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 <net/net_namespace.h>
38 #ifdef CONFIG_DCB
39 #include <net/dcbnl.h>
40 #endif
41 #include <net/netprio_cgroup.h>
42 #include <net/xdp.h>
43
44 #include <linux/netdev_features.h>
45 #include <linux/neighbour.h>
46 #include <uapi/linux/netdevice.h>
47 #include <uapi/linux/if_bonding.h>
48 #include <uapi/linux/pkt_cls.h>
49 #include <linux/hashtable.h>
50 #include <linux/rbtree.h>
51 #include <linux/android_kabi.h>
52
53 struct netpoll_info;
54 struct device;
55 struct ethtool_ops;
56 struct phy_device;
57 struct dsa_port;
58 struct ip_tunnel_parm;
59 struct macsec_context;
60 struct macsec_ops;
61
62 struct sfp_bus;
63 /* 802.11 specific */
64 struct wireless_dev;
65 /* 802.15.4 specific */
66 struct wpan_dev;
67 struct mpls_dev;
68 /* UDP Tunnel offloads */
69 struct udp_tunnel_info;
70 struct udp_tunnel_nic_info;
71 struct udp_tunnel_nic;
72 struct bpf_prog;
73 struct xdp_buff;
74
75 void synchronize_net(void);
76 void netdev_set_default_ethtool_ops(struct net_device *dev,
77 const struct ethtool_ops *ops);
78
79 /* Backlog congestion levels */
80 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
81 #define NET_RX_DROP 1 /* packet dropped */
82
83 #define MAX_NEST_DEV 8
84
85 /*
86 * Transmit return codes: transmit return codes originate from three different
87 * namespaces:
88 *
89 * - qdisc return codes
90 * - driver transmit return codes
91 * - errno values
92 *
93 * Drivers are allowed to return any one of those in their hard_start_xmit()
94 * function. Real network devices commonly used with qdiscs should only return
95 * the driver transmit return codes though - when qdiscs are used, the actual
96 * transmission happens asynchronously, so the value is not propagated to
97 * higher layers. Virtual network devices transmit synchronously; in this case
98 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
99 * others are propagated to higher layers.
100 */
101
102 /* qdisc ->enqueue() return codes. */
103 #define NET_XMIT_SUCCESS 0x00
104 #define NET_XMIT_DROP 0x01 /* skb dropped */
105 #define NET_XMIT_CN 0x02 /* congestion notification */
106 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
107
108 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
109 * indicates that the device will soon be dropping packets, or already drops
110 * some packets of the same priority; prompting us to send less aggressively. */
111 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
112 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
113
114 /* Driver transmit return codes */
115 #define NETDEV_TX_MASK 0xf0
116
117 enum netdev_tx {
118 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
119 NETDEV_TX_OK = 0x00, /* driver took care of packet */
120 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
121 };
122 typedef enum netdev_tx netdev_tx_t;
123
124 /*
125 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
126 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
127 */
dev_xmit_complete(int rc)128 static inline bool dev_xmit_complete(int rc)
129 {
130 /*
131 * Positive cases with an skb consumed by a driver:
132 * - successful transmission (rc == NETDEV_TX_OK)
133 * - error while transmitting (rc < 0)
134 * - error while queueing to a different device (rc & NET_XMIT_MASK)
135 */
136 if (likely(rc < NET_XMIT_MASK))
137 return true;
138
139 return false;
140 }
141
142 /*
143 * Compute the worst-case header length according to the protocols
144 * used.
145 */
146
147 #if defined(CONFIG_HYPERV_NET)
148 # define LL_MAX_HEADER 128
149 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
150 # if defined(CONFIG_MAC80211_MESH)
151 # define LL_MAX_HEADER 128
152 # else
153 # define LL_MAX_HEADER 96
154 # endif
155 #else
156 # define LL_MAX_HEADER 32
157 #endif
158
159 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
160 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
161 #define MAX_HEADER LL_MAX_HEADER
162 #else
163 #define MAX_HEADER (LL_MAX_HEADER + 48)
164 #endif
165
166 /*
167 * Old network device statistics. Fields are native words
168 * (unsigned long) so they can be read and written atomically.
169 */
170
171 #define NET_DEV_STAT(FIELD) \
172 union { \
173 unsigned long FIELD; \
174 atomic_long_t __##FIELD; \
175 }
176
177 struct net_device_stats {
178 NET_DEV_STAT(rx_packets);
179 NET_DEV_STAT(tx_packets);
180 NET_DEV_STAT(rx_bytes);
181 NET_DEV_STAT(tx_bytes);
182 NET_DEV_STAT(rx_errors);
183 NET_DEV_STAT(tx_errors);
184 NET_DEV_STAT(rx_dropped);
185 NET_DEV_STAT(tx_dropped);
186 NET_DEV_STAT(multicast);
187 NET_DEV_STAT(collisions);
188 NET_DEV_STAT(rx_length_errors);
189 NET_DEV_STAT(rx_over_errors);
190 NET_DEV_STAT(rx_crc_errors);
191 NET_DEV_STAT(rx_frame_errors);
192 NET_DEV_STAT(rx_fifo_errors);
193 NET_DEV_STAT(rx_missed_errors);
194 NET_DEV_STAT(tx_aborted_errors);
195 NET_DEV_STAT(tx_carrier_errors);
196 NET_DEV_STAT(tx_fifo_errors);
197 NET_DEV_STAT(tx_heartbeat_errors);
198 NET_DEV_STAT(tx_window_errors);
199 NET_DEV_STAT(rx_compressed);
200 NET_DEV_STAT(tx_compressed);
201 };
202 #undef NET_DEV_STAT
203
204
205 #include <linux/cache.h>
206 #include <linux/skbuff.h>
207
208 #ifdef CONFIG_RPS
209 #include <linux/static_key.h>
210 extern struct static_key_false rps_needed;
211 extern struct static_key_false rfs_needed;
212 #endif
213
214 struct neighbour;
215 struct neigh_parms;
216 struct sk_buff;
217
218 struct netdev_hw_addr {
219 struct list_head list;
220 struct rb_node node;
221 unsigned char addr[MAX_ADDR_LEN];
222 unsigned char type;
223 #define NETDEV_HW_ADDR_T_LAN 1
224 #define NETDEV_HW_ADDR_T_SAN 2
225 #define NETDEV_HW_ADDR_T_UNICAST 3
226 #define NETDEV_HW_ADDR_T_MULTICAST 4
227 bool global_use;
228 int sync_cnt;
229 int refcount;
230 int synced;
231 struct rcu_head rcu_head;
232 };
233
234 struct netdev_hw_addr_list {
235 struct list_head list;
236 int count;
237
238 /* Auxiliary tree for faster lookup on addition and deletion */
239 struct rb_root tree;
240 };
241
242 #define netdev_hw_addr_list_count(l) ((l)->count)
243 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
244 #define netdev_hw_addr_list_for_each(ha, l) \
245 list_for_each_entry(ha, &(l)->list, list)
246
247 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
248 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
249 #define netdev_for_each_uc_addr(ha, dev) \
250 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
251
252 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
253 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
254 #define netdev_for_each_mc_addr(ha, dev) \
255 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
256
257 struct hh_cache {
258 unsigned int hh_len;
259 seqlock_t hh_lock;
260
261 /* cached hardware header; allow for machine alignment needs. */
262 #define HH_DATA_MOD 16
263 #define HH_DATA_OFF(__len) \
264 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
265 #define HH_DATA_ALIGN(__len) \
266 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
267 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
268 };
269
270 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
271 * Alternative is:
272 * dev->hard_header_len ? (dev->hard_header_len +
273 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
274 *
275 * We could use other alignment values, but we must maintain the
276 * relationship HH alignment <= LL alignment.
277 */
278 #define LL_RESERVED_SPACE(dev) \
279 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
280 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
281 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
282 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
283 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
284
285 struct header_ops {
286 int (*create) (struct sk_buff *skb, struct net_device *dev,
287 unsigned short type, const void *daddr,
288 const void *saddr, unsigned int len);
289 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
290 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
291 void (*cache_update)(struct hh_cache *hh,
292 const struct net_device *dev,
293 const unsigned char *haddr);
294 bool (*validate)(const char *ll_header, unsigned int len);
295 __be16 (*parse_protocol)(const struct sk_buff *skb);
296
297 ANDROID_KABI_RESERVE(1);
298 ANDROID_KABI_RESERVE(2);
299 };
300
301 /* These flag bits are private to the generic network queueing
302 * layer; they may not be explicitly referenced by any other
303 * code.
304 */
305
306 enum netdev_state_t {
307 __LINK_STATE_START,
308 __LINK_STATE_PRESENT,
309 __LINK_STATE_NOCARRIER,
310 __LINK_STATE_LINKWATCH_PENDING,
311 __LINK_STATE_DORMANT,
312 __LINK_STATE_TESTING,
313 };
314
315
316 struct gro_list {
317 struct list_head list;
318 int count;
319 };
320
321 /*
322 * size of gro hash buckets, must less than bit number of
323 * napi_struct::gro_bitmask
324 */
325 #define GRO_HASH_BUCKETS 8
326
327 /*
328 * Structure for NAPI scheduling similar to tasklet but with weighting
329 */
330 struct napi_struct {
331 /* The poll_list must only be managed by the entity which
332 * changes the state of the NAPI_STATE_SCHED bit. This means
333 * whoever atomically sets that bit can add this napi_struct
334 * to the per-CPU poll_list, and whoever clears that bit
335 * can remove from the list right before clearing the bit.
336 */
337 struct list_head poll_list;
338
339 unsigned long state;
340 int weight;
341 int defer_hard_irqs_count;
342 unsigned long gro_bitmask;
343 int (*poll)(struct napi_struct *, int);
344 #ifdef CONFIG_NETPOLL
345 int poll_owner;
346 #endif
347 struct net_device *dev;
348 struct gro_list gro_hash[GRO_HASH_BUCKETS];
349 struct sk_buff *skb;
350 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
351 int rx_count; /* length of rx_list */
352 struct hrtimer timer;
353 struct list_head dev_list;
354 struct hlist_node napi_hash_node;
355 unsigned int napi_id;
356 struct task_struct *thread;
357
358 ANDROID_KABI_RESERVE(1);
359 ANDROID_KABI_RESERVE(2);
360 ANDROID_KABI_RESERVE(3);
361 ANDROID_KABI_RESERVE(4);
362 };
363
364 enum {
365 NAPI_STATE_SCHED, /* Poll is scheduled */
366 NAPI_STATE_MISSED, /* reschedule a napi */
367 NAPI_STATE_DISABLE, /* Disable pending */
368 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
369 NAPI_STATE_LISTED, /* NAPI added to system lists */
370 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
371 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
372 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
373 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
374 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
375 };
376
377 enum {
378 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
379 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
380 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
381 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
382 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
383 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
384 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
385 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
386 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
387 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
388 };
389
390 enum gro_result {
391 GRO_MERGED,
392 GRO_MERGED_FREE,
393 GRO_HELD,
394 GRO_NORMAL,
395 GRO_CONSUMED,
396 };
397 typedef enum gro_result gro_result_t;
398
399 /*
400 * enum rx_handler_result - Possible return values for rx_handlers.
401 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
402 * further.
403 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
404 * case skb->dev was changed by rx_handler.
405 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
406 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
407 *
408 * rx_handlers are functions called from inside __netif_receive_skb(), to do
409 * special processing of the skb, prior to delivery to protocol handlers.
410 *
411 * Currently, a net_device can only have a single rx_handler registered. Trying
412 * to register a second rx_handler will return -EBUSY.
413 *
414 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
415 * To unregister a rx_handler on a net_device, use
416 * netdev_rx_handler_unregister().
417 *
418 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
419 * do with the skb.
420 *
421 * If the rx_handler consumed the skb in some way, it should return
422 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
423 * the skb to be delivered in some other way.
424 *
425 * If the rx_handler changed skb->dev, to divert the skb to another
426 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
427 * new device will be called if it exists.
428 *
429 * If the rx_handler decides the skb should be ignored, it should return
430 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
431 * are registered on exact device (ptype->dev == skb->dev).
432 *
433 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
434 * delivered, it should return RX_HANDLER_PASS.
435 *
436 * A device without a registered rx_handler will behave as if rx_handler
437 * returned RX_HANDLER_PASS.
438 */
439
440 enum rx_handler_result {
441 RX_HANDLER_CONSUMED,
442 RX_HANDLER_ANOTHER,
443 RX_HANDLER_EXACT,
444 RX_HANDLER_PASS,
445 };
446 typedef enum rx_handler_result rx_handler_result_t;
447 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
448
449 void __napi_schedule(struct napi_struct *n);
450 void __napi_schedule_irqoff(struct napi_struct *n);
451
napi_disable_pending(struct napi_struct * n)452 static inline bool napi_disable_pending(struct napi_struct *n)
453 {
454 return test_bit(NAPI_STATE_DISABLE, &n->state);
455 }
456
napi_prefer_busy_poll(struct napi_struct * n)457 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
458 {
459 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
460 }
461
462 bool napi_schedule_prep(struct napi_struct *n);
463
464 /**
465 * napi_schedule - schedule NAPI poll
466 * @n: NAPI context
467 *
468 * Schedule NAPI poll routine to be called if it is not already
469 * running.
470 */
napi_schedule(struct napi_struct * n)471 static inline void napi_schedule(struct napi_struct *n)
472 {
473 if (napi_schedule_prep(n))
474 __napi_schedule(n);
475 }
476
477 /**
478 * napi_schedule_irqoff - schedule NAPI poll
479 * @n: NAPI context
480 *
481 * Variant of napi_schedule(), assuming hard irqs are masked.
482 */
napi_schedule_irqoff(struct napi_struct * n)483 static inline void napi_schedule_irqoff(struct napi_struct *n)
484 {
485 if (napi_schedule_prep(n))
486 __napi_schedule_irqoff(n);
487 }
488
489 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
napi_reschedule(struct napi_struct * napi)490 static inline bool napi_reschedule(struct napi_struct *napi)
491 {
492 if (napi_schedule_prep(napi)) {
493 __napi_schedule(napi);
494 return true;
495 }
496 return false;
497 }
498
499 bool napi_complete_done(struct napi_struct *n, int work_done);
500 /**
501 * napi_complete - NAPI processing complete
502 * @n: NAPI context
503 *
504 * Mark NAPI processing as complete.
505 * Consider using napi_complete_done() instead.
506 * Return false if device should avoid rearming interrupts.
507 */
napi_complete(struct napi_struct * n)508 static inline bool napi_complete(struct napi_struct *n)
509 {
510 return napi_complete_done(n, 0);
511 }
512
513 int dev_set_threaded(struct net_device *dev, bool threaded);
514
515 /**
516 * napi_disable - prevent NAPI from scheduling
517 * @n: NAPI context
518 *
519 * Stop NAPI from being scheduled on this context.
520 * Waits till any outstanding processing completes.
521 */
522 void napi_disable(struct napi_struct *n);
523
524 void napi_enable(struct napi_struct *n);
525
526 /**
527 * napi_synchronize - wait until NAPI is not running
528 * @n: NAPI context
529 *
530 * Wait until NAPI is done being scheduled on this context.
531 * Waits till any outstanding processing completes but
532 * does not disable future activations.
533 */
napi_synchronize(const struct napi_struct * n)534 static inline void napi_synchronize(const struct napi_struct *n)
535 {
536 if (IS_ENABLED(CONFIG_SMP))
537 while (test_bit(NAPI_STATE_SCHED, &n->state))
538 msleep(1);
539 else
540 barrier();
541 }
542
543 /**
544 * napi_if_scheduled_mark_missed - if napi is running, set the
545 * NAPIF_STATE_MISSED
546 * @n: NAPI context
547 *
548 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
549 * NAPI is scheduled.
550 **/
napi_if_scheduled_mark_missed(struct napi_struct * n)551 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
552 {
553 unsigned long val, new;
554
555 do {
556 val = READ_ONCE(n->state);
557 if (val & NAPIF_STATE_DISABLE)
558 return true;
559
560 if (!(val & NAPIF_STATE_SCHED))
561 return false;
562
563 new = val | NAPIF_STATE_MISSED;
564 } while (cmpxchg(&n->state, val, new) != val);
565
566 return true;
567 }
568
569 enum netdev_queue_state_t {
570 __QUEUE_STATE_DRV_XOFF,
571 __QUEUE_STATE_STACK_XOFF,
572 __QUEUE_STATE_FROZEN,
573 };
574
575 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
576 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
577 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
578
579 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
580 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
581 QUEUE_STATE_FROZEN)
582 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
583 QUEUE_STATE_FROZEN)
584
585 /*
586 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
587 * netif_tx_* functions below are used to manipulate this flag. The
588 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
589 * queue independently. The netif_xmit_*stopped functions below are called
590 * to check if the queue has been stopped by the driver or stack (either
591 * of the XOFF bits are set in the state). Drivers should not need to call
592 * netif_xmit*stopped functions, they should only be using netif_tx_*.
593 */
594
595 struct netdev_queue {
596 /*
597 * read-mostly part
598 */
599 struct net_device *dev;
600 struct Qdisc __rcu *qdisc;
601 struct Qdisc *qdisc_sleeping;
602 #ifdef CONFIG_SYSFS
603 struct kobject kobj;
604 #endif
605 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 int numa_node;
607 #endif
608 unsigned long tx_maxrate;
609 /*
610 * Number of TX timeouts for this queue
611 * (/sys/class/net/DEV/Q/trans_timeout)
612 */
613 unsigned long trans_timeout;
614
615 /* Subordinate device that the queue has been assigned to */
616 struct net_device *sb_dev;
617 #ifdef CONFIG_XDP_SOCKETS
618 struct xsk_buff_pool *pool;
619 #endif
620 /*
621 * write-mostly part
622 */
623 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
624 int xmit_lock_owner;
625 /*
626 * Time (in jiffies) of last Tx
627 */
628 unsigned long trans_start;
629
630 unsigned long state;
631
632 #ifdef CONFIG_BQL
633 struct dql dql;
634 #endif
635
636 ANDROID_KABI_RESERVE(1);
637 ANDROID_KABI_RESERVE(2);
638 ANDROID_KABI_RESERVE(3);
639 ANDROID_KABI_RESERVE(4);
640 } ____cacheline_aligned_in_smp;
641
642 extern int sysctl_fb_tunnels_only_for_init_net;
643 extern int sysctl_devconf_inherit_init_net;
644
645 /*
646 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
647 * == 1 : For initns only
648 * == 2 : For none.
649 */
net_has_fallback_tunnels(const struct net * net)650 static inline bool net_has_fallback_tunnels(const struct net *net)
651 {
652 #if IS_ENABLED(CONFIG_SYSCTL)
653 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
654
655 return !fb_tunnels_only_for_init_net ||
656 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
657 #else
658 return true;
659 #endif
660 }
661
net_inherit_devconf(void)662 static inline int net_inherit_devconf(void)
663 {
664 #if IS_ENABLED(CONFIG_SYSCTL)
665 return READ_ONCE(sysctl_devconf_inherit_init_net);
666 #else
667 return 0;
668 #endif
669 }
670
netdev_queue_numa_node_read(const struct netdev_queue * q)671 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
672 {
673 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
674 return q->numa_node;
675 #else
676 return NUMA_NO_NODE;
677 #endif
678 }
679
netdev_queue_numa_node_write(struct netdev_queue * q,int node)680 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
681 {
682 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
683 q->numa_node = node;
684 #endif
685 }
686
687 #ifdef CONFIG_RPS
688 /*
689 * This structure holds an RPS map which can be of variable length. The
690 * map is an array of CPUs.
691 */
692 struct rps_map {
693 unsigned int len;
694 struct rcu_head rcu;
695 u16 cpus[];
696 };
697 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
698
699 /*
700 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
701 * tail pointer for that CPU's input queue at the time of last enqueue, and
702 * a hardware filter index.
703 */
704 struct rps_dev_flow {
705 u16 cpu;
706 u16 filter;
707 unsigned int last_qtail;
708 };
709 #define RPS_NO_FILTER 0xffff
710
711 /*
712 * The rps_dev_flow_table structure contains a table of flow mappings.
713 */
714 struct rps_dev_flow_table {
715 unsigned int mask;
716 struct rcu_head rcu;
717 struct rps_dev_flow flows[];
718 };
719 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
720 ((_num) * sizeof(struct rps_dev_flow)))
721
722 /*
723 * The rps_sock_flow_table contains mappings of flows to the last CPU
724 * on which they were processed by the application (set in recvmsg).
725 * Each entry is a 32bit value. Upper part is the high-order bits
726 * of flow hash, lower part is CPU number.
727 * rps_cpu_mask is used to partition the space, depending on number of
728 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
729 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
730 * meaning we use 32-6=26 bits for the hash.
731 */
732 struct rps_sock_flow_table {
733 u32 mask;
734
735 u32 ents[] ____cacheline_aligned_in_smp;
736 };
737 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
738
739 #define RPS_NO_CPU 0xffff
740
741 extern u32 rps_cpu_mask;
742 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
743
rps_record_sock_flow(struct rps_sock_flow_table * table,u32 hash)744 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
745 u32 hash)
746 {
747 if (table && hash) {
748 unsigned int index = hash & table->mask;
749 u32 val = hash & ~rps_cpu_mask;
750
751 /* We only give a hint, preemption can change CPU under us */
752 val |= raw_smp_processor_id();
753
754 /* The following WRITE_ONCE() is paired with the READ_ONCE()
755 * here, and another one in get_rps_cpu().
756 */
757 if (READ_ONCE(table->ents[index]) != val)
758 WRITE_ONCE(table->ents[index], val);
759 }
760 }
761
762 #ifdef CONFIG_RFS_ACCEL
763 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
764 u16 filter_id);
765 #endif
766 #endif /* CONFIG_RPS */
767
768 /* This structure contains an instance of an RX queue. */
769 struct netdev_rx_queue {
770 struct xdp_rxq_info xdp_rxq;
771 #ifdef CONFIG_RPS
772 struct rps_map __rcu *rps_map;
773 struct rps_dev_flow_table __rcu *rps_flow_table;
774 #endif
775 struct kobject kobj;
776 struct net_device *dev;
777 #ifdef CONFIG_XDP_SOCKETS
778 struct xsk_buff_pool *pool;
779 #endif
780
781 ANDROID_KABI_RESERVE(1);
782 ANDROID_KABI_RESERVE(2);
783 ANDROID_KABI_RESERVE(3);
784 ANDROID_KABI_RESERVE(4);
785 } ____cacheline_aligned_in_smp;
786
787 /*
788 * RX queue sysfs structures and functions.
789 */
790 struct rx_queue_attribute {
791 struct attribute attr;
792 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
793 ssize_t (*store)(struct netdev_rx_queue *queue,
794 const char *buf, size_t len);
795 };
796
797 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
798 enum xps_map_type {
799 XPS_CPUS = 0,
800 XPS_RXQS,
801 XPS_MAPS_MAX,
802 };
803
804 #ifdef CONFIG_XPS
805 /*
806 * This structure holds an XPS map which can be of variable length. The
807 * map is an array of queues.
808 */
809 struct xps_map {
810 unsigned int len;
811 unsigned int alloc_len;
812 struct rcu_head rcu;
813 u16 queues[];
814 };
815 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
816 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
817 - sizeof(struct xps_map)) / sizeof(u16))
818
819 /*
820 * This structure holds all XPS maps for device. Maps are indexed by CPU.
821 *
822 * We keep track of the number of cpus/rxqs used when the struct is allocated,
823 * in nr_ids. This will help not accessing out-of-bound memory.
824 *
825 * We keep track of the number of traffic classes used when the struct is
826 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
827 * not crossing its upper bound, as the original dev->num_tc can be updated in
828 * the meantime.
829 */
830 struct xps_dev_maps {
831 struct rcu_head rcu;
832 unsigned int nr_ids;
833 s16 num_tc;
834 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
835 };
836
837 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
838 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
839
840 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
841 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
842
843 #endif /* CONFIG_XPS */
844
845 #define TC_MAX_QUEUE 16
846 #define TC_BITMASK 15
847 /* HW offloaded queuing disciplines txq count and offset maps */
848 struct netdev_tc_txq {
849 u16 count;
850 u16 offset;
851 };
852
853 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
854 /*
855 * This structure is to hold information about the device
856 * configured to run FCoE protocol stack.
857 */
858 struct netdev_fcoe_hbainfo {
859 char manufacturer[64];
860 char serial_number[64];
861 char hardware_version[64];
862 char driver_version[64];
863 char optionrom_version[64];
864 char firmware_version[64];
865 char model[256];
866 char model_description[256];
867 };
868 #endif
869
870 #define MAX_PHYS_ITEM_ID_LEN 32
871
872 /* This structure holds a unique identifier to identify some
873 * physical item (port for example) used by a netdevice.
874 */
875 struct netdev_phys_item_id {
876 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
877 unsigned char id_len;
878 };
879
netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)880 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
881 struct netdev_phys_item_id *b)
882 {
883 return a->id_len == b->id_len &&
884 memcmp(a->id, b->id, a->id_len) == 0;
885 }
886
887 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
888 struct sk_buff *skb,
889 struct net_device *sb_dev);
890
891 enum net_device_path_type {
892 DEV_PATH_ETHERNET = 0,
893 DEV_PATH_VLAN,
894 DEV_PATH_BRIDGE,
895 DEV_PATH_PPPOE,
896 DEV_PATH_DSA,
897 };
898
899 struct net_device_path {
900 enum net_device_path_type type;
901 const struct net_device *dev;
902 union {
903 struct {
904 u16 id;
905 __be16 proto;
906 u8 h_dest[ETH_ALEN];
907 } encap;
908 struct {
909 enum {
910 DEV_PATH_BR_VLAN_KEEP,
911 DEV_PATH_BR_VLAN_TAG,
912 DEV_PATH_BR_VLAN_UNTAG,
913 DEV_PATH_BR_VLAN_UNTAG_HW,
914 } vlan_mode;
915 u16 vlan_id;
916 __be16 vlan_proto;
917 } bridge;
918 struct {
919 int port;
920 u16 proto;
921 } dsa;
922 };
923 };
924
925 #define NET_DEVICE_PATH_STACK_MAX 5
926 #define NET_DEVICE_PATH_VLAN_MAX 2
927
928 struct net_device_path_stack {
929 int num_paths;
930 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
931 };
932
933 struct net_device_path_ctx {
934 const struct net_device *dev;
935 u8 daddr[ETH_ALEN];
936
937 int num_vlans;
938 struct {
939 u16 id;
940 __be16 proto;
941 } vlan[NET_DEVICE_PATH_VLAN_MAX];
942 };
943
944 enum tc_setup_type {
945 TC_SETUP_QDISC_MQPRIO,
946 TC_SETUP_CLSU32,
947 TC_SETUP_CLSFLOWER,
948 TC_SETUP_CLSMATCHALL,
949 TC_SETUP_CLSBPF,
950 TC_SETUP_BLOCK,
951 TC_SETUP_QDISC_CBS,
952 TC_SETUP_QDISC_RED,
953 TC_SETUP_QDISC_PRIO,
954 TC_SETUP_QDISC_MQ,
955 TC_SETUP_QDISC_ETF,
956 TC_SETUP_ROOT_QDISC,
957 TC_SETUP_QDISC_GRED,
958 TC_SETUP_QDISC_TAPRIO,
959 TC_SETUP_FT,
960 TC_SETUP_QDISC_ETS,
961 TC_SETUP_QDISC_TBF,
962 TC_SETUP_QDISC_FIFO,
963 TC_SETUP_QDISC_HTB,
964 };
965
966 /* These structures hold the attributes of bpf state that are being passed
967 * to the netdevice through the bpf op.
968 */
969 enum bpf_netdev_command {
970 /* Set or clear a bpf program used in the earliest stages of packet
971 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
972 * is responsible for calling bpf_prog_put on any old progs that are
973 * stored. In case of error, the callee need not release the new prog
974 * reference, but on success it takes ownership and must bpf_prog_put
975 * when it is no longer used.
976 */
977 XDP_SETUP_PROG,
978 XDP_SETUP_PROG_HW,
979 /* BPF program for offload callbacks, invoked at program load time. */
980 BPF_OFFLOAD_MAP_ALLOC,
981 BPF_OFFLOAD_MAP_FREE,
982 XDP_SETUP_XSK_POOL,
983 };
984
985 struct bpf_prog_offload_ops;
986 struct netlink_ext_ack;
987 struct xdp_umem;
988 struct xdp_dev_bulk_queue;
989 struct bpf_xdp_link;
990
991 enum bpf_xdp_mode {
992 XDP_MODE_SKB = 0,
993 XDP_MODE_DRV = 1,
994 XDP_MODE_HW = 2,
995 __MAX_XDP_MODE
996 };
997
998 struct bpf_xdp_entity {
999 struct bpf_prog *prog;
1000 struct bpf_xdp_link *link;
1001 };
1002
1003 struct netdev_bpf {
1004 enum bpf_netdev_command command;
1005 union {
1006 /* XDP_SETUP_PROG */
1007 struct {
1008 u32 flags;
1009 struct bpf_prog *prog;
1010 struct netlink_ext_ack *extack;
1011 };
1012 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1013 struct {
1014 struct bpf_offloaded_map *offmap;
1015 };
1016 /* XDP_SETUP_XSK_POOL */
1017 struct {
1018 struct xsk_buff_pool *pool;
1019 u16 queue_id;
1020 } xsk;
1021 };
1022 };
1023
1024 /* Flags for ndo_xsk_wakeup. */
1025 #define XDP_WAKEUP_RX (1 << 0)
1026 #define XDP_WAKEUP_TX (1 << 1)
1027
1028 #ifdef CONFIG_XFRM_OFFLOAD
1029 struct xfrmdev_ops {
1030 int (*xdo_dev_state_add) (struct xfrm_state *x);
1031 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1032 void (*xdo_dev_state_free) (struct xfrm_state *x);
1033 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1034 struct xfrm_state *x);
1035 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1036
1037 ANDROID_KABI_RESERVE(1);
1038 ANDROID_KABI_RESERVE(2);
1039 ANDROID_KABI_RESERVE(3);
1040 ANDROID_KABI_RESERVE(4);
1041 };
1042 #endif
1043
1044 struct dev_ifalias {
1045 struct rcu_head rcuhead;
1046 char ifalias[];
1047 };
1048
1049 struct devlink;
1050 struct tlsdev_ops;
1051
1052 struct netdev_name_node {
1053 struct hlist_node hlist;
1054 struct list_head list;
1055 struct net_device *dev;
1056 const char *name;
1057 };
1058
1059 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
1060 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
1061
1062 struct netdev_net_notifier {
1063 struct list_head list;
1064 struct notifier_block *nb;
1065 };
1066
1067 /*
1068 * This structure defines the management hooks for network devices.
1069 * The following hooks can be defined; unless noted otherwise, they are
1070 * optional and can be filled with a null pointer.
1071 *
1072 * int (*ndo_init)(struct net_device *dev);
1073 * This function is called once when a network device is registered.
1074 * The network device can use this for any late stage initialization
1075 * or semantic validation. It can fail with an error code which will
1076 * be propagated back to register_netdev.
1077 *
1078 * void (*ndo_uninit)(struct net_device *dev);
1079 * This function is called when device is unregistered or when registration
1080 * fails. It is not called if init fails.
1081 *
1082 * int (*ndo_open)(struct net_device *dev);
1083 * This function is called when a network device transitions to the up
1084 * state.
1085 *
1086 * int (*ndo_stop)(struct net_device *dev);
1087 * This function is called when a network device transitions to the down
1088 * state.
1089 *
1090 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1091 * struct net_device *dev);
1092 * Called when a packet needs to be transmitted.
1093 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1094 * the queue before that can happen; it's for obsolete devices and weird
1095 * corner cases, but the stack really does a non-trivial amount
1096 * of useless work if you return NETDEV_TX_BUSY.
1097 * Required; cannot be NULL.
1098 *
1099 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1100 * struct net_device *dev
1101 * netdev_features_t features);
1102 * Called by core transmit path to determine if device is capable of
1103 * performing offload operations on a given packet. This is to give
1104 * the device an opportunity to implement any restrictions that cannot
1105 * be otherwise expressed by feature flags. The check is called with
1106 * the set of features that the stack has calculated and it returns
1107 * those the driver believes to be appropriate.
1108 *
1109 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1110 * struct net_device *sb_dev);
1111 * Called to decide which queue to use when device supports multiple
1112 * transmit queues.
1113 *
1114 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1115 * This function is called to allow device receiver to make
1116 * changes to configuration when multicast or promiscuous is enabled.
1117 *
1118 * void (*ndo_set_rx_mode)(struct net_device *dev);
1119 * This function is called device changes address list filtering.
1120 * If driver handles unicast address filtering, it should set
1121 * IFF_UNICAST_FLT in its priv_flags.
1122 *
1123 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1124 * This function is called when the Media Access Control address
1125 * needs to be changed. If this interface is not defined, the
1126 * MAC address can not be changed.
1127 *
1128 * int (*ndo_validate_addr)(struct net_device *dev);
1129 * Test if Media Access Control address is valid for the device.
1130 *
1131 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1132 * Old-style ioctl entry point. This is used internally by the
1133 * appletalk and ieee802154 subsystems but is no longer called by
1134 * the device ioctl handler.
1135 *
1136 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1137 * Used by the bonding driver for its device specific ioctls:
1138 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1139 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1140 *
1141 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1142 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1143 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1144 *
1145 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1146 * Used to set network devices bus interface parameters. This interface
1147 * is retained for legacy reasons; new devices should use the bus
1148 * interface (PCI) for low level management.
1149 *
1150 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1151 * Called when a user wants to change the Maximum Transfer Unit
1152 * of a device.
1153 *
1154 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1155 * Callback used when the transmitter has not made any progress
1156 * for dev->watchdog ticks.
1157 *
1158 * void (*ndo_get_stats64)(struct net_device *dev,
1159 * struct rtnl_link_stats64 *storage);
1160 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1161 * Called when a user wants to get the network device usage
1162 * statistics. Drivers must do one of the following:
1163 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1164 * rtnl_link_stats64 structure passed by the caller.
1165 * 2. Define @ndo_get_stats to update a net_device_stats structure
1166 * (which should normally be dev->stats) and return a pointer to
1167 * it. The structure may be changed asynchronously only if each
1168 * field is written atomically.
1169 * 3. Update dev->stats asynchronously and atomically, and define
1170 * neither operation.
1171 *
1172 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1173 * Return true if this device supports offload stats of this attr_id.
1174 *
1175 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1176 * void *attr_data)
1177 * Get statistics for offload operations by attr_id. Write it into the
1178 * attr_data pointer.
1179 *
1180 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1181 * If device supports VLAN filtering this function is called when a
1182 * VLAN id is registered.
1183 *
1184 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1185 * If device supports VLAN filtering this function is called when a
1186 * VLAN id is unregistered.
1187 *
1188 * void (*ndo_poll_controller)(struct net_device *dev);
1189 *
1190 * SR-IOV management functions.
1191 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1192 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1193 * u8 qos, __be16 proto);
1194 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1195 * int max_tx_rate);
1196 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1197 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1198 * int (*ndo_get_vf_config)(struct net_device *dev,
1199 * int vf, struct ifla_vf_info *ivf);
1200 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1201 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1202 * struct nlattr *port[]);
1203 *
1204 * Enable or disable the VF ability to query its RSS Redirection Table and
1205 * Hash Key. This is needed since on some devices VF share this information
1206 * with PF and querying it may introduce a theoretical security risk.
1207 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1208 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1209 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1210 * void *type_data);
1211 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1212 * This is always called from the stack with the rtnl lock held and netif
1213 * tx queues stopped. This allows the netdevice to perform queue
1214 * management safely.
1215 *
1216 * Fiber Channel over Ethernet (FCoE) offload functions.
1217 * int (*ndo_fcoe_enable)(struct net_device *dev);
1218 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1219 * so the underlying device can perform whatever needed configuration or
1220 * initialization to support acceleration of FCoE traffic.
1221 *
1222 * int (*ndo_fcoe_disable)(struct net_device *dev);
1223 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1224 * so the underlying device can perform whatever needed clean-ups to
1225 * stop supporting acceleration of FCoE traffic.
1226 *
1227 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1228 * struct scatterlist *sgl, unsigned int sgc);
1229 * Called when the FCoE Initiator wants to initialize an I/O that
1230 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1231 * perform necessary setup and returns 1 to indicate the device is set up
1232 * successfully to perform DDP on this I/O, otherwise this returns 0.
1233 *
1234 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1235 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1236 * indicated by the FC exchange id 'xid', so the underlying device can
1237 * clean up and reuse resources for later DDP requests.
1238 *
1239 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1240 * struct scatterlist *sgl, unsigned int sgc);
1241 * Called when the FCoE Target wants to initialize an I/O that
1242 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1243 * perform necessary setup and returns 1 to indicate the device is set up
1244 * successfully to perform DDP on this I/O, otherwise this returns 0.
1245 *
1246 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1247 * struct netdev_fcoe_hbainfo *hbainfo);
1248 * Called when the FCoE Protocol stack wants information on the underlying
1249 * device. This information is utilized by the FCoE protocol stack to
1250 * register attributes with Fiber Channel management service as per the
1251 * FC-GS Fabric Device Management Information(FDMI) specification.
1252 *
1253 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1254 * Called when the underlying device wants to override default World Wide
1255 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1256 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1257 * protocol stack to use.
1258 *
1259 * RFS acceleration.
1260 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1261 * u16 rxq_index, u32 flow_id);
1262 * Set hardware filter for RFS. rxq_index is the target queue index;
1263 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1264 * Return the filter ID on success, or a negative error code.
1265 *
1266 * Slave management functions (for bridge, bonding, etc).
1267 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1268 * Called to make another netdev an underling.
1269 *
1270 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1271 * Called to release previously enslaved netdev.
1272 *
1273 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1274 * struct sk_buff *skb,
1275 * bool all_slaves);
1276 * Get the xmit slave of master device. If all_slaves is true, function
1277 * assume all the slaves can transmit.
1278 *
1279 * Feature/offload setting functions.
1280 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1281 * netdev_features_t features);
1282 * Adjusts the requested feature flags according to device-specific
1283 * constraints, and returns the resulting flags. Must not modify
1284 * the device state.
1285 *
1286 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1287 * Called to update device configuration to new features. Passed
1288 * feature set might be less than what was returned by ndo_fix_features()).
1289 * Must return >0 or -errno if it changed dev->features itself.
1290 *
1291 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1292 * struct net_device *dev,
1293 * const unsigned char *addr, u16 vid, u16 flags,
1294 * struct netlink_ext_ack *extack);
1295 * Adds an FDB entry to dev for addr.
1296 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1297 * struct net_device *dev,
1298 * const unsigned char *addr, u16 vid)
1299 * Deletes the FDB entry from dev coresponding to addr.
1300 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1301 * struct net_device *dev, struct net_device *filter_dev,
1302 * int *idx)
1303 * Used to add FDB entries to dump requests. Implementers should add
1304 * entries to skb and update idx with the number of entries.
1305 *
1306 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1307 * u16 flags, struct netlink_ext_ack *extack)
1308 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1309 * struct net_device *dev, u32 filter_mask,
1310 * int nlflags)
1311 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1312 * u16 flags);
1313 *
1314 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1315 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1316 * which do not represent real hardware may define this to allow their
1317 * userspace components to manage their virtual carrier state. Devices
1318 * that determine carrier state from physical hardware properties (eg
1319 * network cables) or protocol-dependent mechanisms (eg
1320 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1321 *
1322 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1323 * struct netdev_phys_item_id *ppid);
1324 * Called to get ID of physical port of this device. If driver does
1325 * not implement this, it is assumed that the hw is not able to have
1326 * multiple net devices on single physical port.
1327 *
1328 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1329 * struct netdev_phys_item_id *ppid)
1330 * Called to get the parent ID of the physical port of this device.
1331 *
1332 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1333 * struct net_device *dev)
1334 * Called by upper layer devices to accelerate switching or other
1335 * station functionality into hardware. 'pdev is the lowerdev
1336 * to use for the offload and 'dev' is the net device that will
1337 * back the offload. Returns a pointer to the private structure
1338 * the upper layer will maintain.
1339 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1340 * Called by upper layer device to delete the station created
1341 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1342 * the station and priv is the structure returned by the add
1343 * operation.
1344 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1345 * int queue_index, u32 maxrate);
1346 * Called when a user wants to set a max-rate limitation of specific
1347 * TX queue.
1348 * int (*ndo_get_iflink)(const struct net_device *dev);
1349 * Called to get the iflink value of this device.
1350 * void (*ndo_change_proto_down)(struct net_device *dev,
1351 * bool proto_down);
1352 * This function is used to pass protocol port error state information
1353 * to the switch driver. The switch driver can react to the proto_down
1354 * by doing a phys down on the associated switch port.
1355 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1356 * This function is used to get egress tunnel information for given skb.
1357 * This is useful for retrieving outer tunnel header parameters while
1358 * sampling packet.
1359 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1360 * This function is used to specify the headroom that the skb must
1361 * consider when allocation skb during packet reception. Setting
1362 * appropriate rx headroom value allows avoiding skb head copy on
1363 * forward. Setting a negative value resets the rx headroom to the
1364 * default value.
1365 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1366 * This function is used to set or query state related to XDP on the
1367 * netdevice and manage BPF offload. See definition of
1368 * enum bpf_netdev_command for details.
1369 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1370 * u32 flags);
1371 * This function is used to submit @n XDP packets for transmit on a
1372 * netdevice. Returns number of frames successfully transmitted, frames
1373 * that got dropped are freed/returned via xdp_return_frame().
1374 * Returns negative number, means general error invoking ndo, meaning
1375 * no frames were xmit'ed and core-caller will free all frames.
1376 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1377 * struct xdp_buff *xdp);
1378 * Get the xmit slave of master device based on the xdp_buff.
1379 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1380 * This function is used to wake up the softirq, ksoftirqd or kthread
1381 * responsible for sending and/or receiving packets on a specific
1382 * queue id bound to an AF_XDP socket. The flags field specifies if
1383 * only RX, only Tx, or both should be woken up using the flags
1384 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1385 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1386 * Get devlink port instance associated with a given netdev.
1387 * Called with a reference on the netdevice and devlink locks only,
1388 * rtnl_lock is not held.
1389 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1390 * int cmd);
1391 * Add, change, delete or get information on an IPv4 tunnel.
1392 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1393 * If a device is paired with a peer device, return the peer instance.
1394 * The caller must be under RCU read context.
1395 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1396 * Get the forwarding path to reach the real device from the HW destination address
1397 */
1398 struct net_device_ops {
1399 int (*ndo_init)(struct net_device *dev);
1400 void (*ndo_uninit)(struct net_device *dev);
1401 int (*ndo_open)(struct net_device *dev);
1402 int (*ndo_stop)(struct net_device *dev);
1403 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1404 struct net_device *dev);
1405 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1406 struct net_device *dev,
1407 netdev_features_t features);
1408 u16 (*ndo_select_queue)(struct net_device *dev,
1409 struct sk_buff *skb,
1410 struct net_device *sb_dev);
1411 void (*ndo_change_rx_flags)(struct net_device *dev,
1412 int flags);
1413 void (*ndo_set_rx_mode)(struct net_device *dev);
1414 int (*ndo_set_mac_address)(struct net_device *dev,
1415 void *addr);
1416 int (*ndo_validate_addr)(struct net_device *dev);
1417 int (*ndo_do_ioctl)(struct net_device *dev,
1418 struct ifreq *ifr, int cmd);
1419 int (*ndo_eth_ioctl)(struct net_device *dev,
1420 struct ifreq *ifr, int cmd);
1421 int (*ndo_siocbond)(struct net_device *dev,
1422 struct ifreq *ifr, int cmd);
1423 int (*ndo_siocwandev)(struct net_device *dev,
1424 struct if_settings *ifs);
1425 int (*ndo_siocdevprivate)(struct net_device *dev,
1426 struct ifreq *ifr,
1427 void __user *data, int cmd);
1428 int (*ndo_set_config)(struct net_device *dev,
1429 struct ifmap *map);
1430 int (*ndo_change_mtu)(struct net_device *dev,
1431 int new_mtu);
1432 int (*ndo_neigh_setup)(struct net_device *dev,
1433 struct neigh_parms *);
1434 void (*ndo_tx_timeout) (struct net_device *dev,
1435 unsigned int txqueue);
1436
1437 void (*ndo_get_stats64)(struct net_device *dev,
1438 struct rtnl_link_stats64 *storage);
1439 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1440 int (*ndo_get_offload_stats)(int attr_id,
1441 const struct net_device *dev,
1442 void *attr_data);
1443 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1444
1445 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1446 __be16 proto, u16 vid);
1447 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1448 __be16 proto, u16 vid);
1449 #ifdef CONFIG_NET_POLL_CONTROLLER
1450 void (*ndo_poll_controller)(struct net_device *dev);
1451 int (*ndo_netpoll_setup)(struct net_device *dev,
1452 struct netpoll_info *info);
1453 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1454 #endif
1455 int (*ndo_set_vf_mac)(struct net_device *dev,
1456 int queue, u8 *mac);
1457 int (*ndo_set_vf_vlan)(struct net_device *dev,
1458 int queue, u16 vlan,
1459 u8 qos, __be16 proto);
1460 int (*ndo_set_vf_rate)(struct net_device *dev,
1461 int vf, int min_tx_rate,
1462 int max_tx_rate);
1463 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1464 int vf, bool setting);
1465 int (*ndo_set_vf_trust)(struct net_device *dev,
1466 int vf, bool setting);
1467 int (*ndo_get_vf_config)(struct net_device *dev,
1468 int vf,
1469 struct ifla_vf_info *ivf);
1470 int (*ndo_set_vf_link_state)(struct net_device *dev,
1471 int vf, int link_state);
1472 int (*ndo_get_vf_stats)(struct net_device *dev,
1473 int vf,
1474 struct ifla_vf_stats
1475 *vf_stats);
1476 int (*ndo_set_vf_port)(struct net_device *dev,
1477 int vf,
1478 struct nlattr *port[]);
1479 int (*ndo_get_vf_port)(struct net_device *dev,
1480 int vf, struct sk_buff *skb);
1481 int (*ndo_get_vf_guid)(struct net_device *dev,
1482 int vf,
1483 struct ifla_vf_guid *node_guid,
1484 struct ifla_vf_guid *port_guid);
1485 int (*ndo_set_vf_guid)(struct net_device *dev,
1486 int vf, u64 guid,
1487 int guid_type);
1488 int (*ndo_set_vf_rss_query_en)(
1489 struct net_device *dev,
1490 int vf, bool setting);
1491 int (*ndo_setup_tc)(struct net_device *dev,
1492 enum tc_setup_type type,
1493 void *type_data);
1494 #if IS_ENABLED(CONFIG_FCOE)
1495 int (*ndo_fcoe_enable)(struct net_device *dev);
1496 int (*ndo_fcoe_disable)(struct net_device *dev);
1497 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1498 u16 xid,
1499 struct scatterlist *sgl,
1500 unsigned int sgc);
1501 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1502 u16 xid);
1503 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1504 u16 xid,
1505 struct scatterlist *sgl,
1506 unsigned int sgc);
1507 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1508 struct netdev_fcoe_hbainfo *hbainfo);
1509 #endif
1510
1511 #if IS_ENABLED(CONFIG_LIBFCOE)
1512 #define NETDEV_FCOE_WWNN 0
1513 #define NETDEV_FCOE_WWPN 1
1514 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1515 u64 *wwn, int type);
1516 #endif
1517
1518 #ifdef CONFIG_RFS_ACCEL
1519 int (*ndo_rx_flow_steer)(struct net_device *dev,
1520 const struct sk_buff *skb,
1521 u16 rxq_index,
1522 u32 flow_id);
1523 #endif
1524 int (*ndo_add_slave)(struct net_device *dev,
1525 struct net_device *slave_dev,
1526 struct netlink_ext_ack *extack);
1527 int (*ndo_del_slave)(struct net_device *dev,
1528 struct net_device *slave_dev);
1529 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1530 struct sk_buff *skb,
1531 bool all_slaves);
1532 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1533 struct sock *sk);
1534 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1535 netdev_features_t features);
1536 int (*ndo_set_features)(struct net_device *dev,
1537 netdev_features_t features);
1538 int (*ndo_neigh_construct)(struct net_device *dev,
1539 struct neighbour *n);
1540 void (*ndo_neigh_destroy)(struct net_device *dev,
1541 struct neighbour *n);
1542
1543 int (*ndo_fdb_add)(struct ndmsg *ndm,
1544 struct nlattr *tb[],
1545 struct net_device *dev,
1546 const unsigned char *addr,
1547 u16 vid,
1548 u16 flags,
1549 struct netlink_ext_ack *extack);
1550 int (*ndo_fdb_del)(struct ndmsg *ndm,
1551 struct nlattr *tb[],
1552 struct net_device *dev,
1553 const unsigned char *addr,
1554 u16 vid);
1555 int (*ndo_fdb_dump)(struct sk_buff *skb,
1556 struct netlink_callback *cb,
1557 struct net_device *dev,
1558 struct net_device *filter_dev,
1559 int *idx);
1560 int (*ndo_fdb_get)(struct sk_buff *skb,
1561 struct nlattr *tb[],
1562 struct net_device *dev,
1563 const unsigned char *addr,
1564 u16 vid, u32 portid, u32 seq,
1565 struct netlink_ext_ack *extack);
1566 int (*ndo_bridge_setlink)(struct net_device *dev,
1567 struct nlmsghdr *nlh,
1568 u16 flags,
1569 struct netlink_ext_ack *extack);
1570 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1571 u32 pid, u32 seq,
1572 struct net_device *dev,
1573 u32 filter_mask,
1574 int nlflags);
1575 int (*ndo_bridge_dellink)(struct net_device *dev,
1576 struct nlmsghdr *nlh,
1577 u16 flags);
1578 int (*ndo_change_carrier)(struct net_device *dev,
1579 bool new_carrier);
1580 int (*ndo_get_phys_port_id)(struct net_device *dev,
1581 struct netdev_phys_item_id *ppid);
1582 int (*ndo_get_port_parent_id)(struct net_device *dev,
1583 struct netdev_phys_item_id *ppid);
1584 int (*ndo_get_phys_port_name)(struct net_device *dev,
1585 char *name, size_t len);
1586 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1587 struct net_device *dev);
1588 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1589 void *priv);
1590
1591 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1592 int queue_index,
1593 u32 maxrate);
1594 int (*ndo_get_iflink)(const struct net_device *dev);
1595 int (*ndo_change_proto_down)(struct net_device *dev,
1596 bool proto_down);
1597 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1598 struct sk_buff *skb);
1599 void (*ndo_set_rx_headroom)(struct net_device *dev,
1600 int needed_headroom);
1601 int (*ndo_bpf)(struct net_device *dev,
1602 struct netdev_bpf *bpf);
1603 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1604 struct xdp_frame **xdp,
1605 u32 flags);
1606 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1607 struct xdp_buff *xdp);
1608 int (*ndo_xsk_wakeup)(struct net_device *dev,
1609 u32 queue_id, u32 flags);
1610 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1611 int (*ndo_tunnel_ctl)(struct net_device *dev,
1612 struct ip_tunnel_parm *p, int cmd);
1613 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1614 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1615 struct net_device_path *path);
1616
1617 ANDROID_KABI_RESERVE(1);
1618 ANDROID_KABI_RESERVE(2);
1619 ANDROID_KABI_RESERVE(3);
1620 ANDROID_KABI_RESERVE(4);
1621 ANDROID_KABI_RESERVE(5);
1622 ANDROID_KABI_RESERVE(6);
1623 ANDROID_KABI_RESERVE(7);
1624 ANDROID_KABI_RESERVE(8);
1625 };
1626
1627 /**
1628 * enum netdev_priv_flags - &struct net_device priv_flags
1629 *
1630 * These are the &struct net_device, they are only set internally
1631 * by drivers and used in the kernel. These flags are invisible to
1632 * userspace; this means that the order of these flags can change
1633 * during any kernel release.
1634 *
1635 * You should have a pretty good reason to be extending these flags.
1636 *
1637 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1638 * @IFF_EBRIDGE: Ethernet bridging device
1639 * @IFF_BONDING: bonding master or slave
1640 * @IFF_ISATAP: ISATAP interface (RFC4214)
1641 * @IFF_WAN_HDLC: WAN HDLC device
1642 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1643 * release skb->dst
1644 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1645 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1646 * @IFF_MACVLAN_PORT: device used as macvlan port
1647 * @IFF_BRIDGE_PORT: device used as bridge port
1648 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1649 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1650 * @IFF_UNICAST_FLT: Supports unicast filtering
1651 * @IFF_TEAM_PORT: device used as team port
1652 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1653 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1654 * change when it's running
1655 * @IFF_MACVLAN: Macvlan device
1656 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1657 * underlying stacked devices
1658 * @IFF_L3MDEV_MASTER: device is an L3 master device
1659 * @IFF_NO_QUEUE: device can run without qdisc attached
1660 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1661 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1662 * @IFF_TEAM: device is a team device
1663 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1664 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1665 * entity (i.e. the master device for bridged veth)
1666 * @IFF_MACSEC: device is a MACsec device
1667 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1668 * @IFF_FAILOVER: device is a failover master device
1669 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1670 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1671 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1672 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1673 * skb_headlen(skb) == 0 (data starts from frag0)
1674 */
1675 enum netdev_priv_flags {
1676 IFF_802_1Q_VLAN = 1<<0,
1677 IFF_EBRIDGE = 1<<1,
1678 IFF_BONDING = 1<<2,
1679 IFF_ISATAP = 1<<3,
1680 IFF_WAN_HDLC = 1<<4,
1681 IFF_XMIT_DST_RELEASE = 1<<5,
1682 IFF_DONT_BRIDGE = 1<<6,
1683 IFF_DISABLE_NETPOLL = 1<<7,
1684 IFF_MACVLAN_PORT = 1<<8,
1685 IFF_BRIDGE_PORT = 1<<9,
1686 IFF_OVS_DATAPATH = 1<<10,
1687 IFF_TX_SKB_SHARING = 1<<11,
1688 IFF_UNICAST_FLT = 1<<12,
1689 IFF_TEAM_PORT = 1<<13,
1690 IFF_SUPP_NOFCS = 1<<14,
1691 IFF_LIVE_ADDR_CHANGE = 1<<15,
1692 IFF_MACVLAN = 1<<16,
1693 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1694 IFF_L3MDEV_MASTER = 1<<18,
1695 IFF_NO_QUEUE = 1<<19,
1696 IFF_OPENVSWITCH = 1<<20,
1697 IFF_L3MDEV_SLAVE = 1<<21,
1698 IFF_TEAM = 1<<22,
1699 IFF_RXFH_CONFIGURED = 1<<23,
1700 IFF_PHONY_HEADROOM = 1<<24,
1701 IFF_MACSEC = 1<<25,
1702 IFF_NO_RX_HANDLER = 1<<26,
1703 IFF_FAILOVER = 1<<27,
1704 IFF_FAILOVER_SLAVE = 1<<28,
1705 IFF_L3MDEV_RX_HANDLER = 1<<29,
1706 IFF_LIVE_RENAME_OK = 1<<30,
1707 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1708 };
1709
1710 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1711 #define IFF_EBRIDGE IFF_EBRIDGE
1712 #define IFF_BONDING IFF_BONDING
1713 #define IFF_ISATAP IFF_ISATAP
1714 #define IFF_WAN_HDLC IFF_WAN_HDLC
1715 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1716 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1717 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1718 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1719 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1720 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1721 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1722 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1723 #define IFF_TEAM_PORT IFF_TEAM_PORT
1724 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1725 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1726 #define IFF_MACVLAN IFF_MACVLAN
1727 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1728 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1729 #define IFF_NO_QUEUE IFF_NO_QUEUE
1730 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1731 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1732 #define IFF_TEAM IFF_TEAM
1733 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1734 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1735 #define IFF_MACSEC IFF_MACSEC
1736 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1737 #define IFF_FAILOVER IFF_FAILOVER
1738 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1739 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1740 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1741 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1742
1743 /* Specifies the type of the struct net_device::ml_priv pointer */
1744 enum netdev_ml_priv_type {
1745 ML_PRIV_NONE,
1746 ML_PRIV_CAN,
1747 };
1748
1749 /**
1750 * struct net_device - The DEVICE structure.
1751 *
1752 * Actually, this whole structure is a big mistake. It mixes I/O
1753 * data with strictly "high-level" data, and it has to know about
1754 * almost every data structure used in the INET module.
1755 *
1756 * @name: This is the first field of the "visible" part of this structure
1757 * (i.e. as seen by users in the "Space.c" file). It is the name
1758 * of the interface.
1759 *
1760 * @name_node: Name hashlist node
1761 * @ifalias: SNMP alias
1762 * @mem_end: Shared memory end
1763 * @mem_start: Shared memory start
1764 * @base_addr: Device I/O address
1765 * @irq: Device IRQ number
1766 *
1767 * @state: Generic network queuing layer state, see netdev_state_t
1768 * @dev_list: The global list of network devices
1769 * @napi_list: List entry used for polling NAPI devices
1770 * @unreg_list: List entry when we are unregistering the
1771 * device; see the function unregister_netdev
1772 * @close_list: List entry used when we are closing the device
1773 * @ptype_all: Device-specific packet handlers for all protocols
1774 * @ptype_specific: Device-specific, protocol-specific packet handlers
1775 *
1776 * @adj_list: Directly linked devices, like slaves for bonding
1777 * @features: Currently active device features
1778 * @hw_features: User-changeable features
1779 *
1780 * @wanted_features: User-requested features
1781 * @vlan_features: Mask of features inheritable by VLAN devices
1782 *
1783 * @hw_enc_features: Mask of features inherited by encapsulating devices
1784 * This field indicates what encapsulation
1785 * offloads the hardware is capable of doing,
1786 * and drivers will need to set them appropriately.
1787 *
1788 * @mpls_features: Mask of features inheritable by MPLS
1789 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1790 *
1791 * @ifindex: interface index
1792 * @group: The group the device belongs to
1793 *
1794 * @stats: Statistics struct, which was left as a legacy, use
1795 * rtnl_link_stats64 instead
1796 *
1797 * @rx_dropped: Dropped packets by core network,
1798 * do not use this in drivers
1799 * @tx_dropped: Dropped packets by core network,
1800 * do not use this in drivers
1801 * @rx_nohandler: nohandler dropped packets by core network on
1802 * inactive devices, do not use this in drivers
1803 * @carrier_up_count: Number of times the carrier has been up
1804 * @carrier_down_count: Number of times the carrier has been down
1805 *
1806 * @wireless_handlers: List of functions to handle Wireless Extensions,
1807 * instead of ioctl,
1808 * see <net/iw_handler.h> for details.
1809 * @wireless_data: Instance data managed by the core of wireless extensions
1810 *
1811 * @netdev_ops: Includes several pointers to callbacks,
1812 * if one wants to override the ndo_*() functions
1813 * @ethtool_ops: Management operations
1814 * @l3mdev_ops: Layer 3 master device operations
1815 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1816 * discovery handling. Necessary for e.g. 6LoWPAN.
1817 * @xfrmdev_ops: Transformation offload operations
1818 * @tlsdev_ops: Transport Layer Security offload operations
1819 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1820 * of Layer 2 headers.
1821 *
1822 * @flags: Interface flags (a la BSD)
1823 * @priv_flags: Like 'flags' but invisible to userspace,
1824 * see if.h for the definitions
1825 * @gflags: Global flags ( kept as legacy )
1826 * @padded: How much padding added by alloc_netdev()
1827 * @operstate: RFC2863 operstate
1828 * @link_mode: Mapping policy to operstate
1829 * @if_port: Selectable AUI, TP, ...
1830 * @dma: DMA channel
1831 * @mtu: Interface MTU value
1832 * @min_mtu: Interface Minimum MTU value
1833 * @max_mtu: Interface Maximum MTU value
1834 * @type: Interface hardware type
1835 * @hard_header_len: Maximum hardware header length.
1836 * @min_header_len: Minimum hardware header length
1837 *
1838 * @needed_headroom: Extra headroom the hardware may need, but not in all
1839 * cases can this be guaranteed
1840 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1841 * cases can this be guaranteed. Some cases also use
1842 * LL_MAX_HEADER instead to allocate the skb
1843 *
1844 * interface address info:
1845 *
1846 * @perm_addr: Permanent hw address
1847 * @addr_assign_type: Hw address assignment type
1848 * @addr_len: Hardware address length
1849 * @upper_level: Maximum depth level of upper devices.
1850 * @lower_level: Maximum depth level of lower devices.
1851 * @neigh_priv_len: Used in neigh_alloc()
1852 * @dev_id: Used to differentiate devices that share
1853 * the same link layer address
1854 * @dev_port: Used to differentiate devices that share
1855 * the same function
1856 * @addr_list_lock: XXX: need comments on this one
1857 * @name_assign_type: network interface name assignment type
1858 * @uc_promisc: Counter that indicates promiscuous mode
1859 * has been enabled due to the need to listen to
1860 * additional unicast addresses in a device that
1861 * does not implement ndo_set_rx_mode()
1862 * @uc: unicast mac addresses
1863 * @mc: multicast mac addresses
1864 * @dev_addrs: list of device hw addresses
1865 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1866 * @promiscuity: Number of times the NIC is told to work in
1867 * promiscuous mode; if it becomes 0 the NIC will
1868 * exit promiscuous mode
1869 * @allmulti: Counter, enables or disables allmulticast mode
1870 *
1871 * @vlan_info: VLAN info
1872 * @dsa_ptr: dsa specific data
1873 * @tipc_ptr: TIPC specific data
1874 * @atalk_ptr: AppleTalk link
1875 * @ip_ptr: IPv4 specific data
1876 * @ip6_ptr: IPv6 specific data
1877 * @ax25_ptr: AX.25 specific data
1878 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1879 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1880 * device struct
1881 * @mpls_ptr: mpls_dev struct pointer
1882 * @mctp_ptr: MCTP specific data
1883 *
1884 * @dev_addr: Hw address (before bcast,
1885 * because most packets are unicast)
1886 *
1887 * @_rx: Array of RX queues
1888 * @num_rx_queues: Number of RX queues
1889 * allocated at register_netdev() time
1890 * @real_num_rx_queues: Number of RX queues currently active in device
1891 * @xdp_prog: XDP sockets filter program pointer
1892 * @gro_flush_timeout: timeout for GRO layer in NAPI
1893 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1894 * allow to avoid NIC hard IRQ, on busy queues.
1895 *
1896 * @rx_handler: handler for received packets
1897 * @rx_handler_data: XXX: need comments on this one
1898 * @miniq_ingress: ingress/clsact qdisc specific data for
1899 * ingress processing
1900 * @ingress_queue: XXX: need comments on this one
1901 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1902 * @broadcast: hw bcast address
1903 *
1904 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1905 * indexed by RX queue number. Assigned by driver.
1906 * This must only be set if the ndo_rx_flow_steer
1907 * operation is defined
1908 * @index_hlist: Device index hash chain
1909 *
1910 * @_tx: Array of TX queues
1911 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1912 * @real_num_tx_queues: Number of TX queues currently active in device
1913 * @qdisc: Root qdisc from userspace point of view
1914 * @tx_queue_len: Max frames per queue allowed
1915 * @tx_global_lock: XXX: need comments on this one
1916 * @xdp_bulkq: XDP device bulk queue
1917 * @xps_maps: all CPUs/RXQs maps for XPS device
1918 *
1919 * @xps_maps: XXX: need comments on this one
1920 * @miniq_egress: clsact qdisc specific data for
1921 * egress processing
1922 * @qdisc_hash: qdisc hash table
1923 * @watchdog_timeo: Represents the timeout that is used by
1924 * the watchdog (see dev_watchdog())
1925 * @watchdog_timer: List of timers
1926 *
1927 * @proto_down_reason: reason a netdev interface is held down
1928 * @pcpu_refcnt: Number of references to this device
1929 * @dev_refcnt: Number of references to this device
1930 * @todo_list: Delayed register/unregister
1931 * @link_watch_list: XXX: need comments on this one
1932 *
1933 * @reg_state: Register/unregister state machine
1934 * @dismantle: Device is going to be freed
1935 * @rtnl_link_state: This enum represents the phases of creating
1936 * a new link
1937 *
1938 * @needs_free_netdev: Should unregister perform free_netdev?
1939 * @priv_destructor: Called from unregister
1940 * @npinfo: XXX: need comments on this one
1941 * @nd_net: Network namespace this network device is inside
1942 *
1943 * @ml_priv: Mid-layer private
1944 * @ml_priv_type: Mid-layer private type
1945 * @lstats: Loopback statistics
1946 * @tstats: Tunnel statistics
1947 * @dstats: Dummy statistics
1948 * @vstats: Virtual ethernet statistics
1949 *
1950 * @garp_port: GARP
1951 * @mrp_port: MRP
1952 *
1953 * @dev: Class/net/name entry
1954 * @sysfs_groups: Space for optional device, statistics and wireless
1955 * sysfs groups
1956 *
1957 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1958 * @rtnl_link_ops: Rtnl_link_ops
1959 *
1960 * @gso_max_size: Maximum size of generic segmentation offload
1961 * @gso_max_segs: Maximum number of segments that can be passed to the
1962 * NIC for GSO
1963 *
1964 * @dcbnl_ops: Data Center Bridging netlink ops
1965 * @num_tc: Number of traffic classes in the net device
1966 * @tc_to_txq: XXX: need comments on this one
1967 * @prio_tc_map: XXX: need comments on this one
1968 *
1969 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1970 *
1971 * @priomap: XXX: need comments on this one
1972 * @phydev: Physical device may attach itself
1973 * for hardware timestamping
1974 * @sfp_bus: attached &struct sfp_bus structure.
1975 *
1976 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1977 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1978 *
1979 * @proto_down: protocol port state information can be sent to the
1980 * switch driver and used to set the phys state of the
1981 * switch port.
1982 *
1983 * @wol_enabled: Wake-on-LAN is enabled
1984 *
1985 * @threaded: napi threaded mode is enabled
1986 *
1987 * @net_notifier_list: List of per-net netdev notifier block
1988 * that follow this device when it is moved
1989 * to another network namespace.
1990 *
1991 * @macsec_ops: MACsec offloading ops
1992 *
1993 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1994 * offload capabilities of the device
1995 * @udp_tunnel_nic: UDP tunnel offload state
1996 * @xdp_state: stores info on attached XDP BPF programs
1997 *
1998 * @nested_level: Used as as a parameter of spin_lock_nested() of
1999 * dev->addr_list_lock.
2000 * @unlink_list: As netif_addr_lock() can be called recursively,
2001 * keep a list of interfaces to be deleted.
2002 *
2003 * FIXME: cleanup struct net_device such that network protocol info
2004 * moves out.
2005 */
2006
2007 struct net_device {
2008 char name[IFNAMSIZ];
2009 struct netdev_name_node *name_node;
2010 struct dev_ifalias __rcu *ifalias;
2011 /*
2012 * I/O specific fields
2013 * FIXME: Merge these and struct ifmap into one
2014 */
2015 unsigned long mem_end;
2016 unsigned long mem_start;
2017 unsigned long base_addr;
2018
2019 /*
2020 * Some hardware also needs these fields (state,dev_list,
2021 * napi_list,unreg_list,close_list) but they are not
2022 * part of the usual set specified in Space.c.
2023 */
2024
2025 unsigned long state;
2026
2027 struct list_head dev_list;
2028 struct list_head napi_list;
2029 struct list_head unreg_list;
2030 struct list_head close_list;
2031 struct list_head ptype_all;
2032 struct list_head ptype_specific;
2033
2034 struct {
2035 struct list_head upper;
2036 struct list_head lower;
2037 } adj_list;
2038
2039 /* Read-mostly cache-line for fast-path access */
2040 unsigned int flags;
2041 unsigned int priv_flags;
2042 const struct net_device_ops *netdev_ops;
2043 int ifindex;
2044 unsigned short gflags;
2045 unsigned short hard_header_len;
2046
2047 /* Note : dev->mtu is often read without holding a lock.
2048 * Writers usually hold RTNL.
2049 * It is recommended to use READ_ONCE() to annotate the reads,
2050 * and to use WRITE_ONCE() to annotate the writes.
2051 */
2052 unsigned int mtu;
2053 unsigned short needed_headroom;
2054 unsigned short needed_tailroom;
2055
2056 netdev_features_t features;
2057 netdev_features_t hw_features;
2058 netdev_features_t wanted_features;
2059 netdev_features_t vlan_features;
2060 netdev_features_t hw_enc_features;
2061 netdev_features_t mpls_features;
2062 netdev_features_t gso_partial_features;
2063
2064 unsigned int min_mtu;
2065 unsigned int max_mtu;
2066 unsigned short type;
2067 unsigned char min_header_len;
2068 unsigned char name_assign_type;
2069
2070 int group;
2071
2072 struct net_device_stats stats; /* not used by modern drivers */
2073
2074 atomic_long_t rx_dropped;
2075 atomic_long_t tx_dropped;
2076 atomic_long_t rx_nohandler;
2077
2078 /* Stats to monitor link on/off, flapping */
2079 atomic_t carrier_up_count;
2080 atomic_t carrier_down_count;
2081
2082 #ifdef CONFIG_WIRELESS_EXT
2083 const struct iw_handler_def *wireless_handlers;
2084 struct iw_public_data *wireless_data;
2085 #endif
2086 const struct ethtool_ops *ethtool_ops;
2087 #ifdef CONFIG_NET_L3_MASTER_DEV
2088 const struct l3mdev_ops *l3mdev_ops;
2089 #endif
2090 #if IS_ENABLED(CONFIG_IPV6)
2091 const struct ndisc_ops *ndisc_ops;
2092 #endif
2093
2094 #ifdef CONFIG_XFRM_OFFLOAD
2095 const struct xfrmdev_ops *xfrmdev_ops;
2096 #endif
2097
2098 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2099 const struct tlsdev_ops *tlsdev_ops;
2100 #endif
2101
2102 const struct header_ops *header_ops;
2103
2104 unsigned char operstate;
2105 unsigned char link_mode;
2106
2107 unsigned char if_port;
2108 unsigned char dma;
2109
2110 /* Interface address info. */
2111 unsigned char perm_addr[MAX_ADDR_LEN];
2112 unsigned char addr_assign_type;
2113 unsigned char addr_len;
2114 unsigned char upper_level;
2115 unsigned char lower_level;
2116
2117 unsigned short neigh_priv_len;
2118 unsigned short dev_id;
2119 unsigned short dev_port;
2120 unsigned short padded;
2121
2122 spinlock_t addr_list_lock;
2123 int irq;
2124
2125 struct netdev_hw_addr_list uc;
2126 struct netdev_hw_addr_list mc;
2127 struct netdev_hw_addr_list dev_addrs;
2128
2129 #ifdef CONFIG_SYSFS
2130 struct kset *queues_kset;
2131 #endif
2132 #ifdef CONFIG_LOCKDEP
2133 struct list_head unlink_list;
2134 #endif
2135 unsigned int promiscuity;
2136 unsigned int allmulti;
2137 bool uc_promisc;
2138 #ifdef CONFIG_LOCKDEP
2139 unsigned char nested_level;
2140 #endif
2141
2142
2143 /* Protocol-specific pointers */
2144
2145 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2146 struct vlan_info __rcu *vlan_info;
2147 #endif
2148 #if IS_ENABLED(CONFIG_NET_DSA)
2149 struct dsa_port *dsa_ptr;
2150 #endif
2151 #if IS_ENABLED(CONFIG_TIPC)
2152 struct tipc_bearer __rcu *tipc_ptr;
2153 #endif
2154 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2155 void *atalk_ptr;
2156 #endif
2157 struct in_device __rcu *ip_ptr;
2158 struct inet6_dev __rcu *ip6_ptr;
2159 #if IS_ENABLED(CONFIG_AX25)
2160 void *ax25_ptr;
2161 #endif
2162 struct wireless_dev *ieee80211_ptr;
2163 struct wpan_dev *ieee802154_ptr;
2164 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2165 struct mpls_dev __rcu *mpls_ptr;
2166 #endif
2167 #if IS_ENABLED(CONFIG_MCTP)
2168 struct mctp_dev __rcu *mctp_ptr;
2169 #endif
2170
2171 /*
2172 * Cache lines mostly used on receive path (including eth_type_trans())
2173 */
2174 /* Interface address info used in eth_type_trans() */
2175 unsigned char *dev_addr;
2176
2177 struct netdev_rx_queue *_rx;
2178 unsigned int num_rx_queues;
2179 unsigned int real_num_rx_queues;
2180
2181 struct bpf_prog __rcu *xdp_prog;
2182 unsigned long gro_flush_timeout;
2183 int napi_defer_hard_irqs;
2184 rx_handler_func_t __rcu *rx_handler;
2185 void __rcu *rx_handler_data;
2186
2187 #ifdef CONFIG_NET_CLS_ACT
2188 struct mini_Qdisc __rcu *miniq_ingress;
2189 #endif
2190 struct netdev_queue __rcu *ingress_queue;
2191 #ifdef CONFIG_NETFILTER_INGRESS
2192 struct nf_hook_entries __rcu *nf_hooks_ingress;
2193 #endif
2194
2195 unsigned char broadcast[MAX_ADDR_LEN];
2196 #ifdef CONFIG_RFS_ACCEL
2197 struct cpu_rmap *rx_cpu_rmap;
2198 #endif
2199 struct hlist_node index_hlist;
2200
2201 /*
2202 * Cache lines mostly used on transmit path
2203 */
2204 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2205 unsigned int num_tx_queues;
2206 unsigned int real_num_tx_queues;
2207 struct Qdisc __rcu *qdisc;
2208 unsigned int tx_queue_len;
2209 spinlock_t tx_global_lock;
2210
2211 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2212
2213 #ifdef CONFIG_XPS
2214 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2215 #endif
2216 #ifdef CONFIG_NET_CLS_ACT
2217 struct mini_Qdisc __rcu *miniq_egress;
2218 #endif
2219
2220 #ifdef CONFIG_NET_SCHED
2221 DECLARE_HASHTABLE (qdisc_hash, 4);
2222 #endif
2223 /* These may be needed for future network-power-down code. */
2224 struct timer_list watchdog_timer;
2225 int watchdog_timeo;
2226
2227 u32 proto_down_reason;
2228
2229 struct list_head todo_list;
2230
2231 #ifdef CONFIG_PCPU_DEV_REFCNT
2232 int __percpu *pcpu_refcnt;
2233 #else
2234 refcount_t dev_refcnt;
2235 #endif
2236
2237 struct list_head link_watch_list;
2238
2239 enum { NETREG_UNINITIALIZED=0,
2240 NETREG_REGISTERED, /* completed register_netdevice */
2241 NETREG_UNREGISTERING, /* called unregister_netdevice */
2242 NETREG_UNREGISTERED, /* completed unregister todo */
2243 NETREG_RELEASED, /* called free_netdev */
2244 NETREG_DUMMY, /* dummy device for NAPI poll */
2245 } reg_state:8;
2246
2247 bool dismantle;
2248
2249 enum {
2250 RTNL_LINK_INITIALIZED,
2251 RTNL_LINK_INITIALIZING,
2252 } rtnl_link_state:16;
2253
2254 bool needs_free_netdev;
2255 void (*priv_destructor)(struct net_device *dev);
2256
2257 #ifdef CONFIG_NETPOLL
2258 struct netpoll_info __rcu *npinfo;
2259 #endif
2260
2261 possible_net_t nd_net;
2262
2263 /* mid-layer private */
2264 void *ml_priv;
2265 enum netdev_ml_priv_type ml_priv_type;
2266
2267 union {
2268 struct pcpu_lstats __percpu *lstats;
2269 struct pcpu_sw_netstats __percpu *tstats;
2270 struct pcpu_dstats __percpu *dstats;
2271 };
2272
2273 #if IS_ENABLED(CONFIG_GARP)
2274 struct garp_port __rcu *garp_port;
2275 #endif
2276 #if IS_ENABLED(CONFIG_MRP)
2277 struct mrp_port __rcu *mrp_port;
2278 #endif
2279
2280 struct device dev;
2281 const struct attribute_group *sysfs_groups[4];
2282 const struct attribute_group *sysfs_rx_queue_group;
2283
2284 const struct rtnl_link_ops *rtnl_link_ops;
2285
2286 /* for setting kernel sock attribute on TCP connection setup */
2287 #define GSO_MAX_SIZE 65536
2288 unsigned int gso_max_size;
2289 #define GSO_MAX_SEGS 65535
2290 u16 gso_max_segs;
2291
2292 #ifdef CONFIG_DCB
2293 const struct dcbnl_rtnl_ops *dcbnl_ops;
2294 #endif
2295 s16 num_tc;
2296 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2297 u8 prio_tc_map[TC_BITMASK + 1];
2298
2299 #if IS_ENABLED(CONFIG_FCOE)
2300 unsigned int fcoe_ddp_xid;
2301 #endif
2302 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2303 struct netprio_map __rcu *priomap;
2304 #endif
2305 struct phy_device *phydev;
2306 struct sfp_bus *sfp_bus;
2307 struct lock_class_key *qdisc_tx_busylock;
2308 struct lock_class_key *qdisc_running_key;
2309 bool proto_down;
2310 unsigned wol_enabled:1;
2311 unsigned threaded:1;
2312
2313 struct list_head net_notifier_list;
2314
2315 #if IS_ENABLED(CONFIG_MACSEC)
2316 /* MACsec management functions */
2317 const struct macsec_ops *macsec_ops;
2318 #endif
2319 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2320 struct udp_tunnel_nic *udp_tunnel_nic;
2321
2322 /* protected by rtnl_lock */
2323 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2324
2325 ANDROID_KABI_RESERVE(1);
2326 ANDROID_KABI_RESERVE(2);
2327 ANDROID_KABI_RESERVE(3);
2328 ANDROID_KABI_RESERVE(4);
2329 ANDROID_KABI_RESERVE(5);
2330 ANDROID_KABI_RESERVE(6);
2331 ANDROID_KABI_RESERVE(7);
2332 ANDROID_KABI_RESERVE(8);
2333 };
2334 #define to_net_dev(d) container_of(d, struct net_device, dev)
2335
netif_elide_gro(const struct net_device * dev)2336 static inline bool netif_elide_gro(const struct net_device *dev)
2337 {
2338 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2339 return true;
2340 return false;
2341 }
2342
2343 #define NETDEV_ALIGN 32
2344
2345 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)2346 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2347 {
2348 return dev->prio_tc_map[prio & TC_BITMASK];
2349 }
2350
2351 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)2352 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2353 {
2354 if (tc >= dev->num_tc)
2355 return -EINVAL;
2356
2357 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2358 return 0;
2359 }
2360
2361 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2362 void netdev_reset_tc(struct net_device *dev);
2363 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2364 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2365
2366 static inline
netdev_get_num_tc(struct net_device * dev)2367 int netdev_get_num_tc(struct net_device *dev)
2368 {
2369 return dev->num_tc;
2370 }
2371
net_prefetch(void * p)2372 static inline void net_prefetch(void *p)
2373 {
2374 prefetch(p);
2375 #if L1_CACHE_BYTES < 128
2376 prefetch((u8 *)p + L1_CACHE_BYTES);
2377 #endif
2378 }
2379
net_prefetchw(void * p)2380 static inline void net_prefetchw(void *p)
2381 {
2382 prefetchw(p);
2383 #if L1_CACHE_BYTES < 128
2384 prefetchw((u8 *)p + L1_CACHE_BYTES);
2385 #endif
2386 }
2387
2388 void netdev_unbind_sb_channel(struct net_device *dev,
2389 struct net_device *sb_dev);
2390 int netdev_bind_sb_channel_queue(struct net_device *dev,
2391 struct net_device *sb_dev,
2392 u8 tc, u16 count, u16 offset);
2393 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
netdev_get_sb_channel(struct net_device * dev)2394 static inline int netdev_get_sb_channel(struct net_device *dev)
2395 {
2396 return max_t(int, -dev->num_tc, 0);
2397 }
2398
2399 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)2400 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2401 unsigned int index)
2402 {
2403 return &dev->_tx[index];
2404 }
2405
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)2406 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2407 const struct sk_buff *skb)
2408 {
2409 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2410 }
2411
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)2412 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2413 void (*f)(struct net_device *,
2414 struct netdev_queue *,
2415 void *),
2416 void *arg)
2417 {
2418 unsigned int i;
2419
2420 for (i = 0; i < dev->num_tx_queues; i++)
2421 f(dev, &dev->_tx[i], arg);
2422 }
2423
2424 #define netdev_lockdep_set_classes(dev) \
2425 { \
2426 static struct lock_class_key qdisc_tx_busylock_key; \
2427 static struct lock_class_key qdisc_running_key; \
2428 static struct lock_class_key qdisc_xmit_lock_key; \
2429 static struct lock_class_key dev_addr_list_lock_key; \
2430 unsigned int i; \
2431 \
2432 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2433 (dev)->qdisc_running_key = &qdisc_running_key; \
2434 lockdep_set_class(&(dev)->addr_list_lock, \
2435 &dev_addr_list_lock_key); \
2436 for (i = 0; i < (dev)->num_tx_queues; i++) \
2437 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2438 &qdisc_xmit_lock_key); \
2439 }
2440
2441 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2442 struct net_device *sb_dev);
2443 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2444 struct sk_buff *skb,
2445 struct net_device *sb_dev);
2446
2447 /* returns the headroom that the master device needs to take in account
2448 * when forwarding to this dev
2449 */
netdev_get_fwd_headroom(struct net_device * dev)2450 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2451 {
2452 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2453 }
2454
netdev_set_rx_headroom(struct net_device * dev,int new_hr)2455 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2456 {
2457 if (dev->netdev_ops->ndo_set_rx_headroom)
2458 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2459 }
2460
2461 /* set the device rx headroom to the dev's default */
netdev_reset_rx_headroom(struct net_device * dev)2462 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2463 {
2464 netdev_set_rx_headroom(dev, -1);
2465 }
2466
netdev_get_ml_priv(struct net_device * dev,enum netdev_ml_priv_type type)2467 static inline void *netdev_get_ml_priv(struct net_device *dev,
2468 enum netdev_ml_priv_type type)
2469 {
2470 if (dev->ml_priv_type != type)
2471 return NULL;
2472
2473 return dev->ml_priv;
2474 }
2475
netdev_set_ml_priv(struct net_device * dev,void * ml_priv,enum netdev_ml_priv_type type)2476 static inline void netdev_set_ml_priv(struct net_device *dev,
2477 void *ml_priv,
2478 enum netdev_ml_priv_type type)
2479 {
2480 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2481 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2482 dev->ml_priv_type, type);
2483 WARN(!dev->ml_priv_type && dev->ml_priv,
2484 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2485
2486 dev->ml_priv = ml_priv;
2487 dev->ml_priv_type = type;
2488 }
2489
2490 /*
2491 * Net namespace inlines
2492 */
2493 static inline
dev_net(const struct net_device * dev)2494 struct net *dev_net(const struct net_device *dev)
2495 {
2496 return read_pnet(&dev->nd_net);
2497 }
2498
2499 static inline
dev_net_set(struct net_device * dev,struct net * net)2500 void dev_net_set(struct net_device *dev, struct net *net)
2501 {
2502 write_pnet(&dev->nd_net, net);
2503 }
2504
2505 /**
2506 * netdev_priv - access network device private data
2507 * @dev: network device
2508 *
2509 * Get network device private data
2510 */
netdev_priv(const struct net_device * dev)2511 static inline void *netdev_priv(const struct net_device *dev)
2512 {
2513 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2514 }
2515
2516 /* Set the sysfs physical device reference for the network logical device
2517 * if set prior to registration will cause a symlink during initialization.
2518 */
2519 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2520
2521 /* Set the sysfs device type for the network logical device to allow
2522 * fine-grained identification of different network device types. For
2523 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2524 */
2525 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2526
2527 /* Default NAPI poll() weight
2528 * Device drivers are strongly advised to not use bigger value
2529 */
2530 #define NAPI_POLL_WEIGHT 64
2531
2532 /**
2533 * netif_napi_add - initialize a NAPI context
2534 * @dev: network device
2535 * @napi: NAPI context
2536 * @poll: polling function
2537 * @weight: default weight
2538 *
2539 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2540 * *any* of the other NAPI-related functions.
2541 */
2542 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2543 int (*poll)(struct napi_struct *, int), int weight);
2544
2545 /**
2546 * netif_tx_napi_add - initialize a NAPI context
2547 * @dev: network device
2548 * @napi: NAPI context
2549 * @poll: polling function
2550 * @weight: default weight
2551 *
2552 * This variant of netif_napi_add() should be used from drivers using NAPI
2553 * to exclusively poll a TX queue.
2554 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2555 */
netif_tx_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2556 static inline void netif_tx_napi_add(struct net_device *dev,
2557 struct napi_struct *napi,
2558 int (*poll)(struct napi_struct *, int),
2559 int weight)
2560 {
2561 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2562 netif_napi_add(dev, napi, poll, weight);
2563 }
2564
2565 /**
2566 * __netif_napi_del - remove a NAPI context
2567 * @napi: NAPI context
2568 *
2569 * Warning: caller must observe RCU grace period before freeing memory
2570 * containing @napi. Drivers might want to call this helper to combine
2571 * all the needed RCU grace periods into a single one.
2572 */
2573 void __netif_napi_del(struct napi_struct *napi);
2574
2575 /**
2576 * netif_napi_del - remove a NAPI context
2577 * @napi: NAPI context
2578 *
2579 * netif_napi_del() removes a NAPI context from the network device NAPI list
2580 */
netif_napi_del(struct napi_struct * napi)2581 static inline void netif_napi_del(struct napi_struct *napi)
2582 {
2583 __netif_napi_del(napi);
2584 synchronize_net();
2585 }
2586
2587 struct napi_gro_cb {
2588 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2589 void *frag0;
2590
2591 /* Length of frag0. */
2592 unsigned int frag0_len;
2593
2594 /* This indicates where we are processing relative to skb->data. */
2595 int data_offset;
2596
2597 /* This is non-zero if the packet cannot be merged with the new skb. */
2598 u16 flush;
2599
2600 /* Save the IP ID here and check when we get to the transport layer */
2601 u16 flush_id;
2602
2603 /* Number of segments aggregated. */
2604 u16 count;
2605
2606 /* Start offset for remote checksum offload */
2607 u16 gro_remcsum_start;
2608
2609 /* jiffies when first packet was created/queued */
2610 unsigned long age;
2611
2612 /* Used in ipv6_gro_receive() and foo-over-udp */
2613 u16 proto;
2614
2615 /* This is non-zero if the packet may be of the same flow. */
2616 u8 same_flow:1;
2617
2618 /* Used in tunnel GRO receive */
2619 u8 encap_mark:1;
2620
2621 /* GRO checksum is valid */
2622 u8 csum_valid:1;
2623
2624 /* Number of checksums via CHECKSUM_UNNECESSARY */
2625 u8 csum_cnt:3;
2626
2627 /* Free the skb? */
2628 u8 free:2;
2629 #define NAPI_GRO_FREE 1
2630 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2631
2632 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2633 u8 is_ipv6:1;
2634
2635 /* Used in GRE, set in fou/gue_gro_receive */
2636 u8 is_fou:1;
2637
2638 /* Used to determine if flush_id can be ignored */
2639 u8 is_atomic:1;
2640
2641 /* Number of gro_receive callbacks this packet already went through */
2642 u8 recursion_counter:4;
2643
2644 /* GRO is done by frag_list pointer chaining. */
2645 u8 is_flist:1;
2646
2647 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2648 __wsum csum;
2649
2650 /* used in skb_gro_receive() slow path */
2651 struct sk_buff *last;
2652 };
2653
2654 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2655
2656 #define GRO_RECURSION_LIMIT 15
gro_recursion_inc_test(struct sk_buff * skb)2657 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2658 {
2659 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2660 }
2661
2662 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)2663 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2664 struct list_head *head,
2665 struct sk_buff *skb)
2666 {
2667 if (unlikely(gro_recursion_inc_test(skb))) {
2668 NAPI_GRO_CB(skb)->flush |= 1;
2669 return NULL;
2670 }
2671
2672 return cb(head, skb);
2673 }
2674
2675 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2676 struct sk_buff *);
call_gro_receive_sk(gro_receive_sk_t cb,struct sock * sk,struct list_head * head,struct sk_buff * skb)2677 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2678 struct sock *sk,
2679 struct list_head *head,
2680 struct sk_buff *skb)
2681 {
2682 if (unlikely(gro_recursion_inc_test(skb))) {
2683 NAPI_GRO_CB(skb)->flush |= 1;
2684 return NULL;
2685 }
2686
2687 return cb(sk, head, skb);
2688 }
2689
2690 struct packet_type {
2691 __be16 type; /* This is really htons(ether_type). */
2692 bool ignore_outgoing;
2693 struct net_device *dev; /* NULL is wildcarded here */
2694 int (*func) (struct sk_buff *,
2695 struct net_device *,
2696 struct packet_type *,
2697 struct net_device *);
2698 void (*list_func) (struct list_head *,
2699 struct packet_type *,
2700 struct net_device *);
2701 bool (*id_match)(struct packet_type *ptype,
2702 struct sock *sk);
2703 struct net *af_packet_net;
2704 void *af_packet_priv;
2705 struct list_head list;
2706
2707 ANDROID_KABI_RESERVE(1);
2708 ANDROID_KABI_RESERVE(2);
2709 ANDROID_KABI_RESERVE(3);
2710 ANDROID_KABI_RESERVE(4);
2711 };
2712
2713 struct offload_callbacks {
2714 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2715 netdev_features_t features);
2716 struct sk_buff *(*gro_receive)(struct list_head *head,
2717 struct sk_buff *skb);
2718 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2719 };
2720
2721 struct packet_offload {
2722 __be16 type; /* This is really htons(ether_type). */
2723 u16 priority;
2724 struct offload_callbacks callbacks;
2725 struct list_head list;
2726 };
2727
2728 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2729 struct pcpu_sw_netstats {
2730 u64 rx_packets;
2731 u64 rx_bytes;
2732 u64 tx_packets;
2733 u64 tx_bytes;
2734 struct u64_stats_sync syncp;
2735 } __aligned(4 * sizeof(u64));
2736
2737 struct pcpu_lstats {
2738 u64_stats_t packets;
2739 u64_stats_t bytes;
2740 struct u64_stats_sync syncp;
2741 } __aligned(2 * sizeof(u64));
2742
2743 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2744
dev_sw_netstats_rx_add(struct net_device * dev,unsigned int len)2745 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2746 {
2747 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2748
2749 u64_stats_update_begin(&tstats->syncp);
2750 tstats->rx_bytes += len;
2751 tstats->rx_packets++;
2752 u64_stats_update_end(&tstats->syncp);
2753 }
2754
dev_sw_netstats_tx_add(struct net_device * dev,unsigned int packets,unsigned int len)2755 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2756 unsigned int packets,
2757 unsigned int len)
2758 {
2759 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2760
2761 u64_stats_update_begin(&tstats->syncp);
2762 tstats->tx_bytes += len;
2763 tstats->tx_packets += packets;
2764 u64_stats_update_end(&tstats->syncp);
2765 }
2766
dev_lstats_add(struct net_device * dev,unsigned int len)2767 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2768 {
2769 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2770
2771 u64_stats_update_begin(&lstats->syncp);
2772 u64_stats_add(&lstats->bytes, len);
2773 u64_stats_inc(&lstats->packets);
2774 u64_stats_update_end(&lstats->syncp);
2775 }
2776
2777 #define __netdev_alloc_pcpu_stats(type, gfp) \
2778 ({ \
2779 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2780 if (pcpu_stats) { \
2781 int __cpu; \
2782 for_each_possible_cpu(__cpu) { \
2783 typeof(type) *stat; \
2784 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2785 u64_stats_init(&stat->syncp); \
2786 } \
2787 } \
2788 pcpu_stats; \
2789 })
2790
2791 #define netdev_alloc_pcpu_stats(type) \
2792 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2793
2794 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2795 ({ \
2796 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2797 if (pcpu_stats) { \
2798 int __cpu; \
2799 for_each_possible_cpu(__cpu) { \
2800 typeof(type) *stat; \
2801 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2802 u64_stats_init(&stat->syncp); \
2803 } \
2804 } \
2805 pcpu_stats; \
2806 })
2807
2808 enum netdev_lag_tx_type {
2809 NETDEV_LAG_TX_TYPE_UNKNOWN,
2810 NETDEV_LAG_TX_TYPE_RANDOM,
2811 NETDEV_LAG_TX_TYPE_BROADCAST,
2812 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2813 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2814 NETDEV_LAG_TX_TYPE_HASH,
2815 };
2816
2817 enum netdev_lag_hash {
2818 NETDEV_LAG_HASH_NONE,
2819 NETDEV_LAG_HASH_L2,
2820 NETDEV_LAG_HASH_L34,
2821 NETDEV_LAG_HASH_L23,
2822 NETDEV_LAG_HASH_E23,
2823 NETDEV_LAG_HASH_E34,
2824 NETDEV_LAG_HASH_VLAN_SRCMAC,
2825 NETDEV_LAG_HASH_UNKNOWN,
2826 };
2827
2828 struct netdev_lag_upper_info {
2829 enum netdev_lag_tx_type tx_type;
2830 enum netdev_lag_hash hash_type;
2831 };
2832
2833 struct netdev_lag_lower_state_info {
2834 u8 link_up : 1,
2835 tx_enabled : 1;
2836 };
2837
2838 #include <linux/notifier.h>
2839
2840 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2841 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2842 * adding new types.
2843 */
2844 enum netdev_cmd {
2845 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2846 NETDEV_DOWN,
2847 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2848 detected a hardware crash and restarted
2849 - we can use this eg to kick tcp sessions
2850 once done */
2851 NETDEV_CHANGE, /* Notify device state change */
2852 NETDEV_REGISTER,
2853 NETDEV_UNREGISTER,
2854 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2855 NETDEV_CHANGEADDR, /* notify after the address change */
2856 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2857 NETDEV_GOING_DOWN,
2858 NETDEV_CHANGENAME,
2859 NETDEV_FEAT_CHANGE,
2860 NETDEV_BONDING_FAILOVER,
2861 NETDEV_PRE_UP,
2862 NETDEV_PRE_TYPE_CHANGE,
2863 NETDEV_POST_TYPE_CHANGE,
2864 NETDEV_POST_INIT,
2865 NETDEV_RELEASE,
2866 NETDEV_NOTIFY_PEERS,
2867 NETDEV_JOIN,
2868 NETDEV_CHANGEUPPER,
2869 NETDEV_RESEND_IGMP,
2870 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2871 NETDEV_CHANGEINFODATA,
2872 NETDEV_BONDING_INFO,
2873 NETDEV_PRECHANGEUPPER,
2874 NETDEV_CHANGELOWERSTATE,
2875 NETDEV_UDP_TUNNEL_PUSH_INFO,
2876 NETDEV_UDP_TUNNEL_DROP_INFO,
2877 NETDEV_CHANGE_TX_QUEUE_LEN,
2878 NETDEV_CVLAN_FILTER_PUSH_INFO,
2879 NETDEV_CVLAN_FILTER_DROP_INFO,
2880 NETDEV_SVLAN_FILTER_PUSH_INFO,
2881 NETDEV_SVLAN_FILTER_DROP_INFO,
2882 };
2883 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2884
2885 int register_netdevice_notifier(struct notifier_block *nb);
2886 int unregister_netdevice_notifier(struct notifier_block *nb);
2887 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2888 int unregister_netdevice_notifier_net(struct net *net,
2889 struct notifier_block *nb);
2890 int register_netdevice_notifier_dev_net(struct net_device *dev,
2891 struct notifier_block *nb,
2892 struct netdev_net_notifier *nn);
2893 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2894 struct notifier_block *nb,
2895 struct netdev_net_notifier *nn);
2896
2897 struct netdev_notifier_info {
2898 struct net_device *dev;
2899 struct netlink_ext_ack *extack;
2900 };
2901
2902 struct netdev_notifier_info_ext {
2903 struct netdev_notifier_info info; /* must be first */
2904 union {
2905 u32 mtu;
2906 } ext;
2907 };
2908
2909 struct netdev_notifier_change_info {
2910 struct netdev_notifier_info info; /* must be first */
2911 unsigned int flags_changed;
2912 };
2913
2914 struct netdev_notifier_changeupper_info {
2915 struct netdev_notifier_info info; /* must be first */
2916 struct net_device *upper_dev; /* new upper dev */
2917 bool master; /* is upper dev master */
2918 bool linking; /* is the notification for link or unlink */
2919 void *upper_info; /* upper dev info */
2920 };
2921
2922 struct netdev_notifier_changelowerstate_info {
2923 struct netdev_notifier_info info; /* must be first */
2924 void *lower_state_info; /* is lower dev state */
2925 };
2926
2927 struct netdev_notifier_pre_changeaddr_info {
2928 struct netdev_notifier_info info; /* must be first */
2929 const unsigned char *dev_addr;
2930 };
2931
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)2932 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2933 struct net_device *dev)
2934 {
2935 info->dev = dev;
2936 info->extack = NULL;
2937 }
2938
2939 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)2940 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2941 {
2942 return info->dev;
2943 }
2944
2945 static inline struct netlink_ext_ack *
netdev_notifier_info_to_extack(const struct netdev_notifier_info * info)2946 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2947 {
2948 return info->extack;
2949 }
2950
2951 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2952
2953
2954 extern rwlock_t dev_base_lock; /* Device list lock */
2955
2956 #define for_each_netdev(net, d) \
2957 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2958 #define for_each_netdev_reverse(net, d) \
2959 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2960 #define for_each_netdev_rcu(net, d) \
2961 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2962 #define for_each_netdev_safe(net, d, n) \
2963 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2964 #define for_each_netdev_continue(net, d) \
2965 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2966 #define for_each_netdev_continue_reverse(net, d) \
2967 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2968 dev_list)
2969 #define for_each_netdev_continue_rcu(net, d) \
2970 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2971 #define for_each_netdev_in_bond_rcu(bond, slave) \
2972 for_each_netdev_rcu(&init_net, slave) \
2973 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2974 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2975
next_net_device(struct net_device * dev)2976 static inline struct net_device *next_net_device(struct net_device *dev)
2977 {
2978 struct list_head *lh;
2979 struct net *net;
2980
2981 net = dev_net(dev);
2982 lh = dev->dev_list.next;
2983 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2984 }
2985
next_net_device_rcu(struct net_device * dev)2986 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2987 {
2988 struct list_head *lh;
2989 struct net *net;
2990
2991 net = dev_net(dev);
2992 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2993 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2994 }
2995
first_net_device(struct net * net)2996 static inline struct net_device *first_net_device(struct net *net)
2997 {
2998 return list_empty(&net->dev_base_head) ? NULL :
2999 net_device_entry(net->dev_base_head.next);
3000 }
3001
first_net_device_rcu(struct net * net)3002 static inline struct net_device *first_net_device_rcu(struct net *net)
3003 {
3004 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3005
3006 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3007 }
3008
3009 int netdev_boot_setup_check(struct net_device *dev);
3010 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3011 const char *hwaddr);
3012 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3013 void dev_add_pack(struct packet_type *pt);
3014 void dev_remove_pack(struct packet_type *pt);
3015 void __dev_remove_pack(struct packet_type *pt);
3016 void dev_add_offload(struct packet_offload *po);
3017 void dev_remove_offload(struct packet_offload *po);
3018
3019 int dev_get_iflink(const struct net_device *dev);
3020 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3021 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3022 struct net_device_path_stack *stack);
3023 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3024 unsigned short mask);
3025 struct net_device *dev_get_by_name(struct net *net, const char *name);
3026 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3027 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3028 bool netdev_name_in_use(struct net *net, const char *name);
3029 int dev_alloc_name(struct net_device *dev, const char *name);
3030 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3031 void dev_close(struct net_device *dev);
3032 void dev_close_many(struct list_head *head, bool unlink);
3033 void dev_disable_lro(struct net_device *dev);
3034 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3035 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3036 struct net_device *sb_dev);
3037 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3038 struct net_device *sb_dev);
3039
3040 int dev_queue_xmit(struct sk_buff *skb);
3041 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
3042 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3043
dev_direct_xmit(struct sk_buff * skb,u16 queue_id)3044 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3045 {
3046 int ret;
3047
3048 ret = __dev_direct_xmit(skb, queue_id);
3049 if (!dev_xmit_complete(ret))
3050 kfree_skb(skb);
3051 return ret;
3052 }
3053
3054 int register_netdevice(struct net_device *dev);
3055 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3056 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)3057 static inline void unregister_netdevice(struct net_device *dev)
3058 {
3059 unregister_netdevice_queue(dev, NULL);
3060 }
3061
3062 int netdev_refcnt_read(const struct net_device *dev);
3063 void free_netdev(struct net_device *dev);
3064 void netdev_freemem(struct net_device *dev);
3065 int init_dummy_netdev(struct net_device *dev);
3066
3067 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3068 struct sk_buff *skb,
3069 bool all_slaves);
3070 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3071 struct sock *sk);
3072 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3073 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3074 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3075 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3076 int netdev_get_name(struct net *net, char *name, int ifindex);
3077 int dev_restart(struct net_device *dev);
3078 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
3079 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
3080
skb_gro_offset(const struct sk_buff * skb)3081 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
3082 {
3083 return NAPI_GRO_CB(skb)->data_offset;
3084 }
3085
skb_gro_len(const struct sk_buff * skb)3086 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
3087 {
3088 return skb->len - NAPI_GRO_CB(skb)->data_offset;
3089 }
3090
skb_gro_pull(struct sk_buff * skb,unsigned int len)3091 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
3092 {
3093 NAPI_GRO_CB(skb)->data_offset += len;
3094 }
3095
skb_gro_header_fast(struct sk_buff * skb,unsigned int offset)3096 static inline void *skb_gro_header_fast(struct sk_buff *skb,
3097 unsigned int offset)
3098 {
3099 return NAPI_GRO_CB(skb)->frag0 + offset;
3100 }
3101
skb_gro_header_hard(struct sk_buff * skb,unsigned int hlen)3102 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
3103 {
3104 return NAPI_GRO_CB(skb)->frag0_len < hlen;
3105 }
3106
skb_gro_frag0_invalidate(struct sk_buff * skb)3107 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
3108 {
3109 NAPI_GRO_CB(skb)->frag0 = NULL;
3110 NAPI_GRO_CB(skb)->frag0_len = 0;
3111 }
3112
skb_gro_header_slow(struct sk_buff * skb,unsigned int hlen,unsigned int offset)3113 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
3114 unsigned int offset)
3115 {
3116 if (!pskb_may_pull(skb, hlen))
3117 return NULL;
3118
3119 skb_gro_frag0_invalidate(skb);
3120 return skb->data + offset;
3121 }
3122
skb_gro_network_header(struct sk_buff * skb)3123 static inline void *skb_gro_network_header(struct sk_buff *skb)
3124 {
3125 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
3126 skb_network_offset(skb);
3127 }
3128
skb_gro_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3129 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
3130 const void *start, unsigned int len)
3131 {
3132 if (NAPI_GRO_CB(skb)->csum_valid)
3133 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
3134 csum_partial(start, len, 0));
3135 }
3136
3137 /* GRO checksum functions. These are logical equivalents of the normal
3138 * checksum functions (in skbuff.h) except that they operate on the GRO
3139 * offsets and fields in sk_buff.
3140 */
3141
3142 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
3143
skb_at_gro_remcsum_start(struct sk_buff * skb)3144 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
3145 {
3146 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
3147 }
3148
__skb_gro_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)3149 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
3150 bool zero_okay,
3151 __sum16 check)
3152 {
3153 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
3154 skb_checksum_start_offset(skb) <
3155 skb_gro_offset(skb)) &&
3156 !skb_at_gro_remcsum_start(skb) &&
3157 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3158 (!zero_okay || check));
3159 }
3160
__skb_gro_checksum_validate_complete(struct sk_buff * skb,__wsum psum)3161 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
3162 __wsum psum)
3163 {
3164 if (NAPI_GRO_CB(skb)->csum_valid &&
3165 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3166 return 0;
3167
3168 NAPI_GRO_CB(skb)->csum = psum;
3169
3170 return __skb_gro_checksum_complete(skb);
3171 }
3172
skb_gro_incr_csum_unnecessary(struct sk_buff * skb)3173 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3174 {
3175 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3176 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3177 NAPI_GRO_CB(skb)->csum_cnt--;
3178 } else {
3179 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3180 * verified a new top level checksum or an encapsulated one
3181 * during GRO. This saves work if we fallback to normal path.
3182 */
3183 __skb_incr_checksum_unnecessary(skb);
3184 }
3185 }
3186
3187 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3188 compute_pseudo) \
3189 ({ \
3190 __sum16 __ret = 0; \
3191 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3192 __ret = __skb_gro_checksum_validate_complete(skb, \
3193 compute_pseudo(skb, proto)); \
3194 if (!__ret) \
3195 skb_gro_incr_csum_unnecessary(skb); \
3196 __ret; \
3197 })
3198
3199 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3200 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3201
3202 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3203 compute_pseudo) \
3204 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3205
3206 #define skb_gro_checksum_simple_validate(skb) \
3207 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3208
__skb_gro_checksum_convert_check(struct sk_buff * skb)3209 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3210 {
3211 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3212 !NAPI_GRO_CB(skb)->csum_valid);
3213 }
3214
__skb_gro_checksum_convert(struct sk_buff * skb,__wsum pseudo)3215 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3216 __wsum pseudo)
3217 {
3218 NAPI_GRO_CB(skb)->csum = ~pseudo;
3219 NAPI_GRO_CB(skb)->csum_valid = 1;
3220 }
3221
3222 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3223 do { \
3224 if (__skb_gro_checksum_convert_check(skb)) \
3225 __skb_gro_checksum_convert(skb, \
3226 compute_pseudo(skb, proto)); \
3227 } while (0)
3228
3229 struct gro_remcsum {
3230 int offset;
3231 __wsum delta;
3232 };
3233
skb_gro_remcsum_init(struct gro_remcsum * grc)3234 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3235 {
3236 grc->offset = 0;
3237 grc->delta = 0;
3238 }
3239
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)3240 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3241 unsigned int off, size_t hdrlen,
3242 int start, int offset,
3243 struct gro_remcsum *grc,
3244 bool nopartial)
3245 {
3246 __wsum delta;
3247 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3248
3249 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3250
3251 if (!nopartial) {
3252 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3253 return ptr;
3254 }
3255
3256 ptr = skb_gro_header_fast(skb, off);
3257 if (skb_gro_header_hard(skb, off + plen)) {
3258 ptr = skb_gro_header_slow(skb, off + plen, off);
3259 if (!ptr)
3260 return NULL;
3261 }
3262
3263 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3264 start, offset);
3265
3266 /* Adjust skb->csum since we changed the packet */
3267 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3268
3269 grc->offset = off + hdrlen + offset;
3270 grc->delta = delta;
3271
3272 return ptr;
3273 }
3274
skb_gro_remcsum_cleanup(struct sk_buff * skb,struct gro_remcsum * grc)3275 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3276 struct gro_remcsum *grc)
3277 {
3278 void *ptr;
3279 size_t plen = grc->offset + sizeof(u16);
3280
3281 if (!grc->delta)
3282 return;
3283
3284 ptr = skb_gro_header_fast(skb, grc->offset);
3285 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3286 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3287 if (!ptr)
3288 return;
3289 }
3290
3291 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3292 }
3293
3294 #ifdef CONFIG_XFRM_OFFLOAD
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)3295 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3296 {
3297 if (PTR_ERR(pp) != -EINPROGRESS)
3298 NAPI_GRO_CB(skb)->flush |= flush;
3299 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)3300 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3301 struct sk_buff *pp,
3302 int flush,
3303 struct gro_remcsum *grc)
3304 {
3305 if (PTR_ERR(pp) != -EINPROGRESS) {
3306 NAPI_GRO_CB(skb)->flush |= flush;
3307 skb_gro_remcsum_cleanup(skb, grc);
3308 skb->remcsum_offload = 0;
3309 }
3310 }
3311 #else
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)3312 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3313 {
3314 NAPI_GRO_CB(skb)->flush |= flush;
3315 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)3316 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3317 struct sk_buff *pp,
3318 int flush,
3319 struct gro_remcsum *grc)
3320 {
3321 NAPI_GRO_CB(skb)->flush |= flush;
3322 skb_gro_remcsum_cleanup(skb, grc);
3323 skb->remcsum_offload = 0;
3324 }
3325 #endif
3326
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)3327 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3328 unsigned short type,
3329 const void *daddr, const void *saddr,
3330 unsigned int len)
3331 {
3332 if (!dev->header_ops || !dev->header_ops->create)
3333 return 0;
3334
3335 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3336 }
3337
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)3338 static inline int dev_parse_header(const struct sk_buff *skb,
3339 unsigned char *haddr)
3340 {
3341 const struct net_device *dev = skb->dev;
3342
3343 if (!dev->header_ops || !dev->header_ops->parse)
3344 return 0;
3345 return dev->header_ops->parse(skb, haddr);
3346 }
3347
dev_parse_header_protocol(const struct sk_buff * skb)3348 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3349 {
3350 const struct net_device *dev = skb->dev;
3351
3352 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3353 return 0;
3354 return dev->header_ops->parse_protocol(skb);
3355 }
3356
3357 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)3358 static inline bool dev_validate_header(const struct net_device *dev,
3359 char *ll_header, int len)
3360 {
3361 if (likely(len >= dev->hard_header_len))
3362 return true;
3363 if (len < dev->min_header_len)
3364 return false;
3365
3366 if (capable(CAP_SYS_RAWIO)) {
3367 memset(ll_header + len, 0, dev->hard_header_len - len);
3368 return true;
3369 }
3370
3371 if (dev->header_ops && dev->header_ops->validate)
3372 return dev->header_ops->validate(ll_header, len);
3373
3374 return false;
3375 }
3376
dev_has_header(const struct net_device * dev)3377 static inline bool dev_has_header(const struct net_device *dev)
3378 {
3379 return dev->header_ops && dev->header_ops->create;
3380 }
3381
3382 #ifdef CONFIG_NET_FLOW_LIMIT
3383 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3384 struct sd_flow_limit {
3385 u64 count;
3386 unsigned int num_buckets;
3387 unsigned int history_head;
3388 u16 history[FLOW_LIMIT_HISTORY];
3389 u8 buckets[];
3390 };
3391
3392 extern int netdev_flow_limit_table_len;
3393 #endif /* CONFIG_NET_FLOW_LIMIT */
3394
3395 /*
3396 * Incoming packets are placed on per-CPU queues
3397 */
3398 struct softnet_data {
3399 struct list_head poll_list;
3400 struct sk_buff_head process_queue;
3401
3402 /* stats */
3403 unsigned int processed;
3404 unsigned int time_squeeze;
3405 unsigned int received_rps;
3406 #ifdef CONFIG_RPS
3407 struct softnet_data *rps_ipi_list;
3408 #endif
3409 #ifdef CONFIG_NET_FLOW_LIMIT
3410 struct sd_flow_limit __rcu *flow_limit;
3411 #endif
3412 struct Qdisc *output_queue;
3413 struct Qdisc **output_queue_tailp;
3414 struct sk_buff *completion_queue;
3415 #ifdef CONFIG_XFRM_OFFLOAD
3416 struct sk_buff_head xfrm_backlog;
3417 #endif
3418 /* written and read only by owning cpu: */
3419 struct {
3420 u16 recursion;
3421 u8 more;
3422 } xmit;
3423 #ifdef CONFIG_RPS
3424 /* input_queue_head should be written by cpu owning this struct,
3425 * and only read by other cpus. Worth using a cache line.
3426 */
3427 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3428
3429 /* Elements below can be accessed between CPUs for RPS/RFS */
3430 call_single_data_t csd ____cacheline_aligned_in_smp;
3431 struct softnet_data *rps_ipi_next;
3432 unsigned int cpu;
3433 unsigned int input_queue_tail;
3434 #endif
3435 unsigned int dropped;
3436 struct sk_buff_head input_pkt_queue;
3437 struct napi_struct backlog;
3438
3439 };
3440
input_queue_head_incr(struct softnet_data * sd)3441 static inline void input_queue_head_incr(struct softnet_data *sd)
3442 {
3443 #ifdef CONFIG_RPS
3444 sd->input_queue_head++;
3445 #endif
3446 }
3447
input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)3448 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3449 unsigned int *qtail)
3450 {
3451 #ifdef CONFIG_RPS
3452 *qtail = ++sd->input_queue_tail;
3453 #endif
3454 }
3455
3456 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3457
dev_recursion_level(void)3458 static inline int dev_recursion_level(void)
3459 {
3460 return this_cpu_read(softnet_data.xmit.recursion);
3461 }
3462
3463 #define XMIT_RECURSION_LIMIT 8
dev_xmit_recursion(void)3464 static inline bool dev_xmit_recursion(void)
3465 {
3466 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3467 XMIT_RECURSION_LIMIT);
3468 }
3469
dev_xmit_recursion_inc(void)3470 static inline void dev_xmit_recursion_inc(void)
3471 {
3472 __this_cpu_inc(softnet_data.xmit.recursion);
3473 }
3474
dev_xmit_recursion_dec(void)3475 static inline void dev_xmit_recursion_dec(void)
3476 {
3477 __this_cpu_dec(softnet_data.xmit.recursion);
3478 }
3479
3480 void __netif_schedule(struct Qdisc *q);
3481 void netif_schedule_queue(struct netdev_queue *txq);
3482
netif_tx_schedule_all(struct net_device * dev)3483 static inline void netif_tx_schedule_all(struct net_device *dev)
3484 {
3485 unsigned int i;
3486
3487 for (i = 0; i < dev->num_tx_queues; i++)
3488 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3489 }
3490
netif_tx_start_queue(struct netdev_queue * dev_queue)3491 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3492 {
3493 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3494 }
3495
3496 /**
3497 * netif_start_queue - allow transmit
3498 * @dev: network device
3499 *
3500 * Allow upper layers to call the device hard_start_xmit routine.
3501 */
netif_start_queue(struct net_device * dev)3502 static inline void netif_start_queue(struct net_device *dev)
3503 {
3504 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3505 }
3506
netif_tx_start_all_queues(struct net_device * dev)3507 static inline void netif_tx_start_all_queues(struct net_device *dev)
3508 {
3509 unsigned int i;
3510
3511 for (i = 0; i < dev->num_tx_queues; i++) {
3512 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3513 netif_tx_start_queue(txq);
3514 }
3515 }
3516
3517 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3518
3519 /**
3520 * netif_wake_queue - restart transmit
3521 * @dev: network device
3522 *
3523 * Allow upper layers to call the device hard_start_xmit routine.
3524 * Used for flow control when transmit resources are available.
3525 */
netif_wake_queue(struct net_device * dev)3526 static inline void netif_wake_queue(struct net_device *dev)
3527 {
3528 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3529 }
3530
netif_tx_wake_all_queues(struct net_device * dev)3531 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3532 {
3533 unsigned int i;
3534
3535 for (i = 0; i < dev->num_tx_queues; i++) {
3536 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3537 netif_tx_wake_queue(txq);
3538 }
3539 }
3540
netif_tx_stop_queue(struct netdev_queue * dev_queue)3541 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3542 {
3543 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3544 }
3545
3546 /**
3547 * netif_stop_queue - stop transmitted packets
3548 * @dev: network device
3549 *
3550 * Stop upper layers calling the device hard_start_xmit routine.
3551 * Used for flow control when transmit resources are unavailable.
3552 */
netif_stop_queue(struct net_device * dev)3553 static inline void netif_stop_queue(struct net_device *dev)
3554 {
3555 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3556 }
3557
3558 void netif_tx_stop_all_queues(struct net_device *dev);
3559
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)3560 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3561 {
3562 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3563 }
3564
3565 /**
3566 * netif_queue_stopped - test if transmit queue is flowblocked
3567 * @dev: network device
3568 *
3569 * Test if transmit queue on device is currently unable to send.
3570 */
netif_queue_stopped(const struct net_device * dev)3571 static inline bool netif_queue_stopped(const struct net_device *dev)
3572 {
3573 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3574 }
3575
netif_xmit_stopped(const struct netdev_queue * dev_queue)3576 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3577 {
3578 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3579 }
3580
3581 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)3582 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3583 {
3584 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3585 }
3586
3587 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)3588 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3589 {
3590 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3591 }
3592
3593 /**
3594 * netdev_queue_set_dql_min_limit - set dql minimum limit
3595 * @dev_queue: pointer to transmit queue
3596 * @min_limit: dql minimum limit
3597 *
3598 * Forces xmit_more() to return true until the minimum threshold
3599 * defined by @min_limit is reached (or until the tx queue is
3600 * empty). Warning: to be use with care, misuse will impact the
3601 * latency.
3602 */
netdev_queue_set_dql_min_limit(struct netdev_queue * dev_queue,unsigned int min_limit)3603 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3604 unsigned int min_limit)
3605 {
3606 #ifdef CONFIG_BQL
3607 dev_queue->dql.min_limit = min_limit;
3608 #endif
3609 }
3610
3611 /**
3612 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3613 * @dev_queue: pointer to transmit queue
3614 *
3615 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3616 * to give appropriate hint to the CPU.
3617 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)3618 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3619 {
3620 #ifdef CONFIG_BQL
3621 prefetchw(&dev_queue->dql.num_queued);
3622 #endif
3623 }
3624
3625 /**
3626 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3627 * @dev_queue: pointer to transmit queue
3628 *
3629 * BQL enabled drivers might use this helper in their TX completion path,
3630 * to give appropriate hint to the CPU.
3631 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3632 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3633 {
3634 #ifdef CONFIG_BQL
3635 prefetchw(&dev_queue->dql.limit);
3636 #endif
3637 }
3638
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3639 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3640 unsigned int bytes)
3641 {
3642 #ifdef CONFIG_BQL
3643 dql_queued(&dev_queue->dql, bytes);
3644
3645 if (likely(dql_avail(&dev_queue->dql) >= 0))
3646 return;
3647
3648 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3649
3650 /*
3651 * The XOFF flag must be set before checking the dql_avail below,
3652 * because in netdev_tx_completed_queue we update the dql_completed
3653 * before checking the XOFF flag.
3654 */
3655 smp_mb();
3656
3657 /* check again in case another CPU has just made room avail */
3658 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3659 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3660 #endif
3661 }
3662
3663 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3664 * that they should not test BQL status themselves.
3665 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3666 * skb of a batch.
3667 * Returns true if the doorbell must be used to kick the NIC.
3668 */
__netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes,bool xmit_more)3669 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3670 unsigned int bytes,
3671 bool xmit_more)
3672 {
3673 if (xmit_more) {
3674 #ifdef CONFIG_BQL
3675 dql_queued(&dev_queue->dql, bytes);
3676 #endif
3677 return netif_tx_queue_stopped(dev_queue);
3678 }
3679 netdev_tx_sent_queue(dev_queue, bytes);
3680 return true;
3681 }
3682
3683 /**
3684 * netdev_sent_queue - report the number of bytes queued to hardware
3685 * @dev: network device
3686 * @bytes: number of bytes queued to the hardware device queue
3687 *
3688 * Report the number of bytes queued for sending/completion to the network
3689 * device hardware queue. @bytes should be a good approximation and should
3690 * exactly match netdev_completed_queue() @bytes
3691 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)3692 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3693 {
3694 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3695 }
3696
__netdev_sent_queue(struct net_device * dev,unsigned int bytes,bool xmit_more)3697 static inline bool __netdev_sent_queue(struct net_device *dev,
3698 unsigned int bytes,
3699 bool xmit_more)
3700 {
3701 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3702 xmit_more);
3703 }
3704
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3705 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3706 unsigned int pkts, unsigned int bytes)
3707 {
3708 #ifdef CONFIG_BQL
3709 if (unlikely(!bytes))
3710 return;
3711
3712 dql_completed(&dev_queue->dql, bytes);
3713
3714 /*
3715 * Without the memory barrier there is a small possiblity that
3716 * netdev_tx_sent_queue will miss the update and cause the queue to
3717 * be stopped forever
3718 */
3719 smp_mb();
3720
3721 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3722 return;
3723
3724 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3725 netif_schedule_queue(dev_queue);
3726 #endif
3727 }
3728
3729 /**
3730 * netdev_completed_queue - report bytes and packets completed by device
3731 * @dev: network device
3732 * @pkts: actual number of packets sent over the medium
3733 * @bytes: actual number of bytes sent over the medium
3734 *
3735 * Report the number of bytes and packets transmitted by the network device
3736 * hardware queue over the physical medium, @bytes must exactly match the
3737 * @bytes amount passed to netdev_sent_queue()
3738 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3739 static inline void netdev_completed_queue(struct net_device *dev,
3740 unsigned int pkts, unsigned int bytes)
3741 {
3742 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3743 }
3744
netdev_tx_reset_queue(struct netdev_queue * q)3745 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3746 {
3747 #ifdef CONFIG_BQL
3748 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3749 dql_reset(&q->dql);
3750 #endif
3751 }
3752
3753 /**
3754 * netdev_reset_queue - reset the packets and bytes count of a network device
3755 * @dev_queue: network device
3756 *
3757 * Reset the bytes and packet count of a network device and clear the
3758 * software flow control OFF bit for this network device
3759 */
netdev_reset_queue(struct net_device * dev_queue)3760 static inline void netdev_reset_queue(struct net_device *dev_queue)
3761 {
3762 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3763 }
3764
3765 /**
3766 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3767 * @dev: network device
3768 * @queue_index: given tx queue index
3769 *
3770 * Returns 0 if given tx queue index >= number of device tx queues,
3771 * otherwise returns the originally passed tx queue index.
3772 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3773 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3774 {
3775 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3776 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3777 dev->name, queue_index,
3778 dev->real_num_tx_queues);
3779 return 0;
3780 }
3781
3782 return queue_index;
3783 }
3784
3785 /**
3786 * netif_running - test if up
3787 * @dev: network device
3788 *
3789 * Test if the device has been brought up.
3790 */
netif_running(const struct net_device * dev)3791 static inline bool netif_running(const struct net_device *dev)
3792 {
3793 return test_bit(__LINK_STATE_START, &dev->state);
3794 }
3795
3796 /*
3797 * Routines to manage the subqueues on a device. We only need start,
3798 * stop, and a check if it's stopped. All other device management is
3799 * done at the overall netdevice level.
3800 * Also test the device if we're multiqueue.
3801 */
3802
3803 /**
3804 * netif_start_subqueue - allow sending packets on subqueue
3805 * @dev: network device
3806 * @queue_index: sub queue index
3807 *
3808 * Start individual transmit queue of a device with multiple transmit queues.
3809 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)3810 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3811 {
3812 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3813
3814 netif_tx_start_queue(txq);
3815 }
3816
3817 /**
3818 * netif_stop_subqueue - stop sending packets on subqueue
3819 * @dev: network device
3820 * @queue_index: sub queue index
3821 *
3822 * Stop individual transmit queue of a device with multiple transmit queues.
3823 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)3824 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3825 {
3826 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3827 netif_tx_stop_queue(txq);
3828 }
3829
3830 /**
3831 * __netif_subqueue_stopped - test status of subqueue
3832 * @dev: network device
3833 * @queue_index: sub queue index
3834 *
3835 * Check individual transmit queue of a device with multiple transmit queues.
3836 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3837 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3838 u16 queue_index)
3839 {
3840 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3841
3842 return netif_tx_queue_stopped(txq);
3843 }
3844
3845 /**
3846 * netif_subqueue_stopped - test status of subqueue
3847 * @dev: network device
3848 * @skb: sub queue buffer pointer
3849 *
3850 * Check individual transmit queue of a device with multiple transmit queues.
3851 */
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3852 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3853 struct sk_buff *skb)
3854 {
3855 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3856 }
3857
3858 /**
3859 * netif_wake_subqueue - allow sending packets on subqueue
3860 * @dev: network device
3861 * @queue_index: sub queue index
3862 *
3863 * Resume individual transmit queue of a device with multiple transmit queues.
3864 */
netif_wake_subqueue(struct net_device * dev,u16 queue_index)3865 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3866 {
3867 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3868
3869 netif_tx_wake_queue(txq);
3870 }
3871
3872 #ifdef CONFIG_XPS
3873 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3874 u16 index);
3875 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3876 u16 index, enum xps_map_type type);
3877
3878 /**
3879 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3880 * @j: CPU/Rx queue index
3881 * @mask: bitmask of all cpus/rx queues
3882 * @nr_bits: number of bits in the bitmask
3883 *
3884 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3885 */
netif_attr_test_mask(unsigned long j,const unsigned long * mask,unsigned int nr_bits)3886 static inline bool netif_attr_test_mask(unsigned long j,
3887 const unsigned long *mask,
3888 unsigned int nr_bits)
3889 {
3890 cpu_max_bits_warn(j, nr_bits);
3891 return test_bit(j, mask);
3892 }
3893
3894 /**
3895 * netif_attr_test_online - Test for online CPU/Rx queue
3896 * @j: CPU/Rx queue index
3897 * @online_mask: bitmask for CPUs/Rx queues that are online
3898 * @nr_bits: number of bits in the bitmask
3899 *
3900 * Returns true if a CPU/Rx queue is online.
3901 */
netif_attr_test_online(unsigned long j,const unsigned long * online_mask,unsigned int nr_bits)3902 static inline bool netif_attr_test_online(unsigned long j,
3903 const unsigned long *online_mask,
3904 unsigned int nr_bits)
3905 {
3906 cpu_max_bits_warn(j, nr_bits);
3907
3908 if (online_mask)
3909 return test_bit(j, online_mask);
3910
3911 return (j < nr_bits);
3912 }
3913
3914 /**
3915 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3916 * @n: CPU/Rx queue index
3917 * @srcp: the cpumask/Rx queue mask pointer
3918 * @nr_bits: number of bits in the bitmask
3919 *
3920 * Returns >= nr_bits if no further CPUs/Rx queues set.
3921 */
netif_attrmask_next(int n,const unsigned long * srcp,unsigned int nr_bits)3922 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3923 unsigned int nr_bits)
3924 {
3925 /* -1 is a legal arg here. */
3926 if (n != -1)
3927 cpu_max_bits_warn(n, nr_bits);
3928
3929 if (srcp)
3930 return find_next_bit(srcp, nr_bits, n + 1);
3931
3932 return n + 1;
3933 }
3934
3935 /**
3936 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3937 * @n: CPU/Rx queue index
3938 * @src1p: the first CPUs/Rx queues mask pointer
3939 * @src2p: the second CPUs/Rx queues mask pointer
3940 * @nr_bits: number of bits in the bitmask
3941 *
3942 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3943 */
netif_attrmask_next_and(int n,const unsigned long * src1p,const unsigned long * src2p,unsigned int nr_bits)3944 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3945 const unsigned long *src2p,
3946 unsigned int nr_bits)
3947 {
3948 /* -1 is a legal arg here. */
3949 if (n != -1)
3950 cpu_max_bits_warn(n, nr_bits);
3951
3952 if (src1p && src2p)
3953 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3954 else if (src1p)
3955 return find_next_bit(src1p, nr_bits, n + 1);
3956 else if (src2p)
3957 return find_next_bit(src2p, nr_bits, n + 1);
3958
3959 return n + 1;
3960 }
3961 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3962 static inline int netif_set_xps_queue(struct net_device *dev,
3963 const struct cpumask *mask,
3964 u16 index)
3965 {
3966 return 0;
3967 }
3968
__netif_set_xps_queue(struct net_device * dev,const unsigned long * mask,u16 index,enum xps_map_type type)3969 static inline int __netif_set_xps_queue(struct net_device *dev,
3970 const unsigned long *mask,
3971 u16 index, enum xps_map_type type)
3972 {
3973 return 0;
3974 }
3975 #endif
3976
3977 /**
3978 * netif_is_multiqueue - test if device has multiple transmit queues
3979 * @dev: network device
3980 *
3981 * Check if device has multiple transmit queues
3982 */
netif_is_multiqueue(const struct net_device * dev)3983 static inline bool netif_is_multiqueue(const struct net_device *dev)
3984 {
3985 return dev->num_tx_queues > 1;
3986 }
3987
3988 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3989
3990 #ifdef CONFIG_SYSFS
3991 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3992 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxqs)3993 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3994 unsigned int rxqs)
3995 {
3996 dev->real_num_rx_queues = rxqs;
3997 return 0;
3998 }
3999 #endif
4000 int netif_set_real_num_queues(struct net_device *dev,
4001 unsigned int txq, unsigned int rxq);
4002
4003 static inline struct netdev_rx_queue *
__netif_get_rx_queue(struct net_device * dev,unsigned int rxq)4004 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
4005 {
4006 return dev->_rx + rxq;
4007 }
4008
4009 #ifdef CONFIG_SYSFS
get_netdev_rx_queue_index(struct netdev_rx_queue * queue)4010 static inline unsigned int get_netdev_rx_queue_index(
4011 struct netdev_rx_queue *queue)
4012 {
4013 struct net_device *dev = queue->dev;
4014 int index = queue - dev->_rx;
4015
4016 BUG_ON(index >= dev->num_rx_queues);
4017 return index;
4018 }
4019 #endif
4020
4021 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
4022 int netif_get_num_default_rss_queues(void);
4023
4024 enum skb_free_reason {
4025 SKB_REASON_CONSUMED,
4026 SKB_REASON_DROPPED,
4027 };
4028
4029 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
4030 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
4031
4032 /*
4033 * It is not allowed to call kfree_skb() or consume_skb() from hardware
4034 * interrupt context or with hardware interrupts being disabled.
4035 * (in_hardirq() || irqs_disabled())
4036 *
4037 * We provide four helpers that can be used in following contexts :
4038 *
4039 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
4040 * replacing kfree_skb(skb)
4041 *
4042 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
4043 * Typically used in place of consume_skb(skb) in TX completion path
4044 *
4045 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
4046 * replacing kfree_skb(skb)
4047 *
4048 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
4049 * and consumed a packet. Used in place of consume_skb(skb)
4050 */
dev_kfree_skb_irq(struct sk_buff * skb)4051 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
4052 {
4053 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
4054 }
4055
dev_consume_skb_irq(struct sk_buff * skb)4056 static inline void dev_consume_skb_irq(struct sk_buff *skb)
4057 {
4058 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
4059 }
4060
dev_kfree_skb_any(struct sk_buff * skb)4061 static inline void dev_kfree_skb_any(struct sk_buff *skb)
4062 {
4063 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
4064 }
4065
dev_consume_skb_any(struct sk_buff * skb)4066 static inline void dev_consume_skb_any(struct sk_buff *skb)
4067 {
4068 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
4069 }
4070
4071 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
4072 struct bpf_prog *xdp_prog);
4073 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
4074 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
4075 int netif_rx(struct sk_buff *skb);
4076 int netif_rx_ni(struct sk_buff *skb);
4077 int netif_rx_any_context(struct sk_buff *skb);
4078 int netif_receive_skb(struct sk_buff *skb);
4079 int netif_receive_skb_core(struct sk_buff *skb);
4080 void netif_receive_skb_list(struct list_head *head);
4081 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
4082 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
4083 struct sk_buff *napi_get_frags(struct napi_struct *napi);
4084 gro_result_t napi_gro_frags(struct napi_struct *napi);
4085 struct packet_offload *gro_find_receive_by_type(__be16 type);
4086 struct packet_offload *gro_find_complete_by_type(__be16 type);
4087
napi_free_frags(struct napi_struct * napi)4088 static inline void napi_free_frags(struct napi_struct *napi)
4089 {
4090 kfree_skb(napi->skb);
4091 napi->skb = NULL;
4092 }
4093
4094 bool netdev_is_rx_handler_busy(struct net_device *dev);
4095 int netdev_rx_handler_register(struct net_device *dev,
4096 rx_handler_func_t *rx_handler,
4097 void *rx_handler_data);
4098 void netdev_rx_handler_unregister(struct net_device *dev);
4099
4100 bool dev_valid_name(const char *name);
is_socket_ioctl_cmd(unsigned int cmd)4101 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
4102 {
4103 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
4104 }
4105 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
4106 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
4107 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4108 void __user *data, bool *need_copyout);
4109 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4110 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4111 unsigned int dev_get_flags(const struct net_device *);
4112 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4113 struct netlink_ext_ack *extack);
4114 int dev_change_flags(struct net_device *dev, unsigned int flags,
4115 struct netlink_ext_ack *extack);
4116 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
4117 unsigned int gchanges);
4118 int dev_change_name(struct net_device *, const char *);
4119 int dev_set_alias(struct net_device *, const char *, size_t);
4120 int dev_get_alias(const struct net_device *, char *, size_t);
4121 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4122 const char *pat, int new_ifindex);
4123 static inline
dev_change_net_namespace(struct net_device * dev,struct net * net,const char * pat)4124 int dev_change_net_namespace(struct net_device *dev, struct net *net,
4125 const char *pat)
4126 {
4127 return __dev_change_net_namespace(dev, net, pat, 0);
4128 }
4129 int __dev_set_mtu(struct net_device *, int);
4130 int dev_validate_mtu(struct net_device *dev, int mtu,
4131 struct netlink_ext_ack *extack);
4132 int dev_set_mtu_ext(struct net_device *dev, int mtu,
4133 struct netlink_ext_ack *extack);
4134 int dev_set_mtu(struct net_device *, int);
4135 int dev_change_tx_queue_len(struct net_device *, unsigned long);
4136 void dev_set_group(struct net_device *, int);
4137 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4138 struct netlink_ext_ack *extack);
4139 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4140 struct netlink_ext_ack *extack);
4141 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4142 struct netlink_ext_ack *extack);
4143 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4144 int dev_change_carrier(struct net_device *, bool new_carrier);
4145 int dev_get_phys_port_id(struct net_device *dev,
4146 struct netdev_phys_item_id *ppid);
4147 int dev_get_phys_port_name(struct net_device *dev,
4148 char *name, size_t len);
4149 int dev_get_port_parent_id(struct net_device *dev,
4150 struct netdev_phys_item_id *ppid, bool recurse);
4151 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4152 int dev_change_proto_down(struct net_device *dev, bool proto_down);
4153 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
4154 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
4155 u32 value);
4156 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4157 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4158 struct netdev_queue *txq, int *ret);
4159
4160 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
4161 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
4162 int fd, int expected_fd, u32 flags);
4163 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4164 u8 dev_xdp_prog_count(struct net_device *dev);
4165 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4166
4167 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4168 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4169 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4170 bool is_skb_forwardable(const struct net_device *dev,
4171 const struct sk_buff *skb);
4172
__is_skb_forwardable(const struct net_device * dev,const struct sk_buff * skb,const bool check_mtu)4173 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4174 const struct sk_buff *skb,
4175 const bool check_mtu)
4176 {
4177 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4178 unsigned int len;
4179
4180 if (!(dev->flags & IFF_UP))
4181 return false;
4182
4183 if (!check_mtu)
4184 return true;
4185
4186 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4187 if (skb->len <= len)
4188 return true;
4189
4190 /* if TSO is enabled, we don't care about the length as the packet
4191 * could be forwarded without being segmented before
4192 */
4193 if (skb_is_gso(skb))
4194 return true;
4195
4196 return false;
4197 }
4198
____dev_forward_skb(struct net_device * dev,struct sk_buff * skb,const bool check_mtu)4199 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4200 struct sk_buff *skb,
4201 const bool check_mtu)
4202 {
4203 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4204 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4205 atomic_long_inc(&dev->rx_dropped);
4206 kfree_skb(skb);
4207 return NET_RX_DROP;
4208 }
4209
4210 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4211 skb->priority = 0;
4212 return 0;
4213 }
4214
4215 bool dev_nit_active(struct net_device *dev);
4216 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4217
4218 extern int netdev_budget;
4219 extern unsigned int netdev_budget_usecs;
4220
4221 /* Called by rtnetlink.c:rtnl_unlock() */
4222 void netdev_run_todo(void);
4223
4224 /**
4225 * dev_put - release reference to device
4226 * @dev: network device
4227 *
4228 * Release reference to device to allow it to be freed.
4229 */
dev_put(struct net_device * dev)4230 static inline void dev_put(struct net_device *dev)
4231 {
4232 if (dev) {
4233 #ifdef CONFIG_PCPU_DEV_REFCNT
4234 this_cpu_dec(*dev->pcpu_refcnt);
4235 #else
4236 refcount_dec(&dev->dev_refcnt);
4237 #endif
4238 }
4239 }
4240
4241 /**
4242 * dev_hold - get reference to device
4243 * @dev: network device
4244 *
4245 * Hold reference to device to keep it from being freed.
4246 */
dev_hold(struct net_device * dev)4247 static inline void dev_hold(struct net_device *dev)
4248 {
4249 if (dev) {
4250 #ifdef CONFIG_PCPU_DEV_REFCNT
4251 this_cpu_inc(*dev->pcpu_refcnt);
4252 #else
4253 refcount_inc(&dev->dev_refcnt);
4254 #endif
4255 }
4256 }
4257
4258 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4259 * and _off may be called from IRQ context, but it is caller
4260 * who is responsible for serialization of these calls.
4261 *
4262 * The name carrier is inappropriate, these functions should really be
4263 * called netif_lowerlayer_*() because they represent the state of any
4264 * kind of lower layer not just hardware media.
4265 */
4266
4267 void linkwatch_init_dev(struct net_device *dev);
4268 void linkwatch_fire_event(struct net_device *dev);
4269 void linkwatch_forget_dev(struct net_device *dev);
4270
4271 /**
4272 * netif_carrier_ok - test if carrier present
4273 * @dev: network device
4274 *
4275 * Check if carrier is present on device
4276 */
netif_carrier_ok(const struct net_device * dev)4277 static inline bool netif_carrier_ok(const struct net_device *dev)
4278 {
4279 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4280 }
4281
4282 unsigned long dev_trans_start(struct net_device *dev);
4283
4284 void __netdev_watchdog_up(struct net_device *dev);
4285
4286 void netif_carrier_on(struct net_device *dev);
4287 void netif_carrier_off(struct net_device *dev);
4288 void netif_carrier_event(struct net_device *dev);
4289
4290 /**
4291 * netif_dormant_on - mark device as dormant.
4292 * @dev: network device
4293 *
4294 * Mark device as dormant (as per RFC2863).
4295 *
4296 * The dormant state indicates that the relevant interface is not
4297 * actually in a condition to pass packets (i.e., it is not 'up') but is
4298 * in a "pending" state, waiting for some external event. For "on-
4299 * demand" interfaces, this new state identifies the situation where the
4300 * interface is waiting for events to place it in the up state.
4301 */
netif_dormant_on(struct net_device * dev)4302 static inline void netif_dormant_on(struct net_device *dev)
4303 {
4304 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4305 linkwatch_fire_event(dev);
4306 }
4307
4308 /**
4309 * netif_dormant_off - set device as not dormant.
4310 * @dev: network device
4311 *
4312 * Device is not in dormant state.
4313 */
netif_dormant_off(struct net_device * dev)4314 static inline void netif_dormant_off(struct net_device *dev)
4315 {
4316 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4317 linkwatch_fire_event(dev);
4318 }
4319
4320 /**
4321 * netif_dormant - test if device is dormant
4322 * @dev: network device
4323 *
4324 * Check if device is dormant.
4325 */
netif_dormant(const struct net_device * dev)4326 static inline bool netif_dormant(const struct net_device *dev)
4327 {
4328 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4329 }
4330
4331
4332 /**
4333 * netif_testing_on - mark device as under test.
4334 * @dev: network device
4335 *
4336 * Mark device as under test (as per RFC2863).
4337 *
4338 * The testing state indicates that some test(s) must be performed on
4339 * the interface. After completion, of the test, the interface state
4340 * will change to up, dormant, or down, as appropriate.
4341 */
netif_testing_on(struct net_device * dev)4342 static inline void netif_testing_on(struct net_device *dev)
4343 {
4344 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4345 linkwatch_fire_event(dev);
4346 }
4347
4348 /**
4349 * netif_testing_off - set device as not under test.
4350 * @dev: network device
4351 *
4352 * Device is not in testing state.
4353 */
netif_testing_off(struct net_device * dev)4354 static inline void netif_testing_off(struct net_device *dev)
4355 {
4356 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4357 linkwatch_fire_event(dev);
4358 }
4359
4360 /**
4361 * netif_testing - test if device is under test
4362 * @dev: network device
4363 *
4364 * Check if device is under test
4365 */
netif_testing(const struct net_device * dev)4366 static inline bool netif_testing(const struct net_device *dev)
4367 {
4368 return test_bit(__LINK_STATE_TESTING, &dev->state);
4369 }
4370
4371
4372 /**
4373 * netif_oper_up - test if device is operational
4374 * @dev: network device
4375 *
4376 * Check if carrier is operational
4377 */
netif_oper_up(const struct net_device * dev)4378 static inline bool netif_oper_up(const struct net_device *dev)
4379 {
4380 return (dev->operstate == IF_OPER_UP ||
4381 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4382 }
4383
4384 /**
4385 * netif_device_present - is device available or removed
4386 * @dev: network device
4387 *
4388 * Check if device has not been removed from system.
4389 */
netif_device_present(const struct net_device * dev)4390 static inline bool netif_device_present(const struct net_device *dev)
4391 {
4392 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4393 }
4394
4395 void netif_device_detach(struct net_device *dev);
4396
4397 void netif_device_attach(struct net_device *dev);
4398
4399 /*
4400 * Network interface message level settings
4401 */
4402
4403 enum {
4404 NETIF_MSG_DRV_BIT,
4405 NETIF_MSG_PROBE_BIT,
4406 NETIF_MSG_LINK_BIT,
4407 NETIF_MSG_TIMER_BIT,
4408 NETIF_MSG_IFDOWN_BIT,
4409 NETIF_MSG_IFUP_BIT,
4410 NETIF_MSG_RX_ERR_BIT,
4411 NETIF_MSG_TX_ERR_BIT,
4412 NETIF_MSG_TX_QUEUED_BIT,
4413 NETIF_MSG_INTR_BIT,
4414 NETIF_MSG_TX_DONE_BIT,
4415 NETIF_MSG_RX_STATUS_BIT,
4416 NETIF_MSG_PKTDATA_BIT,
4417 NETIF_MSG_HW_BIT,
4418 NETIF_MSG_WOL_BIT,
4419
4420 /* When you add a new bit above, update netif_msg_class_names array
4421 * in net/ethtool/common.c
4422 */
4423 NETIF_MSG_CLASS_COUNT,
4424 };
4425 /* Both ethtool_ops interface and internal driver implementation use u32 */
4426 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4427
4428 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4429 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4430
4431 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4432 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4433 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4434 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4435 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4436 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4437 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4438 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4439 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4440 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4441 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4442 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4443 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4444 #define NETIF_MSG_HW __NETIF_MSG(HW)
4445 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4446
4447 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4448 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4449 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4450 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4451 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4452 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4453 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4454 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4455 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4456 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4457 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4458 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4459 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4460 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4461 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4462
netif_msg_init(int debug_value,int default_msg_enable_bits)4463 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4464 {
4465 /* use default */
4466 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4467 return default_msg_enable_bits;
4468 if (debug_value == 0) /* no output */
4469 return 0;
4470 /* set low N bits */
4471 return (1U << debug_value) - 1;
4472 }
4473
__netif_tx_lock(struct netdev_queue * txq,int cpu)4474 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4475 {
4476 spin_lock(&txq->_xmit_lock);
4477 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4478 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4479 }
4480
__netif_tx_acquire(struct netdev_queue * txq)4481 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4482 {
4483 __acquire(&txq->_xmit_lock);
4484 return true;
4485 }
4486
__netif_tx_release(struct netdev_queue * txq)4487 static inline void __netif_tx_release(struct netdev_queue *txq)
4488 {
4489 __release(&txq->_xmit_lock);
4490 }
4491
__netif_tx_lock_bh(struct netdev_queue * txq)4492 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4493 {
4494 spin_lock_bh(&txq->_xmit_lock);
4495 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4496 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4497 }
4498
__netif_tx_trylock(struct netdev_queue * txq)4499 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4500 {
4501 bool ok = spin_trylock(&txq->_xmit_lock);
4502
4503 if (likely(ok)) {
4504 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4505 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4506 }
4507 return ok;
4508 }
4509
__netif_tx_unlock(struct netdev_queue * txq)4510 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4511 {
4512 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4513 WRITE_ONCE(txq->xmit_lock_owner, -1);
4514 spin_unlock(&txq->_xmit_lock);
4515 }
4516
__netif_tx_unlock_bh(struct netdev_queue * txq)4517 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4518 {
4519 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4520 WRITE_ONCE(txq->xmit_lock_owner, -1);
4521 spin_unlock_bh(&txq->_xmit_lock);
4522 }
4523
txq_trans_update(struct netdev_queue * txq)4524 static inline void txq_trans_update(struct netdev_queue *txq)
4525 {
4526 if (txq->xmit_lock_owner != -1)
4527 txq->trans_start = jiffies;
4528 }
4529
4530 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
netif_trans_update(struct net_device * dev)4531 static inline void netif_trans_update(struct net_device *dev)
4532 {
4533 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4534
4535 if (txq->trans_start != jiffies)
4536 txq->trans_start = jiffies;
4537 }
4538
4539 /**
4540 * netif_tx_lock - grab network device transmit lock
4541 * @dev: network device
4542 *
4543 * Get network device transmit lock
4544 */
netif_tx_lock(struct net_device * dev)4545 static inline void netif_tx_lock(struct net_device *dev)
4546 {
4547 unsigned int i;
4548 int cpu;
4549
4550 spin_lock(&dev->tx_global_lock);
4551 cpu = smp_processor_id();
4552 for (i = 0; i < dev->num_tx_queues; i++) {
4553 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4554
4555 /* We are the only thread of execution doing a
4556 * freeze, but we have to grab the _xmit_lock in
4557 * order to synchronize with threads which are in
4558 * the ->hard_start_xmit() handler and already
4559 * checked the frozen bit.
4560 */
4561 __netif_tx_lock(txq, cpu);
4562 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4563 __netif_tx_unlock(txq);
4564 }
4565 }
4566
netif_tx_lock_bh(struct net_device * dev)4567 static inline void netif_tx_lock_bh(struct net_device *dev)
4568 {
4569 local_bh_disable();
4570 netif_tx_lock(dev);
4571 }
4572
netif_tx_unlock(struct net_device * dev)4573 static inline void netif_tx_unlock(struct net_device *dev)
4574 {
4575 unsigned int i;
4576
4577 for (i = 0; i < dev->num_tx_queues; i++) {
4578 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4579
4580 /* No need to grab the _xmit_lock here. If the
4581 * queue is not stopped for another reason, we
4582 * force a schedule.
4583 */
4584 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4585 netif_schedule_queue(txq);
4586 }
4587 spin_unlock(&dev->tx_global_lock);
4588 }
4589
netif_tx_unlock_bh(struct net_device * dev)4590 static inline void netif_tx_unlock_bh(struct net_device *dev)
4591 {
4592 netif_tx_unlock(dev);
4593 local_bh_enable();
4594 }
4595
4596 #define HARD_TX_LOCK(dev, txq, cpu) { \
4597 if ((dev->features & NETIF_F_LLTX) == 0) { \
4598 __netif_tx_lock(txq, cpu); \
4599 } else { \
4600 __netif_tx_acquire(txq); \
4601 } \
4602 }
4603
4604 #define HARD_TX_TRYLOCK(dev, txq) \
4605 (((dev->features & NETIF_F_LLTX) == 0) ? \
4606 __netif_tx_trylock(txq) : \
4607 __netif_tx_acquire(txq))
4608
4609 #define HARD_TX_UNLOCK(dev, txq) { \
4610 if ((dev->features & NETIF_F_LLTX) == 0) { \
4611 __netif_tx_unlock(txq); \
4612 } else { \
4613 __netif_tx_release(txq); \
4614 } \
4615 }
4616
netif_tx_disable(struct net_device * dev)4617 static inline void netif_tx_disable(struct net_device *dev)
4618 {
4619 unsigned int i;
4620 int cpu;
4621
4622 local_bh_disable();
4623 cpu = smp_processor_id();
4624 spin_lock(&dev->tx_global_lock);
4625 for (i = 0; i < dev->num_tx_queues; i++) {
4626 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4627
4628 __netif_tx_lock(txq, cpu);
4629 netif_tx_stop_queue(txq);
4630 __netif_tx_unlock(txq);
4631 }
4632 spin_unlock(&dev->tx_global_lock);
4633 local_bh_enable();
4634 }
4635
netif_addr_lock(struct net_device * dev)4636 static inline void netif_addr_lock(struct net_device *dev)
4637 {
4638 unsigned char nest_level = 0;
4639
4640 #ifdef CONFIG_LOCKDEP
4641 nest_level = dev->nested_level;
4642 #endif
4643 spin_lock_nested(&dev->addr_list_lock, nest_level);
4644 }
4645
netif_addr_lock_bh(struct net_device * dev)4646 static inline void netif_addr_lock_bh(struct net_device *dev)
4647 {
4648 unsigned char nest_level = 0;
4649
4650 #ifdef CONFIG_LOCKDEP
4651 nest_level = dev->nested_level;
4652 #endif
4653 local_bh_disable();
4654 spin_lock_nested(&dev->addr_list_lock, nest_level);
4655 }
4656
netif_addr_unlock(struct net_device * dev)4657 static inline void netif_addr_unlock(struct net_device *dev)
4658 {
4659 spin_unlock(&dev->addr_list_lock);
4660 }
4661
netif_addr_unlock_bh(struct net_device * dev)4662 static inline void netif_addr_unlock_bh(struct net_device *dev)
4663 {
4664 spin_unlock_bh(&dev->addr_list_lock);
4665 }
4666
4667 /*
4668 * dev_addrs walker. Should be used only for read access. Call with
4669 * rcu_read_lock held.
4670 */
4671 #define for_each_dev_addr(dev, ha) \
4672 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4673
4674 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4675
4676 void ether_setup(struct net_device *dev);
4677
4678 /* Support for loadable net-drivers */
4679 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4680 unsigned char name_assign_type,
4681 void (*setup)(struct net_device *),
4682 unsigned int txqs, unsigned int rxqs);
4683 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4684 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4685
4686 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4687 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4688 count)
4689
4690 int register_netdev(struct net_device *dev);
4691 void unregister_netdev(struct net_device *dev);
4692
4693 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4694
4695 /* General hardware address lists handling functions */
4696 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4697 struct netdev_hw_addr_list *from_list, int addr_len);
4698 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4699 struct netdev_hw_addr_list *from_list, int addr_len);
4700 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4701 struct net_device *dev,
4702 int (*sync)(struct net_device *, const unsigned char *),
4703 int (*unsync)(struct net_device *,
4704 const unsigned char *));
4705 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4706 struct net_device *dev,
4707 int (*sync)(struct net_device *,
4708 const unsigned char *, int),
4709 int (*unsync)(struct net_device *,
4710 const unsigned char *, int));
4711 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4712 struct net_device *dev,
4713 int (*unsync)(struct net_device *,
4714 const unsigned char *, int));
4715 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4716 struct net_device *dev,
4717 int (*unsync)(struct net_device *,
4718 const unsigned char *));
4719 void __hw_addr_init(struct netdev_hw_addr_list *list);
4720
4721 /* Functions used for device addresses handling */
4722 static inline void
__dev_addr_set(struct net_device * dev,const u8 * addr,size_t len)4723 __dev_addr_set(struct net_device *dev, const u8 *addr, size_t len)
4724 {
4725 memcpy(dev->dev_addr, addr, len);
4726 }
4727
dev_addr_set(struct net_device * dev,const u8 * addr)4728 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4729 {
4730 __dev_addr_set(dev, addr, dev->addr_len);
4731 }
4732
4733 static inline void
dev_addr_mod(struct net_device * dev,unsigned int offset,const u8 * addr,size_t len)4734 dev_addr_mod(struct net_device *dev, unsigned int offset,
4735 const u8 *addr, size_t len)
4736 {
4737 memcpy(&dev->dev_addr[offset], addr, len);
4738 }
4739
4740 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4741 unsigned char addr_type);
4742 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4743 unsigned char addr_type);
4744 void dev_addr_flush(struct net_device *dev);
4745 int dev_addr_init(struct net_device *dev);
4746
4747 /* Functions used for unicast addresses handling */
4748 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4749 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4750 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4751 int dev_uc_sync(struct net_device *to, struct net_device *from);
4752 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4753 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4754 void dev_uc_flush(struct net_device *dev);
4755 void dev_uc_init(struct net_device *dev);
4756
4757 /**
4758 * __dev_uc_sync - Synchonize device's unicast list
4759 * @dev: device to sync
4760 * @sync: function to call if address should be added
4761 * @unsync: function to call if address should be removed
4762 *
4763 * Add newly added addresses to the interface, and release
4764 * addresses that have been deleted.
4765 */
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4766 static inline int __dev_uc_sync(struct net_device *dev,
4767 int (*sync)(struct net_device *,
4768 const unsigned char *),
4769 int (*unsync)(struct net_device *,
4770 const unsigned char *))
4771 {
4772 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4773 }
4774
4775 /**
4776 * __dev_uc_unsync - Remove synchronized addresses from device
4777 * @dev: device to sync
4778 * @unsync: function to call if address should be removed
4779 *
4780 * Remove all addresses that were added to the device by dev_uc_sync().
4781 */
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4782 static inline void __dev_uc_unsync(struct net_device *dev,
4783 int (*unsync)(struct net_device *,
4784 const unsigned char *))
4785 {
4786 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4787 }
4788
4789 /* Functions used for multicast addresses handling */
4790 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4791 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4792 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4793 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4794 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4795 int dev_mc_sync(struct net_device *to, struct net_device *from);
4796 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4797 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4798 void dev_mc_flush(struct net_device *dev);
4799 void dev_mc_init(struct net_device *dev);
4800
4801 /**
4802 * __dev_mc_sync - Synchonize device's multicast list
4803 * @dev: device to sync
4804 * @sync: function to call if address should be added
4805 * @unsync: function to call if address should be removed
4806 *
4807 * Add newly added addresses to the interface, and release
4808 * addresses that have been deleted.
4809 */
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4810 static inline int __dev_mc_sync(struct net_device *dev,
4811 int (*sync)(struct net_device *,
4812 const unsigned char *),
4813 int (*unsync)(struct net_device *,
4814 const unsigned char *))
4815 {
4816 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4817 }
4818
4819 /**
4820 * __dev_mc_unsync - Remove synchronized addresses from device
4821 * @dev: device to sync
4822 * @unsync: function to call if address should be removed
4823 *
4824 * Remove all addresses that were added to the device by dev_mc_sync().
4825 */
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4826 static inline void __dev_mc_unsync(struct net_device *dev,
4827 int (*unsync)(struct net_device *,
4828 const unsigned char *))
4829 {
4830 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4831 }
4832
4833 /* Functions used for secondary unicast and multicast support */
4834 void dev_set_rx_mode(struct net_device *dev);
4835 void __dev_set_rx_mode(struct net_device *dev);
4836 int dev_set_promiscuity(struct net_device *dev, int inc);
4837 int dev_set_allmulti(struct net_device *dev, int inc);
4838 void netdev_state_change(struct net_device *dev);
4839 void __netdev_notify_peers(struct net_device *dev);
4840 void netdev_notify_peers(struct net_device *dev);
4841 void netdev_features_change(struct net_device *dev);
4842 /* Load a device via the kmod */
4843 void dev_load(struct net *net, const char *name);
4844 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4845 struct rtnl_link_stats64 *storage);
4846 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4847 const struct net_device_stats *netdev_stats);
4848 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4849 const struct pcpu_sw_netstats __percpu *netstats);
4850 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4851
4852 extern int netdev_max_backlog;
4853 extern int netdev_tstamp_prequeue;
4854 extern int netdev_unregister_timeout_secs;
4855 extern int weight_p;
4856 extern int dev_weight_rx_bias;
4857 extern int dev_weight_tx_bias;
4858 extern int dev_rx_weight;
4859 extern int dev_tx_weight;
4860 extern int gro_normal_batch;
4861
4862 enum {
4863 NESTED_SYNC_IMM_BIT,
4864 NESTED_SYNC_TODO_BIT,
4865 };
4866
4867 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4868 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4869
4870 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4871 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4872
4873 struct netdev_nested_priv {
4874 unsigned char flags;
4875 void *data;
4876 };
4877
4878 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4879 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4880 struct list_head **iter);
4881 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4882 struct list_head **iter);
4883
4884 #ifdef CONFIG_LOCKDEP
4885 static LIST_HEAD(net_unlink_list);
4886
net_unlink_todo(struct net_device * dev)4887 static inline void net_unlink_todo(struct net_device *dev)
4888 {
4889 if (list_empty(&dev->unlink_list))
4890 list_add_tail(&dev->unlink_list, &net_unlink_list);
4891 }
4892 #endif
4893
4894 /* iterate through upper list, must be called under RCU read lock */
4895 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4896 for (iter = &(dev)->adj_list.upper, \
4897 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4898 updev; \
4899 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4900
4901 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4902 int (*fn)(struct net_device *upper_dev,
4903 struct netdev_nested_priv *priv),
4904 struct netdev_nested_priv *priv);
4905
4906 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4907 struct net_device *upper_dev);
4908
4909 bool netdev_has_any_upper_dev(struct net_device *dev);
4910
4911 void *netdev_lower_get_next_private(struct net_device *dev,
4912 struct list_head **iter);
4913 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4914 struct list_head **iter);
4915
4916 #define netdev_for_each_lower_private(dev, priv, iter) \
4917 for (iter = (dev)->adj_list.lower.next, \
4918 priv = netdev_lower_get_next_private(dev, &(iter)); \
4919 priv; \
4920 priv = netdev_lower_get_next_private(dev, &(iter)))
4921
4922 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4923 for (iter = &(dev)->adj_list.lower, \
4924 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4925 priv; \
4926 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4927
4928 void *netdev_lower_get_next(struct net_device *dev,
4929 struct list_head **iter);
4930
4931 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4932 for (iter = (dev)->adj_list.lower.next, \
4933 ldev = netdev_lower_get_next(dev, &(iter)); \
4934 ldev; \
4935 ldev = netdev_lower_get_next(dev, &(iter)))
4936
4937 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4938 struct list_head **iter);
4939 int netdev_walk_all_lower_dev(struct net_device *dev,
4940 int (*fn)(struct net_device *lower_dev,
4941 struct netdev_nested_priv *priv),
4942 struct netdev_nested_priv *priv);
4943 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4944 int (*fn)(struct net_device *lower_dev,
4945 struct netdev_nested_priv *priv),
4946 struct netdev_nested_priv *priv);
4947
4948 void *netdev_adjacent_get_private(struct list_head *adj_list);
4949 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4950 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4951 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4952 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4953 struct netlink_ext_ack *extack);
4954 int netdev_master_upper_dev_link(struct net_device *dev,
4955 struct net_device *upper_dev,
4956 void *upper_priv, void *upper_info,
4957 struct netlink_ext_ack *extack);
4958 void netdev_upper_dev_unlink(struct net_device *dev,
4959 struct net_device *upper_dev);
4960 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4961 struct net_device *new_dev,
4962 struct net_device *dev,
4963 struct netlink_ext_ack *extack);
4964 void netdev_adjacent_change_commit(struct net_device *old_dev,
4965 struct net_device *new_dev,
4966 struct net_device *dev);
4967 void netdev_adjacent_change_abort(struct net_device *old_dev,
4968 struct net_device *new_dev,
4969 struct net_device *dev);
4970 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4971 void *netdev_lower_dev_get_private(struct net_device *dev,
4972 struct net_device *lower_dev);
4973 void netdev_lower_state_changed(struct net_device *lower_dev,
4974 void *lower_state_info);
4975
4976 /* RSS keys are 40 or 52 bytes long */
4977 #define NETDEV_RSS_KEY_LEN 52
4978 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4979 void netdev_rss_key_fill(void *buffer, size_t len);
4980
4981 int skb_checksum_help(struct sk_buff *skb);
4982 int skb_crc32c_csum_help(struct sk_buff *skb);
4983 int skb_csum_hwoffload_help(struct sk_buff *skb,
4984 const netdev_features_t features);
4985
4986 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4987 netdev_features_t features, bool tx_path);
4988 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4989 netdev_features_t features);
4990
4991 struct netdev_bonding_info {
4992 ifslave slave;
4993 ifbond master;
4994 };
4995
4996 struct netdev_notifier_bonding_info {
4997 struct netdev_notifier_info info; /* must be first */
4998 struct netdev_bonding_info bonding_info;
4999 };
5000
5001 void netdev_bonding_info_change(struct net_device *dev,
5002 struct netdev_bonding_info *bonding_info);
5003
5004 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
5005 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
5006 #else
ethtool_notify(struct net_device * dev,unsigned int cmd,const void * data)5007 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
5008 const void *data)
5009 {
5010 }
5011 #endif
5012
5013 static inline
skb_gso_segment(struct sk_buff * skb,netdev_features_t features)5014 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
5015 {
5016 return __skb_gso_segment(skb, features, true);
5017 }
5018 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
5019
can_checksum_protocol(netdev_features_t features,__be16 protocol)5020 static inline bool can_checksum_protocol(netdev_features_t features,
5021 __be16 protocol)
5022 {
5023 if (protocol == htons(ETH_P_FCOE))
5024 return !!(features & NETIF_F_FCOE_CRC);
5025
5026 /* Assume this is an IP checksum (not SCTP CRC) */
5027
5028 if (features & NETIF_F_HW_CSUM) {
5029 /* Can checksum everything */
5030 return true;
5031 }
5032
5033 switch (protocol) {
5034 case htons(ETH_P_IP):
5035 return !!(features & NETIF_F_IP_CSUM);
5036 case htons(ETH_P_IPV6):
5037 return !!(features & NETIF_F_IPV6_CSUM);
5038 default:
5039 return false;
5040 }
5041 }
5042
5043 #ifdef CONFIG_BUG
5044 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
5045 #else
netdev_rx_csum_fault(struct net_device * dev,struct sk_buff * skb)5046 static inline void netdev_rx_csum_fault(struct net_device *dev,
5047 struct sk_buff *skb)
5048 {
5049 }
5050 #endif
5051 /* rx skb timestamps */
5052 void net_enable_timestamp(void);
5053 void net_disable_timestamp(void);
5054
5055 #ifdef CONFIG_PROC_FS
5056 int __init dev_proc_init(void);
5057 #else
5058 #define dev_proc_init() 0
5059 #endif
5060
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)5061 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
5062 struct sk_buff *skb, struct net_device *dev,
5063 bool more)
5064 {
5065 __this_cpu_write(softnet_data.xmit.more, more);
5066 return ops->ndo_start_xmit(skb, dev);
5067 }
5068
netdev_xmit_more(void)5069 static inline bool netdev_xmit_more(void)
5070 {
5071 return __this_cpu_read(softnet_data.xmit.more);
5072 }
5073
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)5074 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
5075 struct netdev_queue *txq, bool more)
5076 {
5077 const struct net_device_ops *ops = dev->netdev_ops;
5078 netdev_tx_t rc;
5079
5080 rc = __netdev_start_xmit(ops, skb, dev, more);
5081 if (rc == NETDEV_TX_OK)
5082 txq_trans_update(txq);
5083
5084 return rc;
5085 }
5086
5087 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
5088 const void *ns);
5089 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
5090 const void *ns);
5091
5092 extern const struct kobj_ns_type_operations net_ns_type_operations;
5093
5094 const char *netdev_drivername(const struct net_device *dev);
5095
5096 void linkwatch_run_queue(void);
5097
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)5098 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
5099 netdev_features_t f2)
5100 {
5101 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
5102 if (f1 & NETIF_F_HW_CSUM)
5103 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5104 else
5105 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5106 }
5107
5108 return f1 & f2;
5109 }
5110
netdev_get_wanted_features(struct net_device * dev)5111 static inline netdev_features_t netdev_get_wanted_features(
5112 struct net_device *dev)
5113 {
5114 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5115 }
5116 netdev_features_t netdev_increment_features(netdev_features_t all,
5117 netdev_features_t one, netdev_features_t mask);
5118
5119 /* Allow TSO being used on stacked device :
5120 * Performing the GSO segmentation before last device
5121 * is a performance improvement.
5122 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)5123 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5124 netdev_features_t mask)
5125 {
5126 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5127 }
5128
5129 int __netdev_update_features(struct net_device *dev);
5130 void netdev_update_features(struct net_device *dev);
5131 void netdev_change_features(struct net_device *dev);
5132
5133 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5134 struct net_device *dev);
5135
5136 netdev_features_t passthru_features_check(struct sk_buff *skb,
5137 struct net_device *dev,
5138 netdev_features_t features);
5139 netdev_features_t netif_skb_features(struct sk_buff *skb);
5140
net_gso_ok(netdev_features_t features,int gso_type)5141 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5142 {
5143 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5144
5145 /* check flags correspondence */
5146 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5147 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5148 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5149 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5150 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5151 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5152 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5153 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5154 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5155 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5156 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5157 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5158 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5159 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5160 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5161 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5162 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5163 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5164 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5165
5166 return (features & feature) == feature;
5167 }
5168
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)5169 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5170 {
5171 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5172 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5173 }
5174
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)5175 static inline bool netif_needs_gso(struct sk_buff *skb,
5176 netdev_features_t features)
5177 {
5178 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5179 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5180 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5181 }
5182
netif_set_gso_max_size(struct net_device * dev,unsigned int size)5183 static inline void netif_set_gso_max_size(struct net_device *dev,
5184 unsigned int size)
5185 {
5186 dev->gso_max_size = size;
5187 }
5188
skb_gso_error_unwind(struct sk_buff * skb,__be16 protocol,int pulled_hlen,u16 mac_offset,int mac_len)5189 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5190 int pulled_hlen, u16 mac_offset,
5191 int mac_len)
5192 {
5193 skb->protocol = protocol;
5194 skb->encapsulation = 1;
5195 skb_push(skb, pulled_hlen);
5196 skb_reset_transport_header(skb);
5197 skb->mac_header = mac_offset;
5198 skb->network_header = skb->mac_header + mac_len;
5199 skb->mac_len = mac_len;
5200 }
5201
netif_is_macsec(const struct net_device * dev)5202 static inline bool netif_is_macsec(const struct net_device *dev)
5203 {
5204 return dev->priv_flags & IFF_MACSEC;
5205 }
5206
netif_is_macvlan(const struct net_device * dev)5207 static inline bool netif_is_macvlan(const struct net_device *dev)
5208 {
5209 return dev->priv_flags & IFF_MACVLAN;
5210 }
5211
netif_is_macvlan_port(const struct net_device * dev)5212 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5213 {
5214 return dev->priv_flags & IFF_MACVLAN_PORT;
5215 }
5216
netif_is_bond_master(const struct net_device * dev)5217 static inline bool netif_is_bond_master(const struct net_device *dev)
5218 {
5219 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5220 }
5221
netif_is_bond_slave(const struct net_device * dev)5222 static inline bool netif_is_bond_slave(const struct net_device *dev)
5223 {
5224 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5225 }
5226
netif_supports_nofcs(struct net_device * dev)5227 static inline bool netif_supports_nofcs(struct net_device *dev)
5228 {
5229 return dev->priv_flags & IFF_SUPP_NOFCS;
5230 }
5231
netif_has_l3_rx_handler(const struct net_device * dev)5232 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5233 {
5234 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5235 }
5236
netif_is_l3_master(const struct net_device * dev)5237 static inline bool netif_is_l3_master(const struct net_device *dev)
5238 {
5239 return dev->priv_flags & IFF_L3MDEV_MASTER;
5240 }
5241
netif_is_l3_slave(const struct net_device * dev)5242 static inline bool netif_is_l3_slave(const struct net_device *dev)
5243 {
5244 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5245 }
5246
dev_sdif(const struct net_device * dev)5247 static inline int dev_sdif(const struct net_device *dev)
5248 {
5249 #ifdef CONFIG_NET_L3_MASTER_DEV
5250 if (netif_is_l3_slave(dev))
5251 return dev->ifindex;
5252 #endif
5253 return 0;
5254 }
5255
netif_is_bridge_master(const struct net_device * dev)5256 static inline bool netif_is_bridge_master(const struct net_device *dev)
5257 {
5258 return dev->priv_flags & IFF_EBRIDGE;
5259 }
5260
netif_is_bridge_port(const struct net_device * dev)5261 static inline bool netif_is_bridge_port(const struct net_device *dev)
5262 {
5263 return dev->priv_flags & IFF_BRIDGE_PORT;
5264 }
5265
netif_is_ovs_master(const struct net_device * dev)5266 static inline bool netif_is_ovs_master(const struct net_device *dev)
5267 {
5268 return dev->priv_flags & IFF_OPENVSWITCH;
5269 }
5270
netif_is_ovs_port(const struct net_device * dev)5271 static inline bool netif_is_ovs_port(const struct net_device *dev)
5272 {
5273 return dev->priv_flags & IFF_OVS_DATAPATH;
5274 }
5275
netif_is_any_bridge_port(const struct net_device * dev)5276 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5277 {
5278 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5279 }
5280
netif_is_team_master(const struct net_device * dev)5281 static inline bool netif_is_team_master(const struct net_device *dev)
5282 {
5283 return dev->priv_flags & IFF_TEAM;
5284 }
5285
netif_is_team_port(const struct net_device * dev)5286 static inline bool netif_is_team_port(const struct net_device *dev)
5287 {
5288 return dev->priv_flags & IFF_TEAM_PORT;
5289 }
5290
netif_is_lag_master(const struct net_device * dev)5291 static inline bool netif_is_lag_master(const struct net_device *dev)
5292 {
5293 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5294 }
5295
netif_is_lag_port(const struct net_device * dev)5296 static inline bool netif_is_lag_port(const struct net_device *dev)
5297 {
5298 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5299 }
5300
netif_is_rxfh_configured(const struct net_device * dev)5301 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5302 {
5303 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5304 }
5305
netif_is_failover(const struct net_device * dev)5306 static inline bool netif_is_failover(const struct net_device *dev)
5307 {
5308 return dev->priv_flags & IFF_FAILOVER;
5309 }
5310
netif_is_failover_slave(const struct net_device * dev)5311 static inline bool netif_is_failover_slave(const struct net_device *dev)
5312 {
5313 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5314 }
5315
5316 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)5317 static inline void netif_keep_dst(struct net_device *dev)
5318 {
5319 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5320 }
5321
5322 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
netif_reduces_vlan_mtu(struct net_device * dev)5323 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5324 {
5325 /* TODO: reserve and use an additional IFF bit, if we get more users */
5326 return dev->priv_flags & IFF_MACSEC;
5327 }
5328
5329 extern struct pernet_operations __net_initdata loopback_net_ops;
5330
5331 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5332
5333 /* netdev_printk helpers, similar to dev_printk */
5334
netdev_name(const struct net_device * dev)5335 static inline const char *netdev_name(const struct net_device *dev)
5336 {
5337 if (!dev->name[0] || strchr(dev->name, '%'))
5338 return "(unnamed net_device)";
5339 return dev->name;
5340 }
5341
netdev_unregistering(const struct net_device * dev)5342 static inline bool netdev_unregistering(const struct net_device *dev)
5343 {
5344 return dev->reg_state == NETREG_UNREGISTERING;
5345 }
5346
netdev_reg_state(const struct net_device * dev)5347 static inline const char *netdev_reg_state(const struct net_device *dev)
5348 {
5349 switch (dev->reg_state) {
5350 case NETREG_UNINITIALIZED: return " (uninitialized)";
5351 case NETREG_REGISTERED: return "";
5352 case NETREG_UNREGISTERING: return " (unregistering)";
5353 case NETREG_UNREGISTERED: return " (unregistered)";
5354 case NETREG_RELEASED: return " (released)";
5355 case NETREG_DUMMY: return " (dummy)";
5356 }
5357
5358 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5359 return " (unknown)";
5360 }
5361
5362 __printf(3, 4) __cold
5363 void netdev_printk(const char *level, const struct net_device *dev,
5364 const char *format, ...);
5365 __printf(2, 3) __cold
5366 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5367 __printf(2, 3) __cold
5368 void netdev_alert(const struct net_device *dev, const char *format, ...);
5369 __printf(2, 3) __cold
5370 void netdev_crit(const struct net_device *dev, const char *format, ...);
5371 __printf(2, 3) __cold
5372 void netdev_err(const struct net_device *dev, const char *format, ...);
5373 __printf(2, 3) __cold
5374 void netdev_warn(const struct net_device *dev, const char *format, ...);
5375 __printf(2, 3) __cold
5376 void netdev_notice(const struct net_device *dev, const char *format, ...);
5377 __printf(2, 3) __cold
5378 void netdev_info(const struct net_device *dev, const char *format, ...);
5379
5380 #define netdev_level_once(level, dev, fmt, ...) \
5381 do { \
5382 static bool __print_once __read_mostly; \
5383 \
5384 if (!__print_once) { \
5385 __print_once = true; \
5386 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5387 } \
5388 } while (0)
5389
5390 #define netdev_emerg_once(dev, fmt, ...) \
5391 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5392 #define netdev_alert_once(dev, fmt, ...) \
5393 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5394 #define netdev_crit_once(dev, fmt, ...) \
5395 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5396 #define netdev_err_once(dev, fmt, ...) \
5397 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5398 #define netdev_warn_once(dev, fmt, ...) \
5399 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5400 #define netdev_notice_once(dev, fmt, ...) \
5401 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5402 #define netdev_info_once(dev, fmt, ...) \
5403 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5404
5405 #define MODULE_ALIAS_NETDEV(device) \
5406 MODULE_ALIAS("netdev-" device)
5407
5408 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5409 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5410 #define netdev_dbg(__dev, format, args...) \
5411 do { \
5412 dynamic_netdev_dbg(__dev, format, ##args); \
5413 } while (0)
5414 #elif defined(DEBUG)
5415 #define netdev_dbg(__dev, format, args...) \
5416 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5417 #else
5418 #define netdev_dbg(__dev, format, args...) \
5419 ({ \
5420 if (0) \
5421 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5422 })
5423 #endif
5424
5425 #if defined(VERBOSE_DEBUG)
5426 #define netdev_vdbg netdev_dbg
5427 #else
5428
5429 #define netdev_vdbg(dev, format, args...) \
5430 ({ \
5431 if (0) \
5432 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5433 0; \
5434 })
5435 #endif
5436
5437 /*
5438 * netdev_WARN() acts like dev_printk(), but with the key difference
5439 * of using a WARN/WARN_ON to get the message out, including the
5440 * file/line information and a backtrace.
5441 */
5442 #define netdev_WARN(dev, format, args...) \
5443 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5444 netdev_reg_state(dev), ##args)
5445
5446 #define netdev_WARN_ONCE(dev, format, args...) \
5447 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5448 netdev_reg_state(dev), ##args)
5449
5450 /* netif printk helpers, similar to netdev_printk */
5451
5452 #define netif_printk(priv, type, level, dev, fmt, args...) \
5453 do { \
5454 if (netif_msg_##type(priv)) \
5455 netdev_printk(level, (dev), fmt, ##args); \
5456 } while (0)
5457
5458 #define netif_level(level, priv, type, dev, fmt, args...) \
5459 do { \
5460 if (netif_msg_##type(priv)) \
5461 netdev_##level(dev, fmt, ##args); \
5462 } while (0)
5463
5464 #define netif_emerg(priv, type, dev, fmt, args...) \
5465 netif_level(emerg, priv, type, dev, fmt, ##args)
5466 #define netif_alert(priv, type, dev, fmt, args...) \
5467 netif_level(alert, priv, type, dev, fmt, ##args)
5468 #define netif_crit(priv, type, dev, fmt, args...) \
5469 netif_level(crit, priv, type, dev, fmt, ##args)
5470 #define netif_err(priv, type, dev, fmt, args...) \
5471 netif_level(err, priv, type, dev, fmt, ##args)
5472 #define netif_warn(priv, type, dev, fmt, args...) \
5473 netif_level(warn, priv, type, dev, fmt, ##args)
5474 #define netif_notice(priv, type, dev, fmt, args...) \
5475 netif_level(notice, priv, type, dev, fmt, ##args)
5476 #define netif_info(priv, type, dev, fmt, args...) \
5477 netif_level(info, priv, type, dev, fmt, ##args)
5478
5479 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5480 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5481 #define netif_dbg(priv, type, netdev, format, args...) \
5482 do { \
5483 if (netif_msg_##type(priv)) \
5484 dynamic_netdev_dbg(netdev, format, ##args); \
5485 } while (0)
5486 #elif defined(DEBUG)
5487 #define netif_dbg(priv, type, dev, format, args...) \
5488 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5489 #else
5490 #define netif_dbg(priv, type, dev, format, args...) \
5491 ({ \
5492 if (0) \
5493 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5494 0; \
5495 })
5496 #endif
5497
5498 /* if @cond then downgrade to debug, else print at @level */
5499 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5500 do { \
5501 if (cond) \
5502 netif_dbg(priv, type, netdev, fmt, ##args); \
5503 else \
5504 netif_ ## level(priv, type, netdev, fmt, ##args); \
5505 } while (0)
5506
5507 #if defined(VERBOSE_DEBUG)
5508 #define netif_vdbg netif_dbg
5509 #else
5510 #define netif_vdbg(priv, type, dev, format, args...) \
5511 ({ \
5512 if (0) \
5513 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5514 0; \
5515 })
5516 #endif
5517
5518 /*
5519 * The list of packet types we will receive (as opposed to discard)
5520 * and the routines to invoke.
5521 *
5522 * Why 16. Because with 16 the only overlap we get on a hash of the
5523 * low nibble of the protocol value is RARP/SNAP/X.25.
5524 *
5525 * 0800 IP
5526 * 0001 802.3
5527 * 0002 AX.25
5528 * 0004 802.2
5529 * 8035 RARP
5530 * 0005 SNAP
5531 * 0805 X.25
5532 * 0806 ARP
5533 * 8137 IPX
5534 * 0009 Localtalk
5535 * 86DD IPv6
5536 */
5537 #define PTYPE_HASH_SIZE (16)
5538 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5539
5540 extern struct list_head ptype_all __read_mostly;
5541 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5542
5543 extern struct net_device *blackhole_netdev;
5544
5545 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5546 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5547 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5548 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5549 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5550
5551 #endif /* _LINUX_NETDEVICE_H */
5552