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