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 + READ_ONCE((dev)->needed_headroom)) \
274 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
275 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
276 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
277 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
278
279 struct header_ops {
280 int (*create) (struct sk_buff *skb, struct net_device *dev,
281 unsigned short type, const void *daddr,
282 const void *saddr, unsigned int len);
283 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
284 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
285 void (*cache_update)(struct hh_cache *hh,
286 const struct net_device *dev,
287 const unsigned char *haddr);
288 bool (*validate)(const char *ll_header, unsigned int len);
289 __be16 (*parse_protocol)(const struct sk_buff *skb);
290 };
291
292 /* These flag bits are private to the generic network queueing
293 * layer; they may not be explicitly referenced by any other
294 * code.
295 */
296
297 enum netdev_state_t {
298 __LINK_STATE_START,
299 __LINK_STATE_PRESENT,
300 __LINK_STATE_NOCARRIER,
301 __LINK_STATE_LINKWATCH_PENDING,
302 __LINK_STATE_DORMANT,
303 __LINK_STATE_TESTING,
304 };
305
306
307 /*
308 * This structure holds boot-time configured netdevice settings. They
309 * are then used in the device probing.
310 */
311 struct netdev_boot_setup {
312 char name[IFNAMSIZ];
313 struct ifmap map;
314 };
315 #define NETDEV_BOOT_SETUP_MAX 8
316
317 int __init netdev_boot_setup(char *str);
318
319 struct gro_list {
320 struct list_head list;
321 int count;
322 };
323
324 /*
325 * size of gro hash buckets, must less than bit number of
326 * napi_struct::gro_bitmask
327 */
328 #define GRO_HASH_BUCKETS 8
329
330 /*
331 * Structure for NAPI scheduling similar to tasklet but with weighting
332 */
333 struct napi_struct {
334 /* The poll_list must only be managed by the entity which
335 * changes the state of the NAPI_STATE_SCHED bit. This means
336 * whoever atomically sets that bit can add this napi_struct
337 * to the per-CPU poll_list, and whoever clears that bit
338 * can remove from the list right before clearing the bit.
339 */
340 struct list_head poll_list;
341
342 unsigned long state;
343 int weight;
344 u32 defer_hard_irqs_count;
345 unsigned long gro_bitmask;
346 int (*poll)(struct napi_struct *, int);
347 #ifdef CONFIG_NETPOLL
348 int poll_owner;
349 #endif
350 struct net_device *dev;
351 struct gro_list gro_hash[GRO_HASH_BUCKETS];
352 struct sk_buff *skb;
353 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
354 int rx_count; /* length of rx_list */
355 struct hrtimer timer;
356 struct list_head dev_list;
357 struct hlist_node napi_hash_node;
358 unsigned int napi_id;
359 };
360
361 enum {
362 NAPI_STATE_SCHED, /* Poll is scheduled */
363 NAPI_STATE_MISSED, /* reschedule a napi */
364 NAPI_STATE_DISABLE, /* Disable pending */
365 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
366 NAPI_STATE_LISTED, /* NAPI added to system lists */
367 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
368 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
369 };
370
371 enum {
372 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
373 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
374 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
375 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
376 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
377 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
378 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
379 };
380
381 enum gro_result {
382 GRO_MERGED,
383 GRO_MERGED_FREE,
384 GRO_HELD,
385 GRO_NORMAL,
386 GRO_DROP,
387 GRO_CONSUMED,
388 };
389 typedef enum gro_result gro_result_t;
390
391 /*
392 * enum rx_handler_result - Possible return values for rx_handlers.
393 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
394 * further.
395 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
396 * case skb->dev was changed by rx_handler.
397 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
398 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
399 *
400 * rx_handlers are functions called from inside __netif_receive_skb(), to do
401 * special processing of the skb, prior to delivery to protocol handlers.
402 *
403 * Currently, a net_device can only have a single rx_handler registered. Trying
404 * to register a second rx_handler will return -EBUSY.
405 *
406 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
407 * To unregister a rx_handler on a net_device, use
408 * netdev_rx_handler_unregister().
409 *
410 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
411 * do with the skb.
412 *
413 * If the rx_handler consumed the skb in some way, it should return
414 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
415 * the skb to be delivered in some other way.
416 *
417 * If the rx_handler changed skb->dev, to divert the skb to another
418 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
419 * new device will be called if it exists.
420 *
421 * If the rx_handler decides the skb should be ignored, it should return
422 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
423 * are registered on exact device (ptype->dev == skb->dev).
424 *
425 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
426 * delivered, it should return RX_HANDLER_PASS.
427 *
428 * A device without a registered rx_handler will behave as if rx_handler
429 * returned RX_HANDLER_PASS.
430 */
431
432 enum rx_handler_result {
433 RX_HANDLER_CONSUMED,
434 RX_HANDLER_ANOTHER,
435 RX_HANDLER_EXACT,
436 RX_HANDLER_PASS,
437 };
438 typedef enum rx_handler_result rx_handler_result_t;
439 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
440
441 void __napi_schedule(struct napi_struct *n);
442 void __napi_schedule_irqoff(struct napi_struct *n);
443
napi_disable_pending(struct napi_struct * n)444 static inline bool napi_disable_pending(struct napi_struct *n)
445 {
446 return test_bit(NAPI_STATE_DISABLE, &n->state);
447 }
448
449 bool napi_schedule_prep(struct napi_struct *n);
450
451 /**
452 * napi_schedule - schedule NAPI poll
453 * @n: NAPI context
454 *
455 * Schedule NAPI poll routine to be called if it is not already
456 * running.
457 */
napi_schedule(struct napi_struct * n)458 static inline void napi_schedule(struct napi_struct *n)
459 {
460 if (napi_schedule_prep(n))
461 __napi_schedule(n);
462 }
463
464 /**
465 * napi_schedule_irqoff - schedule NAPI poll
466 * @n: NAPI context
467 *
468 * Variant of napi_schedule(), assuming hard irqs are masked.
469 */
napi_schedule_irqoff(struct napi_struct * n)470 static inline void napi_schedule_irqoff(struct napi_struct *n)
471 {
472 if (napi_schedule_prep(n))
473 __napi_schedule_irqoff(n);
474 }
475
476 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
napi_reschedule(struct napi_struct * napi)477 static inline bool napi_reschedule(struct napi_struct *napi)
478 {
479 if (napi_schedule_prep(napi)) {
480 __napi_schedule(napi);
481 return true;
482 }
483 return false;
484 }
485
486 bool napi_complete_done(struct napi_struct *n, int work_done);
487 /**
488 * napi_complete - NAPI processing complete
489 * @n: NAPI context
490 *
491 * Mark NAPI processing as complete.
492 * Consider using napi_complete_done() instead.
493 * Return false if device should avoid rearming interrupts.
494 */
napi_complete(struct napi_struct * n)495 static inline bool napi_complete(struct napi_struct *n)
496 {
497 return napi_complete_done(n, 0);
498 }
499
500 /**
501 * napi_disable - prevent NAPI from scheduling
502 * @n: NAPI context
503 *
504 * Stop NAPI from being scheduled on this context.
505 * Waits till any outstanding processing completes.
506 */
507 void napi_disable(struct napi_struct *n);
508
509 /**
510 * napi_enable - enable NAPI scheduling
511 * @n: NAPI context
512 *
513 * Resume NAPI from being scheduled on this context.
514 * Must be paired with napi_disable.
515 */
napi_enable(struct napi_struct * n)516 static inline void napi_enable(struct napi_struct *n)
517 {
518 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
519 smp_mb__before_atomic();
520 clear_bit(NAPI_STATE_SCHED, &n->state);
521 clear_bit(NAPI_STATE_NPSVC, &n->state);
522 }
523
524 /**
525 * napi_synchronize - wait until NAPI is not running
526 * @n: NAPI context
527 *
528 * Wait until NAPI is done being scheduled on this context.
529 * Waits till any outstanding processing completes but
530 * does not disable future activations.
531 */
napi_synchronize(const struct napi_struct * n)532 static inline void napi_synchronize(const struct napi_struct *n)
533 {
534 if (IS_ENABLED(CONFIG_SMP))
535 while (test_bit(NAPI_STATE_SCHED, &n->state))
536 msleep(1);
537 else
538 barrier();
539 }
540
541 /**
542 * napi_if_scheduled_mark_missed - if napi is running, set the
543 * NAPIF_STATE_MISSED
544 * @n: NAPI context
545 *
546 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
547 * NAPI is scheduled.
548 **/
napi_if_scheduled_mark_missed(struct napi_struct * n)549 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
550 {
551 unsigned long val, new;
552
553 do {
554 val = READ_ONCE(n->state);
555 if (val & NAPIF_STATE_DISABLE)
556 return true;
557
558 if (!(val & NAPIF_STATE_SCHED))
559 return false;
560
561 new = val | NAPIF_STATE_MISSED;
562 } while (cmpxchg(&n->state, val, new) != val);
563
564 return true;
565 }
566
567 enum netdev_queue_state_t {
568 __QUEUE_STATE_DRV_XOFF,
569 __QUEUE_STATE_STACK_XOFF,
570 __QUEUE_STATE_FROZEN,
571 };
572
573 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
574 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
575 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
576
577 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
578 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
579 QUEUE_STATE_FROZEN)
580 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
581 QUEUE_STATE_FROZEN)
582
583 /*
584 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
585 * netif_tx_* functions below are used to manipulate this flag. The
586 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
587 * queue independently. The netif_xmit_*stopped functions below are called
588 * to check if the queue has been stopped by the driver or stack (either
589 * of the XOFF bits are set in the state). Drivers should not need to call
590 * netif_xmit*stopped functions, they should only be using netif_tx_*.
591 */
592
593 struct netdev_queue {
594 /*
595 * read-mostly part
596 */
597 struct net_device *dev;
598 struct Qdisc __rcu *qdisc;
599 struct Qdisc *qdisc_sleeping;
600 #ifdef CONFIG_SYSFS
601 struct kobject kobj;
602 #endif
603 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
604 int numa_node;
605 #endif
606 unsigned long tx_maxrate;
607 /*
608 * Number of TX timeouts for this queue
609 * (/sys/class/net/DEV/Q/trans_timeout)
610 */
611 unsigned long trans_timeout;
612
613 /* Subordinate device that the queue has been assigned to */
614 struct net_device *sb_dev;
615 #ifdef CONFIG_XDP_SOCKETS
616 struct xsk_buff_pool *pool;
617 #endif
618 /*
619 * write-mostly part
620 */
621 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
622 int xmit_lock_owner;
623 /*
624 * Time (in jiffies) of last Tx
625 */
626 unsigned long trans_start;
627
628 unsigned long state;
629
630 #ifdef CONFIG_BQL
631 struct dql dql;
632 #endif
633 } ____cacheline_aligned_in_smp;
634
635 extern int sysctl_fb_tunnels_only_for_init_net;
636 extern int sysctl_devconf_inherit_init_net;
637
638 /*
639 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
640 * == 1 : For initns only
641 * == 2 : For none.
642 */
net_has_fallback_tunnels(const struct net * net)643 static inline bool net_has_fallback_tunnels(const struct net *net)
644 {
645 #if IS_ENABLED(CONFIG_SYSCTL)
646 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
647
648 return !fb_tunnels_only_for_init_net ||
649 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
650 #else
651 return true;
652 #endif
653 }
654
net_inherit_devconf(void)655 static inline int net_inherit_devconf(void)
656 {
657 #if IS_ENABLED(CONFIG_SYSCTL)
658 return READ_ONCE(sysctl_devconf_inherit_init_net);
659 #else
660 return 0;
661 #endif
662 }
663
netdev_queue_numa_node_read(const struct netdev_queue * q)664 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
665 {
666 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
667 return q->numa_node;
668 #else
669 return NUMA_NO_NODE;
670 #endif
671 }
672
netdev_queue_numa_node_write(struct netdev_queue * q,int node)673 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
674 {
675 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
676 q->numa_node = node;
677 #endif
678 }
679
680 #ifdef CONFIG_RPS
681 /*
682 * This structure holds an RPS map which can be of variable length. The
683 * map is an array of CPUs.
684 */
685 struct rps_map {
686 unsigned int len;
687 struct rcu_head rcu;
688 u16 cpus[];
689 };
690 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
691
692 /*
693 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
694 * tail pointer for that CPU's input queue at the time of last enqueue, and
695 * a hardware filter index.
696 */
697 struct rps_dev_flow {
698 u16 cpu;
699 u16 filter;
700 unsigned int last_qtail;
701 };
702 #define RPS_NO_FILTER 0xffff
703
704 /*
705 * The rps_dev_flow_table structure contains a table of flow mappings.
706 */
707 struct rps_dev_flow_table {
708 unsigned int mask;
709 struct rcu_head rcu;
710 struct rps_dev_flow flows[];
711 };
712 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
713 ((_num) * sizeof(struct rps_dev_flow)))
714
715 /*
716 * The rps_sock_flow_table contains mappings of flows to the last CPU
717 * on which they were processed by the application (set in recvmsg).
718 * Each entry is a 32bit value. Upper part is the high-order bits
719 * of flow hash, lower part is CPU number.
720 * rps_cpu_mask is used to partition the space, depending on number of
721 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
722 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
723 * meaning we use 32-6=26 bits for the hash.
724 */
725 struct rps_sock_flow_table {
726 u32 mask;
727
728 u32 ents[] ____cacheline_aligned_in_smp;
729 };
730 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
731
732 #define RPS_NO_CPU 0xffff
733
734 extern u32 rps_cpu_mask;
735 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
736
rps_record_sock_flow(struct rps_sock_flow_table * table,u32 hash)737 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
738 u32 hash)
739 {
740 if (table && hash) {
741 unsigned int index = hash & table->mask;
742 u32 val = hash & ~rps_cpu_mask;
743
744 /* We only give a hint, preemption can change CPU under us */
745 val |= raw_smp_processor_id();
746
747 /* The following WRITE_ONCE() is paired with the READ_ONCE()
748 * here, and another one in get_rps_cpu().
749 */
750 if (READ_ONCE(table->ents[index]) != val)
751 WRITE_ONCE(table->ents[index], val);
752 }
753 }
754
755 #ifdef CONFIG_RFS_ACCEL
756 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
757 u16 filter_id);
758 #endif
759 #endif /* CONFIG_RPS */
760
761 /* This structure contains an instance of an RX queue. */
762 struct netdev_rx_queue {
763 #ifdef CONFIG_RPS
764 struct rps_map __rcu *rps_map;
765 struct rps_dev_flow_table __rcu *rps_flow_table;
766 #endif
767 struct kobject kobj;
768 struct net_device *dev;
769 struct xdp_rxq_info xdp_rxq;
770 #ifdef CONFIG_XDP_SOCKETS
771 struct xsk_buff_pool *pool;
772 #endif
773 } ____cacheline_aligned_in_smp;
774
775 /*
776 * RX queue sysfs structures and functions.
777 */
778 struct rx_queue_attribute {
779 struct attribute attr;
780 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
781 ssize_t (*store)(struct netdev_rx_queue *queue,
782 const char *buf, size_t len);
783 };
784
785 #ifdef CONFIG_XPS
786 /*
787 * This structure holds an XPS map which can be of variable length. The
788 * map is an array of queues.
789 */
790 struct xps_map {
791 unsigned int len;
792 unsigned int alloc_len;
793 struct rcu_head rcu;
794 u16 queues[];
795 };
796 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
797 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
798 - sizeof(struct xps_map)) / sizeof(u16))
799
800 /*
801 * This structure holds all XPS maps for device. Maps are indexed by CPU.
802 */
803 struct xps_dev_maps {
804 struct rcu_head rcu;
805 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
806 };
807
808 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
809 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
810
811 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
812 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
813
814 #endif /* CONFIG_XPS */
815
816 #define TC_MAX_QUEUE 16
817 #define TC_BITMASK 15
818 /* HW offloaded queuing disciplines txq count and offset maps */
819 struct netdev_tc_txq {
820 u16 count;
821 u16 offset;
822 };
823
824 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
825 /*
826 * This structure is to hold information about the device
827 * configured to run FCoE protocol stack.
828 */
829 struct netdev_fcoe_hbainfo {
830 char manufacturer[64];
831 char serial_number[64];
832 char hardware_version[64];
833 char driver_version[64];
834 char optionrom_version[64];
835 char firmware_version[64];
836 char model[256];
837 char model_description[256];
838 };
839 #endif
840
841 #define MAX_PHYS_ITEM_ID_LEN 32
842
843 /* This structure holds a unique identifier to identify some
844 * physical item (port for example) used by a netdevice.
845 */
846 struct netdev_phys_item_id {
847 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
848 unsigned char id_len;
849 };
850
netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)851 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
852 struct netdev_phys_item_id *b)
853 {
854 return a->id_len == b->id_len &&
855 memcmp(a->id, b->id, a->id_len) == 0;
856 }
857
858 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
859 struct sk_buff *skb,
860 struct net_device *sb_dev);
861
862 enum tc_setup_type {
863 TC_SETUP_QDISC_MQPRIO,
864 TC_SETUP_CLSU32,
865 TC_SETUP_CLSFLOWER,
866 TC_SETUP_CLSMATCHALL,
867 TC_SETUP_CLSBPF,
868 TC_SETUP_BLOCK,
869 TC_SETUP_QDISC_CBS,
870 TC_SETUP_QDISC_RED,
871 TC_SETUP_QDISC_PRIO,
872 TC_SETUP_QDISC_MQ,
873 TC_SETUP_QDISC_ETF,
874 TC_SETUP_ROOT_QDISC,
875 TC_SETUP_QDISC_GRED,
876 TC_SETUP_QDISC_TAPRIO,
877 TC_SETUP_FT,
878 TC_SETUP_QDISC_ETS,
879 TC_SETUP_QDISC_TBF,
880 TC_SETUP_QDISC_FIFO,
881 };
882
883 /* These structures hold the attributes of bpf state that are being passed
884 * to the netdevice through the bpf op.
885 */
886 enum bpf_netdev_command {
887 /* Set or clear a bpf program used in the earliest stages of packet
888 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
889 * is responsible for calling bpf_prog_put on any old progs that are
890 * stored. In case of error, the callee need not release the new prog
891 * reference, but on success it takes ownership and must bpf_prog_put
892 * when it is no longer used.
893 */
894 XDP_SETUP_PROG,
895 XDP_SETUP_PROG_HW,
896 /* BPF program for offload callbacks, invoked at program load time. */
897 BPF_OFFLOAD_MAP_ALLOC,
898 BPF_OFFLOAD_MAP_FREE,
899 XDP_SETUP_XSK_POOL,
900 };
901
902 struct bpf_prog_offload_ops;
903 struct netlink_ext_ack;
904 struct xdp_umem;
905 struct xdp_dev_bulk_queue;
906 struct bpf_xdp_link;
907
908 enum bpf_xdp_mode {
909 XDP_MODE_SKB = 0,
910 XDP_MODE_DRV = 1,
911 XDP_MODE_HW = 2,
912 __MAX_XDP_MODE
913 };
914
915 struct bpf_xdp_entity {
916 struct bpf_prog *prog;
917 struct bpf_xdp_link *link;
918 };
919
920 struct netdev_bpf {
921 enum bpf_netdev_command command;
922 union {
923 /* XDP_SETUP_PROG */
924 struct {
925 u32 flags;
926 struct bpf_prog *prog;
927 struct netlink_ext_ack *extack;
928 };
929 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
930 struct {
931 struct bpf_offloaded_map *offmap;
932 };
933 /* XDP_SETUP_XSK_POOL */
934 struct {
935 struct xsk_buff_pool *pool;
936 u16 queue_id;
937 } xsk;
938 };
939 };
940
941 /* Flags for ndo_xsk_wakeup. */
942 #define XDP_WAKEUP_RX (1 << 0)
943 #define XDP_WAKEUP_TX (1 << 1)
944
945 #ifdef CONFIG_XFRM_OFFLOAD
946 struct xfrmdev_ops {
947 int (*xdo_dev_state_add) (struct xfrm_state *x);
948 void (*xdo_dev_state_delete) (struct xfrm_state *x);
949 void (*xdo_dev_state_free) (struct xfrm_state *x);
950 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
951 struct xfrm_state *x);
952 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
953 };
954 #endif
955
956 struct dev_ifalias {
957 struct rcu_head rcuhead;
958 char ifalias[];
959 };
960
961 struct devlink;
962 struct tlsdev_ops;
963
964 struct netdev_name_node {
965 struct hlist_node hlist;
966 struct list_head list;
967 struct net_device *dev;
968 const char *name;
969 };
970
971 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
972 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
973
974 struct netdev_net_notifier {
975 struct list_head list;
976 struct notifier_block *nb;
977 };
978
979 /*
980 * This structure defines the management hooks for network devices.
981 * The following hooks can be defined; unless noted otherwise, they are
982 * optional and can be filled with a null pointer.
983 *
984 * int (*ndo_init)(struct net_device *dev);
985 * This function is called once when a network device is registered.
986 * The network device can use this for any late stage initialization
987 * or semantic validation. It can fail with an error code which will
988 * be propagated back to register_netdev.
989 *
990 * void (*ndo_uninit)(struct net_device *dev);
991 * This function is called when device is unregistered or when registration
992 * fails. It is not called if init fails.
993 *
994 * int (*ndo_open)(struct net_device *dev);
995 * This function is called when a network device transitions to the up
996 * state.
997 *
998 * int (*ndo_stop)(struct net_device *dev);
999 * This function is called when a network device transitions to the down
1000 * state.
1001 *
1002 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1003 * struct net_device *dev);
1004 * Called when a packet needs to be transmitted.
1005 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1006 * the queue before that can happen; it's for obsolete devices and weird
1007 * corner cases, but the stack really does a non-trivial amount
1008 * of useless work if you return NETDEV_TX_BUSY.
1009 * Required; cannot be NULL.
1010 *
1011 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1012 * struct net_device *dev
1013 * netdev_features_t features);
1014 * Called by core transmit path to determine if device is capable of
1015 * performing offload operations on a given packet. This is to give
1016 * the device an opportunity to implement any restrictions that cannot
1017 * be otherwise expressed by feature flags. The check is called with
1018 * the set of features that the stack has calculated and it returns
1019 * those the driver believes to be appropriate.
1020 *
1021 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1022 * struct net_device *sb_dev);
1023 * Called to decide which queue to use when device supports multiple
1024 * transmit queues.
1025 *
1026 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1027 * This function is called to allow device receiver to make
1028 * changes to configuration when multicast or promiscuous is enabled.
1029 *
1030 * void (*ndo_set_rx_mode)(struct net_device *dev);
1031 * This function is called device changes address list filtering.
1032 * If driver handles unicast address filtering, it should set
1033 * IFF_UNICAST_FLT in its priv_flags.
1034 *
1035 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1036 * This function is called when the Media Access Control address
1037 * needs to be changed. If this interface is not defined, the
1038 * MAC address can not be changed.
1039 *
1040 * int (*ndo_validate_addr)(struct net_device *dev);
1041 * Test if Media Access Control address is valid for the device.
1042 *
1043 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1044 * Called when a user requests an ioctl which can't be handled by
1045 * the generic interface code. If not defined ioctls return
1046 * not supported error code.
1047 *
1048 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1049 * Used to set network devices bus interface parameters. This interface
1050 * is retained for legacy reasons; new devices should use the bus
1051 * interface (PCI) for low level management.
1052 *
1053 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1054 * Called when a user wants to change the Maximum Transfer Unit
1055 * of a device.
1056 *
1057 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1058 * Callback used when the transmitter has not made any progress
1059 * for dev->watchdog ticks.
1060 *
1061 * void (*ndo_get_stats64)(struct net_device *dev,
1062 * struct rtnl_link_stats64 *storage);
1063 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1064 * Called when a user wants to get the network device usage
1065 * statistics. Drivers must do one of the following:
1066 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1067 * rtnl_link_stats64 structure passed by the caller.
1068 * 2. Define @ndo_get_stats to update a net_device_stats structure
1069 * (which should normally be dev->stats) and return a pointer to
1070 * it. The structure may be changed asynchronously only if each
1071 * field is written atomically.
1072 * 3. Update dev->stats asynchronously and atomically, and define
1073 * neither operation.
1074 *
1075 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1076 * Return true if this device supports offload stats of this attr_id.
1077 *
1078 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1079 * void *attr_data)
1080 * Get statistics for offload operations by attr_id. Write it into the
1081 * attr_data pointer.
1082 *
1083 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1084 * If device supports VLAN filtering this function is called when a
1085 * VLAN id is registered.
1086 *
1087 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1088 * If device supports VLAN filtering this function is called when a
1089 * VLAN id is unregistered.
1090 *
1091 * void (*ndo_poll_controller)(struct net_device *dev);
1092 *
1093 * SR-IOV management functions.
1094 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1095 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1096 * u8 qos, __be16 proto);
1097 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1098 * int max_tx_rate);
1099 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1100 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1101 * int (*ndo_get_vf_config)(struct net_device *dev,
1102 * int vf, struct ifla_vf_info *ivf);
1103 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1104 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1105 * struct nlattr *port[]);
1106 *
1107 * Enable or disable the VF ability to query its RSS Redirection Table and
1108 * Hash Key. This is needed since on some devices VF share this information
1109 * with PF and querying it may introduce a theoretical security risk.
1110 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1111 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1112 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1113 * void *type_data);
1114 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1115 * This is always called from the stack with the rtnl lock held and netif
1116 * tx queues stopped. This allows the netdevice to perform queue
1117 * management safely.
1118 *
1119 * Fiber Channel over Ethernet (FCoE) offload functions.
1120 * int (*ndo_fcoe_enable)(struct net_device *dev);
1121 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1122 * so the underlying device can perform whatever needed configuration or
1123 * initialization to support acceleration of FCoE traffic.
1124 *
1125 * int (*ndo_fcoe_disable)(struct net_device *dev);
1126 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1127 * so the underlying device can perform whatever needed clean-ups to
1128 * stop supporting acceleration of FCoE traffic.
1129 *
1130 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1131 * struct scatterlist *sgl, unsigned int sgc);
1132 * Called when the FCoE Initiator wants to initialize an I/O that
1133 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1134 * perform necessary setup and returns 1 to indicate the device is set up
1135 * successfully to perform DDP on this I/O, otherwise this returns 0.
1136 *
1137 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1138 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1139 * indicated by the FC exchange id 'xid', so the underlying device can
1140 * clean up and reuse resources for later DDP requests.
1141 *
1142 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1143 * struct scatterlist *sgl, unsigned int sgc);
1144 * Called when the FCoE Target wants to initialize an I/O that
1145 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1146 * perform necessary setup and returns 1 to indicate the device is set up
1147 * successfully to perform DDP on this I/O, otherwise this returns 0.
1148 *
1149 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1150 * struct netdev_fcoe_hbainfo *hbainfo);
1151 * Called when the FCoE Protocol stack wants information on the underlying
1152 * device. This information is utilized by the FCoE protocol stack to
1153 * register attributes with Fiber Channel management service as per the
1154 * FC-GS Fabric Device Management Information(FDMI) specification.
1155 *
1156 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1157 * Called when the underlying device wants to override default World Wide
1158 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1159 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1160 * protocol stack to use.
1161 *
1162 * RFS acceleration.
1163 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1164 * u16 rxq_index, u32 flow_id);
1165 * Set hardware filter for RFS. rxq_index is the target queue index;
1166 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1167 * Return the filter ID on success, or a negative error code.
1168 *
1169 * Slave management functions (for bridge, bonding, etc).
1170 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1171 * Called to make another netdev an underling.
1172 *
1173 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1174 * Called to release previously enslaved netdev.
1175 *
1176 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1177 * struct sk_buff *skb,
1178 * bool all_slaves);
1179 * Get the xmit slave of master device. If all_slaves is true, function
1180 * assume all the slaves can transmit.
1181 *
1182 * Feature/offload setting functions.
1183 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1184 * netdev_features_t features);
1185 * Adjusts the requested feature flags according to device-specific
1186 * constraints, and returns the resulting flags. Must not modify
1187 * the device state.
1188 *
1189 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1190 * Called to update device configuration to new features. Passed
1191 * feature set might be less than what was returned by ndo_fix_features()).
1192 * Must return >0 or -errno if it changed dev->features itself.
1193 *
1194 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1195 * struct net_device *dev,
1196 * const unsigned char *addr, u16 vid, u16 flags,
1197 * struct netlink_ext_ack *extack);
1198 * Adds an FDB entry to dev for addr.
1199 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1200 * struct net_device *dev,
1201 * const unsigned char *addr, u16 vid)
1202 * Deletes the FDB entry from dev coresponding to addr.
1203 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1204 * struct net_device *dev, struct net_device *filter_dev,
1205 * int *idx)
1206 * Used to add FDB entries to dump requests. Implementers should add
1207 * entries to skb and update idx with the number of entries.
1208 *
1209 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1210 * u16 flags, struct netlink_ext_ack *extack)
1211 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1212 * struct net_device *dev, u32 filter_mask,
1213 * int nlflags)
1214 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1215 * u16 flags);
1216 *
1217 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1218 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1219 * which do not represent real hardware may define this to allow their
1220 * userspace components to manage their virtual carrier state. Devices
1221 * that determine carrier state from physical hardware properties (eg
1222 * network cables) or protocol-dependent mechanisms (eg
1223 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1224 *
1225 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1226 * struct netdev_phys_item_id *ppid);
1227 * Called to get ID of physical port of this device. If driver does
1228 * not implement this, it is assumed that the hw is not able to have
1229 * multiple net devices on single physical port.
1230 *
1231 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1232 * struct netdev_phys_item_id *ppid)
1233 * Called to get the parent ID of the physical port of this device.
1234 *
1235 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1236 * struct udp_tunnel_info *ti);
1237 * Called by UDP tunnel to notify a driver about the UDP port and socket
1238 * address family that a UDP tunnel is listnening to. It is called only
1239 * when a new port starts listening. The operation is protected by the
1240 * RTNL.
1241 *
1242 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1243 * struct udp_tunnel_info *ti);
1244 * Called by UDP tunnel to notify the driver about a UDP port and socket
1245 * address family that the UDP tunnel is not listening to anymore. The
1246 * operation is protected by the RTNL.
1247 *
1248 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1249 * struct net_device *dev)
1250 * Called by upper layer devices to accelerate switching or other
1251 * station functionality into hardware. 'pdev is the lowerdev
1252 * to use for the offload and 'dev' is the net device that will
1253 * back the offload. Returns a pointer to the private structure
1254 * the upper layer will maintain.
1255 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1256 * Called by upper layer device to delete the station created
1257 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1258 * the station and priv is the structure returned by the add
1259 * operation.
1260 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1261 * int queue_index, u32 maxrate);
1262 * Called when a user wants to set a max-rate limitation of specific
1263 * TX queue.
1264 * int (*ndo_get_iflink)(const struct net_device *dev);
1265 * Called to get the iflink value of this device.
1266 * void (*ndo_change_proto_down)(struct net_device *dev,
1267 * bool proto_down);
1268 * This function is used to pass protocol port error state information
1269 * to the switch driver. The switch driver can react to the proto_down
1270 * by doing a phys down on the associated switch port.
1271 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1272 * This function is used to get egress tunnel information for given skb.
1273 * This is useful for retrieving outer tunnel header parameters while
1274 * sampling packet.
1275 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1276 * This function is used to specify the headroom that the skb must
1277 * consider when allocation skb during packet reception. Setting
1278 * appropriate rx headroom value allows avoiding skb head copy on
1279 * forward. Setting a negative value resets the rx headroom to the
1280 * default value.
1281 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1282 * This function is used to set or query state related to XDP on the
1283 * netdevice and manage BPF offload. See definition of
1284 * enum bpf_netdev_command for details.
1285 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1286 * u32 flags);
1287 * This function is used to submit @n XDP packets for transmit on a
1288 * netdevice. Returns number of frames successfully transmitted, frames
1289 * that got dropped are freed/returned via xdp_return_frame().
1290 * Returns negative number, means general error invoking ndo, meaning
1291 * no frames were xmit'ed and core-caller will free all frames.
1292 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1293 * This function is used to wake up the softirq, ksoftirqd or kthread
1294 * responsible for sending and/or receiving packets on a specific
1295 * queue id bound to an AF_XDP socket. The flags field specifies if
1296 * only RX, only Tx, or both should be woken up using the flags
1297 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1298 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1299 * Get devlink port instance associated with a given netdev.
1300 * Called with a reference on the netdevice and devlink locks only,
1301 * rtnl_lock is not held.
1302 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1303 * int cmd);
1304 * Add, change, delete or get information on an IPv4 tunnel.
1305 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1306 * If a device is paired with a peer device, return the peer instance.
1307 * The caller must be under RCU read context.
1308 */
1309 struct net_device_ops {
1310 int (*ndo_init)(struct net_device *dev);
1311 void (*ndo_uninit)(struct net_device *dev);
1312 int (*ndo_open)(struct net_device *dev);
1313 int (*ndo_stop)(struct net_device *dev);
1314 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1315 struct net_device *dev);
1316 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1317 struct net_device *dev,
1318 netdev_features_t features);
1319 u16 (*ndo_select_queue)(struct net_device *dev,
1320 struct sk_buff *skb,
1321 struct net_device *sb_dev);
1322 void (*ndo_change_rx_flags)(struct net_device *dev,
1323 int flags);
1324 void (*ndo_set_rx_mode)(struct net_device *dev);
1325 int (*ndo_set_mac_address)(struct net_device *dev,
1326 void *addr);
1327 int (*ndo_validate_addr)(struct net_device *dev);
1328 int (*ndo_do_ioctl)(struct net_device *dev,
1329 struct ifreq *ifr, int cmd);
1330 int (*ndo_set_config)(struct net_device *dev,
1331 struct ifmap *map);
1332 int (*ndo_change_mtu)(struct net_device *dev,
1333 int new_mtu);
1334 int (*ndo_neigh_setup)(struct net_device *dev,
1335 struct neigh_parms *);
1336 void (*ndo_tx_timeout) (struct net_device *dev,
1337 unsigned int txqueue);
1338
1339 void (*ndo_get_stats64)(struct net_device *dev,
1340 struct rtnl_link_stats64 *storage);
1341 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1342 int (*ndo_get_offload_stats)(int attr_id,
1343 const struct net_device *dev,
1344 void *attr_data);
1345 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1346
1347 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1348 __be16 proto, u16 vid);
1349 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1350 __be16 proto, u16 vid);
1351 #ifdef CONFIG_NET_POLL_CONTROLLER
1352 void (*ndo_poll_controller)(struct net_device *dev);
1353 int (*ndo_netpoll_setup)(struct net_device *dev,
1354 struct netpoll_info *info);
1355 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1356 #endif
1357 int (*ndo_set_vf_mac)(struct net_device *dev,
1358 int queue, u8 *mac);
1359 int (*ndo_set_vf_vlan)(struct net_device *dev,
1360 int queue, u16 vlan,
1361 u8 qos, __be16 proto);
1362 int (*ndo_set_vf_rate)(struct net_device *dev,
1363 int vf, int min_tx_rate,
1364 int max_tx_rate);
1365 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1366 int vf, bool setting);
1367 int (*ndo_set_vf_trust)(struct net_device *dev,
1368 int vf, bool setting);
1369 int (*ndo_get_vf_config)(struct net_device *dev,
1370 int vf,
1371 struct ifla_vf_info *ivf);
1372 int (*ndo_set_vf_link_state)(struct net_device *dev,
1373 int vf, int link_state);
1374 int (*ndo_get_vf_stats)(struct net_device *dev,
1375 int vf,
1376 struct ifla_vf_stats
1377 *vf_stats);
1378 int (*ndo_set_vf_port)(struct net_device *dev,
1379 int vf,
1380 struct nlattr *port[]);
1381 int (*ndo_get_vf_port)(struct net_device *dev,
1382 int vf, struct sk_buff *skb);
1383 int (*ndo_get_vf_guid)(struct net_device *dev,
1384 int vf,
1385 struct ifla_vf_guid *node_guid,
1386 struct ifla_vf_guid *port_guid);
1387 int (*ndo_set_vf_guid)(struct net_device *dev,
1388 int vf, u64 guid,
1389 int guid_type);
1390 int (*ndo_set_vf_rss_query_en)(
1391 struct net_device *dev,
1392 int vf, bool setting);
1393 int (*ndo_setup_tc)(struct net_device *dev,
1394 enum tc_setup_type type,
1395 void *type_data);
1396 #if IS_ENABLED(CONFIG_FCOE)
1397 int (*ndo_fcoe_enable)(struct net_device *dev);
1398 int (*ndo_fcoe_disable)(struct net_device *dev);
1399 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1400 u16 xid,
1401 struct scatterlist *sgl,
1402 unsigned int sgc);
1403 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1404 u16 xid);
1405 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1406 u16 xid,
1407 struct scatterlist *sgl,
1408 unsigned int sgc);
1409 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1410 struct netdev_fcoe_hbainfo *hbainfo);
1411 #endif
1412
1413 #if IS_ENABLED(CONFIG_LIBFCOE)
1414 #define NETDEV_FCOE_WWNN 0
1415 #define NETDEV_FCOE_WWPN 1
1416 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1417 u64 *wwn, int type);
1418 #endif
1419
1420 #ifdef CONFIG_RFS_ACCEL
1421 int (*ndo_rx_flow_steer)(struct net_device *dev,
1422 const struct sk_buff *skb,
1423 u16 rxq_index,
1424 u32 flow_id);
1425 #endif
1426 int (*ndo_add_slave)(struct net_device *dev,
1427 struct net_device *slave_dev,
1428 struct netlink_ext_ack *extack);
1429 int (*ndo_del_slave)(struct net_device *dev,
1430 struct net_device *slave_dev);
1431 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1432 struct sk_buff *skb,
1433 bool all_slaves);
1434 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1435 netdev_features_t features);
1436 int (*ndo_set_features)(struct net_device *dev,
1437 netdev_features_t features);
1438 int (*ndo_neigh_construct)(struct net_device *dev,
1439 struct neighbour *n);
1440 void (*ndo_neigh_destroy)(struct net_device *dev,
1441 struct neighbour *n);
1442
1443 int (*ndo_fdb_add)(struct ndmsg *ndm,
1444 struct nlattr *tb[],
1445 struct net_device *dev,
1446 const unsigned char *addr,
1447 u16 vid,
1448 u16 flags,
1449 struct netlink_ext_ack *extack);
1450 int (*ndo_fdb_del)(struct ndmsg *ndm,
1451 struct nlattr *tb[],
1452 struct net_device *dev,
1453 const unsigned char *addr,
1454 u16 vid);
1455 int (*ndo_fdb_dump)(struct sk_buff *skb,
1456 struct netlink_callback *cb,
1457 struct net_device *dev,
1458 struct net_device *filter_dev,
1459 int *idx);
1460 int (*ndo_fdb_get)(struct sk_buff *skb,
1461 struct nlattr *tb[],
1462 struct net_device *dev,
1463 const unsigned char *addr,
1464 u16 vid, u32 portid, u32 seq,
1465 struct netlink_ext_ack *extack);
1466 int (*ndo_bridge_setlink)(struct net_device *dev,
1467 struct nlmsghdr *nlh,
1468 u16 flags,
1469 struct netlink_ext_ack *extack);
1470 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1471 u32 pid, u32 seq,
1472 struct net_device *dev,
1473 u32 filter_mask,
1474 int nlflags);
1475 int (*ndo_bridge_dellink)(struct net_device *dev,
1476 struct nlmsghdr *nlh,
1477 u16 flags);
1478 int (*ndo_change_carrier)(struct net_device *dev,
1479 bool new_carrier);
1480 int (*ndo_get_phys_port_id)(struct net_device *dev,
1481 struct netdev_phys_item_id *ppid);
1482 int (*ndo_get_port_parent_id)(struct net_device *dev,
1483 struct netdev_phys_item_id *ppid);
1484 int (*ndo_get_phys_port_name)(struct net_device *dev,
1485 char *name, size_t len);
1486 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1487 struct udp_tunnel_info *ti);
1488 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1489 struct udp_tunnel_info *ti);
1490 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1491 struct net_device *dev);
1492 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1493 void *priv);
1494
1495 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1496 int queue_index,
1497 u32 maxrate);
1498 int (*ndo_get_iflink)(const struct net_device *dev);
1499 int (*ndo_change_proto_down)(struct net_device *dev,
1500 bool proto_down);
1501 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1502 struct sk_buff *skb);
1503 void (*ndo_set_rx_headroom)(struct net_device *dev,
1504 int needed_headroom);
1505 int (*ndo_bpf)(struct net_device *dev,
1506 struct netdev_bpf *bpf);
1507 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1508 struct xdp_frame **xdp,
1509 u32 flags);
1510 int (*ndo_xsk_wakeup)(struct net_device *dev,
1511 u32 queue_id, u32 flags);
1512 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1513 int (*ndo_tunnel_ctl)(struct net_device *dev,
1514 struct ip_tunnel_parm *p, int cmd);
1515 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1516 };
1517
1518 /**
1519 * enum net_device_priv_flags - &struct net_device priv_flags
1520 *
1521 * These are the &struct net_device, they are only set internally
1522 * by drivers and used in the kernel. These flags are invisible to
1523 * userspace; this means that the order of these flags can change
1524 * during any kernel release.
1525 *
1526 * You should have a pretty good reason to be extending these flags.
1527 *
1528 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1529 * @IFF_EBRIDGE: Ethernet bridging device
1530 * @IFF_BONDING: bonding master or slave
1531 * @IFF_ISATAP: ISATAP interface (RFC4214)
1532 * @IFF_WAN_HDLC: WAN HDLC device
1533 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1534 * release skb->dst
1535 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1536 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1537 * @IFF_MACVLAN_PORT: device used as macvlan port
1538 * @IFF_BRIDGE_PORT: device used as bridge port
1539 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1540 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1541 * @IFF_UNICAST_FLT: Supports unicast filtering
1542 * @IFF_TEAM_PORT: device used as team port
1543 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1544 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1545 * change when it's running
1546 * @IFF_MACVLAN: Macvlan device
1547 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1548 * underlying stacked devices
1549 * @IFF_L3MDEV_MASTER: device is an L3 master device
1550 * @IFF_NO_QUEUE: device can run without qdisc attached
1551 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1552 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1553 * @IFF_TEAM: device is a team device
1554 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1555 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1556 * entity (i.e. the master device for bridged veth)
1557 * @IFF_MACSEC: device is a MACsec device
1558 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1559 * @IFF_FAILOVER: device is a failover master device
1560 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1561 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1562 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1563 */
1564 enum netdev_priv_flags {
1565 IFF_802_1Q_VLAN = 1<<0,
1566 IFF_EBRIDGE = 1<<1,
1567 IFF_BONDING = 1<<2,
1568 IFF_ISATAP = 1<<3,
1569 IFF_WAN_HDLC = 1<<4,
1570 IFF_XMIT_DST_RELEASE = 1<<5,
1571 IFF_DONT_BRIDGE = 1<<6,
1572 IFF_DISABLE_NETPOLL = 1<<7,
1573 IFF_MACVLAN_PORT = 1<<8,
1574 IFF_BRIDGE_PORT = 1<<9,
1575 IFF_OVS_DATAPATH = 1<<10,
1576 IFF_TX_SKB_SHARING = 1<<11,
1577 IFF_UNICAST_FLT = 1<<12,
1578 IFF_TEAM_PORT = 1<<13,
1579 IFF_SUPP_NOFCS = 1<<14,
1580 IFF_LIVE_ADDR_CHANGE = 1<<15,
1581 IFF_MACVLAN = 1<<16,
1582 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1583 IFF_L3MDEV_MASTER = 1<<18,
1584 IFF_NO_QUEUE = 1<<19,
1585 IFF_OPENVSWITCH = 1<<20,
1586 IFF_L3MDEV_SLAVE = 1<<21,
1587 IFF_TEAM = 1<<22,
1588 IFF_RXFH_CONFIGURED = 1<<23,
1589 IFF_PHONY_HEADROOM = 1<<24,
1590 IFF_MACSEC = 1<<25,
1591 IFF_NO_RX_HANDLER = 1<<26,
1592 IFF_FAILOVER = 1<<27,
1593 IFF_FAILOVER_SLAVE = 1<<28,
1594 IFF_L3MDEV_RX_HANDLER = 1<<29,
1595 IFF_LIVE_RENAME_OK = 1<<30,
1596 };
1597
1598 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1599 #define IFF_EBRIDGE IFF_EBRIDGE
1600 #define IFF_BONDING IFF_BONDING
1601 #define IFF_ISATAP IFF_ISATAP
1602 #define IFF_WAN_HDLC IFF_WAN_HDLC
1603 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1604 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1605 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1606 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1607 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1608 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1609 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1610 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1611 #define IFF_TEAM_PORT IFF_TEAM_PORT
1612 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1613 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1614 #define IFF_MACVLAN IFF_MACVLAN
1615 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1616 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1617 #define IFF_NO_QUEUE IFF_NO_QUEUE
1618 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1619 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1620 #define IFF_TEAM IFF_TEAM
1621 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1622 #define IFF_MACSEC IFF_MACSEC
1623 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1624 #define IFF_FAILOVER IFF_FAILOVER
1625 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1626 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1627 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1628
1629 /* Specifies the type of the struct net_device::ml_priv pointer */
1630 enum netdev_ml_priv_type {
1631 ML_PRIV_NONE,
1632 ML_PRIV_CAN,
1633 };
1634
1635 /**
1636 * struct net_device - The DEVICE structure.
1637 *
1638 * Actually, this whole structure is a big mistake. It mixes I/O
1639 * data with strictly "high-level" data, and it has to know about
1640 * almost every data structure used in the INET module.
1641 *
1642 * @name: This is the first field of the "visible" part of this structure
1643 * (i.e. as seen by users in the "Space.c" file). It is the name
1644 * of the interface.
1645 *
1646 * @name_node: Name hashlist node
1647 * @ifalias: SNMP alias
1648 * @mem_end: Shared memory end
1649 * @mem_start: Shared memory start
1650 * @base_addr: Device I/O address
1651 * @irq: Device IRQ number
1652 *
1653 * @state: Generic network queuing layer state, see netdev_state_t
1654 * @dev_list: The global list of network devices
1655 * @napi_list: List entry used for polling NAPI devices
1656 * @unreg_list: List entry when we are unregistering the
1657 * device; see the function unregister_netdev
1658 * @close_list: List entry used when we are closing the device
1659 * @ptype_all: Device-specific packet handlers for all protocols
1660 * @ptype_specific: Device-specific, protocol-specific packet handlers
1661 *
1662 * @adj_list: Directly linked devices, like slaves for bonding
1663 * @features: Currently active device features
1664 * @hw_features: User-changeable features
1665 *
1666 * @wanted_features: User-requested features
1667 * @vlan_features: Mask of features inheritable by VLAN devices
1668 *
1669 * @hw_enc_features: Mask of features inherited by encapsulating devices
1670 * This field indicates what encapsulation
1671 * offloads the hardware is capable of doing,
1672 * and drivers will need to set them appropriately.
1673 *
1674 * @mpls_features: Mask of features inheritable by MPLS
1675 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1676 *
1677 * @ifindex: interface index
1678 * @group: The group the device belongs to
1679 *
1680 * @stats: Statistics struct, which was left as a legacy, use
1681 * rtnl_link_stats64 instead
1682 *
1683 * @rx_dropped: Dropped packets by core network,
1684 * do not use this in drivers
1685 * @tx_dropped: Dropped packets by core network,
1686 * do not use this in drivers
1687 * @rx_nohandler: nohandler dropped packets by core network on
1688 * inactive devices, do not use this in drivers
1689 * @carrier_up_count: Number of times the carrier has been up
1690 * @carrier_down_count: Number of times the carrier has been down
1691 *
1692 * @wireless_handlers: List of functions to handle Wireless Extensions,
1693 * instead of ioctl,
1694 * see <net/iw_handler.h> for details.
1695 * @wireless_data: Instance data managed by the core of wireless extensions
1696 *
1697 * @netdev_ops: Includes several pointers to callbacks,
1698 * if one wants to override the ndo_*() functions
1699 * @ethtool_ops: Management operations
1700 * @l3mdev_ops: Layer 3 master device operations
1701 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1702 * discovery handling. Necessary for e.g. 6LoWPAN.
1703 * @xfrmdev_ops: Transformation offload operations
1704 * @tlsdev_ops: Transport Layer Security offload operations
1705 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1706 * of Layer 2 headers.
1707 *
1708 * @flags: Interface flags (a la BSD)
1709 * @priv_flags: Like 'flags' but invisible to userspace,
1710 * see if.h for the definitions
1711 * @gflags: Global flags ( kept as legacy )
1712 * @padded: How much padding added by alloc_netdev()
1713 * @operstate: RFC2863 operstate
1714 * @link_mode: Mapping policy to operstate
1715 * @if_port: Selectable AUI, TP, ...
1716 * @dma: DMA channel
1717 * @mtu: Interface MTU value
1718 * @min_mtu: Interface Minimum MTU value
1719 * @max_mtu: Interface Maximum MTU value
1720 * @type: Interface hardware type
1721 * @hard_header_len: Maximum hardware header length.
1722 * @min_header_len: Minimum hardware header length
1723 *
1724 * @needed_headroom: Extra headroom the hardware may need, but not in all
1725 * cases can this be guaranteed
1726 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1727 * cases can this be guaranteed. Some cases also use
1728 * LL_MAX_HEADER instead to allocate the skb
1729 *
1730 * interface address info:
1731 *
1732 * @perm_addr: Permanent hw address
1733 * @addr_assign_type: Hw address assignment type
1734 * @addr_len: Hardware address length
1735 * @upper_level: Maximum depth level of upper devices.
1736 * @lower_level: Maximum depth level of lower devices.
1737 * @neigh_priv_len: Used in neigh_alloc()
1738 * @dev_id: Used to differentiate devices that share
1739 * the same link layer address
1740 * @dev_port: Used to differentiate devices that share
1741 * the same function
1742 * @addr_list_lock: XXX: need comments on this one
1743 * @name_assign_type: network interface name assignment type
1744 * @uc_promisc: Counter that indicates promiscuous mode
1745 * has been enabled due to the need to listen to
1746 * additional unicast addresses in a device that
1747 * does not implement ndo_set_rx_mode()
1748 * @uc: unicast mac addresses
1749 * @mc: multicast mac addresses
1750 * @dev_addrs: list of device hw addresses
1751 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1752 * @promiscuity: Number of times the NIC is told to work in
1753 * promiscuous mode; if it becomes 0 the NIC will
1754 * exit promiscuous mode
1755 * @allmulti: Counter, enables or disables allmulticast mode
1756 *
1757 * @vlan_info: VLAN info
1758 * @dsa_ptr: dsa specific data
1759 * @tipc_ptr: TIPC specific data
1760 * @atalk_ptr: AppleTalk link
1761 * @ip_ptr: IPv4 specific data
1762 * @ip6_ptr: IPv6 specific data
1763 * @ax25_ptr: AX.25 specific data
1764 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1765 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1766 * device struct
1767 * @mpls_ptr: mpls_dev struct pointer
1768 *
1769 * @dev_addr: Hw address (before bcast,
1770 * because most packets are unicast)
1771 *
1772 * @_rx: Array of RX queues
1773 * @num_rx_queues: Number of RX queues
1774 * allocated at register_netdev() time
1775 * @real_num_rx_queues: Number of RX queues currently active in device
1776 * @xdp_prog: XDP sockets filter program pointer
1777 * @gro_flush_timeout: timeout for GRO layer in NAPI
1778 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1779 * allow to avoid NIC hard IRQ, on busy queues.
1780 *
1781 * @rx_handler: handler for received packets
1782 * @rx_handler_data: XXX: need comments on this one
1783 * @miniq_ingress: ingress/clsact qdisc specific data for
1784 * ingress processing
1785 * @ingress_queue: XXX: need comments on this one
1786 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1787 * @broadcast: hw bcast address
1788 *
1789 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1790 * indexed by RX queue number. Assigned by driver.
1791 * This must only be set if the ndo_rx_flow_steer
1792 * operation is defined
1793 * @index_hlist: Device index hash chain
1794 *
1795 * @_tx: Array of TX queues
1796 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1797 * @real_num_tx_queues: Number of TX queues currently active in device
1798 * @qdisc: Root qdisc from userspace point of view
1799 * @tx_queue_len: Max frames per queue allowed
1800 * @tx_global_lock: XXX: need comments on this one
1801 * @xdp_bulkq: XDP device bulk queue
1802 * @xps_cpus_map: all CPUs map for XPS device
1803 * @xps_rxqs_map: all RXQs map for XPS device
1804 *
1805 * @xps_maps: XXX: need comments on this one
1806 * @miniq_egress: clsact qdisc specific data for
1807 * egress processing
1808 * @qdisc_hash: qdisc hash table
1809 * @watchdog_timeo: Represents the timeout that is used by
1810 * the watchdog (see dev_watchdog())
1811 * @watchdog_timer: List of timers
1812 *
1813 * @proto_down_reason: reason a netdev interface is held down
1814 * @pcpu_refcnt: Number of references to this device
1815 * @todo_list: Delayed register/unregister
1816 * @link_watch_list: XXX: need comments on this one
1817 *
1818 * @reg_state: Register/unregister state machine
1819 * @dismantle: Device is going to be freed
1820 * @rtnl_link_state: This enum represents the phases of creating
1821 * a new link
1822 *
1823 * @needs_free_netdev: Should unregister perform free_netdev?
1824 * @priv_destructor: Called from unregister
1825 * @npinfo: XXX: need comments on this one
1826 * @nd_net: Network namespace this network device is inside
1827 *
1828 * @ml_priv: Mid-layer private
1829 * @ml_priv_type: Mid-layer private type
1830 * @lstats: Loopback statistics
1831 * @tstats: Tunnel statistics
1832 * @dstats: Dummy statistics
1833 * @vstats: Virtual ethernet statistics
1834 *
1835 * @garp_port: GARP
1836 * @mrp_port: MRP
1837 *
1838 * @dev: Class/net/name entry
1839 * @sysfs_groups: Space for optional device, statistics and wireless
1840 * sysfs groups
1841 *
1842 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1843 * @rtnl_link_ops: Rtnl_link_ops
1844 *
1845 * @gso_max_size: Maximum size of generic segmentation offload
1846 * @gso_max_segs: Maximum number of segments that can be passed to the
1847 * NIC for GSO
1848 *
1849 * @dcbnl_ops: Data Center Bridging netlink ops
1850 * @num_tc: Number of traffic classes in the net device
1851 * @tc_to_txq: XXX: need comments on this one
1852 * @prio_tc_map: XXX: need comments on this one
1853 *
1854 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1855 *
1856 * @priomap: XXX: need comments on this one
1857 * @phydev: Physical device may attach itself
1858 * for hardware timestamping
1859 * @sfp_bus: attached &struct sfp_bus structure.
1860 *
1861 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1862 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1863 *
1864 * @proto_down: protocol port state information can be sent to the
1865 * switch driver and used to set the phys state of the
1866 * switch port.
1867 *
1868 * @wol_enabled: Wake-on-LAN is enabled
1869 *
1870 * @net_notifier_list: List of per-net netdev notifier block
1871 * that follow this device when it is moved
1872 * to another network namespace.
1873 *
1874 * @macsec_ops: MACsec offloading ops
1875 *
1876 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1877 * offload capabilities of the device
1878 * @udp_tunnel_nic: UDP tunnel offload state
1879 * @xdp_state: stores info on attached XDP BPF programs
1880 *
1881 * @nested_level: Used as as a parameter of spin_lock_nested() of
1882 * dev->addr_list_lock.
1883 * @unlink_list: As netif_addr_lock() can be called recursively,
1884 * keep a list of interfaces to be deleted.
1885 *
1886 * FIXME: cleanup struct net_device such that network protocol info
1887 * moves out.
1888 */
1889
1890 struct net_device {
1891 char name[IFNAMSIZ];
1892 struct netdev_name_node *name_node;
1893 struct dev_ifalias __rcu *ifalias;
1894 /*
1895 * I/O specific fields
1896 * FIXME: Merge these and struct ifmap into one
1897 */
1898 unsigned long mem_end;
1899 unsigned long mem_start;
1900 unsigned long base_addr;
1901 int irq;
1902
1903 /*
1904 * Some hardware also needs these fields (state,dev_list,
1905 * napi_list,unreg_list,close_list) but they are not
1906 * part of the usual set specified in Space.c.
1907 */
1908
1909 unsigned long state;
1910
1911 struct list_head dev_list;
1912 struct list_head napi_list;
1913 struct list_head unreg_list;
1914 struct list_head close_list;
1915 struct list_head ptype_all;
1916 struct list_head ptype_specific;
1917
1918 struct {
1919 struct list_head upper;
1920 struct list_head lower;
1921 } adj_list;
1922
1923 netdev_features_t features;
1924 netdev_features_t hw_features;
1925 netdev_features_t wanted_features;
1926 netdev_features_t vlan_features;
1927 netdev_features_t hw_enc_features;
1928 netdev_features_t mpls_features;
1929 netdev_features_t gso_partial_features;
1930
1931 int ifindex;
1932 int group;
1933
1934 struct net_device_stats stats;
1935
1936 atomic_long_t rx_dropped;
1937 atomic_long_t tx_dropped;
1938 atomic_long_t rx_nohandler;
1939
1940 /* Stats to monitor link on/off, flapping */
1941 atomic_t carrier_up_count;
1942 atomic_t carrier_down_count;
1943
1944 #ifdef CONFIG_WIRELESS_EXT
1945 const struct iw_handler_def *wireless_handlers;
1946 struct iw_public_data *wireless_data;
1947 #endif
1948 const struct net_device_ops *netdev_ops;
1949 const struct ethtool_ops *ethtool_ops;
1950 #ifdef CONFIG_NET_L3_MASTER_DEV
1951 const struct l3mdev_ops *l3mdev_ops;
1952 #endif
1953 #if IS_ENABLED(CONFIG_IPV6)
1954 const struct ndisc_ops *ndisc_ops;
1955 #endif
1956
1957 #ifdef CONFIG_XFRM_OFFLOAD
1958 const struct xfrmdev_ops *xfrmdev_ops;
1959 #endif
1960
1961 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1962 const struct tlsdev_ops *tlsdev_ops;
1963 #endif
1964
1965 const struct header_ops *header_ops;
1966
1967 unsigned int flags;
1968 unsigned int priv_flags;
1969
1970 unsigned short gflags;
1971 unsigned short padded;
1972
1973 unsigned char operstate;
1974 unsigned char link_mode;
1975
1976 unsigned char if_port;
1977 unsigned char dma;
1978
1979 /* Note : dev->mtu is often read without holding a lock.
1980 * Writers usually hold RTNL.
1981 * It is recommended to use READ_ONCE() to annotate the reads,
1982 * and to use WRITE_ONCE() to annotate the writes.
1983 */
1984 unsigned int mtu;
1985 unsigned int min_mtu;
1986 unsigned int max_mtu;
1987 unsigned short type;
1988 unsigned short hard_header_len;
1989 unsigned char min_header_len;
1990 unsigned char name_assign_type;
1991
1992 unsigned short needed_headroom;
1993 unsigned short needed_tailroom;
1994
1995 /* Interface address info. */
1996 unsigned char perm_addr[MAX_ADDR_LEN];
1997 unsigned char addr_assign_type;
1998 unsigned char addr_len;
1999 unsigned char upper_level;
2000 unsigned char lower_level;
2001
2002 unsigned short neigh_priv_len;
2003 unsigned short dev_id;
2004 unsigned short dev_port;
2005 spinlock_t addr_list_lock;
2006
2007 struct netdev_hw_addr_list uc;
2008 struct netdev_hw_addr_list mc;
2009 struct netdev_hw_addr_list dev_addrs;
2010
2011 #ifdef CONFIG_SYSFS
2012 struct kset *queues_kset;
2013 #endif
2014 #ifdef CONFIG_LOCKDEP
2015 struct list_head unlink_list;
2016 #endif
2017 unsigned int promiscuity;
2018 unsigned int allmulti;
2019 bool uc_promisc;
2020 #ifdef CONFIG_LOCKDEP
2021 unsigned char nested_level;
2022 #endif
2023
2024
2025 /* Protocol-specific pointers */
2026
2027 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2028 struct vlan_info __rcu *vlan_info;
2029 #endif
2030 #if IS_ENABLED(CONFIG_NET_DSA)
2031 struct dsa_port *dsa_ptr;
2032 #endif
2033 #if IS_ENABLED(CONFIG_TIPC)
2034 struct tipc_bearer __rcu *tipc_ptr;
2035 #endif
2036 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2037 void *atalk_ptr;
2038 #endif
2039 struct in_device __rcu *ip_ptr;
2040 struct inet6_dev __rcu *ip6_ptr;
2041 #if IS_ENABLED(CONFIG_NEWIP)
2042 struct ninet_dev __rcu *nip_ptr; /* NIP */
2043 #endif
2044 #if IS_ENABLED(CONFIG_AX25)
2045 void *ax25_ptr;
2046 #endif
2047 struct wireless_dev *ieee80211_ptr;
2048 struct wpan_dev *ieee802154_ptr;
2049 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2050 struct mpls_dev __rcu *mpls_ptr;
2051 #endif
2052
2053 /*
2054 * Cache lines mostly used on receive path (including eth_type_trans())
2055 */
2056 /* Interface address info used in eth_type_trans() */
2057 unsigned char *dev_addr;
2058
2059 struct netdev_rx_queue *_rx;
2060 unsigned int num_rx_queues;
2061 unsigned int real_num_rx_queues;
2062
2063 struct bpf_prog __rcu *xdp_prog;
2064 unsigned long gro_flush_timeout;
2065 u32 napi_defer_hard_irqs;
2066 rx_handler_func_t __rcu *rx_handler;
2067 void __rcu *rx_handler_data;
2068
2069 #ifdef CONFIG_NET_CLS_ACT
2070 struct mini_Qdisc __rcu *miniq_ingress;
2071 #endif
2072 struct netdev_queue __rcu *ingress_queue;
2073 #ifdef CONFIG_NETFILTER_INGRESS
2074 struct nf_hook_entries __rcu *nf_hooks_ingress;
2075 #endif
2076
2077 unsigned char broadcast[MAX_ADDR_LEN];
2078 #ifdef CONFIG_RFS_ACCEL
2079 struct cpu_rmap *rx_cpu_rmap;
2080 #endif
2081 struct hlist_node index_hlist;
2082
2083 /*
2084 * Cache lines mostly used on transmit path
2085 */
2086 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2087 unsigned int num_tx_queues;
2088 unsigned int real_num_tx_queues;
2089 struct Qdisc __rcu *qdisc;
2090 unsigned int tx_queue_len;
2091 spinlock_t tx_global_lock;
2092
2093 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2094
2095 #ifdef CONFIG_XPS
2096 struct xps_dev_maps __rcu *xps_cpus_map;
2097 struct xps_dev_maps __rcu *xps_rxqs_map;
2098 #endif
2099 #ifdef CONFIG_NET_CLS_ACT
2100 struct mini_Qdisc __rcu *miniq_egress;
2101 #endif
2102
2103 #ifdef CONFIG_NET_SCHED
2104 DECLARE_HASHTABLE (qdisc_hash, 4);
2105 #endif
2106 /* These may be needed for future network-power-down code. */
2107 struct timer_list watchdog_timer;
2108 int watchdog_timeo;
2109
2110 u32 proto_down_reason;
2111
2112 struct list_head todo_list;
2113 int __percpu *pcpu_refcnt;
2114
2115 struct list_head link_watch_list;
2116
2117 enum { NETREG_UNINITIALIZED=0,
2118 NETREG_REGISTERED, /* completed register_netdevice */
2119 NETREG_UNREGISTERING, /* called unregister_netdevice */
2120 NETREG_UNREGISTERED, /* completed unregister todo */
2121 NETREG_RELEASED, /* called free_netdev */
2122 NETREG_DUMMY, /* dummy device for NAPI poll */
2123 } reg_state:8;
2124
2125 bool dismantle;
2126
2127 enum {
2128 RTNL_LINK_INITIALIZED,
2129 RTNL_LINK_INITIALIZING,
2130 } rtnl_link_state:16;
2131
2132 bool needs_free_netdev;
2133 void (*priv_destructor)(struct net_device *dev);
2134
2135 #ifdef CONFIG_NETPOLL
2136 struct netpoll_info __rcu *npinfo;
2137 #endif
2138
2139 possible_net_t nd_net;
2140
2141 /* mid-layer private */
2142 void *ml_priv;
2143 enum netdev_ml_priv_type ml_priv_type;
2144
2145 union {
2146 struct pcpu_lstats __percpu *lstats;
2147 struct pcpu_sw_netstats __percpu *tstats;
2148 struct pcpu_dstats __percpu *dstats;
2149 };
2150
2151 #if IS_ENABLED(CONFIG_GARP)
2152 struct garp_port __rcu *garp_port;
2153 #endif
2154 #if IS_ENABLED(CONFIG_MRP)
2155 struct mrp_port __rcu *mrp_port;
2156 #endif
2157
2158 struct device dev;
2159 const struct attribute_group *sysfs_groups[4];
2160 const struct attribute_group *sysfs_rx_queue_group;
2161
2162 const struct rtnl_link_ops *rtnl_link_ops;
2163
2164 /* for setting kernel sock attribute on TCP connection setup */
2165 #define GSO_MAX_SIZE 65536
2166 unsigned int gso_max_size;
2167 #define GSO_MAX_SEGS 65535
2168 u16 gso_max_segs;
2169
2170 #ifdef CONFIG_DCB
2171 const struct dcbnl_rtnl_ops *dcbnl_ops;
2172 #endif
2173 s16 num_tc;
2174 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2175 u8 prio_tc_map[TC_BITMASK + 1];
2176
2177 #if IS_ENABLED(CONFIG_FCOE)
2178 unsigned int fcoe_ddp_xid;
2179 #endif
2180 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2181 struct netprio_map __rcu *priomap;
2182 #endif
2183 struct phy_device *phydev;
2184 struct sfp_bus *sfp_bus;
2185 struct lock_class_key *qdisc_tx_busylock;
2186 struct lock_class_key *qdisc_running_key;
2187 bool proto_down;
2188 unsigned wol_enabled:1;
2189
2190 struct list_head net_notifier_list;
2191
2192 #if IS_ENABLED(CONFIG_MACSEC)
2193 /* MACsec management functions */
2194 const struct macsec_ops *macsec_ops;
2195 #endif
2196 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2197 struct udp_tunnel_nic *udp_tunnel_nic;
2198
2199 /* protected by rtnl_lock */
2200 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2201 };
2202 #define to_net_dev(d) container_of(d, struct net_device, dev)
2203
netif_elide_gro(const struct net_device * dev)2204 static inline bool netif_elide_gro(const struct net_device *dev)
2205 {
2206 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2207 return true;
2208 return false;
2209 }
2210
2211 #define NETDEV_ALIGN 32
2212
2213 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)2214 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2215 {
2216 return dev->prio_tc_map[prio & TC_BITMASK];
2217 }
2218
2219 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)2220 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2221 {
2222 if (tc >= dev->num_tc)
2223 return -EINVAL;
2224
2225 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2226 return 0;
2227 }
2228
2229 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2230 void netdev_reset_tc(struct net_device *dev);
2231 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2232 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2233
2234 static inline
netdev_get_num_tc(struct net_device * dev)2235 int netdev_get_num_tc(struct net_device *dev)
2236 {
2237 return dev->num_tc;
2238 }
2239
net_prefetch(void * p)2240 static inline void net_prefetch(void *p)
2241 {
2242 prefetch(p);
2243 #if L1_CACHE_BYTES < 128
2244 prefetch((u8 *)p + L1_CACHE_BYTES);
2245 #endif
2246 }
2247
net_prefetchw(void * p)2248 static inline void net_prefetchw(void *p)
2249 {
2250 prefetchw(p);
2251 #if L1_CACHE_BYTES < 128
2252 prefetchw((u8 *)p + L1_CACHE_BYTES);
2253 #endif
2254 }
2255
2256 void netdev_unbind_sb_channel(struct net_device *dev,
2257 struct net_device *sb_dev);
2258 int netdev_bind_sb_channel_queue(struct net_device *dev,
2259 struct net_device *sb_dev,
2260 u8 tc, u16 count, u16 offset);
2261 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
netdev_get_sb_channel(struct net_device * dev)2262 static inline int netdev_get_sb_channel(struct net_device *dev)
2263 {
2264 return max_t(int, -dev->num_tc, 0);
2265 }
2266
2267 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)2268 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2269 unsigned int index)
2270 {
2271 return &dev->_tx[index];
2272 }
2273
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)2274 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2275 const struct sk_buff *skb)
2276 {
2277 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2278 }
2279
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)2280 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2281 void (*f)(struct net_device *,
2282 struct netdev_queue *,
2283 void *),
2284 void *arg)
2285 {
2286 unsigned int i;
2287
2288 for (i = 0; i < dev->num_tx_queues; i++)
2289 f(dev, &dev->_tx[i], arg);
2290 }
2291
2292 #define netdev_lockdep_set_classes(dev) \
2293 { \
2294 static struct lock_class_key qdisc_tx_busylock_key; \
2295 static struct lock_class_key qdisc_running_key; \
2296 static struct lock_class_key qdisc_xmit_lock_key; \
2297 static struct lock_class_key dev_addr_list_lock_key; \
2298 unsigned int i; \
2299 \
2300 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2301 (dev)->qdisc_running_key = &qdisc_running_key; \
2302 lockdep_set_class(&(dev)->addr_list_lock, \
2303 &dev_addr_list_lock_key); \
2304 for (i = 0; i < (dev)->num_tx_queues; i++) \
2305 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2306 &qdisc_xmit_lock_key); \
2307 }
2308
2309 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2310 struct net_device *sb_dev);
2311 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2312 struct sk_buff *skb,
2313 struct net_device *sb_dev);
2314
2315 /* returns the headroom that the master device needs to take in account
2316 * when forwarding to this dev
2317 */
netdev_get_fwd_headroom(struct net_device * dev)2318 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2319 {
2320 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2321 }
2322
netdev_set_rx_headroom(struct net_device * dev,int new_hr)2323 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2324 {
2325 if (dev->netdev_ops->ndo_set_rx_headroom)
2326 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2327 }
2328
2329 /* set the device rx headroom to the dev's default */
netdev_reset_rx_headroom(struct net_device * dev)2330 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2331 {
2332 netdev_set_rx_headroom(dev, -1);
2333 }
2334
netdev_get_ml_priv(struct net_device * dev,enum netdev_ml_priv_type type)2335 static inline void *netdev_get_ml_priv(struct net_device *dev,
2336 enum netdev_ml_priv_type type)
2337 {
2338 if (dev->ml_priv_type != type)
2339 return NULL;
2340
2341 return dev->ml_priv;
2342 }
2343
netdev_set_ml_priv(struct net_device * dev,void * ml_priv,enum netdev_ml_priv_type type)2344 static inline void netdev_set_ml_priv(struct net_device *dev,
2345 void *ml_priv,
2346 enum netdev_ml_priv_type type)
2347 {
2348 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2349 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2350 dev->ml_priv_type, type);
2351 WARN(!dev->ml_priv_type && dev->ml_priv,
2352 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2353
2354 dev->ml_priv = ml_priv;
2355 dev->ml_priv_type = type;
2356 }
2357
2358 /*
2359 * Net namespace inlines
2360 */
2361 static inline
dev_net(const struct net_device * dev)2362 struct net *dev_net(const struct net_device *dev)
2363 {
2364 return read_pnet(&dev->nd_net);
2365 }
2366
2367 static inline
dev_net_rcu(const struct net_device * dev)2368 struct net *dev_net_rcu(const struct net_device *dev)
2369 {
2370 return read_pnet_rcu(&dev->nd_net);
2371 }
2372
2373 static inline
dev_net_set(struct net_device * dev,struct net * net)2374 void dev_net_set(struct net_device *dev, struct net *net)
2375 {
2376 write_pnet(&dev->nd_net, net);
2377 }
2378
2379 /**
2380 * netdev_priv - access network device private data
2381 * @dev: network device
2382 *
2383 * Get network device private data
2384 */
netdev_priv(const struct net_device * dev)2385 static inline void *netdev_priv(const struct net_device *dev)
2386 {
2387 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2388 }
2389
2390 /* Set the sysfs physical device reference for the network logical device
2391 * if set prior to registration will cause a symlink during initialization.
2392 */
2393 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2394
2395 /* Set the sysfs device type for the network logical device to allow
2396 * fine-grained identification of different network device types. For
2397 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2398 */
2399 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2400
2401 /* Default NAPI poll() weight
2402 * Device drivers are strongly advised to not use bigger value
2403 */
2404 #define NAPI_POLL_WEIGHT 64
2405
2406 /**
2407 * netif_napi_add - initialize a NAPI context
2408 * @dev: network device
2409 * @napi: NAPI context
2410 * @poll: polling function
2411 * @weight: default weight
2412 *
2413 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2414 * *any* of the other NAPI-related functions.
2415 */
2416 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2417 int (*poll)(struct napi_struct *, int), int weight);
2418
2419 /**
2420 * netif_tx_napi_add - initialize a NAPI context
2421 * @dev: network device
2422 * @napi: NAPI context
2423 * @poll: polling function
2424 * @weight: default weight
2425 *
2426 * This variant of netif_napi_add() should be used from drivers using NAPI
2427 * to exclusively poll a TX queue.
2428 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2429 */
netif_tx_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2430 static inline void netif_tx_napi_add(struct net_device *dev,
2431 struct napi_struct *napi,
2432 int (*poll)(struct napi_struct *, int),
2433 int weight)
2434 {
2435 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2436 netif_napi_add(dev, napi, poll, weight);
2437 }
2438
2439 /**
2440 * __netif_napi_del - remove a NAPI context
2441 * @napi: NAPI context
2442 *
2443 * Warning: caller must observe RCU grace period before freeing memory
2444 * containing @napi. Drivers might want to call this helper to combine
2445 * all the needed RCU grace periods into a single one.
2446 */
2447 void __netif_napi_del(struct napi_struct *napi);
2448
2449 /**
2450 * netif_napi_del - remove a NAPI context
2451 * @napi: NAPI context
2452 *
2453 * netif_napi_del() removes a NAPI context from the network device NAPI list
2454 */
netif_napi_del(struct napi_struct * napi)2455 static inline void netif_napi_del(struct napi_struct *napi)
2456 {
2457 __netif_napi_del(napi);
2458 synchronize_net();
2459 }
2460
2461 struct napi_gro_cb {
2462 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2463 void *frag0;
2464
2465 /* Length of frag0. */
2466 unsigned int frag0_len;
2467
2468 /* This indicates where we are processing relative to skb->data. */
2469 int data_offset;
2470
2471 /* This is non-zero if the packet cannot be merged with the new skb. */
2472 u16 flush;
2473
2474 /* Save the IP ID here and check when we get to the transport layer */
2475 u16 flush_id;
2476
2477 /* Number of segments aggregated. */
2478 u16 count;
2479
2480 /* Start offset for remote checksum offload */
2481 u16 gro_remcsum_start;
2482
2483 /* jiffies when first packet was created/queued */
2484 unsigned long age;
2485
2486 /* Used in ipv6_gro_receive() and foo-over-udp */
2487 u16 proto;
2488
2489 /* This is non-zero if the packet may be of the same flow. */
2490 u8 same_flow:1;
2491
2492 /* Used in tunnel GRO receive */
2493 u8 encap_mark:1;
2494
2495 /* GRO checksum is valid */
2496 u8 csum_valid:1;
2497
2498 /* Number of checksums via CHECKSUM_UNNECESSARY */
2499 u8 csum_cnt:3;
2500
2501 /* Free the skb? */
2502 u8 free:2;
2503 #define NAPI_GRO_FREE 1
2504 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2505
2506 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2507 u8 is_ipv6:1;
2508
2509 /* Used in GRE, set in fou/gue_gro_receive */
2510 u8 is_fou:1;
2511
2512 /* Used to determine if flush_id can be ignored */
2513 u8 is_atomic:1;
2514
2515 /* Number of gro_receive callbacks this packet already went through */
2516 u8 recursion_counter:4;
2517
2518 /* GRO is done by frag_list pointer chaining. */
2519 u8 is_flist:1;
2520
2521 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2522 __wsum csum;
2523
2524 /* used in skb_gro_receive() slow path */
2525 struct sk_buff *last;
2526 };
2527
2528 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2529
2530 #define GRO_RECURSION_LIMIT 15
gro_recursion_inc_test(struct sk_buff * skb)2531 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2532 {
2533 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2534 }
2535
2536 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)2537 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2538 struct list_head *head,
2539 struct sk_buff *skb)
2540 {
2541 if (unlikely(gro_recursion_inc_test(skb))) {
2542 NAPI_GRO_CB(skb)->flush |= 1;
2543 return NULL;
2544 }
2545
2546 return cb(head, skb);
2547 }
2548
2549 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2550 struct sk_buff *);
call_gro_receive_sk(gro_receive_sk_t cb,struct sock * sk,struct list_head * head,struct sk_buff * skb)2551 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2552 struct sock *sk,
2553 struct list_head *head,
2554 struct sk_buff *skb)
2555 {
2556 if (unlikely(gro_recursion_inc_test(skb))) {
2557 NAPI_GRO_CB(skb)->flush |= 1;
2558 return NULL;
2559 }
2560
2561 return cb(sk, head, skb);
2562 }
2563
2564 struct packet_type {
2565 __be16 type; /* This is really htons(ether_type). */
2566 bool ignore_outgoing;
2567 struct net_device *dev; /* NULL is wildcarded here */
2568 int (*func) (struct sk_buff *,
2569 struct net_device *,
2570 struct packet_type *,
2571 struct net_device *);
2572 void (*list_func) (struct list_head *,
2573 struct packet_type *,
2574 struct net_device *);
2575 bool (*id_match)(struct packet_type *ptype,
2576 struct sock *sk);
2577 struct net *af_packet_net;
2578 void *af_packet_priv;
2579 struct list_head list;
2580 };
2581
2582 struct offload_callbacks {
2583 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2584 netdev_features_t features);
2585 struct sk_buff *(*gro_receive)(struct list_head *head,
2586 struct sk_buff *skb);
2587 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2588 };
2589
2590 struct packet_offload {
2591 __be16 type; /* This is really htons(ether_type). */
2592 u16 priority;
2593 struct offload_callbacks callbacks;
2594 struct list_head list;
2595 };
2596
2597 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2598 struct pcpu_sw_netstats {
2599 u64 rx_packets;
2600 u64 rx_bytes;
2601 u64 tx_packets;
2602 u64 tx_bytes;
2603 struct u64_stats_sync syncp;
2604 } __aligned(4 * sizeof(u64));
2605
2606 struct pcpu_lstats {
2607 u64_stats_t packets;
2608 u64_stats_t bytes;
2609 struct u64_stats_sync syncp;
2610 } __aligned(2 * sizeof(u64));
2611
2612 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2613
dev_sw_netstats_rx_add(struct net_device * dev,unsigned int len)2614 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2615 {
2616 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2617
2618 u64_stats_update_begin(&tstats->syncp);
2619 tstats->rx_bytes += len;
2620 tstats->rx_packets++;
2621 u64_stats_update_end(&tstats->syncp);
2622 }
2623
dev_lstats_add(struct net_device * dev,unsigned int len)2624 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2625 {
2626 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2627
2628 u64_stats_update_begin(&lstats->syncp);
2629 u64_stats_add(&lstats->bytes, len);
2630 u64_stats_inc(&lstats->packets);
2631 u64_stats_update_end(&lstats->syncp);
2632 }
2633
2634 #define __netdev_alloc_pcpu_stats(type, gfp) \
2635 ({ \
2636 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2637 if (pcpu_stats) { \
2638 int __cpu; \
2639 for_each_possible_cpu(__cpu) { \
2640 typeof(type) *stat; \
2641 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2642 u64_stats_init(&stat->syncp); \
2643 } \
2644 } \
2645 pcpu_stats; \
2646 })
2647
2648 #define netdev_alloc_pcpu_stats(type) \
2649 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2650
2651 enum netdev_lag_tx_type {
2652 NETDEV_LAG_TX_TYPE_UNKNOWN,
2653 NETDEV_LAG_TX_TYPE_RANDOM,
2654 NETDEV_LAG_TX_TYPE_BROADCAST,
2655 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2656 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2657 NETDEV_LAG_TX_TYPE_HASH,
2658 };
2659
2660 enum netdev_lag_hash {
2661 NETDEV_LAG_HASH_NONE,
2662 NETDEV_LAG_HASH_L2,
2663 NETDEV_LAG_HASH_L34,
2664 NETDEV_LAG_HASH_L23,
2665 NETDEV_LAG_HASH_E23,
2666 NETDEV_LAG_HASH_E34,
2667 NETDEV_LAG_HASH_UNKNOWN,
2668 };
2669
2670 struct netdev_lag_upper_info {
2671 enum netdev_lag_tx_type tx_type;
2672 enum netdev_lag_hash hash_type;
2673 };
2674
2675 struct netdev_lag_lower_state_info {
2676 u8 link_up : 1,
2677 tx_enabled : 1;
2678 };
2679
2680 #include <linux/notifier.h>
2681
2682 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2683 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2684 * adding new types.
2685 */
2686 enum netdev_cmd {
2687 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2688 NETDEV_DOWN,
2689 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2690 detected a hardware crash and restarted
2691 - we can use this eg to kick tcp sessions
2692 once done */
2693 NETDEV_CHANGE, /* Notify device state change */
2694 NETDEV_REGISTER,
2695 NETDEV_UNREGISTER,
2696 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2697 NETDEV_CHANGEADDR, /* notify after the address change */
2698 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2699 NETDEV_GOING_DOWN,
2700 NETDEV_CHANGENAME,
2701 NETDEV_FEAT_CHANGE,
2702 NETDEV_BONDING_FAILOVER,
2703 NETDEV_PRE_UP,
2704 NETDEV_PRE_TYPE_CHANGE,
2705 NETDEV_POST_TYPE_CHANGE,
2706 NETDEV_POST_INIT,
2707 NETDEV_RELEASE,
2708 NETDEV_NOTIFY_PEERS,
2709 NETDEV_JOIN,
2710 NETDEV_CHANGEUPPER,
2711 NETDEV_RESEND_IGMP,
2712 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2713 NETDEV_CHANGEINFODATA,
2714 NETDEV_BONDING_INFO,
2715 NETDEV_PRECHANGEUPPER,
2716 NETDEV_CHANGELOWERSTATE,
2717 NETDEV_UDP_TUNNEL_PUSH_INFO,
2718 NETDEV_UDP_TUNNEL_DROP_INFO,
2719 NETDEV_CHANGE_TX_QUEUE_LEN,
2720 NETDEV_CVLAN_FILTER_PUSH_INFO,
2721 NETDEV_CVLAN_FILTER_DROP_INFO,
2722 NETDEV_SVLAN_FILTER_PUSH_INFO,
2723 NETDEV_SVLAN_FILTER_DROP_INFO,
2724 };
2725 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2726
2727 int register_netdevice_notifier(struct notifier_block *nb);
2728 int unregister_netdevice_notifier(struct notifier_block *nb);
2729 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2730 int unregister_netdevice_notifier_net(struct net *net,
2731 struct notifier_block *nb);
2732 int register_netdevice_notifier_dev_net(struct net_device *dev,
2733 struct notifier_block *nb,
2734 struct netdev_net_notifier *nn);
2735 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2736 struct notifier_block *nb,
2737 struct netdev_net_notifier *nn);
2738
2739 struct netdev_notifier_info {
2740 struct net_device *dev;
2741 struct netlink_ext_ack *extack;
2742 };
2743
2744 struct netdev_notifier_info_ext {
2745 struct netdev_notifier_info info; /* must be first */
2746 union {
2747 u32 mtu;
2748 } ext;
2749 };
2750
2751 struct netdev_notifier_change_info {
2752 struct netdev_notifier_info info; /* must be first */
2753 unsigned int flags_changed;
2754 };
2755
2756 struct netdev_notifier_changeupper_info {
2757 struct netdev_notifier_info info; /* must be first */
2758 struct net_device *upper_dev; /* new upper dev */
2759 bool master; /* is upper dev master */
2760 bool linking; /* is the notification for link or unlink */
2761 void *upper_info; /* upper dev info */
2762 };
2763
2764 struct netdev_notifier_changelowerstate_info {
2765 struct netdev_notifier_info info; /* must be first */
2766 void *lower_state_info; /* is lower dev state */
2767 };
2768
2769 struct netdev_notifier_pre_changeaddr_info {
2770 struct netdev_notifier_info info; /* must be first */
2771 const unsigned char *dev_addr;
2772 };
2773
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)2774 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2775 struct net_device *dev)
2776 {
2777 info->dev = dev;
2778 info->extack = NULL;
2779 }
2780
2781 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)2782 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2783 {
2784 return info->dev;
2785 }
2786
2787 static inline struct netlink_ext_ack *
netdev_notifier_info_to_extack(const struct netdev_notifier_info * info)2788 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2789 {
2790 return info->extack;
2791 }
2792
2793 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2794
2795
2796 extern rwlock_t dev_base_lock; /* Device list lock */
2797
2798 #define for_each_netdev(net, d) \
2799 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2800 #define for_each_netdev_reverse(net, d) \
2801 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2802 #define for_each_netdev_rcu(net, d) \
2803 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2804 #define for_each_netdev_safe(net, d, n) \
2805 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2806 #define for_each_netdev_continue(net, d) \
2807 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2808 #define for_each_netdev_continue_reverse(net, d) \
2809 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2810 dev_list)
2811 #define for_each_netdev_continue_rcu(net, d) \
2812 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2813 #define for_each_netdev_in_bond_rcu(bond, slave) \
2814 for_each_netdev_rcu(&init_net, slave) \
2815 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2816 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2817
next_net_device(struct net_device * dev)2818 static inline struct net_device *next_net_device(struct net_device *dev)
2819 {
2820 struct list_head *lh;
2821 struct net *net;
2822
2823 net = dev_net(dev);
2824 lh = dev->dev_list.next;
2825 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2826 }
2827
next_net_device_rcu(struct net_device * dev)2828 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2829 {
2830 struct list_head *lh;
2831 struct net *net;
2832
2833 net = dev_net(dev);
2834 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2835 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2836 }
2837
first_net_device(struct net * net)2838 static inline struct net_device *first_net_device(struct net *net)
2839 {
2840 return list_empty(&net->dev_base_head) ? NULL :
2841 net_device_entry(net->dev_base_head.next);
2842 }
2843
first_net_device_rcu(struct net * net)2844 static inline struct net_device *first_net_device_rcu(struct net *net)
2845 {
2846 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2847
2848 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2849 }
2850
2851 int netdev_boot_setup_check(struct net_device *dev);
2852 unsigned long netdev_boot_base(const char *prefix, int unit);
2853 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2854 const char *hwaddr);
2855 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2856 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2857 void dev_add_pack(struct packet_type *pt);
2858 void dev_remove_pack(struct packet_type *pt);
2859 void __dev_remove_pack(struct packet_type *pt);
2860 void dev_add_offload(struct packet_offload *po);
2861 void dev_remove_offload(struct packet_offload *po);
2862
2863 int dev_get_iflink(const struct net_device *dev);
2864 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2865 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2866 unsigned short mask);
2867 struct net_device *dev_get_by_name(struct net *net, const char *name);
2868 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2869 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2870 int dev_alloc_name(struct net_device *dev, const char *name);
2871 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2872 void dev_close(struct net_device *dev);
2873 void dev_close_many(struct list_head *head, bool unlink);
2874 void dev_disable_lro(struct net_device *dev);
2875 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2876 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2877 struct net_device *sb_dev);
2878 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2879 struct net_device *sb_dev);
2880
2881 int dev_queue_xmit(struct sk_buff *skb);
2882 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2883 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2884
dev_direct_xmit(struct sk_buff * skb,u16 queue_id)2885 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2886 {
2887 int ret;
2888
2889 ret = __dev_direct_xmit(skb, queue_id);
2890 if (!dev_xmit_complete(ret))
2891 kfree_skb(skb);
2892 return ret;
2893 }
2894
2895 int register_netdevice(struct net_device *dev);
2896 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2897 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)2898 static inline void unregister_netdevice(struct net_device *dev)
2899 {
2900 unregister_netdevice_queue(dev, NULL);
2901 }
2902
2903 int netdev_refcnt_read(const struct net_device *dev);
2904 void free_netdev(struct net_device *dev);
2905 void netdev_freemem(struct net_device *dev);
2906 int init_dummy_netdev(struct net_device *dev);
2907
2908 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2909 struct sk_buff *skb,
2910 bool all_slaves);
2911 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2912 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2913 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2914 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2915 int netdev_get_name(struct net *net, char *name, int ifindex);
2916 int dev_restart(struct net_device *dev);
2917 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2918 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2919
skb_gro_offset(const struct sk_buff * skb)2920 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2921 {
2922 return NAPI_GRO_CB(skb)->data_offset;
2923 }
2924
skb_gro_len(const struct sk_buff * skb)2925 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2926 {
2927 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2928 }
2929
skb_gro_pull(struct sk_buff * skb,unsigned int len)2930 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2931 {
2932 NAPI_GRO_CB(skb)->data_offset += len;
2933 }
2934
skb_gro_header_fast(struct sk_buff * skb,unsigned int offset)2935 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2936 unsigned int offset)
2937 {
2938 return NAPI_GRO_CB(skb)->frag0 + offset;
2939 }
2940
skb_gro_header_hard(struct sk_buff * skb,unsigned int hlen)2941 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2942 {
2943 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2944 }
2945
skb_gro_frag0_invalidate(struct sk_buff * skb)2946 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2947 {
2948 NAPI_GRO_CB(skb)->frag0 = NULL;
2949 NAPI_GRO_CB(skb)->frag0_len = 0;
2950 }
2951
skb_gro_header_slow(struct sk_buff * skb,unsigned int hlen,unsigned int offset)2952 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2953 unsigned int offset)
2954 {
2955 if (!pskb_may_pull(skb, hlen))
2956 return NULL;
2957
2958 skb_gro_frag0_invalidate(skb);
2959 return skb->data + offset;
2960 }
2961
skb_gro_network_header(struct sk_buff * skb)2962 static inline void *skb_gro_network_header(struct sk_buff *skb)
2963 {
2964 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2965 skb_network_offset(skb);
2966 }
2967
skb_gro_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)2968 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2969 const void *start, unsigned int len)
2970 {
2971 if (NAPI_GRO_CB(skb)->csum_valid)
2972 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2973 csum_partial(start, len, 0));
2974 }
2975
2976 /* GRO checksum functions. These are logical equivalents of the normal
2977 * checksum functions (in skbuff.h) except that they operate on the GRO
2978 * offsets and fields in sk_buff.
2979 */
2980
2981 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2982
skb_at_gro_remcsum_start(struct sk_buff * skb)2983 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2984 {
2985 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2986 }
2987
__skb_gro_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)2988 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2989 bool zero_okay,
2990 __sum16 check)
2991 {
2992 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2993 skb_checksum_start_offset(skb) <
2994 skb_gro_offset(skb)) &&
2995 !skb_at_gro_remcsum_start(skb) &&
2996 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2997 (!zero_okay || check));
2998 }
2999
__skb_gro_checksum_validate_complete(struct sk_buff * skb,__wsum psum)3000 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
3001 __wsum psum)
3002 {
3003 if (NAPI_GRO_CB(skb)->csum_valid &&
3004 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3005 return 0;
3006
3007 NAPI_GRO_CB(skb)->csum = psum;
3008
3009 return __skb_gro_checksum_complete(skb);
3010 }
3011
skb_gro_incr_csum_unnecessary(struct sk_buff * skb)3012 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3013 {
3014 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3015 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3016 NAPI_GRO_CB(skb)->csum_cnt--;
3017 } else {
3018 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3019 * verified a new top level checksum or an encapsulated one
3020 * during GRO. This saves work if we fallback to normal path.
3021 */
3022 __skb_incr_checksum_unnecessary(skb);
3023 }
3024 }
3025
3026 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3027 compute_pseudo) \
3028 ({ \
3029 __sum16 __ret = 0; \
3030 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3031 __ret = __skb_gro_checksum_validate_complete(skb, \
3032 compute_pseudo(skb, proto)); \
3033 if (!__ret) \
3034 skb_gro_incr_csum_unnecessary(skb); \
3035 __ret; \
3036 })
3037
3038 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3039 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3040
3041 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3042 compute_pseudo) \
3043 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3044
3045 #define skb_gro_checksum_simple_validate(skb) \
3046 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3047
__skb_gro_checksum_convert_check(struct sk_buff * skb)3048 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3049 {
3050 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3051 !NAPI_GRO_CB(skb)->csum_valid);
3052 }
3053
__skb_gro_checksum_convert(struct sk_buff * skb,__wsum pseudo)3054 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3055 __wsum pseudo)
3056 {
3057 NAPI_GRO_CB(skb)->csum = ~pseudo;
3058 NAPI_GRO_CB(skb)->csum_valid = 1;
3059 }
3060
3061 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3062 do { \
3063 if (__skb_gro_checksum_convert_check(skb)) \
3064 __skb_gro_checksum_convert(skb, \
3065 compute_pseudo(skb, proto)); \
3066 } while (0)
3067
3068 struct gro_remcsum {
3069 int offset;
3070 __wsum delta;
3071 };
3072
skb_gro_remcsum_init(struct gro_remcsum * grc)3073 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3074 {
3075 grc->offset = 0;
3076 grc->delta = 0;
3077 }
3078
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)3079 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3080 unsigned int off, size_t hdrlen,
3081 int start, int offset,
3082 struct gro_remcsum *grc,
3083 bool nopartial)
3084 {
3085 __wsum delta;
3086 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3087
3088 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3089
3090 if (!nopartial) {
3091 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3092 return ptr;
3093 }
3094
3095 ptr = skb_gro_header_fast(skb, off);
3096 if (skb_gro_header_hard(skb, off + plen)) {
3097 ptr = skb_gro_header_slow(skb, off + plen, off);
3098 if (!ptr)
3099 return NULL;
3100 }
3101
3102 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3103 start, offset);
3104
3105 /* Adjust skb->csum since we changed the packet */
3106 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3107
3108 grc->offset = off + hdrlen + offset;
3109 grc->delta = delta;
3110
3111 return ptr;
3112 }
3113
skb_gro_remcsum_cleanup(struct sk_buff * skb,struct gro_remcsum * grc)3114 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3115 struct gro_remcsum *grc)
3116 {
3117 void *ptr;
3118 size_t plen = grc->offset + sizeof(u16);
3119
3120 if (!grc->delta)
3121 return;
3122
3123 ptr = skb_gro_header_fast(skb, grc->offset);
3124 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3125 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3126 if (!ptr)
3127 return;
3128 }
3129
3130 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3131 }
3132
3133 #ifdef CONFIG_XFRM_OFFLOAD
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)3134 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3135 {
3136 if (PTR_ERR(pp) != -EINPROGRESS)
3137 NAPI_GRO_CB(skb)->flush |= flush;
3138 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)3139 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3140 struct sk_buff *pp,
3141 int flush,
3142 struct gro_remcsum *grc)
3143 {
3144 if (PTR_ERR(pp) != -EINPROGRESS) {
3145 NAPI_GRO_CB(skb)->flush |= flush;
3146 skb_gro_remcsum_cleanup(skb, grc);
3147 skb->remcsum_offload = 0;
3148 }
3149 }
3150 #else
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)3151 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3152 {
3153 NAPI_GRO_CB(skb)->flush |= flush;
3154 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)3155 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3156 struct sk_buff *pp,
3157 int flush,
3158 struct gro_remcsum *grc)
3159 {
3160 NAPI_GRO_CB(skb)->flush |= flush;
3161 skb_gro_remcsum_cleanup(skb, grc);
3162 skb->remcsum_offload = 0;
3163 }
3164 #endif
3165
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)3166 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3167 unsigned short type,
3168 const void *daddr, const void *saddr,
3169 unsigned int len)
3170 {
3171 if (!dev->header_ops || !dev->header_ops->create)
3172 return 0;
3173
3174 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3175 }
3176
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)3177 static inline int dev_parse_header(const struct sk_buff *skb,
3178 unsigned char *haddr)
3179 {
3180 const struct net_device *dev = skb->dev;
3181
3182 if (!dev->header_ops || !dev->header_ops->parse)
3183 return 0;
3184 return dev->header_ops->parse(skb, haddr);
3185 }
3186
dev_parse_header_protocol(const struct sk_buff * skb)3187 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3188 {
3189 const struct net_device *dev = skb->dev;
3190
3191 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3192 return 0;
3193 return dev->header_ops->parse_protocol(skb);
3194 }
3195
3196 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)3197 static inline bool dev_validate_header(const struct net_device *dev,
3198 char *ll_header, int len)
3199 {
3200 if (likely(len >= dev->hard_header_len))
3201 return true;
3202 if (len < dev->min_header_len)
3203 return false;
3204
3205 if (capable(CAP_SYS_RAWIO)) {
3206 memset(ll_header + len, 0, dev->hard_header_len - len);
3207 return true;
3208 }
3209
3210 if (dev->header_ops && dev->header_ops->validate)
3211 return dev->header_ops->validate(ll_header, len);
3212
3213 return false;
3214 }
3215
dev_has_header(const struct net_device * dev)3216 static inline bool dev_has_header(const struct net_device *dev)
3217 {
3218 return dev->header_ops && dev->header_ops->create;
3219 }
3220
3221 #ifdef CONFIG_NET_FLOW_LIMIT
3222 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3223 struct sd_flow_limit {
3224 u64 count;
3225 unsigned int num_buckets;
3226 unsigned int history_head;
3227 u16 history[FLOW_LIMIT_HISTORY];
3228 u8 buckets[];
3229 };
3230
3231 extern int netdev_flow_limit_table_len;
3232 #endif /* CONFIG_NET_FLOW_LIMIT */
3233
3234 /*
3235 * Incoming packets are placed on per-CPU queues
3236 */
3237 struct softnet_data {
3238 struct list_head poll_list;
3239 struct sk_buff_head process_queue;
3240
3241 /* stats */
3242 unsigned int processed;
3243 unsigned int time_squeeze;
3244 unsigned int received_rps;
3245 #ifdef CONFIG_RPS
3246 struct softnet_data *rps_ipi_list;
3247 #endif
3248 #ifdef CONFIG_NET_FLOW_LIMIT
3249 struct sd_flow_limit __rcu *flow_limit;
3250 #endif
3251 struct Qdisc *output_queue;
3252 struct Qdisc **output_queue_tailp;
3253 struct sk_buff *completion_queue;
3254 #ifdef CONFIG_XFRM_OFFLOAD
3255 struct sk_buff_head xfrm_backlog;
3256 #endif
3257 /* written and read only by owning cpu: */
3258 struct {
3259 u16 recursion;
3260 u8 more;
3261 } xmit;
3262 #ifdef CONFIG_RPS
3263 /* input_queue_head should be written by cpu owning this struct,
3264 * and only read by other cpus. Worth using a cache line.
3265 */
3266 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3267
3268 /* Elements below can be accessed between CPUs for RPS/RFS */
3269 call_single_data_t csd ____cacheline_aligned_in_smp;
3270 struct softnet_data *rps_ipi_next;
3271 unsigned int cpu;
3272 unsigned int input_queue_tail;
3273 #endif
3274 unsigned int dropped;
3275 struct sk_buff_head input_pkt_queue;
3276 struct napi_struct backlog;
3277
3278 };
3279
input_queue_head_incr(struct softnet_data * sd)3280 static inline void input_queue_head_incr(struct softnet_data *sd)
3281 {
3282 #ifdef CONFIG_RPS
3283 sd->input_queue_head++;
3284 #endif
3285 }
3286
input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)3287 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3288 unsigned int *qtail)
3289 {
3290 #ifdef CONFIG_RPS
3291 *qtail = ++sd->input_queue_tail;
3292 #endif
3293 }
3294
3295 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3296
dev_recursion_level(void)3297 static inline int dev_recursion_level(void)
3298 {
3299 return this_cpu_read(softnet_data.xmit.recursion);
3300 }
3301
3302 #define XMIT_RECURSION_LIMIT 8
dev_xmit_recursion(void)3303 static inline bool dev_xmit_recursion(void)
3304 {
3305 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3306 XMIT_RECURSION_LIMIT);
3307 }
3308
dev_xmit_recursion_inc(void)3309 static inline void dev_xmit_recursion_inc(void)
3310 {
3311 __this_cpu_inc(softnet_data.xmit.recursion);
3312 }
3313
dev_xmit_recursion_dec(void)3314 static inline void dev_xmit_recursion_dec(void)
3315 {
3316 __this_cpu_dec(softnet_data.xmit.recursion);
3317 }
3318
3319 void __netif_schedule(struct Qdisc *q);
3320 void netif_schedule_queue(struct netdev_queue *txq);
3321
netif_tx_schedule_all(struct net_device * dev)3322 static inline void netif_tx_schedule_all(struct net_device *dev)
3323 {
3324 unsigned int i;
3325
3326 for (i = 0; i < dev->num_tx_queues; i++)
3327 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3328 }
3329
netif_tx_start_queue(struct netdev_queue * dev_queue)3330 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3331 {
3332 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3333 }
3334
3335 /**
3336 * netif_start_queue - allow transmit
3337 * @dev: network device
3338 *
3339 * Allow upper layers to call the device hard_start_xmit routine.
3340 */
netif_start_queue(struct net_device * dev)3341 static inline void netif_start_queue(struct net_device *dev)
3342 {
3343 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3344 }
3345
netif_tx_start_all_queues(struct net_device * dev)3346 static inline void netif_tx_start_all_queues(struct net_device *dev)
3347 {
3348 unsigned int i;
3349
3350 for (i = 0; i < dev->num_tx_queues; i++) {
3351 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3352 netif_tx_start_queue(txq);
3353 }
3354 }
3355
3356 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3357
3358 /**
3359 * netif_wake_queue - restart transmit
3360 * @dev: network device
3361 *
3362 * Allow upper layers to call the device hard_start_xmit routine.
3363 * Used for flow control when transmit resources are available.
3364 */
netif_wake_queue(struct net_device * dev)3365 static inline void netif_wake_queue(struct net_device *dev)
3366 {
3367 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3368 }
3369
netif_tx_wake_all_queues(struct net_device * dev)3370 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3371 {
3372 unsigned int i;
3373
3374 for (i = 0; i < dev->num_tx_queues; i++) {
3375 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3376 netif_tx_wake_queue(txq);
3377 }
3378 }
3379
netif_tx_stop_queue(struct netdev_queue * dev_queue)3380 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3381 {
3382 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3383 }
3384
3385 /**
3386 * netif_stop_queue - stop transmitted packets
3387 * @dev: network device
3388 *
3389 * Stop upper layers calling the device hard_start_xmit routine.
3390 * Used for flow control when transmit resources are unavailable.
3391 */
netif_stop_queue(struct net_device * dev)3392 static inline void netif_stop_queue(struct net_device *dev)
3393 {
3394 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3395 }
3396
3397 void netif_tx_stop_all_queues(struct net_device *dev);
3398
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)3399 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3400 {
3401 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3402 }
3403
3404 /**
3405 * netif_queue_stopped - test if transmit queue is flowblocked
3406 * @dev: network device
3407 *
3408 * Test if transmit queue on device is currently unable to send.
3409 */
netif_queue_stopped(const struct net_device * dev)3410 static inline bool netif_queue_stopped(const struct net_device *dev)
3411 {
3412 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3413 }
3414
netif_xmit_stopped(const struct netdev_queue * dev_queue)3415 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3416 {
3417 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3418 }
3419
3420 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)3421 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3422 {
3423 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3424 }
3425
3426 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)3427 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3428 {
3429 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3430 }
3431
3432 /**
3433 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3434 * @dev_queue: pointer to transmit queue
3435 *
3436 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3437 * to give appropriate hint to the CPU.
3438 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)3439 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3440 {
3441 #ifdef CONFIG_BQL
3442 prefetchw(&dev_queue->dql.num_queued);
3443 #endif
3444 }
3445
3446 /**
3447 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3448 * @dev_queue: pointer to transmit queue
3449 *
3450 * BQL enabled drivers might use this helper in their TX completion path,
3451 * to give appropriate hint to the CPU.
3452 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3453 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3454 {
3455 #ifdef CONFIG_BQL
3456 prefetchw(&dev_queue->dql.limit);
3457 #endif
3458 }
3459
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3460 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3461 unsigned int bytes)
3462 {
3463 #ifdef CONFIG_BQL
3464 dql_queued(&dev_queue->dql, bytes);
3465
3466 if (likely(dql_avail(&dev_queue->dql) >= 0))
3467 return;
3468
3469 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3470
3471 /*
3472 * The XOFF flag must be set before checking the dql_avail below,
3473 * because in netdev_tx_completed_queue we update the dql_completed
3474 * before checking the XOFF flag.
3475 */
3476 smp_mb();
3477
3478 /* check again in case another CPU has just made room avail */
3479 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3480 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3481 #endif
3482 }
3483
3484 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3485 * that they should not test BQL status themselves.
3486 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3487 * skb of a batch.
3488 * Returns true if the doorbell must be used to kick the NIC.
3489 */
__netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes,bool xmit_more)3490 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3491 unsigned int bytes,
3492 bool xmit_more)
3493 {
3494 if (xmit_more) {
3495 #ifdef CONFIG_BQL
3496 dql_queued(&dev_queue->dql, bytes);
3497 #endif
3498 return netif_tx_queue_stopped(dev_queue);
3499 }
3500 netdev_tx_sent_queue(dev_queue, bytes);
3501 return true;
3502 }
3503
3504 /**
3505 * netdev_sent_queue - report the number of bytes queued to hardware
3506 * @dev: network device
3507 * @bytes: number of bytes queued to the hardware device queue
3508 *
3509 * Report the number of bytes queued for sending/completion to the network
3510 * device hardware queue. @bytes should be a good approximation and should
3511 * exactly match netdev_completed_queue() @bytes
3512 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)3513 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3514 {
3515 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3516 }
3517
__netdev_sent_queue(struct net_device * dev,unsigned int bytes,bool xmit_more)3518 static inline bool __netdev_sent_queue(struct net_device *dev,
3519 unsigned int bytes,
3520 bool xmit_more)
3521 {
3522 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3523 xmit_more);
3524 }
3525
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3526 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3527 unsigned int pkts, unsigned int bytes)
3528 {
3529 #ifdef CONFIG_BQL
3530 if (unlikely(!bytes))
3531 return;
3532
3533 dql_completed(&dev_queue->dql, bytes);
3534
3535 /*
3536 * Without the memory barrier there is a small possiblity that
3537 * netdev_tx_sent_queue will miss the update and cause the queue to
3538 * be stopped forever
3539 */
3540 smp_mb();
3541
3542 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3543 return;
3544
3545 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3546 netif_schedule_queue(dev_queue);
3547 #endif
3548 }
3549
3550 /**
3551 * netdev_completed_queue - report bytes and packets completed by device
3552 * @dev: network device
3553 * @pkts: actual number of packets sent over the medium
3554 * @bytes: actual number of bytes sent over the medium
3555 *
3556 * Report the number of bytes and packets transmitted by the network device
3557 * hardware queue over the physical medium, @bytes must exactly match the
3558 * @bytes amount passed to netdev_sent_queue()
3559 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3560 static inline void netdev_completed_queue(struct net_device *dev,
3561 unsigned int pkts, unsigned int bytes)
3562 {
3563 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3564 }
3565
netdev_tx_reset_queue(struct netdev_queue * q)3566 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3567 {
3568 #ifdef CONFIG_BQL
3569 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3570 dql_reset(&q->dql);
3571 #endif
3572 }
3573
3574 /**
3575 * netdev_reset_queue - reset the packets and bytes count of a network device
3576 * @dev_queue: network device
3577 *
3578 * Reset the bytes and packet count of a network device and clear the
3579 * software flow control OFF bit for this network device
3580 */
netdev_reset_queue(struct net_device * dev_queue)3581 static inline void netdev_reset_queue(struct net_device *dev_queue)
3582 {
3583 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3584 }
3585
3586 /**
3587 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3588 * @dev: network device
3589 * @queue_index: given tx queue index
3590 *
3591 * Returns 0 if given tx queue index >= number of device tx queues,
3592 * otherwise returns the originally passed tx queue index.
3593 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3594 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3595 {
3596 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3597 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3598 dev->name, queue_index,
3599 dev->real_num_tx_queues);
3600 return 0;
3601 }
3602
3603 return queue_index;
3604 }
3605
3606 /**
3607 * netif_running - test if up
3608 * @dev: network device
3609 *
3610 * Test if the device has been brought up.
3611 */
netif_running(const struct net_device * dev)3612 static inline bool netif_running(const struct net_device *dev)
3613 {
3614 return test_bit(__LINK_STATE_START, &dev->state);
3615 }
3616
3617 /*
3618 * Routines to manage the subqueues on a device. We only need start,
3619 * stop, and a check if it's stopped. All other device management is
3620 * done at the overall netdevice level.
3621 * Also test the device if we're multiqueue.
3622 */
3623
3624 /**
3625 * netif_start_subqueue - allow sending packets on subqueue
3626 * @dev: network device
3627 * @queue_index: sub queue index
3628 *
3629 * Start individual transmit queue of a device with multiple transmit queues.
3630 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)3631 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3632 {
3633 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3634
3635 netif_tx_start_queue(txq);
3636 }
3637
3638 /**
3639 * netif_stop_subqueue - stop sending packets on subqueue
3640 * @dev: network device
3641 * @queue_index: sub queue index
3642 *
3643 * Stop individual transmit queue of a device with multiple transmit queues.
3644 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)3645 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3646 {
3647 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3648 netif_tx_stop_queue(txq);
3649 }
3650
3651 /**
3652 * netif_subqueue_stopped - test status of subqueue
3653 * @dev: network device
3654 * @queue_index: sub queue index
3655 *
3656 * Check individual transmit queue of a device with multiple transmit queues.
3657 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3658 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3659 u16 queue_index)
3660 {
3661 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3662
3663 return netif_tx_queue_stopped(txq);
3664 }
3665
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3666 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3667 struct sk_buff *skb)
3668 {
3669 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3670 }
3671
3672 /**
3673 * netif_wake_subqueue - allow sending packets on subqueue
3674 * @dev: network device
3675 * @queue_index: sub queue index
3676 *
3677 * Resume individual transmit queue of a device with multiple transmit queues.
3678 */
netif_wake_subqueue(struct net_device * dev,u16 queue_index)3679 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3680 {
3681 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3682
3683 netif_tx_wake_queue(txq);
3684 }
3685
3686 #ifdef CONFIG_XPS
3687 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3688 u16 index);
3689 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3690 u16 index, bool is_rxqs_map);
3691
3692 /**
3693 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3694 * @j: CPU/Rx queue index
3695 * @mask: bitmask of all cpus/rx queues
3696 * @nr_bits: number of bits in the bitmask
3697 *
3698 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3699 */
netif_attr_test_mask(unsigned long j,const unsigned long * mask,unsigned int nr_bits)3700 static inline bool netif_attr_test_mask(unsigned long j,
3701 const unsigned long *mask,
3702 unsigned int nr_bits)
3703 {
3704 cpu_max_bits_warn(j, nr_bits);
3705 return test_bit(j, mask);
3706 }
3707
3708 /**
3709 * netif_attr_test_online - Test for online CPU/Rx queue
3710 * @j: CPU/Rx queue index
3711 * @online_mask: bitmask for CPUs/Rx queues that are online
3712 * @nr_bits: number of bits in the bitmask
3713 *
3714 * Returns true if a CPU/Rx queue is online.
3715 */
netif_attr_test_online(unsigned long j,const unsigned long * online_mask,unsigned int nr_bits)3716 static inline bool netif_attr_test_online(unsigned long j,
3717 const unsigned long *online_mask,
3718 unsigned int nr_bits)
3719 {
3720 cpu_max_bits_warn(j, nr_bits);
3721
3722 if (online_mask)
3723 return test_bit(j, online_mask);
3724
3725 return (j < nr_bits);
3726 }
3727
3728 /**
3729 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3730 * @n: CPU/Rx queue index
3731 * @srcp: the cpumask/Rx queue mask pointer
3732 * @nr_bits: number of bits in the bitmask
3733 *
3734 * Returns >= nr_bits if no further CPUs/Rx queues set.
3735 */
netif_attrmask_next(int n,const unsigned long * srcp,unsigned int nr_bits)3736 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3737 unsigned int nr_bits)
3738 {
3739 /* -1 is a legal arg here. */
3740 if (n != -1)
3741 cpu_max_bits_warn(n, nr_bits);
3742
3743 if (srcp)
3744 return find_next_bit(srcp, nr_bits, n + 1);
3745
3746 return n + 1;
3747 }
3748
3749 /**
3750 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3751 * @n: CPU/Rx queue index
3752 * @src1p: the first CPUs/Rx queues mask pointer
3753 * @src2p: the second CPUs/Rx queues mask pointer
3754 * @nr_bits: number of bits in the bitmask
3755 *
3756 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3757 */
netif_attrmask_next_and(int n,const unsigned long * src1p,const unsigned long * src2p,unsigned int nr_bits)3758 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3759 const unsigned long *src2p,
3760 unsigned int nr_bits)
3761 {
3762 /* -1 is a legal arg here. */
3763 if (n != -1)
3764 cpu_max_bits_warn(n, nr_bits);
3765
3766 if (src1p && src2p)
3767 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3768 else if (src1p)
3769 return find_next_bit(src1p, nr_bits, n + 1);
3770 else if (src2p)
3771 return find_next_bit(src2p, nr_bits, n + 1);
3772
3773 return n + 1;
3774 }
3775 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3776 static inline int netif_set_xps_queue(struct net_device *dev,
3777 const struct cpumask *mask,
3778 u16 index)
3779 {
3780 return 0;
3781 }
3782
__netif_set_xps_queue(struct net_device * dev,const unsigned long * mask,u16 index,bool is_rxqs_map)3783 static inline int __netif_set_xps_queue(struct net_device *dev,
3784 const unsigned long *mask,
3785 u16 index, bool is_rxqs_map)
3786 {
3787 return 0;
3788 }
3789 #endif
3790
3791 /**
3792 * netif_is_multiqueue - test if device has multiple transmit queues
3793 * @dev: network device
3794 *
3795 * Check if device has multiple transmit queues
3796 */
netif_is_multiqueue(const struct net_device * dev)3797 static inline bool netif_is_multiqueue(const struct net_device *dev)
3798 {
3799 return dev->num_tx_queues > 1;
3800 }
3801
3802 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3803
3804 #ifdef CONFIG_SYSFS
3805 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3806 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxqs)3807 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3808 unsigned int rxqs)
3809 {
3810 dev->real_num_rx_queues = rxqs;
3811 return 0;
3812 }
3813 #endif
3814
3815 static inline struct netdev_rx_queue *
__netif_get_rx_queue(struct net_device * dev,unsigned int rxq)3816 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3817 {
3818 return dev->_rx + rxq;
3819 }
3820
3821 #ifdef CONFIG_SYSFS
get_netdev_rx_queue_index(struct netdev_rx_queue * queue)3822 static inline unsigned int get_netdev_rx_queue_index(
3823 struct netdev_rx_queue *queue)
3824 {
3825 struct net_device *dev = queue->dev;
3826 int index = queue - dev->_rx;
3827
3828 BUG_ON(index >= dev->num_rx_queues);
3829 return index;
3830 }
3831 #endif
3832
3833 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3834 int netif_get_num_default_rss_queues(void);
3835
3836 enum skb_free_reason {
3837 SKB_REASON_CONSUMED,
3838 SKB_REASON_DROPPED,
3839 };
3840
3841 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3842 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3843
3844 /*
3845 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3846 * interrupt context or with hardware interrupts being disabled.
3847 * (in_irq() || irqs_disabled())
3848 *
3849 * We provide four helpers that can be used in following contexts :
3850 *
3851 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3852 * replacing kfree_skb(skb)
3853 *
3854 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3855 * Typically used in place of consume_skb(skb) in TX completion path
3856 *
3857 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3858 * replacing kfree_skb(skb)
3859 *
3860 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3861 * and consumed a packet. Used in place of consume_skb(skb)
3862 */
dev_kfree_skb_irq(struct sk_buff * skb)3863 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3864 {
3865 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3866 }
3867
dev_consume_skb_irq(struct sk_buff * skb)3868 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3869 {
3870 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3871 }
3872
dev_kfree_skb_any(struct sk_buff * skb)3873 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3874 {
3875 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3876 }
3877
dev_consume_skb_any(struct sk_buff * skb)3878 static inline void dev_consume_skb_any(struct sk_buff *skb)
3879 {
3880 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3881 }
3882
3883 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3884 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3885 int netif_rx(struct sk_buff *skb);
3886 int netif_rx_ni(struct sk_buff *skb);
3887 int netif_rx_any_context(struct sk_buff *skb);
3888 int netif_receive_skb(struct sk_buff *skb);
3889 int netif_receive_skb_core(struct sk_buff *skb);
3890 void netif_receive_skb_list(struct list_head *head);
3891 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3892 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3893 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3894 gro_result_t napi_gro_frags(struct napi_struct *napi);
3895 struct packet_offload *gro_find_receive_by_type(__be16 type);
3896 struct packet_offload *gro_find_complete_by_type(__be16 type);
3897
napi_free_frags(struct napi_struct * napi)3898 static inline void napi_free_frags(struct napi_struct *napi)
3899 {
3900 kfree_skb(napi->skb);
3901 napi->skb = NULL;
3902 }
3903
3904 bool netdev_is_rx_handler_busy(struct net_device *dev);
3905 int netdev_rx_handler_register(struct net_device *dev,
3906 rx_handler_func_t *rx_handler,
3907 void *rx_handler_data);
3908 void netdev_rx_handler_unregister(struct net_device *dev);
3909
3910 bool dev_valid_name(const char *name);
is_socket_ioctl_cmd(unsigned int cmd)3911 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3912 {
3913 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3914 }
3915 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3916 bool *need_copyout);
3917 int dev_ifconf(struct net *net, struct ifconf *, int);
3918 int dev_ethtool(struct net *net, struct ifreq *);
3919 unsigned int dev_get_flags(const struct net_device *);
3920 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3921 struct netlink_ext_ack *extack);
3922 int dev_change_flags(struct net_device *dev, unsigned int flags,
3923 struct netlink_ext_ack *extack);
3924 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3925 unsigned int gchanges);
3926 int dev_change_name(struct net_device *, const char *);
3927 int dev_set_alias(struct net_device *, const char *, size_t);
3928 int dev_get_alias(const struct net_device *, char *, size_t);
3929 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3930 int __dev_set_mtu(struct net_device *, int);
3931 int dev_validate_mtu(struct net_device *dev, int mtu,
3932 struct netlink_ext_ack *extack);
3933 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3934 struct netlink_ext_ack *extack);
3935 int dev_set_mtu(struct net_device *, int);
3936 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3937 void dev_set_group(struct net_device *, int);
3938 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3939 struct netlink_ext_ack *extack);
3940 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3941 struct netlink_ext_ack *extack);
3942 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3943 struct netlink_ext_ack *extack);
3944 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3945 int dev_change_carrier(struct net_device *, bool new_carrier);
3946 int dev_get_phys_port_id(struct net_device *dev,
3947 struct netdev_phys_item_id *ppid);
3948 int dev_get_phys_port_name(struct net_device *dev,
3949 char *name, size_t len);
3950 int dev_get_port_parent_id(struct net_device *dev,
3951 struct netdev_phys_item_id *ppid, bool recurse);
3952 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3953 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3954 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3955 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3956 u32 value);
3957 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3958 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3959 struct netdev_queue *txq, int *ret);
3960
3961 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3962 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3963 int fd, int expected_fd, u32 flags);
3964 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3965 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3966
3967 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3968
3969 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3970 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3971 bool is_skb_forwardable(const struct net_device *dev,
3972 const struct sk_buff *skb);
3973
____dev_forward_skb(struct net_device * dev,struct sk_buff * skb)3974 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3975 struct sk_buff *skb)
3976 {
3977 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3978 unlikely(!is_skb_forwardable(dev, skb))) {
3979 atomic_long_inc(&dev->rx_dropped);
3980 kfree_skb(skb);
3981 return NET_RX_DROP;
3982 }
3983
3984 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
3985 skb->priority = 0;
3986 return 0;
3987 }
3988
3989 bool dev_nit_active(struct net_device *dev);
3990 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3991
3992 extern int netdev_budget;
3993 extern unsigned int netdev_budget_usecs;
3994
3995 /* Called by rtnetlink.c:rtnl_unlock() */
3996 void netdev_run_todo(void);
3997
3998 /**
3999 * dev_put - release reference to device
4000 * @dev: network device
4001 *
4002 * Release reference to device to allow it to be freed.
4003 */
dev_put(struct net_device * dev)4004 static inline void dev_put(struct net_device *dev)
4005 {
4006 if (dev)
4007 this_cpu_dec(*dev->pcpu_refcnt);
4008 }
4009
4010 /**
4011 * dev_hold - get reference to device
4012 * @dev: network device
4013 *
4014 * Hold reference to device to keep it from being freed.
4015 */
dev_hold(struct net_device * dev)4016 static inline void dev_hold(struct net_device *dev)
4017 {
4018 if (dev)
4019 this_cpu_inc(*dev->pcpu_refcnt);
4020 }
4021
4022 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4023 * and _off may be called from IRQ context, but it is caller
4024 * who is responsible for serialization of these calls.
4025 *
4026 * The name carrier is inappropriate, these functions should really be
4027 * called netif_lowerlayer_*() because they represent the state of any
4028 * kind of lower layer not just hardware media.
4029 */
4030
4031 void linkwatch_init_dev(struct net_device *dev);
4032 void linkwatch_fire_event(struct net_device *dev);
4033 void linkwatch_forget_dev(struct net_device *dev);
4034
4035 /**
4036 * netif_carrier_ok - test if carrier present
4037 * @dev: network device
4038 *
4039 * Check if carrier is present on device
4040 */
netif_carrier_ok(const struct net_device * dev)4041 static inline bool netif_carrier_ok(const struct net_device *dev)
4042 {
4043 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4044 }
4045
4046 unsigned long dev_trans_start(struct net_device *dev);
4047
4048 void __netdev_watchdog_up(struct net_device *dev);
4049
4050 void netif_carrier_on(struct net_device *dev);
4051
4052 void netif_carrier_off(struct net_device *dev);
4053
4054 /**
4055 * netif_dormant_on - mark device as dormant.
4056 * @dev: network device
4057 *
4058 * Mark device as dormant (as per RFC2863).
4059 *
4060 * The dormant state indicates that the relevant interface is not
4061 * actually in a condition to pass packets (i.e., it is not 'up') but is
4062 * in a "pending" state, waiting for some external event. For "on-
4063 * demand" interfaces, this new state identifies the situation where the
4064 * interface is waiting for events to place it in the up state.
4065 */
netif_dormant_on(struct net_device * dev)4066 static inline void netif_dormant_on(struct net_device *dev)
4067 {
4068 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4069 linkwatch_fire_event(dev);
4070 }
4071
4072 /**
4073 * netif_dormant_off - set device as not dormant.
4074 * @dev: network device
4075 *
4076 * Device is not in dormant state.
4077 */
netif_dormant_off(struct net_device * dev)4078 static inline void netif_dormant_off(struct net_device *dev)
4079 {
4080 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4081 linkwatch_fire_event(dev);
4082 }
4083
4084 /**
4085 * netif_dormant - test if device is dormant
4086 * @dev: network device
4087 *
4088 * Check if device is dormant.
4089 */
netif_dormant(const struct net_device * dev)4090 static inline bool netif_dormant(const struct net_device *dev)
4091 {
4092 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4093 }
4094
4095
4096 /**
4097 * netif_testing_on - mark device as under test.
4098 * @dev: network device
4099 *
4100 * Mark device as under test (as per RFC2863).
4101 *
4102 * The testing state indicates that some test(s) must be performed on
4103 * the interface. After completion, of the test, the interface state
4104 * will change to up, dormant, or down, as appropriate.
4105 */
netif_testing_on(struct net_device * dev)4106 static inline void netif_testing_on(struct net_device *dev)
4107 {
4108 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4109 linkwatch_fire_event(dev);
4110 }
4111
4112 /**
4113 * netif_testing_off - set device as not under test.
4114 * @dev: network device
4115 *
4116 * Device is not in testing state.
4117 */
netif_testing_off(struct net_device * dev)4118 static inline void netif_testing_off(struct net_device *dev)
4119 {
4120 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4121 linkwatch_fire_event(dev);
4122 }
4123
4124 /**
4125 * netif_testing - test if device is under test
4126 * @dev: network device
4127 *
4128 * Check if device is under test
4129 */
netif_testing(const struct net_device * dev)4130 static inline bool netif_testing(const struct net_device *dev)
4131 {
4132 return test_bit(__LINK_STATE_TESTING, &dev->state);
4133 }
4134
4135
4136 /**
4137 * netif_oper_up - test if device is operational
4138 * @dev: network device
4139 *
4140 * Check if carrier is operational
4141 */
netif_oper_up(const struct net_device * dev)4142 static inline bool netif_oper_up(const struct net_device *dev)
4143 {
4144 return (dev->operstate == IF_OPER_UP ||
4145 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4146 }
4147
4148 /**
4149 * netif_device_present - is device available or removed
4150 * @dev: network device
4151 *
4152 * Check if device has not been removed from system.
4153 */
netif_device_present(struct net_device * dev)4154 static inline bool netif_device_present(struct net_device *dev)
4155 {
4156 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4157 }
4158
4159 void netif_device_detach(struct net_device *dev);
4160
4161 void netif_device_attach(struct net_device *dev);
4162
4163 /*
4164 * Network interface message level settings
4165 */
4166
4167 enum {
4168 NETIF_MSG_DRV_BIT,
4169 NETIF_MSG_PROBE_BIT,
4170 NETIF_MSG_LINK_BIT,
4171 NETIF_MSG_TIMER_BIT,
4172 NETIF_MSG_IFDOWN_BIT,
4173 NETIF_MSG_IFUP_BIT,
4174 NETIF_MSG_RX_ERR_BIT,
4175 NETIF_MSG_TX_ERR_BIT,
4176 NETIF_MSG_TX_QUEUED_BIT,
4177 NETIF_MSG_INTR_BIT,
4178 NETIF_MSG_TX_DONE_BIT,
4179 NETIF_MSG_RX_STATUS_BIT,
4180 NETIF_MSG_PKTDATA_BIT,
4181 NETIF_MSG_HW_BIT,
4182 NETIF_MSG_WOL_BIT,
4183
4184 /* When you add a new bit above, update netif_msg_class_names array
4185 * in net/ethtool/common.c
4186 */
4187 NETIF_MSG_CLASS_COUNT,
4188 };
4189 /* Both ethtool_ops interface and internal driver implementation use u32 */
4190 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4191
4192 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4193 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4194
4195 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4196 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4197 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4198 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4199 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4200 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4201 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4202 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4203 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4204 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4205 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4206 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4207 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4208 #define NETIF_MSG_HW __NETIF_MSG(HW)
4209 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4210
4211 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4212 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4213 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4214 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4215 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4216 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4217 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4218 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4219 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4220 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4221 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4222 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4223 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4224 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4225 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4226
netif_msg_init(int debug_value,int default_msg_enable_bits)4227 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4228 {
4229 /* use default */
4230 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4231 return default_msg_enable_bits;
4232 if (debug_value == 0) /* no output */
4233 return 0;
4234 /* set low N bits */
4235 return (1U << debug_value) - 1;
4236 }
4237
__netif_tx_lock(struct netdev_queue * txq,int cpu)4238 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4239 {
4240 spin_lock(&txq->_xmit_lock);
4241 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4242 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4243 }
4244
__netif_tx_acquire(struct netdev_queue * txq)4245 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4246 {
4247 __acquire(&txq->_xmit_lock);
4248 return true;
4249 }
4250
__netif_tx_release(struct netdev_queue * txq)4251 static inline void __netif_tx_release(struct netdev_queue *txq)
4252 {
4253 __release(&txq->_xmit_lock);
4254 }
4255
__netif_tx_lock_bh(struct netdev_queue * txq)4256 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4257 {
4258 spin_lock_bh(&txq->_xmit_lock);
4259 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4260 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4261 }
4262
__netif_tx_trylock(struct netdev_queue * txq)4263 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4264 {
4265 bool ok = spin_trylock(&txq->_xmit_lock);
4266
4267 if (likely(ok)) {
4268 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4269 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4270 }
4271 return ok;
4272 }
4273
__netif_tx_unlock(struct netdev_queue * txq)4274 static inline void __netif_tx_unlock(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(&txq->_xmit_lock);
4279 }
4280
__netif_tx_unlock_bh(struct netdev_queue * txq)4281 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4282 {
4283 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4284 WRITE_ONCE(txq->xmit_lock_owner, -1);
4285 spin_unlock_bh(&txq->_xmit_lock);
4286 }
4287
txq_trans_update(struct netdev_queue * txq)4288 static inline void txq_trans_update(struct netdev_queue *txq)
4289 {
4290 if (txq->xmit_lock_owner != -1)
4291 txq->trans_start = jiffies;
4292 }
4293
4294 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
netif_trans_update(struct net_device * dev)4295 static inline void netif_trans_update(struct net_device *dev)
4296 {
4297 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4298
4299 if (txq->trans_start != jiffies)
4300 txq->trans_start = jiffies;
4301 }
4302
4303 /**
4304 * netif_tx_lock - grab network device transmit lock
4305 * @dev: network device
4306 *
4307 * Get network device transmit lock
4308 */
netif_tx_lock(struct net_device * dev)4309 static inline void netif_tx_lock(struct net_device *dev)
4310 {
4311 unsigned int i;
4312 int cpu;
4313
4314 spin_lock(&dev->tx_global_lock);
4315 cpu = smp_processor_id();
4316 for (i = 0; i < dev->num_tx_queues; i++) {
4317 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4318
4319 /* We are the only thread of execution doing a
4320 * freeze, but we have to grab the _xmit_lock in
4321 * order to synchronize with threads which are in
4322 * the ->hard_start_xmit() handler and already
4323 * checked the frozen bit.
4324 */
4325 __netif_tx_lock(txq, cpu);
4326 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4327 __netif_tx_unlock(txq);
4328 }
4329 }
4330
netif_tx_lock_bh(struct net_device * dev)4331 static inline void netif_tx_lock_bh(struct net_device *dev)
4332 {
4333 local_bh_disable();
4334 netif_tx_lock(dev);
4335 }
4336
netif_tx_unlock(struct net_device * dev)4337 static inline void netif_tx_unlock(struct net_device *dev)
4338 {
4339 unsigned int i;
4340
4341 for (i = 0; i < dev->num_tx_queues; i++) {
4342 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4343
4344 /* No need to grab the _xmit_lock here. If the
4345 * queue is not stopped for another reason, we
4346 * force a schedule.
4347 */
4348 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4349 netif_schedule_queue(txq);
4350 }
4351 spin_unlock(&dev->tx_global_lock);
4352 }
4353
netif_tx_unlock_bh(struct net_device * dev)4354 static inline void netif_tx_unlock_bh(struct net_device *dev)
4355 {
4356 netif_tx_unlock(dev);
4357 local_bh_enable();
4358 }
4359
4360 #define HARD_TX_LOCK(dev, txq, cpu) { \
4361 if ((dev->features & NETIF_F_LLTX) == 0) { \
4362 __netif_tx_lock(txq, cpu); \
4363 } else { \
4364 __netif_tx_acquire(txq); \
4365 } \
4366 }
4367
4368 #define HARD_TX_TRYLOCK(dev, txq) \
4369 (((dev->features & NETIF_F_LLTX) == 0) ? \
4370 __netif_tx_trylock(txq) : \
4371 __netif_tx_acquire(txq))
4372
4373 #define HARD_TX_UNLOCK(dev, txq) { \
4374 if ((dev->features & NETIF_F_LLTX) == 0) { \
4375 __netif_tx_unlock(txq); \
4376 } else { \
4377 __netif_tx_release(txq); \
4378 } \
4379 }
4380
netif_tx_disable(struct net_device * dev)4381 static inline void netif_tx_disable(struct net_device *dev)
4382 {
4383 unsigned int i;
4384 int cpu;
4385
4386 local_bh_disable();
4387 cpu = smp_processor_id();
4388 spin_lock(&dev->tx_global_lock);
4389 for (i = 0; i < dev->num_tx_queues; i++) {
4390 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4391
4392 __netif_tx_lock(txq, cpu);
4393 netif_tx_stop_queue(txq);
4394 __netif_tx_unlock(txq);
4395 }
4396 spin_unlock(&dev->tx_global_lock);
4397 local_bh_enable();
4398 }
4399
netif_addr_lock(struct net_device * dev)4400 static inline void netif_addr_lock(struct net_device *dev)
4401 {
4402 unsigned char nest_level = 0;
4403
4404 #ifdef CONFIG_LOCKDEP
4405 nest_level = dev->nested_level;
4406 #endif
4407 spin_lock_nested(&dev->addr_list_lock, nest_level);
4408 }
4409
netif_addr_lock_bh(struct net_device * dev)4410 static inline void netif_addr_lock_bh(struct net_device *dev)
4411 {
4412 unsigned char nest_level = 0;
4413
4414 #ifdef CONFIG_LOCKDEP
4415 nest_level = dev->nested_level;
4416 #endif
4417 local_bh_disable();
4418 spin_lock_nested(&dev->addr_list_lock, nest_level);
4419 }
4420
netif_addr_unlock(struct net_device * dev)4421 static inline void netif_addr_unlock(struct net_device *dev)
4422 {
4423 spin_unlock(&dev->addr_list_lock);
4424 }
4425
netif_addr_unlock_bh(struct net_device * dev)4426 static inline void netif_addr_unlock_bh(struct net_device *dev)
4427 {
4428 spin_unlock_bh(&dev->addr_list_lock);
4429 }
4430
4431 /*
4432 * dev_addrs walker. Should be used only for read access. Call with
4433 * rcu_read_lock held.
4434 */
4435 #define for_each_dev_addr(dev, ha) \
4436 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4437
4438 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4439
4440 void ether_setup(struct net_device *dev);
4441
4442 /* Support for loadable net-drivers */
4443 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4444 unsigned char name_assign_type,
4445 void (*setup)(struct net_device *),
4446 unsigned int txqs, unsigned int rxqs);
4447 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4448 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4449
4450 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4451 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4452 count)
4453
4454 int register_netdev(struct net_device *dev);
4455 void unregister_netdev(struct net_device *dev);
4456
4457 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4458
4459 /* General hardware address lists handling functions */
4460 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4461 struct netdev_hw_addr_list *from_list, int addr_len);
4462 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4463 struct netdev_hw_addr_list *from_list, int addr_len);
4464 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4465 struct net_device *dev,
4466 int (*sync)(struct net_device *, const unsigned char *),
4467 int (*unsync)(struct net_device *,
4468 const unsigned char *));
4469 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4470 struct net_device *dev,
4471 int (*sync)(struct net_device *,
4472 const unsigned char *, int),
4473 int (*unsync)(struct net_device *,
4474 const unsigned char *, int));
4475 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4476 struct net_device *dev,
4477 int (*unsync)(struct net_device *,
4478 const unsigned char *, int));
4479 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4480 struct net_device *dev,
4481 int (*unsync)(struct net_device *,
4482 const unsigned char *));
4483 void __hw_addr_init(struct netdev_hw_addr_list *list);
4484
4485 /* Functions used for device addresses handling */
4486 static inline void
__dev_addr_set(struct net_device * dev,const u8 * addr,size_t len)4487 __dev_addr_set(struct net_device *dev, const u8 *addr, size_t len)
4488 {
4489 memcpy(dev->dev_addr, addr, len);
4490 }
4491
dev_addr_set(struct net_device * dev,const u8 * addr)4492 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4493 {
4494 __dev_addr_set(dev, addr, dev->addr_len);
4495 }
4496
4497 static inline void
dev_addr_mod(struct net_device * dev,unsigned int offset,const u8 * addr,size_t len)4498 dev_addr_mod(struct net_device *dev, unsigned int offset,
4499 const u8 *addr, size_t len)
4500 {
4501 memcpy(&dev->dev_addr[offset], addr, len);
4502 }
4503
4504 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4505 unsigned char addr_type);
4506 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4507 unsigned char addr_type);
4508 void dev_addr_flush(struct net_device *dev);
4509 int dev_addr_init(struct net_device *dev);
4510
4511 /* Functions used for unicast addresses handling */
4512 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4513 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4514 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4515 int dev_uc_sync(struct net_device *to, struct net_device *from);
4516 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4517 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4518 void dev_uc_flush(struct net_device *dev);
4519 void dev_uc_init(struct net_device *dev);
4520
4521 /**
4522 * __dev_uc_sync - Synchonize device's unicast list
4523 * @dev: device to sync
4524 * @sync: function to call if address should be added
4525 * @unsync: function to call if address should be removed
4526 *
4527 * Add newly added addresses to the interface, and release
4528 * addresses that have been deleted.
4529 */
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4530 static inline int __dev_uc_sync(struct net_device *dev,
4531 int (*sync)(struct net_device *,
4532 const unsigned char *),
4533 int (*unsync)(struct net_device *,
4534 const unsigned char *))
4535 {
4536 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4537 }
4538
4539 /**
4540 * __dev_uc_unsync - Remove synchronized addresses from device
4541 * @dev: device to sync
4542 * @unsync: function to call if address should be removed
4543 *
4544 * Remove all addresses that were added to the device by dev_uc_sync().
4545 */
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4546 static inline void __dev_uc_unsync(struct net_device *dev,
4547 int (*unsync)(struct net_device *,
4548 const unsigned char *))
4549 {
4550 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4551 }
4552
4553 /* Functions used for multicast addresses handling */
4554 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4555 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4556 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4557 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4558 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4559 int dev_mc_sync(struct net_device *to, struct net_device *from);
4560 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4561 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4562 void dev_mc_flush(struct net_device *dev);
4563 void dev_mc_init(struct net_device *dev);
4564
4565 /**
4566 * __dev_mc_sync - Synchonize device's multicast list
4567 * @dev: device to sync
4568 * @sync: function to call if address should be added
4569 * @unsync: function to call if address should be removed
4570 *
4571 * Add newly added addresses to the interface, and release
4572 * addresses that have been deleted.
4573 */
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4574 static inline int __dev_mc_sync(struct net_device *dev,
4575 int (*sync)(struct net_device *,
4576 const unsigned char *),
4577 int (*unsync)(struct net_device *,
4578 const unsigned char *))
4579 {
4580 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4581 }
4582
4583 /**
4584 * __dev_mc_unsync - Remove synchronized addresses from device
4585 * @dev: device to sync
4586 * @unsync: function to call if address should be removed
4587 *
4588 * Remove all addresses that were added to the device by dev_mc_sync().
4589 */
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4590 static inline void __dev_mc_unsync(struct net_device *dev,
4591 int (*unsync)(struct net_device *,
4592 const unsigned char *))
4593 {
4594 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4595 }
4596
4597 /* Functions used for secondary unicast and multicast support */
4598 void dev_set_rx_mode(struct net_device *dev);
4599 void __dev_set_rx_mode(struct net_device *dev);
4600 int dev_set_promiscuity(struct net_device *dev, int inc);
4601 int dev_set_allmulti(struct net_device *dev, int inc);
4602 void netdev_state_change(struct net_device *dev);
4603 void netdev_notify_peers(struct net_device *dev);
4604 void netdev_features_change(struct net_device *dev);
4605 /* Load a device via the kmod */
4606 void dev_load(struct net *net, const char *name);
4607 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4608 struct rtnl_link_stats64 *storage);
4609 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4610 const struct net_device_stats *netdev_stats);
4611 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4612 const struct pcpu_sw_netstats __percpu *netstats);
4613
4614 extern int netdev_max_backlog;
4615 extern int netdev_tstamp_prequeue;
4616 extern int weight_p;
4617 extern int dev_weight_rx_bias;
4618 extern int dev_weight_tx_bias;
4619 extern int dev_rx_weight;
4620 extern int dev_tx_weight;
4621 extern int gro_normal_batch;
4622
4623 enum {
4624 NESTED_SYNC_IMM_BIT,
4625 NESTED_SYNC_TODO_BIT,
4626 };
4627
4628 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4629 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4630
4631 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4632 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4633
4634 struct netdev_nested_priv {
4635 unsigned char flags;
4636 void *data;
4637 };
4638
4639 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4640 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4641 struct list_head **iter);
4642 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4643 struct list_head **iter);
4644
4645 #ifdef CONFIG_LOCKDEP
4646 static LIST_HEAD(net_unlink_list);
4647
net_unlink_todo(struct net_device * dev)4648 static inline void net_unlink_todo(struct net_device *dev)
4649 {
4650 if (list_empty(&dev->unlink_list))
4651 list_add_tail(&dev->unlink_list, &net_unlink_list);
4652 }
4653 #endif
4654
4655 /* iterate through upper list, must be called under RCU read lock */
4656 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4657 for (iter = &(dev)->adj_list.upper, \
4658 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4659 updev; \
4660 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4661
4662 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4663 int (*fn)(struct net_device *upper_dev,
4664 struct netdev_nested_priv *priv),
4665 struct netdev_nested_priv *priv);
4666
4667 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4668 struct net_device *upper_dev);
4669
4670 bool netdev_has_any_upper_dev(struct net_device *dev);
4671
4672 void *netdev_lower_get_next_private(struct net_device *dev,
4673 struct list_head **iter);
4674 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4675 struct list_head **iter);
4676
4677 #define netdev_for_each_lower_private(dev, priv, iter) \
4678 for (iter = (dev)->adj_list.lower.next, \
4679 priv = netdev_lower_get_next_private(dev, &(iter)); \
4680 priv; \
4681 priv = netdev_lower_get_next_private(dev, &(iter)))
4682
4683 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4684 for (iter = &(dev)->adj_list.lower, \
4685 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4686 priv; \
4687 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4688
4689 void *netdev_lower_get_next(struct net_device *dev,
4690 struct list_head **iter);
4691
4692 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4693 for (iter = (dev)->adj_list.lower.next, \
4694 ldev = netdev_lower_get_next(dev, &(iter)); \
4695 ldev; \
4696 ldev = netdev_lower_get_next(dev, &(iter)))
4697
4698 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4699 struct list_head **iter);
4700 int netdev_walk_all_lower_dev(struct net_device *dev,
4701 int (*fn)(struct net_device *lower_dev,
4702 struct netdev_nested_priv *priv),
4703 struct netdev_nested_priv *priv);
4704 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4705 int (*fn)(struct net_device *lower_dev,
4706 struct netdev_nested_priv *priv),
4707 struct netdev_nested_priv *priv);
4708
4709 void *netdev_adjacent_get_private(struct list_head *adj_list);
4710 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4711 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4712 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4713 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4714 struct netlink_ext_ack *extack);
4715 int netdev_master_upper_dev_link(struct net_device *dev,
4716 struct net_device *upper_dev,
4717 void *upper_priv, void *upper_info,
4718 struct netlink_ext_ack *extack);
4719 void netdev_upper_dev_unlink(struct net_device *dev,
4720 struct net_device *upper_dev);
4721 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4722 struct net_device *new_dev,
4723 struct net_device *dev,
4724 struct netlink_ext_ack *extack);
4725 void netdev_adjacent_change_commit(struct net_device *old_dev,
4726 struct net_device *new_dev,
4727 struct net_device *dev);
4728 void netdev_adjacent_change_abort(struct net_device *old_dev,
4729 struct net_device *new_dev,
4730 struct net_device *dev);
4731 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4732 void *netdev_lower_dev_get_private(struct net_device *dev,
4733 struct net_device *lower_dev);
4734 void netdev_lower_state_changed(struct net_device *lower_dev,
4735 void *lower_state_info);
4736
4737 /* RSS keys are 40 or 52 bytes long */
4738 #define NETDEV_RSS_KEY_LEN 52
4739 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4740 void netdev_rss_key_fill(void *buffer, size_t len);
4741
4742 int skb_checksum_help(struct sk_buff *skb);
4743 int skb_crc32c_csum_help(struct sk_buff *skb);
4744 int skb_csum_hwoffload_help(struct sk_buff *skb,
4745 const netdev_features_t features);
4746
4747 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4748 netdev_features_t features, bool tx_path);
4749 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4750 netdev_features_t features);
4751
4752 struct netdev_bonding_info {
4753 ifslave slave;
4754 ifbond master;
4755 };
4756
4757 struct netdev_notifier_bonding_info {
4758 struct netdev_notifier_info info; /* must be first */
4759 struct netdev_bonding_info bonding_info;
4760 };
4761
4762 void netdev_bonding_info_change(struct net_device *dev,
4763 struct netdev_bonding_info *bonding_info);
4764
4765 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4766 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4767 #else
ethtool_notify(struct net_device * dev,unsigned int cmd,const void * data)4768 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4769 const void *data)
4770 {
4771 }
4772 #endif
4773
4774 static inline
skb_gso_segment(struct sk_buff * skb,netdev_features_t features)4775 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4776 {
4777 return __skb_gso_segment(skb, features, true);
4778 }
4779 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4780
can_checksum_protocol(netdev_features_t features,__be16 protocol)4781 static inline bool can_checksum_protocol(netdev_features_t features,
4782 __be16 protocol)
4783 {
4784 if (protocol == htons(ETH_P_FCOE))
4785 return !!(features & NETIF_F_FCOE_CRC);
4786
4787 /* Assume this is an IP checksum (not SCTP CRC) */
4788
4789 if (features & NETIF_F_HW_CSUM) {
4790 /* Can checksum everything */
4791 return true;
4792 }
4793
4794 switch (protocol) {
4795 case htons(ETH_P_IP):
4796 return !!(features & NETIF_F_IP_CSUM);
4797 case htons(ETH_P_IPV6):
4798 return !!(features & NETIF_F_IPV6_CSUM);
4799 default:
4800 return false;
4801 }
4802 }
4803
4804 #ifdef CONFIG_BUG
4805 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4806 #else
netdev_rx_csum_fault(struct net_device * dev,struct sk_buff * skb)4807 static inline void netdev_rx_csum_fault(struct net_device *dev,
4808 struct sk_buff *skb)
4809 {
4810 }
4811 #endif
4812 /* rx skb timestamps */
4813 void net_enable_timestamp(void);
4814 void net_disable_timestamp(void);
4815
4816 #ifdef CONFIG_PROC_FS
4817 int __init dev_proc_init(void);
4818 #else
4819 #define dev_proc_init() 0
4820 #endif
4821
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)4822 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4823 struct sk_buff *skb, struct net_device *dev,
4824 bool more)
4825 {
4826 __this_cpu_write(softnet_data.xmit.more, more);
4827 return ops->ndo_start_xmit(skb, dev);
4828 }
4829
netdev_xmit_more(void)4830 static inline bool netdev_xmit_more(void)
4831 {
4832 return __this_cpu_read(softnet_data.xmit.more);
4833 }
4834
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)4835 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4836 struct netdev_queue *txq, bool more)
4837 {
4838 const struct net_device_ops *ops = dev->netdev_ops;
4839 netdev_tx_t rc;
4840
4841 rc = __netdev_start_xmit(ops, skb, dev, more);
4842 if (rc == NETDEV_TX_OK)
4843 txq_trans_update(txq);
4844
4845 return rc;
4846 }
4847
4848 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4849 const void *ns);
4850 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4851 const void *ns);
4852
4853 extern const struct kobj_ns_type_operations net_ns_type_operations;
4854
4855 const char *netdev_drivername(const struct net_device *dev);
4856
4857 void linkwatch_run_queue(void);
4858
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)4859 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4860 netdev_features_t f2)
4861 {
4862 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4863 if (f1 & NETIF_F_HW_CSUM)
4864 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4865 else
4866 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4867 }
4868
4869 return f1 & f2;
4870 }
4871
netdev_get_wanted_features(struct net_device * dev)4872 static inline netdev_features_t netdev_get_wanted_features(
4873 struct net_device *dev)
4874 {
4875 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4876 }
4877 netdev_features_t netdev_increment_features(netdev_features_t all,
4878 netdev_features_t one, netdev_features_t mask);
4879
4880 /* Allow TSO being used on stacked device :
4881 * Performing the GSO segmentation before last device
4882 * is a performance improvement.
4883 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)4884 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4885 netdev_features_t mask)
4886 {
4887 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4888 }
4889
4890 int __netdev_update_features(struct net_device *dev);
4891 void netdev_update_features(struct net_device *dev);
4892 void netdev_change_features(struct net_device *dev);
4893
4894 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4895 struct net_device *dev);
4896
4897 netdev_features_t passthru_features_check(struct sk_buff *skb,
4898 struct net_device *dev,
4899 netdev_features_t features);
4900 netdev_features_t netif_skb_features(struct sk_buff *skb);
4901
net_gso_ok(netdev_features_t features,int gso_type)4902 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4903 {
4904 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4905
4906 /* check flags correspondence */
4907 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4908 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4909 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4910 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4911 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4912 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4913 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4914 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4915 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4916 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4917 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4918 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4919 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4920 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4921 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4922 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4923 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4924 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4925 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4926
4927 return (features & feature) == feature;
4928 }
4929
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)4930 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4931 {
4932 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4933 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4934 }
4935
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)4936 static inline bool netif_needs_gso(struct sk_buff *skb,
4937 netdev_features_t features)
4938 {
4939 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4940 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4941 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4942 }
4943
netif_set_gso_max_size(struct net_device * dev,unsigned int size)4944 static inline void netif_set_gso_max_size(struct net_device *dev,
4945 unsigned int size)
4946 {
4947 dev->gso_max_size = size;
4948 }
4949
skb_gso_error_unwind(struct sk_buff * skb,__be16 protocol,int pulled_hlen,u16 mac_offset,int mac_len)4950 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4951 int pulled_hlen, u16 mac_offset,
4952 int mac_len)
4953 {
4954 skb->protocol = protocol;
4955 skb->encapsulation = 1;
4956 skb_push(skb, pulled_hlen);
4957 skb_reset_transport_header(skb);
4958 skb->mac_header = mac_offset;
4959 skb->network_header = skb->mac_header + mac_len;
4960 skb->mac_len = mac_len;
4961 }
4962
netif_is_macsec(const struct net_device * dev)4963 static inline bool netif_is_macsec(const struct net_device *dev)
4964 {
4965 return dev->priv_flags & IFF_MACSEC;
4966 }
4967
netif_is_macvlan(const struct net_device * dev)4968 static inline bool netif_is_macvlan(const struct net_device *dev)
4969 {
4970 return dev->priv_flags & IFF_MACVLAN;
4971 }
4972
netif_is_macvlan_port(const struct net_device * dev)4973 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4974 {
4975 return dev->priv_flags & IFF_MACVLAN_PORT;
4976 }
4977
netif_is_bond_master(const struct net_device * dev)4978 static inline bool netif_is_bond_master(const struct net_device *dev)
4979 {
4980 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4981 }
4982
netif_is_bond_slave(const struct net_device * dev)4983 static inline bool netif_is_bond_slave(const struct net_device *dev)
4984 {
4985 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4986 }
4987
netif_supports_nofcs(struct net_device * dev)4988 static inline bool netif_supports_nofcs(struct net_device *dev)
4989 {
4990 return dev->priv_flags & IFF_SUPP_NOFCS;
4991 }
4992
netif_has_l3_rx_handler(const struct net_device * dev)4993 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4994 {
4995 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4996 }
4997
netif_is_l3_master(const struct net_device * dev)4998 static inline bool netif_is_l3_master(const struct net_device *dev)
4999 {
5000 return dev->priv_flags & IFF_L3MDEV_MASTER;
5001 }
5002
netif_is_l3_slave(const struct net_device * dev)5003 static inline bool netif_is_l3_slave(const struct net_device *dev)
5004 {
5005 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5006 }
5007
netif_is_bridge_master(const struct net_device * dev)5008 static inline bool netif_is_bridge_master(const struct net_device *dev)
5009 {
5010 return dev->priv_flags & IFF_EBRIDGE;
5011 }
5012
netif_is_bridge_port(const struct net_device * dev)5013 static inline bool netif_is_bridge_port(const struct net_device *dev)
5014 {
5015 return dev->priv_flags & IFF_BRIDGE_PORT;
5016 }
5017
netif_is_ovs_master(const struct net_device * dev)5018 static inline bool netif_is_ovs_master(const struct net_device *dev)
5019 {
5020 return dev->priv_flags & IFF_OPENVSWITCH;
5021 }
5022
netif_is_ovs_port(const struct net_device * dev)5023 static inline bool netif_is_ovs_port(const struct net_device *dev)
5024 {
5025 return dev->priv_flags & IFF_OVS_DATAPATH;
5026 }
5027
netif_is_any_bridge_port(const struct net_device * dev)5028 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5029 {
5030 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5031 }
5032
netif_is_team_master(const struct net_device * dev)5033 static inline bool netif_is_team_master(const struct net_device *dev)
5034 {
5035 return dev->priv_flags & IFF_TEAM;
5036 }
5037
netif_is_team_port(const struct net_device * dev)5038 static inline bool netif_is_team_port(const struct net_device *dev)
5039 {
5040 return dev->priv_flags & IFF_TEAM_PORT;
5041 }
5042
netif_is_lag_master(const struct net_device * dev)5043 static inline bool netif_is_lag_master(const struct net_device *dev)
5044 {
5045 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5046 }
5047
netif_is_lag_port(const struct net_device * dev)5048 static inline bool netif_is_lag_port(const struct net_device *dev)
5049 {
5050 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5051 }
5052
netif_is_rxfh_configured(const struct net_device * dev)5053 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5054 {
5055 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5056 }
5057
netif_is_failover(const struct net_device * dev)5058 static inline bool netif_is_failover(const struct net_device *dev)
5059 {
5060 return dev->priv_flags & IFF_FAILOVER;
5061 }
5062
netif_is_failover_slave(const struct net_device * dev)5063 static inline bool netif_is_failover_slave(const struct net_device *dev)
5064 {
5065 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5066 }
5067
5068 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)5069 static inline void netif_keep_dst(struct net_device *dev)
5070 {
5071 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5072 }
5073
5074 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
netif_reduces_vlan_mtu(struct net_device * dev)5075 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5076 {
5077 /* TODO: reserve and use an additional IFF bit, if we get more users */
5078 return dev->priv_flags & IFF_MACSEC;
5079 }
5080
5081 extern struct pernet_operations __net_initdata loopback_net_ops;
5082
5083 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5084
5085 /* netdev_printk helpers, similar to dev_printk */
5086
netdev_name(const struct net_device * dev)5087 static inline const char *netdev_name(const struct net_device *dev)
5088 {
5089 if (!dev->name[0] || strchr(dev->name, '%'))
5090 return "(unnamed net_device)";
5091 return dev->name;
5092 }
5093
netdev_unregistering(const struct net_device * dev)5094 static inline bool netdev_unregistering(const struct net_device *dev)
5095 {
5096 return dev->reg_state == NETREG_UNREGISTERING;
5097 }
5098
netdev_reg_state(const struct net_device * dev)5099 static inline const char *netdev_reg_state(const struct net_device *dev)
5100 {
5101 switch (dev->reg_state) {
5102 case NETREG_UNINITIALIZED: return " (uninitialized)";
5103 case NETREG_REGISTERED: return "";
5104 case NETREG_UNREGISTERING: return " (unregistering)";
5105 case NETREG_UNREGISTERED: return " (unregistered)";
5106 case NETREG_RELEASED: return " (released)";
5107 case NETREG_DUMMY: return " (dummy)";
5108 }
5109
5110 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5111 return " (unknown)";
5112 }
5113
5114 __printf(3, 4) __cold
5115 void netdev_printk(const char *level, const struct net_device *dev,
5116 const char *format, ...);
5117 __printf(2, 3) __cold
5118 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5119 __printf(2, 3) __cold
5120 void netdev_alert(const struct net_device *dev, const char *format, ...);
5121 __printf(2, 3) __cold
5122 void netdev_crit(const struct net_device *dev, const char *format, ...);
5123 __printf(2, 3) __cold
5124 void netdev_err(const struct net_device *dev, const char *format, ...);
5125 __printf(2, 3) __cold
5126 void netdev_warn(const struct net_device *dev, const char *format, ...);
5127 __printf(2, 3) __cold
5128 void netdev_notice(const struct net_device *dev, const char *format, ...);
5129 __printf(2, 3) __cold
5130 void netdev_info(const struct net_device *dev, const char *format, ...);
5131
5132 #define netdev_level_once(level, dev, fmt, ...) \
5133 do { \
5134 static bool __print_once __read_mostly; \
5135 \
5136 if (!__print_once) { \
5137 __print_once = true; \
5138 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5139 } \
5140 } while (0)
5141
5142 #define netdev_emerg_once(dev, fmt, ...) \
5143 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5144 #define netdev_alert_once(dev, fmt, ...) \
5145 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5146 #define netdev_crit_once(dev, fmt, ...) \
5147 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5148 #define netdev_err_once(dev, fmt, ...) \
5149 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5150 #define netdev_warn_once(dev, fmt, ...) \
5151 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5152 #define netdev_notice_once(dev, fmt, ...) \
5153 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5154 #define netdev_info_once(dev, fmt, ...) \
5155 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5156
5157 #define MODULE_ALIAS_NETDEV(device) \
5158 MODULE_ALIAS("netdev-" device)
5159
5160 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5161 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5162 #define netdev_dbg(__dev, format, args...) \
5163 do { \
5164 dynamic_netdev_dbg(__dev, format, ##args); \
5165 } while (0)
5166 #elif defined(DEBUG)
5167 #define netdev_dbg(__dev, format, args...) \
5168 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5169 #else
5170 #define netdev_dbg(__dev, format, args...) \
5171 ({ \
5172 if (0) \
5173 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5174 })
5175 #endif
5176
5177 #if defined(VERBOSE_DEBUG)
5178 #define netdev_vdbg netdev_dbg
5179 #else
5180
5181 #define netdev_vdbg(dev, format, args...) \
5182 ({ \
5183 if (0) \
5184 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5185 0; \
5186 })
5187 #endif
5188
5189 /*
5190 * netdev_WARN() acts like dev_printk(), but with the key difference
5191 * of using a WARN/WARN_ON to get the message out, including the
5192 * file/line information and a backtrace.
5193 */
5194 #define netdev_WARN(dev, format, args...) \
5195 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5196 netdev_reg_state(dev), ##args)
5197
5198 #define netdev_WARN_ONCE(dev, format, args...) \
5199 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5200 netdev_reg_state(dev), ##args)
5201
5202 /* netif printk helpers, similar to netdev_printk */
5203
5204 #define netif_printk(priv, type, level, dev, fmt, args...) \
5205 do { \
5206 if (netif_msg_##type(priv)) \
5207 netdev_printk(level, (dev), fmt, ##args); \
5208 } while (0)
5209
5210 #define netif_level(level, priv, type, dev, fmt, args...) \
5211 do { \
5212 if (netif_msg_##type(priv)) \
5213 netdev_##level(dev, fmt, ##args); \
5214 } while (0)
5215
5216 #define netif_emerg(priv, type, dev, fmt, args...) \
5217 netif_level(emerg, priv, type, dev, fmt, ##args)
5218 #define netif_alert(priv, type, dev, fmt, args...) \
5219 netif_level(alert, priv, type, dev, fmt, ##args)
5220 #define netif_crit(priv, type, dev, fmt, args...) \
5221 netif_level(crit, priv, type, dev, fmt, ##args)
5222 #define netif_err(priv, type, dev, fmt, args...) \
5223 netif_level(err, priv, type, dev, fmt, ##args)
5224 #define netif_warn(priv, type, dev, fmt, args...) \
5225 netif_level(warn, priv, type, dev, fmt, ##args)
5226 #define netif_notice(priv, type, dev, fmt, args...) \
5227 netif_level(notice, priv, type, dev, fmt, ##args)
5228 #define netif_info(priv, type, dev, fmt, args...) \
5229 netif_level(info, priv, type, dev, fmt, ##args)
5230
5231 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5232 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5233 #define netif_dbg(priv, type, netdev, format, args...) \
5234 do { \
5235 if (netif_msg_##type(priv)) \
5236 dynamic_netdev_dbg(netdev, format, ##args); \
5237 } while (0)
5238 #elif defined(DEBUG)
5239 #define netif_dbg(priv, type, dev, format, args...) \
5240 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5241 #else
5242 #define netif_dbg(priv, type, dev, format, args...) \
5243 ({ \
5244 if (0) \
5245 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5246 0; \
5247 })
5248 #endif
5249
5250 /* if @cond then downgrade to debug, else print at @level */
5251 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5252 do { \
5253 if (cond) \
5254 netif_dbg(priv, type, netdev, fmt, ##args); \
5255 else \
5256 netif_ ## level(priv, type, netdev, fmt, ##args); \
5257 } while (0)
5258
5259 #if defined(VERBOSE_DEBUG)
5260 #define netif_vdbg netif_dbg
5261 #else
5262 #define netif_vdbg(priv, type, dev, format, args...) \
5263 ({ \
5264 if (0) \
5265 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5266 0; \
5267 })
5268 #endif
5269
5270 /*
5271 * The list of packet types we will receive (as opposed to discard)
5272 * and the routines to invoke.
5273 *
5274 * Why 16. Because with 16 the only overlap we get on a hash of the
5275 * low nibble of the protocol value is RARP/SNAP/X.25.
5276 *
5277 * 0800 IP
5278 * 0001 802.3
5279 * 0002 AX.25
5280 * 0004 802.2
5281 * 8035 RARP
5282 * 0005 SNAP
5283 * 0805 X.25
5284 * 0806 ARP
5285 * 8137 IPX
5286 * 0009 Localtalk
5287 * 86DD IPv6
5288 */
5289 #define PTYPE_HASH_SIZE (16)
5290 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5291
5292 extern struct net_device *blackhole_netdev;
5293
5294 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5295 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5296 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5297 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5298 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5299
5300 #endif /* _LINUX_NETDEVICE_H */
5301