3875 * Network interface message level settings
3876 */
3877
3878 enum {
3879 NETIF_MSG_DRV = 0x0001,
3880 NETIF_MSG_PROBE = 0x0002,
3881 NETIF_MSG_LINK = 0x0004,
3882 NETIF_MSG_TIMER = 0x0008,
3883 NETIF_MSG_IFDOWN = 0x0010,
3884 NETIF_MSG_IFUP = 0x0020,
3885 NETIF_MSG_RX_ERR = 0x0040,
3886 NETIF_MSG_TX_ERR = 0x0080,
3887 NETIF_MSG_TX_QUEUED = 0x0100,
3888 NETIF_MSG_INTR = 0x0200,
3889 NETIF_MSG_TX_DONE = 0x0400,
3890 NETIF_MSG_RX_STATUS = 0x0800,
3891 NETIF_MSG_PKTDATA = 0x1000,
3892 NETIF_MSG_HW = 0x2000,
3893 NETIF_MSG_WOL = 0x4000,
3894 };
3895
3896 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3897 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3898 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3899 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3900 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3901 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3902 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3903 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3904 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3905 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3906 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3907 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3908 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3909 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3910 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3911
3912 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3913 {
3914 /* use default */
3915 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3916 return default_msg_enable_bits;
3917 if (debug_value == 0) /* no output */
3918 return 0;
3919 /* set low N bits */
3920 return (1U << debug_value) - 1;
3921 }
3922
3923 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3924 {
3925 spin_lock(&txq->_xmit_lock);
3926 txq->xmit_lock_owner = cpu;
3927 }
3928
3929 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3930 {
3931 __acquire(&txq->_xmit_lock);
3932 return true;
3933 }
3934
3935 static inline void __netif_tx_release(struct netdev_queue *txq)
3936 {
3937 __release(&txq->_xmit_lock);
3938 }
3939
3940 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3941 {
3942 spin_lock_bh(&txq->_xmit_lock);
3943 txq->xmit_lock_owner = smp_processor_id();
3944 }
3945
3946 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3947 {
3948 bool ok = spin_trylock(&txq->_xmit_lock);
3949 if (likely(ok))
3950 txq->xmit_lock_owner = smp_processor_id();
3951 return ok;
3952 }
3953
3954 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3955 {
3956 txq->xmit_lock_owner = -1;
3957 spin_unlock(&txq->_xmit_lock);
3958 }
3959
3960 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3961 {
3962 txq->xmit_lock_owner = -1;
3963 spin_unlock_bh(&txq->_xmit_lock);
3964 }
3965
3966 static inline void txq_trans_update(struct netdev_queue *txq)
3967 {
3968 if (txq->xmit_lock_owner != -1)
3969 txq->trans_start = jiffies;
3970 }
3971
3972 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3973 static inline void netif_trans_update(struct net_device *dev)
3974 {
3975 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3976
3977 if (txq->trans_start != jiffies)
3978 txq->trans_start = jiffies;
3979 }
3980
3981 /**
3982 * netif_tx_lock - grab network device transmit lock
3983 * @dev: network device
3984 *
3985 * Get network device transmit lock
3986 */
3987 static inline void netif_tx_lock(struct net_device *dev)
3988 {
3989 unsigned int i;
3990 int cpu;
3991
3992 spin_lock(&dev->tx_global_lock);
3993 cpu = smp_processor_id();
3994 for (i = 0; i < dev->num_tx_queues; i++) {
3995 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3996
3997 /* We are the only thread of execution doing a
3998 * freeze, but we have to grab the _xmit_lock in
3999 * order to synchronize with threads which are in
4000 * the ->hard_start_xmit() handler and already
4001 * checked the frozen bit.
4002 */
4003 __netif_tx_lock(txq, cpu);
4004 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4005 __netif_tx_unlock(txq);
4006 }
4007 }
4008
4009 static inline void netif_tx_lock_bh(struct net_device *dev)
4010 {
4011 local_bh_disable();
4012 netif_tx_lock(dev);
4013 }
4014
4015 static inline void netif_tx_unlock(struct net_device *dev)
4016 {
4017 unsigned int i;
4018
4019 for (i = 0; i < dev->num_tx_queues; i++) {
4020 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4021
4022 /* No need to grab the _xmit_lock here. If the
4023 * queue is not stopped for another reason, we
4024 * force a schedule.
4025 */
4026 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4027 netif_schedule_queue(txq);
4028 }
4029 spin_unlock(&dev->tx_global_lock);
4030 }
4031
4032 static inline void netif_tx_unlock_bh(struct net_device *dev)
4033 {
4034 netif_tx_unlock(dev);
4035 local_bh_enable();
4036 }
4037
4038 #define HARD_TX_LOCK(dev, txq, cpu) { \
4039 if ((dev->features & NETIF_F_LLTX) == 0) { \
4040 __netif_tx_lock(txq, cpu); \
4041 } else { \
4042 __netif_tx_acquire(txq); \
4043 } \
4044 }
4045
4046 #define HARD_TX_TRYLOCK(dev, txq) \
4047 (((dev->features & NETIF_F_LLTX) == 0) ? \
4048 __netif_tx_trylock(txq) : \
4049 __netif_tx_acquire(txq))
4050
4051 #define HARD_TX_UNLOCK(dev, txq) { \
4052 if ((dev->features & NETIF_F_LLTX) == 0) { \
4053 __netif_tx_unlock(txq); \
4054 } else { \
4055 __netif_tx_release(txq); \
4056 } \
4057 }
4058
4059 static inline void netif_tx_disable(struct net_device *dev)
4060 {
4061 unsigned int i;
4062 int cpu;
4063
4064 local_bh_disable();
4065 cpu = smp_processor_id();
4066 for (i = 0; i < dev->num_tx_queues; i++) {
4067 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4068
4069 __netif_tx_lock(txq, cpu);
4070 netif_tx_stop_queue(txq);
4071 __netif_tx_unlock(txq);
4072 }
4073 local_bh_enable();
4074 }
4075
4076 static inline void netif_addr_lock(struct net_device *dev)
4077 {
4078 spin_lock(&dev->addr_list_lock);
4079 }
4080
4081 static inline void netif_addr_lock_bh(struct net_device *dev)
4082 {
4083 spin_lock_bh(&dev->addr_list_lock);
4084 }
4085
4086 static inline void netif_addr_unlock(struct net_device *dev)
4087 {
4088 spin_unlock(&dev->addr_list_lock);
4089 }
4090
4091 static inline void netif_addr_unlock_bh(struct net_device *dev)
4092 {
4093 spin_unlock_bh(&dev->addr_list_lock);
4094 }
4095
4096 /*
4097 * dev_addrs walker. Should be used only for read access. Call with
4098 * rcu_read_lock held.
4099 */
4100 #define for_each_dev_addr(dev, ha) \
4101 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4102
4103 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4104
4105 void ether_setup(struct net_device *dev);
4106
4107 /* Support for loadable net-drivers */
4108 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4109 unsigned char name_assign_type,
4110 void (*setup)(struct net_device *),
4111 unsigned int txqs, unsigned int rxqs);
4112 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4113 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4114
4115 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4116 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4117 count)
4118
4119 int register_netdev(struct net_device *dev);
4120 void unregister_netdev(struct net_device *dev);
4121
4122 /* General hardware address lists handling functions */
4123 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4124 struct netdev_hw_addr_list *from_list, int addr_len);
4125 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4126 struct netdev_hw_addr_list *from_list, int addr_len);
4127 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4128 struct net_device *dev,
4129 int (*sync)(struct net_device *, const unsigned char *),
4130 int (*unsync)(struct net_device *,
4131 const unsigned char *));
4132 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4133 struct net_device *dev,
4134 int (*sync)(struct net_device *,
4135 const unsigned char *, int),
4136 int (*unsync)(struct net_device *,
4137 const unsigned char *, int));
4138 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4139 struct net_device *dev,
4140 int (*unsync)(struct net_device *,
4141 const unsigned char *, int));
4142 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4143 struct net_device *dev,
4144 int (*unsync)(struct net_device *,
4145 const unsigned char *));
4146 void __hw_addr_init(struct netdev_hw_addr_list *list);
4147
4148 /* Functions used for device addresses handling */
4149 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4150 unsigned char addr_type);
4151 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4152 unsigned char addr_type);
4153 void dev_addr_flush(struct net_device *dev);
4154 int dev_addr_init(struct net_device *dev);
4155
4156 /* Functions used for unicast addresses handling */
4157 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4158 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4159 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4160 int dev_uc_sync(struct net_device *to, struct net_device *from);
4161 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4162 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4163 void dev_uc_flush(struct net_device *dev);
4164 void dev_uc_init(struct net_device *dev);
4165
4166 /**
4167 * __dev_uc_sync - Synchonize device's unicast list
4168 * @dev: device to sync
4169 * @sync: function to call if address should be added
4170 * @unsync: function to call if address should be removed
4171 *
4172 * Add newly added addresses to the interface, and release
4173 * addresses that have been deleted.
4174 */
4175 static inline int __dev_uc_sync(struct net_device *dev,
4176 int (*sync)(struct net_device *,
4177 const unsigned char *),
4178 int (*unsync)(struct net_device *,
4179 const unsigned char *))
4180 {
4181 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4182 }
4183
4184 /**
4185 * __dev_uc_unsync - Remove synchronized addresses from device
4186 * @dev: device to sync
4187 * @unsync: function to call if address should be removed
4188 *
4189 * Remove all addresses that were added to the device by dev_uc_sync().
4190 */
4191 static inline void __dev_uc_unsync(struct net_device *dev,
4192 int (*unsync)(struct net_device *,
4193 const unsigned char *))
4194 {
4195 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4196 }
4197
4198 /* Functions used for multicast addresses handling */
4199 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4200 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4201 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4202 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4203 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4204 int dev_mc_sync(struct net_device *to, struct net_device *from);
4205 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4206 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4207 void dev_mc_flush(struct net_device *dev);
4208 void dev_mc_init(struct net_device *dev);
4209
4210 /**
4211 * __dev_mc_sync - Synchonize device's multicast list
4212 * @dev: device to sync
4213 * @sync: function to call if address should be added
4214 * @unsync: function to call if address should be removed
4215 *
4216 * Add newly added addresses to the interface, and release
4217 * addresses that have been deleted.
4218 */
4219 static inline int __dev_mc_sync(struct net_device *dev,
4220 int (*sync)(struct net_device *,
4221 const unsigned char *),
4222 int (*unsync)(struct net_device *,
4223 const unsigned char *))
4224 {
4225 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4226 }
4227
4228 /**
4229 * __dev_mc_unsync - Remove synchronized addresses from device
4230 * @dev: device to sync
4231 * @unsync: function to call if address should be removed
4232 *
4233 * Remove all addresses that were added to the device by dev_mc_sync().
4234 */
4235 static inline void __dev_mc_unsync(struct net_device *dev,
4236 int (*unsync)(struct net_device *,
4237 const unsigned char *))
4238 {
4239 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4240 }
4241
4242 /* Functions used for secondary unicast and multicast support */
4243 void dev_set_rx_mode(struct net_device *dev);
4244 void __dev_set_rx_mode(struct net_device *dev);
4245 int dev_set_promiscuity(struct net_device *dev, int inc);
4246 int dev_set_allmulti(struct net_device *dev, int inc);
4247 void netdev_state_change(struct net_device *dev);
4248 void netdev_notify_peers(struct net_device *dev);
4249 void netdev_features_change(struct net_device *dev);
4250 /* Load a device via the kmod */
4251 void dev_load(struct net *net, const char *name);
4252 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4253 struct rtnl_link_stats64 *storage);
4254 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4255 const struct net_device_stats *netdev_stats);
4256
4257 extern int netdev_max_backlog;
4258 extern int netdev_tstamp_prequeue;
4259 extern int weight_p;
4260 extern int dev_weight_rx_bias;
4261 extern int dev_weight_tx_bias;
4262 extern int dev_rx_weight;
4263 extern int dev_tx_weight;
4264 extern int gro_normal_batch;
4265
4266 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4267 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4268 struct list_head **iter);
4269 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4270 struct list_head **iter);
4271
4272 /* iterate through upper list, must be called under RCU read lock */
4273 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4274 for (iter = &(dev)->adj_list.upper, \
4275 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4276 updev; \
4277 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4278
4279 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4280 int (*fn)(struct net_device *upper_dev,
4281 void *data),
4282 void *data);
4283
4284 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4285 struct net_device *upper_dev);
4286
4287 bool netdev_has_any_upper_dev(struct net_device *dev);
4288
4289 void *netdev_lower_get_next_private(struct net_device *dev,
4290 struct list_head **iter);
4291 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4292 struct list_head **iter);
4293
4294 #define netdev_for_each_lower_private(dev, priv, iter) \
4295 for (iter = (dev)->adj_list.lower.next, \
4296 priv = netdev_lower_get_next_private(dev, &(iter)); \
4297 priv; \
4298 priv = netdev_lower_get_next_private(dev, &(iter)))
4299
4300 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4301 for (iter = &(dev)->adj_list.lower, \
4302 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4303 priv; \
4304 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4305
4306 void *netdev_lower_get_next(struct net_device *dev,
4307 struct list_head **iter);
4308
4309 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4310 for (iter = (dev)->adj_list.lower.next, \
4311 ldev = netdev_lower_get_next(dev, &(iter)); \
4312 ldev; \
4313 ldev = netdev_lower_get_next(dev, &(iter)))
4314
4315 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4316 struct list_head **iter);
4317 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4318 struct list_head **iter);
4319
4320 int netdev_walk_all_lower_dev(struct net_device *dev,
4321 int (*fn)(struct net_device *lower_dev,
4322 void *data),
4323 void *data);
4324 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4325 int (*fn)(struct net_device *lower_dev,
4326 void *data),
4327 void *data);
4328
4329 void *netdev_adjacent_get_private(struct list_head *adj_list);
4330 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4331 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4332 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4333 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4334 struct netlink_ext_ack *extack);
4335 int netdev_master_upper_dev_link(struct net_device *dev,
4336 struct net_device *upper_dev,
4337 void *upper_priv, void *upper_info,
4338 struct netlink_ext_ack *extack);
4339 void netdev_upper_dev_unlink(struct net_device *dev,
4340 struct net_device *upper_dev);
4341 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4342 struct net_device *new_dev,
4343 struct net_device *dev,
4344 struct netlink_ext_ack *extack);
4345 void netdev_adjacent_change_commit(struct net_device *old_dev,
4346 struct net_device *new_dev,
4347 struct net_device *dev);
4348 void netdev_adjacent_change_abort(struct net_device *old_dev,
4349 struct net_device *new_dev,
4350 struct net_device *dev);
4351 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4352 void *netdev_lower_dev_get_private(struct net_device *dev,
4353 struct net_device *lower_dev);
4354 void netdev_lower_state_changed(struct net_device *lower_dev,
4355 void *lower_state_info);
4356
4357 /* RSS keys are 40 or 52 bytes long */
4358 #define NETDEV_RSS_KEY_LEN 52
4359 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4360 void netdev_rss_key_fill(void *buffer, size_t len);
4361
4362 int skb_checksum_help(struct sk_buff *skb);
4363 int skb_crc32c_csum_help(struct sk_buff *skb);
4364 int skb_csum_hwoffload_help(struct sk_buff *skb,
4365 const netdev_features_t features);
4366
4367 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4368 netdev_features_t features, bool tx_path);
4369 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4370 netdev_features_t features);
4371
4372 struct netdev_bonding_info {
4373 ifslave slave;
4374 ifbond master;
4375 };
4376
4377 struct netdev_notifier_bonding_info {
4378 struct netdev_notifier_info info; /* must be first */
4379 struct netdev_bonding_info bonding_info;
4380 };
4381
4382 void netdev_bonding_info_change(struct net_device *dev,
4383 struct netdev_bonding_info *bonding_info);
4384
4385 static inline
4386 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4387 {
4388 return __skb_gso_segment(skb, features, true);
4389 }
4390 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4391
4392 static inline bool can_checksum_protocol(netdev_features_t features,
4393 __be16 protocol)
4394 {
4395 if (protocol == htons(ETH_P_FCOE))
4396 return !!(features & NETIF_F_FCOE_CRC);
4397
4398 /* Assume this is an IP checksum (not SCTP CRC) */
4399
4400 if (features & NETIF_F_HW_CSUM) {
4401 /* Can checksum everything */
4402 return true;
4403 }
4404
4405 switch (protocol) {
4406 case htons(ETH_P_IP):
4407 return !!(features & NETIF_F_IP_CSUM);
4408 case htons(ETH_P_IPV6):
4409 return !!(features & NETIF_F_IPV6_CSUM);
4410 default:
4411 return false;
4412 }
4413 }
4414
4415 #ifdef CONFIG_BUG
4416 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4417 #else
4418 static inline void netdev_rx_csum_fault(struct net_device *dev,
4419 struct sk_buff *skb)
4420 {
4421 }
4422 #endif
4423 /* rx skb timestamps */
4424 void net_enable_timestamp(void);
4425 void net_disable_timestamp(void);
4426
4427 #ifdef CONFIG_PROC_FS
4428 int __init dev_proc_init(void);
4429 #else
4430 #define dev_proc_init() 0
4431 #endif
4432
4433 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4434 struct sk_buff *skb, struct net_device *dev,
4435 bool more)
4436 {
4437 __this_cpu_write(softnet_data.xmit.more, more);
4438 return ops->ndo_start_xmit(skb, dev);
4439 }
4440
4441 static inline bool netdev_xmit_more(void)
4442 {
4443 return __this_cpu_read(softnet_data.xmit.more);
4444 }
4445
4446 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4447 struct netdev_queue *txq, bool more)
4448 {
4449 const struct net_device_ops *ops = dev->netdev_ops;
4450 netdev_tx_t rc;
4451
4452 rc = __netdev_start_xmit(ops, skb, dev, more);
4453 if (rc == NETDEV_TX_OK)
4454 txq_trans_update(txq);
4455
4456 return rc;
4457 }
4458
4459 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4460 const void *ns);
4461 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4462 const void *ns);
4463
4464 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4465 {
4466 return netdev_class_create_file_ns(class_attr, NULL);
4467 }
4468
4469 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4470 {
4471 netdev_class_remove_file_ns(class_attr, NULL);
4472 }
4473
4474 extern const struct kobj_ns_type_operations net_ns_type_operations;
4475
4476 const char *netdev_drivername(const struct net_device *dev);
4477
4478 void linkwatch_run_queue(void);
4479
4480 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4481 netdev_features_t f2)
4482 {
4483 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4484 if (f1 & NETIF_F_HW_CSUM)
4485 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4486 else
4487 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4488 }
4489
4490 return f1 & f2;
4491 }
4492
4493 static inline netdev_features_t netdev_get_wanted_features(
4494 struct net_device *dev)
4495 {
4496 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4497 }
4498 netdev_features_t netdev_increment_features(netdev_features_t all,
4499 netdev_features_t one, netdev_features_t mask);
4500
4501 /* Allow TSO being used on stacked device :
4502 * Performing the GSO segmentation before last device
4503 * is a performance improvement.
4504 */
4505 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4506 netdev_features_t mask)
4507 {
4508 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4509 }
4510
4511 int __netdev_update_features(struct net_device *dev);
4512 void netdev_update_features(struct net_device *dev);
4513 void netdev_change_features(struct net_device *dev);
4514
4515 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4516 struct net_device *dev);
4517
4518 netdev_features_t passthru_features_check(struct sk_buff *skb,
4519 struct net_device *dev,
4520 netdev_features_t features);
4521 netdev_features_t netif_skb_features(struct sk_buff *skb);
4522
4523 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4524 {
4525 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4526
4527 /* check flags correspondence */
4528 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4529 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4530 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4531 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4532 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4533 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4534 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4535 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4536 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4537 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4538 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4539 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4540 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4541 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4542 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4543 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4544 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4545 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4546
4547 return (features & feature) == feature;
4548 }
4549
4550 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4551 {
4552 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4553 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4554 }
4555
4556 static inline bool netif_needs_gso(struct sk_buff *skb,
4557 netdev_features_t features)
4558 {
4559 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4560 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4561 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4562 }
4563
4564 static inline void netif_set_gso_max_size(struct net_device *dev,
4565 unsigned int size)
4566 {
4567 dev->gso_max_size = size;
4568 }
4569
4570 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4571 int pulled_hlen, u16 mac_offset,
4572 int mac_len)
4573 {
4574 skb->protocol = protocol;
4575 skb->encapsulation = 1;
4576 skb_push(skb, pulled_hlen);
4577 skb_reset_transport_header(skb);
4578 skb->mac_header = mac_offset;
4579 skb->network_header = skb->mac_header + mac_len;
4580 skb->mac_len = mac_len;
4581 }
4582
4583 static inline bool netif_is_macsec(const struct net_device *dev)
4584 {
4585 return dev->priv_flags & IFF_MACSEC;
4586 }
4587
4588 static inline bool netif_is_macvlan(const struct net_device *dev)
4589 {
4590 return dev->priv_flags & IFF_MACVLAN;
4591 }
4592
4593 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4594 {
4595 return dev->priv_flags & IFF_MACVLAN_PORT;
4596 }
4597
4598 static inline bool netif_is_bond_master(const struct net_device *dev)
4599 {
4600 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4601 }
4602
4603 static inline bool netif_is_bond_slave(const struct net_device *dev)
4604 {
4605 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4606 }
4607
4608 static inline bool netif_supports_nofcs(struct net_device *dev)
4609 {
4610 return dev->priv_flags & IFF_SUPP_NOFCS;
4611 }
4612
4613 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4614 {
4615 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4616 }
4617
4618 static inline bool netif_is_l3_master(const struct net_device *dev)
4619 {
4620 return dev->priv_flags & IFF_L3MDEV_MASTER;
4621 }
4622
4623 static inline bool netif_is_l3_slave(const struct net_device *dev)
4624 {
4625 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4626 }
4627
4628 static inline bool netif_is_bridge_master(const struct net_device *dev)
4629 {
4630 return dev->priv_flags & IFF_EBRIDGE;
4631 }
4632
4633 static inline bool netif_is_bridge_port(const struct net_device *dev)
4634 {
4635 return dev->priv_flags & IFF_BRIDGE_PORT;
4636 }
4637
4638 static inline bool netif_is_ovs_master(const struct net_device *dev)
4639 {
4640 return dev->priv_flags & IFF_OPENVSWITCH;
4641 }
4642
4643 static inline bool netif_is_ovs_port(const struct net_device *dev)
4644 {
4645 return dev->priv_flags & IFF_OVS_DATAPATH;
4646 }
4647
4648 static inline bool netif_is_team_master(const struct net_device *dev)
4649 {
4650 return dev->priv_flags & IFF_TEAM;
4651 }
4652
4653 static inline bool netif_is_team_port(const struct net_device *dev)
4654 {
4655 return dev->priv_flags & IFF_TEAM_PORT;
4656 }
4657
4658 static inline bool netif_is_lag_master(const struct net_device *dev)
4659 {
4660 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4661 }
4662
4663 static inline bool netif_is_lag_port(const struct net_device *dev)
4664 {
4665 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4666 }
4667
4668 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4669 {
4670 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4671 }
4672
4673 static inline bool netif_is_failover(const struct net_device *dev)
4674 {
4675 return dev->priv_flags & IFF_FAILOVER;
4676 }
4677
4678 static inline bool netif_is_failover_slave(const struct net_device *dev)
4679 {
4680 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4681 }
4682
4683 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4684 static inline void netif_keep_dst(struct net_device *dev)
4685 {
4686 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4687 }
4688
4689 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4690 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4691 {
4692 /* TODO: reserve and use an additional IFF bit, if we get more users */
4693 return dev->priv_flags & IFF_MACSEC;
4694 }
4695
4696 extern struct pernet_operations __net_initdata loopback_net_ops;
4697
4698 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4699
4700 /* netdev_printk helpers, similar to dev_printk */
4701
4702 static inline const char *netdev_name(const struct net_device *dev)
4703 {
4704 if (!dev->name[0] || strchr(dev->name, '%'))
4705 return "(unnamed net_device)";
4706 return dev->name;
4707 }
4708
4709 static inline bool netdev_unregistering(const struct net_device *dev)
4710 {
4711 return dev->reg_state == NETREG_UNREGISTERING;
4712 }
4713
4714 static inline const char *netdev_reg_state(const struct net_device *dev)
4715 {
4716 switch (dev->reg_state) {
4717 case NETREG_UNINITIALIZED: return " (uninitialized)";
4718 case NETREG_REGISTERED: return "";
4719 case NETREG_UNREGISTERING: return " (unregistering)";
4720 case NETREG_UNREGISTERED: return " (unregistered)";
4721 case NETREG_RELEASED: return " (released)";
4722 case NETREG_DUMMY: return " (dummy)";
4723 }
4724
4725 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4726 return " (unknown)";
4727 }
4728
4729 __printf(3, 4) __cold
4730 void netdev_printk(const char *level, const struct net_device *dev,
4731 const char *format, ...);
4732 __printf(2, 3) __cold
4733 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4734 __printf(2, 3) __cold
4735 void netdev_alert(const struct net_device *dev, const char *format, ...);
4736 __printf(2, 3) __cold
4737 void netdev_crit(const struct net_device *dev, const char *format, ...);
4738 __printf(2, 3) __cold
4739 void netdev_err(const struct net_device *dev, const char *format, ...);
4740 __printf(2, 3) __cold
4741 void netdev_warn(const struct net_device *dev, const char *format, ...);
4742 __printf(2, 3) __cold
4743 void netdev_notice(const struct net_device *dev, const char *format, ...);
4744 __printf(2, 3) __cold
4745 void netdev_info(const struct net_device *dev, const char *format, ...);
4746
4747 #define netdev_level_once(level, dev, fmt, ...) \
4748 do { \
4749 static bool __print_once __read_mostly; \
4750 \
4751 if (!__print_once) { \
4752 __print_once = true; \
4753 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4754 } \
4755 } while (0)
4756
4757 #define netdev_emerg_once(dev, fmt, ...) \
4758 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4759 #define netdev_alert_once(dev, fmt, ...) \
4760 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4761 #define netdev_crit_once(dev, fmt, ...) \
4762 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4763 #define netdev_err_once(dev, fmt, ...) \
4764 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4765 #define netdev_warn_once(dev, fmt, ...) \
4766 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4767 #define netdev_notice_once(dev, fmt, ...) \
4768 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4769 #define netdev_info_once(dev, fmt, ...) \
4770 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4771
4772 #define MODULE_ALIAS_NETDEV(device) \
4773 MODULE_ALIAS("netdev-" device)
4774
4775 #if defined(CONFIG_DYNAMIC_DEBUG)
4776 #define netdev_dbg(__dev, format, args...) \
4777 do { \
4778 dynamic_netdev_dbg(__dev, format, ##args); \
4779 } while (0)
4780 #elif defined(DEBUG)
4781 #define netdev_dbg(__dev, format, args...) \
4782 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4783 #else
4784 #define netdev_dbg(__dev, format, args...) \
4785 ({ \
4786 if (0) \
4787 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4788 })
4789 #endif
4790
4791 #if defined(VERBOSE_DEBUG)
4792 #define netdev_vdbg netdev_dbg
4793 #else
4794
4795 #define netdev_vdbg(dev, format, args...) \
4796 ({ \
4797 if (0) \
4798 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4799 0; \
4800 })
4801 #endif
4802
4803 /*
4804 * netdev_WARN() acts like dev_printk(), but with the key difference
4805 * of using a WARN/WARN_ON to get the message out, including the
4806 * file/line information and a backtrace.
4807 */
4808 #define netdev_WARN(dev, format, args...) \
4809 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4810 netdev_reg_state(dev), ##args)
4811
4812 #define netdev_WARN_ONCE(dev, format, args...) \
4813 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4814 netdev_reg_state(dev), ##args)
4815
4816 /* netif printk helpers, similar to netdev_printk */
4817
4818 #define netif_printk(priv, type, level, dev, fmt, args...) \
4819 do { \
4820 if (netif_msg_##type(priv)) \
4821 netdev_printk(level, (dev), fmt, ##args); \
4822 } while (0)
4823
4824 #define netif_level(level, priv, type, dev, fmt, args...) \
4825 do { \
4826 if (netif_msg_##type(priv)) \
4827 netdev_##level(dev, fmt, ##args); \
4828 } while (0)
4829
4830 #define netif_emerg(priv, type, dev, fmt, args...) \
4831 netif_level(emerg, priv, type, dev, fmt, ##args)
4832 #define netif_alert(priv, type, dev, fmt, args...) \
4833 netif_level(alert, priv, type, dev, fmt, ##args)
4834 #define netif_crit(priv, type, dev, fmt, args...) \
4835 netif_level(crit, priv, type, dev, fmt, ##args)
4836 #define netif_err(priv, type, dev, fmt, args...) \
4837 netif_level(err, priv, type, dev, fmt, ##args)
4838 #define netif_warn(priv, type, dev, fmt, args...) \
4839 netif_level(warn, priv, type, dev, fmt, ##args)
4840 #define netif_notice(priv, type, dev, fmt, args...) \
4841 netif_level(notice, priv, type, dev, fmt, ##args)
4842 #define netif_info(priv, type, dev, fmt, args...) \
4843 netif_level(info, priv, type, dev, fmt, ##args)
4844
4845 #if defined(CONFIG_DYNAMIC_DEBUG)
4846 #define netif_dbg(priv, type, netdev, format, args...) \
4847 do { \
4848 if (netif_msg_##type(priv)) \
4849 dynamic_netdev_dbg(netdev, format, ##args); \
4850 } while (0)
4851 #elif defined(DEBUG)
4852 #define netif_dbg(priv, type, dev, format, args...) \
4853 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4854 #else
4855 #define netif_dbg(priv, type, dev, format, args...) \
4856 ({ \
4857 if (0) \
4858 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4859 0; \
4860 })
4861 #endif
4862
4863 /* if @cond then downgrade to debug, else print at @level */
4864 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4865 do { \
4866 if (cond) \
4867 netif_dbg(priv, type, netdev, fmt, ##args); \
4868 else \
4869 netif_ ## level(priv, type, netdev, fmt, ##args); \
4870 } while (0)
4871
4872 #if defined(VERBOSE_DEBUG)
4873 #define netif_vdbg netif_dbg
4874 #else
4875 #define netif_vdbg(priv, type, dev, format, args...) \
4876 ({ \
4877 if (0) \
4878 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4879 0; \
4880 })
4881 #endif
4882
4883 /*
4884 * The list of packet types we will receive (as opposed to discard)
4885 * and the routines to invoke.
4886 *
4887 * Why 16. Because with 16 the only overlap we get on a hash of the
4888 * low nibble of the protocol value is RARP/SNAP/X.25.
4889 *
4890 * 0800 IP
4891 * 0001 802.3
4892 * 0002 AX.25
4893 * 0004 802.2
4894 * 8035 RARP
4895 * 0005 SNAP
4896 * 0805 X.25
4897 * 0806 ARP
4898 * 8137 IPX
4899 * 0009 Localtalk
4900 * 86DD IPv6
4901 */
4902 #define PTYPE_HASH_SIZE (16)
4903 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4904
4905 extern struct net_device *blackhole_netdev;
4906
4907 #endif /* _LINUX_NETDEVICE_H */