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 #include <asm/local.h>
32
33 #include <linux/percpu.h>
34 #include <linux/rculist.h>
35 #include <linux/workqueue.h>
36 #include <linux/dynamic_queue_limits.h>
37
38 #include <net/net_namespace.h>
39 #ifdef CONFIG_DCB
40 #include <net/dcbnl.h>
41 #endif
42 #include <net/netprio_cgroup.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 <uapi/linux/netdev.h>
50 #include <linux/hashtable.h>
51 #include <linux/rbtree.h>
52 #include <net/net_trackers.h>
53 #include <net/net_debug.h>
54 #include <net/dropreason-core.h>
55 #include <linux/android_kabi.h>
56
57 struct netpoll_info;
58 struct device;
59 struct ethtool_ops;
60 struct kernel_hwtstamp_config;
61 struct phy_device;
62 struct dsa_port;
63 struct ip_tunnel_parm;
64 struct macsec_context;
65 struct macsec_ops;
66 struct netdev_name_node;
67 struct sd_flow_limit;
68 struct sfp_bus;
69 /* 802.11 specific */
70 struct wireless_dev;
71 /* 802.15.4 specific */
72 struct wpan_dev;
73 struct mpls_dev;
74 /* UDP Tunnel offloads */
75 struct udp_tunnel_info;
76 struct udp_tunnel_nic_info;
77 struct udp_tunnel_nic;
78 struct bpf_prog;
79 struct xdp_buff;
80 struct xdp_frame;
81 struct xdp_metadata_ops;
82 struct xdp_md;
83
84 typedef u32 xdp_features_t;
85
86 void synchronize_net(void);
87 void netdev_set_default_ethtool_ops(struct net_device *dev,
88 const struct ethtool_ops *ops);
89 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
90
91 /* Backlog congestion levels */
92 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
93 #define NET_RX_DROP 1 /* packet dropped */
94
95 #define MAX_NEST_DEV 8
96
97 /*
98 * Transmit return codes: transmit return codes originate from three different
99 * namespaces:
100 *
101 * - qdisc return codes
102 * - driver transmit return codes
103 * - errno values
104 *
105 * Drivers are allowed to return any one of those in their hard_start_xmit()
106 * function. Real network devices commonly used with qdiscs should only return
107 * the driver transmit return codes though - when qdiscs are used, the actual
108 * transmission happens asynchronously, so the value is not propagated to
109 * higher layers. Virtual network devices transmit synchronously; in this case
110 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
111 * others are propagated to higher layers.
112 */
113
114 /* qdisc ->enqueue() return codes. */
115 #define NET_XMIT_SUCCESS 0x00
116 #define NET_XMIT_DROP 0x01 /* skb dropped */
117 #define NET_XMIT_CN 0x02 /* congestion notification */
118 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
119
120 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
121 * indicates that the device will soon be dropping packets, or already drops
122 * some packets of the same priority; prompting us to send less aggressively. */
123 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
124 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
125
126 /* Driver transmit return codes */
127 #define NETDEV_TX_MASK 0xf0
128
129 enum netdev_tx {
130 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
131 NETDEV_TX_OK = 0x00, /* driver took care of packet */
132 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
133 };
134 typedef enum netdev_tx netdev_tx_t;
135
136 /*
137 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
138 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
139 */
dev_xmit_complete(int rc)140 static inline bool dev_xmit_complete(int rc)
141 {
142 /*
143 * Positive cases with an skb consumed by a driver:
144 * - successful transmission (rc == NETDEV_TX_OK)
145 * - error while transmitting (rc < 0)
146 * - error while queueing to a different device (rc & NET_XMIT_MASK)
147 */
148 if (likely(rc < NET_XMIT_MASK))
149 return true;
150
151 return false;
152 }
153
154 /*
155 * Compute the worst-case header length according to the protocols
156 * used.
157 */
158
159 #if defined(CONFIG_HYPERV_NET)
160 # define LL_MAX_HEADER 128
161 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
162 # if defined(CONFIG_MAC80211_MESH)
163 # define LL_MAX_HEADER 128
164 # else
165 # define LL_MAX_HEADER 96
166 # endif
167 #else
168 # define LL_MAX_HEADER 32
169 #endif
170
171 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
172 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
173 #define MAX_HEADER LL_MAX_HEADER
174 #else
175 #define MAX_HEADER (LL_MAX_HEADER + 48)
176 #endif
177
178 /*
179 * Old network device statistics. Fields are native words
180 * (unsigned long) so they can be read and written atomically.
181 */
182
183 #define NET_DEV_STAT(FIELD) \
184 union { \
185 unsigned long FIELD; \
186 atomic_long_t __##FIELD; \
187 }
188
189 struct net_device_stats {
190 NET_DEV_STAT(rx_packets);
191 NET_DEV_STAT(tx_packets);
192 NET_DEV_STAT(rx_bytes);
193 NET_DEV_STAT(tx_bytes);
194 NET_DEV_STAT(rx_errors);
195 NET_DEV_STAT(tx_errors);
196 NET_DEV_STAT(rx_dropped);
197 NET_DEV_STAT(tx_dropped);
198 NET_DEV_STAT(multicast);
199 NET_DEV_STAT(collisions);
200 NET_DEV_STAT(rx_length_errors);
201 NET_DEV_STAT(rx_over_errors);
202 NET_DEV_STAT(rx_crc_errors);
203 NET_DEV_STAT(rx_frame_errors);
204 NET_DEV_STAT(rx_fifo_errors);
205 NET_DEV_STAT(rx_missed_errors);
206 NET_DEV_STAT(tx_aborted_errors);
207 NET_DEV_STAT(tx_carrier_errors);
208 NET_DEV_STAT(tx_fifo_errors);
209 NET_DEV_STAT(tx_heartbeat_errors);
210 NET_DEV_STAT(tx_window_errors);
211 NET_DEV_STAT(rx_compressed);
212 NET_DEV_STAT(tx_compressed);
213 };
214 #undef NET_DEV_STAT
215
216 /* per-cpu stats, allocated on demand.
217 * Try to fit them in a single cache line, for dev_get_stats() sake.
218 */
219 struct net_device_core_stats {
220 unsigned long rx_dropped;
221 unsigned long tx_dropped;
222 unsigned long rx_nohandler;
223 unsigned long rx_otherhost_dropped;
224 } __aligned(4 * sizeof(unsigned long));
225
226 #include <linux/cache.h>
227 #include <linux/skbuff.h>
228
229 #ifdef CONFIG_RPS
230 #include <linux/static_key.h>
231 extern struct static_key_false rps_needed;
232 extern struct static_key_false rfs_needed;
233 #endif
234
235 struct neighbour;
236 struct neigh_parms;
237 struct sk_buff;
238
239 struct netdev_hw_addr {
240 struct list_head list;
241 struct rb_node node;
242 unsigned char addr[MAX_ADDR_LEN];
243 unsigned char type;
244 #define NETDEV_HW_ADDR_T_LAN 1
245 #define NETDEV_HW_ADDR_T_SAN 2
246 #define NETDEV_HW_ADDR_T_UNICAST 3
247 #define NETDEV_HW_ADDR_T_MULTICAST 4
248 bool global_use;
249 int sync_cnt;
250 int refcount;
251 int synced;
252 struct rcu_head rcu_head;
253 };
254
255 struct netdev_hw_addr_list {
256 struct list_head list;
257 int count;
258
259 /* Auxiliary tree for faster lookup on addition and deletion */
260 struct rb_root tree;
261 };
262
263 #define netdev_hw_addr_list_count(l) ((l)->count)
264 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
265 #define netdev_hw_addr_list_for_each(ha, l) \
266 list_for_each_entry(ha, &(l)->list, list)
267
268 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
269 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
270 #define netdev_for_each_uc_addr(ha, dev) \
271 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
272 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
273 netdev_for_each_uc_addr((_ha), (_dev)) \
274 if ((_ha)->sync_cnt)
275
276 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
277 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
278 #define netdev_for_each_mc_addr(ha, dev) \
279 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
280 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
281 netdev_for_each_mc_addr((_ha), (_dev)) \
282 if ((_ha)->sync_cnt)
283
284 struct hh_cache {
285 unsigned int hh_len;
286 seqlock_t hh_lock;
287
288 /* cached hardware header; allow for machine alignment needs. */
289 #define HH_DATA_MOD 16
290 #define HH_DATA_OFF(__len) \
291 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
292 #define HH_DATA_ALIGN(__len) \
293 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
294 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
295 };
296
297 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
298 * Alternative is:
299 * dev->hard_header_len ? (dev->hard_header_len +
300 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
301 *
302 * We could use other alignment values, but we must maintain the
303 * relationship HH alignment <= LL alignment.
304 */
305 #define LL_RESERVED_SPACE(dev) \
306 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
307 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
308 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
309 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
310 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
311
312 struct header_ops {
313 int (*create) (struct sk_buff *skb, struct net_device *dev,
314 unsigned short type, const void *daddr,
315 const void *saddr, unsigned int len);
316 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
317 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
318 void (*cache_update)(struct hh_cache *hh,
319 const struct net_device *dev,
320 const unsigned char *haddr);
321 bool (*validate)(const char *ll_header, unsigned int len);
322 __be16 (*parse_protocol)(const struct sk_buff *skb);
323
324 ANDROID_KABI_RESERVE(1);
325 ANDROID_KABI_RESERVE(2);
326 };
327
328 /* These flag bits are private to the generic network queueing
329 * layer; they may not be explicitly referenced by any other
330 * code.
331 */
332
333 enum netdev_state_t {
334 __LINK_STATE_START,
335 __LINK_STATE_PRESENT,
336 __LINK_STATE_NOCARRIER,
337 __LINK_STATE_LINKWATCH_PENDING,
338 __LINK_STATE_DORMANT,
339 __LINK_STATE_TESTING,
340 };
341
342 struct gro_list {
343 struct list_head list;
344 int count;
345 };
346
347 /*
348 * size of gro hash buckets, must less than bit number of
349 * napi_struct::gro_bitmask
350 */
351 #define GRO_HASH_BUCKETS 8
352
353 /*
354 * Structure for NAPI scheduling similar to tasklet but with weighting
355 */
356 struct napi_struct {
357 /* The poll_list must only be managed by the entity which
358 * changes the state of the NAPI_STATE_SCHED bit. This means
359 * whoever atomically sets that bit can add this napi_struct
360 * to the per-CPU poll_list, and whoever clears that bit
361 * can remove from the list right before clearing the bit.
362 */
363 struct list_head poll_list;
364
365 unsigned long state;
366 int weight;
367 int defer_hard_irqs_count;
368 unsigned long gro_bitmask;
369 int (*poll)(struct napi_struct *, int);
370 #ifdef CONFIG_NETPOLL
371 /* CPU actively polling if netpoll is configured */
372 int poll_owner;
373 #endif
374 /* CPU on which NAPI has been scheduled for processing */
375 int list_owner;
376 struct net_device *dev;
377 struct gro_list gro_hash[GRO_HASH_BUCKETS];
378 struct sk_buff *skb;
379 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
380 int rx_count; /* length of rx_list */
381 unsigned int napi_id;
382 struct hrtimer timer;
383 struct task_struct *thread;
384 /* control-path-only fields follow */
385 struct list_head dev_list;
386 struct hlist_node napi_hash_node;
387
388 ANDROID_KABI_RESERVE(1);
389 ANDROID_KABI_RESERVE(2);
390 ANDROID_KABI_RESERVE(3);
391 ANDROID_KABI_RESERVE(4);
392 };
393
394 enum {
395 NAPI_STATE_SCHED, /* Poll is scheduled */
396 NAPI_STATE_MISSED, /* reschedule a napi */
397 NAPI_STATE_DISABLE, /* Disable pending */
398 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
399 NAPI_STATE_LISTED, /* NAPI added to system lists */
400 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
401 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
402 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
403 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
404 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
405 };
406
407 enum {
408 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
409 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
410 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
411 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
412 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
413 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
414 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
415 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
416 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
417 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
418 };
419
420 enum gro_result {
421 GRO_MERGED,
422 GRO_MERGED_FREE,
423 GRO_HELD,
424 GRO_NORMAL,
425 GRO_CONSUMED,
426 };
427 typedef enum gro_result gro_result_t;
428
429 /*
430 * enum rx_handler_result - Possible return values for rx_handlers.
431 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
432 * further.
433 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
434 * case skb->dev was changed by rx_handler.
435 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
436 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
437 *
438 * rx_handlers are functions called from inside __netif_receive_skb(), to do
439 * special processing of the skb, prior to delivery to protocol handlers.
440 *
441 * Currently, a net_device can only have a single rx_handler registered. Trying
442 * to register a second rx_handler will return -EBUSY.
443 *
444 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
445 * To unregister a rx_handler on a net_device, use
446 * netdev_rx_handler_unregister().
447 *
448 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
449 * do with the skb.
450 *
451 * If the rx_handler consumed the skb in some way, it should return
452 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
453 * the skb to be delivered in some other way.
454 *
455 * If the rx_handler changed skb->dev, to divert the skb to another
456 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
457 * new device will be called if it exists.
458 *
459 * If the rx_handler decides the skb should be ignored, it should return
460 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
461 * are registered on exact device (ptype->dev == skb->dev).
462 *
463 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
464 * delivered, it should return RX_HANDLER_PASS.
465 *
466 * A device without a registered rx_handler will behave as if rx_handler
467 * returned RX_HANDLER_PASS.
468 */
469
470 enum rx_handler_result {
471 RX_HANDLER_CONSUMED,
472 RX_HANDLER_ANOTHER,
473 RX_HANDLER_EXACT,
474 RX_HANDLER_PASS,
475 };
476 typedef enum rx_handler_result rx_handler_result_t;
477 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
478
479 void __napi_schedule(struct napi_struct *n);
480 void __napi_schedule_irqoff(struct napi_struct *n);
481
napi_disable_pending(struct napi_struct * n)482 static inline bool napi_disable_pending(struct napi_struct *n)
483 {
484 return test_bit(NAPI_STATE_DISABLE, &n->state);
485 }
486
napi_prefer_busy_poll(struct napi_struct * n)487 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
488 {
489 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
490 }
491
492 bool napi_schedule_prep(struct napi_struct *n);
493
494 /**
495 * napi_schedule - schedule NAPI poll
496 * @n: NAPI context
497 *
498 * Schedule NAPI poll routine to be called if it is not already
499 * running.
500 */
napi_schedule(struct napi_struct * n)501 static inline void napi_schedule(struct napi_struct *n)
502 {
503 if (napi_schedule_prep(n))
504 __napi_schedule(n);
505 }
506
507 /**
508 * napi_schedule_irqoff - schedule NAPI poll
509 * @n: NAPI context
510 *
511 * Variant of napi_schedule(), assuming hard irqs are masked.
512 */
napi_schedule_irqoff(struct napi_struct * n)513 static inline void napi_schedule_irqoff(struct napi_struct *n)
514 {
515 if (napi_schedule_prep(n))
516 __napi_schedule_irqoff(n);
517 }
518
519 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
napi_reschedule(struct napi_struct * napi)520 static inline bool napi_reschedule(struct napi_struct *napi)
521 {
522 if (napi_schedule_prep(napi)) {
523 __napi_schedule(napi);
524 return true;
525 }
526 return false;
527 }
528
529 /**
530 * napi_complete_done - NAPI processing complete
531 * @n: NAPI context
532 * @work_done: number of packets processed
533 *
534 * Mark NAPI processing as complete. Should only be called if poll budget
535 * has not been completely consumed.
536 * Prefer over napi_complete().
537 * Return false if device should avoid rearming interrupts.
538 */
539 bool napi_complete_done(struct napi_struct *n, int work_done);
540
napi_complete(struct napi_struct * n)541 static inline bool napi_complete(struct napi_struct *n)
542 {
543 return napi_complete_done(n, 0);
544 }
545
546 int dev_set_threaded(struct net_device *dev, bool threaded);
547
548 /**
549 * napi_disable - prevent NAPI from scheduling
550 * @n: NAPI context
551 *
552 * Stop NAPI from being scheduled on this context.
553 * Waits till any outstanding processing completes.
554 */
555 void napi_disable(struct napi_struct *n);
556
557 void napi_enable(struct napi_struct *n);
558
559 /**
560 * napi_synchronize - wait until NAPI is not running
561 * @n: NAPI context
562 *
563 * Wait until NAPI is done being scheduled on this context.
564 * Waits till any outstanding processing completes but
565 * does not disable future activations.
566 */
napi_synchronize(const struct napi_struct * n)567 static inline void napi_synchronize(const struct napi_struct *n)
568 {
569 if (IS_ENABLED(CONFIG_SMP))
570 while (test_bit(NAPI_STATE_SCHED, &n->state))
571 msleep(1);
572 else
573 barrier();
574 }
575
576 /**
577 * napi_if_scheduled_mark_missed - if napi is running, set the
578 * NAPIF_STATE_MISSED
579 * @n: NAPI context
580 *
581 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
582 * NAPI is scheduled.
583 **/
napi_if_scheduled_mark_missed(struct napi_struct * n)584 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
585 {
586 unsigned long val, new;
587
588 val = READ_ONCE(n->state);
589 do {
590 if (val & NAPIF_STATE_DISABLE)
591 return true;
592
593 if (!(val & NAPIF_STATE_SCHED))
594 return false;
595
596 new = val | NAPIF_STATE_MISSED;
597 } while (!try_cmpxchg(&n->state, &val, new));
598
599 return true;
600 }
601
602 enum netdev_queue_state_t {
603 __QUEUE_STATE_DRV_XOFF,
604 __QUEUE_STATE_STACK_XOFF,
605 __QUEUE_STATE_FROZEN,
606 };
607
608 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
609 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
610 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
611
612 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
613 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
614 QUEUE_STATE_FROZEN)
615 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
616 QUEUE_STATE_FROZEN)
617
618 /*
619 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
620 * netif_tx_* functions below are used to manipulate this flag. The
621 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
622 * queue independently. The netif_xmit_*stopped functions below are called
623 * to check if the queue has been stopped by the driver or stack (either
624 * of the XOFF bits are set in the state). Drivers should not need to call
625 * netif_xmit*stopped functions, they should only be using netif_tx_*.
626 */
627
628 struct netdev_queue {
629 /*
630 * read-mostly part
631 */
632 struct net_device *dev;
633 netdevice_tracker dev_tracker;
634
635 struct Qdisc __rcu *qdisc;
636 struct Qdisc __rcu *qdisc_sleeping;
637 #ifdef CONFIG_SYSFS
638 struct kobject kobj;
639 #endif
640 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
641 int numa_node;
642 #endif
643 unsigned long tx_maxrate;
644 /*
645 * Number of TX timeouts for this queue
646 * (/sys/class/net/DEV/Q/trans_timeout)
647 */
648 atomic_long_t trans_timeout;
649
650 /* Subordinate device that the queue has been assigned to */
651 struct net_device *sb_dev;
652 #ifdef CONFIG_XDP_SOCKETS
653 struct xsk_buff_pool *pool;
654 #endif
655 /*
656 * write-mostly part
657 */
658 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
659 int xmit_lock_owner;
660 /*
661 * Time (in jiffies) of last Tx
662 */
663 unsigned long trans_start;
664
665 unsigned long state;
666
667 #ifdef CONFIG_BQL
668 struct dql dql;
669 #endif
670
671 ANDROID_KABI_RESERVE(1);
672 ANDROID_KABI_RESERVE(2);
673 ANDROID_KABI_RESERVE(3);
674 ANDROID_KABI_RESERVE(4);
675 } ____cacheline_aligned_in_smp;
676
677 extern int sysctl_fb_tunnels_only_for_init_net;
678 extern int sysctl_devconf_inherit_init_net;
679
680 /*
681 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
682 * == 1 : For initns only
683 * == 2 : For none.
684 */
net_has_fallback_tunnels(const struct net * net)685 static inline bool net_has_fallback_tunnels(const struct net *net)
686 {
687 #if IS_ENABLED(CONFIG_SYSCTL)
688 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
689
690 return !fb_tunnels_only_for_init_net ||
691 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
692 #else
693 return true;
694 #endif
695 }
696
net_inherit_devconf(void)697 static inline int net_inherit_devconf(void)
698 {
699 #if IS_ENABLED(CONFIG_SYSCTL)
700 return READ_ONCE(sysctl_devconf_inherit_init_net);
701 #else
702 return 0;
703 #endif
704 }
705
netdev_queue_numa_node_read(const struct netdev_queue * q)706 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
707 {
708 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
709 return q->numa_node;
710 #else
711 return NUMA_NO_NODE;
712 #endif
713 }
714
netdev_queue_numa_node_write(struct netdev_queue * q,int node)715 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
716 {
717 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
718 q->numa_node = node;
719 #endif
720 }
721
722 #ifdef CONFIG_RPS
723 /*
724 * This structure holds an RPS map which can be of variable length. The
725 * map is an array of CPUs.
726 */
727 struct rps_map {
728 unsigned int len;
729 struct rcu_head rcu;
730 u16 cpus[];
731 };
732 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
733
734 /*
735 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
736 * tail pointer for that CPU's input queue at the time of last enqueue, and
737 * a hardware filter index.
738 */
739 struct rps_dev_flow {
740 u16 cpu;
741 u16 filter;
742 unsigned int last_qtail;
743 };
744 #define RPS_NO_FILTER 0xffff
745
746 /*
747 * The rps_dev_flow_table structure contains a table of flow mappings.
748 */
749 struct rps_dev_flow_table {
750 unsigned int mask;
751 struct rcu_head rcu;
752 struct rps_dev_flow flows[];
753 };
754 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
755 ((_num) * sizeof(struct rps_dev_flow)))
756
757 /*
758 * The rps_sock_flow_table contains mappings of flows to the last CPU
759 * on which they were processed by the application (set in recvmsg).
760 * Each entry is a 32bit value. Upper part is the high-order bits
761 * of flow hash, lower part is CPU number.
762 * rps_cpu_mask is used to partition the space, depending on number of
763 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
764 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
765 * meaning we use 32-6=26 bits for the hash.
766 */
767 struct rps_sock_flow_table {
768 u32 mask;
769
770 u32 ents[] ____cacheline_aligned_in_smp;
771 };
772 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
773
774 #define RPS_NO_CPU 0xffff
775
776 extern u32 rps_cpu_mask;
777 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
778
rps_record_sock_flow(struct rps_sock_flow_table * table,u32 hash)779 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
780 u32 hash)
781 {
782 if (table && hash) {
783 unsigned int index = hash & table->mask;
784 u32 val = hash & ~rps_cpu_mask;
785
786 /* We only give a hint, preemption can change CPU under us */
787 val |= raw_smp_processor_id();
788
789 /* The following WRITE_ONCE() is paired with the READ_ONCE()
790 * here, and another one in get_rps_cpu().
791 */
792 if (READ_ONCE(table->ents[index]) != val)
793 WRITE_ONCE(table->ents[index], val);
794 }
795 }
796
797 #ifdef CONFIG_RFS_ACCEL
798 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
799 u16 filter_id);
800 #endif
801 #endif /* CONFIG_RPS */
802
803 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
804 enum xps_map_type {
805 XPS_CPUS = 0,
806 XPS_RXQS,
807 XPS_MAPS_MAX,
808 };
809
810 #ifdef CONFIG_XPS
811 /*
812 * This structure holds an XPS map which can be of variable length. The
813 * map is an array of queues.
814 */
815 struct xps_map {
816 unsigned int len;
817 unsigned int alloc_len;
818 struct rcu_head rcu;
819 u16 queues[];
820 };
821 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
822 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
823 - sizeof(struct xps_map)) / sizeof(u16))
824
825 /*
826 * This structure holds all XPS maps for device. Maps are indexed by CPU.
827 *
828 * We keep track of the number of cpus/rxqs used when the struct is allocated,
829 * in nr_ids. This will help not accessing out-of-bound memory.
830 *
831 * We keep track of the number of traffic classes used when the struct is
832 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
833 * not crossing its upper bound, as the original dev->num_tc can be updated in
834 * the meantime.
835 */
836 struct xps_dev_maps {
837 struct rcu_head rcu;
838 unsigned int nr_ids;
839 s16 num_tc;
840 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
841 };
842
843 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
844 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
845
846 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
847 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
848
849 #endif /* CONFIG_XPS */
850
851 #define TC_MAX_QUEUE 16
852 #define TC_BITMASK 15
853 /* HW offloaded queuing disciplines txq count and offset maps */
854 struct netdev_tc_txq {
855 u16 count;
856 u16 offset;
857 };
858
859 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
860 /*
861 * This structure is to hold information about the device
862 * configured to run FCoE protocol stack.
863 */
864 struct netdev_fcoe_hbainfo {
865 char manufacturer[64];
866 char serial_number[64];
867 char hardware_version[64];
868 char driver_version[64];
869 char optionrom_version[64];
870 char firmware_version[64];
871 char model[256];
872 char model_description[256];
873 };
874 #endif
875
876 #define MAX_PHYS_ITEM_ID_LEN 32
877
878 /* This structure holds a unique identifier to identify some
879 * physical item (port for example) used by a netdevice.
880 */
881 struct netdev_phys_item_id {
882 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
883 unsigned char id_len;
884 };
885
netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)886 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
887 struct netdev_phys_item_id *b)
888 {
889 return a->id_len == b->id_len &&
890 memcmp(a->id, b->id, a->id_len) == 0;
891 }
892
893 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
894 struct sk_buff *skb,
895 struct net_device *sb_dev);
896
897 enum net_device_path_type {
898 DEV_PATH_ETHERNET = 0,
899 DEV_PATH_VLAN,
900 DEV_PATH_BRIDGE,
901 DEV_PATH_PPPOE,
902 DEV_PATH_DSA,
903 DEV_PATH_MTK_WDMA,
904 };
905
906 struct net_device_path {
907 enum net_device_path_type type;
908 const struct net_device *dev;
909 union {
910 struct {
911 u16 id;
912 __be16 proto;
913 u8 h_dest[ETH_ALEN];
914 } encap;
915 struct {
916 enum {
917 DEV_PATH_BR_VLAN_KEEP,
918 DEV_PATH_BR_VLAN_TAG,
919 DEV_PATH_BR_VLAN_UNTAG,
920 DEV_PATH_BR_VLAN_UNTAG_HW,
921 } vlan_mode;
922 u16 vlan_id;
923 __be16 vlan_proto;
924 } bridge;
925 struct {
926 int port;
927 u16 proto;
928 } dsa;
929 struct {
930 u8 wdma_idx;
931 u8 queue;
932 u16 wcid;
933 u8 bss;
934 } mtk_wdma;
935 };
936 };
937
938 #define NET_DEVICE_PATH_STACK_MAX 5
939 #define NET_DEVICE_PATH_VLAN_MAX 2
940
941 struct net_device_path_stack {
942 int num_paths;
943 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
944 };
945
946 struct net_device_path_ctx {
947 const struct net_device *dev;
948 u8 daddr[ETH_ALEN];
949
950 int num_vlans;
951 struct {
952 u16 id;
953 __be16 proto;
954 } vlan[NET_DEVICE_PATH_VLAN_MAX];
955 };
956
957 enum tc_setup_type {
958 TC_QUERY_CAPS,
959 TC_SETUP_QDISC_MQPRIO,
960 TC_SETUP_CLSU32,
961 TC_SETUP_CLSFLOWER,
962 TC_SETUP_CLSMATCHALL,
963 TC_SETUP_CLSBPF,
964 TC_SETUP_BLOCK,
965 TC_SETUP_QDISC_CBS,
966 TC_SETUP_QDISC_RED,
967 TC_SETUP_QDISC_PRIO,
968 TC_SETUP_QDISC_MQ,
969 TC_SETUP_QDISC_ETF,
970 TC_SETUP_ROOT_QDISC,
971 TC_SETUP_QDISC_GRED,
972 TC_SETUP_QDISC_TAPRIO,
973 TC_SETUP_FT,
974 TC_SETUP_QDISC_ETS,
975 TC_SETUP_QDISC_TBF,
976 TC_SETUP_QDISC_FIFO,
977 TC_SETUP_QDISC_HTB,
978 TC_SETUP_ACT,
979 };
980
981 /* These structures hold the attributes of bpf state that are being passed
982 * to the netdevice through the bpf op.
983 */
984 enum bpf_netdev_command {
985 /* Set or clear a bpf program used in the earliest stages of packet
986 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
987 * is responsible for calling bpf_prog_put on any old progs that are
988 * stored. In case of error, the callee need not release the new prog
989 * reference, but on success it takes ownership and must bpf_prog_put
990 * when it is no longer used.
991 */
992 XDP_SETUP_PROG,
993 XDP_SETUP_PROG_HW,
994 /* BPF program for offload callbacks, invoked at program load time. */
995 BPF_OFFLOAD_MAP_ALLOC,
996 BPF_OFFLOAD_MAP_FREE,
997 XDP_SETUP_XSK_POOL,
998 };
999
1000 struct bpf_prog_offload_ops;
1001 struct netlink_ext_ack;
1002 struct xdp_umem;
1003 struct xdp_dev_bulk_queue;
1004 struct bpf_xdp_link;
1005
1006 enum bpf_xdp_mode {
1007 XDP_MODE_SKB = 0,
1008 XDP_MODE_DRV = 1,
1009 XDP_MODE_HW = 2,
1010 __MAX_XDP_MODE
1011 };
1012
1013 struct bpf_xdp_entity {
1014 struct bpf_prog *prog;
1015 struct bpf_xdp_link *link;
1016 };
1017
1018 struct netdev_bpf {
1019 enum bpf_netdev_command command;
1020 union {
1021 /* XDP_SETUP_PROG */
1022 struct {
1023 u32 flags;
1024 struct bpf_prog *prog;
1025 struct netlink_ext_ack *extack;
1026 };
1027 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1028 struct {
1029 struct bpf_offloaded_map *offmap;
1030 };
1031 /* XDP_SETUP_XSK_POOL */
1032 struct {
1033 struct xsk_buff_pool *pool;
1034 u16 queue_id;
1035 } xsk;
1036 };
1037 };
1038
1039 /* Flags for ndo_xsk_wakeup. */
1040 #define XDP_WAKEUP_RX (1 << 0)
1041 #define XDP_WAKEUP_TX (1 << 1)
1042
1043 #ifdef CONFIG_XFRM_OFFLOAD
1044 struct xfrmdev_ops {
1045 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1046 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1047 void (*xdo_dev_state_free) (struct xfrm_state *x);
1048 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1049 struct xfrm_state *x);
1050 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1051 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1052 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1053 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1054 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1055
1056 ANDROID_KABI_RESERVE(1);
1057 ANDROID_KABI_RESERVE(2);
1058 ANDROID_KABI_RESERVE(3);
1059 ANDROID_KABI_RESERVE(4);
1060 };
1061 #endif
1062
1063 struct dev_ifalias {
1064 struct rcu_head rcuhead;
1065 char ifalias[];
1066 };
1067
1068 struct devlink;
1069 struct tlsdev_ops;
1070
1071 struct netdev_net_notifier {
1072 struct list_head list;
1073 struct notifier_block *nb;
1074 };
1075
1076 /*
1077 * This structure defines the management hooks for network devices.
1078 * The following hooks can be defined; unless noted otherwise, they are
1079 * optional and can be filled with a null pointer.
1080 *
1081 * int (*ndo_init)(struct net_device *dev);
1082 * This function is called once when a network device is registered.
1083 * The network device can use this for any late stage initialization
1084 * or semantic validation. It can fail with an error code which will
1085 * be propagated back to register_netdev.
1086 *
1087 * void (*ndo_uninit)(struct net_device *dev);
1088 * This function is called when device is unregistered or when registration
1089 * fails. It is not called if init fails.
1090 *
1091 * int (*ndo_open)(struct net_device *dev);
1092 * This function is called when a network device transitions to the up
1093 * state.
1094 *
1095 * int (*ndo_stop)(struct net_device *dev);
1096 * This function is called when a network device transitions to the down
1097 * state.
1098 *
1099 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1100 * struct net_device *dev);
1101 * Called when a packet needs to be transmitted.
1102 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1103 * the queue before that can happen; it's for obsolete devices and weird
1104 * corner cases, but the stack really does a non-trivial amount
1105 * of useless work if you return NETDEV_TX_BUSY.
1106 * Required; cannot be NULL.
1107 *
1108 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1109 * struct net_device *dev
1110 * netdev_features_t features);
1111 * Called by core transmit path to determine if device is capable of
1112 * performing offload operations on a given packet. This is to give
1113 * the device an opportunity to implement any restrictions that cannot
1114 * be otherwise expressed by feature flags. The check is called with
1115 * the set of features that the stack has calculated and it returns
1116 * those the driver believes to be appropriate.
1117 *
1118 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1119 * struct net_device *sb_dev);
1120 * Called to decide which queue to use when device supports multiple
1121 * transmit queues.
1122 *
1123 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1124 * This function is called to allow device receiver to make
1125 * changes to configuration when multicast or promiscuous is enabled.
1126 *
1127 * void (*ndo_set_rx_mode)(struct net_device *dev);
1128 * This function is called device changes address list filtering.
1129 * If driver handles unicast address filtering, it should set
1130 * IFF_UNICAST_FLT in its priv_flags.
1131 *
1132 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1133 * This function is called when the Media Access Control address
1134 * needs to be changed. If this interface is not defined, the
1135 * MAC address can not be changed.
1136 *
1137 * int (*ndo_validate_addr)(struct net_device *dev);
1138 * Test if Media Access Control address is valid for the device.
1139 *
1140 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1141 * Old-style ioctl entry point. This is used internally by the
1142 * appletalk and ieee802154 subsystems but is no longer called by
1143 * the device ioctl handler.
1144 *
1145 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1146 * Used by the bonding driver for its device specific ioctls:
1147 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1148 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1149 *
1150 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1151 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1152 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1153 *
1154 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1155 * Used to set network devices bus interface parameters. This interface
1156 * is retained for legacy reasons; new devices should use the bus
1157 * interface (PCI) for low level management.
1158 *
1159 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1160 * Called when a user wants to change the Maximum Transfer Unit
1161 * of a device.
1162 *
1163 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1164 * Callback used when the transmitter has not made any progress
1165 * for dev->watchdog ticks.
1166 *
1167 * void (*ndo_get_stats64)(struct net_device *dev,
1168 * struct rtnl_link_stats64 *storage);
1169 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1170 * Called when a user wants to get the network device usage
1171 * statistics. Drivers must do one of the following:
1172 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1173 * rtnl_link_stats64 structure passed by the caller.
1174 * 2. Define @ndo_get_stats to update a net_device_stats structure
1175 * (which should normally be dev->stats) and return a pointer to
1176 * it. The structure may be changed asynchronously only if each
1177 * field is written atomically.
1178 * 3. Update dev->stats asynchronously and atomically, and define
1179 * neither operation.
1180 *
1181 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1182 * Return true if this device supports offload stats of this attr_id.
1183 *
1184 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1185 * void *attr_data)
1186 * Get statistics for offload operations by attr_id. Write it into the
1187 * attr_data pointer.
1188 *
1189 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1190 * If device supports VLAN filtering this function is called when a
1191 * VLAN id is registered.
1192 *
1193 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1194 * If device supports VLAN filtering this function is called when a
1195 * VLAN id is unregistered.
1196 *
1197 * void (*ndo_poll_controller)(struct net_device *dev);
1198 *
1199 * SR-IOV management functions.
1200 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1201 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1202 * u8 qos, __be16 proto);
1203 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1204 * int max_tx_rate);
1205 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1206 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1207 * int (*ndo_get_vf_config)(struct net_device *dev,
1208 * int vf, struct ifla_vf_info *ivf);
1209 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1210 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1211 * struct nlattr *port[]);
1212 *
1213 * Enable or disable the VF ability to query its RSS Redirection Table and
1214 * Hash Key. This is needed since on some devices VF share this information
1215 * with PF and querying it may introduce a theoretical security risk.
1216 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1217 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1218 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1219 * void *type_data);
1220 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1221 * This is always called from the stack with the rtnl lock held and netif
1222 * tx queues stopped. This allows the netdevice to perform queue
1223 * management safely.
1224 *
1225 * Fiber Channel over Ethernet (FCoE) offload functions.
1226 * int (*ndo_fcoe_enable)(struct net_device *dev);
1227 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1228 * so the underlying device can perform whatever needed configuration or
1229 * initialization to support acceleration of FCoE traffic.
1230 *
1231 * int (*ndo_fcoe_disable)(struct net_device *dev);
1232 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1233 * so the underlying device can perform whatever needed clean-ups to
1234 * stop supporting acceleration of FCoE traffic.
1235 *
1236 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1237 * struct scatterlist *sgl, unsigned int sgc);
1238 * Called when the FCoE Initiator wants to initialize an I/O that
1239 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1240 * perform necessary setup and returns 1 to indicate the device is set up
1241 * successfully to perform DDP on this I/O, otherwise this returns 0.
1242 *
1243 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1244 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1245 * indicated by the FC exchange id 'xid', so the underlying device can
1246 * clean up and reuse resources for later DDP requests.
1247 *
1248 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1249 * struct scatterlist *sgl, unsigned int sgc);
1250 * Called when the FCoE Target wants to initialize an I/O that
1251 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1252 * perform necessary setup and returns 1 to indicate the device is set up
1253 * successfully to perform DDP on this I/O, otherwise this returns 0.
1254 *
1255 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1256 * struct netdev_fcoe_hbainfo *hbainfo);
1257 * Called when the FCoE Protocol stack wants information on the underlying
1258 * device. This information is utilized by the FCoE protocol stack to
1259 * register attributes with Fiber Channel management service as per the
1260 * FC-GS Fabric Device Management Information(FDMI) specification.
1261 *
1262 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1263 * Called when the underlying device wants to override default World Wide
1264 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1265 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1266 * protocol stack to use.
1267 *
1268 * RFS acceleration.
1269 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1270 * u16 rxq_index, u32 flow_id);
1271 * Set hardware filter for RFS. rxq_index is the target queue index;
1272 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1273 * Return the filter ID on success, or a negative error code.
1274 *
1275 * Slave management functions (for bridge, bonding, etc).
1276 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1277 * Called to make another netdev an underling.
1278 *
1279 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1280 * Called to release previously enslaved netdev.
1281 *
1282 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1283 * struct sk_buff *skb,
1284 * bool all_slaves);
1285 * Get the xmit slave of master device. If all_slaves is true, function
1286 * assume all the slaves can transmit.
1287 *
1288 * Feature/offload setting functions.
1289 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1290 * netdev_features_t features);
1291 * Adjusts the requested feature flags according to device-specific
1292 * constraints, and returns the resulting flags. Must not modify
1293 * the device state.
1294 *
1295 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1296 * Called to update device configuration to new features. Passed
1297 * feature set might be less than what was returned by ndo_fix_features()).
1298 * Must return >0 or -errno if it changed dev->features itself.
1299 *
1300 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1301 * struct net_device *dev,
1302 * const unsigned char *addr, u16 vid, u16 flags,
1303 * struct netlink_ext_ack *extack);
1304 * Adds an FDB entry to dev for addr.
1305 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1306 * struct net_device *dev,
1307 * const unsigned char *addr, u16 vid)
1308 * Deletes the FDB entry from dev coresponding to addr.
1309 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1310 * struct net_device *dev,
1311 * u16 vid,
1312 * struct netlink_ext_ack *extack);
1313 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1314 * struct net_device *dev, struct net_device *filter_dev,
1315 * int *idx)
1316 * Used to add FDB entries to dump requests. Implementers should add
1317 * entries to skb and update idx with the number of entries.
1318 *
1319 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1320 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1321 * Adds an MDB entry to dev.
1322 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1323 * struct netlink_ext_ack *extack);
1324 * Deletes the MDB entry from dev.
1325 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1326 * struct netlink_callback *cb);
1327 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1328 * callback is used by core rtnetlink code.
1329 *
1330 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1331 * u16 flags, struct netlink_ext_ack *extack)
1332 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1333 * struct net_device *dev, u32 filter_mask,
1334 * int nlflags)
1335 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1336 * u16 flags);
1337 *
1338 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1339 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1340 * which do not represent real hardware may define this to allow their
1341 * userspace components to manage their virtual carrier state. Devices
1342 * that determine carrier state from physical hardware properties (eg
1343 * network cables) or protocol-dependent mechanisms (eg
1344 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1345 *
1346 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1347 * struct netdev_phys_item_id *ppid);
1348 * Called to get ID of physical port of this device. If driver does
1349 * not implement this, it is assumed that the hw is not able to have
1350 * multiple net devices on single physical port.
1351 *
1352 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1353 * struct netdev_phys_item_id *ppid)
1354 * Called to get the parent ID of the physical port of this device.
1355 *
1356 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1357 * struct net_device *dev)
1358 * Called by upper layer devices to accelerate switching or other
1359 * station functionality into hardware. 'pdev is the lowerdev
1360 * to use for the offload and 'dev' is the net device that will
1361 * back the offload. Returns a pointer to the private structure
1362 * the upper layer will maintain.
1363 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1364 * Called by upper layer device to delete the station created
1365 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1366 * the station and priv is the structure returned by the add
1367 * operation.
1368 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1369 * int queue_index, u32 maxrate);
1370 * Called when a user wants to set a max-rate limitation of specific
1371 * TX queue.
1372 * int (*ndo_get_iflink)(const struct net_device *dev);
1373 * Called to get the iflink value of this device.
1374 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1375 * This function is used to get egress tunnel information for given skb.
1376 * This is useful for retrieving outer tunnel header parameters while
1377 * sampling packet.
1378 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1379 * This function is used to specify the headroom that the skb must
1380 * consider when allocation skb during packet reception. Setting
1381 * appropriate rx headroom value allows avoiding skb head copy on
1382 * forward. Setting a negative value resets the rx headroom to the
1383 * default value.
1384 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1385 * This function is used to set or query state related to XDP on the
1386 * netdevice and manage BPF offload. See definition of
1387 * enum bpf_netdev_command for details.
1388 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1389 * u32 flags);
1390 * This function is used to submit @n XDP packets for transmit on a
1391 * netdevice. Returns number of frames successfully transmitted, frames
1392 * that got dropped are freed/returned via xdp_return_frame().
1393 * Returns negative number, means general error invoking ndo, meaning
1394 * no frames were xmit'ed and core-caller will free all frames.
1395 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1396 * struct xdp_buff *xdp);
1397 * Get the xmit slave of master device based on the xdp_buff.
1398 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1399 * This function is used to wake up the softirq, ksoftirqd or kthread
1400 * responsible for sending and/or receiving packets on a specific
1401 * queue id bound to an AF_XDP socket. The flags field specifies if
1402 * only RX, only Tx, or both should be woken up using the flags
1403 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1404 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1405 * int cmd);
1406 * Add, change, delete or get information on an IPv4 tunnel.
1407 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1408 * If a device is paired with a peer device, return the peer instance.
1409 * The caller must be under RCU read context.
1410 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1411 * Get the forwarding path to reach the real device from the HW destination address
1412 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1413 * const struct skb_shared_hwtstamps *hwtstamps,
1414 * bool cycles);
1415 * Get hardware timestamp based on normal/adjustable time or free running
1416 * cycle counter. This function is required if physical clock supports a
1417 * free running cycle counter.
1418 *
1419 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1420 * struct kernel_hwtstamp_config *kernel_config);
1421 * Get the currently configured hardware timestamping parameters for the
1422 * NIC device.
1423 *
1424 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1425 * struct kernel_hwtstamp_config *kernel_config,
1426 * struct netlink_ext_ack *extack);
1427 * Change the hardware timestamping parameters for NIC device.
1428 */
1429 struct net_device_ops {
1430 int (*ndo_init)(struct net_device *dev);
1431 void (*ndo_uninit)(struct net_device *dev);
1432 int (*ndo_open)(struct net_device *dev);
1433 int (*ndo_stop)(struct net_device *dev);
1434 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1435 struct net_device *dev);
1436 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1437 struct net_device *dev,
1438 netdev_features_t features);
1439 u16 (*ndo_select_queue)(struct net_device *dev,
1440 struct sk_buff *skb,
1441 struct net_device *sb_dev);
1442 void (*ndo_change_rx_flags)(struct net_device *dev,
1443 int flags);
1444 void (*ndo_set_rx_mode)(struct net_device *dev);
1445 int (*ndo_set_mac_address)(struct net_device *dev,
1446 void *addr);
1447 int (*ndo_validate_addr)(struct net_device *dev);
1448 int (*ndo_do_ioctl)(struct net_device *dev,
1449 struct ifreq *ifr, int cmd);
1450 int (*ndo_eth_ioctl)(struct net_device *dev,
1451 struct ifreq *ifr, int cmd);
1452 int (*ndo_siocbond)(struct net_device *dev,
1453 struct ifreq *ifr, int cmd);
1454 int (*ndo_siocwandev)(struct net_device *dev,
1455 struct if_settings *ifs);
1456 int (*ndo_siocdevprivate)(struct net_device *dev,
1457 struct ifreq *ifr,
1458 void __user *data, int cmd);
1459 int (*ndo_set_config)(struct net_device *dev,
1460 struct ifmap *map);
1461 int (*ndo_change_mtu)(struct net_device *dev,
1462 int new_mtu);
1463 int (*ndo_neigh_setup)(struct net_device *dev,
1464 struct neigh_parms *);
1465 void (*ndo_tx_timeout) (struct net_device *dev,
1466 unsigned int txqueue);
1467
1468 void (*ndo_get_stats64)(struct net_device *dev,
1469 struct rtnl_link_stats64 *storage);
1470 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1471 int (*ndo_get_offload_stats)(int attr_id,
1472 const struct net_device *dev,
1473 void *attr_data);
1474 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1475
1476 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1477 __be16 proto, u16 vid);
1478 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1479 __be16 proto, u16 vid);
1480 #ifdef CONFIG_NET_POLL_CONTROLLER
1481 void (*ndo_poll_controller)(struct net_device *dev);
1482 int (*ndo_netpoll_setup)(struct net_device *dev,
1483 struct netpoll_info *info);
1484 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1485 #endif
1486 int (*ndo_set_vf_mac)(struct net_device *dev,
1487 int queue, u8 *mac);
1488 int (*ndo_set_vf_vlan)(struct net_device *dev,
1489 int queue, u16 vlan,
1490 u8 qos, __be16 proto);
1491 int (*ndo_set_vf_rate)(struct net_device *dev,
1492 int vf, int min_tx_rate,
1493 int max_tx_rate);
1494 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1495 int vf, bool setting);
1496 int (*ndo_set_vf_trust)(struct net_device *dev,
1497 int vf, bool setting);
1498 int (*ndo_get_vf_config)(struct net_device *dev,
1499 int vf,
1500 struct ifla_vf_info *ivf);
1501 int (*ndo_set_vf_link_state)(struct net_device *dev,
1502 int vf, int link_state);
1503 int (*ndo_get_vf_stats)(struct net_device *dev,
1504 int vf,
1505 struct ifla_vf_stats
1506 *vf_stats);
1507 int (*ndo_set_vf_port)(struct net_device *dev,
1508 int vf,
1509 struct nlattr *port[]);
1510 int (*ndo_get_vf_port)(struct net_device *dev,
1511 int vf, struct sk_buff *skb);
1512 int (*ndo_get_vf_guid)(struct net_device *dev,
1513 int vf,
1514 struct ifla_vf_guid *node_guid,
1515 struct ifla_vf_guid *port_guid);
1516 int (*ndo_set_vf_guid)(struct net_device *dev,
1517 int vf, u64 guid,
1518 int guid_type);
1519 int (*ndo_set_vf_rss_query_en)(
1520 struct net_device *dev,
1521 int vf, bool setting);
1522 int (*ndo_setup_tc)(struct net_device *dev,
1523 enum tc_setup_type type,
1524 void *type_data);
1525 #if IS_ENABLED(CONFIG_FCOE)
1526 int (*ndo_fcoe_enable)(struct net_device *dev);
1527 int (*ndo_fcoe_disable)(struct net_device *dev);
1528 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1529 u16 xid,
1530 struct scatterlist *sgl,
1531 unsigned int sgc);
1532 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1533 u16 xid);
1534 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1535 u16 xid,
1536 struct scatterlist *sgl,
1537 unsigned int sgc);
1538 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1539 struct netdev_fcoe_hbainfo *hbainfo);
1540 #endif
1541
1542 #if IS_ENABLED(CONFIG_LIBFCOE)
1543 #define NETDEV_FCOE_WWNN 0
1544 #define NETDEV_FCOE_WWPN 1
1545 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1546 u64 *wwn, int type);
1547 #endif
1548
1549 #ifdef CONFIG_RFS_ACCEL
1550 int (*ndo_rx_flow_steer)(struct net_device *dev,
1551 const struct sk_buff *skb,
1552 u16 rxq_index,
1553 u32 flow_id);
1554 #endif
1555 int (*ndo_add_slave)(struct net_device *dev,
1556 struct net_device *slave_dev,
1557 struct netlink_ext_ack *extack);
1558 int (*ndo_del_slave)(struct net_device *dev,
1559 struct net_device *slave_dev);
1560 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1561 struct sk_buff *skb,
1562 bool all_slaves);
1563 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1564 struct sock *sk);
1565 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1566 netdev_features_t features);
1567 int (*ndo_set_features)(struct net_device *dev,
1568 netdev_features_t features);
1569 int (*ndo_neigh_construct)(struct net_device *dev,
1570 struct neighbour *n);
1571 void (*ndo_neigh_destroy)(struct net_device *dev,
1572 struct neighbour *n);
1573
1574 int (*ndo_fdb_add)(struct ndmsg *ndm,
1575 struct nlattr *tb[],
1576 struct net_device *dev,
1577 const unsigned char *addr,
1578 u16 vid,
1579 u16 flags,
1580 struct netlink_ext_ack *extack);
1581 int (*ndo_fdb_del)(struct ndmsg *ndm,
1582 struct nlattr *tb[],
1583 struct net_device *dev,
1584 const unsigned char *addr,
1585 u16 vid, struct netlink_ext_ack *extack);
1586 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1587 struct nlattr *tb[],
1588 struct net_device *dev,
1589 u16 vid,
1590 struct netlink_ext_ack *extack);
1591 int (*ndo_fdb_dump)(struct sk_buff *skb,
1592 struct netlink_callback *cb,
1593 struct net_device *dev,
1594 struct net_device *filter_dev,
1595 int *idx);
1596 int (*ndo_fdb_get)(struct sk_buff *skb,
1597 struct nlattr *tb[],
1598 struct net_device *dev,
1599 const unsigned char *addr,
1600 u16 vid, u32 portid, u32 seq,
1601 struct netlink_ext_ack *extack);
1602 int (*ndo_mdb_add)(struct net_device *dev,
1603 struct nlattr *tb[],
1604 u16 nlmsg_flags,
1605 struct netlink_ext_ack *extack);
1606 int (*ndo_mdb_del)(struct net_device *dev,
1607 struct nlattr *tb[],
1608 struct netlink_ext_ack *extack);
1609 int (*ndo_mdb_dump)(struct net_device *dev,
1610 struct sk_buff *skb,
1611 struct netlink_callback *cb);
1612 int (*ndo_bridge_setlink)(struct net_device *dev,
1613 struct nlmsghdr *nlh,
1614 u16 flags,
1615 struct netlink_ext_ack *extack);
1616 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1617 u32 pid, u32 seq,
1618 struct net_device *dev,
1619 u32 filter_mask,
1620 int nlflags);
1621 int (*ndo_bridge_dellink)(struct net_device *dev,
1622 struct nlmsghdr *nlh,
1623 u16 flags);
1624 int (*ndo_change_carrier)(struct net_device *dev,
1625 bool new_carrier);
1626 int (*ndo_get_phys_port_id)(struct net_device *dev,
1627 struct netdev_phys_item_id *ppid);
1628 int (*ndo_get_port_parent_id)(struct net_device *dev,
1629 struct netdev_phys_item_id *ppid);
1630 int (*ndo_get_phys_port_name)(struct net_device *dev,
1631 char *name, size_t len);
1632 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1633 struct net_device *dev);
1634 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1635 void *priv);
1636
1637 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1638 int queue_index,
1639 u32 maxrate);
1640 int (*ndo_get_iflink)(const struct net_device *dev);
1641 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1642 struct sk_buff *skb);
1643 void (*ndo_set_rx_headroom)(struct net_device *dev,
1644 int needed_headroom);
1645 int (*ndo_bpf)(struct net_device *dev,
1646 struct netdev_bpf *bpf);
1647 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1648 struct xdp_frame **xdp,
1649 u32 flags);
1650 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1651 struct xdp_buff *xdp);
1652 int (*ndo_xsk_wakeup)(struct net_device *dev,
1653 u32 queue_id, u32 flags);
1654 int (*ndo_tunnel_ctl)(struct net_device *dev,
1655 struct ip_tunnel_parm *p, int cmd);
1656 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1657 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1658 struct net_device_path *path);
1659 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1660 const struct skb_shared_hwtstamps *hwtstamps,
1661 bool cycles);
1662 int (*ndo_hwtstamp_get)(struct net_device *dev,
1663 struct kernel_hwtstamp_config *kernel_config);
1664 int (*ndo_hwtstamp_set)(struct net_device *dev,
1665 struct kernel_hwtstamp_config *kernel_config,
1666 struct netlink_ext_ack *extack);
1667 ANDROID_KABI_RESERVE(1);
1668 ANDROID_KABI_RESERVE(2);
1669 ANDROID_KABI_RESERVE(3);
1670 ANDROID_KABI_RESERVE(4);
1671 ANDROID_KABI_RESERVE(5);
1672 ANDROID_KABI_RESERVE(6);
1673 ANDROID_KABI_RESERVE(7);
1674 ANDROID_KABI_RESERVE(8);
1675 };
1676
1677 /**
1678 * enum netdev_priv_flags - &struct net_device priv_flags
1679 *
1680 * These are the &struct net_device, they are only set internally
1681 * by drivers and used in the kernel. These flags are invisible to
1682 * userspace; this means that the order of these flags can change
1683 * during any kernel release.
1684 *
1685 * You should have a pretty good reason to be extending these flags.
1686 *
1687 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1688 * @IFF_EBRIDGE: Ethernet bridging device
1689 * @IFF_BONDING: bonding master or slave
1690 * @IFF_ISATAP: ISATAP interface (RFC4214)
1691 * @IFF_WAN_HDLC: WAN HDLC device
1692 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1693 * release skb->dst
1694 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1695 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1696 * @IFF_MACVLAN_PORT: device used as macvlan port
1697 * @IFF_BRIDGE_PORT: device used as bridge port
1698 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1699 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1700 * @IFF_UNICAST_FLT: Supports unicast filtering
1701 * @IFF_TEAM_PORT: device used as team port
1702 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1703 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1704 * change when it's running
1705 * @IFF_MACVLAN: Macvlan device
1706 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1707 * underlying stacked devices
1708 * @IFF_L3MDEV_MASTER: device is an L3 master device
1709 * @IFF_NO_QUEUE: device can run without qdisc attached
1710 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1711 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1712 * @IFF_TEAM: device is a team device
1713 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1714 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1715 * entity (i.e. the master device for bridged veth)
1716 * @IFF_MACSEC: device is a MACsec device
1717 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1718 * @IFF_FAILOVER: device is a failover master device
1719 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1720 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1721 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1722 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1723 * skb_headlen(skb) == 0 (data starts from frag0)
1724 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1725 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1726 * ndo_hwtstamp_set() for all timestamp requests regardless of source,
1727 * even if those aren't HWTSTAMP_SOURCE_NETDEV.
1728 */
1729 enum netdev_priv_flags {
1730 IFF_802_1Q_VLAN = 1<<0,
1731 IFF_EBRIDGE = 1<<1,
1732 IFF_BONDING = 1<<2,
1733 IFF_ISATAP = 1<<3,
1734 IFF_WAN_HDLC = 1<<4,
1735 IFF_XMIT_DST_RELEASE = 1<<5,
1736 IFF_DONT_BRIDGE = 1<<6,
1737 IFF_DISABLE_NETPOLL = 1<<7,
1738 IFF_MACVLAN_PORT = 1<<8,
1739 IFF_BRIDGE_PORT = 1<<9,
1740 IFF_OVS_DATAPATH = 1<<10,
1741 IFF_TX_SKB_SHARING = 1<<11,
1742 IFF_UNICAST_FLT = 1<<12,
1743 IFF_TEAM_PORT = 1<<13,
1744 IFF_SUPP_NOFCS = 1<<14,
1745 IFF_LIVE_ADDR_CHANGE = 1<<15,
1746 IFF_MACVLAN = 1<<16,
1747 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1748 IFF_L3MDEV_MASTER = 1<<18,
1749 IFF_NO_QUEUE = 1<<19,
1750 IFF_OPENVSWITCH = 1<<20,
1751 IFF_L3MDEV_SLAVE = 1<<21,
1752 IFF_TEAM = 1<<22,
1753 IFF_RXFH_CONFIGURED = 1<<23,
1754 IFF_PHONY_HEADROOM = 1<<24,
1755 IFF_MACSEC = 1<<25,
1756 IFF_NO_RX_HANDLER = 1<<26,
1757 IFF_FAILOVER = 1<<27,
1758 IFF_FAILOVER_SLAVE = 1<<28,
1759 IFF_L3MDEV_RX_HANDLER = 1<<29,
1760 IFF_NO_ADDRCONF = BIT_ULL(30),
1761 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1762 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1763 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33),
1764 };
1765
1766 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1767 #define IFF_EBRIDGE IFF_EBRIDGE
1768 #define IFF_BONDING IFF_BONDING
1769 #define IFF_ISATAP IFF_ISATAP
1770 #define IFF_WAN_HDLC IFF_WAN_HDLC
1771 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1772 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1773 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1774 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1775 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1776 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1777 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1778 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1779 #define IFF_TEAM_PORT IFF_TEAM_PORT
1780 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1781 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1782 #define IFF_MACVLAN IFF_MACVLAN
1783 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1784 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1785 #define IFF_NO_QUEUE IFF_NO_QUEUE
1786 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1787 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1788 #define IFF_TEAM IFF_TEAM
1789 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1790 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1791 #define IFF_MACSEC IFF_MACSEC
1792 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1793 #define IFF_FAILOVER IFF_FAILOVER
1794 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1795 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1796 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1797
1798 /* Specifies the type of the struct net_device::ml_priv pointer */
1799 enum netdev_ml_priv_type {
1800 ML_PRIV_NONE,
1801 ML_PRIV_CAN,
1802 };
1803
1804 enum netdev_stat_type {
1805 NETDEV_PCPU_STAT_NONE,
1806 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1807 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1808 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1809 };
1810
1811 /**
1812 * struct net_device - The DEVICE structure.
1813 *
1814 * Actually, this whole structure is a big mistake. It mixes I/O
1815 * data with strictly "high-level" data, and it has to know about
1816 * almost every data structure used in the INET module.
1817 *
1818 * @name: This is the first field of the "visible" part of this structure
1819 * (i.e. as seen by users in the "Space.c" file). It is the name
1820 * of the interface.
1821 *
1822 * @name_node: Name hashlist node
1823 * @ifalias: SNMP alias
1824 * @mem_end: Shared memory end
1825 * @mem_start: Shared memory start
1826 * @base_addr: Device I/O address
1827 * @irq: Device IRQ number
1828 *
1829 * @state: Generic network queuing layer state, see netdev_state_t
1830 * @dev_list: The global list of network devices
1831 * @napi_list: List entry used for polling NAPI devices
1832 * @unreg_list: List entry when we are unregistering the
1833 * device; see the function unregister_netdev
1834 * @close_list: List entry used when we are closing the device
1835 * @ptype_all: Device-specific packet handlers for all protocols
1836 * @ptype_specific: Device-specific, protocol-specific packet handlers
1837 *
1838 * @adj_list: Directly linked devices, like slaves for bonding
1839 * @features: Currently active device features
1840 * @hw_features: User-changeable features
1841 *
1842 * @wanted_features: User-requested features
1843 * @vlan_features: Mask of features inheritable by VLAN devices
1844 *
1845 * @hw_enc_features: Mask of features inherited by encapsulating devices
1846 * This field indicates what encapsulation
1847 * offloads the hardware is capable of doing,
1848 * and drivers will need to set them appropriately.
1849 *
1850 * @mpls_features: Mask of features inheritable by MPLS
1851 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1852 *
1853 * @ifindex: interface index
1854 * @group: The group the device belongs to
1855 *
1856 * @stats: Statistics struct, which was left as a legacy, use
1857 * rtnl_link_stats64 instead
1858 *
1859 * @core_stats: core networking counters,
1860 * do not use this in drivers
1861 * @carrier_up_count: Number of times the carrier has been up
1862 * @carrier_down_count: Number of times the carrier has been down
1863 *
1864 * @wireless_handlers: List of functions to handle Wireless Extensions,
1865 * instead of ioctl,
1866 * see <net/iw_handler.h> for details.
1867 * @wireless_data: Instance data managed by the core of wireless extensions
1868 *
1869 * @netdev_ops: Includes several pointers to callbacks,
1870 * if one wants to override the ndo_*() functions
1871 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1872 * @ethtool_ops: Management operations
1873 * @l3mdev_ops: Layer 3 master device operations
1874 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1875 * discovery handling. Necessary for e.g. 6LoWPAN.
1876 * @xfrmdev_ops: Transformation offload operations
1877 * @tlsdev_ops: Transport Layer Security offload operations
1878 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1879 * of Layer 2 headers.
1880 *
1881 * @flags: Interface flags (a la BSD)
1882 * @xdp_features: XDP capability supported by the device
1883 * @priv_flags: Like 'flags' but invisible to userspace,
1884 * see if.h for the definitions
1885 * @gflags: Global flags ( kept as legacy )
1886 * @padded: How much padding added by alloc_netdev()
1887 * @operstate: RFC2863 operstate
1888 * @link_mode: Mapping policy to operstate
1889 * @if_port: Selectable AUI, TP, ...
1890 * @dma: DMA channel
1891 * @mtu: Interface MTU value
1892 * @min_mtu: Interface Minimum MTU value
1893 * @max_mtu: Interface Maximum MTU value
1894 * @type: Interface hardware type
1895 * @hard_header_len: Maximum hardware header length.
1896 * @min_header_len: Minimum hardware header length
1897 *
1898 * @needed_headroom: Extra headroom the hardware may need, but not in all
1899 * cases can this be guaranteed
1900 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1901 * cases can this be guaranteed. Some cases also use
1902 * LL_MAX_HEADER instead to allocate the skb
1903 *
1904 * interface address info:
1905 *
1906 * @perm_addr: Permanent hw address
1907 * @addr_assign_type: Hw address assignment type
1908 * @addr_len: Hardware address length
1909 * @upper_level: Maximum depth level of upper devices.
1910 * @lower_level: Maximum depth level of lower devices.
1911 * @neigh_priv_len: Used in neigh_alloc()
1912 * @dev_id: Used to differentiate devices that share
1913 * the same link layer address
1914 * @dev_port: Used to differentiate devices that share
1915 * the same function
1916 * @addr_list_lock: XXX: need comments on this one
1917 * @name_assign_type: network interface name assignment type
1918 * @uc_promisc: Counter that indicates promiscuous mode
1919 * has been enabled due to the need to listen to
1920 * additional unicast addresses in a device that
1921 * does not implement ndo_set_rx_mode()
1922 * @uc: unicast mac addresses
1923 * @mc: multicast mac addresses
1924 * @dev_addrs: list of device hw addresses
1925 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1926 * @promiscuity: Number of times the NIC is told to work in
1927 * promiscuous mode; if it becomes 0 the NIC will
1928 * exit promiscuous mode
1929 * @allmulti: Counter, enables or disables allmulticast mode
1930 *
1931 * @vlan_info: VLAN info
1932 * @dsa_ptr: dsa specific data
1933 * @tipc_ptr: TIPC specific data
1934 * @atalk_ptr: AppleTalk link
1935 * @ip_ptr: IPv4 specific data
1936 * @ip6_ptr: IPv6 specific data
1937 * @ax25_ptr: AX.25 specific data
1938 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1939 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1940 * device struct
1941 * @mpls_ptr: mpls_dev struct pointer
1942 * @mctp_ptr: MCTP specific data
1943 *
1944 * @dev_addr: Hw address (before bcast,
1945 * because most packets are unicast)
1946 *
1947 * @_rx: Array of RX queues
1948 * @num_rx_queues: Number of RX queues
1949 * allocated at register_netdev() time
1950 * @real_num_rx_queues: Number of RX queues currently active in device
1951 * @xdp_prog: XDP sockets filter program pointer
1952 * @gro_flush_timeout: timeout for GRO layer in NAPI
1953 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1954 * allow to avoid NIC hard IRQ, on busy queues.
1955 *
1956 * @rx_handler: handler for received packets
1957 * @rx_handler_data: XXX: need comments on this one
1958 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1959 * @ingress_queue: XXX: need comments on this one
1960 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1961 * @broadcast: hw bcast address
1962 *
1963 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1964 * indexed by RX queue number. Assigned by driver.
1965 * This must only be set if the ndo_rx_flow_steer
1966 * operation is defined
1967 * @index_hlist: Device index hash chain
1968 *
1969 * @_tx: Array of TX queues
1970 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1971 * @real_num_tx_queues: Number of TX queues currently active in device
1972 * @qdisc: Root qdisc from userspace point of view
1973 * @tx_queue_len: Max frames per queue allowed
1974 * @tx_global_lock: XXX: need comments on this one
1975 * @xdp_bulkq: XDP device bulk queue
1976 * @xps_maps: all CPUs/RXQs maps for XPS device
1977 *
1978 * @xps_maps: XXX: need comments on this one
1979 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1980 * @nf_hooks_egress: netfilter hooks executed for egress packets
1981 * @qdisc_hash: qdisc hash table
1982 * @watchdog_timeo: Represents the timeout that is used by
1983 * the watchdog (see dev_watchdog())
1984 * @watchdog_timer: List of timers
1985 *
1986 * @proto_down_reason: reason a netdev interface is held down
1987 * @pcpu_refcnt: Number of references to this device
1988 * @dev_refcnt: Number of references to this device
1989 * @refcnt_tracker: Tracker directory for tracked references to this device
1990 * @todo_list: Delayed register/unregister
1991 * @link_watch_list: XXX: need comments on this one
1992 *
1993 * @reg_state: Register/unregister state machine
1994 * @dismantle: Device is going to be freed
1995 * @rtnl_link_state: This enum represents the phases of creating
1996 * a new link
1997 *
1998 * @needs_free_netdev: Should unregister perform free_netdev?
1999 * @priv_destructor: Called from unregister
2000 * @npinfo: XXX: need comments on this one
2001 * @nd_net: Network namespace this network device is inside
2002 *
2003 * @ml_priv: Mid-layer private
2004 * @ml_priv_type: Mid-layer private type
2005 *
2006 * @pcpu_stat_type: Type of device statistics which the core should
2007 * allocate/free: none, lstats, tstats, dstats. none
2008 * means the driver is handling statistics allocation/
2009 * freeing internally.
2010 * @lstats: Loopback statistics: packets, bytes
2011 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes
2012 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes
2013 *
2014 * @garp_port: GARP
2015 * @mrp_port: MRP
2016 *
2017 * @dm_private: Drop monitor private
2018 *
2019 * @dev: Class/net/name entry
2020 * @sysfs_groups: Space for optional device, statistics and wireless
2021 * sysfs groups
2022 *
2023 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
2024 * @rtnl_link_ops: Rtnl_link_ops
2025 *
2026 * @gso_max_size: Maximum size of generic segmentation offload
2027 * @tso_max_size: Device (as in HW) limit on the max TSO request size
2028 * @gso_max_segs: Maximum number of segments that can be passed to the
2029 * NIC for GSO
2030 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
2031 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
2032 * for IPv4.
2033 *
2034 * @dcbnl_ops: Data Center Bridging netlink ops
2035 * @num_tc: Number of traffic classes in the net device
2036 * @tc_to_txq: XXX: need comments on this one
2037 * @prio_tc_map: XXX: need comments on this one
2038 *
2039 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
2040 *
2041 * @priomap: XXX: need comments on this one
2042 * @phydev: Physical device may attach itself
2043 * for hardware timestamping
2044 * @sfp_bus: attached &struct sfp_bus structure.
2045 *
2046 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2047 *
2048 * @proto_down: protocol port state information can be sent to the
2049 * switch driver and used to set the phys state of the
2050 * switch port.
2051 *
2052 * @wol_enabled: Wake-on-LAN is enabled
2053 *
2054 * @threaded: napi threaded mode is enabled
2055 *
2056 * @net_notifier_list: List of per-net netdev notifier block
2057 * that follow this device when it is moved
2058 * to another network namespace.
2059 *
2060 * @macsec_ops: MACsec offloading ops
2061 *
2062 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2063 * offload capabilities of the device
2064 * @udp_tunnel_nic: UDP tunnel offload state
2065 * @xdp_state: stores info on attached XDP BPF programs
2066 *
2067 * @nested_level: Used as a parameter of spin_lock_nested() of
2068 * dev->addr_list_lock.
2069 * @unlink_list: As netif_addr_lock() can be called recursively,
2070 * keep a list of interfaces to be deleted.
2071 * @gro_max_size: Maximum size of aggregated packet in generic
2072 * receive offload (GRO)
2073 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2074 * receive offload (GRO), for IPv4.
2075 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2076 * zero copy driver
2077 *
2078 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2079 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2080 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2081 * @dev_registered_tracker: tracker for reference held while
2082 * registered
2083 * @offload_xstats_l3: L3 HW stats for this netdevice.
2084 *
2085 * @devlink_port: Pointer to related devlink port structure.
2086 * Assigned by a driver before netdev registration using
2087 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2088 * during the time netdevice is registered.
2089 *
2090 * FIXME: cleanup struct net_device such that network protocol info
2091 * moves out.
2092 */
2093
2094 struct net_device {
2095 char name[IFNAMSIZ];
2096 struct netdev_name_node *name_node;
2097 struct dev_ifalias __rcu *ifalias;
2098 /*
2099 * I/O specific fields
2100 * FIXME: Merge these and struct ifmap into one
2101 */
2102 unsigned long mem_end;
2103 unsigned long mem_start;
2104 unsigned long base_addr;
2105
2106 /*
2107 * Some hardware also needs these fields (state,dev_list,
2108 * napi_list,unreg_list,close_list) but they are not
2109 * part of the usual set specified in Space.c.
2110 */
2111
2112 unsigned long state;
2113
2114 struct list_head dev_list;
2115 struct list_head napi_list;
2116 struct list_head unreg_list;
2117 struct list_head close_list;
2118 struct list_head ptype_all;
2119 struct list_head ptype_specific;
2120
2121 struct {
2122 struct list_head upper;
2123 struct list_head lower;
2124 } adj_list;
2125
2126 /* Read-mostly cache-line for fast-path access */
2127 unsigned int flags;
2128 xdp_features_t xdp_features;
2129 unsigned long long priv_flags;
2130 const struct net_device_ops *netdev_ops;
2131 const struct xdp_metadata_ops *xdp_metadata_ops;
2132 int ifindex;
2133 unsigned short gflags;
2134 unsigned short hard_header_len;
2135
2136 /* Note : dev->mtu is often read without holding a lock.
2137 * Writers usually hold RTNL.
2138 * It is recommended to use READ_ONCE() to annotate the reads,
2139 * and to use WRITE_ONCE() to annotate the writes.
2140 */
2141 unsigned int mtu;
2142 unsigned short needed_headroom;
2143 unsigned short needed_tailroom;
2144
2145 netdev_features_t features;
2146 netdev_features_t hw_features;
2147 netdev_features_t wanted_features;
2148 netdev_features_t vlan_features;
2149 netdev_features_t hw_enc_features;
2150 netdev_features_t mpls_features;
2151 netdev_features_t gso_partial_features;
2152
2153 unsigned int min_mtu;
2154 unsigned int max_mtu;
2155 unsigned short type;
2156 unsigned char min_header_len;
2157 unsigned char name_assign_type;
2158
2159 int group;
2160
2161 struct net_device_stats stats; /* not used by modern drivers */
2162
2163 struct net_device_core_stats __percpu *core_stats;
2164
2165 /* Stats to monitor link on/off, flapping */
2166 atomic_t carrier_up_count;
2167 atomic_t carrier_down_count;
2168
2169 /* Android KMI hack to allow vendors to have their own wifi changes in modules */
2170 const struct iw_handler_def *wireless_handlers;
2171 struct iw_public_data *wireless_data;
2172 const struct ethtool_ops *ethtool_ops;
2173 #ifdef CONFIG_NET_L3_MASTER_DEV
2174 const struct l3mdev_ops *l3mdev_ops;
2175 #endif
2176 #if IS_ENABLED(CONFIG_IPV6)
2177 const struct ndisc_ops *ndisc_ops;
2178 #endif
2179
2180 #ifdef CONFIG_XFRM_OFFLOAD
2181 const struct xfrmdev_ops *xfrmdev_ops;
2182 #endif
2183
2184 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2185 const struct tlsdev_ops *tlsdev_ops;
2186 #endif
2187
2188 const struct header_ops *header_ops;
2189
2190 unsigned char operstate;
2191 unsigned char link_mode;
2192
2193 unsigned char if_port;
2194 unsigned char dma;
2195
2196 /* Interface address info. */
2197 unsigned char perm_addr[MAX_ADDR_LEN];
2198 unsigned char addr_assign_type;
2199 unsigned char addr_len;
2200 unsigned char upper_level;
2201 unsigned char lower_level;
2202
2203 unsigned short neigh_priv_len;
2204 unsigned short dev_id;
2205 unsigned short dev_port;
2206 unsigned short padded;
2207
2208 spinlock_t addr_list_lock;
2209 int irq;
2210
2211 struct netdev_hw_addr_list uc;
2212 struct netdev_hw_addr_list mc;
2213 struct netdev_hw_addr_list dev_addrs;
2214
2215 #ifdef CONFIG_SYSFS
2216 struct kset *queues_kset;
2217 #endif
2218 #ifdef CONFIG_LOCKDEP
2219 struct list_head unlink_list;
2220 #endif
2221 unsigned int promiscuity;
2222 unsigned int allmulti;
2223 bool uc_promisc;
2224 #ifdef CONFIG_LOCKDEP
2225 unsigned char nested_level;
2226 #endif
2227
2228
2229 /* Protocol-specific pointers */
2230
2231 struct in_device __rcu *ip_ptr;
2232 struct inet6_dev __rcu *ip6_ptr;
2233 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2234 struct vlan_info __rcu *vlan_info;
2235 #endif
2236 #if IS_ENABLED(CONFIG_NET_DSA)
2237 struct dsa_port *dsa_ptr;
2238 #endif
2239 #if IS_ENABLED(CONFIG_TIPC)
2240 struct tipc_bearer __rcu *tipc_ptr;
2241 #endif
2242 #if IS_ENABLED(CONFIG_ATALK)
2243 void *atalk_ptr;
2244 #endif
2245 #if IS_ENABLED(CONFIG_AX25)
2246 void *ax25_ptr;
2247 #endif
2248 /* Android KMI hack to allow vendors to have their own wifi changes in modules */
2249 struct wireless_dev *ieee80211_ptr;
2250 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2251 struct wpan_dev *ieee802154_ptr;
2252 #endif
2253 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2254 struct mpls_dev __rcu *mpls_ptr;
2255 #endif
2256 #if IS_ENABLED(CONFIG_MCTP)
2257 struct mctp_dev __rcu *mctp_ptr;
2258 #endif
2259
2260 /*
2261 * Cache lines mostly used on receive path (including eth_type_trans())
2262 */
2263 /* Interface address info used in eth_type_trans() */
2264 const unsigned char *dev_addr;
2265
2266 struct netdev_rx_queue *_rx;
2267 unsigned int num_rx_queues;
2268 unsigned int real_num_rx_queues;
2269
2270 struct bpf_prog __rcu *xdp_prog;
2271 unsigned long gro_flush_timeout;
2272 int napi_defer_hard_irqs;
2273 #define GRO_LEGACY_MAX_SIZE 65536u
2274 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2275 * and shinfo->gso_segs is a 16bit field.
2276 */
2277 #define GRO_MAX_SIZE (8 * 65535u)
2278 unsigned int gro_max_size;
2279 unsigned int gro_ipv4_max_size;
2280 unsigned int xdp_zc_max_segs;
2281 rx_handler_func_t __rcu *rx_handler;
2282 void __rcu *rx_handler_data;
2283 #ifdef CONFIG_NET_XGRESS
2284 struct bpf_mprog_entry __rcu *tcx_ingress;
2285 #endif
2286 struct netdev_queue __rcu *ingress_queue;
2287 #ifdef CONFIG_NETFILTER_INGRESS
2288 struct nf_hook_entries __rcu *nf_hooks_ingress;
2289 #endif
2290
2291 unsigned char broadcast[MAX_ADDR_LEN];
2292 #ifdef CONFIG_RFS_ACCEL
2293 struct cpu_rmap *rx_cpu_rmap;
2294 #endif
2295 struct hlist_node index_hlist;
2296
2297 /*
2298 * Cache lines mostly used on transmit path
2299 */
2300 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2301 unsigned int num_tx_queues;
2302 unsigned int real_num_tx_queues;
2303 struct Qdisc __rcu *qdisc;
2304 unsigned int tx_queue_len;
2305 spinlock_t tx_global_lock;
2306
2307 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2308
2309 #ifdef CONFIG_XPS
2310 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2311 #endif
2312 #ifdef CONFIG_NET_XGRESS
2313 struct bpf_mprog_entry __rcu *tcx_egress;
2314 #endif
2315 #ifdef CONFIG_NETFILTER_EGRESS
2316 struct nf_hook_entries __rcu *nf_hooks_egress;
2317 #endif
2318
2319 #ifdef CONFIG_NET_SCHED
2320 DECLARE_HASHTABLE (qdisc_hash, 4);
2321 #endif
2322 /* These may be needed for future network-power-down code. */
2323 struct timer_list watchdog_timer;
2324 int watchdog_timeo;
2325
2326 u32 proto_down_reason;
2327
2328 struct list_head todo_list;
2329
2330 #ifdef CONFIG_PCPU_DEV_REFCNT
2331 int __percpu *pcpu_refcnt;
2332 #else
2333 refcount_t dev_refcnt;
2334 #endif
2335 struct ref_tracker_dir refcnt_tracker;
2336
2337 struct list_head link_watch_list;
2338
2339 enum { NETREG_UNINITIALIZED=0,
2340 NETREG_REGISTERED, /* completed register_netdevice */
2341 NETREG_UNREGISTERING, /* called unregister_netdevice */
2342 NETREG_UNREGISTERED, /* completed unregister todo */
2343 NETREG_RELEASED, /* called free_netdev */
2344 NETREG_DUMMY, /* dummy device for NAPI poll */
2345 } reg_state:8;
2346
2347 bool dismantle;
2348
2349 enum {
2350 RTNL_LINK_INITIALIZED,
2351 RTNL_LINK_INITIALIZING,
2352 } rtnl_link_state:16;
2353
2354 bool needs_free_netdev;
2355 void (*priv_destructor)(struct net_device *dev);
2356
2357 #ifdef CONFIG_NETPOLL
2358 struct netpoll_info __rcu *npinfo;
2359 #endif
2360
2361 possible_net_t nd_net;
2362
2363 /* mid-layer private */
2364 void *ml_priv;
2365 enum netdev_ml_priv_type ml_priv_type;
2366
2367 enum netdev_stat_type pcpu_stat_type:8;
2368 union {
2369 struct pcpu_lstats __percpu *lstats;
2370 struct pcpu_sw_netstats __percpu *tstats;
2371 struct pcpu_dstats __percpu *dstats;
2372 };
2373
2374 #if IS_ENABLED(CONFIG_GARP)
2375 struct garp_port __rcu *garp_port;
2376 #endif
2377 #if IS_ENABLED(CONFIG_MRP)
2378 struct mrp_port __rcu *mrp_port;
2379 #endif
2380 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2381 struct dm_hw_stat_delta __rcu *dm_private;
2382 #endif
2383 struct device dev;
2384 const struct attribute_group *sysfs_groups[4];
2385 const struct attribute_group *sysfs_rx_queue_group;
2386
2387 const struct rtnl_link_ops *rtnl_link_ops;
2388
2389 /* for setting kernel sock attribute on TCP connection setup */
2390 #define GSO_MAX_SEGS 65535u
2391 #define GSO_LEGACY_MAX_SIZE 65536u
2392 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2393 * and shinfo->gso_segs is a 16bit field.
2394 */
2395 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2396
2397 unsigned int gso_max_size;
2398 #define TSO_LEGACY_MAX_SIZE 65536
2399 #define TSO_MAX_SIZE UINT_MAX
2400 unsigned int tso_max_size;
2401 u16 gso_max_segs;
2402 #define TSO_MAX_SEGS U16_MAX
2403 u16 tso_max_segs;
2404 unsigned int gso_ipv4_max_size;
2405
2406 #ifdef CONFIG_DCB
2407 const struct dcbnl_rtnl_ops *dcbnl_ops;
2408 #endif
2409 s16 num_tc;
2410 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2411 u8 prio_tc_map[TC_BITMASK + 1];
2412
2413 #if IS_ENABLED(CONFIG_FCOE)
2414 unsigned int fcoe_ddp_xid;
2415 #endif
2416 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2417 struct netprio_map __rcu *priomap;
2418 #endif
2419 struct phy_device *phydev;
2420 struct sfp_bus *sfp_bus;
2421 struct lock_class_key *qdisc_tx_busylock;
2422 bool proto_down;
2423 unsigned wol_enabled:1;
2424 unsigned threaded:1;
2425
2426 struct list_head net_notifier_list;
2427
2428 #if IS_ENABLED(CONFIG_MACSEC)
2429 /* MACsec management functions */
2430 const struct macsec_ops *macsec_ops;
2431 #endif
2432 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2433 struct udp_tunnel_nic *udp_tunnel_nic;
2434
2435 /* protected by rtnl_lock */
2436 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2437
2438 u8 dev_addr_shadow[MAX_ADDR_LEN];
2439 netdevice_tracker linkwatch_dev_tracker;
2440 netdevice_tracker watchdog_dev_tracker;
2441 netdevice_tracker dev_registered_tracker;
2442 struct rtnl_hw_stats64 *offload_xstats_l3;
2443
2444 struct devlink_port *devlink_port;
2445
2446 ANDROID_KABI_RESERVE(1);
2447 ANDROID_KABI_RESERVE(2);
2448 ANDROID_KABI_RESERVE(3);
2449 ANDROID_KABI_RESERVE(4);
2450 ANDROID_KABI_RESERVE(5);
2451 ANDROID_KABI_RESERVE(6);
2452 ANDROID_KABI_RESERVE(7);
2453 ANDROID_KABI_RESERVE(8);
2454 };
2455 #define to_net_dev(d) container_of(d, struct net_device, dev)
2456
2457 /*
2458 * Driver should use this to assign devlink port instance to a netdevice
2459 * before it registers the netdevice. Therefore devlink_port is static
2460 * during the netdev lifetime after it is registered.
2461 */
2462 #define SET_NETDEV_DEVLINK_PORT(dev, port) \
2463 ({ \
2464 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2465 ((dev)->devlink_port = (port)); \
2466 })
2467
netif_elide_gro(const struct net_device * dev)2468 static inline bool netif_elide_gro(const struct net_device *dev)
2469 {
2470 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2471 return true;
2472 return false;
2473 }
2474
2475 #define NETDEV_ALIGN 32
2476
2477 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)2478 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2479 {
2480 return dev->prio_tc_map[prio & TC_BITMASK];
2481 }
2482
2483 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)2484 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2485 {
2486 if (tc >= dev->num_tc)
2487 return -EINVAL;
2488
2489 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2490 return 0;
2491 }
2492
2493 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2494 void netdev_reset_tc(struct net_device *dev);
2495 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2496 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2497
2498 static inline
netdev_get_num_tc(struct net_device * dev)2499 int netdev_get_num_tc(struct net_device *dev)
2500 {
2501 return dev->num_tc;
2502 }
2503
net_prefetch(void * p)2504 static inline void net_prefetch(void *p)
2505 {
2506 prefetch(p);
2507 #if L1_CACHE_BYTES < 128
2508 prefetch((u8 *)p + L1_CACHE_BYTES);
2509 #endif
2510 }
2511
net_prefetchw(void * p)2512 static inline void net_prefetchw(void *p)
2513 {
2514 prefetchw(p);
2515 #if L1_CACHE_BYTES < 128
2516 prefetchw((u8 *)p + L1_CACHE_BYTES);
2517 #endif
2518 }
2519
2520 void netdev_unbind_sb_channel(struct net_device *dev,
2521 struct net_device *sb_dev);
2522 int netdev_bind_sb_channel_queue(struct net_device *dev,
2523 struct net_device *sb_dev,
2524 u8 tc, u16 count, u16 offset);
2525 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
netdev_get_sb_channel(struct net_device * dev)2526 static inline int netdev_get_sb_channel(struct net_device *dev)
2527 {
2528 return max_t(int, -dev->num_tc, 0);
2529 }
2530
2531 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)2532 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2533 unsigned int index)
2534 {
2535 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2536 return &dev->_tx[index];
2537 }
2538
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)2539 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2540 const struct sk_buff *skb)
2541 {
2542 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2543 }
2544
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)2545 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2546 void (*f)(struct net_device *,
2547 struct netdev_queue *,
2548 void *),
2549 void *arg)
2550 {
2551 unsigned int i;
2552
2553 for (i = 0; i < dev->num_tx_queues; i++)
2554 f(dev, &dev->_tx[i], arg);
2555 }
2556
2557 #define netdev_lockdep_set_classes(dev) \
2558 { \
2559 static struct lock_class_key qdisc_tx_busylock_key; \
2560 static struct lock_class_key qdisc_xmit_lock_key; \
2561 static struct lock_class_key dev_addr_list_lock_key; \
2562 unsigned int i; \
2563 \
2564 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2565 lockdep_set_class(&(dev)->addr_list_lock, \
2566 &dev_addr_list_lock_key); \
2567 for (i = 0; i < (dev)->num_tx_queues; i++) \
2568 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2569 &qdisc_xmit_lock_key); \
2570 }
2571
2572 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2573 struct net_device *sb_dev);
2574 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2575 struct sk_buff *skb,
2576 struct net_device *sb_dev);
2577
2578 /* returns the headroom that the master device needs to take in account
2579 * when forwarding to this dev
2580 */
netdev_get_fwd_headroom(struct net_device * dev)2581 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2582 {
2583 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2584 }
2585
netdev_set_rx_headroom(struct net_device * dev,int new_hr)2586 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2587 {
2588 if (dev->netdev_ops->ndo_set_rx_headroom)
2589 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2590 }
2591
2592 /* set the device rx headroom to the dev's default */
netdev_reset_rx_headroom(struct net_device * dev)2593 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2594 {
2595 netdev_set_rx_headroom(dev, -1);
2596 }
2597
netdev_get_ml_priv(struct net_device * dev,enum netdev_ml_priv_type type)2598 static inline void *netdev_get_ml_priv(struct net_device *dev,
2599 enum netdev_ml_priv_type type)
2600 {
2601 if (dev->ml_priv_type != type)
2602 return NULL;
2603
2604 return dev->ml_priv;
2605 }
2606
netdev_set_ml_priv(struct net_device * dev,void * ml_priv,enum netdev_ml_priv_type type)2607 static inline void netdev_set_ml_priv(struct net_device *dev,
2608 void *ml_priv,
2609 enum netdev_ml_priv_type type)
2610 {
2611 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2612 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2613 dev->ml_priv_type, type);
2614 WARN(!dev->ml_priv_type && dev->ml_priv,
2615 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2616
2617 dev->ml_priv = ml_priv;
2618 dev->ml_priv_type = type;
2619 }
2620
2621 /*
2622 * Net namespace inlines
2623 */
2624 static inline
dev_net(const struct net_device * dev)2625 struct net *dev_net(const struct net_device *dev)
2626 {
2627 return read_pnet(&dev->nd_net);
2628 }
2629
2630 static inline
dev_net_set(struct net_device * dev,struct net * net)2631 void dev_net_set(struct net_device *dev, struct net *net)
2632 {
2633 write_pnet(&dev->nd_net, net);
2634 }
2635
2636 /**
2637 * netdev_priv - access network device private data
2638 * @dev: network device
2639 *
2640 * Get network device private data
2641 */
netdev_priv(const struct net_device * dev)2642 static inline void *netdev_priv(const struct net_device *dev)
2643 {
2644 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2645 }
2646
2647 /* Set the sysfs physical device reference for the network logical device
2648 * if set prior to registration will cause a symlink during initialization.
2649 */
2650 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2651
2652 /* Set the sysfs device type for the network logical device to allow
2653 * fine-grained identification of different network device types. For
2654 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2655 */
2656 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2657
2658 /* Default NAPI poll() weight
2659 * Device drivers are strongly advised to not use bigger value
2660 */
2661 #define NAPI_POLL_WEIGHT 64
2662
2663 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2664 int (*poll)(struct napi_struct *, int), int weight);
2665
2666 /**
2667 * netif_napi_add() - initialize a NAPI context
2668 * @dev: network device
2669 * @napi: NAPI context
2670 * @poll: polling function
2671 *
2672 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2673 * *any* of the other NAPI-related functions.
2674 */
2675 static inline void
netif_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int))2676 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2677 int (*poll)(struct napi_struct *, int))
2678 {
2679 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2680 }
2681
2682 static inline void
netif_napi_add_tx_weight(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2683 netif_napi_add_tx_weight(struct net_device *dev,
2684 struct napi_struct *napi,
2685 int (*poll)(struct napi_struct *, int),
2686 int weight)
2687 {
2688 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2689 netif_napi_add_weight(dev, napi, poll, weight);
2690 }
2691
2692 /**
2693 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2694 * @dev: network device
2695 * @napi: NAPI context
2696 * @poll: polling function
2697 *
2698 * This variant of netif_napi_add() should be used from drivers using NAPI
2699 * to exclusively poll a TX queue.
2700 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2701 */
netif_napi_add_tx(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int))2702 static inline void netif_napi_add_tx(struct net_device *dev,
2703 struct napi_struct *napi,
2704 int (*poll)(struct napi_struct *, int))
2705 {
2706 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2707 }
2708
2709 /**
2710 * __netif_napi_del - remove a NAPI context
2711 * @napi: NAPI context
2712 *
2713 * Warning: caller must observe RCU grace period before freeing memory
2714 * containing @napi. Drivers might want to call this helper to combine
2715 * all the needed RCU grace periods into a single one.
2716 */
2717 void __netif_napi_del(struct napi_struct *napi);
2718
2719 /**
2720 * netif_napi_del - remove a NAPI context
2721 * @napi: NAPI context
2722 *
2723 * netif_napi_del() removes a NAPI context from the network device NAPI list
2724 */
netif_napi_del(struct napi_struct * napi)2725 static inline void netif_napi_del(struct napi_struct *napi)
2726 {
2727 __netif_napi_del(napi);
2728 synchronize_net();
2729 }
2730
2731 struct packet_type {
2732 __be16 type; /* This is really htons(ether_type). */
2733 bool ignore_outgoing;
2734 struct net_device *dev; /* NULL is wildcarded here */
2735 netdevice_tracker dev_tracker;
2736 int (*func) (struct sk_buff *,
2737 struct net_device *,
2738 struct packet_type *,
2739 struct net_device *);
2740 void (*list_func) (struct list_head *,
2741 struct packet_type *,
2742 struct net_device *);
2743 bool (*id_match)(struct packet_type *ptype,
2744 struct sock *sk);
2745 struct net *af_packet_net;
2746 void *af_packet_priv;
2747 struct list_head list;
2748
2749 ANDROID_KABI_RESERVE(1);
2750 ANDROID_KABI_RESERVE(2);
2751 ANDROID_KABI_RESERVE(3);
2752 ANDROID_KABI_RESERVE(4);
2753 };
2754
2755 struct offload_callbacks {
2756 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2757 netdev_features_t features);
2758 struct sk_buff *(*gro_receive)(struct list_head *head,
2759 struct sk_buff *skb);
2760 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2761 };
2762
2763 struct packet_offload {
2764 __be16 type; /* This is really htons(ether_type). */
2765 u16 priority;
2766 struct offload_callbacks callbacks;
2767 struct list_head list;
2768 };
2769
2770 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2771 struct pcpu_sw_netstats {
2772 u64_stats_t rx_packets;
2773 u64_stats_t rx_bytes;
2774 u64_stats_t tx_packets;
2775 u64_stats_t tx_bytes;
2776 struct u64_stats_sync syncp;
2777 } __aligned(4 * sizeof(u64));
2778
2779 struct pcpu_dstats {
2780 u64 rx_packets;
2781 u64 rx_bytes;
2782 u64 rx_drops;
2783 u64 tx_packets;
2784 u64 tx_bytes;
2785 u64 tx_drops;
2786 struct u64_stats_sync syncp;
2787 } __aligned(8 * sizeof(u64));
2788
2789 struct pcpu_lstats {
2790 u64_stats_t packets;
2791 u64_stats_t bytes;
2792 struct u64_stats_sync syncp;
2793 } __aligned(2 * sizeof(u64));
2794
2795 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2796
dev_sw_netstats_rx_add(struct net_device * dev,unsigned int len)2797 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2798 {
2799 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2800
2801 u64_stats_update_begin(&tstats->syncp);
2802 u64_stats_add(&tstats->rx_bytes, len);
2803 u64_stats_inc(&tstats->rx_packets);
2804 u64_stats_update_end(&tstats->syncp);
2805 }
2806
dev_sw_netstats_tx_add(struct net_device * dev,unsigned int packets,unsigned int len)2807 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2808 unsigned int packets,
2809 unsigned int len)
2810 {
2811 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2812
2813 u64_stats_update_begin(&tstats->syncp);
2814 u64_stats_add(&tstats->tx_bytes, len);
2815 u64_stats_add(&tstats->tx_packets, packets);
2816 u64_stats_update_end(&tstats->syncp);
2817 }
2818
dev_lstats_add(struct net_device * dev,unsigned int len)2819 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2820 {
2821 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2822
2823 u64_stats_update_begin(&lstats->syncp);
2824 u64_stats_add(&lstats->bytes, len);
2825 u64_stats_inc(&lstats->packets);
2826 u64_stats_update_end(&lstats->syncp);
2827 }
2828
2829 #define __netdev_alloc_pcpu_stats(type, gfp) \
2830 ({ \
2831 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2832 if (pcpu_stats) { \
2833 int __cpu; \
2834 for_each_possible_cpu(__cpu) { \
2835 typeof(type) *stat; \
2836 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2837 u64_stats_init(&stat->syncp); \
2838 } \
2839 } \
2840 pcpu_stats; \
2841 })
2842
2843 #define netdev_alloc_pcpu_stats(type) \
2844 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2845
2846 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2847 ({ \
2848 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2849 if (pcpu_stats) { \
2850 int __cpu; \
2851 for_each_possible_cpu(__cpu) { \
2852 typeof(type) *stat; \
2853 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2854 u64_stats_init(&stat->syncp); \
2855 } \
2856 } \
2857 pcpu_stats; \
2858 })
2859
2860 enum netdev_lag_tx_type {
2861 NETDEV_LAG_TX_TYPE_UNKNOWN,
2862 NETDEV_LAG_TX_TYPE_RANDOM,
2863 NETDEV_LAG_TX_TYPE_BROADCAST,
2864 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2865 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2866 NETDEV_LAG_TX_TYPE_HASH,
2867 };
2868
2869 enum netdev_lag_hash {
2870 NETDEV_LAG_HASH_NONE,
2871 NETDEV_LAG_HASH_L2,
2872 NETDEV_LAG_HASH_L34,
2873 NETDEV_LAG_HASH_L23,
2874 NETDEV_LAG_HASH_E23,
2875 NETDEV_LAG_HASH_E34,
2876 NETDEV_LAG_HASH_VLAN_SRCMAC,
2877 NETDEV_LAG_HASH_UNKNOWN,
2878 };
2879
2880 struct netdev_lag_upper_info {
2881 enum netdev_lag_tx_type tx_type;
2882 enum netdev_lag_hash hash_type;
2883 };
2884
2885 struct netdev_lag_lower_state_info {
2886 u8 link_up : 1,
2887 tx_enabled : 1;
2888 };
2889
2890 #include <linux/notifier.h>
2891
2892 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2893 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2894 * adding new types.
2895 */
2896 enum netdev_cmd {
2897 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2898 NETDEV_DOWN,
2899 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2900 detected a hardware crash and restarted
2901 - we can use this eg to kick tcp sessions
2902 once done */
2903 NETDEV_CHANGE, /* Notify device state change */
2904 NETDEV_REGISTER,
2905 NETDEV_UNREGISTER,
2906 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2907 NETDEV_CHANGEADDR, /* notify after the address change */
2908 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2909 NETDEV_GOING_DOWN,
2910 NETDEV_CHANGENAME,
2911 NETDEV_FEAT_CHANGE,
2912 NETDEV_BONDING_FAILOVER,
2913 NETDEV_PRE_UP,
2914 NETDEV_PRE_TYPE_CHANGE,
2915 NETDEV_POST_TYPE_CHANGE,
2916 NETDEV_POST_INIT,
2917 NETDEV_PRE_UNINIT,
2918 NETDEV_RELEASE,
2919 NETDEV_NOTIFY_PEERS,
2920 NETDEV_JOIN,
2921 NETDEV_CHANGEUPPER,
2922 NETDEV_RESEND_IGMP,
2923 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2924 NETDEV_CHANGEINFODATA,
2925 NETDEV_BONDING_INFO,
2926 NETDEV_PRECHANGEUPPER,
2927 NETDEV_CHANGELOWERSTATE,
2928 NETDEV_UDP_TUNNEL_PUSH_INFO,
2929 NETDEV_UDP_TUNNEL_DROP_INFO,
2930 NETDEV_CHANGE_TX_QUEUE_LEN,
2931 NETDEV_CVLAN_FILTER_PUSH_INFO,
2932 NETDEV_CVLAN_FILTER_DROP_INFO,
2933 NETDEV_SVLAN_FILTER_PUSH_INFO,
2934 NETDEV_SVLAN_FILTER_DROP_INFO,
2935 NETDEV_OFFLOAD_XSTATS_ENABLE,
2936 NETDEV_OFFLOAD_XSTATS_DISABLE,
2937 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2938 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2939 NETDEV_XDP_FEAT_CHANGE,
2940 };
2941 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2942
2943 int register_netdevice_notifier(struct notifier_block *nb);
2944 int unregister_netdevice_notifier(struct notifier_block *nb);
2945 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2946 int unregister_netdevice_notifier_net(struct net *net,
2947 struct notifier_block *nb);
2948 int register_netdevice_notifier_dev_net(struct net_device *dev,
2949 struct notifier_block *nb,
2950 struct netdev_net_notifier *nn);
2951 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2952 struct notifier_block *nb,
2953 struct netdev_net_notifier *nn);
2954
2955 struct netdev_notifier_info {
2956 struct net_device *dev;
2957 struct netlink_ext_ack *extack;
2958 };
2959
2960 struct netdev_notifier_info_ext {
2961 struct netdev_notifier_info info; /* must be first */
2962 union {
2963 u32 mtu;
2964 } ext;
2965 };
2966
2967 struct netdev_notifier_change_info {
2968 struct netdev_notifier_info info; /* must be first */
2969 unsigned int flags_changed;
2970 };
2971
2972 struct netdev_notifier_changeupper_info {
2973 struct netdev_notifier_info info; /* must be first */
2974 struct net_device *upper_dev; /* new upper dev */
2975 bool master; /* is upper dev master */
2976 bool linking; /* is the notification for link or unlink */
2977 void *upper_info; /* upper dev info */
2978 };
2979
2980 struct netdev_notifier_changelowerstate_info {
2981 struct netdev_notifier_info info; /* must be first */
2982 void *lower_state_info; /* is lower dev state */
2983 };
2984
2985 struct netdev_notifier_pre_changeaddr_info {
2986 struct netdev_notifier_info info; /* must be first */
2987 const unsigned char *dev_addr;
2988 };
2989
2990 enum netdev_offload_xstats_type {
2991 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2992 };
2993
2994 struct netdev_notifier_offload_xstats_info {
2995 struct netdev_notifier_info info; /* must be first */
2996 enum netdev_offload_xstats_type type;
2997
2998 union {
2999 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
3000 struct netdev_notifier_offload_xstats_rd *report_delta;
3001 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
3002 struct netdev_notifier_offload_xstats_ru *report_used;
3003 };
3004 };
3005
3006 int netdev_offload_xstats_enable(struct net_device *dev,
3007 enum netdev_offload_xstats_type type,
3008 struct netlink_ext_ack *extack);
3009 int netdev_offload_xstats_disable(struct net_device *dev,
3010 enum netdev_offload_xstats_type type);
3011 bool netdev_offload_xstats_enabled(const struct net_device *dev,
3012 enum netdev_offload_xstats_type type);
3013 int netdev_offload_xstats_get(struct net_device *dev,
3014 enum netdev_offload_xstats_type type,
3015 struct rtnl_hw_stats64 *stats, bool *used,
3016 struct netlink_ext_ack *extack);
3017 void
3018 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3019 const struct rtnl_hw_stats64 *stats);
3020 void
3021 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3022 void netdev_offload_xstats_push_delta(struct net_device *dev,
3023 enum netdev_offload_xstats_type type,
3024 const struct rtnl_hw_stats64 *stats);
3025
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)3026 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3027 struct net_device *dev)
3028 {
3029 info->dev = dev;
3030 info->extack = NULL;
3031 }
3032
3033 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)3034 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3035 {
3036 return info->dev;
3037 }
3038
3039 static inline struct netlink_ext_ack *
netdev_notifier_info_to_extack(const struct netdev_notifier_info * info)3040 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3041 {
3042 return info->extack;
3043 }
3044
3045 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3046 int call_netdevice_notifiers_info(unsigned long val,
3047 struct netdev_notifier_info *info);
3048
3049 extern rwlock_t dev_base_lock; /* Device list lock */
3050
3051 #define for_each_netdev(net, d) \
3052 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3053 #define for_each_netdev_reverse(net, d) \
3054 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3055 #define for_each_netdev_rcu(net, d) \
3056 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3057 #define for_each_netdev_safe(net, d, n) \
3058 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3059 #define for_each_netdev_continue(net, d) \
3060 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3061 #define for_each_netdev_continue_reverse(net, d) \
3062 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3063 dev_list)
3064 #define for_each_netdev_continue_rcu(net, d) \
3065 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3066 #define for_each_netdev_in_bond_rcu(bond, slave) \
3067 for_each_netdev_rcu(&init_net, slave) \
3068 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3069 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3070
3071 #define for_each_netdev_dump(net, d, ifindex) \
3072 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3073
next_net_device(struct net_device * dev)3074 static inline struct net_device *next_net_device(struct net_device *dev)
3075 {
3076 struct list_head *lh;
3077 struct net *net;
3078
3079 net = dev_net(dev);
3080 lh = dev->dev_list.next;
3081 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3082 }
3083
next_net_device_rcu(struct net_device * dev)3084 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3085 {
3086 struct list_head *lh;
3087 struct net *net;
3088
3089 net = dev_net(dev);
3090 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3091 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3092 }
3093
first_net_device(struct net * net)3094 static inline struct net_device *first_net_device(struct net *net)
3095 {
3096 return list_empty(&net->dev_base_head) ? NULL :
3097 net_device_entry(net->dev_base_head.next);
3098 }
3099
first_net_device_rcu(struct net * net)3100 static inline struct net_device *first_net_device_rcu(struct net *net)
3101 {
3102 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3103
3104 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3105 }
3106
3107 int netdev_boot_setup_check(struct net_device *dev);
3108 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3109 const char *hwaddr);
3110 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3111 void dev_add_pack(struct packet_type *pt);
3112 void dev_remove_pack(struct packet_type *pt);
3113 void __dev_remove_pack(struct packet_type *pt);
3114 void dev_add_offload(struct packet_offload *po);
3115 void dev_remove_offload(struct packet_offload *po);
3116
3117 int dev_get_iflink(const struct net_device *dev);
3118 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3119 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3120 struct net_device_path_stack *stack);
3121 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3122 unsigned short mask);
3123 struct net_device *dev_get_by_name(struct net *net, const char *name);
3124 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3125 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3126 bool netdev_name_in_use(struct net *net, const char *name);
3127 int dev_alloc_name(struct net_device *dev, const char *name);
3128 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3129 void dev_close(struct net_device *dev);
3130 void dev_close_many(struct list_head *head, bool unlink);
3131 void dev_disable_lro(struct net_device *dev);
3132 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3133 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3134 struct net_device *sb_dev);
3135 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3136 struct net_device *sb_dev);
3137
3138 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3139 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3140
dev_queue_xmit(struct sk_buff * skb)3141 static inline int dev_queue_xmit(struct sk_buff *skb)
3142 {
3143 return __dev_queue_xmit(skb, NULL);
3144 }
3145
dev_queue_xmit_accel(struct sk_buff * skb,struct net_device * sb_dev)3146 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3147 struct net_device *sb_dev)
3148 {
3149 return __dev_queue_xmit(skb, sb_dev);
3150 }
3151
dev_direct_xmit(struct sk_buff * skb,u16 queue_id)3152 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3153 {
3154 int ret;
3155
3156 ret = __dev_direct_xmit(skb, queue_id);
3157 if (!dev_xmit_complete(ret))
3158 kfree_skb(skb);
3159 return ret;
3160 }
3161
3162 int register_netdevice(struct net_device *dev);
3163 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3164 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)3165 static inline void unregister_netdevice(struct net_device *dev)
3166 {
3167 unregister_netdevice_queue(dev, NULL);
3168 }
3169
3170 int netdev_refcnt_read(const struct net_device *dev);
3171 void free_netdev(struct net_device *dev);
3172 void netdev_freemem(struct net_device *dev);
3173 int init_dummy_netdev(struct net_device *dev);
3174
3175 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3176 struct sk_buff *skb,
3177 bool all_slaves);
3178 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3179 struct sock *sk);
3180 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3181 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3182 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3183 netdevice_tracker *tracker, gfp_t gfp);
3184 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3185 netdevice_tracker *tracker, gfp_t gfp);
3186 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3187 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3188
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)3189 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3190 unsigned short type,
3191 const void *daddr, const void *saddr,
3192 unsigned int len)
3193 {
3194 if (!dev->header_ops || !dev->header_ops->create)
3195 return 0;
3196
3197 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3198 }
3199
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)3200 static inline int dev_parse_header(const struct sk_buff *skb,
3201 unsigned char *haddr)
3202 {
3203 const struct net_device *dev = skb->dev;
3204
3205 if (!dev->header_ops || !dev->header_ops->parse)
3206 return 0;
3207 return dev->header_ops->parse(skb, haddr);
3208 }
3209
dev_parse_header_protocol(const struct sk_buff * skb)3210 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3211 {
3212 const struct net_device *dev = skb->dev;
3213
3214 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3215 return 0;
3216 return dev->header_ops->parse_protocol(skb);
3217 }
3218
3219 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)3220 static inline bool dev_validate_header(const struct net_device *dev,
3221 char *ll_header, int len)
3222 {
3223 if (likely(len >= dev->hard_header_len))
3224 return true;
3225 if (len < dev->min_header_len)
3226 return false;
3227
3228 if (capable(CAP_SYS_RAWIO)) {
3229 memset(ll_header + len, 0, dev->hard_header_len - len);
3230 return true;
3231 }
3232
3233 if (dev->header_ops && dev->header_ops->validate)
3234 return dev->header_ops->validate(ll_header, len);
3235
3236 return false;
3237 }
3238
dev_has_header(const struct net_device * dev)3239 static inline bool dev_has_header(const struct net_device *dev)
3240 {
3241 return dev->header_ops && dev->header_ops->create;
3242 }
3243
3244 /*
3245 * Incoming packets are placed on per-CPU queues
3246 */
3247 struct softnet_data {
3248 struct list_head poll_list;
3249 struct sk_buff_head process_queue;
3250
3251 /* stats */
3252 unsigned int processed;
3253 unsigned int time_squeeze;
3254 #ifdef CONFIG_RPS
3255 struct softnet_data *rps_ipi_list;
3256 #endif
3257
3258 bool in_net_rx_action;
3259 bool in_napi_threaded_poll;
3260
3261 #ifdef CONFIG_NET_FLOW_LIMIT
3262 struct sd_flow_limit __rcu *flow_limit;
3263 #endif
3264 struct Qdisc *output_queue;
3265 struct Qdisc **output_queue_tailp;
3266 struct sk_buff *completion_queue;
3267 #ifdef CONFIG_XFRM_OFFLOAD
3268 struct sk_buff_head xfrm_backlog;
3269 #endif
3270 /* written and read only by owning cpu: */
3271 struct {
3272 u16 recursion;
3273 u8 more;
3274 #ifdef CONFIG_NET_EGRESS
3275 u8 skip_txqueue;
3276 #endif
3277 } xmit;
3278 #ifdef CONFIG_RPS
3279 /* input_queue_head should be written by cpu owning this struct,
3280 * and only read by other cpus. Worth using a cache line.
3281 */
3282 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3283
3284 /* Elements below can be accessed between CPUs for RPS/RFS */
3285 call_single_data_t csd ____cacheline_aligned_in_smp;
3286 struct softnet_data *rps_ipi_next;
3287 unsigned int cpu;
3288 unsigned int input_queue_tail;
3289 #endif
3290 unsigned int received_rps;
3291 unsigned int dropped;
3292 struct sk_buff_head input_pkt_queue;
3293 struct napi_struct backlog;
3294
3295 /* Another possibly contended cache line */
3296 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3297 int defer_count;
3298 int defer_ipi_scheduled;
3299 struct sk_buff *defer_list;
3300 call_single_data_t defer_csd;
3301 };
3302
input_queue_head_incr(struct softnet_data * sd)3303 static inline void input_queue_head_incr(struct softnet_data *sd)
3304 {
3305 #ifdef CONFIG_RPS
3306 sd->input_queue_head++;
3307 #endif
3308 }
3309
input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)3310 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3311 unsigned int *qtail)
3312 {
3313 #ifdef CONFIG_RPS
3314 *qtail = ++sd->input_queue_tail;
3315 #endif
3316 }
3317
3318 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3319
dev_recursion_level(void)3320 static inline int dev_recursion_level(void)
3321 {
3322 return this_cpu_read(softnet_data.xmit.recursion);
3323 }
3324
3325 #define XMIT_RECURSION_LIMIT 8
dev_xmit_recursion(void)3326 static inline bool dev_xmit_recursion(void)
3327 {
3328 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3329 XMIT_RECURSION_LIMIT);
3330 }
3331
dev_xmit_recursion_inc(void)3332 static inline void dev_xmit_recursion_inc(void)
3333 {
3334 __this_cpu_inc(softnet_data.xmit.recursion);
3335 }
3336
dev_xmit_recursion_dec(void)3337 static inline void dev_xmit_recursion_dec(void)
3338 {
3339 __this_cpu_dec(softnet_data.xmit.recursion);
3340 }
3341
3342 void __netif_schedule(struct Qdisc *q);
3343 void netif_schedule_queue(struct netdev_queue *txq);
3344
netif_tx_schedule_all(struct net_device * dev)3345 static inline void netif_tx_schedule_all(struct net_device *dev)
3346 {
3347 unsigned int i;
3348
3349 for (i = 0; i < dev->num_tx_queues; i++)
3350 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3351 }
3352
netif_tx_start_queue(struct netdev_queue * dev_queue)3353 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3354 {
3355 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3356 }
3357
3358 /**
3359 * netif_start_queue - allow transmit
3360 * @dev: network device
3361 *
3362 * Allow upper layers to call the device hard_start_xmit routine.
3363 */
netif_start_queue(struct net_device * dev)3364 static inline void netif_start_queue(struct net_device *dev)
3365 {
3366 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3367 }
3368
netif_tx_start_all_queues(struct net_device * dev)3369 static inline void netif_tx_start_all_queues(struct net_device *dev)
3370 {
3371 unsigned int i;
3372
3373 for (i = 0; i < dev->num_tx_queues; i++) {
3374 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3375 netif_tx_start_queue(txq);
3376 }
3377 }
3378
3379 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3380
3381 /**
3382 * netif_wake_queue - restart transmit
3383 * @dev: network device
3384 *
3385 * Allow upper layers to call the device hard_start_xmit routine.
3386 * Used for flow control when transmit resources are available.
3387 */
netif_wake_queue(struct net_device * dev)3388 static inline void netif_wake_queue(struct net_device *dev)
3389 {
3390 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3391 }
3392
netif_tx_wake_all_queues(struct net_device * dev)3393 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3394 {
3395 unsigned int i;
3396
3397 for (i = 0; i < dev->num_tx_queues; i++) {
3398 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3399 netif_tx_wake_queue(txq);
3400 }
3401 }
3402
netif_tx_stop_queue(struct netdev_queue * dev_queue)3403 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3404 {
3405 /* Must be an atomic op see netif_txq_try_stop() */
3406 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3407 }
3408
3409 /**
3410 * netif_stop_queue - stop transmitted packets
3411 * @dev: network device
3412 *
3413 * Stop upper layers calling the device hard_start_xmit routine.
3414 * Used for flow control when transmit resources are unavailable.
3415 */
netif_stop_queue(struct net_device * dev)3416 static inline void netif_stop_queue(struct net_device *dev)
3417 {
3418 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3419 }
3420
3421 void netif_tx_stop_all_queues(struct net_device *dev);
3422
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)3423 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3424 {
3425 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3426 }
3427
3428 /**
3429 * netif_queue_stopped - test if transmit queue is flowblocked
3430 * @dev: network device
3431 *
3432 * Test if transmit queue on device is currently unable to send.
3433 */
netif_queue_stopped(const struct net_device * dev)3434 static inline bool netif_queue_stopped(const struct net_device *dev)
3435 {
3436 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3437 }
3438
netif_xmit_stopped(const struct netdev_queue * dev_queue)3439 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3440 {
3441 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3442 }
3443
3444 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)3445 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3446 {
3447 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3448 }
3449
3450 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)3451 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3452 {
3453 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3454 }
3455
3456 /**
3457 * netdev_queue_set_dql_min_limit - set dql minimum limit
3458 * @dev_queue: pointer to transmit queue
3459 * @min_limit: dql minimum limit
3460 *
3461 * Forces xmit_more() to return true until the minimum threshold
3462 * defined by @min_limit is reached (or until the tx queue is
3463 * empty). Warning: to be use with care, misuse will impact the
3464 * latency.
3465 */
netdev_queue_set_dql_min_limit(struct netdev_queue * dev_queue,unsigned int min_limit)3466 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3467 unsigned int min_limit)
3468 {
3469 #ifdef CONFIG_BQL
3470 dev_queue->dql.min_limit = min_limit;
3471 #endif
3472 }
3473
3474 /**
3475 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3476 * @dev_queue: pointer to transmit queue
3477 *
3478 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3479 * to give appropriate hint to the CPU.
3480 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)3481 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3482 {
3483 #ifdef CONFIG_BQL
3484 prefetchw(&dev_queue->dql.num_queued);
3485 #endif
3486 }
3487
3488 /**
3489 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3490 * @dev_queue: pointer to transmit queue
3491 *
3492 * BQL enabled drivers might use this helper in their TX completion path,
3493 * to give appropriate hint to the CPU.
3494 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3495 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3496 {
3497 #ifdef CONFIG_BQL
3498 prefetchw(&dev_queue->dql.limit);
3499 #endif
3500 }
3501
3502 /**
3503 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3504 * @dev_queue: network device queue
3505 * @bytes: number of bytes queued to the device queue
3506 *
3507 * Report the number of bytes queued for sending/completion to the network
3508 * device hardware queue. @bytes should be a good approximation and should
3509 * exactly match netdev_completed_queue() @bytes.
3510 * This is typically called once per packet, from ndo_start_xmit().
3511 */
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3512 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3513 unsigned int bytes)
3514 {
3515 #ifdef CONFIG_BQL
3516 dql_queued(&dev_queue->dql, bytes);
3517
3518 if (likely(dql_avail(&dev_queue->dql) >= 0))
3519 return;
3520
3521 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3522
3523 /*
3524 * The XOFF flag must be set before checking the dql_avail below,
3525 * because in netdev_tx_completed_queue we update the dql_completed
3526 * before checking the XOFF flag.
3527 */
3528 smp_mb();
3529
3530 /* check again in case another CPU has just made room avail */
3531 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3532 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3533 #endif
3534 }
3535
3536 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3537 * that they should not test BQL status themselves.
3538 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3539 * skb of a batch.
3540 * Returns true if the doorbell must be used to kick the NIC.
3541 */
__netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes,bool xmit_more)3542 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3543 unsigned int bytes,
3544 bool xmit_more)
3545 {
3546 if (xmit_more) {
3547 #ifdef CONFIG_BQL
3548 dql_queued(&dev_queue->dql, bytes);
3549 #endif
3550 return netif_tx_queue_stopped(dev_queue);
3551 }
3552 netdev_tx_sent_queue(dev_queue, bytes);
3553 return true;
3554 }
3555
3556 /**
3557 * netdev_sent_queue - report the number of bytes queued to hardware
3558 * @dev: network device
3559 * @bytes: number of bytes queued to the hardware device queue
3560 *
3561 * Report the number of bytes queued for sending/completion to the network
3562 * device hardware queue#0. @bytes should be a good approximation and should
3563 * exactly match netdev_completed_queue() @bytes.
3564 * This is typically called once per packet, from ndo_start_xmit().
3565 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)3566 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3567 {
3568 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3569 }
3570
__netdev_sent_queue(struct net_device * dev,unsigned int bytes,bool xmit_more)3571 static inline bool __netdev_sent_queue(struct net_device *dev,
3572 unsigned int bytes,
3573 bool xmit_more)
3574 {
3575 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3576 xmit_more);
3577 }
3578
3579 /**
3580 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3581 * @dev_queue: network device queue
3582 * @pkts: number of packets (currently ignored)
3583 * @bytes: number of bytes dequeued from the device queue
3584 *
3585 * Must be called at most once per TX completion round (and not per
3586 * individual packet), so that BQL can adjust its limits appropriately.
3587 */
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3588 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3589 unsigned int pkts, unsigned int bytes)
3590 {
3591 #ifdef CONFIG_BQL
3592 if (unlikely(!bytes))
3593 return;
3594
3595 dql_completed(&dev_queue->dql, bytes);
3596
3597 /*
3598 * Without the memory barrier there is a small possiblity that
3599 * netdev_tx_sent_queue will miss the update and cause the queue to
3600 * be stopped forever
3601 */
3602 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3603
3604 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3605 return;
3606
3607 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3608 netif_schedule_queue(dev_queue);
3609 #endif
3610 }
3611
3612 /**
3613 * netdev_completed_queue - report bytes and packets completed by device
3614 * @dev: network device
3615 * @pkts: actual number of packets sent over the medium
3616 * @bytes: actual number of bytes sent over the medium
3617 *
3618 * Report the number of bytes and packets transmitted by the network device
3619 * hardware queue over the physical medium, @bytes must exactly match the
3620 * @bytes amount passed to netdev_sent_queue()
3621 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3622 static inline void netdev_completed_queue(struct net_device *dev,
3623 unsigned int pkts, unsigned int bytes)
3624 {
3625 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3626 }
3627
netdev_tx_reset_queue(struct netdev_queue * q)3628 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3629 {
3630 #ifdef CONFIG_BQL
3631 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3632 dql_reset(&q->dql);
3633 #endif
3634 }
3635
3636 /**
3637 * netdev_reset_queue - reset the packets and bytes count of a network device
3638 * @dev_queue: network device
3639 *
3640 * Reset the bytes and packet count of a network device and clear the
3641 * software flow control OFF bit for this network device
3642 */
netdev_reset_queue(struct net_device * dev_queue)3643 static inline void netdev_reset_queue(struct net_device *dev_queue)
3644 {
3645 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3646 }
3647
3648 /**
3649 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3650 * @dev: network device
3651 * @queue_index: given tx queue index
3652 *
3653 * Returns 0 if given tx queue index >= number of device tx queues,
3654 * otherwise returns the originally passed tx queue index.
3655 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3656 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3657 {
3658 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3659 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3660 dev->name, queue_index,
3661 dev->real_num_tx_queues);
3662 return 0;
3663 }
3664
3665 return queue_index;
3666 }
3667
3668 /**
3669 * netif_running - test if up
3670 * @dev: network device
3671 *
3672 * Test if the device has been brought up.
3673 */
netif_running(const struct net_device * dev)3674 static inline bool netif_running(const struct net_device *dev)
3675 {
3676 return test_bit(__LINK_STATE_START, &dev->state);
3677 }
3678
3679 /*
3680 * Routines to manage the subqueues on a device. We only need start,
3681 * stop, and a check if it's stopped. All other device management is
3682 * done at the overall netdevice level.
3683 * Also test the device if we're multiqueue.
3684 */
3685
3686 /**
3687 * netif_start_subqueue - allow sending packets on subqueue
3688 * @dev: network device
3689 * @queue_index: sub queue index
3690 *
3691 * Start individual transmit queue of a device with multiple transmit queues.
3692 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)3693 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3694 {
3695 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3696
3697 netif_tx_start_queue(txq);
3698 }
3699
3700 /**
3701 * netif_stop_subqueue - stop sending packets on subqueue
3702 * @dev: network device
3703 * @queue_index: sub queue index
3704 *
3705 * Stop individual transmit queue of a device with multiple transmit queues.
3706 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)3707 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3708 {
3709 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3710 netif_tx_stop_queue(txq);
3711 }
3712
3713 /**
3714 * __netif_subqueue_stopped - test status of subqueue
3715 * @dev: network device
3716 * @queue_index: sub queue index
3717 *
3718 * Check individual transmit queue of a device with multiple transmit queues.
3719 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3720 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3721 u16 queue_index)
3722 {
3723 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3724
3725 return netif_tx_queue_stopped(txq);
3726 }
3727
3728 /**
3729 * netif_subqueue_stopped - test status of subqueue
3730 * @dev: network device
3731 * @skb: sub queue buffer pointer
3732 *
3733 * Check individual transmit queue of a device with multiple transmit queues.
3734 */
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3735 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3736 struct sk_buff *skb)
3737 {
3738 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3739 }
3740
3741 /**
3742 * netif_wake_subqueue - allow sending packets on subqueue
3743 * @dev: network device
3744 * @queue_index: sub queue index
3745 *
3746 * Resume individual transmit queue of a device with multiple transmit queues.
3747 */
netif_wake_subqueue(struct net_device * dev,u16 queue_index)3748 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3749 {
3750 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3751
3752 netif_tx_wake_queue(txq);
3753 }
3754
3755 #ifdef CONFIG_XPS
3756 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3757 u16 index);
3758 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3759 u16 index, enum xps_map_type type);
3760
3761 /**
3762 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3763 * @j: CPU/Rx queue index
3764 * @mask: bitmask of all cpus/rx queues
3765 * @nr_bits: number of bits in the bitmask
3766 *
3767 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3768 */
netif_attr_test_mask(unsigned long j,const unsigned long * mask,unsigned int nr_bits)3769 static inline bool netif_attr_test_mask(unsigned long j,
3770 const unsigned long *mask,
3771 unsigned int nr_bits)
3772 {
3773 cpu_max_bits_warn(j, nr_bits);
3774 return test_bit(j, mask);
3775 }
3776
3777 /**
3778 * netif_attr_test_online - Test for online CPU/Rx queue
3779 * @j: CPU/Rx queue index
3780 * @online_mask: bitmask for CPUs/Rx queues that are online
3781 * @nr_bits: number of bits in the bitmask
3782 *
3783 * Returns true if a CPU/Rx queue is online.
3784 */
netif_attr_test_online(unsigned long j,const unsigned long * online_mask,unsigned int nr_bits)3785 static inline bool netif_attr_test_online(unsigned long j,
3786 const unsigned long *online_mask,
3787 unsigned int nr_bits)
3788 {
3789 cpu_max_bits_warn(j, nr_bits);
3790
3791 if (online_mask)
3792 return test_bit(j, online_mask);
3793
3794 return (j < nr_bits);
3795 }
3796
3797 /**
3798 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3799 * @n: CPU/Rx queue index
3800 * @srcp: the cpumask/Rx queue mask pointer
3801 * @nr_bits: number of bits in the bitmask
3802 *
3803 * Returns >= nr_bits if no further CPUs/Rx queues set.
3804 */
netif_attrmask_next(int n,const unsigned long * srcp,unsigned int nr_bits)3805 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3806 unsigned int nr_bits)
3807 {
3808 /* -1 is a legal arg here. */
3809 if (n != -1)
3810 cpu_max_bits_warn(n, nr_bits);
3811
3812 if (srcp)
3813 return find_next_bit(srcp, nr_bits, n + 1);
3814
3815 return n + 1;
3816 }
3817
3818 /**
3819 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3820 * @n: CPU/Rx queue index
3821 * @src1p: the first CPUs/Rx queues mask pointer
3822 * @src2p: the second CPUs/Rx queues mask pointer
3823 * @nr_bits: number of bits in the bitmask
3824 *
3825 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3826 */
netif_attrmask_next_and(int n,const unsigned long * src1p,const unsigned long * src2p,unsigned int nr_bits)3827 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3828 const unsigned long *src2p,
3829 unsigned int nr_bits)
3830 {
3831 /* -1 is a legal arg here. */
3832 if (n != -1)
3833 cpu_max_bits_warn(n, nr_bits);
3834
3835 if (src1p && src2p)
3836 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3837 else if (src1p)
3838 return find_next_bit(src1p, nr_bits, n + 1);
3839 else if (src2p)
3840 return find_next_bit(src2p, nr_bits, n + 1);
3841
3842 return n + 1;
3843 }
3844 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3845 static inline int netif_set_xps_queue(struct net_device *dev,
3846 const struct cpumask *mask,
3847 u16 index)
3848 {
3849 return 0;
3850 }
3851
__netif_set_xps_queue(struct net_device * dev,const unsigned long * mask,u16 index,enum xps_map_type type)3852 static inline int __netif_set_xps_queue(struct net_device *dev,
3853 const unsigned long *mask,
3854 u16 index, enum xps_map_type type)
3855 {
3856 return 0;
3857 }
3858 #endif
3859
3860 /**
3861 * netif_is_multiqueue - test if device has multiple transmit queues
3862 * @dev: network device
3863 *
3864 * Check if device has multiple transmit queues
3865 */
netif_is_multiqueue(const struct net_device * dev)3866 static inline bool netif_is_multiqueue(const struct net_device *dev)
3867 {
3868 return dev->num_tx_queues > 1;
3869 }
3870
3871 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3872
3873 #ifdef CONFIG_SYSFS
3874 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3875 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxqs)3876 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3877 unsigned int rxqs)
3878 {
3879 dev->real_num_rx_queues = rxqs;
3880 return 0;
3881 }
3882 #endif
3883 int netif_set_real_num_queues(struct net_device *dev,
3884 unsigned int txq, unsigned int rxq);
3885
3886 int netif_get_num_default_rss_queues(void);
3887
3888 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3889 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3890
3891 /*
3892 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3893 * interrupt context or with hardware interrupts being disabled.
3894 * (in_hardirq() || irqs_disabled())
3895 *
3896 * We provide four helpers that can be used in following contexts :
3897 *
3898 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3899 * replacing kfree_skb(skb)
3900 *
3901 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3902 * Typically used in place of consume_skb(skb) in TX completion path
3903 *
3904 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3905 * replacing kfree_skb(skb)
3906 *
3907 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3908 * and consumed a packet. Used in place of consume_skb(skb)
3909 */
dev_kfree_skb_irq(struct sk_buff * skb)3910 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3911 {
3912 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3913 }
3914
dev_consume_skb_irq(struct sk_buff * skb)3915 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3916 {
3917 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3918 }
3919
dev_kfree_skb_any(struct sk_buff * skb)3920 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3921 {
3922 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3923 }
3924
dev_consume_skb_any(struct sk_buff * skb)3925 static inline void dev_consume_skb_any(struct sk_buff *skb)
3926 {
3927 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3928 }
3929
3930 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3931 struct bpf_prog *xdp_prog);
3932 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3933 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3934 int netif_rx(struct sk_buff *skb);
3935 int __netif_rx(struct sk_buff *skb);
3936
3937 int netif_receive_skb(struct sk_buff *skb);
3938 int netif_receive_skb_core(struct sk_buff *skb);
3939 void netif_receive_skb_list_internal(struct list_head *head);
3940 void netif_receive_skb_list(struct list_head *head);
3941 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3942 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3943 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3944 void napi_get_frags_check(struct napi_struct *napi);
3945 gro_result_t napi_gro_frags(struct napi_struct *napi);
3946 struct packet_offload *gro_find_receive_by_type(__be16 type);
3947 struct packet_offload *gro_find_complete_by_type(__be16 type);
3948
napi_free_frags(struct napi_struct * napi)3949 static inline void napi_free_frags(struct napi_struct *napi)
3950 {
3951 kfree_skb(napi->skb);
3952 napi->skb = NULL;
3953 }
3954
3955 bool netdev_is_rx_handler_busy(struct net_device *dev);
3956 int netdev_rx_handler_register(struct net_device *dev,
3957 rx_handler_func_t *rx_handler,
3958 void *rx_handler_data);
3959 void netdev_rx_handler_unregister(struct net_device *dev);
3960
3961 bool dev_valid_name(const char *name);
is_socket_ioctl_cmd(unsigned int cmd)3962 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3963 {
3964 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3965 }
3966 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3967 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3968 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3969 void __user *data, bool *need_copyout);
3970 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3971 int generic_hwtstamp_get_lower(struct net_device *dev,
3972 struct kernel_hwtstamp_config *kernel_cfg);
3973 int generic_hwtstamp_set_lower(struct net_device *dev,
3974 struct kernel_hwtstamp_config *kernel_cfg,
3975 struct netlink_ext_ack *extack);
3976 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3977 unsigned int dev_get_flags(const struct net_device *);
3978 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3979 struct netlink_ext_ack *extack);
3980 int dev_change_flags(struct net_device *dev, unsigned int flags,
3981 struct netlink_ext_ack *extack);
3982 int dev_set_alias(struct net_device *, const char *, size_t);
3983 int dev_get_alias(const struct net_device *, char *, size_t);
3984 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3985 const char *pat, int new_ifindex);
3986 static inline
dev_change_net_namespace(struct net_device * dev,struct net * net,const char * pat)3987 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3988 const char *pat)
3989 {
3990 return __dev_change_net_namespace(dev, net, pat, 0);
3991 }
3992 int __dev_set_mtu(struct net_device *, int);
3993 int dev_set_mtu(struct net_device *, int);
3994 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3995 struct netlink_ext_ack *extack);
3996 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3997 struct netlink_ext_ack *extack);
3998 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3999 struct netlink_ext_ack *extack);
4000 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4001 int dev_get_port_parent_id(struct net_device *dev,
4002 struct netdev_phys_item_id *ppid, bool recurse);
4003 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4004 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4005 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4006 struct netdev_queue *txq, int *ret);
4007
4008 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4009 u8 dev_xdp_prog_count(struct net_device *dev);
4010 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4011
4012 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4013 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4014 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4015 bool is_skb_forwardable(const struct net_device *dev,
4016 const struct sk_buff *skb);
4017
__is_skb_forwardable(const struct net_device * dev,const struct sk_buff * skb,const bool check_mtu)4018 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4019 const struct sk_buff *skb,
4020 const bool check_mtu)
4021 {
4022 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4023 unsigned int len;
4024
4025 if (!(dev->flags & IFF_UP))
4026 return false;
4027
4028 if (!check_mtu)
4029 return true;
4030
4031 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4032 if (skb->len <= len)
4033 return true;
4034
4035 /* if TSO is enabled, we don't care about the length as the packet
4036 * could be forwarded without being segmented before
4037 */
4038 if (skb_is_gso(skb))
4039 return true;
4040
4041 return false;
4042 }
4043
4044 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
4045
dev_core_stats(struct net_device * dev)4046 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
4047 {
4048 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
4049 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
4050
4051 if (likely(p))
4052 return p;
4053
4054 return netdev_core_stats_alloc(dev);
4055 }
4056
4057 #define DEV_CORE_STATS_INC(FIELD) \
4058 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4059 { \
4060 struct net_device_core_stats __percpu *p; \
4061 \
4062 p = dev_core_stats(dev); \
4063 if (p) \
4064 this_cpu_inc(p->FIELD); \
4065 }
4066 DEV_CORE_STATS_INC(rx_dropped)
DEV_CORE_STATS_INC(tx_dropped)4067 DEV_CORE_STATS_INC(tx_dropped)
4068 DEV_CORE_STATS_INC(rx_nohandler)
4069 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4070
4071 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4072 struct sk_buff *skb,
4073 const bool check_mtu)
4074 {
4075 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4076 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4077 dev_core_stats_rx_dropped_inc(dev);
4078 kfree_skb(skb);
4079 return NET_RX_DROP;
4080 }
4081
4082 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4083 skb->priority = 0;
4084 return 0;
4085 }
4086
4087 bool dev_nit_active(struct net_device *dev);
4088 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4089
__dev_put(struct net_device * dev)4090 static inline void __dev_put(struct net_device *dev)
4091 {
4092 if (dev) {
4093 #ifdef CONFIG_PCPU_DEV_REFCNT
4094 this_cpu_dec(*dev->pcpu_refcnt);
4095 #else
4096 refcount_dec(&dev->dev_refcnt);
4097 #endif
4098 }
4099 }
4100
__dev_hold(struct net_device * dev)4101 static inline void __dev_hold(struct net_device *dev)
4102 {
4103 if (dev) {
4104 #ifdef CONFIG_PCPU_DEV_REFCNT
4105 this_cpu_inc(*dev->pcpu_refcnt);
4106 #else
4107 refcount_inc(&dev->dev_refcnt);
4108 #endif
4109 }
4110 }
4111
__netdev_tracker_alloc(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4112 static inline void __netdev_tracker_alloc(struct net_device *dev,
4113 netdevice_tracker *tracker,
4114 gfp_t gfp)
4115 {
4116 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4117 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4118 #endif
4119 }
4120
4121 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4122 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4123 */
netdev_tracker_alloc(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4124 static inline void netdev_tracker_alloc(struct net_device *dev,
4125 netdevice_tracker *tracker, gfp_t gfp)
4126 {
4127 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4128 refcount_dec(&dev->refcnt_tracker.no_tracker);
4129 __netdev_tracker_alloc(dev, tracker, gfp);
4130 #endif
4131 }
4132
netdev_tracker_free(struct net_device * dev,netdevice_tracker * tracker)4133 static inline void netdev_tracker_free(struct net_device *dev,
4134 netdevice_tracker *tracker)
4135 {
4136 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4137 ref_tracker_free(&dev->refcnt_tracker, tracker);
4138 #endif
4139 }
4140
netdev_hold(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4141 static inline void netdev_hold(struct net_device *dev,
4142 netdevice_tracker *tracker, gfp_t gfp)
4143 {
4144 if (dev) {
4145 __dev_hold(dev);
4146 __netdev_tracker_alloc(dev, tracker, gfp);
4147 }
4148 }
4149
netdev_put(struct net_device * dev,netdevice_tracker * tracker)4150 static inline void netdev_put(struct net_device *dev,
4151 netdevice_tracker *tracker)
4152 {
4153 if (dev) {
4154 netdev_tracker_free(dev, tracker);
4155 __dev_put(dev);
4156 }
4157 }
4158
4159 /**
4160 * dev_hold - get reference to device
4161 * @dev: network device
4162 *
4163 * Hold reference to device to keep it from being freed.
4164 * Try using netdev_hold() instead.
4165 */
dev_hold(struct net_device * dev)4166 static inline void dev_hold(struct net_device *dev)
4167 {
4168 netdev_hold(dev, NULL, GFP_ATOMIC);
4169 }
4170
4171 /**
4172 * dev_put - release reference to device
4173 * @dev: network device
4174 *
4175 * Release reference to device to allow it to be freed.
4176 * Try using netdev_put() instead.
4177 */
dev_put(struct net_device * dev)4178 static inline void dev_put(struct net_device *dev)
4179 {
4180 netdev_put(dev, NULL);
4181 }
4182
netdev_ref_replace(struct net_device * odev,struct net_device * ndev,netdevice_tracker * tracker,gfp_t gfp)4183 static inline void netdev_ref_replace(struct net_device *odev,
4184 struct net_device *ndev,
4185 netdevice_tracker *tracker,
4186 gfp_t gfp)
4187 {
4188 if (odev)
4189 netdev_tracker_free(odev, tracker);
4190
4191 __dev_hold(ndev);
4192 __dev_put(odev);
4193
4194 if (ndev)
4195 __netdev_tracker_alloc(ndev, tracker, gfp);
4196 }
4197
4198 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4199 * and _off may be called from IRQ context, but it is caller
4200 * who is responsible for serialization of these calls.
4201 *
4202 * The name carrier is inappropriate, these functions should really be
4203 * called netif_lowerlayer_*() because they represent the state of any
4204 * kind of lower layer not just hardware media.
4205 */
4206 void linkwatch_fire_event(struct net_device *dev);
4207
4208 /**
4209 * netif_carrier_ok - test if carrier present
4210 * @dev: network device
4211 *
4212 * Check if carrier is present on device
4213 */
netif_carrier_ok(const struct net_device * dev)4214 static inline bool netif_carrier_ok(const struct net_device *dev)
4215 {
4216 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4217 }
4218
4219 unsigned long dev_trans_start(struct net_device *dev);
4220
4221 void __netdev_watchdog_up(struct net_device *dev);
4222
4223 void netif_carrier_on(struct net_device *dev);
4224 void netif_carrier_off(struct net_device *dev);
4225 void netif_carrier_event(struct net_device *dev);
4226
4227 /**
4228 * netif_dormant_on - mark device as dormant.
4229 * @dev: network device
4230 *
4231 * Mark device as dormant (as per RFC2863).
4232 *
4233 * The dormant state indicates that the relevant interface is not
4234 * actually in a condition to pass packets (i.e., it is not 'up') but is
4235 * in a "pending" state, waiting for some external event. For "on-
4236 * demand" interfaces, this new state identifies the situation where the
4237 * interface is waiting for events to place it in the up state.
4238 */
netif_dormant_on(struct net_device * dev)4239 static inline void netif_dormant_on(struct net_device *dev)
4240 {
4241 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4242 linkwatch_fire_event(dev);
4243 }
4244
4245 /**
4246 * netif_dormant_off - set device as not dormant.
4247 * @dev: network device
4248 *
4249 * Device is not in dormant state.
4250 */
netif_dormant_off(struct net_device * dev)4251 static inline void netif_dormant_off(struct net_device *dev)
4252 {
4253 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4254 linkwatch_fire_event(dev);
4255 }
4256
4257 /**
4258 * netif_dormant - test if device is dormant
4259 * @dev: network device
4260 *
4261 * Check if device is dormant.
4262 */
netif_dormant(const struct net_device * dev)4263 static inline bool netif_dormant(const struct net_device *dev)
4264 {
4265 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4266 }
4267
4268
4269 /**
4270 * netif_testing_on - mark device as under test.
4271 * @dev: network device
4272 *
4273 * Mark device as under test (as per RFC2863).
4274 *
4275 * The testing state indicates that some test(s) must be performed on
4276 * the interface. After completion, of the test, the interface state
4277 * will change to up, dormant, or down, as appropriate.
4278 */
netif_testing_on(struct net_device * dev)4279 static inline void netif_testing_on(struct net_device *dev)
4280 {
4281 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4282 linkwatch_fire_event(dev);
4283 }
4284
4285 /**
4286 * netif_testing_off - set device as not under test.
4287 * @dev: network device
4288 *
4289 * Device is not in testing state.
4290 */
netif_testing_off(struct net_device * dev)4291 static inline void netif_testing_off(struct net_device *dev)
4292 {
4293 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4294 linkwatch_fire_event(dev);
4295 }
4296
4297 /**
4298 * netif_testing - test if device is under test
4299 * @dev: network device
4300 *
4301 * Check if device is under test
4302 */
netif_testing(const struct net_device * dev)4303 static inline bool netif_testing(const struct net_device *dev)
4304 {
4305 return test_bit(__LINK_STATE_TESTING, &dev->state);
4306 }
4307
4308
4309 /**
4310 * netif_oper_up - test if device is operational
4311 * @dev: network device
4312 *
4313 * Check if carrier is operational
4314 */
netif_oper_up(const struct net_device * dev)4315 static inline bool netif_oper_up(const struct net_device *dev)
4316 {
4317 return (dev->operstate == IF_OPER_UP ||
4318 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4319 }
4320
4321 /**
4322 * netif_device_present - is device available or removed
4323 * @dev: network device
4324 *
4325 * Check if device has not been removed from system.
4326 */
netif_device_present(const struct net_device * dev)4327 static inline bool netif_device_present(const struct net_device *dev)
4328 {
4329 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4330 }
4331
4332 void netif_device_detach(struct net_device *dev);
4333
4334 void netif_device_attach(struct net_device *dev);
4335
4336 /*
4337 * Network interface message level settings
4338 */
4339
4340 enum {
4341 NETIF_MSG_DRV_BIT,
4342 NETIF_MSG_PROBE_BIT,
4343 NETIF_MSG_LINK_BIT,
4344 NETIF_MSG_TIMER_BIT,
4345 NETIF_MSG_IFDOWN_BIT,
4346 NETIF_MSG_IFUP_BIT,
4347 NETIF_MSG_RX_ERR_BIT,
4348 NETIF_MSG_TX_ERR_BIT,
4349 NETIF_MSG_TX_QUEUED_BIT,
4350 NETIF_MSG_INTR_BIT,
4351 NETIF_MSG_TX_DONE_BIT,
4352 NETIF_MSG_RX_STATUS_BIT,
4353 NETIF_MSG_PKTDATA_BIT,
4354 NETIF_MSG_HW_BIT,
4355 NETIF_MSG_WOL_BIT,
4356
4357 /* When you add a new bit above, update netif_msg_class_names array
4358 * in net/ethtool/common.c
4359 */
4360 NETIF_MSG_CLASS_COUNT,
4361 };
4362 /* Both ethtool_ops interface and internal driver implementation use u32 */
4363 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4364
4365 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4366 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4367
4368 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4369 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4370 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4371 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4372 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4373 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4374 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4375 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4376 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4377 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4378 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4379 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4380 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4381 #define NETIF_MSG_HW __NETIF_MSG(HW)
4382 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4383
4384 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4385 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4386 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4387 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4388 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4389 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4390 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4391 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4392 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4393 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4394 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4395 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4396 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4397 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4398 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4399
netif_msg_init(int debug_value,int default_msg_enable_bits)4400 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4401 {
4402 /* use default */
4403 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4404 return default_msg_enable_bits;
4405 if (debug_value == 0) /* no output */
4406 return 0;
4407 /* set low N bits */
4408 return (1U << debug_value) - 1;
4409 }
4410
__netif_tx_lock(struct netdev_queue * txq,int cpu)4411 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4412 {
4413 spin_lock(&txq->_xmit_lock);
4414 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4415 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4416 }
4417
__netif_tx_acquire(struct netdev_queue * txq)4418 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4419 {
4420 __acquire(&txq->_xmit_lock);
4421 return true;
4422 }
4423
__netif_tx_release(struct netdev_queue * txq)4424 static inline void __netif_tx_release(struct netdev_queue *txq)
4425 {
4426 __release(&txq->_xmit_lock);
4427 }
4428
__netif_tx_lock_bh(struct netdev_queue * txq)4429 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4430 {
4431 spin_lock_bh(&txq->_xmit_lock);
4432 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4433 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4434 }
4435
__netif_tx_trylock(struct netdev_queue * txq)4436 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4437 {
4438 bool ok = spin_trylock(&txq->_xmit_lock);
4439
4440 if (likely(ok)) {
4441 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4442 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4443 }
4444 return ok;
4445 }
4446
__netif_tx_unlock(struct netdev_queue * txq)4447 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4448 {
4449 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4450 WRITE_ONCE(txq->xmit_lock_owner, -1);
4451 spin_unlock(&txq->_xmit_lock);
4452 }
4453
__netif_tx_unlock_bh(struct netdev_queue * txq)4454 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4455 {
4456 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4457 WRITE_ONCE(txq->xmit_lock_owner, -1);
4458 spin_unlock_bh(&txq->_xmit_lock);
4459 }
4460
4461 /*
4462 * txq->trans_start can be read locklessly from dev_watchdog()
4463 */
txq_trans_update(struct netdev_queue * txq)4464 static inline void txq_trans_update(struct netdev_queue *txq)
4465 {
4466 if (txq->xmit_lock_owner != -1)
4467 WRITE_ONCE(txq->trans_start, jiffies);
4468 }
4469
txq_trans_cond_update(struct netdev_queue * txq)4470 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4471 {
4472 unsigned long now = jiffies;
4473
4474 if (READ_ONCE(txq->trans_start) != now)
4475 WRITE_ONCE(txq->trans_start, now);
4476 }
4477
4478 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
netif_trans_update(struct net_device * dev)4479 static inline void netif_trans_update(struct net_device *dev)
4480 {
4481 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4482
4483 txq_trans_cond_update(txq);
4484 }
4485
4486 /**
4487 * netif_tx_lock - grab network device transmit lock
4488 * @dev: network device
4489 *
4490 * Get network device transmit lock
4491 */
4492 void netif_tx_lock(struct net_device *dev);
4493
netif_tx_lock_bh(struct net_device * dev)4494 static inline void netif_tx_lock_bh(struct net_device *dev)
4495 {
4496 local_bh_disable();
4497 netif_tx_lock(dev);
4498 }
4499
4500 void netif_tx_unlock(struct net_device *dev);
4501
netif_tx_unlock_bh(struct net_device * dev)4502 static inline void netif_tx_unlock_bh(struct net_device *dev)
4503 {
4504 netif_tx_unlock(dev);
4505 local_bh_enable();
4506 }
4507
4508 #define HARD_TX_LOCK(dev, txq, cpu) { \
4509 if ((dev->features & NETIF_F_LLTX) == 0) { \
4510 __netif_tx_lock(txq, cpu); \
4511 } else { \
4512 __netif_tx_acquire(txq); \
4513 } \
4514 }
4515
4516 #define HARD_TX_TRYLOCK(dev, txq) \
4517 (((dev->features & NETIF_F_LLTX) == 0) ? \
4518 __netif_tx_trylock(txq) : \
4519 __netif_tx_acquire(txq))
4520
4521 #define HARD_TX_UNLOCK(dev, txq) { \
4522 if ((dev->features & NETIF_F_LLTX) == 0) { \
4523 __netif_tx_unlock(txq); \
4524 } else { \
4525 __netif_tx_release(txq); \
4526 } \
4527 }
4528
netif_tx_disable(struct net_device * dev)4529 static inline void netif_tx_disable(struct net_device *dev)
4530 {
4531 unsigned int i;
4532 int cpu;
4533
4534 local_bh_disable();
4535 cpu = smp_processor_id();
4536 spin_lock(&dev->tx_global_lock);
4537 for (i = 0; i < dev->num_tx_queues; i++) {
4538 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4539
4540 __netif_tx_lock(txq, cpu);
4541 netif_tx_stop_queue(txq);
4542 __netif_tx_unlock(txq);
4543 }
4544 spin_unlock(&dev->tx_global_lock);
4545 local_bh_enable();
4546 }
4547
netif_addr_lock(struct net_device * dev)4548 static inline void netif_addr_lock(struct net_device *dev)
4549 {
4550 unsigned char nest_level = 0;
4551
4552 #ifdef CONFIG_LOCKDEP
4553 nest_level = dev->nested_level;
4554 #endif
4555 spin_lock_nested(&dev->addr_list_lock, nest_level);
4556 }
4557
netif_addr_lock_bh(struct net_device * dev)4558 static inline void netif_addr_lock_bh(struct net_device *dev)
4559 {
4560 unsigned char nest_level = 0;
4561
4562 #ifdef CONFIG_LOCKDEP
4563 nest_level = dev->nested_level;
4564 #endif
4565 local_bh_disable();
4566 spin_lock_nested(&dev->addr_list_lock, nest_level);
4567 }
4568
netif_addr_unlock(struct net_device * dev)4569 static inline void netif_addr_unlock(struct net_device *dev)
4570 {
4571 spin_unlock(&dev->addr_list_lock);
4572 }
4573
netif_addr_unlock_bh(struct net_device * dev)4574 static inline void netif_addr_unlock_bh(struct net_device *dev)
4575 {
4576 spin_unlock_bh(&dev->addr_list_lock);
4577 }
4578
4579 /*
4580 * dev_addrs walker. Should be used only for read access. Call with
4581 * rcu_read_lock held.
4582 */
4583 #define for_each_dev_addr(dev, ha) \
4584 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4585
4586 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4587
4588 void ether_setup(struct net_device *dev);
4589
4590 /* Support for loadable net-drivers */
4591 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4592 unsigned char name_assign_type,
4593 void (*setup)(struct net_device *),
4594 unsigned int txqs, unsigned int rxqs);
4595 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4596 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4597
4598 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4599 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4600 count)
4601
4602 int register_netdev(struct net_device *dev);
4603 void unregister_netdev(struct net_device *dev);
4604
4605 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4606
4607 /* General hardware address lists handling functions */
4608 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4609 struct netdev_hw_addr_list *from_list, int addr_len);
4610 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4611 struct netdev_hw_addr_list *from_list, int addr_len);
4612 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4613 struct net_device *dev,
4614 int (*sync)(struct net_device *, const unsigned char *),
4615 int (*unsync)(struct net_device *,
4616 const unsigned char *));
4617 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4618 struct net_device *dev,
4619 int (*sync)(struct net_device *,
4620 const unsigned char *, int),
4621 int (*unsync)(struct net_device *,
4622 const unsigned char *, int));
4623 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4624 struct net_device *dev,
4625 int (*unsync)(struct net_device *,
4626 const unsigned char *, int));
4627 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4628 struct net_device *dev,
4629 int (*unsync)(struct net_device *,
4630 const unsigned char *));
4631 void __hw_addr_init(struct netdev_hw_addr_list *list);
4632
4633 /* Functions used for device addresses handling */
4634 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4635 const void *addr, size_t len);
4636
4637 static inline void
__dev_addr_set(struct net_device * dev,const void * addr,size_t len)4638 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4639 {
4640 dev_addr_mod(dev, 0, addr, len);
4641 }
4642
dev_addr_set(struct net_device * dev,const u8 * addr)4643 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4644 {
4645 __dev_addr_set(dev, addr, dev->addr_len);
4646 }
4647
4648 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4649 unsigned char addr_type);
4650 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4651 unsigned char addr_type);
4652
4653 /* Functions used for unicast addresses handling */
4654 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4655 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4656 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4657 int dev_uc_sync(struct net_device *to, struct net_device *from);
4658 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4659 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4660 void dev_uc_flush(struct net_device *dev);
4661 void dev_uc_init(struct net_device *dev);
4662
4663 /**
4664 * __dev_uc_sync - Synchonize device's unicast list
4665 * @dev: device to sync
4666 * @sync: function to call if address should be added
4667 * @unsync: function to call if address should be removed
4668 *
4669 * Add newly added addresses to the interface, and release
4670 * addresses that have been deleted.
4671 */
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4672 static inline int __dev_uc_sync(struct net_device *dev,
4673 int (*sync)(struct net_device *,
4674 const unsigned char *),
4675 int (*unsync)(struct net_device *,
4676 const unsigned char *))
4677 {
4678 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4679 }
4680
4681 /**
4682 * __dev_uc_unsync - Remove synchronized addresses from device
4683 * @dev: device to sync
4684 * @unsync: function to call if address should be removed
4685 *
4686 * Remove all addresses that were added to the device by dev_uc_sync().
4687 */
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4688 static inline void __dev_uc_unsync(struct net_device *dev,
4689 int (*unsync)(struct net_device *,
4690 const unsigned char *))
4691 {
4692 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4693 }
4694
4695 /* Functions used for multicast addresses handling */
4696 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4697 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4698 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4699 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4700 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4701 int dev_mc_sync(struct net_device *to, struct net_device *from);
4702 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4703 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4704 void dev_mc_flush(struct net_device *dev);
4705 void dev_mc_init(struct net_device *dev);
4706
4707 /**
4708 * __dev_mc_sync - Synchonize device's multicast list
4709 * @dev: device to sync
4710 * @sync: function to call if address should be added
4711 * @unsync: function to call if address should be removed
4712 *
4713 * Add newly added addresses to the interface, and release
4714 * addresses that have been deleted.
4715 */
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4716 static inline int __dev_mc_sync(struct net_device *dev,
4717 int (*sync)(struct net_device *,
4718 const unsigned char *),
4719 int (*unsync)(struct net_device *,
4720 const unsigned char *))
4721 {
4722 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4723 }
4724
4725 /**
4726 * __dev_mc_unsync - Remove synchronized addresses from device
4727 * @dev: device to sync
4728 * @unsync: function to call if address should be removed
4729 *
4730 * Remove all addresses that were added to the device by dev_mc_sync().
4731 */
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4732 static inline void __dev_mc_unsync(struct net_device *dev,
4733 int (*unsync)(struct net_device *,
4734 const unsigned char *))
4735 {
4736 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4737 }
4738
4739 /* Functions used for secondary unicast and multicast support */
4740 void dev_set_rx_mode(struct net_device *dev);
4741 int dev_set_promiscuity(struct net_device *dev, int inc);
4742 int dev_set_allmulti(struct net_device *dev, int inc);
4743 void netdev_state_change(struct net_device *dev);
4744 void __netdev_notify_peers(struct net_device *dev);
4745 void netdev_notify_peers(struct net_device *dev);
4746 void netdev_features_change(struct net_device *dev);
4747 /* Load a device via the kmod */
4748 void dev_load(struct net *net, const char *name);
4749 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4750 struct rtnl_link_stats64 *storage);
4751 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4752 const struct net_device_stats *netdev_stats);
4753 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4754 const struct pcpu_sw_netstats __percpu *netstats);
4755 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4756
4757 extern int netdev_max_backlog;
4758 extern int dev_rx_weight;
4759 extern int dev_tx_weight;
4760 extern int gro_normal_batch;
4761
4762 enum {
4763 NESTED_SYNC_IMM_BIT,
4764 NESTED_SYNC_TODO_BIT,
4765 };
4766
4767 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4768 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4769
4770 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4771 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4772
4773 struct netdev_nested_priv {
4774 unsigned char flags;
4775 void *data;
4776 };
4777
4778 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4779 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4780 struct list_head **iter);
4781
4782 /* iterate through upper list, must be called under RCU read lock */
4783 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4784 for (iter = &(dev)->adj_list.upper, \
4785 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4786 updev; \
4787 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4788
4789 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4790 int (*fn)(struct net_device *upper_dev,
4791 struct netdev_nested_priv *priv),
4792 struct netdev_nested_priv *priv);
4793
4794 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4795 struct net_device *upper_dev);
4796
4797 bool netdev_has_any_upper_dev(struct net_device *dev);
4798
4799 void *netdev_lower_get_next_private(struct net_device *dev,
4800 struct list_head **iter);
4801 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4802 struct list_head **iter);
4803
4804 #define netdev_for_each_lower_private(dev, priv, iter) \
4805 for (iter = (dev)->adj_list.lower.next, \
4806 priv = netdev_lower_get_next_private(dev, &(iter)); \
4807 priv; \
4808 priv = netdev_lower_get_next_private(dev, &(iter)))
4809
4810 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4811 for (iter = &(dev)->adj_list.lower, \
4812 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4813 priv; \
4814 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4815
4816 void *netdev_lower_get_next(struct net_device *dev,
4817 struct list_head **iter);
4818
4819 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4820 for (iter = (dev)->adj_list.lower.next, \
4821 ldev = netdev_lower_get_next(dev, &(iter)); \
4822 ldev; \
4823 ldev = netdev_lower_get_next(dev, &(iter)))
4824
4825 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4826 struct list_head **iter);
4827 int netdev_walk_all_lower_dev(struct net_device *dev,
4828 int (*fn)(struct net_device *lower_dev,
4829 struct netdev_nested_priv *priv),
4830 struct netdev_nested_priv *priv);
4831 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4832 int (*fn)(struct net_device *lower_dev,
4833 struct netdev_nested_priv *priv),
4834 struct netdev_nested_priv *priv);
4835
4836 void *netdev_adjacent_get_private(struct list_head *adj_list);
4837 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4838 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4839 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4840 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4841 struct netlink_ext_ack *extack);
4842 int netdev_master_upper_dev_link(struct net_device *dev,
4843 struct net_device *upper_dev,
4844 void *upper_priv, void *upper_info,
4845 struct netlink_ext_ack *extack);
4846 void netdev_upper_dev_unlink(struct net_device *dev,
4847 struct net_device *upper_dev);
4848 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4849 struct net_device *new_dev,
4850 struct net_device *dev,
4851 struct netlink_ext_ack *extack);
4852 void netdev_adjacent_change_commit(struct net_device *old_dev,
4853 struct net_device *new_dev,
4854 struct net_device *dev);
4855 void netdev_adjacent_change_abort(struct net_device *old_dev,
4856 struct net_device *new_dev,
4857 struct net_device *dev);
4858 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4859 void *netdev_lower_dev_get_private(struct net_device *dev,
4860 struct net_device *lower_dev);
4861 void netdev_lower_state_changed(struct net_device *lower_dev,
4862 void *lower_state_info);
4863
4864 /* RSS keys are 40 or 52 bytes long */
4865 #define NETDEV_RSS_KEY_LEN 52
4866 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4867 void netdev_rss_key_fill(void *buffer, size_t len);
4868
4869 int skb_checksum_help(struct sk_buff *skb);
4870 int skb_crc32c_csum_help(struct sk_buff *skb);
4871 int skb_csum_hwoffload_help(struct sk_buff *skb,
4872 const netdev_features_t features);
4873
4874 struct netdev_bonding_info {
4875 ifslave slave;
4876 ifbond master;
4877 };
4878
4879 struct netdev_notifier_bonding_info {
4880 struct netdev_notifier_info info; /* must be first */
4881 struct netdev_bonding_info bonding_info;
4882 };
4883
4884 void netdev_bonding_info_change(struct net_device *dev,
4885 struct netdev_bonding_info *bonding_info);
4886
4887 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4888 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4889 #else
ethtool_notify(struct net_device * dev,unsigned int cmd,const void * data)4890 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4891 const void *data)
4892 {
4893 }
4894 #endif
4895
4896 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4897
can_checksum_protocol(netdev_features_t features,__be16 protocol)4898 static inline bool can_checksum_protocol(netdev_features_t features,
4899 __be16 protocol)
4900 {
4901 if (protocol == htons(ETH_P_FCOE))
4902 return !!(features & NETIF_F_FCOE_CRC);
4903
4904 /* Assume this is an IP checksum (not SCTP CRC) */
4905
4906 if (features & NETIF_F_HW_CSUM) {
4907 /* Can checksum everything */
4908 return true;
4909 }
4910
4911 switch (protocol) {
4912 case htons(ETH_P_IP):
4913 return !!(features & NETIF_F_IP_CSUM);
4914 case htons(ETH_P_IPV6):
4915 return !!(features & NETIF_F_IPV6_CSUM);
4916 default:
4917 return false;
4918 }
4919 }
4920
4921 #ifdef CONFIG_BUG
4922 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4923 #else
netdev_rx_csum_fault(struct net_device * dev,struct sk_buff * skb)4924 static inline void netdev_rx_csum_fault(struct net_device *dev,
4925 struct sk_buff *skb)
4926 {
4927 }
4928 #endif
4929 /* rx skb timestamps */
4930 void net_enable_timestamp(void);
4931 void net_disable_timestamp(void);
4932
netdev_get_tstamp(struct net_device * dev,const struct skb_shared_hwtstamps * hwtstamps,bool cycles)4933 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4934 const struct skb_shared_hwtstamps *hwtstamps,
4935 bool cycles)
4936 {
4937 const struct net_device_ops *ops = dev->netdev_ops;
4938
4939 if (ops->ndo_get_tstamp)
4940 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4941
4942 return hwtstamps->hwtstamp;
4943 }
4944
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)4945 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4946 struct sk_buff *skb, struct net_device *dev,
4947 bool more)
4948 {
4949 __this_cpu_write(softnet_data.xmit.more, more);
4950 return ops->ndo_start_xmit(skb, dev);
4951 }
4952
netdev_xmit_more(void)4953 static inline bool netdev_xmit_more(void)
4954 {
4955 return __this_cpu_read(softnet_data.xmit.more);
4956 }
4957
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)4958 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4959 struct netdev_queue *txq, bool more)
4960 {
4961 const struct net_device_ops *ops = dev->netdev_ops;
4962 netdev_tx_t rc;
4963
4964 rc = __netdev_start_xmit(ops, skb, dev, more);
4965 if (rc == NETDEV_TX_OK)
4966 txq_trans_update(txq);
4967
4968 return rc;
4969 }
4970
4971 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4972 const void *ns);
4973 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4974 const void *ns);
4975
4976 extern const struct kobj_ns_type_operations net_ns_type_operations;
4977
4978 const char *netdev_drivername(const struct net_device *dev);
4979
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)4980 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4981 netdev_features_t f2)
4982 {
4983 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4984 if (f1 & NETIF_F_HW_CSUM)
4985 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4986 else
4987 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4988 }
4989
4990 return f1 & f2;
4991 }
4992
netdev_get_wanted_features(struct net_device * dev)4993 static inline netdev_features_t netdev_get_wanted_features(
4994 struct net_device *dev)
4995 {
4996 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4997 }
4998 netdev_features_t netdev_increment_features(netdev_features_t all,
4999 netdev_features_t one, netdev_features_t mask);
5000
5001 /* Allow TSO being used on stacked device :
5002 * Performing the GSO segmentation before last device
5003 * is a performance improvement.
5004 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)5005 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5006 netdev_features_t mask)
5007 {
5008 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5009 }
5010
5011 int __netdev_update_features(struct net_device *dev);
5012 void netdev_update_features(struct net_device *dev);
5013 void netdev_change_features(struct net_device *dev);
5014
5015 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5016 struct net_device *dev);
5017
5018 netdev_features_t passthru_features_check(struct sk_buff *skb,
5019 struct net_device *dev,
5020 netdev_features_t features);
5021 netdev_features_t netif_skb_features(struct sk_buff *skb);
5022 void skb_warn_bad_offload(const struct sk_buff *skb);
5023
net_gso_ok(netdev_features_t features,int gso_type)5024 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5025 {
5026 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5027
5028 /* check flags correspondence */
5029 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5030 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5031 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5032 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5033 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5034 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5035 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5036 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5037 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5038 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5039 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5040 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5041 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5042 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5043 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5044 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5045 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5046 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5047 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5048
5049 return (features & feature) == feature;
5050 }
5051
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)5052 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5053 {
5054 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5055 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5056 }
5057
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)5058 static inline bool netif_needs_gso(struct sk_buff *skb,
5059 netdev_features_t features)
5060 {
5061 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5062 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5063 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5064 }
5065
5066 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5067 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5068 void netif_inherit_tso_max(struct net_device *to,
5069 const struct net_device *from);
5070
netif_is_macsec(const struct net_device * dev)5071 static inline bool netif_is_macsec(const struct net_device *dev)
5072 {
5073 return dev->priv_flags & IFF_MACSEC;
5074 }
5075
netif_is_macvlan(const struct net_device * dev)5076 static inline bool netif_is_macvlan(const struct net_device *dev)
5077 {
5078 return dev->priv_flags & IFF_MACVLAN;
5079 }
5080
netif_is_macvlan_port(const struct net_device * dev)5081 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5082 {
5083 return dev->priv_flags & IFF_MACVLAN_PORT;
5084 }
5085
netif_is_bond_master(const struct net_device * dev)5086 static inline bool netif_is_bond_master(const struct net_device *dev)
5087 {
5088 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5089 }
5090
netif_is_bond_slave(const struct net_device * dev)5091 static inline bool netif_is_bond_slave(const struct net_device *dev)
5092 {
5093 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5094 }
5095
netif_supports_nofcs(struct net_device * dev)5096 static inline bool netif_supports_nofcs(struct net_device *dev)
5097 {
5098 return dev->priv_flags & IFF_SUPP_NOFCS;
5099 }
5100
netif_has_l3_rx_handler(const struct net_device * dev)5101 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5102 {
5103 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5104 }
5105
netif_is_l3_master(const struct net_device * dev)5106 static inline bool netif_is_l3_master(const struct net_device *dev)
5107 {
5108 return dev->priv_flags & IFF_L3MDEV_MASTER;
5109 }
5110
netif_is_l3_slave(const struct net_device * dev)5111 static inline bool netif_is_l3_slave(const struct net_device *dev)
5112 {
5113 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5114 }
5115
dev_sdif(const struct net_device * dev)5116 static inline int dev_sdif(const struct net_device *dev)
5117 {
5118 #ifdef CONFIG_NET_L3_MASTER_DEV
5119 if (netif_is_l3_slave(dev))
5120 return dev->ifindex;
5121 #endif
5122 return 0;
5123 }
5124
netif_is_bridge_master(const struct net_device * dev)5125 static inline bool netif_is_bridge_master(const struct net_device *dev)
5126 {
5127 return dev->priv_flags & IFF_EBRIDGE;
5128 }
5129
netif_is_bridge_port(const struct net_device * dev)5130 static inline bool netif_is_bridge_port(const struct net_device *dev)
5131 {
5132 return dev->priv_flags & IFF_BRIDGE_PORT;
5133 }
5134
netif_is_ovs_master(const struct net_device * dev)5135 static inline bool netif_is_ovs_master(const struct net_device *dev)
5136 {
5137 return dev->priv_flags & IFF_OPENVSWITCH;
5138 }
5139
netif_is_ovs_port(const struct net_device * dev)5140 static inline bool netif_is_ovs_port(const struct net_device *dev)
5141 {
5142 return dev->priv_flags & IFF_OVS_DATAPATH;
5143 }
5144
netif_is_any_bridge_master(const struct net_device * dev)5145 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5146 {
5147 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5148 }
5149
netif_is_any_bridge_port(const struct net_device * dev)5150 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5151 {
5152 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5153 }
5154
netif_is_team_master(const struct net_device * dev)5155 static inline bool netif_is_team_master(const struct net_device *dev)
5156 {
5157 return dev->priv_flags & IFF_TEAM;
5158 }
5159
netif_is_team_port(const struct net_device * dev)5160 static inline bool netif_is_team_port(const struct net_device *dev)
5161 {
5162 return dev->priv_flags & IFF_TEAM_PORT;
5163 }
5164
netif_is_lag_master(const struct net_device * dev)5165 static inline bool netif_is_lag_master(const struct net_device *dev)
5166 {
5167 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5168 }
5169
netif_is_lag_port(const struct net_device * dev)5170 static inline bool netif_is_lag_port(const struct net_device *dev)
5171 {
5172 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5173 }
5174
netif_is_rxfh_configured(const struct net_device * dev)5175 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5176 {
5177 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5178 }
5179
netif_is_failover(const struct net_device * dev)5180 static inline bool netif_is_failover(const struct net_device *dev)
5181 {
5182 return dev->priv_flags & IFF_FAILOVER;
5183 }
5184
netif_is_failover_slave(const struct net_device * dev)5185 static inline bool netif_is_failover_slave(const struct net_device *dev)
5186 {
5187 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5188 }
5189
5190 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)5191 static inline void netif_keep_dst(struct net_device *dev)
5192 {
5193 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5194 }
5195
5196 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
netif_reduces_vlan_mtu(struct net_device * dev)5197 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5198 {
5199 /* TODO: reserve and use an additional IFF bit, if we get more users */
5200 return netif_is_macsec(dev);
5201 }
5202
5203 extern struct pernet_operations __net_initdata loopback_net_ops;
5204
5205 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5206
5207 /* netdev_printk helpers, similar to dev_printk */
5208
netdev_name(const struct net_device * dev)5209 static inline const char *netdev_name(const struct net_device *dev)
5210 {
5211 if (!dev->name[0] || strchr(dev->name, '%'))
5212 return "(unnamed net_device)";
5213 return dev->name;
5214 }
5215
netdev_reg_state(const struct net_device * dev)5216 static inline const char *netdev_reg_state(const struct net_device *dev)
5217 {
5218 switch (dev->reg_state) {
5219 case NETREG_UNINITIALIZED: return " (uninitialized)";
5220 case NETREG_REGISTERED: return "";
5221 case NETREG_UNREGISTERING: return " (unregistering)";
5222 case NETREG_UNREGISTERED: return " (unregistered)";
5223 case NETREG_RELEASED: return " (released)";
5224 case NETREG_DUMMY: return " (dummy)";
5225 }
5226
5227 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5228 return " (unknown)";
5229 }
5230
5231 #define MODULE_ALIAS_NETDEV(device) \
5232 MODULE_ALIAS("netdev-" device)
5233
5234 /*
5235 * netdev_WARN() acts like dev_printk(), but with the key difference
5236 * of using a WARN/WARN_ON to get the message out, including the
5237 * file/line information and a backtrace.
5238 */
5239 #define netdev_WARN(dev, format, args...) \
5240 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5241 netdev_reg_state(dev), ##args)
5242
5243 #define netdev_WARN_ONCE(dev, format, args...) \
5244 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5245 netdev_reg_state(dev), ##args)
5246
5247 /*
5248 * The list of packet types we will receive (as opposed to discard)
5249 * and the routines to invoke.
5250 *
5251 * Why 16. Because with 16 the only overlap we get on a hash of the
5252 * low nibble of the protocol value is RARP/SNAP/X.25.
5253 *
5254 * 0800 IP
5255 * 0001 802.3
5256 * 0002 AX.25
5257 * 0004 802.2
5258 * 8035 RARP
5259 * 0005 SNAP
5260 * 0805 X.25
5261 * 0806 ARP
5262 * 8137 IPX
5263 * 0009 Localtalk
5264 * 86DD IPv6
5265 */
5266 #define PTYPE_HASH_SIZE (16)
5267 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5268
5269 extern struct list_head ptype_all __read_mostly;
5270 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5271
5272 extern struct net_device *blackhole_netdev;
5273
5274 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5275 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5276 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5277 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5278 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5279
5280 #endif /* _LINUX_NETDEVICE_H */
5281