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 struct ip_tunnel_parm;
57 struct macsec_context;
58 struct macsec_ops;
59
60 struct sfp_bus;
61 /* 802.11 specific */
62 struct wireless_dev;
63 /* 802.15.4 specific */
64 struct wpan_dev;
65 struct mpls_dev;
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info;
68 struct udp_tunnel_nic_info;
69 struct udp_tunnel_nic;
70 struct bpf_prog;
71 struct xdp_buff;
72
73 void synchronize_net(void);
74 void netdev_set_default_ethtool_ops(struct net_device *dev,
75 const struct ethtool_ops *ops);
76
77 /* Backlog congestion levels */
78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
79 #define NET_RX_DROP 1 /* packet dropped */
80
81 #define MAX_NEST_DEV 8
82
83 /*
84 * Transmit return codes: transmit return codes originate from three different
85 * namespaces:
86 *
87 * - qdisc return codes
88 * - driver transmit return codes
89 * - errno values
90 *
91 * Drivers are allowed to return any one of those in their hard_start_xmit()
92 * function. Real network devices commonly used with qdiscs should only return
93 * the driver transmit return codes though - when qdiscs are used, the actual
94 * transmission happens asynchronously, so the value is not propagated to
95 * higher layers. Virtual network devices transmit synchronously; in this case
96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
97 * others are propagated to higher layers.
98 */
99
100 /* qdisc ->enqueue() return codes. */
101 #define NET_XMIT_SUCCESS 0x00
102 #define NET_XMIT_DROP 0x01 /* skb dropped */
103 #define NET_XMIT_CN 0x02 /* congestion notification */
104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
105
106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
107 * indicates that the device will soon be dropping packets, or already drops
108 * some packets of the same priority; prompting us to send less aggressively. */
109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
111
112 /* Driver transmit return codes */
113 #define NETDEV_TX_MASK 0xf0
114
115 enum netdev_tx {
116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
117 NETDEV_TX_OK = 0x00, /* driver took care of packet */
118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
119 };
120 typedef enum netdev_tx netdev_tx_t;
121
122 /*
123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
125 */
dev_xmit_complete(int rc)126 static inline bool dev_xmit_complete(int rc)
127 {
128 /*
129 * Positive cases with an skb consumed by a driver:
130 * - successful transmission (rc == NETDEV_TX_OK)
131 * - error while transmitting (rc < 0)
132 * - error while queueing to a different device (rc & NET_XMIT_MASK)
133 */
134 if (likely(rc < NET_XMIT_MASK))
135 return true;
136
137 return false;
138 }
139
140 /*
141 * Compute the worst-case header length according to the protocols
142 * used.
143 */
144
145 #if defined(CONFIG_HYPERV_NET)
146 # define LL_MAX_HEADER 128
147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
148 # if defined(CONFIG_MAC80211_MESH)
149 # define LL_MAX_HEADER 128
150 # else
151 # define LL_MAX_HEADER 96
152 # endif
153 #else
154 # define LL_MAX_HEADER 32
155 #endif
156
157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
159 #define MAX_HEADER LL_MAX_HEADER
160 #else
161 #define MAX_HEADER (LL_MAX_HEADER + 48)
162 #endif
163
164 /*
165 * Old network device statistics. Fields are native words
166 * (unsigned long) so they can be read and written atomically.
167 */
168
169 #define NET_DEV_STAT(FIELD) \
170 union { \
171 unsigned long FIELD; \
172 atomic_long_t __##FIELD; \
173 }
174
175 struct net_device_stats {
176 NET_DEV_STAT(rx_packets);
177 NET_DEV_STAT(tx_packets);
178 NET_DEV_STAT(rx_bytes);
179 NET_DEV_STAT(tx_bytes);
180 NET_DEV_STAT(rx_errors);
181 NET_DEV_STAT(tx_errors);
182 NET_DEV_STAT(rx_dropped);
183 NET_DEV_STAT(tx_dropped);
184 NET_DEV_STAT(multicast);
185 NET_DEV_STAT(collisions);
186 NET_DEV_STAT(rx_length_errors);
187 NET_DEV_STAT(rx_over_errors);
188 NET_DEV_STAT(rx_crc_errors);
189 NET_DEV_STAT(rx_frame_errors);
190 NET_DEV_STAT(rx_fifo_errors);
191 NET_DEV_STAT(rx_missed_errors);
192 NET_DEV_STAT(tx_aborted_errors);
193 NET_DEV_STAT(tx_carrier_errors);
194 NET_DEV_STAT(tx_fifo_errors);
195 NET_DEV_STAT(tx_heartbeat_errors);
196 NET_DEV_STAT(tx_window_errors);
197 NET_DEV_STAT(rx_compressed);
198 NET_DEV_STAT(tx_compressed);
199 };
200 #undef NET_DEV_STAT
201
202
203 #include <linux/cache.h>
204 #include <linux/skbuff.h>
205
206 #ifdef CONFIG_RPS
207 #include <linux/static_key.h>
208 extern struct static_key_false rps_needed;
209 extern struct static_key_false rfs_needed;
210 #endif
211
212 struct neighbour;
213 struct neigh_parms;
214 struct sk_buff;
215
216 struct netdev_hw_addr {
217 struct list_head list;
218 unsigned char addr[MAX_ADDR_LEN];
219 unsigned char type;
220 #define NETDEV_HW_ADDR_T_LAN 1
221 #define NETDEV_HW_ADDR_T_SAN 2
222 #define NETDEV_HW_ADDR_T_UNICAST 3
223 #define NETDEV_HW_ADDR_T_MULTICAST 4
224 bool global_use;
225 int sync_cnt;
226 int refcount;
227 int synced;
228 struct rcu_head rcu_head;
229 };
230
231 struct netdev_hw_addr_list {
232 struct list_head list;
233 int count;
234 };
235
236 #define netdev_hw_addr_list_count(l) ((l)->count)
237 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
238 #define netdev_hw_addr_list_for_each(ha, l) \
239 list_for_each_entry(ha, &(l)->list, list)
240
241 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
242 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
243 #define netdev_for_each_uc_addr(ha, dev) \
244 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
245
246 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
247 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
248 #define netdev_for_each_mc_addr(ha, dev) \
249 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
250
251 struct hh_cache {
252 unsigned int hh_len;
253 seqlock_t hh_lock;
254
255 /* cached hardware header; allow for machine alignment needs. */
256 #define HH_DATA_MOD 16
257 #define HH_DATA_OFF(__len) \
258 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
259 #define HH_DATA_ALIGN(__len) \
260 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
261 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
262 };
263
264 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
265 * Alternative is:
266 * dev->hard_header_len ? (dev->hard_header_len +
267 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
268 *
269 * We could use other alignment values, but we must maintain the
270 * relationship HH alignment <= LL alignment.
271 */
272 #define LL_RESERVED_SPACE(dev) \
273 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
274 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
275 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
276
277 struct header_ops {
278 int (*create) (struct sk_buff *skb, struct net_device *dev,
279 unsigned short type, const void *daddr,
280 const void *saddr, unsigned int len);
281 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
282 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
283 void (*cache_update)(struct hh_cache *hh,
284 const struct net_device *dev,
285 const unsigned char *haddr);
286 bool (*validate)(const char *ll_header, unsigned int len);
287 __be16 (*parse_protocol)(const struct sk_buff *skb);
288 };
289
290 /* These flag bits are private to the generic network queueing
291 * layer; they may not be explicitly referenced by any other
292 * code.
293 */
294
295 enum netdev_state_t {
296 __LINK_STATE_START,
297 __LINK_STATE_PRESENT,
298 __LINK_STATE_NOCARRIER,
299 __LINK_STATE_LINKWATCH_PENDING,
300 __LINK_STATE_DORMANT,
301 __LINK_STATE_TESTING,
302 };
303
304
305 /*
306 * This structure holds boot-time configured netdevice settings. They
307 * are then used in the device probing.
308 */
309 struct netdev_boot_setup {
310 char name[IFNAMSIZ];
311 struct ifmap map;
312 };
313 #define NETDEV_BOOT_SETUP_MAX 8
314
315 int __init netdev_boot_setup(char *str);
316
317 struct gro_list {
318 struct list_head list;
319 int count;
320 };
321
322 /*
323 * size of gro hash buckets, must less than bit number of
324 * napi_struct::gro_bitmask
325 */
326 #define GRO_HASH_BUCKETS 8
327
328 /*
329 * Structure for NAPI scheduling similar to tasklet but with weighting
330 */
331 struct napi_struct {
332 /* The poll_list must only be managed by the entity which
333 * changes the state of the NAPI_STATE_SCHED bit. This means
334 * whoever atomically sets that bit can add this napi_struct
335 * to the per-CPU poll_list, and whoever clears that bit
336 * can remove from the list right before clearing the bit.
337 */
338 struct list_head poll_list;
339
340 unsigned long state;
341 int weight;
342 int defer_hard_irqs_count;
343 unsigned long gro_bitmask;
344 int (*poll)(struct napi_struct *, int);
345 #ifdef CONFIG_NETPOLL
346 int poll_owner;
347 #endif
348 struct net_device *dev;
349 struct gro_list gro_hash[GRO_HASH_BUCKETS];
350 struct sk_buff *skb;
351 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
352 int rx_count; /* length of rx_list */
353 struct hrtimer timer;
354 struct list_head dev_list;
355 struct hlist_node napi_hash_node;
356 unsigned int napi_id;
357 };
358
359 enum {
360 NAPI_STATE_SCHED, /* Poll is scheduled */
361 NAPI_STATE_MISSED, /* reschedule a napi */
362 NAPI_STATE_DISABLE, /* Disable pending */
363 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
364 NAPI_STATE_LISTED, /* NAPI added to system lists */
365 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
366 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
367 };
368
369 enum {
370 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
371 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
372 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
373 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
374 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
375 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
376 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
377 };
378
379 enum gro_result {
380 GRO_MERGED,
381 GRO_MERGED_FREE,
382 GRO_HELD,
383 GRO_NORMAL,
384 GRO_DROP,
385 GRO_CONSUMED,
386 };
387 typedef enum gro_result gro_result_t;
388
389 /*
390 * enum rx_handler_result - Possible return values for rx_handlers.
391 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
392 * further.
393 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
394 * case skb->dev was changed by rx_handler.
395 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
396 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
397 *
398 * rx_handlers are functions called from inside __netif_receive_skb(), to do
399 * special processing of the skb, prior to delivery to protocol handlers.
400 *
401 * Currently, a net_device can only have a single rx_handler registered. Trying
402 * to register a second rx_handler will return -EBUSY.
403 *
404 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
405 * To unregister a rx_handler on a net_device, use
406 * netdev_rx_handler_unregister().
407 *
408 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
409 * do with the skb.
410 *
411 * If the rx_handler consumed the skb in some way, it should return
412 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
413 * the skb to be delivered in some other way.
414 *
415 * If the rx_handler changed skb->dev, to divert the skb to another
416 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
417 * new device will be called if it exists.
418 *
419 * If the rx_handler decides the skb should be ignored, it should return
420 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
421 * are registered on exact device (ptype->dev == skb->dev).
422 *
423 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
424 * delivered, it should return RX_HANDLER_PASS.
425 *
426 * A device without a registered rx_handler will behave as if rx_handler
427 * returned RX_HANDLER_PASS.
428 */
429
430 enum rx_handler_result {
431 RX_HANDLER_CONSUMED,
432 RX_HANDLER_ANOTHER,
433 RX_HANDLER_EXACT,
434 RX_HANDLER_PASS,
435 };
436 typedef enum rx_handler_result rx_handler_result_t;
437 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
438
439 void __napi_schedule(struct napi_struct *n);
440 void __napi_schedule_irqoff(struct napi_struct *n);
441
napi_disable_pending(struct napi_struct * n)442 static inline bool napi_disable_pending(struct napi_struct *n)
443 {
444 return test_bit(NAPI_STATE_DISABLE, &n->state);
445 }
446
447 bool napi_schedule_prep(struct napi_struct *n);
448
449 /**
450 * napi_schedule - schedule NAPI poll
451 * @n: NAPI context
452 *
453 * Schedule NAPI poll routine to be called if it is not already
454 * running.
455 */
napi_schedule(struct napi_struct * n)456 static inline void napi_schedule(struct napi_struct *n)
457 {
458 if (napi_schedule_prep(n))
459 __napi_schedule(n);
460 }
461
462 /**
463 * napi_schedule_irqoff - schedule NAPI poll
464 * @n: NAPI context
465 *
466 * Variant of napi_schedule(), assuming hard irqs are masked.
467 */
napi_schedule_irqoff(struct napi_struct * n)468 static inline void napi_schedule_irqoff(struct napi_struct *n)
469 {
470 if (napi_schedule_prep(n))
471 __napi_schedule_irqoff(n);
472 }
473
474 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
napi_reschedule(struct napi_struct * napi)475 static inline bool napi_reschedule(struct napi_struct *napi)
476 {
477 if (napi_schedule_prep(napi)) {
478 __napi_schedule(napi);
479 return true;
480 }
481 return false;
482 }
483
484 bool napi_complete_done(struct napi_struct *n, int work_done);
485 /**
486 * napi_complete - NAPI processing complete
487 * @n: NAPI context
488 *
489 * Mark NAPI processing as complete.
490 * Consider using napi_complete_done() instead.
491 * Return false if device should avoid rearming interrupts.
492 */
napi_complete(struct napi_struct * n)493 static inline bool napi_complete(struct napi_struct *n)
494 {
495 return napi_complete_done(n, 0);
496 }
497
498 /**
499 * napi_disable - prevent NAPI from scheduling
500 * @n: NAPI context
501 *
502 * Stop NAPI from being scheduled on this context.
503 * Waits till any outstanding processing completes.
504 */
505 void napi_disable(struct napi_struct *n);
506
507 /**
508 * napi_enable - enable NAPI scheduling
509 * @n: NAPI context
510 *
511 * Resume NAPI from being scheduled on this context.
512 * Must be paired with napi_disable.
513 */
napi_enable(struct napi_struct * n)514 static inline void napi_enable(struct napi_struct *n)
515 {
516 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
517 smp_mb__before_atomic();
518 clear_bit(NAPI_STATE_SCHED, &n->state);
519 clear_bit(NAPI_STATE_NPSVC, &n->state);
520 }
521
522 /**
523 * napi_synchronize - wait until NAPI is not running
524 * @n: NAPI context
525 *
526 * Wait until NAPI is done being scheduled on this context.
527 * Waits till any outstanding processing completes but
528 * does not disable future activations.
529 */
napi_synchronize(const struct napi_struct * n)530 static inline void napi_synchronize(const struct napi_struct *n)
531 {
532 if (IS_ENABLED(CONFIG_SMP))
533 while (test_bit(NAPI_STATE_SCHED, &n->state))
534 msleep(1);
535 else
536 barrier();
537 }
538
539 /**
540 * napi_if_scheduled_mark_missed - if napi is running, set the
541 * NAPIF_STATE_MISSED
542 * @n: NAPI context
543 *
544 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
545 * NAPI is scheduled.
546 **/
napi_if_scheduled_mark_missed(struct napi_struct * n)547 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
548 {
549 unsigned long val, new;
550
551 do {
552 val = READ_ONCE(n->state);
553 if (val & NAPIF_STATE_DISABLE)
554 return true;
555
556 if (!(val & NAPIF_STATE_SCHED))
557 return false;
558
559 new = val | NAPIF_STATE_MISSED;
560 } while (cmpxchg(&n->state, val, new) != val);
561
562 return true;
563 }
564
565 enum netdev_queue_state_t {
566 __QUEUE_STATE_DRV_XOFF,
567 __QUEUE_STATE_STACK_XOFF,
568 __QUEUE_STATE_FROZEN,
569 };
570
571 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
572 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
573 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
574
575 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
576 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
577 QUEUE_STATE_FROZEN)
578 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
579 QUEUE_STATE_FROZEN)
580
581 /*
582 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
583 * netif_tx_* functions below are used to manipulate this flag. The
584 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
585 * queue independently. The netif_xmit_*stopped functions below are called
586 * to check if the queue has been stopped by the driver or stack (either
587 * of the XOFF bits are set in the state). Drivers should not need to call
588 * netif_xmit*stopped functions, they should only be using netif_tx_*.
589 */
590
591 struct netdev_queue {
592 /*
593 * read-mostly part
594 */
595 struct net_device *dev;
596 struct Qdisc __rcu *qdisc;
597 struct Qdisc *qdisc_sleeping;
598 #ifdef CONFIG_SYSFS
599 struct kobject kobj;
600 #endif
601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
602 int numa_node;
603 #endif
604 unsigned long tx_maxrate;
605 /*
606 * Number of TX timeouts for this queue
607 * (/sys/class/net/DEV/Q/trans_timeout)
608 */
609 unsigned long trans_timeout;
610
611 /* Subordinate device that the queue has been assigned to */
612 struct net_device *sb_dev;
613 #ifdef CONFIG_XDP_SOCKETS
614 struct xsk_buff_pool *pool;
615 #endif
616 /*
617 * write-mostly part
618 */
619 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
620 int xmit_lock_owner;
621 /*
622 * Time (in jiffies) of last Tx
623 */
624 unsigned long trans_start;
625
626 unsigned long state;
627
628 #ifdef CONFIG_BQL
629 struct dql dql;
630 #endif
631 } ____cacheline_aligned_in_smp;
632
633 extern int sysctl_fb_tunnels_only_for_init_net;
634 extern int sysctl_devconf_inherit_init_net;
635
636 /*
637 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
638 * == 1 : For initns only
639 * == 2 : For none.
640 */
net_has_fallback_tunnels(const struct net * net)641 static inline bool net_has_fallback_tunnels(const struct net *net)
642 {
643 #if IS_ENABLED(CONFIG_SYSCTL)
644 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
645
646 return !fb_tunnels_only_for_init_net ||
647 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
648 #else
649 return true;
650 #endif
651 }
652
net_inherit_devconf(void)653 static inline int net_inherit_devconf(void)
654 {
655 #if IS_ENABLED(CONFIG_SYSCTL)
656 return READ_ONCE(sysctl_devconf_inherit_init_net);
657 #else
658 return 0;
659 #endif
660 }
661
netdev_queue_numa_node_read(const struct netdev_queue * q)662 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
663 {
664 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
665 return q->numa_node;
666 #else
667 return NUMA_NO_NODE;
668 #endif
669 }
670
netdev_queue_numa_node_write(struct netdev_queue * q,int node)671 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
672 {
673 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
674 q->numa_node = node;
675 #endif
676 }
677
678 #ifdef CONFIG_RPS
679 /*
680 * This structure holds an RPS map which can be of variable length. The
681 * map is an array of CPUs.
682 */
683 struct rps_map {
684 unsigned int len;
685 struct rcu_head rcu;
686 u16 cpus[];
687 };
688 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
689
690 /*
691 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
692 * tail pointer for that CPU's input queue at the time of last enqueue, and
693 * a hardware filter index.
694 */
695 struct rps_dev_flow {
696 u16 cpu;
697 u16 filter;
698 unsigned int last_qtail;
699 };
700 #define RPS_NO_FILTER 0xffff
701
702 /*
703 * The rps_dev_flow_table structure contains a table of flow mappings.
704 */
705 struct rps_dev_flow_table {
706 unsigned int mask;
707 struct rcu_head rcu;
708 struct rps_dev_flow flows[];
709 };
710 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
711 ((_num) * sizeof(struct rps_dev_flow)))
712
713 /*
714 * The rps_sock_flow_table contains mappings of flows to the last CPU
715 * on which they were processed by the application (set in recvmsg).
716 * Each entry is a 32bit value. Upper part is the high-order bits
717 * of flow hash, lower part is CPU number.
718 * rps_cpu_mask is used to partition the space, depending on number of
719 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
720 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
721 * meaning we use 32-6=26 bits for the hash.
722 */
723 struct rps_sock_flow_table {
724 u32 mask;
725
726 u32 ents[] ____cacheline_aligned_in_smp;
727 };
728 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
729
730 #define RPS_NO_CPU 0xffff
731
732 extern u32 rps_cpu_mask;
733 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
734
rps_record_sock_flow(struct rps_sock_flow_table * table,u32 hash)735 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
736 u32 hash)
737 {
738 if (table && hash) {
739 unsigned int index = hash & table->mask;
740 u32 val = hash & ~rps_cpu_mask;
741
742 /* We only give a hint, preemption can change CPU under us */
743 val |= raw_smp_processor_id();
744
745 if (table->ents[index] != val)
746 table->ents[index] = val;
747 }
748 }
749
750 #ifdef CONFIG_RFS_ACCEL
751 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
752 u16 filter_id);
753 #endif
754 #endif /* CONFIG_RPS */
755
756 /* This structure contains an instance of an RX queue. */
757 struct netdev_rx_queue {
758 #ifdef CONFIG_RPS
759 struct rps_map __rcu *rps_map;
760 struct rps_dev_flow_table __rcu *rps_flow_table;
761 #endif
762 struct kobject kobj;
763 struct net_device *dev;
764 struct xdp_rxq_info xdp_rxq;
765 #ifdef CONFIG_XDP_SOCKETS
766 struct xsk_buff_pool *pool;
767 #endif
768 } ____cacheline_aligned_in_smp;
769
770 /*
771 * RX queue sysfs structures and functions.
772 */
773 struct rx_queue_attribute {
774 struct attribute attr;
775 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
776 ssize_t (*store)(struct netdev_rx_queue *queue,
777 const char *buf, size_t len);
778 };
779
780 #ifdef CONFIG_XPS
781 /*
782 * This structure holds an XPS map which can be of variable length. The
783 * map is an array of queues.
784 */
785 struct xps_map {
786 unsigned int len;
787 unsigned int alloc_len;
788 struct rcu_head rcu;
789 u16 queues[];
790 };
791 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
792 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
793 - sizeof(struct xps_map)) / sizeof(u16))
794
795 /*
796 * This structure holds all XPS maps for device. Maps are indexed by CPU.
797 */
798 struct xps_dev_maps {
799 struct rcu_head rcu;
800 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
801 };
802
803 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
804 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
805
806 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
807 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
808
809 #endif /* CONFIG_XPS */
810
811 #define TC_MAX_QUEUE 16
812 #define TC_BITMASK 15
813 /* HW offloaded queuing disciplines txq count and offset maps */
814 struct netdev_tc_txq {
815 u16 count;
816 u16 offset;
817 };
818
819 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
820 /*
821 * This structure is to hold information about the device
822 * configured to run FCoE protocol stack.
823 */
824 struct netdev_fcoe_hbainfo {
825 char manufacturer[64];
826 char serial_number[64];
827 char hardware_version[64];
828 char driver_version[64];
829 char optionrom_version[64];
830 char firmware_version[64];
831 char model[256];
832 char model_description[256];
833 };
834 #endif
835
836 #define MAX_PHYS_ITEM_ID_LEN 32
837
838 /* This structure holds a unique identifier to identify some
839 * physical item (port for example) used by a netdevice.
840 */
841 struct netdev_phys_item_id {
842 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
843 unsigned char id_len;
844 };
845
netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)846 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
847 struct netdev_phys_item_id *b)
848 {
849 return a->id_len == b->id_len &&
850 memcmp(a->id, b->id, a->id_len) == 0;
851 }
852
853 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
854 struct sk_buff *skb,
855 struct net_device *sb_dev);
856
857 enum tc_setup_type {
858 TC_SETUP_QDISC_MQPRIO,
859 TC_SETUP_CLSU32,
860 TC_SETUP_CLSFLOWER,
861 TC_SETUP_CLSMATCHALL,
862 TC_SETUP_CLSBPF,
863 TC_SETUP_BLOCK,
864 TC_SETUP_QDISC_CBS,
865 TC_SETUP_QDISC_RED,
866 TC_SETUP_QDISC_PRIO,
867 TC_SETUP_QDISC_MQ,
868 TC_SETUP_QDISC_ETF,
869 TC_SETUP_ROOT_QDISC,
870 TC_SETUP_QDISC_GRED,
871 TC_SETUP_QDISC_TAPRIO,
872 TC_SETUP_FT,
873 TC_SETUP_QDISC_ETS,
874 TC_SETUP_QDISC_TBF,
875 TC_SETUP_QDISC_FIFO,
876 };
877
878 /* These structures hold the attributes of bpf state that are being passed
879 * to the netdevice through the bpf op.
880 */
881 enum bpf_netdev_command {
882 /* Set or clear a bpf program used in the earliest stages of packet
883 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
884 * is responsible for calling bpf_prog_put on any old progs that are
885 * stored. In case of error, the callee need not release the new prog
886 * reference, but on success it takes ownership and must bpf_prog_put
887 * when it is no longer used.
888 */
889 XDP_SETUP_PROG,
890 XDP_SETUP_PROG_HW,
891 /* BPF program for offload callbacks, invoked at program load time. */
892 BPF_OFFLOAD_MAP_ALLOC,
893 BPF_OFFLOAD_MAP_FREE,
894 XDP_SETUP_XSK_POOL,
895 };
896
897 struct bpf_prog_offload_ops;
898 struct netlink_ext_ack;
899 struct xdp_umem;
900 struct xdp_dev_bulk_queue;
901 struct bpf_xdp_link;
902
903 enum bpf_xdp_mode {
904 XDP_MODE_SKB = 0,
905 XDP_MODE_DRV = 1,
906 XDP_MODE_HW = 2,
907 __MAX_XDP_MODE
908 };
909
910 struct bpf_xdp_entity {
911 struct bpf_prog *prog;
912 struct bpf_xdp_link *link;
913 };
914
915 struct netdev_bpf {
916 enum bpf_netdev_command command;
917 union {
918 /* XDP_SETUP_PROG */
919 struct {
920 u32 flags;
921 struct bpf_prog *prog;
922 struct netlink_ext_ack *extack;
923 };
924 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
925 struct {
926 struct bpf_offloaded_map *offmap;
927 };
928 /* XDP_SETUP_XSK_POOL */
929 struct {
930 struct xsk_buff_pool *pool;
931 u16 queue_id;
932 } xsk;
933 };
934 };
935
936 /* Flags for ndo_xsk_wakeup. */
937 #define XDP_WAKEUP_RX (1 << 0)
938 #define XDP_WAKEUP_TX (1 << 1)
939
940 #ifdef CONFIG_XFRM_OFFLOAD
941 struct xfrmdev_ops {
942 int (*xdo_dev_state_add) (struct xfrm_state *x);
943 void (*xdo_dev_state_delete) (struct xfrm_state *x);
944 void (*xdo_dev_state_free) (struct xfrm_state *x);
945 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
946 struct xfrm_state *x);
947 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
948 };
949 #endif
950
951 struct dev_ifalias {
952 struct rcu_head rcuhead;
953 char ifalias[];
954 };
955
956 struct devlink;
957 struct tlsdev_ops;
958
959 struct netdev_name_node {
960 struct hlist_node hlist;
961 struct list_head list;
962 struct net_device *dev;
963 const char *name;
964 };
965
966 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
967 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
968
969 struct netdev_net_notifier {
970 struct list_head list;
971 struct notifier_block *nb;
972 };
973
974 /*
975 * This structure defines the management hooks for network devices.
976 * The following hooks can be defined; unless noted otherwise, they are
977 * optional and can be filled with a null pointer.
978 *
979 * int (*ndo_init)(struct net_device *dev);
980 * This function is called once when a network device is registered.
981 * The network device can use this for any late stage initialization
982 * or semantic validation. It can fail with an error code which will
983 * be propagated back to register_netdev.
984 *
985 * void (*ndo_uninit)(struct net_device *dev);
986 * This function is called when device is unregistered or when registration
987 * fails. It is not called if init fails.
988 *
989 * int (*ndo_open)(struct net_device *dev);
990 * This function is called when a network device transitions to the up
991 * state.
992 *
993 * int (*ndo_stop)(struct net_device *dev);
994 * This function is called when a network device transitions to the down
995 * state.
996 *
997 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
998 * struct net_device *dev);
999 * Called when a packet needs to be transmitted.
1000 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1001 * the queue before that can happen; it's for obsolete devices and weird
1002 * corner cases, but the stack really does a non-trivial amount
1003 * of useless work if you return NETDEV_TX_BUSY.
1004 * Required; cannot be NULL.
1005 *
1006 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1007 * struct net_device *dev
1008 * netdev_features_t features);
1009 * Called by core transmit path to determine if device is capable of
1010 * performing offload operations on a given packet. This is to give
1011 * the device an opportunity to implement any restrictions that cannot
1012 * be otherwise expressed by feature flags. The check is called with
1013 * the set of features that the stack has calculated and it returns
1014 * those the driver believes to be appropriate.
1015 *
1016 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1017 * struct net_device *sb_dev);
1018 * Called to decide which queue to use when device supports multiple
1019 * transmit queues.
1020 *
1021 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1022 * This function is called to allow device receiver to make
1023 * changes to configuration when multicast or promiscuous is enabled.
1024 *
1025 * void (*ndo_set_rx_mode)(struct net_device *dev);
1026 * This function is called device changes address list filtering.
1027 * If driver handles unicast address filtering, it should set
1028 * IFF_UNICAST_FLT in its priv_flags.
1029 *
1030 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1031 * This function is called when the Media Access Control address
1032 * needs to be changed. If this interface is not defined, the
1033 * MAC address can not be changed.
1034 *
1035 * int (*ndo_validate_addr)(struct net_device *dev);
1036 * Test if Media Access Control address is valid for the device.
1037 *
1038 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1039 * Called when a user requests an ioctl which can't be handled by
1040 * the generic interface code. If not defined ioctls return
1041 * not supported error code.
1042 *
1043 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1044 * Used to set network devices bus interface parameters. This interface
1045 * is retained for legacy reasons; new devices should use the bus
1046 * interface (PCI) for low level management.
1047 *
1048 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1049 * Called when a user wants to change the Maximum Transfer Unit
1050 * of a device.
1051 *
1052 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1053 * Callback used when the transmitter has not made any progress
1054 * for dev->watchdog ticks.
1055 *
1056 * void (*ndo_get_stats64)(struct net_device *dev,
1057 * struct rtnl_link_stats64 *storage);
1058 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1059 * Called when a user wants to get the network device usage
1060 * statistics. Drivers must do one of the following:
1061 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1062 * rtnl_link_stats64 structure passed by the caller.
1063 * 2. Define @ndo_get_stats to update a net_device_stats structure
1064 * (which should normally be dev->stats) and return a pointer to
1065 * it. The structure may be changed asynchronously only if each
1066 * field is written atomically.
1067 * 3. Update dev->stats asynchronously and atomically, and define
1068 * neither operation.
1069 *
1070 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1071 * Return true if this device supports offload stats of this attr_id.
1072 *
1073 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1074 * void *attr_data)
1075 * Get statistics for offload operations by attr_id. Write it into the
1076 * attr_data pointer.
1077 *
1078 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1079 * If device supports VLAN filtering this function is called when a
1080 * VLAN id is registered.
1081 *
1082 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1083 * If device supports VLAN filtering this function is called when a
1084 * VLAN id is unregistered.
1085 *
1086 * void (*ndo_poll_controller)(struct net_device *dev);
1087 *
1088 * SR-IOV management functions.
1089 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1090 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1091 * u8 qos, __be16 proto);
1092 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1093 * int max_tx_rate);
1094 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1095 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1096 * int (*ndo_get_vf_config)(struct net_device *dev,
1097 * int vf, struct ifla_vf_info *ivf);
1098 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1099 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1100 * struct nlattr *port[]);
1101 *
1102 * Enable or disable the VF ability to query its RSS Redirection Table and
1103 * Hash Key. This is needed since on some devices VF share this information
1104 * with PF and querying it may introduce a theoretical security risk.
1105 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1106 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1107 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1108 * void *type_data);
1109 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1110 * This is always called from the stack with the rtnl lock held and netif
1111 * tx queues stopped. This allows the netdevice to perform queue
1112 * management safely.
1113 *
1114 * Fiber Channel over Ethernet (FCoE) offload functions.
1115 * int (*ndo_fcoe_enable)(struct net_device *dev);
1116 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1117 * so the underlying device can perform whatever needed configuration or
1118 * initialization to support acceleration of FCoE traffic.
1119 *
1120 * int (*ndo_fcoe_disable)(struct net_device *dev);
1121 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1122 * so the underlying device can perform whatever needed clean-ups to
1123 * stop supporting acceleration of FCoE traffic.
1124 *
1125 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1126 * struct scatterlist *sgl, unsigned int sgc);
1127 * Called when the FCoE Initiator wants to initialize an I/O that
1128 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1129 * perform necessary setup and returns 1 to indicate the device is set up
1130 * successfully to perform DDP on this I/O, otherwise this returns 0.
1131 *
1132 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1133 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1134 * indicated by the FC exchange id 'xid', so the underlying device can
1135 * clean up and reuse resources for later DDP requests.
1136 *
1137 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1138 * struct scatterlist *sgl, unsigned int sgc);
1139 * Called when the FCoE Target wants to initialize an I/O that
1140 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1141 * perform necessary setup and returns 1 to indicate the device is set up
1142 * successfully to perform DDP on this I/O, otherwise this returns 0.
1143 *
1144 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1145 * struct netdev_fcoe_hbainfo *hbainfo);
1146 * Called when the FCoE Protocol stack wants information on the underlying
1147 * device. This information is utilized by the FCoE protocol stack to
1148 * register attributes with Fiber Channel management service as per the
1149 * FC-GS Fabric Device Management Information(FDMI) specification.
1150 *
1151 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1152 * Called when the underlying device wants to override default World Wide
1153 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1154 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1155 * protocol stack to use.
1156 *
1157 * RFS acceleration.
1158 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1159 * u16 rxq_index, u32 flow_id);
1160 * Set hardware filter for RFS. rxq_index is the target queue index;
1161 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1162 * Return the filter ID on success, or a negative error code.
1163 *
1164 * Slave management functions (for bridge, bonding, etc).
1165 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1166 * Called to make another netdev an underling.
1167 *
1168 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1169 * Called to release previously enslaved netdev.
1170 *
1171 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1172 * struct sk_buff *skb,
1173 * bool all_slaves);
1174 * Get the xmit slave of master device. If all_slaves is true, function
1175 * assume all the slaves can transmit.
1176 *
1177 * Feature/offload setting functions.
1178 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1179 * netdev_features_t features);
1180 * Adjusts the requested feature flags according to device-specific
1181 * constraints, and returns the resulting flags. Must not modify
1182 * the device state.
1183 *
1184 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1185 * Called to update device configuration to new features. Passed
1186 * feature set might be less than what was returned by ndo_fix_features()).
1187 * Must return >0 or -errno if it changed dev->features itself.
1188 *
1189 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1190 * struct net_device *dev,
1191 * const unsigned char *addr, u16 vid, u16 flags,
1192 * struct netlink_ext_ack *extack);
1193 * Adds an FDB entry to dev for addr.
1194 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1195 * struct net_device *dev,
1196 * const unsigned char *addr, u16 vid)
1197 * Deletes the FDB entry from dev coresponding to addr.
1198 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1199 * struct net_device *dev, struct net_device *filter_dev,
1200 * int *idx)
1201 * Used to add FDB entries to dump requests. Implementers should add
1202 * entries to skb and update idx with the number of entries.
1203 *
1204 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1205 * u16 flags, struct netlink_ext_ack *extack)
1206 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1207 * struct net_device *dev, u32 filter_mask,
1208 * int nlflags)
1209 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1210 * u16 flags);
1211 *
1212 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1213 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1214 * which do not represent real hardware may define this to allow their
1215 * userspace components to manage their virtual carrier state. Devices
1216 * that determine carrier state from physical hardware properties (eg
1217 * network cables) or protocol-dependent mechanisms (eg
1218 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1219 *
1220 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1221 * struct netdev_phys_item_id *ppid);
1222 * Called to get ID of physical port of this device. If driver does
1223 * not implement this, it is assumed that the hw is not able to have
1224 * multiple net devices on single physical port.
1225 *
1226 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1227 * struct netdev_phys_item_id *ppid)
1228 * Called to get the parent ID of the physical port of this device.
1229 *
1230 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1231 * struct udp_tunnel_info *ti);
1232 * Called by UDP tunnel to notify a driver about the UDP port and socket
1233 * address family that a UDP tunnel is listnening to. It is called only
1234 * when a new port starts listening. The operation is protected by the
1235 * RTNL.
1236 *
1237 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1238 * struct udp_tunnel_info *ti);
1239 * Called by UDP tunnel to notify the driver about a UDP port and socket
1240 * address family that the UDP tunnel is not listening to anymore. The
1241 * operation is protected by the RTNL.
1242 *
1243 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1244 * struct net_device *dev)
1245 * Called by upper layer devices to accelerate switching or other
1246 * station functionality into hardware. 'pdev is the lowerdev
1247 * to use for the offload and 'dev' is the net device that will
1248 * back the offload. Returns a pointer to the private structure
1249 * the upper layer will maintain.
1250 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1251 * Called by upper layer device to delete the station created
1252 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1253 * the station and priv is the structure returned by the add
1254 * operation.
1255 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1256 * int queue_index, u32 maxrate);
1257 * Called when a user wants to set a max-rate limitation of specific
1258 * TX queue.
1259 * int (*ndo_get_iflink)(const struct net_device *dev);
1260 * Called to get the iflink value of this device.
1261 * void (*ndo_change_proto_down)(struct net_device *dev,
1262 * bool proto_down);
1263 * This function is used to pass protocol port error state information
1264 * to the switch driver. The switch driver can react to the proto_down
1265 * by doing a phys down on the associated switch port.
1266 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1267 * This function is used to get egress tunnel information for given skb.
1268 * This is useful for retrieving outer tunnel header parameters while
1269 * sampling packet.
1270 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1271 * This function is used to specify the headroom that the skb must
1272 * consider when allocation skb during packet reception. Setting
1273 * appropriate rx headroom value allows avoiding skb head copy on
1274 * forward. Setting a negative value resets the rx headroom to the
1275 * default value.
1276 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1277 * This function is used to set or query state related to XDP on the
1278 * netdevice and manage BPF offload. See definition of
1279 * enum bpf_netdev_command for details.
1280 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1281 * u32 flags);
1282 * This function is used to submit @n XDP packets for transmit on a
1283 * netdevice. Returns number of frames successfully transmitted, frames
1284 * that got dropped are freed/returned via xdp_return_frame().
1285 * Returns negative number, means general error invoking ndo, meaning
1286 * no frames were xmit'ed and core-caller will free all frames.
1287 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1288 * This function is used to wake up the softirq, ksoftirqd or kthread
1289 * responsible for sending and/or receiving packets on a specific
1290 * queue id bound to an AF_XDP socket. The flags field specifies if
1291 * only RX, only Tx, or both should be woken up using the flags
1292 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1293 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1294 * Get devlink port instance associated with a given netdev.
1295 * Called with a reference on the netdevice and devlink locks only,
1296 * rtnl_lock is not held.
1297 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1298 * int cmd);
1299 * Add, change, delete or get information on an IPv4 tunnel.
1300 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1301 * If a device is paired with a peer device, return the peer instance.
1302 * The caller must be under RCU read context.
1303 */
1304 struct net_device_ops {
1305 int (*ndo_init)(struct net_device *dev);
1306 void (*ndo_uninit)(struct net_device *dev);
1307 int (*ndo_open)(struct net_device *dev);
1308 int (*ndo_stop)(struct net_device *dev);
1309 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1310 struct net_device *dev);
1311 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1312 struct net_device *dev,
1313 netdev_features_t features);
1314 u16 (*ndo_select_queue)(struct net_device *dev,
1315 struct sk_buff *skb,
1316 struct net_device *sb_dev);
1317 void (*ndo_change_rx_flags)(struct net_device *dev,
1318 int flags);
1319 void (*ndo_set_rx_mode)(struct net_device *dev);
1320 int (*ndo_set_mac_address)(struct net_device *dev,
1321 void *addr);
1322 int (*ndo_validate_addr)(struct net_device *dev);
1323 int (*ndo_do_ioctl)(struct net_device *dev,
1324 struct ifreq *ifr, int cmd);
1325 int (*ndo_set_config)(struct net_device *dev,
1326 struct ifmap *map);
1327 int (*ndo_change_mtu)(struct net_device *dev,
1328 int new_mtu);
1329 int (*ndo_neigh_setup)(struct net_device *dev,
1330 struct neigh_parms *);
1331 void (*ndo_tx_timeout) (struct net_device *dev,
1332 unsigned int txqueue);
1333
1334 void (*ndo_get_stats64)(struct net_device *dev,
1335 struct rtnl_link_stats64 *storage);
1336 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1337 int (*ndo_get_offload_stats)(int attr_id,
1338 const struct net_device *dev,
1339 void *attr_data);
1340 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1341
1342 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1343 __be16 proto, u16 vid);
1344 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1345 __be16 proto, u16 vid);
1346 #ifdef CONFIG_NET_POLL_CONTROLLER
1347 void (*ndo_poll_controller)(struct net_device *dev);
1348 int (*ndo_netpoll_setup)(struct net_device *dev,
1349 struct netpoll_info *info);
1350 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1351 #endif
1352 int (*ndo_set_vf_mac)(struct net_device *dev,
1353 int queue, u8 *mac);
1354 int (*ndo_set_vf_vlan)(struct net_device *dev,
1355 int queue, u16 vlan,
1356 u8 qos, __be16 proto);
1357 int (*ndo_set_vf_rate)(struct net_device *dev,
1358 int vf, int min_tx_rate,
1359 int max_tx_rate);
1360 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1361 int vf, bool setting);
1362 int (*ndo_set_vf_trust)(struct net_device *dev,
1363 int vf, bool setting);
1364 int (*ndo_get_vf_config)(struct net_device *dev,
1365 int vf,
1366 struct ifla_vf_info *ivf);
1367 int (*ndo_set_vf_link_state)(struct net_device *dev,
1368 int vf, int link_state);
1369 int (*ndo_get_vf_stats)(struct net_device *dev,
1370 int vf,
1371 struct ifla_vf_stats
1372 *vf_stats);
1373 int (*ndo_set_vf_port)(struct net_device *dev,
1374 int vf,
1375 struct nlattr *port[]);
1376 int (*ndo_get_vf_port)(struct net_device *dev,
1377 int vf, struct sk_buff *skb);
1378 int (*ndo_get_vf_guid)(struct net_device *dev,
1379 int vf,
1380 struct ifla_vf_guid *node_guid,
1381 struct ifla_vf_guid *port_guid);
1382 int (*ndo_set_vf_guid)(struct net_device *dev,
1383 int vf, u64 guid,
1384 int guid_type);
1385 int (*ndo_set_vf_rss_query_en)(
1386 struct net_device *dev,
1387 int vf, bool setting);
1388 int (*ndo_setup_tc)(struct net_device *dev,
1389 enum tc_setup_type type,
1390 void *type_data);
1391 #if IS_ENABLED(CONFIG_FCOE)
1392 int (*ndo_fcoe_enable)(struct net_device *dev);
1393 int (*ndo_fcoe_disable)(struct net_device *dev);
1394 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1395 u16 xid,
1396 struct scatterlist *sgl,
1397 unsigned int sgc);
1398 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1399 u16 xid);
1400 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1401 u16 xid,
1402 struct scatterlist *sgl,
1403 unsigned int sgc);
1404 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1405 struct netdev_fcoe_hbainfo *hbainfo);
1406 #endif
1407
1408 #if IS_ENABLED(CONFIG_LIBFCOE)
1409 #define NETDEV_FCOE_WWNN 0
1410 #define NETDEV_FCOE_WWPN 1
1411 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1412 u64 *wwn, int type);
1413 #endif
1414
1415 #ifdef CONFIG_RFS_ACCEL
1416 int (*ndo_rx_flow_steer)(struct net_device *dev,
1417 const struct sk_buff *skb,
1418 u16 rxq_index,
1419 u32 flow_id);
1420 #endif
1421 int (*ndo_add_slave)(struct net_device *dev,
1422 struct net_device *slave_dev,
1423 struct netlink_ext_ack *extack);
1424 int (*ndo_del_slave)(struct net_device *dev,
1425 struct net_device *slave_dev);
1426 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1427 struct sk_buff *skb,
1428 bool all_slaves);
1429 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1430 netdev_features_t features);
1431 int (*ndo_set_features)(struct net_device *dev,
1432 netdev_features_t features);
1433 int (*ndo_neigh_construct)(struct net_device *dev,
1434 struct neighbour *n);
1435 void (*ndo_neigh_destroy)(struct net_device *dev,
1436 struct neighbour *n);
1437
1438 int (*ndo_fdb_add)(struct ndmsg *ndm,
1439 struct nlattr *tb[],
1440 struct net_device *dev,
1441 const unsigned char *addr,
1442 u16 vid,
1443 u16 flags,
1444 struct netlink_ext_ack *extack);
1445 int (*ndo_fdb_del)(struct ndmsg *ndm,
1446 struct nlattr *tb[],
1447 struct net_device *dev,
1448 const unsigned char *addr,
1449 u16 vid);
1450 int (*ndo_fdb_dump)(struct sk_buff *skb,
1451 struct netlink_callback *cb,
1452 struct net_device *dev,
1453 struct net_device *filter_dev,
1454 int *idx);
1455 int (*ndo_fdb_get)(struct sk_buff *skb,
1456 struct nlattr *tb[],
1457 struct net_device *dev,
1458 const unsigned char *addr,
1459 u16 vid, u32 portid, u32 seq,
1460 struct netlink_ext_ack *extack);
1461 int (*ndo_bridge_setlink)(struct net_device *dev,
1462 struct nlmsghdr *nlh,
1463 u16 flags,
1464 struct netlink_ext_ack *extack);
1465 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1466 u32 pid, u32 seq,
1467 struct net_device *dev,
1468 u32 filter_mask,
1469 int nlflags);
1470 int (*ndo_bridge_dellink)(struct net_device *dev,
1471 struct nlmsghdr *nlh,
1472 u16 flags);
1473 int (*ndo_change_carrier)(struct net_device *dev,
1474 bool new_carrier);
1475 int (*ndo_get_phys_port_id)(struct net_device *dev,
1476 struct netdev_phys_item_id *ppid);
1477 int (*ndo_get_port_parent_id)(struct net_device *dev,
1478 struct netdev_phys_item_id *ppid);
1479 int (*ndo_get_phys_port_name)(struct net_device *dev,
1480 char *name, size_t len);
1481 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1482 struct udp_tunnel_info *ti);
1483 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1484 struct udp_tunnel_info *ti);
1485 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1486 struct net_device *dev);
1487 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1488 void *priv);
1489
1490 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1491 int queue_index,
1492 u32 maxrate);
1493 int (*ndo_get_iflink)(const struct net_device *dev);
1494 int (*ndo_change_proto_down)(struct net_device *dev,
1495 bool proto_down);
1496 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1497 struct sk_buff *skb);
1498 void (*ndo_set_rx_headroom)(struct net_device *dev,
1499 int needed_headroom);
1500 int (*ndo_bpf)(struct net_device *dev,
1501 struct netdev_bpf *bpf);
1502 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1503 struct xdp_frame **xdp,
1504 u32 flags);
1505 int (*ndo_xsk_wakeup)(struct net_device *dev,
1506 u32 queue_id, u32 flags);
1507 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1508 int (*ndo_tunnel_ctl)(struct net_device *dev,
1509 struct ip_tunnel_parm *p, int cmd);
1510 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1511 };
1512
1513 /**
1514 * enum net_device_priv_flags - &struct net_device priv_flags
1515 *
1516 * These are the &struct net_device, they are only set internally
1517 * by drivers and used in the kernel. These flags are invisible to
1518 * userspace; this means that the order of these flags can change
1519 * during any kernel release.
1520 *
1521 * You should have a pretty good reason to be extending these flags.
1522 *
1523 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1524 * @IFF_EBRIDGE: Ethernet bridging device
1525 * @IFF_BONDING: bonding master or slave
1526 * @IFF_ISATAP: ISATAP interface (RFC4214)
1527 * @IFF_WAN_HDLC: WAN HDLC device
1528 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1529 * release skb->dst
1530 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1531 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1532 * @IFF_MACVLAN_PORT: device used as macvlan port
1533 * @IFF_BRIDGE_PORT: device used as bridge port
1534 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1535 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1536 * @IFF_UNICAST_FLT: Supports unicast filtering
1537 * @IFF_TEAM_PORT: device used as team port
1538 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1539 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1540 * change when it's running
1541 * @IFF_MACVLAN: Macvlan device
1542 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1543 * underlying stacked devices
1544 * @IFF_L3MDEV_MASTER: device is an L3 master device
1545 * @IFF_NO_QUEUE: device can run without qdisc attached
1546 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1547 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1548 * @IFF_TEAM: device is a team device
1549 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1550 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1551 * entity (i.e. the master device for bridged veth)
1552 * @IFF_MACSEC: device is a MACsec device
1553 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1554 * @IFF_FAILOVER: device is a failover master device
1555 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1556 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1557 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1558 */
1559 enum netdev_priv_flags {
1560 IFF_802_1Q_VLAN = 1<<0,
1561 IFF_EBRIDGE = 1<<1,
1562 IFF_BONDING = 1<<2,
1563 IFF_ISATAP = 1<<3,
1564 IFF_WAN_HDLC = 1<<4,
1565 IFF_XMIT_DST_RELEASE = 1<<5,
1566 IFF_DONT_BRIDGE = 1<<6,
1567 IFF_DISABLE_NETPOLL = 1<<7,
1568 IFF_MACVLAN_PORT = 1<<8,
1569 IFF_BRIDGE_PORT = 1<<9,
1570 IFF_OVS_DATAPATH = 1<<10,
1571 IFF_TX_SKB_SHARING = 1<<11,
1572 IFF_UNICAST_FLT = 1<<12,
1573 IFF_TEAM_PORT = 1<<13,
1574 IFF_SUPP_NOFCS = 1<<14,
1575 IFF_LIVE_ADDR_CHANGE = 1<<15,
1576 IFF_MACVLAN = 1<<16,
1577 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1578 IFF_L3MDEV_MASTER = 1<<18,
1579 IFF_NO_QUEUE = 1<<19,
1580 IFF_OPENVSWITCH = 1<<20,
1581 IFF_L3MDEV_SLAVE = 1<<21,
1582 IFF_TEAM = 1<<22,
1583 IFF_RXFH_CONFIGURED = 1<<23,
1584 IFF_PHONY_HEADROOM = 1<<24,
1585 IFF_MACSEC = 1<<25,
1586 IFF_NO_RX_HANDLER = 1<<26,
1587 IFF_FAILOVER = 1<<27,
1588 IFF_FAILOVER_SLAVE = 1<<28,
1589 IFF_L3MDEV_RX_HANDLER = 1<<29,
1590 IFF_LIVE_RENAME_OK = 1<<30,
1591 };
1592
1593 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1594 #define IFF_EBRIDGE IFF_EBRIDGE
1595 #define IFF_BONDING IFF_BONDING
1596 #define IFF_ISATAP IFF_ISATAP
1597 #define IFF_WAN_HDLC IFF_WAN_HDLC
1598 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1599 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1600 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1601 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1602 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1603 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1604 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1605 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1606 #define IFF_TEAM_PORT IFF_TEAM_PORT
1607 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1608 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1609 #define IFF_MACVLAN IFF_MACVLAN
1610 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1611 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1612 #define IFF_NO_QUEUE IFF_NO_QUEUE
1613 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1614 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1615 #define IFF_TEAM IFF_TEAM
1616 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1617 #define IFF_MACSEC IFF_MACSEC
1618 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1619 #define IFF_FAILOVER IFF_FAILOVER
1620 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1621 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1622 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1623
1624 /* Specifies the type of the struct net_device::ml_priv pointer */
1625 enum netdev_ml_priv_type {
1626 ML_PRIV_NONE,
1627 ML_PRIV_CAN,
1628 };
1629
1630 /**
1631 * struct net_device - The DEVICE structure.
1632 *
1633 * Actually, this whole structure is a big mistake. It mixes I/O
1634 * data with strictly "high-level" data, and it has to know about
1635 * almost every data structure used in the INET module.
1636 *
1637 * @name: This is the first field of the "visible" part of this structure
1638 * (i.e. as seen by users in the "Space.c" file). It is the name
1639 * of the interface.
1640 *
1641 * @name_node: Name hashlist node
1642 * @ifalias: SNMP alias
1643 * @mem_end: Shared memory end
1644 * @mem_start: Shared memory start
1645 * @base_addr: Device I/O address
1646 * @irq: Device IRQ number
1647 *
1648 * @state: Generic network queuing layer state, see netdev_state_t
1649 * @dev_list: The global list of network devices
1650 * @napi_list: List entry used for polling NAPI devices
1651 * @unreg_list: List entry when we are unregistering the
1652 * device; see the function unregister_netdev
1653 * @close_list: List entry used when we are closing the device
1654 * @ptype_all: Device-specific packet handlers for all protocols
1655 * @ptype_specific: Device-specific, protocol-specific packet handlers
1656 *
1657 * @adj_list: Directly linked devices, like slaves for bonding
1658 * @features: Currently active device features
1659 * @hw_features: User-changeable features
1660 *
1661 * @wanted_features: User-requested features
1662 * @vlan_features: Mask of features inheritable by VLAN devices
1663 *
1664 * @hw_enc_features: Mask of features inherited by encapsulating devices
1665 * This field indicates what encapsulation
1666 * offloads the hardware is capable of doing,
1667 * and drivers will need to set them appropriately.
1668 *
1669 * @mpls_features: Mask of features inheritable by MPLS
1670 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1671 *
1672 * @ifindex: interface index
1673 * @group: The group the device belongs to
1674 *
1675 * @stats: Statistics struct, which was left as a legacy, use
1676 * rtnl_link_stats64 instead
1677 *
1678 * @rx_dropped: Dropped packets by core network,
1679 * do not use this in drivers
1680 * @tx_dropped: Dropped packets by core network,
1681 * do not use this in drivers
1682 * @rx_nohandler: nohandler dropped packets by core network on
1683 * inactive devices, do not use this in drivers
1684 * @carrier_up_count: Number of times the carrier has been up
1685 * @carrier_down_count: Number of times the carrier has been down
1686 *
1687 * @wireless_handlers: List of functions to handle Wireless Extensions,
1688 * instead of ioctl,
1689 * see <net/iw_handler.h> for details.
1690 * @wireless_data: Instance data managed by the core of wireless extensions
1691 *
1692 * @netdev_ops: Includes several pointers to callbacks,
1693 * if one wants to override the ndo_*() functions
1694 * @ethtool_ops: Management operations
1695 * @l3mdev_ops: Layer 3 master device operations
1696 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1697 * discovery handling. Necessary for e.g. 6LoWPAN.
1698 * @xfrmdev_ops: Transformation offload operations
1699 * @tlsdev_ops: Transport Layer Security offload operations
1700 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1701 * of Layer 2 headers.
1702 *
1703 * @flags: Interface flags (a la BSD)
1704 * @priv_flags: Like 'flags' but invisible to userspace,
1705 * see if.h for the definitions
1706 * @gflags: Global flags ( kept as legacy )
1707 * @padded: How much padding added by alloc_netdev()
1708 * @operstate: RFC2863 operstate
1709 * @link_mode: Mapping policy to operstate
1710 * @if_port: Selectable AUI, TP, ...
1711 * @dma: DMA channel
1712 * @mtu: Interface MTU value
1713 * @min_mtu: Interface Minimum MTU value
1714 * @max_mtu: Interface Maximum MTU value
1715 * @type: Interface hardware type
1716 * @hard_header_len: Maximum hardware header length.
1717 * @min_header_len: Minimum hardware header length
1718 *
1719 * @needed_headroom: Extra headroom the hardware may need, but not in all
1720 * cases can this be guaranteed
1721 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1722 * cases can this be guaranteed. Some cases also use
1723 * LL_MAX_HEADER instead to allocate the skb
1724 *
1725 * interface address info:
1726 *
1727 * @perm_addr: Permanent hw address
1728 * @addr_assign_type: Hw address assignment type
1729 * @addr_len: Hardware address length
1730 * @upper_level: Maximum depth level of upper devices.
1731 * @lower_level: Maximum depth level of lower devices.
1732 * @neigh_priv_len: Used in neigh_alloc()
1733 * @dev_id: Used to differentiate devices that share
1734 * the same link layer address
1735 * @dev_port: Used to differentiate devices that share
1736 * the same function
1737 * @addr_list_lock: XXX: need comments on this one
1738 * @name_assign_type: network interface name assignment type
1739 * @uc_promisc: Counter that indicates promiscuous mode
1740 * has been enabled due to the need to listen to
1741 * additional unicast addresses in a device that
1742 * does not implement ndo_set_rx_mode()
1743 * @uc: unicast mac addresses
1744 * @mc: multicast mac addresses
1745 * @dev_addrs: list of device hw addresses
1746 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1747 * @promiscuity: Number of times the NIC is told to work in
1748 * promiscuous mode; if it becomes 0 the NIC will
1749 * exit promiscuous mode
1750 * @allmulti: Counter, enables or disables allmulticast mode
1751 *
1752 * @vlan_info: VLAN info
1753 * @dsa_ptr: dsa specific data
1754 * @tipc_ptr: TIPC specific data
1755 * @atalk_ptr: AppleTalk link
1756 * @ip_ptr: IPv4 specific data
1757 * @dn_ptr: DECnet specific data
1758 * @ip6_ptr: IPv6 specific data
1759 * @ax25_ptr: AX.25 specific data
1760 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1761 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1762 * device struct
1763 * @mpls_ptr: mpls_dev struct pointer
1764 *
1765 * @dev_addr: Hw address (before bcast,
1766 * because most packets are unicast)
1767 *
1768 * @_rx: Array of RX queues
1769 * @num_rx_queues: Number of RX queues
1770 * allocated at register_netdev() time
1771 * @real_num_rx_queues: Number of RX queues currently active in device
1772 * @xdp_prog: XDP sockets filter program pointer
1773 * @gro_flush_timeout: timeout for GRO layer in NAPI
1774 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1775 * allow to avoid NIC hard IRQ, on busy queues.
1776 *
1777 * @rx_handler: handler for received packets
1778 * @rx_handler_data: XXX: need comments on this one
1779 * @miniq_ingress: ingress/clsact qdisc specific data for
1780 * ingress processing
1781 * @ingress_queue: XXX: need comments on this one
1782 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1783 * @broadcast: hw bcast address
1784 *
1785 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1786 * indexed by RX queue number. Assigned by driver.
1787 * This must only be set if the ndo_rx_flow_steer
1788 * operation is defined
1789 * @index_hlist: Device index hash chain
1790 *
1791 * @_tx: Array of TX queues
1792 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1793 * @real_num_tx_queues: Number of TX queues currently active in device
1794 * @qdisc: Root qdisc from userspace point of view
1795 * @tx_queue_len: Max frames per queue allowed
1796 * @tx_global_lock: XXX: need comments on this one
1797 * @xdp_bulkq: XDP device bulk queue
1798 * @xps_cpus_map: all CPUs map for XPS device
1799 * @xps_rxqs_map: all RXQs map for XPS device
1800 *
1801 * @xps_maps: XXX: need comments on this one
1802 * @miniq_egress: clsact qdisc specific data for
1803 * egress processing
1804 * @qdisc_hash: qdisc hash table
1805 * @watchdog_timeo: Represents the timeout that is used by
1806 * the watchdog (see dev_watchdog())
1807 * @watchdog_timer: List of timers
1808 *
1809 * @proto_down_reason: reason a netdev interface is held down
1810 * @pcpu_refcnt: Number of references to this device
1811 * @todo_list: Delayed register/unregister
1812 * @link_watch_list: XXX: need comments on this one
1813 *
1814 * @reg_state: Register/unregister state machine
1815 * @dismantle: Device is going to be freed
1816 * @rtnl_link_state: This enum represents the phases of creating
1817 * a new link
1818 *
1819 * @needs_free_netdev: Should unregister perform free_netdev?
1820 * @priv_destructor: Called from unregister
1821 * @npinfo: XXX: need comments on this one
1822 * @nd_net: Network namespace this network device is inside
1823 *
1824 * @ml_priv: Mid-layer private
1825 * @ml_priv_type: Mid-layer private type
1826 * @lstats: Loopback statistics
1827 * @tstats: Tunnel statistics
1828 * @dstats: Dummy statistics
1829 * @vstats: Virtual ethernet statistics
1830 *
1831 * @garp_port: GARP
1832 * @mrp_port: MRP
1833 *
1834 * @dev: Class/net/name entry
1835 * @sysfs_groups: Space for optional device, statistics and wireless
1836 * sysfs groups
1837 *
1838 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1839 * @rtnl_link_ops: Rtnl_link_ops
1840 *
1841 * @gso_max_size: Maximum size of generic segmentation offload
1842 * @gso_max_segs: Maximum number of segments that can be passed to the
1843 * NIC for GSO
1844 *
1845 * @dcbnl_ops: Data Center Bridging netlink ops
1846 * @num_tc: Number of traffic classes in the net device
1847 * @tc_to_txq: XXX: need comments on this one
1848 * @prio_tc_map: XXX: need comments on this one
1849 *
1850 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1851 *
1852 * @priomap: XXX: need comments on this one
1853 * @phydev: Physical device may attach itself
1854 * for hardware timestamping
1855 * @sfp_bus: attached &struct sfp_bus structure.
1856 *
1857 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1858 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1859 *
1860 * @proto_down: protocol port state information can be sent to the
1861 * switch driver and used to set the phys state of the
1862 * switch port.
1863 *
1864 * @wol_enabled: Wake-on-LAN is enabled
1865 *
1866 * @net_notifier_list: List of per-net netdev notifier block
1867 * that follow this device when it is moved
1868 * to another network namespace.
1869 *
1870 * @macsec_ops: MACsec offloading ops
1871 *
1872 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1873 * offload capabilities of the device
1874 * @udp_tunnel_nic: UDP tunnel offload state
1875 * @xdp_state: stores info on attached XDP BPF programs
1876 *
1877 * @nested_level: Used as as a parameter of spin_lock_nested() of
1878 * dev->addr_list_lock.
1879 * @unlink_list: As netif_addr_lock() can be called recursively,
1880 * keep a list of interfaces to be deleted.
1881 *
1882 * FIXME: cleanup struct net_device such that network protocol info
1883 * moves out.
1884 */
1885
1886 struct net_device {
1887 char name[IFNAMSIZ];
1888 struct netdev_name_node *name_node;
1889 struct dev_ifalias __rcu *ifalias;
1890 /*
1891 * I/O specific fields
1892 * FIXME: Merge these and struct ifmap into one
1893 */
1894 unsigned long mem_end;
1895 unsigned long mem_start;
1896 unsigned long base_addr;
1897 int irq;
1898
1899 /*
1900 * Some hardware also needs these fields (state,dev_list,
1901 * napi_list,unreg_list,close_list) but they are not
1902 * part of the usual set specified in Space.c.
1903 */
1904
1905 unsigned long state;
1906
1907 struct list_head dev_list;
1908 struct list_head napi_list;
1909 struct list_head unreg_list;
1910 struct list_head close_list;
1911 struct list_head ptype_all;
1912 struct list_head ptype_specific;
1913
1914 struct {
1915 struct list_head upper;
1916 struct list_head lower;
1917 } adj_list;
1918
1919 netdev_features_t features;
1920 netdev_features_t hw_features;
1921 netdev_features_t wanted_features;
1922 netdev_features_t vlan_features;
1923 netdev_features_t hw_enc_features;
1924 netdev_features_t mpls_features;
1925 netdev_features_t gso_partial_features;
1926
1927 int ifindex;
1928 int group;
1929
1930 struct net_device_stats stats;
1931
1932 atomic_long_t rx_dropped;
1933 atomic_long_t tx_dropped;
1934 atomic_long_t rx_nohandler;
1935
1936 /* Stats to monitor link on/off, flapping */
1937 atomic_t carrier_up_count;
1938 atomic_t carrier_down_count;
1939
1940 #ifdef CONFIG_WIRELESS_EXT
1941 const struct iw_handler_def *wireless_handlers;
1942 struct iw_public_data *wireless_data;
1943 #endif
1944 const struct net_device_ops *netdev_ops;
1945 const struct ethtool_ops *ethtool_ops;
1946 #ifdef CONFIG_NET_L3_MASTER_DEV
1947 const struct l3mdev_ops *l3mdev_ops;
1948 #endif
1949 #if IS_ENABLED(CONFIG_IPV6)
1950 const struct ndisc_ops *ndisc_ops;
1951 #endif
1952
1953 #ifdef CONFIG_XFRM_OFFLOAD
1954 const struct xfrmdev_ops *xfrmdev_ops;
1955 #endif
1956
1957 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1958 const struct tlsdev_ops *tlsdev_ops;
1959 #endif
1960
1961 const struct header_ops *header_ops;
1962
1963 unsigned int flags;
1964 unsigned int priv_flags;
1965
1966 unsigned short gflags;
1967 unsigned short padded;
1968
1969 unsigned char operstate;
1970 unsigned char link_mode;
1971
1972 unsigned char if_port;
1973 unsigned char dma;
1974
1975 /* Note : dev->mtu is often read without holding a lock.
1976 * Writers usually hold RTNL.
1977 * It is recommended to use READ_ONCE() to annotate the reads,
1978 * and to use WRITE_ONCE() to annotate the writes.
1979 */
1980 unsigned int mtu;
1981 unsigned int min_mtu;
1982 unsigned int max_mtu;
1983 unsigned short type;
1984 unsigned short hard_header_len;
1985 unsigned char min_header_len;
1986 unsigned char name_assign_type;
1987
1988 unsigned short needed_headroom;
1989 unsigned short needed_tailroom;
1990
1991 /* Interface address info. */
1992 unsigned char perm_addr[MAX_ADDR_LEN];
1993 unsigned char addr_assign_type;
1994 unsigned char addr_len;
1995 unsigned char upper_level;
1996 unsigned char lower_level;
1997
1998 unsigned short neigh_priv_len;
1999 unsigned short dev_id;
2000 unsigned short dev_port;
2001 spinlock_t addr_list_lock;
2002
2003 struct netdev_hw_addr_list uc;
2004 struct netdev_hw_addr_list mc;
2005 struct netdev_hw_addr_list dev_addrs;
2006
2007 #ifdef CONFIG_SYSFS
2008 struct kset *queues_kset;
2009 #endif
2010 #ifdef CONFIG_LOCKDEP
2011 struct list_head unlink_list;
2012 #endif
2013 unsigned int promiscuity;
2014 unsigned int allmulti;
2015 bool uc_promisc;
2016 #ifdef CONFIG_LOCKDEP
2017 unsigned char nested_level;
2018 #endif
2019
2020
2021 /* Protocol-specific pointers */
2022
2023 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2024 struct vlan_info __rcu *vlan_info;
2025 #endif
2026 #if IS_ENABLED(CONFIG_NET_DSA)
2027 struct dsa_port *dsa_ptr;
2028 #endif
2029 #if IS_ENABLED(CONFIG_TIPC)
2030 struct tipc_bearer __rcu *tipc_ptr;
2031 #endif
2032 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2033 void *atalk_ptr;
2034 #endif
2035 struct in_device __rcu *ip_ptr;
2036 #if IS_ENABLED(CONFIG_DECNET)
2037 struct dn_dev __rcu *dn_ptr;
2038 #endif
2039 struct inet6_dev __rcu *ip6_ptr;
2040 #if IS_ENABLED(CONFIG_NEWIP)
2041 struct ninet_dev __rcu *nip_ptr; /* NIP */
2042 #endif
2043 #if IS_ENABLED(CONFIG_AX25)
2044 void *ax25_ptr;
2045 #endif
2046 struct wireless_dev *ieee80211_ptr;
2047 struct wpan_dev *ieee802154_ptr;
2048 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2049 struct mpls_dev __rcu *mpls_ptr;
2050 #endif
2051
2052 /*
2053 * Cache lines mostly used on receive path (including eth_type_trans())
2054 */
2055 /* Interface address info used in eth_type_trans() */
2056 unsigned char *dev_addr;
2057
2058 struct netdev_rx_queue *_rx;
2059 unsigned int num_rx_queues;
2060 unsigned int real_num_rx_queues;
2061
2062 struct bpf_prog __rcu *xdp_prog;
2063 unsigned long gro_flush_timeout;
2064 int napi_defer_hard_irqs;
2065 rx_handler_func_t __rcu *rx_handler;
2066 void __rcu *rx_handler_data;
2067
2068 #ifdef CONFIG_NET_CLS_ACT
2069 struct mini_Qdisc __rcu *miniq_ingress;
2070 #endif
2071 struct netdev_queue __rcu *ingress_queue;
2072 #ifdef CONFIG_NETFILTER_INGRESS
2073 struct nf_hook_entries __rcu *nf_hooks_ingress;
2074 #endif
2075
2076 unsigned char broadcast[MAX_ADDR_LEN];
2077 #ifdef CONFIG_RFS_ACCEL
2078 struct cpu_rmap *rx_cpu_rmap;
2079 #endif
2080 struct hlist_node index_hlist;
2081
2082 /*
2083 * Cache lines mostly used on transmit path
2084 */
2085 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2086 unsigned int num_tx_queues;
2087 unsigned int real_num_tx_queues;
2088 struct Qdisc __rcu *qdisc;
2089 unsigned int tx_queue_len;
2090 spinlock_t tx_global_lock;
2091
2092 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2093
2094 #ifdef CONFIG_XPS
2095 struct xps_dev_maps __rcu *xps_cpus_map;
2096 struct xps_dev_maps __rcu *xps_rxqs_map;
2097 #endif
2098 #ifdef CONFIG_NET_CLS_ACT
2099 struct mini_Qdisc __rcu *miniq_egress;
2100 #endif
2101
2102 #ifdef CONFIG_NET_SCHED
2103 DECLARE_HASHTABLE (qdisc_hash, 4);
2104 #endif
2105 /* These may be needed for future network-power-down code. */
2106 struct timer_list watchdog_timer;
2107 int watchdog_timeo;
2108
2109 u32 proto_down_reason;
2110
2111 struct list_head todo_list;
2112 int __percpu *pcpu_refcnt;
2113
2114 struct list_head link_watch_list;
2115
2116 enum { NETREG_UNINITIALIZED=0,
2117 NETREG_REGISTERED, /* completed register_netdevice */
2118 NETREG_UNREGISTERING, /* called unregister_netdevice */
2119 NETREG_UNREGISTERED, /* completed unregister todo */
2120 NETREG_RELEASED, /* called free_netdev */
2121 NETREG_DUMMY, /* dummy device for NAPI poll */
2122 } reg_state:8;
2123
2124 bool dismantle;
2125
2126 enum {
2127 RTNL_LINK_INITIALIZED,
2128 RTNL_LINK_INITIALIZING,
2129 } rtnl_link_state:16;
2130
2131 bool needs_free_netdev;
2132 void (*priv_destructor)(struct net_device *dev);
2133
2134 #ifdef CONFIG_NETPOLL
2135 struct netpoll_info __rcu *npinfo;
2136 #endif
2137
2138 possible_net_t nd_net;
2139
2140 /* mid-layer private */
2141 void *ml_priv;
2142 enum netdev_ml_priv_type ml_priv_type;
2143
2144 union {
2145 struct pcpu_lstats __percpu *lstats;
2146 struct pcpu_sw_netstats __percpu *tstats;
2147 struct pcpu_dstats __percpu *dstats;
2148 };
2149
2150 #if IS_ENABLED(CONFIG_GARP)
2151 struct garp_port __rcu *garp_port;
2152 #endif
2153 #if IS_ENABLED(CONFIG_MRP)
2154 struct mrp_port __rcu *mrp_port;
2155 #endif
2156
2157 struct device dev;
2158 const struct attribute_group *sysfs_groups[4];
2159 const struct attribute_group *sysfs_rx_queue_group;
2160
2161 const struct rtnl_link_ops *rtnl_link_ops;
2162
2163 /* for setting kernel sock attribute on TCP connection setup */
2164 #define GSO_MAX_SIZE 65536
2165 unsigned int gso_max_size;
2166 #define GSO_MAX_SEGS 65535
2167 u16 gso_max_segs;
2168
2169 #ifdef CONFIG_DCB
2170 const struct dcbnl_rtnl_ops *dcbnl_ops;
2171 #endif
2172 s16 num_tc;
2173 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2174 u8 prio_tc_map[TC_BITMASK + 1];
2175
2176 #if IS_ENABLED(CONFIG_FCOE)
2177 unsigned int fcoe_ddp_xid;
2178 #endif
2179 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2180 struct netprio_map __rcu *priomap;
2181 #endif
2182 struct phy_device *phydev;
2183 struct sfp_bus *sfp_bus;
2184 struct lock_class_key *qdisc_tx_busylock;
2185 struct lock_class_key *qdisc_running_key;
2186 bool proto_down;
2187 unsigned wol_enabled:1;
2188
2189 struct list_head net_notifier_list;
2190
2191 #if IS_ENABLED(CONFIG_MACSEC)
2192 /* MACsec management functions */
2193 const struct macsec_ops *macsec_ops;
2194 #endif
2195 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2196 struct udp_tunnel_nic *udp_tunnel_nic;
2197
2198 /* protected by rtnl_lock */
2199 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2200 };
2201 #define to_net_dev(d) container_of(d, struct net_device, dev)
2202
netif_elide_gro(const struct net_device * dev)2203 static inline bool netif_elide_gro(const struct net_device *dev)
2204 {
2205 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2206 return true;
2207 return false;
2208 }
2209
2210 #define NETDEV_ALIGN 32
2211
2212 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)2213 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2214 {
2215 return dev->prio_tc_map[prio & TC_BITMASK];
2216 }
2217
2218 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)2219 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2220 {
2221 if (tc >= dev->num_tc)
2222 return -EINVAL;
2223
2224 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2225 return 0;
2226 }
2227
2228 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2229 void netdev_reset_tc(struct net_device *dev);
2230 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2231 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2232
2233 static inline
netdev_get_num_tc(struct net_device * dev)2234 int netdev_get_num_tc(struct net_device *dev)
2235 {
2236 return dev->num_tc;
2237 }
2238
net_prefetch(void * p)2239 static inline void net_prefetch(void *p)
2240 {
2241 prefetch(p);
2242 #if L1_CACHE_BYTES < 128
2243 prefetch((u8 *)p + L1_CACHE_BYTES);
2244 #endif
2245 }
2246
net_prefetchw(void * p)2247 static inline void net_prefetchw(void *p)
2248 {
2249 prefetchw(p);
2250 #if L1_CACHE_BYTES < 128
2251 prefetchw((u8 *)p + L1_CACHE_BYTES);
2252 #endif
2253 }
2254
2255 void netdev_unbind_sb_channel(struct net_device *dev,
2256 struct net_device *sb_dev);
2257 int netdev_bind_sb_channel_queue(struct net_device *dev,
2258 struct net_device *sb_dev,
2259 u8 tc, u16 count, u16 offset);
2260 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
netdev_get_sb_channel(struct net_device * dev)2261 static inline int netdev_get_sb_channel(struct net_device *dev)
2262 {
2263 return max_t(int, -dev->num_tc, 0);
2264 }
2265
2266 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)2267 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2268 unsigned int index)
2269 {
2270 return &dev->_tx[index];
2271 }
2272
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)2273 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2274 const struct sk_buff *skb)
2275 {
2276 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2277 }
2278
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)2279 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2280 void (*f)(struct net_device *,
2281 struct netdev_queue *,
2282 void *),
2283 void *arg)
2284 {
2285 unsigned int i;
2286
2287 for (i = 0; i < dev->num_tx_queues; i++)
2288 f(dev, &dev->_tx[i], arg);
2289 }
2290
2291 #define netdev_lockdep_set_classes(dev) \
2292 { \
2293 static struct lock_class_key qdisc_tx_busylock_key; \
2294 static struct lock_class_key qdisc_running_key; \
2295 static struct lock_class_key qdisc_xmit_lock_key; \
2296 static struct lock_class_key dev_addr_list_lock_key; \
2297 unsigned int i; \
2298 \
2299 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2300 (dev)->qdisc_running_key = &qdisc_running_key; \
2301 lockdep_set_class(&(dev)->addr_list_lock, \
2302 &dev_addr_list_lock_key); \
2303 for (i = 0; i < (dev)->num_tx_queues; i++) \
2304 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2305 &qdisc_xmit_lock_key); \
2306 }
2307
2308 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2309 struct net_device *sb_dev);
2310 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2311 struct sk_buff *skb,
2312 struct net_device *sb_dev);
2313
2314 /* returns the headroom that the master device needs to take in account
2315 * when forwarding to this dev
2316 */
netdev_get_fwd_headroom(struct net_device * dev)2317 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2318 {
2319 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2320 }
2321
netdev_set_rx_headroom(struct net_device * dev,int new_hr)2322 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2323 {
2324 if (dev->netdev_ops->ndo_set_rx_headroom)
2325 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2326 }
2327
2328 /* set the device rx headroom to the dev's default */
netdev_reset_rx_headroom(struct net_device * dev)2329 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2330 {
2331 netdev_set_rx_headroom(dev, -1);
2332 }
2333
netdev_get_ml_priv(struct net_device * dev,enum netdev_ml_priv_type type)2334 static inline void *netdev_get_ml_priv(struct net_device *dev,
2335 enum netdev_ml_priv_type type)
2336 {
2337 if (dev->ml_priv_type != type)
2338 return NULL;
2339
2340 return dev->ml_priv;
2341 }
2342
netdev_set_ml_priv(struct net_device * dev,void * ml_priv,enum netdev_ml_priv_type type)2343 static inline void netdev_set_ml_priv(struct net_device *dev,
2344 void *ml_priv,
2345 enum netdev_ml_priv_type type)
2346 {
2347 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2348 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2349 dev->ml_priv_type, type);
2350 WARN(!dev->ml_priv_type && dev->ml_priv,
2351 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2352
2353 dev->ml_priv = ml_priv;
2354 dev->ml_priv_type = type;
2355 }
2356
2357 /*
2358 * Net namespace inlines
2359 */
2360 static inline
dev_net(const struct net_device * dev)2361 struct net *dev_net(const struct net_device *dev)
2362 {
2363 return read_pnet(&dev->nd_net);
2364 }
2365
2366 static inline
dev_net_set(struct net_device * dev,struct net * net)2367 void dev_net_set(struct net_device *dev, struct net *net)
2368 {
2369 write_pnet(&dev->nd_net, net);
2370 }
2371
2372 /**
2373 * netdev_priv - access network device private data
2374 * @dev: network device
2375 *
2376 * Get network device private data
2377 */
netdev_priv(const struct net_device * dev)2378 static inline void *netdev_priv(const struct net_device *dev)
2379 {
2380 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2381 }
2382
2383 /* Set the sysfs physical device reference for the network logical device
2384 * if set prior to registration will cause a symlink during initialization.
2385 */
2386 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2387
2388 /* Set the sysfs device type for the network logical device to allow
2389 * fine-grained identification of different network device types. For
2390 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2391 */
2392 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2393
2394 /* Default NAPI poll() weight
2395 * Device drivers are strongly advised to not use bigger value
2396 */
2397 #define NAPI_POLL_WEIGHT 64
2398
2399 /**
2400 * netif_napi_add - initialize a NAPI context
2401 * @dev: network device
2402 * @napi: NAPI context
2403 * @poll: polling function
2404 * @weight: default weight
2405 *
2406 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2407 * *any* of the other NAPI-related functions.
2408 */
2409 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2410 int (*poll)(struct napi_struct *, int), int weight);
2411
2412 /**
2413 * netif_tx_napi_add - initialize a NAPI context
2414 * @dev: network device
2415 * @napi: NAPI context
2416 * @poll: polling function
2417 * @weight: default weight
2418 *
2419 * This variant of netif_napi_add() should be used from drivers using NAPI
2420 * to exclusively poll a TX queue.
2421 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2422 */
netif_tx_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2423 static inline void netif_tx_napi_add(struct net_device *dev,
2424 struct napi_struct *napi,
2425 int (*poll)(struct napi_struct *, int),
2426 int weight)
2427 {
2428 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2429 netif_napi_add(dev, napi, poll, weight);
2430 }
2431
2432 /**
2433 * __netif_napi_del - remove a NAPI context
2434 * @napi: NAPI context
2435 *
2436 * Warning: caller must observe RCU grace period before freeing memory
2437 * containing @napi. Drivers might want to call this helper to combine
2438 * all the needed RCU grace periods into a single one.
2439 */
2440 void __netif_napi_del(struct napi_struct *napi);
2441
2442 /**
2443 * netif_napi_del - remove a NAPI context
2444 * @napi: NAPI context
2445 *
2446 * netif_napi_del() removes a NAPI context from the network device NAPI list
2447 */
netif_napi_del(struct napi_struct * napi)2448 static inline void netif_napi_del(struct napi_struct *napi)
2449 {
2450 __netif_napi_del(napi);
2451 synchronize_net();
2452 }
2453
2454 struct napi_gro_cb {
2455 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2456 void *frag0;
2457
2458 /* Length of frag0. */
2459 unsigned int frag0_len;
2460
2461 /* This indicates where we are processing relative to skb->data. */
2462 int data_offset;
2463
2464 /* This is non-zero if the packet cannot be merged with the new skb. */
2465 u16 flush;
2466
2467 /* Save the IP ID here and check when we get to the transport layer */
2468 u16 flush_id;
2469
2470 /* Number of segments aggregated. */
2471 u16 count;
2472
2473 /* Start offset for remote checksum offload */
2474 u16 gro_remcsum_start;
2475
2476 /* jiffies when first packet was created/queued */
2477 unsigned long age;
2478
2479 /* Used in ipv6_gro_receive() and foo-over-udp */
2480 u16 proto;
2481
2482 /* This is non-zero if the packet may be of the same flow. */
2483 u8 same_flow:1;
2484
2485 /* Used in tunnel GRO receive */
2486 u8 encap_mark:1;
2487
2488 /* GRO checksum is valid */
2489 u8 csum_valid:1;
2490
2491 /* Number of checksums via CHECKSUM_UNNECESSARY */
2492 u8 csum_cnt:3;
2493
2494 /* Free the skb? */
2495 u8 free:2;
2496 #define NAPI_GRO_FREE 1
2497 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2498
2499 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2500 u8 is_ipv6:1;
2501
2502 /* Used in GRE, set in fou/gue_gro_receive */
2503 u8 is_fou:1;
2504
2505 /* Used to determine if flush_id can be ignored */
2506 u8 is_atomic:1;
2507
2508 /* Number of gro_receive callbacks this packet already went through */
2509 u8 recursion_counter:4;
2510
2511 /* GRO is done by frag_list pointer chaining. */
2512 u8 is_flist:1;
2513
2514 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2515 __wsum csum;
2516
2517 /* used in skb_gro_receive() slow path */
2518 struct sk_buff *last;
2519 };
2520
2521 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2522
2523 #define GRO_RECURSION_LIMIT 15
gro_recursion_inc_test(struct sk_buff * skb)2524 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2525 {
2526 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2527 }
2528
2529 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
call_gro_receive(gro_receive_t cb,struct list_head * head,struct sk_buff * skb)2530 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2531 struct list_head *head,
2532 struct sk_buff *skb)
2533 {
2534 if (unlikely(gro_recursion_inc_test(skb))) {
2535 NAPI_GRO_CB(skb)->flush |= 1;
2536 return NULL;
2537 }
2538
2539 return cb(head, skb);
2540 }
2541
2542 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2543 struct sk_buff *);
call_gro_receive_sk(gro_receive_sk_t cb,struct sock * sk,struct list_head * head,struct sk_buff * skb)2544 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2545 struct sock *sk,
2546 struct list_head *head,
2547 struct sk_buff *skb)
2548 {
2549 if (unlikely(gro_recursion_inc_test(skb))) {
2550 NAPI_GRO_CB(skb)->flush |= 1;
2551 return NULL;
2552 }
2553
2554 return cb(sk, head, skb);
2555 }
2556
2557 struct packet_type {
2558 __be16 type; /* This is really htons(ether_type). */
2559 bool ignore_outgoing;
2560 struct net_device *dev; /* NULL is wildcarded here */
2561 int (*func) (struct sk_buff *,
2562 struct net_device *,
2563 struct packet_type *,
2564 struct net_device *);
2565 void (*list_func) (struct list_head *,
2566 struct packet_type *,
2567 struct net_device *);
2568 bool (*id_match)(struct packet_type *ptype,
2569 struct sock *sk);
2570 struct net *af_packet_net;
2571 void *af_packet_priv;
2572 struct list_head list;
2573 };
2574
2575 struct offload_callbacks {
2576 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2577 netdev_features_t features);
2578 struct sk_buff *(*gro_receive)(struct list_head *head,
2579 struct sk_buff *skb);
2580 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2581 };
2582
2583 struct packet_offload {
2584 __be16 type; /* This is really htons(ether_type). */
2585 u16 priority;
2586 struct offload_callbacks callbacks;
2587 struct list_head list;
2588 };
2589
2590 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2591 struct pcpu_sw_netstats {
2592 u64 rx_packets;
2593 u64 rx_bytes;
2594 u64 tx_packets;
2595 u64 tx_bytes;
2596 struct u64_stats_sync syncp;
2597 } __aligned(4 * sizeof(u64));
2598
2599 struct pcpu_lstats {
2600 u64_stats_t packets;
2601 u64_stats_t bytes;
2602 struct u64_stats_sync syncp;
2603 } __aligned(2 * sizeof(u64));
2604
2605 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2606
dev_sw_netstats_rx_add(struct net_device * dev,unsigned int len)2607 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2608 {
2609 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2610
2611 u64_stats_update_begin(&tstats->syncp);
2612 tstats->rx_bytes += len;
2613 tstats->rx_packets++;
2614 u64_stats_update_end(&tstats->syncp);
2615 }
2616
dev_lstats_add(struct net_device * dev,unsigned int len)2617 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2618 {
2619 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2620
2621 u64_stats_update_begin(&lstats->syncp);
2622 u64_stats_add(&lstats->bytes, len);
2623 u64_stats_inc(&lstats->packets);
2624 u64_stats_update_end(&lstats->syncp);
2625 }
2626
2627 #define __netdev_alloc_pcpu_stats(type, gfp) \
2628 ({ \
2629 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2630 if (pcpu_stats) { \
2631 int __cpu; \
2632 for_each_possible_cpu(__cpu) { \
2633 typeof(type) *stat; \
2634 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2635 u64_stats_init(&stat->syncp); \
2636 } \
2637 } \
2638 pcpu_stats; \
2639 })
2640
2641 #define netdev_alloc_pcpu_stats(type) \
2642 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2643
2644 enum netdev_lag_tx_type {
2645 NETDEV_LAG_TX_TYPE_UNKNOWN,
2646 NETDEV_LAG_TX_TYPE_RANDOM,
2647 NETDEV_LAG_TX_TYPE_BROADCAST,
2648 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2649 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2650 NETDEV_LAG_TX_TYPE_HASH,
2651 };
2652
2653 enum netdev_lag_hash {
2654 NETDEV_LAG_HASH_NONE,
2655 NETDEV_LAG_HASH_L2,
2656 NETDEV_LAG_HASH_L34,
2657 NETDEV_LAG_HASH_L23,
2658 NETDEV_LAG_HASH_E23,
2659 NETDEV_LAG_HASH_E34,
2660 NETDEV_LAG_HASH_UNKNOWN,
2661 };
2662
2663 struct netdev_lag_upper_info {
2664 enum netdev_lag_tx_type tx_type;
2665 enum netdev_lag_hash hash_type;
2666 };
2667
2668 struct netdev_lag_lower_state_info {
2669 u8 link_up : 1,
2670 tx_enabled : 1;
2671 };
2672
2673 #include <linux/notifier.h>
2674
2675 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2676 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2677 * adding new types.
2678 */
2679 enum netdev_cmd {
2680 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2681 NETDEV_DOWN,
2682 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2683 detected a hardware crash and restarted
2684 - we can use this eg to kick tcp sessions
2685 once done */
2686 NETDEV_CHANGE, /* Notify device state change */
2687 NETDEV_REGISTER,
2688 NETDEV_UNREGISTER,
2689 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2690 NETDEV_CHANGEADDR, /* notify after the address change */
2691 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2692 NETDEV_GOING_DOWN,
2693 NETDEV_CHANGENAME,
2694 NETDEV_FEAT_CHANGE,
2695 NETDEV_BONDING_FAILOVER,
2696 NETDEV_PRE_UP,
2697 NETDEV_PRE_TYPE_CHANGE,
2698 NETDEV_POST_TYPE_CHANGE,
2699 NETDEV_POST_INIT,
2700 NETDEV_RELEASE,
2701 NETDEV_NOTIFY_PEERS,
2702 NETDEV_JOIN,
2703 NETDEV_CHANGEUPPER,
2704 NETDEV_RESEND_IGMP,
2705 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2706 NETDEV_CHANGEINFODATA,
2707 NETDEV_BONDING_INFO,
2708 NETDEV_PRECHANGEUPPER,
2709 NETDEV_CHANGELOWERSTATE,
2710 NETDEV_UDP_TUNNEL_PUSH_INFO,
2711 NETDEV_UDP_TUNNEL_DROP_INFO,
2712 NETDEV_CHANGE_TX_QUEUE_LEN,
2713 NETDEV_CVLAN_FILTER_PUSH_INFO,
2714 NETDEV_CVLAN_FILTER_DROP_INFO,
2715 NETDEV_SVLAN_FILTER_PUSH_INFO,
2716 NETDEV_SVLAN_FILTER_DROP_INFO,
2717 };
2718 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2719
2720 int register_netdevice_notifier(struct notifier_block *nb);
2721 int unregister_netdevice_notifier(struct notifier_block *nb);
2722 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2723 int unregister_netdevice_notifier_net(struct net *net,
2724 struct notifier_block *nb);
2725 int register_netdevice_notifier_dev_net(struct net_device *dev,
2726 struct notifier_block *nb,
2727 struct netdev_net_notifier *nn);
2728 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2729 struct notifier_block *nb,
2730 struct netdev_net_notifier *nn);
2731
2732 struct netdev_notifier_info {
2733 struct net_device *dev;
2734 struct netlink_ext_ack *extack;
2735 };
2736
2737 struct netdev_notifier_info_ext {
2738 struct netdev_notifier_info info; /* must be first */
2739 union {
2740 u32 mtu;
2741 } ext;
2742 };
2743
2744 struct netdev_notifier_change_info {
2745 struct netdev_notifier_info info; /* must be first */
2746 unsigned int flags_changed;
2747 };
2748
2749 struct netdev_notifier_changeupper_info {
2750 struct netdev_notifier_info info; /* must be first */
2751 struct net_device *upper_dev; /* new upper dev */
2752 bool master; /* is upper dev master */
2753 bool linking; /* is the notification for link or unlink */
2754 void *upper_info; /* upper dev info */
2755 };
2756
2757 struct netdev_notifier_changelowerstate_info {
2758 struct netdev_notifier_info info; /* must be first */
2759 void *lower_state_info; /* is lower dev state */
2760 };
2761
2762 struct netdev_notifier_pre_changeaddr_info {
2763 struct netdev_notifier_info info; /* must be first */
2764 const unsigned char *dev_addr;
2765 };
2766
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)2767 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2768 struct net_device *dev)
2769 {
2770 info->dev = dev;
2771 info->extack = NULL;
2772 }
2773
2774 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)2775 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2776 {
2777 return info->dev;
2778 }
2779
2780 static inline struct netlink_ext_ack *
netdev_notifier_info_to_extack(const struct netdev_notifier_info * info)2781 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2782 {
2783 return info->extack;
2784 }
2785
2786 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2787
2788
2789 extern rwlock_t dev_base_lock; /* Device list lock */
2790
2791 #define for_each_netdev(net, d) \
2792 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2793 #define for_each_netdev_reverse(net, d) \
2794 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2795 #define for_each_netdev_rcu(net, d) \
2796 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2797 #define for_each_netdev_safe(net, d, n) \
2798 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2799 #define for_each_netdev_continue(net, d) \
2800 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2801 #define for_each_netdev_continue_reverse(net, d) \
2802 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2803 dev_list)
2804 #define for_each_netdev_continue_rcu(net, d) \
2805 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2806 #define for_each_netdev_in_bond_rcu(bond, slave) \
2807 for_each_netdev_rcu(&init_net, slave) \
2808 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2809 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2810
next_net_device(struct net_device * dev)2811 static inline struct net_device *next_net_device(struct net_device *dev)
2812 {
2813 struct list_head *lh;
2814 struct net *net;
2815
2816 net = dev_net(dev);
2817 lh = dev->dev_list.next;
2818 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2819 }
2820
next_net_device_rcu(struct net_device * dev)2821 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2822 {
2823 struct list_head *lh;
2824 struct net *net;
2825
2826 net = dev_net(dev);
2827 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2828 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2829 }
2830
first_net_device(struct net * net)2831 static inline struct net_device *first_net_device(struct net *net)
2832 {
2833 return list_empty(&net->dev_base_head) ? NULL :
2834 net_device_entry(net->dev_base_head.next);
2835 }
2836
first_net_device_rcu(struct net * net)2837 static inline struct net_device *first_net_device_rcu(struct net *net)
2838 {
2839 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2840
2841 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2842 }
2843
2844 int netdev_boot_setup_check(struct net_device *dev);
2845 unsigned long netdev_boot_base(const char *prefix, int unit);
2846 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2847 const char *hwaddr);
2848 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2849 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2850 void dev_add_pack(struct packet_type *pt);
2851 void dev_remove_pack(struct packet_type *pt);
2852 void __dev_remove_pack(struct packet_type *pt);
2853 void dev_add_offload(struct packet_offload *po);
2854 void dev_remove_offload(struct packet_offload *po);
2855
2856 int dev_get_iflink(const struct net_device *dev);
2857 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2858 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2859 unsigned short mask);
2860 struct net_device *dev_get_by_name(struct net *net, const char *name);
2861 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2862 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2863 int dev_alloc_name(struct net_device *dev, const char *name);
2864 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2865 void dev_close(struct net_device *dev);
2866 void dev_close_many(struct list_head *head, bool unlink);
2867 void dev_disable_lro(struct net_device *dev);
2868 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2869 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2870 struct net_device *sb_dev);
2871 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2872 struct net_device *sb_dev);
2873
2874 int dev_queue_xmit(struct sk_buff *skb);
2875 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2876 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2877
dev_direct_xmit(struct sk_buff * skb,u16 queue_id)2878 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2879 {
2880 int ret;
2881
2882 ret = __dev_direct_xmit(skb, queue_id);
2883 if (!dev_xmit_complete(ret))
2884 kfree_skb(skb);
2885 return ret;
2886 }
2887
2888 int register_netdevice(struct net_device *dev);
2889 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2890 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)2891 static inline void unregister_netdevice(struct net_device *dev)
2892 {
2893 unregister_netdevice_queue(dev, NULL);
2894 }
2895
2896 int netdev_refcnt_read(const struct net_device *dev);
2897 void free_netdev(struct net_device *dev);
2898 void netdev_freemem(struct net_device *dev);
2899 int init_dummy_netdev(struct net_device *dev);
2900
2901 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2902 struct sk_buff *skb,
2903 bool all_slaves);
2904 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2905 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2906 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2907 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2908 int netdev_get_name(struct net *net, char *name, int ifindex);
2909 int dev_restart(struct net_device *dev);
2910 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2911 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2912
skb_gro_offset(const struct sk_buff * skb)2913 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2914 {
2915 return NAPI_GRO_CB(skb)->data_offset;
2916 }
2917
skb_gro_len(const struct sk_buff * skb)2918 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2919 {
2920 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2921 }
2922
skb_gro_pull(struct sk_buff * skb,unsigned int len)2923 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2924 {
2925 NAPI_GRO_CB(skb)->data_offset += len;
2926 }
2927
skb_gro_header_fast(struct sk_buff * skb,unsigned int offset)2928 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2929 unsigned int offset)
2930 {
2931 return NAPI_GRO_CB(skb)->frag0 + offset;
2932 }
2933
skb_gro_header_hard(struct sk_buff * skb,unsigned int hlen)2934 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2935 {
2936 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2937 }
2938
skb_gro_frag0_invalidate(struct sk_buff * skb)2939 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2940 {
2941 NAPI_GRO_CB(skb)->frag0 = NULL;
2942 NAPI_GRO_CB(skb)->frag0_len = 0;
2943 }
2944
skb_gro_header_slow(struct sk_buff * skb,unsigned int hlen,unsigned int offset)2945 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2946 unsigned int offset)
2947 {
2948 if (!pskb_may_pull(skb, hlen))
2949 return NULL;
2950
2951 skb_gro_frag0_invalidate(skb);
2952 return skb->data + offset;
2953 }
2954
skb_gro_network_header(struct sk_buff * skb)2955 static inline void *skb_gro_network_header(struct sk_buff *skb)
2956 {
2957 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2958 skb_network_offset(skb);
2959 }
2960
skb_gro_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)2961 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2962 const void *start, unsigned int len)
2963 {
2964 if (NAPI_GRO_CB(skb)->csum_valid)
2965 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2966 csum_partial(start, len, 0));
2967 }
2968
2969 /* GRO checksum functions. These are logical equivalents of the normal
2970 * checksum functions (in skbuff.h) except that they operate on the GRO
2971 * offsets and fields in sk_buff.
2972 */
2973
2974 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2975
skb_at_gro_remcsum_start(struct sk_buff * skb)2976 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2977 {
2978 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2979 }
2980
__skb_gro_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)2981 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2982 bool zero_okay,
2983 __sum16 check)
2984 {
2985 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2986 skb_checksum_start_offset(skb) <
2987 skb_gro_offset(skb)) &&
2988 !skb_at_gro_remcsum_start(skb) &&
2989 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2990 (!zero_okay || check));
2991 }
2992
__skb_gro_checksum_validate_complete(struct sk_buff * skb,__wsum psum)2993 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2994 __wsum psum)
2995 {
2996 if (NAPI_GRO_CB(skb)->csum_valid &&
2997 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2998 return 0;
2999
3000 NAPI_GRO_CB(skb)->csum = psum;
3001
3002 return __skb_gro_checksum_complete(skb);
3003 }
3004
skb_gro_incr_csum_unnecessary(struct sk_buff * skb)3005 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3006 {
3007 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3008 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3009 NAPI_GRO_CB(skb)->csum_cnt--;
3010 } else {
3011 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3012 * verified a new top level checksum or an encapsulated one
3013 * during GRO. This saves work if we fallback to normal path.
3014 */
3015 __skb_incr_checksum_unnecessary(skb);
3016 }
3017 }
3018
3019 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3020 compute_pseudo) \
3021 ({ \
3022 __sum16 __ret = 0; \
3023 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3024 __ret = __skb_gro_checksum_validate_complete(skb, \
3025 compute_pseudo(skb, proto)); \
3026 if (!__ret) \
3027 skb_gro_incr_csum_unnecessary(skb); \
3028 __ret; \
3029 })
3030
3031 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3032 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3033
3034 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3035 compute_pseudo) \
3036 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3037
3038 #define skb_gro_checksum_simple_validate(skb) \
3039 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3040
__skb_gro_checksum_convert_check(struct sk_buff * skb)3041 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3042 {
3043 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3044 !NAPI_GRO_CB(skb)->csum_valid);
3045 }
3046
__skb_gro_checksum_convert(struct sk_buff * skb,__wsum pseudo)3047 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3048 __wsum pseudo)
3049 {
3050 NAPI_GRO_CB(skb)->csum = ~pseudo;
3051 NAPI_GRO_CB(skb)->csum_valid = 1;
3052 }
3053
3054 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3055 do { \
3056 if (__skb_gro_checksum_convert_check(skb)) \
3057 __skb_gro_checksum_convert(skb, \
3058 compute_pseudo(skb, proto)); \
3059 } while (0)
3060
3061 struct gro_remcsum {
3062 int offset;
3063 __wsum delta;
3064 };
3065
skb_gro_remcsum_init(struct gro_remcsum * grc)3066 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3067 {
3068 grc->offset = 0;
3069 grc->delta = 0;
3070 }
3071
skb_gro_remcsum_process(struct sk_buff * skb,void * ptr,unsigned int off,size_t hdrlen,int start,int offset,struct gro_remcsum * grc,bool nopartial)3072 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3073 unsigned int off, size_t hdrlen,
3074 int start, int offset,
3075 struct gro_remcsum *grc,
3076 bool nopartial)
3077 {
3078 __wsum delta;
3079 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3080
3081 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3082
3083 if (!nopartial) {
3084 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3085 return ptr;
3086 }
3087
3088 ptr = skb_gro_header_fast(skb, off);
3089 if (skb_gro_header_hard(skb, off + plen)) {
3090 ptr = skb_gro_header_slow(skb, off + plen, off);
3091 if (!ptr)
3092 return NULL;
3093 }
3094
3095 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3096 start, offset);
3097
3098 /* Adjust skb->csum since we changed the packet */
3099 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3100
3101 grc->offset = off + hdrlen + offset;
3102 grc->delta = delta;
3103
3104 return ptr;
3105 }
3106
skb_gro_remcsum_cleanup(struct sk_buff * skb,struct gro_remcsum * grc)3107 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3108 struct gro_remcsum *grc)
3109 {
3110 void *ptr;
3111 size_t plen = grc->offset + sizeof(u16);
3112
3113 if (!grc->delta)
3114 return;
3115
3116 ptr = skb_gro_header_fast(skb, grc->offset);
3117 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3118 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3119 if (!ptr)
3120 return;
3121 }
3122
3123 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3124 }
3125
3126 #ifdef CONFIG_XFRM_OFFLOAD
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)3127 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3128 {
3129 if (PTR_ERR(pp) != -EINPROGRESS)
3130 NAPI_GRO_CB(skb)->flush |= flush;
3131 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)3132 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3133 struct sk_buff *pp,
3134 int flush,
3135 struct gro_remcsum *grc)
3136 {
3137 if (PTR_ERR(pp) != -EINPROGRESS) {
3138 NAPI_GRO_CB(skb)->flush |= flush;
3139 skb_gro_remcsum_cleanup(skb, grc);
3140 skb->remcsum_offload = 0;
3141 }
3142 }
3143 #else
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)3144 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3145 {
3146 NAPI_GRO_CB(skb)->flush |= flush;
3147 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)3148 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3149 struct sk_buff *pp,
3150 int flush,
3151 struct gro_remcsum *grc)
3152 {
3153 NAPI_GRO_CB(skb)->flush |= flush;
3154 skb_gro_remcsum_cleanup(skb, grc);
3155 skb->remcsum_offload = 0;
3156 }
3157 #endif
3158
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)3159 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3160 unsigned short type,
3161 const void *daddr, const void *saddr,
3162 unsigned int len)
3163 {
3164 if (!dev->header_ops || !dev->header_ops->create)
3165 return 0;
3166
3167 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3168 }
3169
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)3170 static inline int dev_parse_header(const struct sk_buff *skb,
3171 unsigned char *haddr)
3172 {
3173 const struct net_device *dev = skb->dev;
3174
3175 if (!dev->header_ops || !dev->header_ops->parse)
3176 return 0;
3177 return dev->header_ops->parse(skb, haddr);
3178 }
3179
dev_parse_header_protocol(const struct sk_buff * skb)3180 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3181 {
3182 const struct net_device *dev = skb->dev;
3183
3184 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3185 return 0;
3186 return dev->header_ops->parse_protocol(skb);
3187 }
3188
3189 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)3190 static inline bool dev_validate_header(const struct net_device *dev,
3191 char *ll_header, int len)
3192 {
3193 if (likely(len >= dev->hard_header_len))
3194 return true;
3195 if (len < dev->min_header_len)
3196 return false;
3197
3198 if (capable(CAP_SYS_RAWIO)) {
3199 memset(ll_header + len, 0, dev->hard_header_len - len);
3200 return true;
3201 }
3202
3203 if (dev->header_ops && dev->header_ops->validate)
3204 return dev->header_ops->validate(ll_header, len);
3205
3206 return false;
3207 }
3208
dev_has_header(const struct net_device * dev)3209 static inline bool dev_has_header(const struct net_device *dev)
3210 {
3211 return dev->header_ops && dev->header_ops->create;
3212 }
3213
3214 #ifdef CONFIG_NET_FLOW_LIMIT
3215 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3216 struct sd_flow_limit {
3217 u64 count;
3218 unsigned int num_buckets;
3219 unsigned int history_head;
3220 u16 history[FLOW_LIMIT_HISTORY];
3221 u8 buckets[];
3222 };
3223
3224 extern int netdev_flow_limit_table_len;
3225 #endif /* CONFIG_NET_FLOW_LIMIT */
3226
3227 /*
3228 * Incoming packets are placed on per-CPU queues
3229 */
3230 struct softnet_data {
3231 struct list_head poll_list;
3232 struct sk_buff_head process_queue;
3233
3234 /* stats */
3235 unsigned int processed;
3236 unsigned int time_squeeze;
3237 unsigned int received_rps;
3238 #ifdef CONFIG_RPS
3239 struct softnet_data *rps_ipi_list;
3240 #endif
3241 #ifdef CONFIG_NET_FLOW_LIMIT
3242 struct sd_flow_limit __rcu *flow_limit;
3243 #endif
3244 struct Qdisc *output_queue;
3245 struct Qdisc **output_queue_tailp;
3246 struct sk_buff *completion_queue;
3247 #ifdef CONFIG_XFRM_OFFLOAD
3248 struct sk_buff_head xfrm_backlog;
3249 #endif
3250 /* written and read only by owning cpu: */
3251 struct {
3252 u16 recursion;
3253 u8 more;
3254 } xmit;
3255 #ifdef CONFIG_RPS
3256 /* input_queue_head should be written by cpu owning this struct,
3257 * and only read by other cpus. Worth using a cache line.
3258 */
3259 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3260
3261 /* Elements below can be accessed between CPUs for RPS/RFS */
3262 call_single_data_t csd ____cacheline_aligned_in_smp;
3263 struct softnet_data *rps_ipi_next;
3264 unsigned int cpu;
3265 unsigned int input_queue_tail;
3266 #endif
3267 unsigned int dropped;
3268 struct sk_buff_head input_pkt_queue;
3269 struct napi_struct backlog;
3270
3271 };
3272
input_queue_head_incr(struct softnet_data * sd)3273 static inline void input_queue_head_incr(struct softnet_data *sd)
3274 {
3275 #ifdef CONFIG_RPS
3276 sd->input_queue_head++;
3277 #endif
3278 }
3279
input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)3280 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3281 unsigned int *qtail)
3282 {
3283 #ifdef CONFIG_RPS
3284 *qtail = ++sd->input_queue_tail;
3285 #endif
3286 }
3287
3288 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3289
dev_recursion_level(void)3290 static inline int dev_recursion_level(void)
3291 {
3292 return this_cpu_read(softnet_data.xmit.recursion);
3293 }
3294
3295 #define XMIT_RECURSION_LIMIT 8
dev_xmit_recursion(void)3296 static inline bool dev_xmit_recursion(void)
3297 {
3298 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3299 XMIT_RECURSION_LIMIT);
3300 }
3301
dev_xmit_recursion_inc(void)3302 static inline void dev_xmit_recursion_inc(void)
3303 {
3304 __this_cpu_inc(softnet_data.xmit.recursion);
3305 }
3306
dev_xmit_recursion_dec(void)3307 static inline void dev_xmit_recursion_dec(void)
3308 {
3309 __this_cpu_dec(softnet_data.xmit.recursion);
3310 }
3311
3312 void __netif_schedule(struct Qdisc *q);
3313 void netif_schedule_queue(struct netdev_queue *txq);
3314
netif_tx_schedule_all(struct net_device * dev)3315 static inline void netif_tx_schedule_all(struct net_device *dev)
3316 {
3317 unsigned int i;
3318
3319 for (i = 0; i < dev->num_tx_queues; i++)
3320 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3321 }
3322
netif_tx_start_queue(struct netdev_queue * dev_queue)3323 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3324 {
3325 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3326 }
3327
3328 /**
3329 * netif_start_queue - allow transmit
3330 * @dev: network device
3331 *
3332 * Allow upper layers to call the device hard_start_xmit routine.
3333 */
netif_start_queue(struct net_device * dev)3334 static inline void netif_start_queue(struct net_device *dev)
3335 {
3336 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3337 }
3338
netif_tx_start_all_queues(struct net_device * dev)3339 static inline void netif_tx_start_all_queues(struct net_device *dev)
3340 {
3341 unsigned int i;
3342
3343 for (i = 0; i < dev->num_tx_queues; i++) {
3344 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3345 netif_tx_start_queue(txq);
3346 }
3347 }
3348
3349 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3350
3351 /**
3352 * netif_wake_queue - restart transmit
3353 * @dev: network device
3354 *
3355 * Allow upper layers to call the device hard_start_xmit routine.
3356 * Used for flow control when transmit resources are available.
3357 */
netif_wake_queue(struct net_device * dev)3358 static inline void netif_wake_queue(struct net_device *dev)
3359 {
3360 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3361 }
3362
netif_tx_wake_all_queues(struct net_device * dev)3363 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3364 {
3365 unsigned int i;
3366
3367 for (i = 0; i < dev->num_tx_queues; i++) {
3368 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3369 netif_tx_wake_queue(txq);
3370 }
3371 }
3372
netif_tx_stop_queue(struct netdev_queue * dev_queue)3373 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3374 {
3375 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3376 }
3377
3378 /**
3379 * netif_stop_queue - stop transmitted packets
3380 * @dev: network device
3381 *
3382 * Stop upper layers calling the device hard_start_xmit routine.
3383 * Used for flow control when transmit resources are unavailable.
3384 */
netif_stop_queue(struct net_device * dev)3385 static inline void netif_stop_queue(struct net_device *dev)
3386 {
3387 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3388 }
3389
3390 void netif_tx_stop_all_queues(struct net_device *dev);
3391
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)3392 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3393 {
3394 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3395 }
3396
3397 /**
3398 * netif_queue_stopped - test if transmit queue is flowblocked
3399 * @dev: network device
3400 *
3401 * Test if transmit queue on device is currently unable to send.
3402 */
netif_queue_stopped(const struct net_device * dev)3403 static inline bool netif_queue_stopped(const struct net_device *dev)
3404 {
3405 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3406 }
3407
netif_xmit_stopped(const struct netdev_queue * dev_queue)3408 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3409 {
3410 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3411 }
3412
3413 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)3414 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3415 {
3416 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3417 }
3418
3419 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)3420 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3421 {
3422 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3423 }
3424
3425 /**
3426 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3427 * @dev_queue: pointer to transmit queue
3428 *
3429 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3430 * to give appropriate hint to the CPU.
3431 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)3432 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3433 {
3434 #ifdef CONFIG_BQL
3435 prefetchw(&dev_queue->dql.num_queued);
3436 #endif
3437 }
3438
3439 /**
3440 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3441 * @dev_queue: pointer to transmit queue
3442 *
3443 * BQL enabled drivers might use this helper in their TX completion path,
3444 * to give appropriate hint to the CPU.
3445 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3446 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3447 {
3448 #ifdef CONFIG_BQL
3449 prefetchw(&dev_queue->dql.limit);
3450 #endif
3451 }
3452
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3453 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3454 unsigned int bytes)
3455 {
3456 #ifdef CONFIG_BQL
3457 dql_queued(&dev_queue->dql, bytes);
3458
3459 if (likely(dql_avail(&dev_queue->dql) >= 0))
3460 return;
3461
3462 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3463
3464 /*
3465 * The XOFF flag must be set before checking the dql_avail below,
3466 * because in netdev_tx_completed_queue we update the dql_completed
3467 * before checking the XOFF flag.
3468 */
3469 smp_mb();
3470
3471 /* check again in case another CPU has just made room avail */
3472 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3473 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3474 #endif
3475 }
3476
3477 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3478 * that they should not test BQL status themselves.
3479 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3480 * skb of a batch.
3481 * Returns true if the doorbell must be used to kick the NIC.
3482 */
__netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes,bool xmit_more)3483 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3484 unsigned int bytes,
3485 bool xmit_more)
3486 {
3487 if (xmit_more) {
3488 #ifdef CONFIG_BQL
3489 dql_queued(&dev_queue->dql, bytes);
3490 #endif
3491 return netif_tx_queue_stopped(dev_queue);
3492 }
3493 netdev_tx_sent_queue(dev_queue, bytes);
3494 return true;
3495 }
3496
3497 /**
3498 * netdev_sent_queue - report the number of bytes queued to hardware
3499 * @dev: network device
3500 * @bytes: number of bytes queued to the hardware device queue
3501 *
3502 * Report the number of bytes queued for sending/completion to the network
3503 * device hardware queue. @bytes should be a good approximation and should
3504 * exactly match netdev_completed_queue() @bytes
3505 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)3506 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3507 {
3508 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3509 }
3510
__netdev_sent_queue(struct net_device * dev,unsigned int bytes,bool xmit_more)3511 static inline bool __netdev_sent_queue(struct net_device *dev,
3512 unsigned int bytes,
3513 bool xmit_more)
3514 {
3515 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3516 xmit_more);
3517 }
3518
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3519 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3520 unsigned int pkts, unsigned int bytes)
3521 {
3522 #ifdef CONFIG_BQL
3523 if (unlikely(!bytes))
3524 return;
3525
3526 dql_completed(&dev_queue->dql, bytes);
3527
3528 /*
3529 * Without the memory barrier there is a small possiblity that
3530 * netdev_tx_sent_queue will miss the update and cause the queue to
3531 * be stopped forever
3532 */
3533 smp_mb();
3534
3535 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3536 return;
3537
3538 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3539 netif_schedule_queue(dev_queue);
3540 #endif
3541 }
3542
3543 /**
3544 * netdev_completed_queue - report bytes and packets completed by device
3545 * @dev: network device
3546 * @pkts: actual number of packets sent over the medium
3547 * @bytes: actual number of bytes sent over the medium
3548 *
3549 * Report the number of bytes and packets transmitted by the network device
3550 * hardware queue over the physical medium, @bytes must exactly match the
3551 * @bytes amount passed to netdev_sent_queue()
3552 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3553 static inline void netdev_completed_queue(struct net_device *dev,
3554 unsigned int pkts, unsigned int bytes)
3555 {
3556 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3557 }
3558
netdev_tx_reset_queue(struct netdev_queue * q)3559 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3560 {
3561 #ifdef CONFIG_BQL
3562 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3563 dql_reset(&q->dql);
3564 #endif
3565 }
3566
3567 /**
3568 * netdev_reset_queue - reset the packets and bytes count of a network device
3569 * @dev_queue: network device
3570 *
3571 * Reset the bytes and packet count of a network device and clear the
3572 * software flow control OFF bit for this network device
3573 */
netdev_reset_queue(struct net_device * dev_queue)3574 static inline void netdev_reset_queue(struct net_device *dev_queue)
3575 {
3576 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3577 }
3578
3579 /**
3580 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3581 * @dev: network device
3582 * @queue_index: given tx queue index
3583 *
3584 * Returns 0 if given tx queue index >= number of device tx queues,
3585 * otherwise returns the originally passed tx queue index.
3586 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3587 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3588 {
3589 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3590 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3591 dev->name, queue_index,
3592 dev->real_num_tx_queues);
3593 return 0;
3594 }
3595
3596 return queue_index;
3597 }
3598
3599 /**
3600 * netif_running - test if up
3601 * @dev: network device
3602 *
3603 * Test if the device has been brought up.
3604 */
netif_running(const struct net_device * dev)3605 static inline bool netif_running(const struct net_device *dev)
3606 {
3607 return test_bit(__LINK_STATE_START, &dev->state);
3608 }
3609
3610 /*
3611 * Routines to manage the subqueues on a device. We only need start,
3612 * stop, and a check if it's stopped. All other device management is
3613 * done at the overall netdevice level.
3614 * Also test the device if we're multiqueue.
3615 */
3616
3617 /**
3618 * netif_start_subqueue - allow sending packets on subqueue
3619 * @dev: network device
3620 * @queue_index: sub queue index
3621 *
3622 * Start individual transmit queue of a device with multiple transmit queues.
3623 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)3624 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3625 {
3626 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3627
3628 netif_tx_start_queue(txq);
3629 }
3630
3631 /**
3632 * netif_stop_subqueue - stop sending packets on subqueue
3633 * @dev: network device
3634 * @queue_index: sub queue index
3635 *
3636 * Stop individual transmit queue of a device with multiple transmit queues.
3637 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)3638 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3639 {
3640 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3641 netif_tx_stop_queue(txq);
3642 }
3643
3644 /**
3645 * netif_subqueue_stopped - test status of subqueue
3646 * @dev: network device
3647 * @queue_index: sub queue index
3648 *
3649 * Check individual transmit queue of a device with multiple transmit queues.
3650 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3651 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3652 u16 queue_index)
3653 {
3654 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3655
3656 return netif_tx_queue_stopped(txq);
3657 }
3658
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3659 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3660 struct sk_buff *skb)
3661 {
3662 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3663 }
3664
3665 /**
3666 * netif_wake_subqueue - allow sending packets on subqueue
3667 * @dev: network device
3668 * @queue_index: sub queue index
3669 *
3670 * Resume individual transmit queue of a device with multiple transmit queues.
3671 */
netif_wake_subqueue(struct net_device * dev,u16 queue_index)3672 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3673 {
3674 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3675
3676 netif_tx_wake_queue(txq);
3677 }
3678
3679 #ifdef CONFIG_XPS
3680 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3681 u16 index);
3682 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3683 u16 index, bool is_rxqs_map);
3684
3685 /**
3686 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3687 * @j: CPU/Rx queue index
3688 * @mask: bitmask of all cpus/rx queues
3689 * @nr_bits: number of bits in the bitmask
3690 *
3691 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3692 */
netif_attr_test_mask(unsigned long j,const unsigned long * mask,unsigned int nr_bits)3693 static inline bool netif_attr_test_mask(unsigned long j,
3694 const unsigned long *mask,
3695 unsigned int nr_bits)
3696 {
3697 cpu_max_bits_warn(j, nr_bits);
3698 return test_bit(j, mask);
3699 }
3700
3701 /**
3702 * netif_attr_test_online - Test for online CPU/Rx queue
3703 * @j: CPU/Rx queue index
3704 * @online_mask: bitmask for CPUs/Rx queues that are online
3705 * @nr_bits: number of bits in the bitmask
3706 *
3707 * Returns true if a CPU/Rx queue is online.
3708 */
netif_attr_test_online(unsigned long j,const unsigned long * online_mask,unsigned int nr_bits)3709 static inline bool netif_attr_test_online(unsigned long j,
3710 const unsigned long *online_mask,
3711 unsigned int nr_bits)
3712 {
3713 cpu_max_bits_warn(j, nr_bits);
3714
3715 if (online_mask)
3716 return test_bit(j, online_mask);
3717
3718 return (j < nr_bits);
3719 }
3720
3721 /**
3722 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3723 * @n: CPU/Rx queue index
3724 * @srcp: the cpumask/Rx queue mask pointer
3725 * @nr_bits: number of bits in the bitmask
3726 *
3727 * Returns >= nr_bits if no further CPUs/Rx queues set.
3728 */
netif_attrmask_next(int n,const unsigned long * srcp,unsigned int nr_bits)3729 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3730 unsigned int nr_bits)
3731 {
3732 /* -1 is a legal arg here. */
3733 if (n != -1)
3734 cpu_max_bits_warn(n, nr_bits);
3735
3736 if (srcp)
3737 return find_next_bit(srcp, nr_bits, n + 1);
3738
3739 return n + 1;
3740 }
3741
3742 /**
3743 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3744 * @n: CPU/Rx queue index
3745 * @src1p: the first CPUs/Rx queues mask pointer
3746 * @src2p: the second CPUs/Rx queues mask pointer
3747 * @nr_bits: number of bits in the bitmask
3748 *
3749 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3750 */
netif_attrmask_next_and(int n,const unsigned long * src1p,const unsigned long * src2p,unsigned int nr_bits)3751 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3752 const unsigned long *src2p,
3753 unsigned int nr_bits)
3754 {
3755 /* -1 is a legal arg here. */
3756 if (n != -1)
3757 cpu_max_bits_warn(n, nr_bits);
3758
3759 if (src1p && src2p)
3760 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3761 else if (src1p)
3762 return find_next_bit(src1p, nr_bits, n + 1);
3763 else if (src2p)
3764 return find_next_bit(src2p, nr_bits, n + 1);
3765
3766 return n + 1;
3767 }
3768 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3769 static inline int netif_set_xps_queue(struct net_device *dev,
3770 const struct cpumask *mask,
3771 u16 index)
3772 {
3773 return 0;
3774 }
3775
__netif_set_xps_queue(struct net_device * dev,const unsigned long * mask,u16 index,bool is_rxqs_map)3776 static inline int __netif_set_xps_queue(struct net_device *dev,
3777 const unsigned long *mask,
3778 u16 index, bool is_rxqs_map)
3779 {
3780 return 0;
3781 }
3782 #endif
3783
3784 /**
3785 * netif_is_multiqueue - test if device has multiple transmit queues
3786 * @dev: network device
3787 *
3788 * Check if device has multiple transmit queues
3789 */
netif_is_multiqueue(const struct net_device * dev)3790 static inline bool netif_is_multiqueue(const struct net_device *dev)
3791 {
3792 return dev->num_tx_queues > 1;
3793 }
3794
3795 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3796
3797 #ifdef CONFIG_SYSFS
3798 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3799 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxqs)3800 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3801 unsigned int rxqs)
3802 {
3803 dev->real_num_rx_queues = rxqs;
3804 return 0;
3805 }
3806 #endif
3807
3808 static inline struct netdev_rx_queue *
__netif_get_rx_queue(struct net_device * dev,unsigned int rxq)3809 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3810 {
3811 return dev->_rx + rxq;
3812 }
3813
3814 #ifdef CONFIG_SYSFS
get_netdev_rx_queue_index(struct netdev_rx_queue * queue)3815 static inline unsigned int get_netdev_rx_queue_index(
3816 struct netdev_rx_queue *queue)
3817 {
3818 struct net_device *dev = queue->dev;
3819 int index = queue - dev->_rx;
3820
3821 BUG_ON(index >= dev->num_rx_queues);
3822 return index;
3823 }
3824 #endif
3825
3826 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3827 int netif_get_num_default_rss_queues(void);
3828
3829 enum skb_free_reason {
3830 SKB_REASON_CONSUMED,
3831 SKB_REASON_DROPPED,
3832 };
3833
3834 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3835 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3836
3837 /*
3838 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3839 * interrupt context or with hardware interrupts being disabled.
3840 * (in_irq() || irqs_disabled())
3841 *
3842 * We provide four helpers that can be used in following contexts :
3843 *
3844 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3845 * replacing kfree_skb(skb)
3846 *
3847 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3848 * Typically used in place of consume_skb(skb) in TX completion path
3849 *
3850 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3851 * replacing kfree_skb(skb)
3852 *
3853 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3854 * and consumed a packet. Used in place of consume_skb(skb)
3855 */
dev_kfree_skb_irq(struct sk_buff * skb)3856 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3857 {
3858 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3859 }
3860
dev_consume_skb_irq(struct sk_buff * skb)3861 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3862 {
3863 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3864 }
3865
dev_kfree_skb_any(struct sk_buff * skb)3866 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3867 {
3868 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3869 }
3870
dev_consume_skb_any(struct sk_buff * skb)3871 static inline void dev_consume_skb_any(struct sk_buff *skb)
3872 {
3873 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3874 }
3875
3876 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3877 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3878 int netif_rx(struct sk_buff *skb);
3879 int netif_rx_ni(struct sk_buff *skb);
3880 int netif_rx_any_context(struct sk_buff *skb);
3881 int netif_receive_skb(struct sk_buff *skb);
3882 int netif_receive_skb_core(struct sk_buff *skb);
3883 void netif_receive_skb_list(struct list_head *head);
3884 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3885 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3886 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3887 gro_result_t napi_gro_frags(struct napi_struct *napi);
3888 struct packet_offload *gro_find_receive_by_type(__be16 type);
3889 struct packet_offload *gro_find_complete_by_type(__be16 type);
3890
napi_free_frags(struct napi_struct * napi)3891 static inline void napi_free_frags(struct napi_struct *napi)
3892 {
3893 kfree_skb(napi->skb);
3894 napi->skb = NULL;
3895 }
3896
3897 bool netdev_is_rx_handler_busy(struct net_device *dev);
3898 int netdev_rx_handler_register(struct net_device *dev,
3899 rx_handler_func_t *rx_handler,
3900 void *rx_handler_data);
3901 void netdev_rx_handler_unregister(struct net_device *dev);
3902
3903 bool dev_valid_name(const char *name);
is_socket_ioctl_cmd(unsigned int cmd)3904 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3905 {
3906 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3907 }
3908 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3909 bool *need_copyout);
3910 int dev_ifconf(struct net *net, struct ifconf *, int);
3911 int dev_ethtool(struct net *net, struct ifreq *);
3912 unsigned int dev_get_flags(const struct net_device *);
3913 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3914 struct netlink_ext_ack *extack);
3915 int dev_change_flags(struct net_device *dev, unsigned int flags,
3916 struct netlink_ext_ack *extack);
3917 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3918 unsigned int gchanges);
3919 int dev_change_name(struct net_device *, const char *);
3920 int dev_set_alias(struct net_device *, const char *, size_t);
3921 int dev_get_alias(const struct net_device *, char *, size_t);
3922 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3923 int __dev_set_mtu(struct net_device *, int);
3924 int dev_validate_mtu(struct net_device *dev, int mtu,
3925 struct netlink_ext_ack *extack);
3926 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3927 struct netlink_ext_ack *extack);
3928 int dev_set_mtu(struct net_device *, int);
3929 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3930 void dev_set_group(struct net_device *, int);
3931 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3932 struct netlink_ext_ack *extack);
3933 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3934 struct netlink_ext_ack *extack);
3935 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3936 struct netlink_ext_ack *extack);
3937 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3938 int dev_change_carrier(struct net_device *, bool new_carrier);
3939 int dev_get_phys_port_id(struct net_device *dev,
3940 struct netdev_phys_item_id *ppid);
3941 int dev_get_phys_port_name(struct net_device *dev,
3942 char *name, size_t len);
3943 int dev_get_port_parent_id(struct net_device *dev,
3944 struct netdev_phys_item_id *ppid, bool recurse);
3945 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3946 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3947 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3948 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3949 u32 value);
3950 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3951 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3952 struct netdev_queue *txq, int *ret);
3953
3954 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3955 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3956 int fd, int expected_fd, u32 flags);
3957 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3958 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3959
3960 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3961
3962 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3963 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3964 bool is_skb_forwardable(const struct net_device *dev,
3965 const struct sk_buff *skb);
3966
____dev_forward_skb(struct net_device * dev,struct sk_buff * skb)3967 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3968 struct sk_buff *skb)
3969 {
3970 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3971 unlikely(!is_skb_forwardable(dev, skb))) {
3972 atomic_long_inc(&dev->rx_dropped);
3973 kfree_skb(skb);
3974 return NET_RX_DROP;
3975 }
3976
3977 skb_scrub_packet(skb, true);
3978 skb->priority = 0;
3979 return 0;
3980 }
3981
3982 bool dev_nit_active(struct net_device *dev);
3983 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3984
3985 extern int netdev_budget;
3986 extern unsigned int netdev_budget_usecs;
3987
3988 /* Called by rtnetlink.c:rtnl_unlock() */
3989 void netdev_run_todo(void);
3990
3991 /**
3992 * dev_put - release reference to device
3993 * @dev: network device
3994 *
3995 * Release reference to device to allow it to be freed.
3996 */
dev_put(struct net_device * dev)3997 static inline void dev_put(struct net_device *dev)
3998 {
3999 if (dev)
4000 this_cpu_dec(*dev->pcpu_refcnt);
4001 }
4002
4003 /**
4004 * dev_hold - get reference to device
4005 * @dev: network device
4006 *
4007 * Hold reference to device to keep it from being freed.
4008 */
dev_hold(struct net_device * dev)4009 static inline void dev_hold(struct net_device *dev)
4010 {
4011 if (dev)
4012 this_cpu_inc(*dev->pcpu_refcnt);
4013 }
4014
4015 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4016 * and _off may be called from IRQ context, but it is caller
4017 * who is responsible for serialization of these calls.
4018 *
4019 * The name carrier is inappropriate, these functions should really be
4020 * called netif_lowerlayer_*() because they represent the state of any
4021 * kind of lower layer not just hardware media.
4022 */
4023
4024 void linkwatch_init_dev(struct net_device *dev);
4025 void linkwatch_fire_event(struct net_device *dev);
4026 void linkwatch_forget_dev(struct net_device *dev);
4027
4028 /**
4029 * netif_carrier_ok - test if carrier present
4030 * @dev: network device
4031 *
4032 * Check if carrier is present on device
4033 */
netif_carrier_ok(const struct net_device * dev)4034 static inline bool netif_carrier_ok(const struct net_device *dev)
4035 {
4036 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4037 }
4038
4039 unsigned long dev_trans_start(struct net_device *dev);
4040
4041 void __netdev_watchdog_up(struct net_device *dev);
4042
4043 void netif_carrier_on(struct net_device *dev);
4044
4045 void netif_carrier_off(struct net_device *dev);
4046
4047 /**
4048 * netif_dormant_on - mark device as dormant.
4049 * @dev: network device
4050 *
4051 * Mark device as dormant (as per RFC2863).
4052 *
4053 * The dormant state indicates that the relevant interface is not
4054 * actually in a condition to pass packets (i.e., it is not 'up') but is
4055 * in a "pending" state, waiting for some external event. For "on-
4056 * demand" interfaces, this new state identifies the situation where the
4057 * interface is waiting for events to place it in the up state.
4058 */
netif_dormant_on(struct net_device * dev)4059 static inline void netif_dormant_on(struct net_device *dev)
4060 {
4061 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4062 linkwatch_fire_event(dev);
4063 }
4064
4065 /**
4066 * netif_dormant_off - set device as not dormant.
4067 * @dev: network device
4068 *
4069 * Device is not in dormant state.
4070 */
netif_dormant_off(struct net_device * dev)4071 static inline void netif_dormant_off(struct net_device *dev)
4072 {
4073 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4074 linkwatch_fire_event(dev);
4075 }
4076
4077 /**
4078 * netif_dormant - test if device is dormant
4079 * @dev: network device
4080 *
4081 * Check if device is dormant.
4082 */
netif_dormant(const struct net_device * dev)4083 static inline bool netif_dormant(const struct net_device *dev)
4084 {
4085 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4086 }
4087
4088
4089 /**
4090 * netif_testing_on - mark device as under test.
4091 * @dev: network device
4092 *
4093 * Mark device as under test (as per RFC2863).
4094 *
4095 * The testing state indicates that some test(s) must be performed on
4096 * the interface. After completion, of the test, the interface state
4097 * will change to up, dormant, or down, as appropriate.
4098 */
netif_testing_on(struct net_device * dev)4099 static inline void netif_testing_on(struct net_device *dev)
4100 {
4101 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4102 linkwatch_fire_event(dev);
4103 }
4104
4105 /**
4106 * netif_testing_off - set device as not under test.
4107 * @dev: network device
4108 *
4109 * Device is not in testing state.
4110 */
netif_testing_off(struct net_device * dev)4111 static inline void netif_testing_off(struct net_device *dev)
4112 {
4113 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4114 linkwatch_fire_event(dev);
4115 }
4116
4117 /**
4118 * netif_testing - test if device is under test
4119 * @dev: network device
4120 *
4121 * Check if device is under test
4122 */
netif_testing(const struct net_device * dev)4123 static inline bool netif_testing(const struct net_device *dev)
4124 {
4125 return test_bit(__LINK_STATE_TESTING, &dev->state);
4126 }
4127
4128
4129 /**
4130 * netif_oper_up - test if device is operational
4131 * @dev: network device
4132 *
4133 * Check if carrier is operational
4134 */
netif_oper_up(const struct net_device * dev)4135 static inline bool netif_oper_up(const struct net_device *dev)
4136 {
4137 return (dev->operstate == IF_OPER_UP ||
4138 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4139 }
4140
4141 /**
4142 * netif_device_present - is device available or removed
4143 * @dev: network device
4144 *
4145 * Check if device has not been removed from system.
4146 */
netif_device_present(struct net_device * dev)4147 static inline bool netif_device_present(struct net_device *dev)
4148 {
4149 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4150 }
4151
4152 void netif_device_detach(struct net_device *dev);
4153
4154 void netif_device_attach(struct net_device *dev);
4155
4156 /*
4157 * Network interface message level settings
4158 */
4159
4160 enum {
4161 NETIF_MSG_DRV_BIT,
4162 NETIF_MSG_PROBE_BIT,
4163 NETIF_MSG_LINK_BIT,
4164 NETIF_MSG_TIMER_BIT,
4165 NETIF_MSG_IFDOWN_BIT,
4166 NETIF_MSG_IFUP_BIT,
4167 NETIF_MSG_RX_ERR_BIT,
4168 NETIF_MSG_TX_ERR_BIT,
4169 NETIF_MSG_TX_QUEUED_BIT,
4170 NETIF_MSG_INTR_BIT,
4171 NETIF_MSG_TX_DONE_BIT,
4172 NETIF_MSG_RX_STATUS_BIT,
4173 NETIF_MSG_PKTDATA_BIT,
4174 NETIF_MSG_HW_BIT,
4175 NETIF_MSG_WOL_BIT,
4176
4177 /* When you add a new bit above, update netif_msg_class_names array
4178 * in net/ethtool/common.c
4179 */
4180 NETIF_MSG_CLASS_COUNT,
4181 };
4182 /* Both ethtool_ops interface and internal driver implementation use u32 */
4183 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4184
4185 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4186 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4187
4188 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4189 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4190 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4191 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4192 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4193 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4194 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4195 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4196 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4197 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4198 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4199 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4200 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4201 #define NETIF_MSG_HW __NETIF_MSG(HW)
4202 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4203
4204 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4205 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4206 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4207 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4208 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4209 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4210 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4211 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4212 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4213 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4214 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4215 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4216 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4217 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4218 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4219
netif_msg_init(int debug_value,int default_msg_enable_bits)4220 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4221 {
4222 /* use default */
4223 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4224 return default_msg_enable_bits;
4225 if (debug_value == 0) /* no output */
4226 return 0;
4227 /* set low N bits */
4228 return (1U << debug_value) - 1;
4229 }
4230
__netif_tx_lock(struct netdev_queue * txq,int cpu)4231 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4232 {
4233 spin_lock(&txq->_xmit_lock);
4234 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4235 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4236 }
4237
__netif_tx_acquire(struct netdev_queue * txq)4238 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4239 {
4240 __acquire(&txq->_xmit_lock);
4241 return true;
4242 }
4243
__netif_tx_release(struct netdev_queue * txq)4244 static inline void __netif_tx_release(struct netdev_queue *txq)
4245 {
4246 __release(&txq->_xmit_lock);
4247 }
4248
__netif_tx_lock_bh(struct netdev_queue * txq)4249 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4250 {
4251 spin_lock_bh(&txq->_xmit_lock);
4252 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4253 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4254 }
4255
__netif_tx_trylock(struct netdev_queue * txq)4256 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4257 {
4258 bool ok = spin_trylock(&txq->_xmit_lock);
4259
4260 if (likely(ok)) {
4261 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4262 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4263 }
4264 return ok;
4265 }
4266
__netif_tx_unlock(struct netdev_queue * txq)4267 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4268 {
4269 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4270 WRITE_ONCE(txq->xmit_lock_owner, -1);
4271 spin_unlock(&txq->_xmit_lock);
4272 }
4273
__netif_tx_unlock_bh(struct netdev_queue * txq)4274 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4275 {
4276 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4277 WRITE_ONCE(txq->xmit_lock_owner, -1);
4278 spin_unlock_bh(&txq->_xmit_lock);
4279 }
4280
txq_trans_update(struct netdev_queue * txq)4281 static inline void txq_trans_update(struct netdev_queue *txq)
4282 {
4283 if (txq->xmit_lock_owner != -1)
4284 txq->trans_start = jiffies;
4285 }
4286
4287 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
netif_trans_update(struct net_device * dev)4288 static inline void netif_trans_update(struct net_device *dev)
4289 {
4290 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4291
4292 if (txq->trans_start != jiffies)
4293 txq->trans_start = jiffies;
4294 }
4295
4296 /**
4297 * netif_tx_lock - grab network device transmit lock
4298 * @dev: network device
4299 *
4300 * Get network device transmit lock
4301 */
netif_tx_lock(struct net_device * dev)4302 static inline void netif_tx_lock(struct net_device *dev)
4303 {
4304 unsigned int i;
4305 int cpu;
4306
4307 spin_lock(&dev->tx_global_lock);
4308 cpu = smp_processor_id();
4309 for (i = 0; i < dev->num_tx_queues; i++) {
4310 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4311
4312 /* We are the only thread of execution doing a
4313 * freeze, but we have to grab the _xmit_lock in
4314 * order to synchronize with threads which are in
4315 * the ->hard_start_xmit() handler and already
4316 * checked the frozen bit.
4317 */
4318 __netif_tx_lock(txq, cpu);
4319 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4320 __netif_tx_unlock(txq);
4321 }
4322 }
4323
netif_tx_lock_bh(struct net_device * dev)4324 static inline void netif_tx_lock_bh(struct net_device *dev)
4325 {
4326 local_bh_disable();
4327 netif_tx_lock(dev);
4328 }
4329
netif_tx_unlock(struct net_device * dev)4330 static inline void netif_tx_unlock(struct net_device *dev)
4331 {
4332 unsigned int i;
4333
4334 for (i = 0; i < dev->num_tx_queues; i++) {
4335 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4336
4337 /* No need to grab the _xmit_lock here. If the
4338 * queue is not stopped for another reason, we
4339 * force a schedule.
4340 */
4341 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4342 netif_schedule_queue(txq);
4343 }
4344 spin_unlock(&dev->tx_global_lock);
4345 }
4346
netif_tx_unlock_bh(struct net_device * dev)4347 static inline void netif_tx_unlock_bh(struct net_device *dev)
4348 {
4349 netif_tx_unlock(dev);
4350 local_bh_enable();
4351 }
4352
4353 #define HARD_TX_LOCK(dev, txq, cpu) { \
4354 if ((dev->features & NETIF_F_LLTX) == 0) { \
4355 __netif_tx_lock(txq, cpu); \
4356 } else { \
4357 __netif_tx_acquire(txq); \
4358 } \
4359 }
4360
4361 #define HARD_TX_TRYLOCK(dev, txq) \
4362 (((dev->features & NETIF_F_LLTX) == 0) ? \
4363 __netif_tx_trylock(txq) : \
4364 __netif_tx_acquire(txq))
4365
4366 #define HARD_TX_UNLOCK(dev, txq) { \
4367 if ((dev->features & NETIF_F_LLTX) == 0) { \
4368 __netif_tx_unlock(txq); \
4369 } else { \
4370 __netif_tx_release(txq); \
4371 } \
4372 }
4373
netif_tx_disable(struct net_device * dev)4374 static inline void netif_tx_disable(struct net_device *dev)
4375 {
4376 unsigned int i;
4377 int cpu;
4378
4379 local_bh_disable();
4380 cpu = smp_processor_id();
4381 spin_lock(&dev->tx_global_lock);
4382 for (i = 0; i < dev->num_tx_queues; i++) {
4383 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4384
4385 __netif_tx_lock(txq, cpu);
4386 netif_tx_stop_queue(txq);
4387 __netif_tx_unlock(txq);
4388 }
4389 spin_unlock(&dev->tx_global_lock);
4390 local_bh_enable();
4391 }
4392
netif_addr_lock(struct net_device * dev)4393 static inline void netif_addr_lock(struct net_device *dev)
4394 {
4395 unsigned char nest_level = 0;
4396
4397 #ifdef CONFIG_LOCKDEP
4398 nest_level = dev->nested_level;
4399 #endif
4400 spin_lock_nested(&dev->addr_list_lock, nest_level);
4401 }
4402
netif_addr_lock_bh(struct net_device * dev)4403 static inline void netif_addr_lock_bh(struct net_device *dev)
4404 {
4405 unsigned char nest_level = 0;
4406
4407 #ifdef CONFIG_LOCKDEP
4408 nest_level = dev->nested_level;
4409 #endif
4410 local_bh_disable();
4411 spin_lock_nested(&dev->addr_list_lock, nest_level);
4412 }
4413
netif_addr_unlock(struct net_device * dev)4414 static inline void netif_addr_unlock(struct net_device *dev)
4415 {
4416 spin_unlock(&dev->addr_list_lock);
4417 }
4418
netif_addr_unlock_bh(struct net_device * dev)4419 static inline void netif_addr_unlock_bh(struct net_device *dev)
4420 {
4421 spin_unlock_bh(&dev->addr_list_lock);
4422 }
4423
4424 /*
4425 * dev_addrs walker. Should be used only for read access. Call with
4426 * rcu_read_lock held.
4427 */
4428 #define for_each_dev_addr(dev, ha) \
4429 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4430
4431 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4432
4433 void ether_setup(struct net_device *dev);
4434
4435 /* Support for loadable net-drivers */
4436 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4437 unsigned char name_assign_type,
4438 void (*setup)(struct net_device *),
4439 unsigned int txqs, unsigned int rxqs);
4440 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4441 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4442
4443 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4444 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4445 count)
4446
4447 int register_netdev(struct net_device *dev);
4448 void unregister_netdev(struct net_device *dev);
4449
4450 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4451
4452 /* General hardware address lists handling functions */
4453 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4454 struct netdev_hw_addr_list *from_list, int addr_len);
4455 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4456 struct netdev_hw_addr_list *from_list, int addr_len);
4457 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4458 struct net_device *dev,
4459 int (*sync)(struct net_device *, const unsigned char *),
4460 int (*unsync)(struct net_device *,
4461 const unsigned char *));
4462 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4463 struct net_device *dev,
4464 int (*sync)(struct net_device *,
4465 const unsigned char *, int),
4466 int (*unsync)(struct net_device *,
4467 const unsigned char *, int));
4468 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4469 struct net_device *dev,
4470 int (*unsync)(struct net_device *,
4471 const unsigned char *, int));
4472 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4473 struct net_device *dev,
4474 int (*unsync)(struct net_device *,
4475 const unsigned char *));
4476 void __hw_addr_init(struct netdev_hw_addr_list *list);
4477
4478 /* Functions used for device addresses handling */
4479 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4480 unsigned char addr_type);
4481 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4482 unsigned char addr_type);
4483 void dev_addr_flush(struct net_device *dev);
4484 int dev_addr_init(struct net_device *dev);
4485
4486 /* Functions used for unicast addresses handling */
4487 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4488 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4489 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4490 int dev_uc_sync(struct net_device *to, struct net_device *from);
4491 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4492 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4493 void dev_uc_flush(struct net_device *dev);
4494 void dev_uc_init(struct net_device *dev);
4495
4496 /**
4497 * __dev_uc_sync - Synchonize device's unicast list
4498 * @dev: device to sync
4499 * @sync: function to call if address should be added
4500 * @unsync: function to call if address should be removed
4501 *
4502 * Add newly added addresses to the interface, and release
4503 * addresses that have been deleted.
4504 */
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4505 static inline int __dev_uc_sync(struct net_device *dev,
4506 int (*sync)(struct net_device *,
4507 const unsigned char *),
4508 int (*unsync)(struct net_device *,
4509 const unsigned char *))
4510 {
4511 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4512 }
4513
4514 /**
4515 * __dev_uc_unsync - Remove synchronized addresses from device
4516 * @dev: device to sync
4517 * @unsync: function to call if address should be removed
4518 *
4519 * Remove all addresses that were added to the device by dev_uc_sync().
4520 */
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4521 static inline void __dev_uc_unsync(struct net_device *dev,
4522 int (*unsync)(struct net_device *,
4523 const unsigned char *))
4524 {
4525 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4526 }
4527
4528 /* Functions used for multicast addresses handling */
4529 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4530 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4531 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4532 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4533 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4534 int dev_mc_sync(struct net_device *to, struct net_device *from);
4535 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4536 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4537 void dev_mc_flush(struct net_device *dev);
4538 void dev_mc_init(struct net_device *dev);
4539
4540 /**
4541 * __dev_mc_sync - Synchonize device's multicast list
4542 * @dev: device to sync
4543 * @sync: function to call if address should be added
4544 * @unsync: function to call if address should be removed
4545 *
4546 * Add newly added addresses to the interface, and release
4547 * addresses that have been deleted.
4548 */
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4549 static inline int __dev_mc_sync(struct net_device *dev,
4550 int (*sync)(struct net_device *,
4551 const unsigned char *),
4552 int (*unsync)(struct net_device *,
4553 const unsigned char *))
4554 {
4555 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4556 }
4557
4558 /**
4559 * __dev_mc_unsync - Remove synchronized addresses from device
4560 * @dev: device to sync
4561 * @unsync: function to call if address should be removed
4562 *
4563 * Remove all addresses that were added to the device by dev_mc_sync().
4564 */
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4565 static inline void __dev_mc_unsync(struct net_device *dev,
4566 int (*unsync)(struct net_device *,
4567 const unsigned char *))
4568 {
4569 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4570 }
4571
4572 /* Functions used for secondary unicast and multicast support */
4573 void dev_set_rx_mode(struct net_device *dev);
4574 void __dev_set_rx_mode(struct net_device *dev);
4575 int dev_set_promiscuity(struct net_device *dev, int inc);
4576 int dev_set_allmulti(struct net_device *dev, int inc);
4577 void netdev_state_change(struct net_device *dev);
4578 void netdev_notify_peers(struct net_device *dev);
4579 void netdev_features_change(struct net_device *dev);
4580 /* Load a device via the kmod */
4581 void dev_load(struct net *net, const char *name);
4582 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4583 struct rtnl_link_stats64 *storage);
4584 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4585 const struct net_device_stats *netdev_stats);
4586 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4587 const struct pcpu_sw_netstats __percpu *netstats);
4588
4589 extern int netdev_max_backlog;
4590 extern int netdev_tstamp_prequeue;
4591 extern int weight_p;
4592 extern int dev_weight_rx_bias;
4593 extern int dev_weight_tx_bias;
4594 extern int dev_rx_weight;
4595 extern int dev_tx_weight;
4596 extern int gro_normal_batch;
4597
4598 enum {
4599 NESTED_SYNC_IMM_BIT,
4600 NESTED_SYNC_TODO_BIT,
4601 };
4602
4603 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4604 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4605
4606 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4607 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4608
4609 struct netdev_nested_priv {
4610 unsigned char flags;
4611 void *data;
4612 };
4613
4614 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4615 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4616 struct list_head **iter);
4617 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4618 struct list_head **iter);
4619
4620 #ifdef CONFIG_LOCKDEP
4621 static LIST_HEAD(net_unlink_list);
4622
net_unlink_todo(struct net_device * dev)4623 static inline void net_unlink_todo(struct net_device *dev)
4624 {
4625 if (list_empty(&dev->unlink_list))
4626 list_add_tail(&dev->unlink_list, &net_unlink_list);
4627 }
4628 #endif
4629
4630 /* iterate through upper list, must be called under RCU read lock */
4631 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4632 for (iter = &(dev)->adj_list.upper, \
4633 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4634 updev; \
4635 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4636
4637 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4638 int (*fn)(struct net_device *upper_dev,
4639 struct netdev_nested_priv *priv),
4640 struct netdev_nested_priv *priv);
4641
4642 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4643 struct net_device *upper_dev);
4644
4645 bool netdev_has_any_upper_dev(struct net_device *dev);
4646
4647 void *netdev_lower_get_next_private(struct net_device *dev,
4648 struct list_head **iter);
4649 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4650 struct list_head **iter);
4651
4652 #define netdev_for_each_lower_private(dev, priv, iter) \
4653 for (iter = (dev)->adj_list.lower.next, \
4654 priv = netdev_lower_get_next_private(dev, &(iter)); \
4655 priv; \
4656 priv = netdev_lower_get_next_private(dev, &(iter)))
4657
4658 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4659 for (iter = &(dev)->adj_list.lower, \
4660 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4661 priv; \
4662 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4663
4664 void *netdev_lower_get_next(struct net_device *dev,
4665 struct list_head **iter);
4666
4667 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4668 for (iter = (dev)->adj_list.lower.next, \
4669 ldev = netdev_lower_get_next(dev, &(iter)); \
4670 ldev; \
4671 ldev = netdev_lower_get_next(dev, &(iter)))
4672
4673 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4674 struct list_head **iter);
4675 int netdev_walk_all_lower_dev(struct net_device *dev,
4676 int (*fn)(struct net_device *lower_dev,
4677 struct netdev_nested_priv *priv),
4678 struct netdev_nested_priv *priv);
4679 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4680 int (*fn)(struct net_device *lower_dev,
4681 struct netdev_nested_priv *priv),
4682 struct netdev_nested_priv *priv);
4683
4684 void *netdev_adjacent_get_private(struct list_head *adj_list);
4685 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4686 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4687 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4688 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4689 struct netlink_ext_ack *extack);
4690 int netdev_master_upper_dev_link(struct net_device *dev,
4691 struct net_device *upper_dev,
4692 void *upper_priv, void *upper_info,
4693 struct netlink_ext_ack *extack);
4694 void netdev_upper_dev_unlink(struct net_device *dev,
4695 struct net_device *upper_dev);
4696 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4697 struct net_device *new_dev,
4698 struct net_device *dev,
4699 struct netlink_ext_ack *extack);
4700 void netdev_adjacent_change_commit(struct net_device *old_dev,
4701 struct net_device *new_dev,
4702 struct net_device *dev);
4703 void netdev_adjacent_change_abort(struct net_device *old_dev,
4704 struct net_device *new_dev,
4705 struct net_device *dev);
4706 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4707 void *netdev_lower_dev_get_private(struct net_device *dev,
4708 struct net_device *lower_dev);
4709 void netdev_lower_state_changed(struct net_device *lower_dev,
4710 void *lower_state_info);
4711
4712 /* RSS keys are 40 or 52 bytes long */
4713 #define NETDEV_RSS_KEY_LEN 52
4714 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4715 void netdev_rss_key_fill(void *buffer, size_t len);
4716
4717 int skb_checksum_help(struct sk_buff *skb);
4718 int skb_crc32c_csum_help(struct sk_buff *skb);
4719 int skb_csum_hwoffload_help(struct sk_buff *skb,
4720 const netdev_features_t features);
4721
4722 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4723 netdev_features_t features, bool tx_path);
4724 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4725 netdev_features_t features);
4726
4727 struct netdev_bonding_info {
4728 ifslave slave;
4729 ifbond master;
4730 };
4731
4732 struct netdev_notifier_bonding_info {
4733 struct netdev_notifier_info info; /* must be first */
4734 struct netdev_bonding_info bonding_info;
4735 };
4736
4737 void netdev_bonding_info_change(struct net_device *dev,
4738 struct netdev_bonding_info *bonding_info);
4739
4740 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4741 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4742 #else
ethtool_notify(struct net_device * dev,unsigned int cmd,const void * data)4743 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4744 const void *data)
4745 {
4746 }
4747 #endif
4748
4749 static inline
skb_gso_segment(struct sk_buff * skb,netdev_features_t features)4750 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4751 {
4752 return __skb_gso_segment(skb, features, true);
4753 }
4754 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4755
can_checksum_protocol(netdev_features_t features,__be16 protocol)4756 static inline bool can_checksum_protocol(netdev_features_t features,
4757 __be16 protocol)
4758 {
4759 if (protocol == htons(ETH_P_FCOE))
4760 return !!(features & NETIF_F_FCOE_CRC);
4761
4762 /* Assume this is an IP checksum (not SCTP CRC) */
4763
4764 if (features & NETIF_F_HW_CSUM) {
4765 /* Can checksum everything */
4766 return true;
4767 }
4768
4769 switch (protocol) {
4770 case htons(ETH_P_IP):
4771 return !!(features & NETIF_F_IP_CSUM);
4772 case htons(ETH_P_IPV6):
4773 return !!(features & NETIF_F_IPV6_CSUM);
4774 default:
4775 return false;
4776 }
4777 }
4778
4779 #ifdef CONFIG_BUG
4780 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4781 #else
netdev_rx_csum_fault(struct net_device * dev,struct sk_buff * skb)4782 static inline void netdev_rx_csum_fault(struct net_device *dev,
4783 struct sk_buff *skb)
4784 {
4785 }
4786 #endif
4787 /* rx skb timestamps */
4788 void net_enable_timestamp(void);
4789 void net_disable_timestamp(void);
4790
4791 #ifdef CONFIG_PROC_FS
4792 int __init dev_proc_init(void);
4793 #else
4794 #define dev_proc_init() 0
4795 #endif
4796
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)4797 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4798 struct sk_buff *skb, struct net_device *dev,
4799 bool more)
4800 {
4801 __this_cpu_write(softnet_data.xmit.more, more);
4802 return ops->ndo_start_xmit(skb, dev);
4803 }
4804
netdev_xmit_more(void)4805 static inline bool netdev_xmit_more(void)
4806 {
4807 return __this_cpu_read(softnet_data.xmit.more);
4808 }
4809
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)4810 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4811 struct netdev_queue *txq, bool more)
4812 {
4813 const struct net_device_ops *ops = dev->netdev_ops;
4814 netdev_tx_t rc;
4815
4816 rc = __netdev_start_xmit(ops, skb, dev, more);
4817 if (rc == NETDEV_TX_OK)
4818 txq_trans_update(txq);
4819
4820 return rc;
4821 }
4822
4823 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4824 const void *ns);
4825 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4826 const void *ns);
4827
4828 extern const struct kobj_ns_type_operations net_ns_type_operations;
4829
4830 const char *netdev_drivername(const struct net_device *dev);
4831
4832 void linkwatch_run_queue(void);
4833
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)4834 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4835 netdev_features_t f2)
4836 {
4837 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4838 if (f1 & NETIF_F_HW_CSUM)
4839 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4840 else
4841 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4842 }
4843
4844 return f1 & f2;
4845 }
4846
netdev_get_wanted_features(struct net_device * dev)4847 static inline netdev_features_t netdev_get_wanted_features(
4848 struct net_device *dev)
4849 {
4850 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4851 }
4852 netdev_features_t netdev_increment_features(netdev_features_t all,
4853 netdev_features_t one, netdev_features_t mask);
4854
4855 /* Allow TSO being used on stacked device :
4856 * Performing the GSO segmentation before last device
4857 * is a performance improvement.
4858 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)4859 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4860 netdev_features_t mask)
4861 {
4862 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4863 }
4864
4865 int __netdev_update_features(struct net_device *dev);
4866 void netdev_update_features(struct net_device *dev);
4867 void netdev_change_features(struct net_device *dev);
4868
4869 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4870 struct net_device *dev);
4871
4872 netdev_features_t passthru_features_check(struct sk_buff *skb,
4873 struct net_device *dev,
4874 netdev_features_t features);
4875 netdev_features_t netif_skb_features(struct sk_buff *skb);
4876
net_gso_ok(netdev_features_t features,int gso_type)4877 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4878 {
4879 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4880
4881 /* check flags correspondence */
4882 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4883 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4884 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4885 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4886 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4887 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4888 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4889 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4890 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4891 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4892 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4893 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4894 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4895 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4896 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4897 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4898 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4899 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4900 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4901
4902 return (features & feature) == feature;
4903 }
4904
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)4905 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4906 {
4907 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4908 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4909 }
4910
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)4911 static inline bool netif_needs_gso(struct sk_buff *skb,
4912 netdev_features_t features)
4913 {
4914 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4915 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4916 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4917 }
4918
netif_set_gso_max_size(struct net_device * dev,unsigned int size)4919 static inline void netif_set_gso_max_size(struct net_device *dev,
4920 unsigned int size)
4921 {
4922 dev->gso_max_size = size;
4923 }
4924
skb_gso_error_unwind(struct sk_buff * skb,__be16 protocol,int pulled_hlen,u16 mac_offset,int mac_len)4925 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4926 int pulled_hlen, u16 mac_offset,
4927 int mac_len)
4928 {
4929 skb->protocol = protocol;
4930 skb->encapsulation = 1;
4931 skb_push(skb, pulled_hlen);
4932 skb_reset_transport_header(skb);
4933 skb->mac_header = mac_offset;
4934 skb->network_header = skb->mac_header + mac_len;
4935 skb->mac_len = mac_len;
4936 }
4937
netif_is_macsec(const struct net_device * dev)4938 static inline bool netif_is_macsec(const struct net_device *dev)
4939 {
4940 return dev->priv_flags & IFF_MACSEC;
4941 }
4942
netif_is_macvlan(const struct net_device * dev)4943 static inline bool netif_is_macvlan(const struct net_device *dev)
4944 {
4945 return dev->priv_flags & IFF_MACVLAN;
4946 }
4947
netif_is_macvlan_port(const struct net_device * dev)4948 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4949 {
4950 return dev->priv_flags & IFF_MACVLAN_PORT;
4951 }
4952
netif_is_bond_master(const struct net_device * dev)4953 static inline bool netif_is_bond_master(const struct net_device *dev)
4954 {
4955 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4956 }
4957
netif_is_bond_slave(const struct net_device * dev)4958 static inline bool netif_is_bond_slave(const struct net_device *dev)
4959 {
4960 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4961 }
4962
netif_supports_nofcs(struct net_device * dev)4963 static inline bool netif_supports_nofcs(struct net_device *dev)
4964 {
4965 return dev->priv_flags & IFF_SUPP_NOFCS;
4966 }
4967
netif_has_l3_rx_handler(const struct net_device * dev)4968 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4969 {
4970 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4971 }
4972
netif_is_l3_master(const struct net_device * dev)4973 static inline bool netif_is_l3_master(const struct net_device *dev)
4974 {
4975 return dev->priv_flags & IFF_L3MDEV_MASTER;
4976 }
4977
netif_is_l3_slave(const struct net_device * dev)4978 static inline bool netif_is_l3_slave(const struct net_device *dev)
4979 {
4980 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4981 }
4982
netif_is_bridge_master(const struct net_device * dev)4983 static inline bool netif_is_bridge_master(const struct net_device *dev)
4984 {
4985 return dev->priv_flags & IFF_EBRIDGE;
4986 }
4987
netif_is_bridge_port(const struct net_device * dev)4988 static inline bool netif_is_bridge_port(const struct net_device *dev)
4989 {
4990 return dev->priv_flags & IFF_BRIDGE_PORT;
4991 }
4992
netif_is_ovs_master(const struct net_device * dev)4993 static inline bool netif_is_ovs_master(const struct net_device *dev)
4994 {
4995 return dev->priv_flags & IFF_OPENVSWITCH;
4996 }
4997
netif_is_ovs_port(const struct net_device * dev)4998 static inline bool netif_is_ovs_port(const struct net_device *dev)
4999 {
5000 return dev->priv_flags & IFF_OVS_DATAPATH;
5001 }
5002
netif_is_any_bridge_port(const struct net_device * dev)5003 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5004 {
5005 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5006 }
5007
netif_is_team_master(const struct net_device * dev)5008 static inline bool netif_is_team_master(const struct net_device *dev)
5009 {
5010 return dev->priv_flags & IFF_TEAM;
5011 }
5012
netif_is_team_port(const struct net_device * dev)5013 static inline bool netif_is_team_port(const struct net_device *dev)
5014 {
5015 return dev->priv_flags & IFF_TEAM_PORT;
5016 }
5017
netif_is_lag_master(const struct net_device * dev)5018 static inline bool netif_is_lag_master(const struct net_device *dev)
5019 {
5020 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5021 }
5022
netif_is_lag_port(const struct net_device * dev)5023 static inline bool netif_is_lag_port(const struct net_device *dev)
5024 {
5025 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5026 }
5027
netif_is_rxfh_configured(const struct net_device * dev)5028 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5029 {
5030 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5031 }
5032
netif_is_failover(const struct net_device * dev)5033 static inline bool netif_is_failover(const struct net_device *dev)
5034 {
5035 return dev->priv_flags & IFF_FAILOVER;
5036 }
5037
netif_is_failover_slave(const struct net_device * dev)5038 static inline bool netif_is_failover_slave(const struct net_device *dev)
5039 {
5040 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5041 }
5042
5043 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)5044 static inline void netif_keep_dst(struct net_device *dev)
5045 {
5046 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5047 }
5048
5049 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
netif_reduces_vlan_mtu(struct net_device * dev)5050 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5051 {
5052 /* TODO: reserve and use an additional IFF bit, if we get more users */
5053 return dev->priv_flags & IFF_MACSEC;
5054 }
5055
5056 extern struct pernet_operations __net_initdata loopback_net_ops;
5057
5058 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5059
5060 /* netdev_printk helpers, similar to dev_printk */
5061
netdev_name(const struct net_device * dev)5062 static inline const char *netdev_name(const struct net_device *dev)
5063 {
5064 if (!dev->name[0] || strchr(dev->name, '%'))
5065 return "(unnamed net_device)";
5066 return dev->name;
5067 }
5068
netdev_unregistering(const struct net_device * dev)5069 static inline bool netdev_unregistering(const struct net_device *dev)
5070 {
5071 return dev->reg_state == NETREG_UNREGISTERING;
5072 }
5073
netdev_reg_state(const struct net_device * dev)5074 static inline const char *netdev_reg_state(const struct net_device *dev)
5075 {
5076 switch (dev->reg_state) {
5077 case NETREG_UNINITIALIZED: return " (uninitialized)";
5078 case NETREG_REGISTERED: return "";
5079 case NETREG_UNREGISTERING: return " (unregistering)";
5080 case NETREG_UNREGISTERED: return " (unregistered)";
5081 case NETREG_RELEASED: return " (released)";
5082 case NETREG_DUMMY: return " (dummy)";
5083 }
5084
5085 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5086 return " (unknown)";
5087 }
5088
5089 __printf(3, 4) __cold
5090 void netdev_printk(const char *level, const struct net_device *dev,
5091 const char *format, ...);
5092 __printf(2, 3) __cold
5093 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5094 __printf(2, 3) __cold
5095 void netdev_alert(const struct net_device *dev, const char *format, ...);
5096 __printf(2, 3) __cold
5097 void netdev_crit(const struct net_device *dev, const char *format, ...);
5098 __printf(2, 3) __cold
5099 void netdev_err(const struct net_device *dev, const char *format, ...);
5100 __printf(2, 3) __cold
5101 void netdev_warn(const struct net_device *dev, const char *format, ...);
5102 __printf(2, 3) __cold
5103 void netdev_notice(const struct net_device *dev, const char *format, ...);
5104 __printf(2, 3) __cold
5105 void netdev_info(const struct net_device *dev, const char *format, ...);
5106
5107 #define netdev_level_once(level, dev, fmt, ...) \
5108 do { \
5109 static bool __print_once __read_mostly; \
5110 \
5111 if (!__print_once) { \
5112 __print_once = true; \
5113 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5114 } \
5115 } while (0)
5116
5117 #define netdev_emerg_once(dev, fmt, ...) \
5118 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5119 #define netdev_alert_once(dev, fmt, ...) \
5120 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5121 #define netdev_crit_once(dev, fmt, ...) \
5122 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5123 #define netdev_err_once(dev, fmt, ...) \
5124 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5125 #define netdev_warn_once(dev, fmt, ...) \
5126 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5127 #define netdev_notice_once(dev, fmt, ...) \
5128 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5129 #define netdev_info_once(dev, fmt, ...) \
5130 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5131
5132 #define MODULE_ALIAS_NETDEV(device) \
5133 MODULE_ALIAS("netdev-" device)
5134
5135 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5136 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5137 #define netdev_dbg(__dev, format, args...) \
5138 do { \
5139 dynamic_netdev_dbg(__dev, format, ##args); \
5140 } while (0)
5141 #elif defined(DEBUG)
5142 #define netdev_dbg(__dev, format, args...) \
5143 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5144 #else
5145 #define netdev_dbg(__dev, format, args...) \
5146 ({ \
5147 if (0) \
5148 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5149 })
5150 #endif
5151
5152 #if defined(VERBOSE_DEBUG)
5153 #define netdev_vdbg netdev_dbg
5154 #else
5155
5156 #define netdev_vdbg(dev, format, args...) \
5157 ({ \
5158 if (0) \
5159 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5160 0; \
5161 })
5162 #endif
5163
5164 /*
5165 * netdev_WARN() acts like dev_printk(), but with the key difference
5166 * of using a WARN/WARN_ON to get the message out, including the
5167 * file/line information and a backtrace.
5168 */
5169 #define netdev_WARN(dev, format, args...) \
5170 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5171 netdev_reg_state(dev), ##args)
5172
5173 #define netdev_WARN_ONCE(dev, format, args...) \
5174 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5175 netdev_reg_state(dev), ##args)
5176
5177 /* netif printk helpers, similar to netdev_printk */
5178
5179 #define netif_printk(priv, type, level, dev, fmt, args...) \
5180 do { \
5181 if (netif_msg_##type(priv)) \
5182 netdev_printk(level, (dev), fmt, ##args); \
5183 } while (0)
5184
5185 #define netif_level(level, priv, type, dev, fmt, args...) \
5186 do { \
5187 if (netif_msg_##type(priv)) \
5188 netdev_##level(dev, fmt, ##args); \
5189 } while (0)
5190
5191 #define netif_emerg(priv, type, dev, fmt, args...) \
5192 netif_level(emerg, priv, type, dev, fmt, ##args)
5193 #define netif_alert(priv, type, dev, fmt, args...) \
5194 netif_level(alert, priv, type, dev, fmt, ##args)
5195 #define netif_crit(priv, type, dev, fmt, args...) \
5196 netif_level(crit, priv, type, dev, fmt, ##args)
5197 #define netif_err(priv, type, dev, fmt, args...) \
5198 netif_level(err, priv, type, dev, fmt, ##args)
5199 #define netif_warn(priv, type, dev, fmt, args...) \
5200 netif_level(warn, priv, type, dev, fmt, ##args)
5201 #define netif_notice(priv, type, dev, fmt, args...) \
5202 netif_level(notice, priv, type, dev, fmt, ##args)
5203 #define netif_info(priv, type, dev, fmt, args...) \
5204 netif_level(info, priv, type, dev, fmt, ##args)
5205
5206 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5207 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5208 #define netif_dbg(priv, type, netdev, format, args...) \
5209 do { \
5210 if (netif_msg_##type(priv)) \
5211 dynamic_netdev_dbg(netdev, format, ##args); \
5212 } while (0)
5213 #elif defined(DEBUG)
5214 #define netif_dbg(priv, type, dev, format, args...) \
5215 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5216 #else
5217 #define netif_dbg(priv, type, dev, format, args...) \
5218 ({ \
5219 if (0) \
5220 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5221 0; \
5222 })
5223 #endif
5224
5225 /* if @cond then downgrade to debug, else print at @level */
5226 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5227 do { \
5228 if (cond) \
5229 netif_dbg(priv, type, netdev, fmt, ##args); \
5230 else \
5231 netif_ ## level(priv, type, netdev, fmt, ##args); \
5232 } while (0)
5233
5234 #if defined(VERBOSE_DEBUG)
5235 #define netif_vdbg netif_dbg
5236 #else
5237 #define netif_vdbg(priv, type, dev, format, args...) \
5238 ({ \
5239 if (0) \
5240 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5241 0; \
5242 })
5243 #endif
5244
5245 /*
5246 * The list of packet types we will receive (as opposed to discard)
5247 * and the routines to invoke.
5248 *
5249 * Why 16. Because with 16 the only overlap we get on a hash of the
5250 * low nibble of the protocol value is RARP/SNAP/X.25.
5251 *
5252 * 0800 IP
5253 * 0001 802.3
5254 * 0002 AX.25
5255 * 0004 802.2
5256 * 8035 RARP
5257 * 0005 SNAP
5258 * 0805 X.25
5259 * 0806 ARP
5260 * 8137 IPX
5261 * 0009 Localtalk
5262 * 86DD IPv6
5263 */
5264 #define PTYPE_HASH_SIZE (16)
5265 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5266
5267 extern struct net_device *blackhole_netdev;
5268
5269 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5270 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5271 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5272 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5273
5274 #endif /* _LINUX_NETDEVICE_H */
5275