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 #include <linux/netdev_features.h>
44 #include <linux/neighbour.h>
45 #include <linux/netdevice_xmit.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_kern;
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 struct ethtool_netdev_state;
84 struct phy_link_topology;
85
86 typedef u32 xdp_features_t;
87
88 void synchronize_net(void);
89 void netdev_set_default_ethtool_ops(struct net_device *dev,
90 const struct ethtool_ops *ops);
91 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
92
93 /* Backlog congestion levels */
94 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
95 #define NET_RX_DROP 1 /* packet dropped */
96
97 #define MAX_NEST_DEV 8
98
99 /*
100 * Transmit return codes: transmit return codes originate from three different
101 * namespaces:
102 *
103 * - qdisc return codes
104 * - driver transmit return codes
105 * - errno values
106 *
107 * Drivers are allowed to return any one of those in their hard_start_xmit()
108 * function. Real network devices commonly used with qdiscs should only return
109 * the driver transmit return codes though - when qdiscs are used, the actual
110 * transmission happens asynchronously, so the value is not propagated to
111 * higher layers. Virtual network devices transmit synchronously; in this case
112 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
113 * others are propagated to higher layers.
114 */
115
116 /* qdisc ->enqueue() return codes. */
117 #define NET_XMIT_SUCCESS 0x00
118 #define NET_XMIT_DROP 0x01 /* skb dropped */
119 #define NET_XMIT_CN 0x02 /* congestion notification */
120 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
121
122 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
123 * indicates that the device will soon be dropping packets, or already drops
124 * some packets of the same priority; prompting us to send less aggressively. */
125 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
126 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
127
128 /* Driver transmit return codes */
129 #define NETDEV_TX_MASK 0xf0
130
131 enum netdev_tx {
132 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
133 NETDEV_TX_OK = 0x00, /* driver took care of packet */
134 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
135 };
136 typedef enum netdev_tx netdev_tx_t;
137
138 /*
139 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
140 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
141 */
dev_xmit_complete(int rc)142 static inline bool dev_xmit_complete(int rc)
143 {
144 /*
145 * Positive cases with an skb consumed by a driver:
146 * - successful transmission (rc == NETDEV_TX_OK)
147 * - error while transmitting (rc < 0)
148 * - error while queueing to a different device (rc & NET_XMIT_MASK)
149 */
150 if (likely(rc < NET_XMIT_MASK))
151 return true;
152
153 return false;
154 }
155
156 /*
157 * Compute the worst-case header length according to the protocols
158 * used.
159 */
160
161 #if defined(CONFIG_HYPERV_NET)
162 # define LL_MAX_HEADER 128
163 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
164 # if defined(CONFIG_MAC80211_MESH)
165 # define LL_MAX_HEADER 128
166 # else
167 # define LL_MAX_HEADER 96
168 # endif
169 #else
170 # define LL_MAX_HEADER 32
171 #endif
172
173 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
174 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
175 #define MAX_HEADER LL_MAX_HEADER
176 #else
177 #define MAX_HEADER (LL_MAX_HEADER + 48)
178 #endif
179
180 /*
181 * Old network device statistics. Fields are native words
182 * (unsigned long) so they can be read and written atomically.
183 */
184
185 #define NET_DEV_STAT(FIELD) \
186 union { \
187 unsigned long FIELD; \
188 atomic_long_t __##FIELD; \
189 }
190
191 struct net_device_stats {
192 NET_DEV_STAT(rx_packets);
193 NET_DEV_STAT(tx_packets);
194 NET_DEV_STAT(rx_bytes);
195 NET_DEV_STAT(tx_bytes);
196 NET_DEV_STAT(rx_errors);
197 NET_DEV_STAT(tx_errors);
198 NET_DEV_STAT(rx_dropped);
199 NET_DEV_STAT(tx_dropped);
200 NET_DEV_STAT(multicast);
201 NET_DEV_STAT(collisions);
202 NET_DEV_STAT(rx_length_errors);
203 NET_DEV_STAT(rx_over_errors);
204 NET_DEV_STAT(rx_crc_errors);
205 NET_DEV_STAT(rx_frame_errors);
206 NET_DEV_STAT(rx_fifo_errors);
207 NET_DEV_STAT(rx_missed_errors);
208 NET_DEV_STAT(tx_aborted_errors);
209 NET_DEV_STAT(tx_carrier_errors);
210 NET_DEV_STAT(tx_fifo_errors);
211 NET_DEV_STAT(tx_heartbeat_errors);
212 NET_DEV_STAT(tx_window_errors);
213 NET_DEV_STAT(rx_compressed);
214 NET_DEV_STAT(tx_compressed);
215 };
216 #undef NET_DEV_STAT
217
218 /* per-cpu stats, allocated on demand.
219 * Try to fit them in a single cache line, for dev_get_stats() sake.
220 */
221 struct net_device_core_stats {
222 unsigned long rx_dropped;
223 unsigned long tx_dropped;
224 unsigned long rx_nohandler;
225 unsigned long rx_otherhost_dropped;
226 } __aligned(4 * sizeof(unsigned long));
227
228 #include <linux/cache.h>
229 #include <linux/skbuff.h>
230
231 struct neighbour;
232 struct neigh_parms;
233 struct sk_buff;
234
235 struct netdev_hw_addr {
236 struct list_head list;
237 struct rb_node node;
238 unsigned char addr[MAX_ADDR_LEN];
239 unsigned char type;
240 #define NETDEV_HW_ADDR_T_LAN 1
241 #define NETDEV_HW_ADDR_T_SAN 2
242 #define NETDEV_HW_ADDR_T_UNICAST 3
243 #define NETDEV_HW_ADDR_T_MULTICAST 4
244 bool global_use;
245 int sync_cnt;
246 int refcount;
247 int synced;
248 struct rcu_head rcu_head;
249 };
250
251 struct netdev_hw_addr_list {
252 struct list_head list;
253 int count;
254
255 /* Auxiliary tree for faster lookup on addition and deletion */
256 struct rb_root tree;
257 };
258
259 #define netdev_hw_addr_list_count(l) ((l)->count)
260 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
261 #define netdev_hw_addr_list_for_each(ha, l) \
262 list_for_each_entry(ha, &(l)->list, list)
263
264 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
265 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
266 #define netdev_for_each_uc_addr(ha, dev) \
267 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
268 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
269 netdev_for_each_uc_addr((_ha), (_dev)) \
270 if ((_ha)->sync_cnt)
271
272 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
273 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
274 #define netdev_for_each_mc_addr(ha, dev) \
275 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
276 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
277 netdev_for_each_mc_addr((_ha), (_dev)) \
278 if ((_ha)->sync_cnt)
279
280 struct hh_cache {
281 unsigned int hh_len;
282 seqlock_t hh_lock;
283
284 /* cached hardware header; allow for machine alignment needs. */
285 #define HH_DATA_MOD 16
286 #define HH_DATA_OFF(__len) \
287 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
288 #define HH_DATA_ALIGN(__len) \
289 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
290 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
291 };
292
293 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
294 * Alternative is:
295 * dev->hard_header_len ? (dev->hard_header_len +
296 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
297 *
298 * We could use other alignment values, but we must maintain the
299 * relationship HH alignment <= LL alignment.
300 */
301 #define LL_RESERVED_SPACE(dev) \
302 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
303 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
304 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
305 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
306 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
307
308 struct header_ops {
309 int (*create) (struct sk_buff *skb, struct net_device *dev,
310 unsigned short type, const void *daddr,
311 const void *saddr, unsigned int len);
312 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
313 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
314 void (*cache_update)(struct hh_cache *hh,
315 const struct net_device *dev,
316 const unsigned char *haddr);
317 bool (*validate)(const char *ll_header, unsigned int len);
318 __be16 (*parse_protocol)(const struct sk_buff *skb);
319
320 ANDROID_KABI_RESERVE(1);
321 ANDROID_KABI_RESERVE(2);
322 };
323
324 /* These flag bits are private to the generic network queueing
325 * layer; they may not be explicitly referenced by any other
326 * code.
327 */
328
329 enum netdev_state_t {
330 __LINK_STATE_START,
331 __LINK_STATE_PRESENT,
332 __LINK_STATE_NOCARRIER,
333 __LINK_STATE_LINKWATCH_PENDING,
334 __LINK_STATE_DORMANT,
335 __LINK_STATE_TESTING,
336 };
337
338 struct gro_list {
339 struct list_head list;
340 int count;
341 };
342
343 /*
344 * size of gro hash buckets, must less than bit number of
345 * napi_struct::gro_bitmask
346 */
347 #define GRO_HASH_BUCKETS 8
348
349 /*
350 * Structure for NAPI scheduling similar to tasklet but with weighting
351 */
352 struct napi_struct {
353 /* The poll_list must only be managed by the entity which
354 * changes the state of the NAPI_STATE_SCHED bit. This means
355 * whoever atomically sets that bit can add this napi_struct
356 * to the per-CPU poll_list, and whoever clears that bit
357 * can remove from the list right before clearing the bit.
358 */
359 struct list_head poll_list;
360
361 unsigned long state;
362 int weight;
363 u32 defer_hard_irqs_count;
364 unsigned long gro_bitmask;
365 int (*poll)(struct napi_struct *, int);
366 #ifdef CONFIG_NETPOLL
367 /* CPU actively polling if netpoll is configured */
368 int poll_owner;
369 #endif
370 /* CPU on which NAPI has been scheduled for processing */
371 int list_owner;
372 struct net_device *dev;
373 struct gro_list gro_hash[GRO_HASH_BUCKETS];
374 struct sk_buff *skb;
375 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
376 int rx_count; /* length of rx_list */
377 unsigned int napi_id;
378 struct hrtimer timer;
379 struct task_struct *thread;
380 /* control-path-only fields follow */
381 struct list_head dev_list;
382 struct hlist_node napi_hash_node;
383 int irq;
384
385 ANDROID_KABI_RESERVE(1);
386 ANDROID_KABI_RESERVE(2);
387 ANDROID_KABI_RESERVE(3);
388 ANDROID_KABI_RESERVE(4);
389 };
390
391 enum {
392 NAPI_STATE_SCHED, /* Poll is scheduled */
393 NAPI_STATE_MISSED, /* reschedule a napi */
394 NAPI_STATE_DISABLE, /* Disable pending */
395 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
396 NAPI_STATE_LISTED, /* NAPI added to system lists */
397 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
398 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
399 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
400 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
401 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
402 };
403
404 enum {
405 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
406 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
407 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
408 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
409 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
410 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
411 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
412 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
413 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
414 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
415 };
416
417 enum gro_result {
418 GRO_MERGED,
419 GRO_MERGED_FREE,
420 GRO_HELD,
421 GRO_NORMAL,
422 GRO_CONSUMED,
423 };
424 typedef enum gro_result gro_result_t;
425
426 /*
427 * enum rx_handler_result - Possible return values for rx_handlers.
428 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
429 * further.
430 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
431 * case skb->dev was changed by rx_handler.
432 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
433 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
434 *
435 * rx_handlers are functions called from inside __netif_receive_skb(), to do
436 * special processing of the skb, prior to delivery to protocol handlers.
437 *
438 * Currently, a net_device can only have a single rx_handler registered. Trying
439 * to register a second rx_handler will return -EBUSY.
440 *
441 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
442 * To unregister a rx_handler on a net_device, use
443 * netdev_rx_handler_unregister().
444 *
445 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
446 * do with the skb.
447 *
448 * If the rx_handler consumed the skb in some way, it should return
449 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
450 * the skb to be delivered in some other way.
451 *
452 * If the rx_handler changed skb->dev, to divert the skb to another
453 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
454 * new device will be called if it exists.
455 *
456 * If the rx_handler decides the skb should be ignored, it should return
457 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
458 * are registered on exact device (ptype->dev == skb->dev).
459 *
460 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
461 * delivered, it should return RX_HANDLER_PASS.
462 *
463 * A device without a registered rx_handler will behave as if rx_handler
464 * returned RX_HANDLER_PASS.
465 */
466
467 enum rx_handler_result {
468 RX_HANDLER_CONSUMED,
469 RX_HANDLER_ANOTHER,
470 RX_HANDLER_EXACT,
471 RX_HANDLER_PASS,
472 };
473 typedef enum rx_handler_result rx_handler_result_t;
474 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
475
476 void __napi_schedule(struct napi_struct *n);
477 void __napi_schedule_irqoff(struct napi_struct *n);
478
napi_disable_pending(struct napi_struct * n)479 static inline bool napi_disable_pending(struct napi_struct *n)
480 {
481 return test_bit(NAPI_STATE_DISABLE, &n->state);
482 }
483
napi_prefer_busy_poll(struct napi_struct * n)484 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
485 {
486 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
487 }
488
489 /**
490 * napi_is_scheduled - test if NAPI is scheduled
491 * @n: NAPI context
492 *
493 * This check is "best-effort". With no locking implemented,
494 * a NAPI can be scheduled or terminate right after this check
495 * and produce not precise results.
496 *
497 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
498 * should not be used normally and napi_schedule should be
499 * used instead.
500 *
501 * Use only if the driver really needs to check if a NAPI
502 * is scheduled for example in the context of delayed timer
503 * that can be skipped if a NAPI is already scheduled.
504 *
505 * Return True if NAPI is scheduled, False otherwise.
506 */
napi_is_scheduled(struct napi_struct * n)507 static inline bool napi_is_scheduled(struct napi_struct *n)
508 {
509 return test_bit(NAPI_STATE_SCHED, &n->state);
510 }
511
512 bool napi_schedule_prep(struct napi_struct *n);
513
514 /**
515 * napi_schedule - schedule NAPI poll
516 * @n: NAPI context
517 *
518 * Schedule NAPI poll routine to be called if it is not already
519 * running.
520 * Return true if we schedule a NAPI or false if not.
521 * Refer to napi_schedule_prep() for additional reason on why
522 * a NAPI might not be scheduled.
523 */
napi_schedule(struct napi_struct * n)524 static inline bool napi_schedule(struct napi_struct *n)
525 {
526 if (napi_schedule_prep(n)) {
527 __napi_schedule(n);
528 return true;
529 }
530
531 return false;
532 }
533
534 /**
535 * napi_schedule_irqoff - schedule NAPI poll
536 * @n: NAPI context
537 *
538 * Variant of napi_schedule(), assuming hard irqs are masked.
539 */
napi_schedule_irqoff(struct napi_struct * n)540 static inline void napi_schedule_irqoff(struct napi_struct *n)
541 {
542 if (napi_schedule_prep(n))
543 __napi_schedule_irqoff(n);
544 }
545
546 /**
547 * napi_complete_done - NAPI processing complete
548 * @n: NAPI context
549 * @work_done: number of packets processed
550 *
551 * Mark NAPI processing as complete. Should only be called if poll budget
552 * has not been completely consumed.
553 * Prefer over napi_complete().
554 * Return false if device should avoid rearming interrupts.
555 */
556 bool napi_complete_done(struct napi_struct *n, int work_done);
557
napi_complete(struct napi_struct * n)558 static inline bool napi_complete(struct napi_struct *n)
559 {
560 return napi_complete_done(n, 0);
561 }
562
563 int dev_set_threaded(struct net_device *dev, bool threaded);
564
565 /**
566 * napi_disable - prevent NAPI from scheduling
567 * @n: NAPI context
568 *
569 * Stop NAPI from being scheduled on this context.
570 * Waits till any outstanding processing completes.
571 */
572 void napi_disable(struct napi_struct *n);
573
574 void napi_enable(struct napi_struct *n);
575
576 /**
577 * napi_synchronize - wait until NAPI is not running
578 * @n: NAPI context
579 *
580 * Wait until NAPI is done being scheduled on this context.
581 * Waits till any outstanding processing completes but
582 * does not disable future activations.
583 */
napi_synchronize(const struct napi_struct * n)584 static inline void napi_synchronize(const struct napi_struct *n)
585 {
586 if (IS_ENABLED(CONFIG_SMP))
587 while (test_bit(NAPI_STATE_SCHED, &n->state))
588 msleep(1);
589 else
590 barrier();
591 }
592
593 /**
594 * napi_if_scheduled_mark_missed - if napi is running, set the
595 * NAPIF_STATE_MISSED
596 * @n: NAPI context
597 *
598 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
599 * NAPI is scheduled.
600 **/
napi_if_scheduled_mark_missed(struct napi_struct * n)601 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
602 {
603 unsigned long val, new;
604
605 val = READ_ONCE(n->state);
606 do {
607 if (val & NAPIF_STATE_DISABLE)
608 return true;
609
610 if (!(val & NAPIF_STATE_SCHED))
611 return false;
612
613 new = val | NAPIF_STATE_MISSED;
614 } while (!try_cmpxchg(&n->state, &val, new));
615
616 return true;
617 }
618
619 enum netdev_queue_state_t {
620 __QUEUE_STATE_DRV_XOFF,
621 __QUEUE_STATE_STACK_XOFF,
622 __QUEUE_STATE_FROZEN,
623 };
624
625 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
626 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
627 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
628
629 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
630 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
631 QUEUE_STATE_FROZEN)
632 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
633 QUEUE_STATE_FROZEN)
634
635 /*
636 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
637 * netif_tx_* functions below are used to manipulate this flag. The
638 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
639 * queue independently. The netif_xmit_*stopped functions below are called
640 * to check if the queue has been stopped by the driver or stack (either
641 * of the XOFF bits are set in the state). Drivers should not need to call
642 * netif_xmit*stopped functions, they should only be using netif_tx_*.
643 */
644
645 struct netdev_queue {
646 /*
647 * read-mostly part
648 */
649 struct net_device *dev;
650 netdevice_tracker dev_tracker;
651
652 struct Qdisc __rcu *qdisc;
653 struct Qdisc __rcu *qdisc_sleeping;
654 #ifdef CONFIG_SYSFS
655 struct kobject kobj;
656 #endif
657 unsigned long tx_maxrate;
658 /*
659 * Number of TX timeouts for this queue
660 * (/sys/class/net/DEV/Q/trans_timeout)
661 */
662 atomic_long_t trans_timeout;
663
664 /* Subordinate device that the queue has been assigned to */
665 struct net_device *sb_dev;
666 #ifdef CONFIG_XDP_SOCKETS
667 struct xsk_buff_pool *pool;
668 #endif
669
670 /*
671 * write-mostly part
672 */
673 #ifdef CONFIG_BQL
674 struct dql dql;
675 #endif
676 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
677 int xmit_lock_owner;
678 /*
679 * Time (in jiffies) of last Tx
680 */
681 unsigned long trans_start;
682
683 unsigned long state;
684
685 /*
686 * slow- / control-path part
687 */
688 /* NAPI instance for the queue
689 * Readers and writers must hold RTNL
690 */
691 struct napi_struct *napi;
692
693 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
694 int numa_node;
695 #endif
696
697 ANDROID_KABI_RESERVE(1);
698 ANDROID_KABI_RESERVE(2);
699 ANDROID_KABI_RESERVE(3);
700 ANDROID_KABI_RESERVE(4);
701 } ____cacheline_aligned_in_smp;
702
703 extern int sysctl_fb_tunnels_only_for_init_net;
704 extern int sysctl_devconf_inherit_init_net;
705
706 /*
707 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
708 * == 1 : For initns only
709 * == 2 : For none.
710 */
net_has_fallback_tunnels(const struct net * net)711 static inline bool net_has_fallback_tunnels(const struct net *net)
712 {
713 #if IS_ENABLED(CONFIG_SYSCTL)
714 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
715
716 return !fb_tunnels_only_for_init_net ||
717 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
718 #else
719 return true;
720 #endif
721 }
722
net_inherit_devconf(void)723 static inline int net_inherit_devconf(void)
724 {
725 #if IS_ENABLED(CONFIG_SYSCTL)
726 return READ_ONCE(sysctl_devconf_inherit_init_net);
727 #else
728 return 0;
729 #endif
730 }
731
netdev_queue_numa_node_read(const struct netdev_queue * q)732 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
733 {
734 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
735 return q->numa_node;
736 #else
737 return NUMA_NO_NODE;
738 #endif
739 }
740
netdev_queue_numa_node_write(struct netdev_queue * q,int node)741 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
742 {
743 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
744 q->numa_node = node;
745 #endif
746 }
747
748 #ifdef CONFIG_RFS_ACCEL
749 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
750 u16 filter_id);
751 #endif
752
753 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
754 enum xps_map_type {
755 XPS_CPUS = 0,
756 XPS_RXQS,
757 XPS_MAPS_MAX,
758 };
759
760 #ifdef CONFIG_XPS
761 /*
762 * This structure holds an XPS map which can be of variable length. The
763 * map is an array of queues.
764 */
765 struct xps_map {
766 unsigned int len;
767 unsigned int alloc_len;
768 struct rcu_head rcu;
769 u16 queues[];
770 };
771 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
772 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
773 - sizeof(struct xps_map)) / sizeof(u16))
774
775 /*
776 * This structure holds all XPS maps for device. Maps are indexed by CPU.
777 *
778 * We keep track of the number of cpus/rxqs used when the struct is allocated,
779 * in nr_ids. This will help not accessing out-of-bound memory.
780 *
781 * We keep track of the number of traffic classes used when the struct is
782 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
783 * not crossing its upper bound, as the original dev->num_tc can be updated in
784 * the meantime.
785 */
786 struct xps_dev_maps {
787 struct rcu_head rcu;
788 unsigned int nr_ids;
789 s16 num_tc;
790 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
791 };
792
793 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
794 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
795
796 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
797 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
798
799 #endif /* CONFIG_XPS */
800
801 #define TC_MAX_QUEUE 16
802 #define TC_BITMASK 15
803 /* HW offloaded queuing disciplines txq count and offset maps */
804 struct netdev_tc_txq {
805 u16 count;
806 u16 offset;
807 };
808
809 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
810 /*
811 * This structure is to hold information about the device
812 * configured to run FCoE protocol stack.
813 */
814 struct netdev_fcoe_hbainfo {
815 char manufacturer[64];
816 char serial_number[64];
817 char hardware_version[64];
818 char driver_version[64];
819 char optionrom_version[64];
820 char firmware_version[64];
821 char model[256];
822 char model_description[256];
823 };
824 #endif
825
826 #define MAX_PHYS_ITEM_ID_LEN 32
827
828 /* This structure holds a unique identifier to identify some
829 * physical item (port for example) used by a netdevice.
830 */
831 struct netdev_phys_item_id {
832 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
833 unsigned char id_len;
834 };
835
netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)836 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
837 struct netdev_phys_item_id *b)
838 {
839 return a->id_len == b->id_len &&
840 memcmp(a->id, b->id, a->id_len) == 0;
841 }
842
843 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
844 struct sk_buff *skb,
845 struct net_device *sb_dev);
846
847 enum net_device_path_type {
848 DEV_PATH_ETHERNET = 0,
849 DEV_PATH_VLAN,
850 DEV_PATH_BRIDGE,
851 DEV_PATH_PPPOE,
852 DEV_PATH_DSA,
853 DEV_PATH_MTK_WDMA,
854 };
855
856 struct net_device_path {
857 enum net_device_path_type type;
858 const struct net_device *dev;
859 union {
860 struct {
861 u16 id;
862 __be16 proto;
863 u8 h_dest[ETH_ALEN];
864 } encap;
865 struct {
866 enum {
867 DEV_PATH_BR_VLAN_KEEP,
868 DEV_PATH_BR_VLAN_TAG,
869 DEV_PATH_BR_VLAN_UNTAG,
870 DEV_PATH_BR_VLAN_UNTAG_HW,
871 } vlan_mode;
872 u16 vlan_id;
873 __be16 vlan_proto;
874 } bridge;
875 struct {
876 int port;
877 u16 proto;
878 } dsa;
879 struct {
880 u8 wdma_idx;
881 u8 queue;
882 u16 wcid;
883 u8 bss;
884 u8 amsdu;
885 } mtk_wdma;
886 };
887 };
888
889 #define NET_DEVICE_PATH_STACK_MAX 5
890 #define NET_DEVICE_PATH_VLAN_MAX 2
891
892 struct net_device_path_stack {
893 int num_paths;
894 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
895 };
896
897 struct net_device_path_ctx {
898 const struct net_device *dev;
899 u8 daddr[ETH_ALEN];
900
901 int num_vlans;
902 struct {
903 u16 id;
904 __be16 proto;
905 } vlan[NET_DEVICE_PATH_VLAN_MAX];
906 };
907
908 enum tc_setup_type {
909 TC_QUERY_CAPS,
910 TC_SETUP_QDISC_MQPRIO,
911 TC_SETUP_CLSU32,
912 TC_SETUP_CLSFLOWER,
913 TC_SETUP_CLSMATCHALL,
914 TC_SETUP_CLSBPF,
915 TC_SETUP_BLOCK,
916 TC_SETUP_QDISC_CBS,
917 TC_SETUP_QDISC_RED,
918 TC_SETUP_QDISC_PRIO,
919 TC_SETUP_QDISC_MQ,
920 TC_SETUP_QDISC_ETF,
921 TC_SETUP_ROOT_QDISC,
922 TC_SETUP_QDISC_GRED,
923 TC_SETUP_QDISC_TAPRIO,
924 TC_SETUP_FT,
925 TC_SETUP_QDISC_ETS,
926 TC_SETUP_QDISC_TBF,
927 TC_SETUP_QDISC_FIFO,
928 TC_SETUP_QDISC_HTB,
929 TC_SETUP_ACT,
930 };
931
932 /* These structures hold the attributes of bpf state that are being passed
933 * to the netdevice through the bpf op.
934 */
935 enum bpf_netdev_command {
936 /* Set or clear a bpf program used in the earliest stages of packet
937 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
938 * is responsible for calling bpf_prog_put on any old progs that are
939 * stored. In case of error, the callee need not release the new prog
940 * reference, but on success it takes ownership and must bpf_prog_put
941 * when it is no longer used.
942 */
943 XDP_SETUP_PROG,
944 XDP_SETUP_PROG_HW,
945 /* BPF program for offload callbacks, invoked at program load time. */
946 BPF_OFFLOAD_MAP_ALLOC,
947 BPF_OFFLOAD_MAP_FREE,
948 XDP_SETUP_XSK_POOL,
949 };
950
951 struct bpf_prog_offload_ops;
952 struct netlink_ext_ack;
953 struct xdp_umem;
954 struct xdp_dev_bulk_queue;
955 struct bpf_xdp_link;
956
957 enum bpf_xdp_mode {
958 XDP_MODE_SKB = 0,
959 XDP_MODE_DRV = 1,
960 XDP_MODE_HW = 2,
961 __MAX_XDP_MODE
962 };
963
964 struct bpf_xdp_entity {
965 struct bpf_prog *prog;
966 struct bpf_xdp_link *link;
967 };
968
969 struct netdev_bpf {
970 enum bpf_netdev_command command;
971 union {
972 /* XDP_SETUP_PROG */
973 struct {
974 u32 flags;
975 struct bpf_prog *prog;
976 struct netlink_ext_ack *extack;
977 };
978 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
979 struct {
980 struct bpf_offloaded_map *offmap;
981 };
982 /* XDP_SETUP_XSK_POOL */
983 struct {
984 struct xsk_buff_pool *pool;
985 u16 queue_id;
986 } xsk;
987 };
988 };
989
990 /* Flags for ndo_xsk_wakeup. */
991 #define XDP_WAKEUP_RX (1 << 0)
992 #define XDP_WAKEUP_TX (1 << 1)
993
994 #ifdef CONFIG_XFRM_OFFLOAD
995 struct xfrmdev_ops {
996 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
997 void (*xdo_dev_state_delete) (struct xfrm_state *x);
998 void (*xdo_dev_state_free) (struct xfrm_state *x);
999 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1000 struct xfrm_state *x);
1001 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1002 void (*xdo_dev_state_update_stats) (struct xfrm_state *x);
1003 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1004 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1005 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1006
1007 ANDROID_KABI_RESERVE(1);
1008 ANDROID_KABI_RESERVE(2);
1009 ANDROID_KABI_RESERVE(3);
1010 ANDROID_KABI_RESERVE(4);
1011 };
1012 #endif
1013
1014 struct dev_ifalias {
1015 struct rcu_head rcuhead;
1016 char ifalias[];
1017 };
1018
1019 struct devlink;
1020 struct tlsdev_ops;
1021
1022 struct netdev_net_notifier {
1023 struct list_head list;
1024 struct notifier_block *nb;
1025 };
1026
1027 /*
1028 * This structure defines the management hooks for network devices.
1029 * The following hooks can be defined; unless noted otherwise, they are
1030 * optional and can be filled with a null pointer.
1031 *
1032 * int (*ndo_init)(struct net_device *dev);
1033 * This function is called once when a network device is registered.
1034 * The network device can use this for any late stage initialization
1035 * or semantic validation. It can fail with an error code which will
1036 * be propagated back to register_netdev.
1037 *
1038 * void (*ndo_uninit)(struct net_device *dev);
1039 * This function is called when device is unregistered or when registration
1040 * fails. It is not called if init fails.
1041 *
1042 * int (*ndo_open)(struct net_device *dev);
1043 * This function is called when a network device transitions to the up
1044 * state.
1045 *
1046 * int (*ndo_stop)(struct net_device *dev);
1047 * This function is called when a network device transitions to the down
1048 * state.
1049 *
1050 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1051 * struct net_device *dev);
1052 * Called when a packet needs to be transmitted.
1053 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1054 * the queue before that can happen; it's for obsolete devices and weird
1055 * corner cases, but the stack really does a non-trivial amount
1056 * of useless work if you return NETDEV_TX_BUSY.
1057 * Required; cannot be NULL.
1058 *
1059 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1060 * struct net_device *dev
1061 * netdev_features_t features);
1062 * Called by core transmit path to determine if device is capable of
1063 * performing offload operations on a given packet. This is to give
1064 * the device an opportunity to implement any restrictions that cannot
1065 * be otherwise expressed by feature flags. The check is called with
1066 * the set of features that the stack has calculated and it returns
1067 * those the driver believes to be appropriate.
1068 *
1069 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1070 * struct net_device *sb_dev);
1071 * Called to decide which queue to use when device supports multiple
1072 * transmit queues.
1073 *
1074 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1075 * This function is called to allow device receiver to make
1076 * changes to configuration when multicast or promiscuous is enabled.
1077 *
1078 * void (*ndo_set_rx_mode)(struct net_device *dev);
1079 * This function is called device changes address list filtering.
1080 * If driver handles unicast address filtering, it should set
1081 * IFF_UNICAST_FLT in its priv_flags.
1082 *
1083 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1084 * This function is called when the Media Access Control address
1085 * needs to be changed. If this interface is not defined, the
1086 * MAC address can not be changed.
1087 *
1088 * int (*ndo_validate_addr)(struct net_device *dev);
1089 * Test if Media Access Control address is valid for the device.
1090 *
1091 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1092 * Old-style ioctl entry point. This is used internally by the
1093 * appletalk and ieee802154 subsystems but is no longer called by
1094 * the device ioctl handler.
1095 *
1096 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1097 * Used by the bonding driver for its device specific ioctls:
1098 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1099 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1100 *
1101 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1102 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1103 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1104 *
1105 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1106 * Used to set network devices bus interface parameters. This interface
1107 * is retained for legacy reasons; new devices should use the bus
1108 * interface (PCI) for low level management.
1109 *
1110 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1111 * Called when a user wants to change the Maximum Transfer Unit
1112 * of a device.
1113 *
1114 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1115 * Callback used when the transmitter has not made any progress
1116 * for dev->watchdog ticks.
1117 *
1118 * void (*ndo_get_stats64)(struct net_device *dev,
1119 * struct rtnl_link_stats64 *storage);
1120 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1121 * Called when a user wants to get the network device usage
1122 * statistics. Drivers must do one of the following:
1123 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1124 * rtnl_link_stats64 structure passed by the caller.
1125 * 2. Define @ndo_get_stats to update a net_device_stats structure
1126 * (which should normally be dev->stats) and return a pointer to
1127 * it. The structure may be changed asynchronously only if each
1128 * field is written atomically.
1129 * 3. Update dev->stats asynchronously and atomically, and define
1130 * neither operation.
1131 *
1132 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1133 * Return true if this device supports offload stats of this attr_id.
1134 *
1135 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1136 * void *attr_data)
1137 * Get statistics for offload operations by attr_id. Write it into the
1138 * attr_data pointer.
1139 *
1140 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1141 * If device supports VLAN filtering this function is called when a
1142 * VLAN id is registered.
1143 *
1144 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1145 * If device supports VLAN filtering this function is called when a
1146 * VLAN id is unregistered.
1147 *
1148 * void (*ndo_poll_controller)(struct net_device *dev);
1149 *
1150 * SR-IOV management functions.
1151 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1152 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1153 * u8 qos, __be16 proto);
1154 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1155 * int max_tx_rate);
1156 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1157 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1158 * int (*ndo_get_vf_config)(struct net_device *dev,
1159 * int vf, struct ifla_vf_info *ivf);
1160 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1161 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1162 * struct nlattr *port[]);
1163 *
1164 * Enable or disable the VF ability to query its RSS Redirection Table and
1165 * Hash Key. This is needed since on some devices VF share this information
1166 * with PF and querying it may introduce a theoretical security risk.
1167 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1168 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1169 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1170 * void *type_data);
1171 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1172 * This is always called from the stack with the rtnl lock held and netif
1173 * tx queues stopped. This allows the netdevice to perform queue
1174 * management safely.
1175 *
1176 * Fiber Channel over Ethernet (FCoE) offload functions.
1177 * int (*ndo_fcoe_enable)(struct net_device *dev);
1178 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1179 * so the underlying device can perform whatever needed configuration or
1180 * initialization to support acceleration of FCoE traffic.
1181 *
1182 * int (*ndo_fcoe_disable)(struct net_device *dev);
1183 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1184 * so the underlying device can perform whatever needed clean-ups to
1185 * stop supporting acceleration of FCoE traffic.
1186 *
1187 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1188 * struct scatterlist *sgl, unsigned int sgc);
1189 * Called when the FCoE Initiator wants to initialize an I/O that
1190 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1191 * perform necessary setup and returns 1 to indicate the device is set up
1192 * successfully to perform DDP on this I/O, otherwise this returns 0.
1193 *
1194 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1195 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1196 * indicated by the FC exchange id 'xid', so the underlying device can
1197 * clean up and reuse resources for later DDP requests.
1198 *
1199 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1200 * struct scatterlist *sgl, unsigned int sgc);
1201 * Called when the FCoE Target wants to initialize an I/O that
1202 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1203 * perform necessary setup and returns 1 to indicate the device is set up
1204 * successfully to perform DDP on this I/O, otherwise this returns 0.
1205 *
1206 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1207 * struct netdev_fcoe_hbainfo *hbainfo);
1208 * Called when the FCoE Protocol stack wants information on the underlying
1209 * device. This information is utilized by the FCoE protocol stack to
1210 * register attributes with Fiber Channel management service as per the
1211 * FC-GS Fabric Device Management Information(FDMI) specification.
1212 *
1213 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1214 * Called when the underlying device wants to override default World Wide
1215 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1216 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1217 * protocol stack to use.
1218 *
1219 * RFS acceleration.
1220 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1221 * u16 rxq_index, u32 flow_id);
1222 * Set hardware filter for RFS. rxq_index is the target queue index;
1223 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1224 * Return the filter ID on success, or a negative error code.
1225 *
1226 * Slave management functions (for bridge, bonding, etc).
1227 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1228 * Called to make another netdev an underling.
1229 *
1230 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1231 * Called to release previously enslaved netdev.
1232 *
1233 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1234 * struct sk_buff *skb,
1235 * bool all_slaves);
1236 * Get the xmit slave of master device. If all_slaves is true, function
1237 * assume all the slaves can transmit.
1238 *
1239 * Feature/offload setting functions.
1240 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1241 * netdev_features_t features);
1242 * Adjusts the requested feature flags according to device-specific
1243 * constraints, and returns the resulting flags. Must not modify
1244 * the device state.
1245 *
1246 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1247 * Called to update device configuration to new features. Passed
1248 * feature set might be less than what was returned by ndo_fix_features()).
1249 * Must return >0 or -errno if it changed dev->features itself.
1250 *
1251 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1252 * struct net_device *dev,
1253 * const unsigned char *addr, u16 vid, u16 flags,
1254 * struct netlink_ext_ack *extack);
1255 * Adds an FDB entry to dev for addr.
1256 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1257 * struct net_device *dev,
1258 * const unsigned char *addr, u16 vid)
1259 * Deletes the FDB entry from dev corresponding to addr.
1260 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1261 * struct netlink_ext_ack *extack);
1262 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1263 * struct net_device *dev, struct net_device *filter_dev,
1264 * int *idx)
1265 * Used to add FDB entries to dump requests. Implementers should add
1266 * entries to skb and update idx with the number of entries.
1267 *
1268 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1269 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1270 * Adds an MDB entry to dev.
1271 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1272 * struct netlink_ext_ack *extack);
1273 * Deletes the MDB entry from dev.
1274 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1275 * struct netlink_ext_ack *extack);
1276 * Bulk deletes MDB entries from dev.
1277 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1278 * struct netlink_callback *cb);
1279 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1280 * callback is used by core rtnetlink code.
1281 *
1282 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1283 * u16 flags, struct netlink_ext_ack *extack)
1284 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1285 * struct net_device *dev, u32 filter_mask,
1286 * int nlflags)
1287 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1288 * u16 flags);
1289 *
1290 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1291 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1292 * which do not represent real hardware may define this to allow their
1293 * userspace components to manage their virtual carrier state. Devices
1294 * that determine carrier state from physical hardware properties (eg
1295 * network cables) or protocol-dependent mechanisms (eg
1296 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1297 *
1298 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1299 * struct netdev_phys_item_id *ppid);
1300 * Called to get ID of physical port of this device. If driver does
1301 * not implement this, it is assumed that the hw is not able to have
1302 * multiple net devices on single physical port.
1303 *
1304 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1305 * struct netdev_phys_item_id *ppid)
1306 * Called to get the parent ID of the physical port of this device.
1307 *
1308 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1309 * struct net_device *dev)
1310 * Called by upper layer devices to accelerate switching or other
1311 * station functionality into hardware. 'pdev is the lowerdev
1312 * to use for the offload and 'dev' is the net device that will
1313 * back the offload. Returns a pointer to the private structure
1314 * the upper layer will maintain.
1315 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1316 * Called by upper layer device to delete the station created
1317 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1318 * the station and priv is the structure returned by the add
1319 * operation.
1320 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1321 * int queue_index, u32 maxrate);
1322 * Called when a user wants to set a max-rate limitation of specific
1323 * TX queue.
1324 * int (*ndo_get_iflink)(const struct net_device *dev);
1325 * Called to get the iflink value of this device.
1326 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1327 * This function is used to get egress tunnel information for given skb.
1328 * This is useful for retrieving outer tunnel header parameters while
1329 * sampling packet.
1330 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1331 * This function is used to specify the headroom that the skb must
1332 * consider when allocation skb during packet reception. Setting
1333 * appropriate rx headroom value allows avoiding skb head copy on
1334 * forward. Setting a negative value resets the rx headroom to the
1335 * default value.
1336 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1337 * This function is used to set or query state related to XDP on the
1338 * netdevice and manage BPF offload. See definition of
1339 * enum bpf_netdev_command for details.
1340 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1341 * u32 flags);
1342 * This function is used to submit @n XDP packets for transmit on a
1343 * netdevice. Returns number of frames successfully transmitted, frames
1344 * that got dropped are freed/returned via xdp_return_frame().
1345 * Returns negative number, means general error invoking ndo, meaning
1346 * no frames were xmit'ed and core-caller will free all frames.
1347 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1348 * struct xdp_buff *xdp);
1349 * Get the xmit slave of master device based on the xdp_buff.
1350 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1351 * This function is used to wake up the softirq, ksoftirqd or kthread
1352 * responsible for sending and/or receiving packets on a specific
1353 * queue id bound to an AF_XDP socket. The flags field specifies if
1354 * only RX, only Tx, or both should be woken up using the flags
1355 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1356 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1357 * int cmd);
1358 * Add, change, delete or get information on an IPv4 tunnel.
1359 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1360 * If a device is paired with a peer device, return the peer instance.
1361 * The caller must be under RCU read context.
1362 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1363 * Get the forwarding path to reach the real device from the HW destination address
1364 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1365 * const struct skb_shared_hwtstamps *hwtstamps,
1366 * bool cycles);
1367 * Get hardware timestamp based on normal/adjustable time or free running
1368 * cycle counter. This function is required if physical clock supports a
1369 * free running cycle counter.
1370 *
1371 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1372 * struct kernel_hwtstamp_config *kernel_config);
1373 * Get the currently configured hardware timestamping parameters for the
1374 * NIC device.
1375 *
1376 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1377 * struct kernel_hwtstamp_config *kernel_config,
1378 * struct netlink_ext_ack *extack);
1379 * Change the hardware timestamping parameters for NIC device.
1380 */
1381 struct net_device_ops {
1382 int (*ndo_init)(struct net_device *dev);
1383 void (*ndo_uninit)(struct net_device *dev);
1384 int (*ndo_open)(struct net_device *dev);
1385 int (*ndo_stop)(struct net_device *dev);
1386 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1387 struct net_device *dev);
1388 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1389 struct net_device *dev,
1390 netdev_features_t features);
1391 u16 (*ndo_select_queue)(struct net_device *dev,
1392 struct sk_buff *skb,
1393 struct net_device *sb_dev);
1394 void (*ndo_change_rx_flags)(struct net_device *dev,
1395 int flags);
1396 void (*ndo_set_rx_mode)(struct net_device *dev);
1397 int (*ndo_set_mac_address)(struct net_device *dev,
1398 void *addr);
1399 int (*ndo_validate_addr)(struct net_device *dev);
1400 int (*ndo_do_ioctl)(struct net_device *dev,
1401 struct ifreq *ifr, int cmd);
1402 int (*ndo_eth_ioctl)(struct net_device *dev,
1403 struct ifreq *ifr, int cmd);
1404 int (*ndo_siocbond)(struct net_device *dev,
1405 struct ifreq *ifr, int cmd);
1406 int (*ndo_siocwandev)(struct net_device *dev,
1407 struct if_settings *ifs);
1408 int (*ndo_siocdevprivate)(struct net_device *dev,
1409 struct ifreq *ifr,
1410 void __user *data, int cmd);
1411 int (*ndo_set_config)(struct net_device *dev,
1412 struct ifmap *map);
1413 int (*ndo_change_mtu)(struct net_device *dev,
1414 int new_mtu);
1415 int (*ndo_neigh_setup)(struct net_device *dev,
1416 struct neigh_parms *);
1417 void (*ndo_tx_timeout) (struct net_device *dev,
1418 unsigned int txqueue);
1419
1420 void (*ndo_get_stats64)(struct net_device *dev,
1421 struct rtnl_link_stats64 *storage);
1422 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1423 int (*ndo_get_offload_stats)(int attr_id,
1424 const struct net_device *dev,
1425 void *attr_data);
1426 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1427
1428 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1429 __be16 proto, u16 vid);
1430 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1431 __be16 proto, u16 vid);
1432 #ifdef CONFIG_NET_POLL_CONTROLLER
1433 void (*ndo_poll_controller)(struct net_device *dev);
1434 int (*ndo_netpoll_setup)(struct net_device *dev,
1435 struct netpoll_info *info);
1436 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1437 #endif
1438 int (*ndo_set_vf_mac)(struct net_device *dev,
1439 int queue, u8 *mac);
1440 int (*ndo_set_vf_vlan)(struct net_device *dev,
1441 int queue, u16 vlan,
1442 u8 qos, __be16 proto);
1443 int (*ndo_set_vf_rate)(struct net_device *dev,
1444 int vf, int min_tx_rate,
1445 int max_tx_rate);
1446 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1447 int vf, bool setting);
1448 int (*ndo_set_vf_trust)(struct net_device *dev,
1449 int vf, bool setting);
1450 int (*ndo_get_vf_config)(struct net_device *dev,
1451 int vf,
1452 struct ifla_vf_info *ivf);
1453 int (*ndo_set_vf_link_state)(struct net_device *dev,
1454 int vf, int link_state);
1455 int (*ndo_get_vf_stats)(struct net_device *dev,
1456 int vf,
1457 struct ifla_vf_stats
1458 *vf_stats);
1459 int (*ndo_set_vf_port)(struct net_device *dev,
1460 int vf,
1461 struct nlattr *port[]);
1462 int (*ndo_get_vf_port)(struct net_device *dev,
1463 int vf, struct sk_buff *skb);
1464 int (*ndo_get_vf_guid)(struct net_device *dev,
1465 int vf,
1466 struct ifla_vf_guid *node_guid,
1467 struct ifla_vf_guid *port_guid);
1468 int (*ndo_set_vf_guid)(struct net_device *dev,
1469 int vf, u64 guid,
1470 int guid_type);
1471 int (*ndo_set_vf_rss_query_en)(
1472 struct net_device *dev,
1473 int vf, bool setting);
1474 int (*ndo_setup_tc)(struct net_device *dev,
1475 enum tc_setup_type type,
1476 void *type_data);
1477 #if IS_ENABLED(CONFIG_FCOE)
1478 int (*ndo_fcoe_enable)(struct net_device *dev);
1479 int (*ndo_fcoe_disable)(struct net_device *dev);
1480 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1481 u16 xid,
1482 struct scatterlist *sgl,
1483 unsigned int sgc);
1484 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1485 u16 xid);
1486 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1487 u16 xid,
1488 struct scatterlist *sgl,
1489 unsigned int sgc);
1490 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1491 struct netdev_fcoe_hbainfo *hbainfo);
1492 #endif
1493
1494 #if IS_ENABLED(CONFIG_LIBFCOE)
1495 #define NETDEV_FCOE_WWNN 0
1496 #define NETDEV_FCOE_WWPN 1
1497 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1498 u64 *wwn, int type);
1499 #endif
1500
1501 #ifdef CONFIG_RFS_ACCEL
1502 int (*ndo_rx_flow_steer)(struct net_device *dev,
1503 const struct sk_buff *skb,
1504 u16 rxq_index,
1505 u32 flow_id);
1506 #endif
1507 int (*ndo_add_slave)(struct net_device *dev,
1508 struct net_device *slave_dev,
1509 struct netlink_ext_ack *extack);
1510 int (*ndo_del_slave)(struct net_device *dev,
1511 struct net_device *slave_dev);
1512 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1513 struct sk_buff *skb,
1514 bool all_slaves);
1515 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1516 struct sock *sk);
1517 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1518 netdev_features_t features);
1519 int (*ndo_set_features)(struct net_device *dev,
1520 netdev_features_t features);
1521 int (*ndo_neigh_construct)(struct net_device *dev,
1522 struct neighbour *n);
1523 void (*ndo_neigh_destroy)(struct net_device *dev,
1524 struct neighbour *n);
1525
1526 int (*ndo_fdb_add)(struct ndmsg *ndm,
1527 struct nlattr *tb[],
1528 struct net_device *dev,
1529 const unsigned char *addr,
1530 u16 vid,
1531 u16 flags,
1532 struct netlink_ext_ack *extack);
1533 int (*ndo_fdb_del)(struct ndmsg *ndm,
1534 struct nlattr *tb[],
1535 struct net_device *dev,
1536 const unsigned char *addr,
1537 u16 vid, struct netlink_ext_ack *extack);
1538 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1539 struct net_device *dev,
1540 struct netlink_ext_ack *extack);
1541 int (*ndo_fdb_dump)(struct sk_buff *skb,
1542 struct netlink_callback *cb,
1543 struct net_device *dev,
1544 struct net_device *filter_dev,
1545 int *idx);
1546 int (*ndo_fdb_get)(struct sk_buff *skb,
1547 struct nlattr *tb[],
1548 struct net_device *dev,
1549 const unsigned char *addr,
1550 u16 vid, u32 portid, u32 seq,
1551 struct netlink_ext_ack *extack);
1552 int (*ndo_mdb_add)(struct net_device *dev,
1553 struct nlattr *tb[],
1554 u16 nlmsg_flags,
1555 struct netlink_ext_ack *extack);
1556 int (*ndo_mdb_del)(struct net_device *dev,
1557 struct nlattr *tb[],
1558 struct netlink_ext_ack *extack);
1559 int (*ndo_mdb_del_bulk)(struct net_device *dev,
1560 struct nlattr *tb[],
1561 struct netlink_ext_ack *extack);
1562 int (*ndo_mdb_dump)(struct net_device *dev,
1563 struct sk_buff *skb,
1564 struct netlink_callback *cb);
1565 int (*ndo_mdb_get)(struct net_device *dev,
1566 struct nlattr *tb[], u32 portid,
1567 u32 seq,
1568 struct netlink_ext_ack *extack);
1569 int (*ndo_bridge_setlink)(struct net_device *dev,
1570 struct nlmsghdr *nlh,
1571 u16 flags,
1572 struct netlink_ext_ack *extack);
1573 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1574 u32 pid, u32 seq,
1575 struct net_device *dev,
1576 u32 filter_mask,
1577 int nlflags);
1578 int (*ndo_bridge_dellink)(struct net_device *dev,
1579 struct nlmsghdr *nlh,
1580 u16 flags);
1581 int (*ndo_change_carrier)(struct net_device *dev,
1582 bool new_carrier);
1583 int (*ndo_get_phys_port_id)(struct net_device *dev,
1584 struct netdev_phys_item_id *ppid);
1585 int (*ndo_get_port_parent_id)(struct net_device *dev,
1586 struct netdev_phys_item_id *ppid);
1587 int (*ndo_get_phys_port_name)(struct net_device *dev,
1588 char *name, size_t len);
1589 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1590 struct net_device *dev);
1591 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1592 void *priv);
1593
1594 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1595 int queue_index,
1596 u32 maxrate);
1597 int (*ndo_get_iflink)(const struct net_device *dev);
1598 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1599 struct sk_buff *skb);
1600 void (*ndo_set_rx_headroom)(struct net_device *dev,
1601 int needed_headroom);
1602 int (*ndo_bpf)(struct net_device *dev,
1603 struct netdev_bpf *bpf);
1604 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1605 struct xdp_frame **xdp,
1606 u32 flags);
1607 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1608 struct xdp_buff *xdp);
1609 int (*ndo_xsk_wakeup)(struct net_device *dev,
1610 u32 queue_id, u32 flags);
1611 int (*ndo_tunnel_ctl)(struct net_device *dev,
1612 struct ip_tunnel_parm_kern *p,
1613 int cmd);
1614 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1615 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1616 struct net_device_path *path);
1617 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1618 const struct skb_shared_hwtstamps *hwtstamps,
1619 bool cycles);
1620 int (*ndo_hwtstamp_get)(struct net_device *dev,
1621 struct kernel_hwtstamp_config *kernel_config);
1622 int (*ndo_hwtstamp_set)(struct net_device *dev,
1623 struct kernel_hwtstamp_config *kernel_config,
1624 struct netlink_ext_ack *extack);
1625 ANDROID_KABI_RESERVE(1);
1626 ANDROID_KABI_RESERVE(2);
1627 ANDROID_KABI_RESERVE(3);
1628 ANDROID_KABI_RESERVE(4);
1629 ANDROID_KABI_RESERVE(5);
1630 ANDROID_KABI_RESERVE(6);
1631 ANDROID_KABI_RESERVE(7);
1632 ANDROID_KABI_RESERVE(8);
1633 };
1634
1635 /**
1636 * enum netdev_priv_flags - &struct net_device priv_flags
1637 *
1638 * These are the &struct net_device, they are only set internally
1639 * by drivers and used in the kernel. These flags are invisible to
1640 * userspace; this means that the order of these flags can change
1641 * during any kernel release.
1642 *
1643 * You should add bitfield booleans after either net_device::priv_flags
1644 * (hotpath) or ::threaded (slowpath) instead of extending these flags.
1645 *
1646 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1647 * @IFF_EBRIDGE: Ethernet bridging device
1648 * @IFF_BONDING: bonding master or slave
1649 * @IFF_ISATAP: ISATAP interface (RFC4214)
1650 * @IFF_WAN_HDLC: WAN HDLC device
1651 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1652 * release skb->dst
1653 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1654 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1655 * @IFF_MACVLAN_PORT: device used as macvlan port
1656 * @IFF_BRIDGE_PORT: device used as bridge port
1657 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1658 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1659 * @IFF_UNICAST_FLT: Supports unicast filtering
1660 * @IFF_TEAM_PORT: device used as team port
1661 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1662 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1663 * change when it's running
1664 * @IFF_MACVLAN: Macvlan device
1665 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1666 * underlying stacked devices
1667 * @IFF_L3MDEV_MASTER: device is an L3 master device
1668 * @IFF_NO_QUEUE: device can run without qdisc attached
1669 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1670 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1671 * @IFF_TEAM: device is a team device
1672 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1673 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1674 * entity (i.e. the master device for bridged veth)
1675 * @IFF_MACSEC: device is a MACsec device
1676 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1677 * @IFF_FAILOVER: device is a failover master device
1678 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1679 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1680 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1681 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1682 * skb_headlen(skb) == 0 (data starts from frag0)
1683 */
1684 enum netdev_priv_flags {
1685 IFF_802_1Q_VLAN = 1<<0,
1686 IFF_EBRIDGE = 1<<1,
1687 IFF_BONDING = 1<<2,
1688 IFF_ISATAP = 1<<3,
1689 IFF_WAN_HDLC = 1<<4,
1690 IFF_XMIT_DST_RELEASE = 1<<5,
1691 IFF_DONT_BRIDGE = 1<<6,
1692 IFF_DISABLE_NETPOLL = 1<<7,
1693 IFF_MACVLAN_PORT = 1<<8,
1694 IFF_BRIDGE_PORT = 1<<9,
1695 IFF_OVS_DATAPATH = 1<<10,
1696 IFF_TX_SKB_SHARING = 1<<11,
1697 IFF_UNICAST_FLT = 1<<12,
1698 IFF_TEAM_PORT = 1<<13,
1699 IFF_SUPP_NOFCS = 1<<14,
1700 IFF_LIVE_ADDR_CHANGE = 1<<15,
1701 IFF_MACVLAN = 1<<16,
1702 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1703 IFF_L3MDEV_MASTER = 1<<18,
1704 IFF_NO_QUEUE = 1<<19,
1705 IFF_OPENVSWITCH = 1<<20,
1706 IFF_L3MDEV_SLAVE = 1<<21,
1707 IFF_TEAM = 1<<22,
1708 IFF_RXFH_CONFIGURED = 1<<23,
1709 IFF_PHONY_HEADROOM = 1<<24,
1710 IFF_MACSEC = 1<<25,
1711 IFF_NO_RX_HANDLER = 1<<26,
1712 IFF_FAILOVER = 1<<27,
1713 IFF_FAILOVER_SLAVE = 1<<28,
1714 IFF_L3MDEV_RX_HANDLER = 1<<29,
1715 IFF_NO_ADDRCONF = BIT_ULL(30),
1716 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1717 };
1718
1719 /* Specifies the type of the struct net_device::ml_priv pointer */
1720 enum netdev_ml_priv_type {
1721 ML_PRIV_NONE,
1722 ML_PRIV_CAN,
1723 };
1724
1725 enum netdev_stat_type {
1726 NETDEV_PCPU_STAT_NONE,
1727 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1728 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1729 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1730 };
1731
1732 enum netdev_reg_state {
1733 NETREG_UNINITIALIZED = 0,
1734 NETREG_REGISTERED, /* completed register_netdevice */
1735 NETREG_UNREGISTERING, /* called unregister_netdevice */
1736 NETREG_UNREGISTERED, /* completed unregister todo */
1737 NETREG_RELEASED, /* called free_netdev */
1738 NETREG_DUMMY, /* dummy device for NAPI poll */
1739 };
1740
1741 /**
1742 * struct net_device - The DEVICE structure.
1743 *
1744 * Actually, this whole structure is a big mistake. It mixes I/O
1745 * data with strictly "high-level" data, and it has to know about
1746 * almost every data structure used in the INET module.
1747 *
1748 * @priv_flags: flags invisible to userspace defined as bits, see
1749 * enum netdev_priv_flags for the definitions
1750 * @lltx: device supports lockless Tx. Deprecated for real HW
1751 * drivers. Mainly used by logical interfaces, such as
1752 * bonding and tunnels
1753 *
1754 * @name: This is the first field of the "visible" part of this structure
1755 * (i.e. as seen by users in the "Space.c" file). It is the name
1756 * of the interface.
1757 *
1758 * @name_node: Name hashlist node
1759 * @ifalias: SNMP alias
1760 * @mem_end: Shared memory end
1761 * @mem_start: Shared memory start
1762 * @base_addr: Device I/O address
1763 * @irq: Device IRQ number
1764 *
1765 * @state: Generic network queuing layer state, see netdev_state_t
1766 * @dev_list: The global list of network devices
1767 * @napi_list: List entry used for polling NAPI devices
1768 * @unreg_list: List entry when we are unregistering the
1769 * device; see the function unregister_netdev
1770 * @close_list: List entry used when we are closing the device
1771 * @ptype_all: Device-specific packet handlers for all protocols
1772 * @ptype_specific: Device-specific, protocol-specific packet handlers
1773 *
1774 * @adj_list: Directly linked devices, like slaves for bonding
1775 * @features: Currently active device features
1776 * @hw_features: User-changeable features
1777 *
1778 * @wanted_features: User-requested features
1779 * @vlan_features: Mask of features inheritable by VLAN devices
1780 *
1781 * @hw_enc_features: Mask of features inherited by encapsulating devices
1782 * This field indicates what encapsulation
1783 * offloads the hardware is capable of doing,
1784 * and drivers will need to set them appropriately.
1785 *
1786 * @mpls_features: Mask of features inheritable by MPLS
1787 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1788 *
1789 * @ifindex: interface index
1790 * @group: The group the device belongs to
1791 *
1792 * @stats: Statistics struct, which was left as a legacy, use
1793 * rtnl_link_stats64 instead
1794 *
1795 * @core_stats: core networking counters,
1796 * do not use this in drivers
1797 * @carrier_up_count: Number of times the carrier has been up
1798 * @carrier_down_count: Number of times the carrier has been down
1799 *
1800 * @wireless_handlers: List of functions to handle Wireless Extensions,
1801 * instead of ioctl,
1802 * see <net/iw_handler.h> for details.
1803 * @wireless_data: Instance data managed by the core of wireless extensions
1804 *
1805 * @netdev_ops: Includes several pointers to callbacks,
1806 * if one wants to override the ndo_*() functions
1807 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1808 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks.
1809 * @ethtool_ops: Management operations
1810 * @l3mdev_ops: Layer 3 master device operations
1811 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1812 * discovery handling. Necessary for e.g. 6LoWPAN.
1813 * @xfrmdev_ops: Transformation offload operations
1814 * @tlsdev_ops: Transport Layer Security offload operations
1815 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1816 * of Layer 2 headers.
1817 *
1818 * @flags: Interface flags (a la BSD)
1819 * @xdp_features: XDP capability supported by the device
1820 * @gflags: Global flags ( kept as legacy )
1821 * @priv_len: Size of the ->priv flexible array
1822 * @priv: Flexible array containing private data
1823 * @operstate: RFC2863 operstate
1824 * @link_mode: Mapping policy to operstate
1825 * @if_port: Selectable AUI, TP, ...
1826 * @dma: DMA channel
1827 * @mtu: Interface MTU value
1828 * @min_mtu: Interface Minimum MTU value
1829 * @max_mtu: Interface Maximum MTU value
1830 * @type: Interface hardware type
1831 * @hard_header_len: Maximum hardware header length.
1832 * @min_header_len: Minimum hardware header length
1833 *
1834 * @needed_headroom: Extra headroom the hardware may need, but not in all
1835 * cases can this be guaranteed
1836 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1837 * cases can this be guaranteed. Some cases also use
1838 * LL_MAX_HEADER instead to allocate the skb
1839 *
1840 * interface address info:
1841 *
1842 * @perm_addr: Permanent hw address
1843 * @addr_assign_type: Hw address assignment type
1844 * @addr_len: Hardware address length
1845 * @upper_level: Maximum depth level of upper devices.
1846 * @lower_level: Maximum depth level of lower devices.
1847 * @neigh_priv_len: Used in neigh_alloc()
1848 * @dev_id: Used to differentiate devices that share
1849 * the same link layer address
1850 * @dev_port: Used to differentiate devices that share
1851 * the same function
1852 * @addr_list_lock: XXX: need comments on this one
1853 * @name_assign_type: network interface name assignment type
1854 * @uc_promisc: Counter that indicates promiscuous mode
1855 * has been enabled due to the need to listen to
1856 * additional unicast addresses in a device that
1857 * does not implement ndo_set_rx_mode()
1858 * @uc: unicast mac addresses
1859 * @mc: multicast mac addresses
1860 * @dev_addrs: list of device hw addresses
1861 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1862 * @promiscuity: Number of times the NIC is told to work in
1863 * promiscuous mode; if it becomes 0 the NIC will
1864 * exit promiscuous mode
1865 * @allmulti: Counter, enables or disables allmulticast mode
1866 *
1867 * @vlan_info: VLAN info
1868 * @dsa_ptr: dsa specific data
1869 * @tipc_ptr: TIPC specific data
1870 * @atalk_ptr: AppleTalk link
1871 * @ip_ptr: IPv4 specific data
1872 * @ip6_ptr: IPv6 specific data
1873 * @ax25_ptr: AX.25 specific data
1874 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1875 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1876 * device struct
1877 * @mpls_ptr: mpls_dev struct pointer
1878 * @mctp_ptr: MCTP specific data
1879 *
1880 * @dev_addr: Hw address (before bcast,
1881 * because most packets are unicast)
1882 *
1883 * @_rx: Array of RX queues
1884 * @num_rx_queues: Number of RX queues
1885 * allocated at register_netdev() time
1886 * @real_num_rx_queues: Number of RX queues currently active in device
1887 * @xdp_prog: XDP sockets filter program pointer
1888 * @gro_flush_timeout: timeout for GRO layer in NAPI
1889 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1890 * allow to avoid NIC hard IRQ, on busy queues.
1891 *
1892 * @rx_handler: handler for received packets
1893 * @rx_handler_data: XXX: need comments on this one
1894 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1895 * @ingress_queue: XXX: need comments on this one
1896 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1897 * @broadcast: hw bcast address
1898 *
1899 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1900 * indexed by RX queue number. Assigned by driver.
1901 * This must only be set if the ndo_rx_flow_steer
1902 * operation is defined
1903 * @index_hlist: Device index hash chain
1904 *
1905 * @_tx: Array of TX queues
1906 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1907 * @real_num_tx_queues: Number of TX queues currently active in device
1908 * @qdisc: Root qdisc from userspace point of view
1909 * @tx_queue_len: Max frames per queue allowed
1910 * @tx_global_lock: XXX: need comments on this one
1911 * @xdp_bulkq: XDP device bulk queue
1912 * @xps_maps: all CPUs/RXQs maps for XPS device
1913 *
1914 * @xps_maps: XXX: need comments on this one
1915 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1916 * @nf_hooks_egress: netfilter hooks executed for egress packets
1917 * @qdisc_hash: qdisc hash table
1918 * @watchdog_timeo: Represents the timeout that is used by
1919 * the watchdog (see dev_watchdog())
1920 * @watchdog_timer: List of timers
1921 *
1922 * @proto_down_reason: reason a netdev interface is held down
1923 * @pcpu_refcnt: Number of references to this device
1924 * @dev_refcnt: Number of references to this device
1925 * @refcnt_tracker: Tracker directory for tracked references to this device
1926 * @todo_list: Delayed register/unregister
1927 * @link_watch_list: XXX: need comments on this one
1928 *
1929 * @reg_state: Register/unregister state machine
1930 * @dismantle: Device is going to be freed
1931 * @rtnl_link_state: This enum represents the phases of creating
1932 * a new link
1933 *
1934 * @needs_free_netdev: Should unregister perform free_netdev?
1935 * @priv_destructor: Called from unregister
1936 * @npinfo: XXX: need comments on this one
1937 * @nd_net: Network namespace this network device is inside
1938 *
1939 * @ml_priv: Mid-layer private
1940 * @ml_priv_type: Mid-layer private type
1941 *
1942 * @pcpu_stat_type: Type of device statistics which the core should
1943 * allocate/free: none, lstats, tstats, dstats. none
1944 * means the driver is handling statistics allocation/
1945 * freeing internally.
1946 * @lstats: Loopback statistics: packets, bytes
1947 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes
1948 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes
1949 *
1950 * @garp_port: GARP
1951 * @mrp_port: MRP
1952 *
1953 * @dm_private: Drop monitor private
1954 *
1955 * @dev: Class/net/name entry
1956 * @sysfs_groups: Space for optional device, statistics and wireless
1957 * sysfs groups
1958 *
1959 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1960 * @rtnl_link_ops: Rtnl_link_ops
1961 * @stat_ops: Optional ops for queue-aware statistics
1962 * @queue_mgmt_ops: Optional ops for queue management
1963 *
1964 * @gso_max_size: Maximum size of generic segmentation offload
1965 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1966 * @gso_max_segs: Maximum number of segments that can be passed to the
1967 * NIC for GSO
1968 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1969 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
1970 * for IPv4.
1971 *
1972 * @dcbnl_ops: Data Center Bridging netlink ops
1973 * @num_tc: Number of traffic classes in the net device
1974 * @tc_to_txq: XXX: need comments on this one
1975 * @prio_tc_map: XXX: need comments on this one
1976 *
1977 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1978 *
1979 * @priomap: XXX: need comments on this one
1980 * @link_topo: Physical link topology tracking attached PHYs
1981 * @phydev: Physical device may attach itself
1982 * for hardware timestamping
1983 * @sfp_bus: attached &struct sfp_bus structure.
1984 *
1985 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1986 *
1987 * @proto_down: protocol port state information can be sent to the
1988 * switch driver and used to set the phys state of the
1989 * switch port.
1990 *
1991 * @threaded: napi threaded mode is enabled
1992 *
1993 * @see_all_hwtstamp_requests: device wants to see calls to
1994 * ndo_hwtstamp_set() for all timestamp requests
1995 * regardless of source, even if those aren't
1996 * HWTSTAMP_SOURCE_NETDEV
1997 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN
1998 * @netns_local: interface can't change network namespaces
1999 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes
2000 *
2001 * @net_notifier_list: List of per-net netdev notifier block
2002 * that follow this device when it is moved
2003 * to another network namespace.
2004 *
2005 * @macsec_ops: MACsec offloading ops
2006 *
2007 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2008 * offload capabilities of the device
2009 * @udp_tunnel_nic: UDP tunnel offload state
2010 * @ethtool: ethtool related state
2011 * @xdp_state: stores info on attached XDP BPF programs
2012 *
2013 * @nested_level: Used as a parameter of spin_lock_nested() of
2014 * dev->addr_list_lock.
2015 * @unlink_list: As netif_addr_lock() can be called recursively,
2016 * keep a list of interfaces to be deleted.
2017 * @gro_max_size: Maximum size of aggregated packet in generic
2018 * receive offload (GRO)
2019 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2020 * receive offload (GRO), for IPv4.
2021 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2022 * zero copy driver
2023 *
2024 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2025 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2026 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2027 * @dev_registered_tracker: tracker for reference held while
2028 * registered
2029 * @offload_xstats_l3: L3 HW stats for this netdevice.
2030 *
2031 * @devlink_port: Pointer to related devlink port structure.
2032 * Assigned by a driver before netdev registration using
2033 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2034 * during the time netdevice is registered.
2035 *
2036 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2037 * where the clock is recovered.
2038 *
2039 * FIXME: cleanup struct net_device such that network protocol info
2040 * moves out.
2041 */
2042
2043 struct net_device {
2044 /* Cacheline organization can be found documented in
2045 * Documentation/networking/net_cachelines/net_device.rst.
2046 * Please update the document when adding new fields.
2047 */
2048
2049 /* TX read-mostly hotpath */
2050 __cacheline_group_begin(net_device_read_tx);
2051 struct_group(priv_flags_fast,
2052 unsigned long priv_flags:32;
2053 unsigned long lltx:1;
2054 );
2055 const struct net_device_ops *netdev_ops;
2056 const struct header_ops *header_ops;
2057 struct netdev_queue *_tx;
2058 netdev_features_t gso_partial_features;
2059 unsigned int real_num_tx_queues;
2060 unsigned int gso_max_size;
2061 unsigned int gso_ipv4_max_size;
2062 u16 gso_max_segs;
2063 s16 num_tc;
2064 /* Note : dev->mtu is often read without holding a lock.
2065 * Writers usually hold RTNL.
2066 * It is recommended to use READ_ONCE() to annotate the reads,
2067 * and to use WRITE_ONCE() to annotate the writes.
2068 */
2069 unsigned int mtu;
2070 unsigned short needed_headroom;
2071 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2072 #ifdef CONFIG_XPS
2073 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2074 #endif
2075 #ifdef CONFIG_NETFILTER_EGRESS
2076 struct nf_hook_entries __rcu *nf_hooks_egress;
2077 #endif
2078 #ifdef CONFIG_NET_XGRESS
2079 struct bpf_mprog_entry __rcu *tcx_egress;
2080 #endif
2081 __cacheline_group_end(net_device_read_tx);
2082
2083 /* TXRX read-mostly hotpath */
2084 __cacheline_group_begin(net_device_read_txrx);
2085 union {
2086 struct pcpu_lstats __percpu *lstats;
2087 struct pcpu_sw_netstats __percpu *tstats;
2088 struct pcpu_dstats __percpu *dstats;
2089 };
2090 unsigned long state;
2091 unsigned int flags;
2092 unsigned short hard_header_len;
2093 netdev_features_t features;
2094 struct inet6_dev __rcu *ip6_ptr;
2095 __cacheline_group_end(net_device_read_txrx);
2096
2097 /* RX read-mostly hotpath */
2098 __cacheline_group_begin(net_device_read_rx);
2099 struct bpf_prog __rcu *xdp_prog;
2100 struct list_head ptype_specific;
2101 int ifindex;
2102 unsigned int real_num_rx_queues;
2103 struct netdev_rx_queue *_rx;
2104 unsigned long gro_flush_timeout;
2105 u32 napi_defer_hard_irqs;
2106 unsigned int gro_max_size;
2107 unsigned int gro_ipv4_max_size;
2108 rx_handler_func_t __rcu *rx_handler;
2109 void __rcu *rx_handler_data;
2110 possible_net_t nd_net;
2111 #ifdef CONFIG_NETPOLL
2112 struct netpoll_info __rcu *npinfo;
2113 #endif
2114 #ifdef CONFIG_NET_XGRESS
2115 struct bpf_mprog_entry __rcu *tcx_ingress;
2116 #endif
2117 __cacheline_group_end(net_device_read_rx);
2118
2119 char name[IFNAMSIZ];
2120 struct netdev_name_node *name_node;
2121 struct dev_ifalias __rcu *ifalias;
2122 /*
2123 * I/O specific fields
2124 * FIXME: Merge these and struct ifmap into one
2125 */
2126 unsigned long mem_end;
2127 unsigned long mem_start;
2128 unsigned long base_addr;
2129
2130 /*
2131 * Some hardware also needs these fields (state,dev_list,
2132 * napi_list,unreg_list,close_list) but they are not
2133 * part of the usual set specified in Space.c.
2134 */
2135
2136
2137 struct list_head dev_list;
2138 struct list_head napi_list;
2139 struct list_head unreg_list;
2140 struct list_head close_list;
2141 struct list_head ptype_all;
2142
2143 struct {
2144 struct list_head upper;
2145 struct list_head lower;
2146 } adj_list;
2147
2148 /* Read-mostly cache-line for fast-path access */
2149 xdp_features_t xdp_features;
2150 const struct xdp_metadata_ops *xdp_metadata_ops;
2151 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2152 unsigned short gflags;
2153
2154 unsigned short needed_tailroom;
2155
2156 netdev_features_t hw_features;
2157 netdev_features_t wanted_features;
2158 netdev_features_t vlan_features;
2159 netdev_features_t hw_enc_features;
2160 netdev_features_t mpls_features;
2161
2162 unsigned int min_mtu;
2163 unsigned int max_mtu;
2164 unsigned short type;
2165 unsigned char min_header_len;
2166 unsigned char name_assign_type;
2167
2168 int group;
2169
2170 struct net_device_stats stats; /* not used by modern drivers */
2171
2172 struct net_device_core_stats __percpu *core_stats;
2173
2174 /* Stats to monitor link on/off, flapping */
2175 atomic_t carrier_up_count;
2176 atomic_t carrier_down_count;
2177
2178 /* Android KMI hack to allow vendors to have their own wifi changes in modules */
2179 const struct iw_handler_def *wireless_handlers;
2180 struct iw_public_data *wireless_data;
2181 const struct ethtool_ops *ethtool_ops;
2182 #ifdef CONFIG_NET_L3_MASTER_DEV
2183 const struct l3mdev_ops *l3mdev_ops;
2184 #endif
2185 #if IS_ENABLED(CONFIG_IPV6)
2186 const struct ndisc_ops *ndisc_ops;
2187 #endif
2188
2189 #ifdef CONFIG_XFRM_OFFLOAD
2190 const struct xfrmdev_ops *xfrmdev_ops;
2191 #endif
2192
2193 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2194 const struct tlsdev_ops *tlsdev_ops;
2195 #endif
2196
2197 unsigned int operstate;
2198 unsigned char link_mode;
2199
2200 unsigned char if_port;
2201 unsigned char dma;
2202
2203 /* Interface address info. */
2204 unsigned char perm_addr[MAX_ADDR_LEN];
2205 unsigned char addr_assign_type;
2206 unsigned char addr_len;
2207 unsigned char upper_level;
2208 unsigned char lower_level;
2209
2210 unsigned short neigh_priv_len;
2211 unsigned short dev_id;
2212 unsigned short dev_port;
2213 int irq;
2214 u32 priv_len;
2215
2216 spinlock_t addr_list_lock;
2217
2218 struct netdev_hw_addr_list uc;
2219 struct netdev_hw_addr_list mc;
2220 struct netdev_hw_addr_list dev_addrs;
2221
2222 #ifdef CONFIG_SYSFS
2223 struct kset *queues_kset;
2224 #endif
2225 #ifdef CONFIG_LOCKDEP
2226 struct list_head unlink_list;
2227 #endif
2228 unsigned int promiscuity;
2229 unsigned int allmulti;
2230 bool uc_promisc;
2231 #ifdef CONFIG_LOCKDEP
2232 unsigned char nested_level;
2233 #endif
2234
2235
2236 /* Protocol-specific pointers */
2237 struct in_device __rcu *ip_ptr;
2238 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2239 struct vlan_info __rcu *vlan_info;
2240 #endif
2241 #if IS_ENABLED(CONFIG_NET_DSA)
2242 struct dsa_port *dsa_ptr;
2243 #endif
2244 #if IS_ENABLED(CONFIG_TIPC)
2245 struct tipc_bearer __rcu *tipc_ptr;
2246 #endif
2247 #if IS_ENABLED(CONFIG_ATALK)
2248 void *atalk_ptr;
2249 #endif
2250 #if IS_ENABLED(CONFIG_AX25)
2251 struct ax25_dev __rcu *ax25_ptr;
2252 #endif
2253 /* Android KMI hack to allow vendors to have their own wifi changes in modules */
2254 struct wireless_dev *ieee80211_ptr;
2255 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2256 struct wpan_dev *ieee802154_ptr;
2257 #endif
2258 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2259 struct mpls_dev __rcu *mpls_ptr;
2260 #endif
2261 #if IS_ENABLED(CONFIG_MCTP)
2262 struct mctp_dev __rcu *mctp_ptr;
2263 #endif
2264
2265 /*
2266 * Cache lines mostly used on receive path (including eth_type_trans())
2267 */
2268 /* Interface address info used in eth_type_trans() */
2269 const unsigned char *dev_addr;
2270
2271 unsigned int num_rx_queues;
2272 #define GRO_LEGACY_MAX_SIZE 65536u
2273 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2274 * and shinfo->gso_segs is a 16bit field.
2275 */
2276 #define GRO_MAX_SIZE (8 * 65535u)
2277 unsigned int xdp_zc_max_segs;
2278 struct netdev_queue __rcu *ingress_queue;
2279 #ifdef CONFIG_NETFILTER_INGRESS
2280 struct nf_hook_entries __rcu *nf_hooks_ingress;
2281 #endif
2282
2283 unsigned char broadcast[MAX_ADDR_LEN];
2284 #ifdef CONFIG_RFS_ACCEL
2285 struct cpu_rmap *rx_cpu_rmap;
2286 #endif
2287 struct hlist_node index_hlist;
2288
2289 /*
2290 * Cache lines mostly used on transmit path
2291 */
2292 unsigned int num_tx_queues;
2293 struct Qdisc __rcu *qdisc;
2294 unsigned int tx_queue_len;
2295 spinlock_t tx_global_lock;
2296
2297 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2298
2299 #ifdef CONFIG_NET_SCHED
2300 DECLARE_HASHTABLE (qdisc_hash, 4);
2301 #endif
2302 /* These may be needed for future network-power-down code. */
2303 struct timer_list watchdog_timer;
2304 int watchdog_timeo;
2305
2306 u32 proto_down_reason;
2307
2308 struct list_head todo_list;
2309
2310 #ifdef CONFIG_PCPU_DEV_REFCNT
2311 int __percpu *pcpu_refcnt;
2312 #else
2313 refcount_t dev_refcnt;
2314 #endif
2315 struct ref_tracker_dir refcnt_tracker;
2316
2317 struct list_head link_watch_list;
2318
2319 u8 reg_state;
2320
2321 bool dismantle;
2322
2323 enum {
2324 RTNL_LINK_INITIALIZED,
2325 RTNL_LINK_INITIALIZING,
2326 } rtnl_link_state:16;
2327
2328 bool needs_free_netdev;
2329 void (*priv_destructor)(struct net_device *dev);
2330
2331 /* mid-layer private */
2332 void *ml_priv;
2333 enum netdev_ml_priv_type ml_priv_type;
2334
2335 enum netdev_stat_type pcpu_stat_type:8;
2336
2337 #if IS_ENABLED(CONFIG_GARP)
2338 struct garp_port __rcu *garp_port;
2339 #endif
2340 #if IS_ENABLED(CONFIG_MRP)
2341 struct mrp_port __rcu *mrp_port;
2342 #endif
2343 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2344 struct dm_hw_stat_delta __rcu *dm_private;
2345 #endif
2346 struct device dev;
2347 const struct attribute_group *sysfs_groups[4];
2348 const struct attribute_group *sysfs_rx_queue_group;
2349
2350 const struct rtnl_link_ops *rtnl_link_ops;
2351
2352 const struct netdev_stat_ops *stat_ops;
2353
2354 const struct netdev_queue_mgmt_ops *queue_mgmt_ops;
2355
2356 /* for setting kernel sock attribute on TCP connection setup */
2357 #define GSO_MAX_SEGS 65535u
2358 #define GSO_LEGACY_MAX_SIZE 65536u
2359 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2360 * and shinfo->gso_segs is a 16bit field.
2361 */
2362 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2363
2364 #define TSO_LEGACY_MAX_SIZE 65536
2365 #define TSO_MAX_SIZE UINT_MAX
2366 unsigned int tso_max_size;
2367 #define TSO_MAX_SEGS U16_MAX
2368 u16 tso_max_segs;
2369
2370 #ifdef CONFIG_DCB
2371 const struct dcbnl_rtnl_ops *dcbnl_ops;
2372 #endif
2373 u8 prio_tc_map[TC_BITMASK + 1];
2374
2375 #if IS_ENABLED(CONFIG_FCOE)
2376 unsigned int fcoe_ddp_xid;
2377 #endif
2378 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2379 struct netprio_map __rcu *priomap;
2380 #endif
2381 struct phy_link_topology *link_topo;
2382 struct phy_device *phydev;
2383 struct sfp_bus *sfp_bus;
2384 struct lock_class_key *qdisc_tx_busylock;
2385 bool proto_down;
2386 bool threaded;
2387
2388 /* priv_flags_slow, ungrouped to save space */
2389 unsigned long see_all_hwtstamp_requests:1;
2390 unsigned long change_proto_down:1;
2391 unsigned long netns_local:1;
2392 unsigned long fcoe_mtu:1;
2393
2394 struct list_head net_notifier_list;
2395
2396 #if IS_ENABLED(CONFIG_MACSEC)
2397 /* MACsec management functions */
2398 const struct macsec_ops *macsec_ops;
2399 #endif
2400 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2401 struct udp_tunnel_nic *udp_tunnel_nic;
2402
2403 struct ethtool_netdev_state *ethtool;
2404
2405 /* protected by rtnl_lock */
2406 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2407
2408 u8 dev_addr_shadow[MAX_ADDR_LEN];
2409 netdevice_tracker linkwatch_dev_tracker;
2410 netdevice_tracker watchdog_dev_tracker;
2411 netdevice_tracker dev_registered_tracker;
2412 struct rtnl_hw_stats64 *offload_xstats_l3;
2413
2414 struct devlink_port *devlink_port;
2415
2416 #if IS_ENABLED(CONFIG_DPLL)
2417 struct dpll_pin __rcu *dpll_pin;
2418 #endif
2419 #if IS_ENABLED(CONFIG_PAGE_POOL)
2420 /** @page_pools: page pools created for this netdevice */
2421 struct hlist_head page_pools;
2422 #endif
2423
2424 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */
2425 struct dim_irq_moder *irq_moder;
2426
2427 ANDROID_KABI_RESERVE(1);
2428 ANDROID_KABI_RESERVE(2);
2429 ANDROID_KABI_RESERVE(3);
2430 ANDROID_KABI_RESERVE(4);
2431 ANDROID_KABI_RESERVE(5);
2432 ANDROID_KABI_RESERVE(6);
2433 ANDROID_KABI_RESERVE(7);
2434 ANDROID_KABI_RESERVE(8);
2435
2436 u8 priv[] ____cacheline_aligned
2437 __counted_by(priv_len);
2438 } ____cacheline_aligned;
2439 #define to_net_dev(d) container_of(d, struct net_device, dev)
2440
2441 /*
2442 * Driver should use this to assign devlink port instance to a netdevice
2443 * before it registers the netdevice. Therefore devlink_port is static
2444 * during the netdev lifetime after it is registered.
2445 */
2446 #define SET_NETDEV_DEVLINK_PORT(dev, port) \
2447 ({ \
2448 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2449 ((dev)->devlink_port = (port)); \
2450 })
2451
netif_elide_gro(const struct net_device * dev)2452 static inline bool netif_elide_gro(const struct net_device *dev)
2453 {
2454 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2455 return true;
2456 return false;
2457 }
2458
2459 #define NETDEV_ALIGN 32
2460
2461 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)2462 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2463 {
2464 return dev->prio_tc_map[prio & TC_BITMASK];
2465 }
2466
2467 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)2468 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2469 {
2470 if (tc >= dev->num_tc)
2471 return -EINVAL;
2472
2473 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2474 return 0;
2475 }
2476
2477 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2478 void netdev_reset_tc(struct net_device *dev);
2479 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2480 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2481
2482 static inline
netdev_get_num_tc(struct net_device * dev)2483 int netdev_get_num_tc(struct net_device *dev)
2484 {
2485 return dev->num_tc;
2486 }
2487
net_prefetch(void * p)2488 static inline void net_prefetch(void *p)
2489 {
2490 prefetch(p);
2491 #if L1_CACHE_BYTES < 128
2492 prefetch((u8 *)p + L1_CACHE_BYTES);
2493 #endif
2494 }
2495
net_prefetchw(void * p)2496 static inline void net_prefetchw(void *p)
2497 {
2498 prefetchw(p);
2499 #if L1_CACHE_BYTES < 128
2500 prefetchw((u8 *)p + L1_CACHE_BYTES);
2501 #endif
2502 }
2503
2504 void netdev_unbind_sb_channel(struct net_device *dev,
2505 struct net_device *sb_dev);
2506 int netdev_bind_sb_channel_queue(struct net_device *dev,
2507 struct net_device *sb_dev,
2508 u8 tc, u16 count, u16 offset);
2509 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
netdev_get_sb_channel(struct net_device * dev)2510 static inline int netdev_get_sb_channel(struct net_device *dev)
2511 {
2512 return max_t(int, -dev->num_tc, 0);
2513 }
2514
2515 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)2516 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2517 unsigned int index)
2518 {
2519 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2520 return &dev->_tx[index];
2521 }
2522
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)2523 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2524 const struct sk_buff *skb)
2525 {
2526 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2527 }
2528
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)2529 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2530 void (*f)(struct net_device *,
2531 struct netdev_queue *,
2532 void *),
2533 void *arg)
2534 {
2535 unsigned int i;
2536
2537 for (i = 0; i < dev->num_tx_queues; i++)
2538 f(dev, &dev->_tx[i], arg);
2539 }
2540
2541 #define netdev_lockdep_set_classes(dev) \
2542 { \
2543 static struct lock_class_key qdisc_tx_busylock_key; \
2544 static struct lock_class_key qdisc_xmit_lock_key; \
2545 static struct lock_class_key dev_addr_list_lock_key; \
2546 unsigned int i; \
2547 \
2548 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2549 lockdep_set_class(&(dev)->addr_list_lock, \
2550 &dev_addr_list_lock_key); \
2551 for (i = 0; i < (dev)->num_tx_queues; i++) \
2552 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2553 &qdisc_xmit_lock_key); \
2554 }
2555
2556 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2557 struct net_device *sb_dev);
2558 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2559 struct sk_buff *skb,
2560 struct net_device *sb_dev);
2561
2562 /* returns the headroom that the master device needs to take in account
2563 * when forwarding to this dev
2564 */
netdev_get_fwd_headroom(struct net_device * dev)2565 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2566 {
2567 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2568 }
2569
netdev_set_rx_headroom(struct net_device * dev,int new_hr)2570 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2571 {
2572 if (dev->netdev_ops->ndo_set_rx_headroom)
2573 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2574 }
2575
2576 /* set the device rx headroom to the dev's default */
netdev_reset_rx_headroom(struct net_device * dev)2577 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2578 {
2579 netdev_set_rx_headroom(dev, -1);
2580 }
2581
netdev_get_ml_priv(struct net_device * dev,enum netdev_ml_priv_type type)2582 static inline void *netdev_get_ml_priv(struct net_device *dev,
2583 enum netdev_ml_priv_type type)
2584 {
2585 if (dev->ml_priv_type != type)
2586 return NULL;
2587
2588 return dev->ml_priv;
2589 }
2590
netdev_set_ml_priv(struct net_device * dev,void * ml_priv,enum netdev_ml_priv_type type)2591 static inline void netdev_set_ml_priv(struct net_device *dev,
2592 void *ml_priv,
2593 enum netdev_ml_priv_type type)
2594 {
2595 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2596 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2597 dev->ml_priv_type, type);
2598 WARN(!dev->ml_priv_type && dev->ml_priv,
2599 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2600
2601 dev->ml_priv = ml_priv;
2602 dev->ml_priv_type = type;
2603 }
2604
2605 /*
2606 * Net namespace inlines
2607 */
2608 static inline
dev_net(const struct net_device * dev)2609 struct net *dev_net(const struct net_device *dev)
2610 {
2611 return read_pnet(&dev->nd_net);
2612 }
2613
2614 static inline
dev_net_rcu(const struct net_device * dev)2615 struct net *dev_net_rcu(const struct net_device *dev)
2616 {
2617 return read_pnet_rcu(&dev->nd_net);
2618 }
2619
2620 static inline
dev_net_set(struct net_device * dev,struct net * net)2621 void dev_net_set(struct net_device *dev, struct net *net)
2622 {
2623 write_pnet(&dev->nd_net, net);
2624 }
2625
2626 /**
2627 * netdev_priv - access network device private data
2628 * @dev: network device
2629 *
2630 * Get network device private data
2631 */
netdev_priv(const struct net_device * dev)2632 static inline void *netdev_priv(const struct net_device *dev)
2633 {
2634 return (void *)dev->priv;
2635 }
2636
2637 /* Set the sysfs physical device reference for the network logical device
2638 * if set prior to registration will cause a symlink during initialization.
2639 */
2640 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2641
2642 /* Set the sysfs device type for the network logical device to allow
2643 * fine-grained identification of different network device types. For
2644 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2645 */
2646 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2647
2648 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2649 enum netdev_queue_type type,
2650 struct napi_struct *napi);
2651
netif_napi_set_irq(struct napi_struct * napi,int irq)2652 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2653 {
2654 napi->irq = irq;
2655 }
2656
2657 /* Default NAPI poll() weight
2658 * Device drivers are strongly advised to not use bigger value
2659 */
2660 #define NAPI_POLL_WEIGHT 64
2661
2662 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2663 int (*poll)(struct napi_struct *, int), int weight);
2664
2665 /**
2666 * netif_napi_add() - initialize a NAPI context
2667 * @dev: network device
2668 * @napi: NAPI context
2669 * @poll: polling function
2670 *
2671 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2672 * *any* of the other NAPI-related functions.
2673 */
2674 static inline void
netif_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int))2675 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2676 int (*poll)(struct napi_struct *, int))
2677 {
2678 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2679 }
2680
2681 static inline void
netif_napi_add_tx_weight(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2682 netif_napi_add_tx_weight(struct net_device *dev,
2683 struct napi_struct *napi,
2684 int (*poll)(struct napi_struct *, int),
2685 int weight)
2686 {
2687 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2688 netif_napi_add_weight(dev, napi, poll, weight);
2689 }
2690
2691 /**
2692 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2693 * @dev: network device
2694 * @napi: NAPI context
2695 * @poll: polling function
2696 *
2697 * This variant of netif_napi_add() should be used from drivers using NAPI
2698 * to exclusively poll a TX queue.
2699 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2700 */
netif_napi_add_tx(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int))2701 static inline void netif_napi_add_tx(struct net_device *dev,
2702 struct napi_struct *napi,
2703 int (*poll)(struct napi_struct *, int))
2704 {
2705 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2706 }
2707
2708 /**
2709 * __netif_napi_del - remove a NAPI context
2710 * @napi: NAPI context
2711 *
2712 * Warning: caller must observe RCU grace period before freeing memory
2713 * containing @napi. Drivers might want to call this helper to combine
2714 * all the needed RCU grace periods into a single one.
2715 */
2716 void __netif_napi_del(struct napi_struct *napi);
2717
2718 /**
2719 * netif_napi_del - remove a NAPI context
2720 * @napi: NAPI context
2721 *
2722 * netif_napi_del() removes a NAPI context from the network device NAPI list
2723 */
netif_napi_del(struct napi_struct * napi)2724 static inline void netif_napi_del(struct napi_struct *napi)
2725 {
2726 __netif_napi_del(napi);
2727 synchronize_net();
2728 }
2729
2730 struct packet_type {
2731 __be16 type; /* This is really htons(ether_type). */
2732 bool ignore_outgoing;
2733 struct net_device *dev; /* NULL is wildcarded here */
2734 netdevice_tracker dev_tracker;
2735 int (*func) (struct sk_buff *,
2736 struct net_device *,
2737 struct packet_type *,
2738 struct net_device *);
2739 void (*list_func) (struct list_head *,
2740 struct packet_type *,
2741 struct net_device *);
2742 bool (*id_match)(struct packet_type *ptype,
2743 struct sock *sk);
2744 struct net *af_packet_net;
2745 void *af_packet_priv;
2746 struct list_head list;
2747
2748 ANDROID_KABI_RESERVE(1);
2749 ANDROID_KABI_RESERVE(2);
2750 ANDROID_KABI_RESERVE(3);
2751 ANDROID_KABI_RESERVE(4);
2752 };
2753
2754 struct offload_callbacks {
2755 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2756 netdev_features_t features);
2757 struct sk_buff *(*gro_receive)(struct list_head *head,
2758 struct sk_buff *skb);
2759 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2760 };
2761
2762 struct packet_offload {
2763 __be16 type; /* This is really htons(ether_type). */
2764 u16 priority;
2765 struct offload_callbacks callbacks;
2766 struct list_head list;
2767 };
2768
2769 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2770 struct pcpu_sw_netstats {
2771 u64_stats_t rx_packets;
2772 u64_stats_t rx_bytes;
2773 u64_stats_t tx_packets;
2774 u64_stats_t tx_bytes;
2775 struct u64_stats_sync syncp;
2776 } __aligned(4 * sizeof(u64));
2777
2778 struct pcpu_dstats {
2779 u64_stats_t rx_packets;
2780 u64_stats_t rx_bytes;
2781 u64_stats_t rx_drops;
2782 u64_stats_t tx_packets;
2783 u64_stats_t tx_bytes;
2784 u64_stats_t tx_drops;
2785 struct u64_stats_sync syncp;
2786 } __aligned(8 * sizeof(u64));
2787
2788 struct pcpu_lstats {
2789 u64_stats_t packets;
2790 u64_stats_t bytes;
2791 struct u64_stats_sync syncp;
2792 } __aligned(2 * sizeof(u64));
2793
2794 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2795
dev_sw_netstats_rx_add(struct net_device * dev,unsigned int len)2796 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2797 {
2798 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2799
2800 u64_stats_update_begin(&tstats->syncp);
2801 u64_stats_add(&tstats->rx_bytes, len);
2802 u64_stats_inc(&tstats->rx_packets);
2803 u64_stats_update_end(&tstats->syncp);
2804 }
2805
dev_sw_netstats_tx_add(struct net_device * dev,unsigned int packets,unsigned int len)2806 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2807 unsigned int packets,
2808 unsigned int len)
2809 {
2810 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2811
2812 u64_stats_update_begin(&tstats->syncp);
2813 u64_stats_add(&tstats->tx_bytes, len);
2814 u64_stats_add(&tstats->tx_packets, packets);
2815 u64_stats_update_end(&tstats->syncp);
2816 }
2817
dev_lstats_add(struct net_device * dev,unsigned int len)2818 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2819 {
2820 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2821
2822 u64_stats_update_begin(&lstats->syncp);
2823 u64_stats_add(&lstats->bytes, len);
2824 u64_stats_inc(&lstats->packets);
2825 u64_stats_update_end(&lstats->syncp);
2826 }
2827
2828 #define __netdev_alloc_pcpu_stats(type, gfp) \
2829 ({ \
2830 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2831 if (pcpu_stats) { \
2832 int __cpu; \
2833 for_each_possible_cpu(__cpu) { \
2834 typeof(type) *stat; \
2835 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2836 u64_stats_init(&stat->syncp); \
2837 } \
2838 } \
2839 pcpu_stats; \
2840 })
2841
2842 #define netdev_alloc_pcpu_stats(type) \
2843 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2844
2845 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2846 ({ \
2847 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2848 if (pcpu_stats) { \
2849 int __cpu; \
2850 for_each_possible_cpu(__cpu) { \
2851 typeof(type) *stat; \
2852 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2853 u64_stats_init(&stat->syncp); \
2854 } \
2855 } \
2856 pcpu_stats; \
2857 })
2858
2859 enum netdev_lag_tx_type {
2860 NETDEV_LAG_TX_TYPE_UNKNOWN,
2861 NETDEV_LAG_TX_TYPE_RANDOM,
2862 NETDEV_LAG_TX_TYPE_BROADCAST,
2863 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2864 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2865 NETDEV_LAG_TX_TYPE_HASH,
2866 };
2867
2868 enum netdev_lag_hash {
2869 NETDEV_LAG_HASH_NONE,
2870 NETDEV_LAG_HASH_L2,
2871 NETDEV_LAG_HASH_L34,
2872 NETDEV_LAG_HASH_L23,
2873 NETDEV_LAG_HASH_E23,
2874 NETDEV_LAG_HASH_E34,
2875 NETDEV_LAG_HASH_VLAN_SRCMAC,
2876 NETDEV_LAG_HASH_UNKNOWN,
2877 };
2878
2879 struct netdev_lag_upper_info {
2880 enum netdev_lag_tx_type tx_type;
2881 enum netdev_lag_hash hash_type;
2882 };
2883
2884 struct netdev_lag_lower_state_info {
2885 u8 link_up : 1,
2886 tx_enabled : 1;
2887 };
2888
2889 #include <linux/notifier.h>
2890
2891 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2892 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2893 * adding new types.
2894 */
2895 enum netdev_cmd {
2896 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2897 NETDEV_DOWN,
2898 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2899 detected a hardware crash and restarted
2900 - we can use this eg to kick tcp sessions
2901 once done */
2902 NETDEV_CHANGE, /* Notify device state change */
2903 NETDEV_REGISTER,
2904 NETDEV_UNREGISTER,
2905 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2906 NETDEV_CHANGEADDR, /* notify after the address change */
2907 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2908 NETDEV_GOING_DOWN,
2909 NETDEV_CHANGENAME,
2910 NETDEV_FEAT_CHANGE,
2911 NETDEV_BONDING_FAILOVER,
2912 NETDEV_PRE_UP,
2913 NETDEV_PRE_TYPE_CHANGE,
2914 NETDEV_POST_TYPE_CHANGE,
2915 NETDEV_POST_INIT,
2916 NETDEV_PRE_UNINIT,
2917 NETDEV_RELEASE,
2918 NETDEV_NOTIFY_PEERS,
2919 NETDEV_JOIN,
2920 NETDEV_CHANGEUPPER,
2921 NETDEV_RESEND_IGMP,
2922 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2923 NETDEV_CHANGEINFODATA,
2924 NETDEV_BONDING_INFO,
2925 NETDEV_PRECHANGEUPPER,
2926 NETDEV_CHANGELOWERSTATE,
2927 NETDEV_UDP_TUNNEL_PUSH_INFO,
2928 NETDEV_UDP_TUNNEL_DROP_INFO,
2929 NETDEV_CHANGE_TX_QUEUE_LEN,
2930 NETDEV_CVLAN_FILTER_PUSH_INFO,
2931 NETDEV_CVLAN_FILTER_DROP_INFO,
2932 NETDEV_SVLAN_FILTER_PUSH_INFO,
2933 NETDEV_SVLAN_FILTER_DROP_INFO,
2934 NETDEV_OFFLOAD_XSTATS_ENABLE,
2935 NETDEV_OFFLOAD_XSTATS_DISABLE,
2936 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2937 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2938 NETDEV_XDP_FEAT_CHANGE,
2939 };
2940 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2941
2942 int register_netdevice_notifier(struct notifier_block *nb);
2943 int unregister_netdevice_notifier(struct notifier_block *nb);
2944 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2945 int unregister_netdevice_notifier_net(struct net *net,
2946 struct notifier_block *nb);
2947 int register_netdevice_notifier_dev_net(struct net_device *dev,
2948 struct notifier_block *nb,
2949 struct netdev_net_notifier *nn);
2950 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2951 struct notifier_block *nb,
2952 struct netdev_net_notifier *nn);
2953
2954 struct netdev_notifier_info {
2955 struct net_device *dev;
2956 struct netlink_ext_ack *extack;
2957 };
2958
2959 struct netdev_notifier_info_ext {
2960 struct netdev_notifier_info info; /* must be first */
2961 union {
2962 u32 mtu;
2963 } ext;
2964 };
2965
2966 struct netdev_notifier_change_info {
2967 struct netdev_notifier_info info; /* must be first */
2968 unsigned int flags_changed;
2969 };
2970
2971 struct netdev_notifier_changeupper_info {
2972 struct netdev_notifier_info info; /* must be first */
2973 struct net_device *upper_dev; /* new upper dev */
2974 bool master; /* is upper dev master */
2975 bool linking; /* is the notification for link or unlink */
2976 void *upper_info; /* upper dev info */
2977 };
2978
2979 struct netdev_notifier_changelowerstate_info {
2980 struct netdev_notifier_info info; /* must be first */
2981 void *lower_state_info; /* is lower dev state */
2982 };
2983
2984 struct netdev_notifier_pre_changeaddr_info {
2985 struct netdev_notifier_info info; /* must be first */
2986 const unsigned char *dev_addr;
2987 };
2988
2989 enum netdev_offload_xstats_type {
2990 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2991 };
2992
2993 struct netdev_notifier_offload_xstats_info {
2994 struct netdev_notifier_info info; /* must be first */
2995 enum netdev_offload_xstats_type type;
2996
2997 union {
2998 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2999 struct netdev_notifier_offload_xstats_rd *report_delta;
3000 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
3001 struct netdev_notifier_offload_xstats_ru *report_used;
3002 };
3003 };
3004
3005 int netdev_offload_xstats_enable(struct net_device *dev,
3006 enum netdev_offload_xstats_type type,
3007 struct netlink_ext_ack *extack);
3008 int netdev_offload_xstats_disable(struct net_device *dev,
3009 enum netdev_offload_xstats_type type);
3010 bool netdev_offload_xstats_enabled(const struct net_device *dev,
3011 enum netdev_offload_xstats_type type);
3012 int netdev_offload_xstats_get(struct net_device *dev,
3013 enum netdev_offload_xstats_type type,
3014 struct rtnl_hw_stats64 *stats, bool *used,
3015 struct netlink_ext_ack *extack);
3016 void
3017 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3018 const struct rtnl_hw_stats64 *stats);
3019 void
3020 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3021 void netdev_offload_xstats_push_delta(struct net_device *dev,
3022 enum netdev_offload_xstats_type type,
3023 const struct rtnl_hw_stats64 *stats);
3024
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)3025 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3026 struct net_device *dev)
3027 {
3028 info->dev = dev;
3029 info->extack = NULL;
3030 }
3031
3032 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)3033 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3034 {
3035 return info->dev;
3036 }
3037
3038 static inline struct netlink_ext_ack *
netdev_notifier_info_to_extack(const struct netdev_notifier_info * info)3039 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3040 {
3041 return info->extack;
3042 }
3043
3044 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3045 int call_netdevice_notifiers_info(unsigned long val,
3046 struct netdev_notifier_info *info);
3047
3048 #define for_each_netdev(net, d) \
3049 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3050 #define for_each_netdev_reverse(net, d) \
3051 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3052 #define for_each_netdev_rcu(net, d) \
3053 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3054 #define for_each_netdev_safe(net, d, n) \
3055 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3056 #define for_each_netdev_continue(net, d) \
3057 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3058 #define for_each_netdev_continue_reverse(net, d) \
3059 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3060 dev_list)
3061 #define for_each_netdev_continue_rcu(net, d) \
3062 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3063 #define for_each_netdev_in_bond_rcu(bond, slave) \
3064 for_each_netdev_rcu(&init_net, slave) \
3065 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3066 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3067
3068 #define for_each_netdev_dump(net, d, ifindex) \
3069 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \
3070 ULONG_MAX, XA_PRESENT)); ifindex++)
3071
next_net_device(struct net_device * dev)3072 static inline struct net_device *next_net_device(struct net_device *dev)
3073 {
3074 struct list_head *lh;
3075 struct net *net;
3076
3077 net = dev_net(dev);
3078 lh = dev->dev_list.next;
3079 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3080 }
3081
next_net_device_rcu(struct net_device * dev)3082 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3083 {
3084 struct list_head *lh;
3085 struct net *net;
3086
3087 net = dev_net(dev);
3088 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3089 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3090 }
3091
first_net_device(struct net * net)3092 static inline struct net_device *first_net_device(struct net *net)
3093 {
3094 return list_empty(&net->dev_base_head) ? NULL :
3095 net_device_entry(net->dev_base_head.next);
3096 }
3097
first_net_device_rcu(struct net * net)3098 static inline struct net_device *first_net_device_rcu(struct net *net)
3099 {
3100 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3101
3102 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3103 }
3104
3105 int netdev_boot_setup_check(struct net_device *dev);
3106 struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type,
3107 const char *hwaddr);
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
3136 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3137 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3138
dev_queue_xmit(struct sk_buff * skb)3139 static inline int dev_queue_xmit(struct sk_buff *skb)
3140 {
3141 return __dev_queue_xmit(skb, NULL);
3142 }
3143
dev_queue_xmit_accel(struct sk_buff * skb,struct net_device * sb_dev)3144 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3145 struct net_device *sb_dev)
3146 {
3147 return __dev_queue_xmit(skb, sb_dev);
3148 }
3149
dev_direct_xmit(struct sk_buff * skb,u16 queue_id)3150 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3151 {
3152 int ret;
3153
3154 ret = __dev_direct_xmit(skb, queue_id);
3155 if (!dev_xmit_complete(ret))
3156 kfree_skb(skb);
3157 return ret;
3158 }
3159
3160 int register_netdevice(struct net_device *dev);
3161 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3162 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)3163 static inline void unregister_netdevice(struct net_device *dev)
3164 {
3165 unregister_netdevice_queue(dev, NULL);
3166 }
3167
3168 int netdev_refcnt_read(const struct net_device *dev);
3169 void free_netdev(struct net_device *dev);
3170 void init_dummy_netdev(struct net_device *dev);
3171
3172 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3173 struct sk_buff *skb,
3174 bool all_slaves);
3175 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3176 struct sock *sk);
3177 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3178 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3179 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3180 netdevice_tracker *tracker, gfp_t gfp);
3181 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3182 netdevice_tracker *tracker, gfp_t gfp);
3183 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3184 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3185 void netdev_copy_name(struct net_device *dev, char *name);
3186
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)3187 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3188 unsigned short type,
3189 const void *daddr, const void *saddr,
3190 unsigned int len)
3191 {
3192 if (!dev->header_ops || !dev->header_ops->create)
3193 return 0;
3194
3195 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3196 }
3197
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)3198 static inline int dev_parse_header(const struct sk_buff *skb,
3199 unsigned char *haddr)
3200 {
3201 const struct net_device *dev = skb->dev;
3202
3203 if (!dev->header_ops || !dev->header_ops->parse)
3204 return 0;
3205 return dev->header_ops->parse(skb, haddr);
3206 }
3207
dev_parse_header_protocol(const struct sk_buff * skb)3208 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3209 {
3210 const struct net_device *dev = skb->dev;
3211
3212 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3213 return 0;
3214 return dev->header_ops->parse_protocol(skb);
3215 }
3216
3217 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)3218 static inline bool dev_validate_header(const struct net_device *dev,
3219 char *ll_header, int len)
3220 {
3221 if (likely(len >= dev->hard_header_len))
3222 return true;
3223 if (len < dev->min_header_len)
3224 return false;
3225
3226 if (capable(CAP_SYS_RAWIO)) {
3227 memset(ll_header + len, 0, dev->hard_header_len - len);
3228 return true;
3229 }
3230
3231 if (dev->header_ops && dev->header_ops->validate)
3232 return dev->header_ops->validate(ll_header, len);
3233
3234 return false;
3235 }
3236
dev_has_header(const struct net_device * dev)3237 static inline bool dev_has_header(const struct net_device *dev)
3238 {
3239 return dev->header_ops && dev->header_ops->create;
3240 }
3241
3242 /*
3243 * Incoming packets are placed on per-CPU queues
3244 */
3245 struct softnet_data {
3246 struct list_head poll_list;
3247 struct sk_buff_head process_queue;
3248 local_lock_t process_queue_bh_lock;
3249
3250 /* stats */
3251 unsigned int processed;
3252 unsigned int time_squeeze;
3253 #ifdef CONFIG_RPS
3254 struct softnet_data *rps_ipi_list;
3255 #endif
3256
3257 unsigned int received_rps;
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 netdev_xmit xmit;
3272 #ifdef CONFIG_RPS
3273 /* input_queue_head should be written by cpu owning this struct,
3274 * and only read by other cpus. Worth using a cache line.
3275 */
3276 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3277
3278 /* Elements below can be accessed between CPUs for RPS/RFS */
3279 call_single_data_t csd ____cacheline_aligned_in_smp;
3280 struct softnet_data *rps_ipi_next;
3281 unsigned int cpu;
3282 unsigned int input_queue_tail;
3283 #endif
3284 struct sk_buff_head input_pkt_queue;
3285 struct napi_struct backlog;
3286
3287 atomic_t dropped ____cacheline_aligned_in_smp;
3288
3289 /* Another possibly contended cache line */
3290 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3291 int defer_count;
3292 int defer_ipi_scheduled;
3293 struct sk_buff *defer_list;
3294 call_single_data_t defer_csd;
3295 };
3296
3297 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3298
3299 #ifndef CONFIG_PREEMPT_RT
dev_recursion_level(void)3300 static inline int dev_recursion_level(void)
3301 {
3302 return this_cpu_read(softnet_data.xmit.recursion);
3303 }
3304 #else
dev_recursion_level(void)3305 static inline int dev_recursion_level(void)
3306 {
3307 return current->net_xmit.recursion;
3308 }
3309
3310 #endif
3311
3312 void __netif_schedule(struct Qdisc *q);
3313 void netif_schedule_queue(struct netdev_queue *txq);
3314
netif_tx_schedule_all(struct net_device * dev)3315 static inline void netif_tx_schedule_all(struct net_device *dev)
3316 {
3317 unsigned int i;
3318
3319 for (i = 0; i < dev->num_tx_queues; i++)
3320 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3321 }
3322
netif_tx_start_queue(struct netdev_queue * dev_queue)3323 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3324 {
3325 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3326 }
3327
3328 /**
3329 * netif_start_queue - allow transmit
3330 * @dev: network device
3331 *
3332 * Allow upper layers to call the device hard_start_xmit routine.
3333 */
netif_start_queue(struct net_device * dev)3334 static inline void netif_start_queue(struct net_device *dev)
3335 {
3336 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3337 }
3338
netif_tx_start_all_queues(struct net_device * dev)3339 static inline void netif_tx_start_all_queues(struct net_device *dev)
3340 {
3341 unsigned int i;
3342
3343 for (i = 0; i < dev->num_tx_queues; i++) {
3344 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3345 netif_tx_start_queue(txq);
3346 }
3347 }
3348
3349 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3350
3351 /**
3352 * netif_wake_queue - restart transmit
3353 * @dev: network device
3354 *
3355 * Allow upper layers to call the device hard_start_xmit routine.
3356 * Used for flow control when transmit resources are available.
3357 */
netif_wake_queue(struct net_device * dev)3358 static inline void netif_wake_queue(struct net_device *dev)
3359 {
3360 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3361 }
3362
netif_tx_wake_all_queues(struct net_device * dev)3363 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3364 {
3365 unsigned int i;
3366
3367 for (i = 0; i < dev->num_tx_queues; i++) {
3368 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3369 netif_tx_wake_queue(txq);
3370 }
3371 }
3372
netif_tx_stop_queue(struct netdev_queue * dev_queue)3373 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3374 {
3375 /* Paired with READ_ONCE() from dev_watchdog() */
3376 WRITE_ONCE(dev_queue->trans_start, jiffies);
3377
3378 /* This barrier is paired with smp_mb() from dev_watchdog() */
3379 smp_mb__before_atomic();
3380
3381 /* Must be an atomic op see netif_txq_try_stop() */
3382 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3383 }
3384
3385 /**
3386 * netif_stop_queue - stop transmitted packets
3387 * @dev: network device
3388 *
3389 * Stop upper layers calling the device hard_start_xmit routine.
3390 * Used for flow control when transmit resources are unavailable.
3391 */
netif_stop_queue(struct net_device * dev)3392 static inline void netif_stop_queue(struct net_device *dev)
3393 {
3394 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3395 }
3396
3397 void netif_tx_stop_all_queues(struct net_device *dev);
3398
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)3399 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3400 {
3401 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3402 }
3403
3404 /**
3405 * netif_queue_stopped - test if transmit queue is flowblocked
3406 * @dev: network device
3407 *
3408 * Test if transmit queue on device is currently unable to send.
3409 */
netif_queue_stopped(const struct net_device * dev)3410 static inline bool netif_queue_stopped(const struct net_device *dev)
3411 {
3412 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3413 }
3414
netif_xmit_stopped(const struct netdev_queue * dev_queue)3415 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3416 {
3417 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3418 }
3419
3420 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)3421 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3422 {
3423 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3424 }
3425
3426 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)3427 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3428 {
3429 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3430 }
3431
3432 /**
3433 * netdev_queue_set_dql_min_limit - set dql minimum limit
3434 * @dev_queue: pointer to transmit queue
3435 * @min_limit: dql minimum limit
3436 *
3437 * Forces xmit_more() to return true until the minimum threshold
3438 * defined by @min_limit is reached (or until the tx queue is
3439 * empty). Warning: to be use with care, misuse will impact the
3440 * latency.
3441 */
netdev_queue_set_dql_min_limit(struct netdev_queue * dev_queue,unsigned int min_limit)3442 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3443 unsigned int min_limit)
3444 {
3445 #ifdef CONFIG_BQL
3446 dev_queue->dql.min_limit = min_limit;
3447 #endif
3448 }
3449
netdev_queue_dql_avail(const struct netdev_queue * txq)3450 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq)
3451 {
3452 #ifdef CONFIG_BQL
3453 /* Non-BQL migrated drivers will return 0, too. */
3454 return dql_avail(&txq->dql);
3455 #else
3456 return 0;
3457 #endif
3458 }
3459
3460 /**
3461 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3462 * @dev_queue: pointer to transmit queue
3463 *
3464 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3465 * to give appropriate hint to the CPU.
3466 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)3467 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3468 {
3469 #ifdef CONFIG_BQL
3470 prefetchw(&dev_queue->dql.num_queued);
3471 #endif
3472 }
3473
3474 /**
3475 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3476 * @dev_queue: pointer to transmit queue
3477 *
3478 * BQL enabled drivers might use this helper in their TX completion path,
3479 * to give appropriate hint to the CPU.
3480 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3481 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3482 {
3483 #ifdef CONFIG_BQL
3484 prefetchw(&dev_queue->dql.limit);
3485 #endif
3486 }
3487
3488 /**
3489 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3490 * @dev_queue: network device queue
3491 * @bytes: number of bytes queued to the device queue
3492 *
3493 * Report the number of bytes queued for sending/completion to the network
3494 * device hardware queue. @bytes should be a good approximation and should
3495 * exactly match netdev_completed_queue() @bytes.
3496 * This is typically called once per packet, from ndo_start_xmit().
3497 */
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3498 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3499 unsigned int bytes)
3500 {
3501 #ifdef CONFIG_BQL
3502 dql_queued(&dev_queue->dql, bytes);
3503
3504 if (likely(dql_avail(&dev_queue->dql) >= 0))
3505 return;
3506
3507 /* Paired with READ_ONCE() from dev_watchdog() */
3508 WRITE_ONCE(dev_queue->trans_start, jiffies);
3509
3510 /* This barrier is paired with smp_mb() from dev_watchdog() */
3511 smp_mb__before_atomic();
3512
3513 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3514
3515 /*
3516 * The XOFF flag must be set before checking the dql_avail below,
3517 * because in netdev_tx_completed_queue we update the dql_completed
3518 * before checking the XOFF flag.
3519 */
3520 smp_mb();
3521
3522 /* check again in case another CPU has just made room avail */
3523 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3524 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3525 #endif
3526 }
3527
3528 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3529 * that they should not test BQL status themselves.
3530 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3531 * skb of a batch.
3532 * Returns true if the doorbell must be used to kick the NIC.
3533 */
__netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes,bool xmit_more)3534 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3535 unsigned int bytes,
3536 bool xmit_more)
3537 {
3538 if (xmit_more) {
3539 #ifdef CONFIG_BQL
3540 dql_queued(&dev_queue->dql, bytes);
3541 #endif
3542 return netif_tx_queue_stopped(dev_queue);
3543 }
3544 netdev_tx_sent_queue(dev_queue, bytes);
3545 return true;
3546 }
3547
3548 /**
3549 * netdev_sent_queue - report the number of bytes queued to hardware
3550 * @dev: network device
3551 * @bytes: number of bytes queued to the hardware device queue
3552 *
3553 * Report the number of bytes queued for sending/completion to the network
3554 * device hardware queue#0. @bytes should be a good approximation and should
3555 * exactly match netdev_completed_queue() @bytes.
3556 * This is typically called once per packet, from ndo_start_xmit().
3557 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)3558 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3559 {
3560 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3561 }
3562
__netdev_sent_queue(struct net_device * dev,unsigned int bytes,bool xmit_more)3563 static inline bool __netdev_sent_queue(struct net_device *dev,
3564 unsigned int bytes,
3565 bool xmit_more)
3566 {
3567 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3568 xmit_more);
3569 }
3570
3571 /**
3572 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3573 * @dev_queue: network device queue
3574 * @pkts: number of packets (currently ignored)
3575 * @bytes: number of bytes dequeued from the device queue
3576 *
3577 * Must be called at most once per TX completion round (and not per
3578 * individual packet), so that BQL can adjust its limits appropriately.
3579 */
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3580 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3581 unsigned int pkts, unsigned int bytes)
3582 {
3583 #ifdef CONFIG_BQL
3584 if (unlikely(!bytes))
3585 return;
3586
3587 dql_completed(&dev_queue->dql, bytes);
3588
3589 /*
3590 * Without the memory barrier there is a small possibility that
3591 * netdev_tx_sent_queue will miss the update and cause the queue to
3592 * be stopped forever
3593 */
3594 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3595
3596 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3597 return;
3598
3599 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3600 netif_schedule_queue(dev_queue);
3601 #endif
3602 }
3603
3604 /**
3605 * netdev_completed_queue - report bytes and packets completed by device
3606 * @dev: network device
3607 * @pkts: actual number of packets sent over the medium
3608 * @bytes: actual number of bytes sent over the medium
3609 *
3610 * Report the number of bytes and packets transmitted by the network device
3611 * hardware queue over the physical medium, @bytes must exactly match the
3612 * @bytes amount passed to netdev_sent_queue()
3613 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3614 static inline void netdev_completed_queue(struct net_device *dev,
3615 unsigned int pkts, unsigned int bytes)
3616 {
3617 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3618 }
3619
netdev_tx_reset_queue(struct netdev_queue * q)3620 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3621 {
3622 #ifdef CONFIG_BQL
3623 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3624 dql_reset(&q->dql);
3625 #endif
3626 }
3627
3628 /**
3629 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue
3630 * @dev: network device
3631 * @qid: stack index of the queue to reset
3632 */
netdev_tx_reset_subqueue(const struct net_device * dev,u32 qid)3633 static inline void netdev_tx_reset_subqueue(const struct net_device *dev,
3634 u32 qid)
3635 {
3636 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid));
3637 }
3638
3639 /**
3640 * netdev_reset_queue - reset the packets and bytes count of a network device
3641 * @dev_queue: network device
3642 *
3643 * Reset the bytes and packet count of a network device and clear the
3644 * software flow control OFF bit for this network device
3645 */
netdev_reset_queue(struct net_device * dev_queue)3646 static inline void netdev_reset_queue(struct net_device *dev_queue)
3647 {
3648 netdev_tx_reset_subqueue(dev_queue, 0);
3649 }
3650
3651 /**
3652 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3653 * @dev: network device
3654 * @queue_index: given tx queue index
3655 *
3656 * Returns 0 if given tx queue index >= number of device tx queues,
3657 * otherwise returns the originally passed tx queue index.
3658 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3659 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3660 {
3661 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3662 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3663 dev->name, queue_index,
3664 dev->real_num_tx_queues);
3665 return 0;
3666 }
3667
3668 return queue_index;
3669 }
3670
3671 /**
3672 * netif_running - test if up
3673 * @dev: network device
3674 *
3675 * Test if the device has been brought up.
3676 */
netif_running(const struct net_device * dev)3677 static inline bool netif_running(const struct net_device *dev)
3678 {
3679 return test_bit(__LINK_STATE_START, &dev->state);
3680 }
3681
3682 /*
3683 * Routines to manage the subqueues on a device. We only need start,
3684 * stop, and a check if it's stopped. All other device management is
3685 * done at the overall netdevice level.
3686 * Also test the device if we're multiqueue.
3687 */
3688
3689 /**
3690 * netif_start_subqueue - allow sending packets on subqueue
3691 * @dev: network device
3692 * @queue_index: sub queue index
3693 *
3694 * Start individual transmit queue of a device with multiple transmit queues.
3695 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)3696 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3697 {
3698 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3699
3700 netif_tx_start_queue(txq);
3701 }
3702
3703 /**
3704 * netif_stop_subqueue - stop sending packets on subqueue
3705 * @dev: network device
3706 * @queue_index: sub queue index
3707 *
3708 * Stop individual transmit queue of a device with multiple transmit queues.
3709 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)3710 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3711 {
3712 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3713 netif_tx_stop_queue(txq);
3714 }
3715
3716 /**
3717 * __netif_subqueue_stopped - test status of subqueue
3718 * @dev: network device
3719 * @queue_index: sub queue index
3720 *
3721 * Check individual transmit queue of a device with multiple transmit queues.
3722 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3723 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3724 u16 queue_index)
3725 {
3726 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3727
3728 return netif_tx_queue_stopped(txq);
3729 }
3730
3731 /**
3732 * netif_subqueue_stopped - test status of subqueue
3733 * @dev: network device
3734 * @skb: sub queue buffer pointer
3735 *
3736 * Check individual transmit queue of a device with multiple transmit queues.
3737 */
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3738 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3739 struct sk_buff *skb)
3740 {
3741 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3742 }
3743
3744 /**
3745 * netif_wake_subqueue - allow sending packets on subqueue
3746 * @dev: network device
3747 * @queue_index: sub queue index
3748 *
3749 * Resume individual transmit queue of a device with multiple transmit queues.
3750 */
netif_wake_subqueue(struct net_device * dev,u16 queue_index)3751 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3752 {
3753 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3754
3755 netif_tx_wake_queue(txq);
3756 }
3757
3758 #ifdef CONFIG_XPS
3759 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3760 u16 index);
3761 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3762 u16 index, enum xps_map_type type);
3763
3764 /**
3765 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3766 * @j: CPU/Rx queue index
3767 * @mask: bitmask of all cpus/rx queues
3768 * @nr_bits: number of bits in the bitmask
3769 *
3770 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3771 */
netif_attr_test_mask(unsigned long j,const unsigned long * mask,unsigned int nr_bits)3772 static inline bool netif_attr_test_mask(unsigned long j,
3773 const unsigned long *mask,
3774 unsigned int nr_bits)
3775 {
3776 cpu_max_bits_warn(j, nr_bits);
3777 return test_bit(j, mask);
3778 }
3779
3780 /**
3781 * netif_attr_test_online - Test for online CPU/Rx queue
3782 * @j: CPU/Rx queue index
3783 * @online_mask: bitmask for CPUs/Rx queues that are online
3784 * @nr_bits: number of bits in the bitmask
3785 *
3786 * Returns true if a CPU/Rx queue is online.
3787 */
netif_attr_test_online(unsigned long j,const unsigned long * online_mask,unsigned int nr_bits)3788 static inline bool netif_attr_test_online(unsigned long j,
3789 const unsigned long *online_mask,
3790 unsigned int nr_bits)
3791 {
3792 cpu_max_bits_warn(j, nr_bits);
3793
3794 if (online_mask)
3795 return test_bit(j, online_mask);
3796
3797 return (j < nr_bits);
3798 }
3799
3800 /**
3801 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3802 * @n: CPU/Rx queue index
3803 * @srcp: the cpumask/Rx queue mask pointer
3804 * @nr_bits: number of bits in the bitmask
3805 *
3806 * Returns >= nr_bits if no further CPUs/Rx queues set.
3807 */
netif_attrmask_next(int n,const unsigned long * srcp,unsigned int nr_bits)3808 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3809 unsigned int nr_bits)
3810 {
3811 /* -1 is a legal arg here. */
3812 if (n != -1)
3813 cpu_max_bits_warn(n, nr_bits);
3814
3815 if (srcp)
3816 return find_next_bit(srcp, nr_bits, n + 1);
3817
3818 return n + 1;
3819 }
3820
3821 /**
3822 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3823 * @n: CPU/Rx queue index
3824 * @src1p: the first CPUs/Rx queues mask pointer
3825 * @src2p: the second CPUs/Rx queues mask pointer
3826 * @nr_bits: number of bits in the bitmask
3827 *
3828 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3829 */
netif_attrmask_next_and(int n,const unsigned long * src1p,const unsigned long * src2p,unsigned int nr_bits)3830 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3831 const unsigned long *src2p,
3832 unsigned int nr_bits)
3833 {
3834 /* -1 is a legal arg here. */
3835 if (n != -1)
3836 cpu_max_bits_warn(n, nr_bits);
3837
3838 if (src1p && src2p)
3839 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3840 else if (src1p)
3841 return find_next_bit(src1p, nr_bits, n + 1);
3842 else if (src2p)
3843 return find_next_bit(src2p, nr_bits, n + 1);
3844
3845 return n + 1;
3846 }
3847 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3848 static inline int netif_set_xps_queue(struct net_device *dev,
3849 const struct cpumask *mask,
3850 u16 index)
3851 {
3852 return 0;
3853 }
3854
__netif_set_xps_queue(struct net_device * dev,const unsigned long * mask,u16 index,enum xps_map_type type)3855 static inline int __netif_set_xps_queue(struct net_device *dev,
3856 const unsigned long *mask,
3857 u16 index, enum xps_map_type type)
3858 {
3859 return 0;
3860 }
3861 #endif
3862
3863 /**
3864 * netif_is_multiqueue - test if device has multiple transmit queues
3865 * @dev: network device
3866 *
3867 * Check if device has multiple transmit queues
3868 */
netif_is_multiqueue(const struct net_device * dev)3869 static inline bool netif_is_multiqueue(const struct net_device *dev)
3870 {
3871 return dev->num_tx_queues > 1;
3872 }
3873
3874 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3875
3876 #ifdef CONFIG_SYSFS
3877 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3878 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxqs)3879 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3880 unsigned int rxqs)
3881 {
3882 dev->real_num_rx_queues = rxqs;
3883 return 0;
3884 }
3885 #endif
3886 int netif_set_real_num_queues(struct net_device *dev,
3887 unsigned int txq, unsigned int rxq);
3888
3889 int netif_get_num_default_rss_queues(void);
3890
3891 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3892 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3893
3894 /*
3895 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3896 * interrupt context or with hardware interrupts being disabled.
3897 * (in_hardirq() || irqs_disabled())
3898 *
3899 * We provide four helpers that can be used in following contexts :
3900 *
3901 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3902 * replacing kfree_skb(skb)
3903 *
3904 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3905 * Typically used in place of consume_skb(skb) in TX completion path
3906 *
3907 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3908 * replacing kfree_skb(skb)
3909 *
3910 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3911 * and consumed a packet. Used in place of consume_skb(skb)
3912 */
dev_kfree_skb_irq(struct sk_buff * skb)3913 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3914 {
3915 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3916 }
3917
dev_consume_skb_irq(struct sk_buff * skb)3918 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3919 {
3920 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3921 }
3922
dev_kfree_skb_any(struct sk_buff * skb)3923 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3924 {
3925 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3926 }
3927
dev_consume_skb_any(struct sk_buff * skb)3928 static inline void dev_consume_skb_any(struct sk_buff *skb)
3929 {
3930 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3931 }
3932
3933 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3934 struct bpf_prog *xdp_prog);
3935 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3936 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb);
3937 int netif_rx(struct sk_buff *skb);
3938 int __netif_rx(struct sk_buff *skb);
3939
3940 int netif_receive_skb(struct sk_buff *skb);
3941 int netif_receive_skb_core(struct sk_buff *skb);
3942 void netif_receive_skb_list_internal(struct list_head *head);
3943 void netif_receive_skb_list(struct list_head *head);
3944 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3945 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3946 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3947 void napi_get_frags_check(struct napi_struct *napi);
3948 gro_result_t napi_gro_frags(struct napi_struct *napi);
3949
napi_free_frags(struct napi_struct * napi)3950 static inline void napi_free_frags(struct napi_struct *napi)
3951 {
3952 kfree_skb(napi->skb);
3953 napi->skb = NULL;
3954 }
3955
3956 bool netdev_is_rx_handler_busy(struct net_device *dev);
3957 int netdev_rx_handler_register(struct net_device *dev,
3958 rx_handler_func_t *rx_handler,
3959 void *rx_handler_data);
3960 void netdev_rx_handler_unregister(struct net_device *dev);
3961
3962 bool dev_valid_name(const char *name);
is_socket_ioctl_cmd(unsigned int cmd)3963 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3964 {
3965 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3966 }
3967 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3968 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3969 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3970 void __user *data, bool *need_copyout);
3971 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3972 int generic_hwtstamp_get_lower(struct net_device *dev,
3973 struct kernel_hwtstamp_config *kernel_cfg);
3974 int generic_hwtstamp_set_lower(struct net_device *dev,
3975 struct kernel_hwtstamp_config *kernel_cfg,
3976 struct netlink_ext_ack *extack);
3977 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3978 unsigned int dev_get_flags(const struct net_device *);
3979 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3980 struct netlink_ext_ack *extack);
3981 int dev_change_flags(struct net_device *dev, unsigned int flags,
3982 struct netlink_ext_ack *extack);
3983 int dev_set_alias(struct net_device *, const char *, size_t);
3984 int dev_get_alias(const struct net_device *, char *, size_t);
3985 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3986 const char *pat, int new_ifindex);
3987 static inline
dev_change_net_namespace(struct net_device * dev,struct net * net,const char * pat)3988 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3989 const char *pat)
3990 {
3991 return __dev_change_net_namespace(dev, net, pat, 0);
3992 }
3993 int __dev_set_mtu(struct net_device *, int);
3994 int dev_set_mtu(struct net_device *, int);
3995 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3996 struct netlink_ext_ack *extack);
3997 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3998 struct netlink_ext_ack *extack);
3999 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4000 struct netlink_ext_ack *extack);
4001 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4002 int dev_get_port_parent_id(struct net_device *dev,
4003 struct netdev_phys_item_id *ppid, bool recurse);
4004 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4005
4006 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4007 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4008 struct netdev_queue *txq, int *ret);
4009
4010 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4011 u8 dev_xdp_prog_count(struct net_device *dev);
4012 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf);
4013 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4014
4015 u32 dev_get_min_mp_channel_count(const struct net_device *dev);
4016
4017 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4018 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4019 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4020 bool is_skb_forwardable(const struct net_device *dev,
4021 const struct sk_buff *skb);
4022
__is_skb_forwardable(const struct net_device * dev,const struct sk_buff * skb,const bool check_mtu)4023 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4024 const struct sk_buff *skb,
4025 const bool check_mtu)
4026 {
4027 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4028 unsigned int len;
4029
4030 if (!(dev->flags & IFF_UP))
4031 return false;
4032
4033 if (!check_mtu)
4034 return true;
4035
4036 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4037 if (skb->len <= len)
4038 return true;
4039
4040 /* if TSO is enabled, we don't care about the length as the packet
4041 * could be forwarded without being segmented before
4042 */
4043 if (skb_is_gso(skb))
4044 return true;
4045
4046 return false;
4047 }
4048
4049 void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4050
4051 #define DEV_CORE_STATS_INC(FIELD) \
4052 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4053 { \
4054 netdev_core_stats_inc(dev, \
4055 offsetof(struct net_device_core_stats, FIELD)); \
4056 }
4057 DEV_CORE_STATS_INC(rx_dropped)
DEV_CORE_STATS_INC(tx_dropped)4058 DEV_CORE_STATS_INC(tx_dropped)
4059 DEV_CORE_STATS_INC(rx_nohandler)
4060 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4061 #undef DEV_CORE_STATS_INC
4062
4063 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4064 struct sk_buff *skb,
4065 const bool check_mtu)
4066 {
4067 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4068 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4069 dev_core_stats_rx_dropped_inc(dev);
4070 kfree_skb(skb);
4071 return NET_RX_DROP;
4072 }
4073
4074 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4075 skb->priority = 0;
4076 return 0;
4077 }
4078
4079 bool dev_nit_active(struct net_device *dev);
4080 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4081
__dev_put(struct net_device * dev)4082 static inline void __dev_put(struct net_device *dev)
4083 {
4084 if (dev) {
4085 #ifdef CONFIG_PCPU_DEV_REFCNT
4086 this_cpu_dec(*dev->pcpu_refcnt);
4087 #else
4088 refcount_dec(&dev->dev_refcnt);
4089 #endif
4090 }
4091 }
4092
__dev_hold(struct net_device * dev)4093 static inline void __dev_hold(struct net_device *dev)
4094 {
4095 if (dev) {
4096 #ifdef CONFIG_PCPU_DEV_REFCNT
4097 this_cpu_inc(*dev->pcpu_refcnt);
4098 #else
4099 refcount_inc(&dev->dev_refcnt);
4100 #endif
4101 }
4102 }
4103
__netdev_tracker_alloc(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4104 static inline void __netdev_tracker_alloc(struct net_device *dev,
4105 netdevice_tracker *tracker,
4106 gfp_t gfp)
4107 {
4108 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4109 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4110 #endif
4111 }
4112
4113 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4114 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4115 */
netdev_tracker_alloc(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4116 static inline void netdev_tracker_alloc(struct net_device *dev,
4117 netdevice_tracker *tracker, gfp_t gfp)
4118 {
4119 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4120 refcount_dec(&dev->refcnt_tracker.no_tracker);
4121 __netdev_tracker_alloc(dev, tracker, gfp);
4122 #endif
4123 }
4124
netdev_tracker_free(struct net_device * dev,netdevice_tracker * tracker)4125 static inline void netdev_tracker_free(struct net_device *dev,
4126 netdevice_tracker *tracker)
4127 {
4128 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4129 ref_tracker_free(&dev->refcnt_tracker, tracker);
4130 #endif
4131 }
4132
netdev_hold(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4133 static inline void netdev_hold(struct net_device *dev,
4134 netdevice_tracker *tracker, gfp_t gfp)
4135 {
4136 if (dev) {
4137 __dev_hold(dev);
4138 __netdev_tracker_alloc(dev, tracker, gfp);
4139 }
4140 }
4141
netdev_put(struct net_device * dev,netdevice_tracker * tracker)4142 static inline void netdev_put(struct net_device *dev,
4143 netdevice_tracker *tracker)
4144 {
4145 if (dev) {
4146 netdev_tracker_free(dev, tracker);
4147 __dev_put(dev);
4148 }
4149 }
4150
4151 /**
4152 * dev_hold - get reference to device
4153 * @dev: network device
4154 *
4155 * Hold reference to device to keep it from being freed.
4156 * Try using netdev_hold() instead.
4157 */
dev_hold(struct net_device * dev)4158 static inline void dev_hold(struct net_device *dev)
4159 {
4160 netdev_hold(dev, NULL, GFP_ATOMIC);
4161 }
4162
4163 /**
4164 * dev_put - release reference to device
4165 * @dev: network device
4166 *
4167 * Release reference to device to allow it to be freed.
4168 * Try using netdev_put() instead.
4169 */
dev_put(struct net_device * dev)4170 static inline void dev_put(struct net_device *dev)
4171 {
4172 netdev_put(dev, NULL);
4173 }
4174
DEFINE_FREE(dev_put,struct net_device *,if (_T)dev_put (_T))4175 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T))
4176
4177 static inline void netdev_ref_replace(struct net_device *odev,
4178 struct net_device *ndev,
4179 netdevice_tracker *tracker,
4180 gfp_t gfp)
4181 {
4182 if (odev)
4183 netdev_tracker_free(odev, tracker);
4184
4185 __dev_hold(ndev);
4186 __dev_put(odev);
4187
4188 if (ndev)
4189 __netdev_tracker_alloc(ndev, tracker, gfp);
4190 }
4191
4192 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4193 * and _off may be called from IRQ context, but it is caller
4194 * who is responsible for serialization of these calls.
4195 *
4196 * The name carrier is inappropriate, these functions should really be
4197 * called netif_lowerlayer_*() because they represent the state of any
4198 * kind of lower layer not just hardware media.
4199 */
4200 void linkwatch_fire_event(struct net_device *dev);
4201
4202 /**
4203 * linkwatch_sync_dev - sync linkwatch for the given device
4204 * @dev: network device to sync linkwatch for
4205 *
4206 * Sync linkwatch for the given device, removing it from the
4207 * pending work list (if queued).
4208 */
4209 void linkwatch_sync_dev(struct net_device *dev);
4210
4211 /**
4212 * netif_carrier_ok - test if carrier present
4213 * @dev: network device
4214 *
4215 * Check if carrier is present on device
4216 */
netif_carrier_ok(const struct net_device * dev)4217 static inline bool netif_carrier_ok(const struct net_device *dev)
4218 {
4219 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4220 }
4221
4222 unsigned long dev_trans_start(struct net_device *dev);
4223
4224 void __netdev_watchdog_up(struct net_device *dev);
4225
4226 void netif_carrier_on(struct net_device *dev);
4227 void netif_carrier_off(struct net_device *dev);
4228 void netif_carrier_event(struct net_device *dev);
4229
4230 /**
4231 * netif_dormant_on - mark device as dormant.
4232 * @dev: network device
4233 *
4234 * Mark device as dormant (as per RFC2863).
4235 *
4236 * The dormant state indicates that the relevant interface is not
4237 * actually in a condition to pass packets (i.e., it is not 'up') but is
4238 * in a "pending" state, waiting for some external event. For "on-
4239 * demand" interfaces, this new state identifies the situation where the
4240 * interface is waiting for events to place it in the up state.
4241 */
netif_dormant_on(struct net_device * dev)4242 static inline void netif_dormant_on(struct net_device *dev)
4243 {
4244 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4245 linkwatch_fire_event(dev);
4246 }
4247
4248 /**
4249 * netif_dormant_off - set device as not dormant.
4250 * @dev: network device
4251 *
4252 * Device is not in dormant state.
4253 */
netif_dormant_off(struct net_device * dev)4254 static inline void netif_dormant_off(struct net_device *dev)
4255 {
4256 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4257 linkwatch_fire_event(dev);
4258 }
4259
4260 /**
4261 * netif_dormant - test if device is dormant
4262 * @dev: network device
4263 *
4264 * Check if device is dormant.
4265 */
netif_dormant(const struct net_device * dev)4266 static inline bool netif_dormant(const struct net_device *dev)
4267 {
4268 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4269 }
4270
4271
4272 /**
4273 * netif_testing_on - mark device as under test.
4274 * @dev: network device
4275 *
4276 * Mark device as under test (as per RFC2863).
4277 *
4278 * The testing state indicates that some test(s) must be performed on
4279 * the interface. After completion, of the test, the interface state
4280 * will change to up, dormant, or down, as appropriate.
4281 */
netif_testing_on(struct net_device * dev)4282 static inline void netif_testing_on(struct net_device *dev)
4283 {
4284 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4285 linkwatch_fire_event(dev);
4286 }
4287
4288 /**
4289 * netif_testing_off - set device as not under test.
4290 * @dev: network device
4291 *
4292 * Device is not in testing state.
4293 */
netif_testing_off(struct net_device * dev)4294 static inline void netif_testing_off(struct net_device *dev)
4295 {
4296 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4297 linkwatch_fire_event(dev);
4298 }
4299
4300 /**
4301 * netif_testing - test if device is under test
4302 * @dev: network device
4303 *
4304 * Check if device is under test
4305 */
netif_testing(const struct net_device * dev)4306 static inline bool netif_testing(const struct net_device *dev)
4307 {
4308 return test_bit(__LINK_STATE_TESTING, &dev->state);
4309 }
4310
4311
4312 /**
4313 * netif_oper_up - test if device is operational
4314 * @dev: network device
4315 *
4316 * Check if carrier is operational
4317 */
netif_oper_up(const struct net_device * dev)4318 static inline bool netif_oper_up(const struct net_device *dev)
4319 {
4320 unsigned int operstate = READ_ONCE(dev->operstate);
4321
4322 return operstate == IF_OPER_UP ||
4323 operstate == IF_OPER_UNKNOWN /* backward compat */;
4324 }
4325
4326 /**
4327 * netif_device_present - is device available or removed
4328 * @dev: network device
4329 *
4330 * Check if device has not been removed from system.
4331 */
netif_device_present(const struct net_device * dev)4332 static inline bool netif_device_present(const struct net_device *dev)
4333 {
4334 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4335 }
4336
4337 void netif_device_detach(struct net_device *dev);
4338
4339 void netif_device_attach(struct net_device *dev);
4340
4341 /*
4342 * Network interface message level settings
4343 */
4344
4345 enum {
4346 NETIF_MSG_DRV_BIT,
4347 NETIF_MSG_PROBE_BIT,
4348 NETIF_MSG_LINK_BIT,
4349 NETIF_MSG_TIMER_BIT,
4350 NETIF_MSG_IFDOWN_BIT,
4351 NETIF_MSG_IFUP_BIT,
4352 NETIF_MSG_RX_ERR_BIT,
4353 NETIF_MSG_TX_ERR_BIT,
4354 NETIF_MSG_TX_QUEUED_BIT,
4355 NETIF_MSG_INTR_BIT,
4356 NETIF_MSG_TX_DONE_BIT,
4357 NETIF_MSG_RX_STATUS_BIT,
4358 NETIF_MSG_PKTDATA_BIT,
4359 NETIF_MSG_HW_BIT,
4360 NETIF_MSG_WOL_BIT,
4361
4362 /* When you add a new bit above, update netif_msg_class_names array
4363 * in net/ethtool/common.c
4364 */
4365 NETIF_MSG_CLASS_COUNT,
4366 };
4367 /* Both ethtool_ops interface and internal driver implementation use u32 */
4368 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4369
4370 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4371 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4372
4373 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4374 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4375 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4376 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4377 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4378 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4379 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4380 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4381 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4382 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4383 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4384 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4385 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4386 #define NETIF_MSG_HW __NETIF_MSG(HW)
4387 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4388
4389 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4390 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4391 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4392 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4393 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4394 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4395 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4396 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4397 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4398 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4399 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4400 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4401 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4402 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4403 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4404
netif_msg_init(int debug_value,int default_msg_enable_bits)4405 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4406 {
4407 /* use default */
4408 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4409 return default_msg_enable_bits;
4410 if (debug_value == 0) /* no output */
4411 return 0;
4412 /* set low N bits */
4413 return (1U << debug_value) - 1;
4414 }
4415
__netif_tx_lock(struct netdev_queue * txq,int cpu)4416 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4417 {
4418 spin_lock(&txq->_xmit_lock);
4419 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4420 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4421 }
4422
__netif_tx_acquire(struct netdev_queue * txq)4423 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4424 {
4425 __acquire(&txq->_xmit_lock);
4426 return true;
4427 }
4428
__netif_tx_release(struct netdev_queue * txq)4429 static inline void __netif_tx_release(struct netdev_queue *txq)
4430 {
4431 __release(&txq->_xmit_lock);
4432 }
4433
__netif_tx_lock_bh(struct netdev_queue * txq)4434 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4435 {
4436 spin_lock_bh(&txq->_xmit_lock);
4437 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4438 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4439 }
4440
__netif_tx_trylock(struct netdev_queue * txq)4441 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4442 {
4443 bool ok = spin_trylock(&txq->_xmit_lock);
4444
4445 if (likely(ok)) {
4446 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4447 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4448 }
4449 return ok;
4450 }
4451
__netif_tx_unlock(struct netdev_queue * txq)4452 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4453 {
4454 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4455 WRITE_ONCE(txq->xmit_lock_owner, -1);
4456 spin_unlock(&txq->_xmit_lock);
4457 }
4458
__netif_tx_unlock_bh(struct netdev_queue * txq)4459 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4460 {
4461 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4462 WRITE_ONCE(txq->xmit_lock_owner, -1);
4463 spin_unlock_bh(&txq->_xmit_lock);
4464 }
4465
4466 /*
4467 * txq->trans_start can be read locklessly from dev_watchdog()
4468 */
txq_trans_update(struct netdev_queue * txq)4469 static inline void txq_trans_update(struct netdev_queue *txq)
4470 {
4471 if (txq->xmit_lock_owner != -1)
4472 WRITE_ONCE(txq->trans_start, jiffies);
4473 }
4474
txq_trans_cond_update(struct netdev_queue * txq)4475 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4476 {
4477 unsigned long now = jiffies;
4478
4479 if (READ_ONCE(txq->trans_start) != now)
4480 WRITE_ONCE(txq->trans_start, now);
4481 }
4482
4483 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
netif_trans_update(struct net_device * dev)4484 static inline void netif_trans_update(struct net_device *dev)
4485 {
4486 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4487
4488 txq_trans_cond_update(txq);
4489 }
4490
4491 /**
4492 * netif_tx_lock - grab network device transmit lock
4493 * @dev: network device
4494 *
4495 * Get network device transmit lock
4496 */
4497 void netif_tx_lock(struct net_device *dev);
4498
netif_tx_lock_bh(struct net_device * dev)4499 static inline void netif_tx_lock_bh(struct net_device *dev)
4500 {
4501 local_bh_disable();
4502 netif_tx_lock(dev);
4503 }
4504
4505 void netif_tx_unlock(struct net_device *dev);
4506
netif_tx_unlock_bh(struct net_device * dev)4507 static inline void netif_tx_unlock_bh(struct net_device *dev)
4508 {
4509 netif_tx_unlock(dev);
4510 local_bh_enable();
4511 }
4512
4513 #define HARD_TX_LOCK(dev, txq, cpu) { \
4514 if (!(dev)->lltx) { \
4515 __netif_tx_lock(txq, cpu); \
4516 } else { \
4517 __netif_tx_acquire(txq); \
4518 } \
4519 }
4520
4521 #define HARD_TX_TRYLOCK(dev, txq) \
4522 (!(dev)->lltx ? \
4523 __netif_tx_trylock(txq) : \
4524 __netif_tx_acquire(txq))
4525
4526 #define HARD_TX_UNLOCK(dev, txq) { \
4527 if (!(dev)->lltx) { \
4528 __netif_tx_unlock(txq); \
4529 } else { \
4530 __netif_tx_release(txq); \
4531 } \
4532 }
4533
netif_tx_disable(struct net_device * dev)4534 static inline void netif_tx_disable(struct net_device *dev)
4535 {
4536 unsigned int i;
4537 int cpu;
4538
4539 local_bh_disable();
4540 cpu = smp_processor_id();
4541 spin_lock(&dev->tx_global_lock);
4542 for (i = 0; i < dev->num_tx_queues; i++) {
4543 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4544
4545 __netif_tx_lock(txq, cpu);
4546 netif_tx_stop_queue(txq);
4547 __netif_tx_unlock(txq);
4548 }
4549 spin_unlock(&dev->tx_global_lock);
4550 local_bh_enable();
4551 }
4552
netif_addr_lock(struct net_device * dev)4553 static inline void netif_addr_lock(struct net_device *dev)
4554 {
4555 unsigned char nest_level = 0;
4556
4557 #ifdef CONFIG_LOCKDEP
4558 nest_level = dev->nested_level;
4559 #endif
4560 spin_lock_nested(&dev->addr_list_lock, nest_level);
4561 }
4562
netif_addr_lock_bh(struct net_device * dev)4563 static inline void netif_addr_lock_bh(struct net_device *dev)
4564 {
4565 unsigned char nest_level = 0;
4566
4567 #ifdef CONFIG_LOCKDEP
4568 nest_level = dev->nested_level;
4569 #endif
4570 local_bh_disable();
4571 spin_lock_nested(&dev->addr_list_lock, nest_level);
4572 }
4573
netif_addr_unlock(struct net_device * dev)4574 static inline void netif_addr_unlock(struct net_device *dev)
4575 {
4576 spin_unlock(&dev->addr_list_lock);
4577 }
4578
netif_addr_unlock_bh(struct net_device * dev)4579 static inline void netif_addr_unlock_bh(struct net_device *dev)
4580 {
4581 spin_unlock_bh(&dev->addr_list_lock);
4582 }
4583
4584 /*
4585 * dev_addrs walker. Should be used only for read access. Call with
4586 * rcu_read_lock held.
4587 */
4588 #define for_each_dev_addr(dev, ha) \
4589 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4590
4591 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4592
4593 void ether_setup(struct net_device *dev);
4594
4595 /* Allocate dummy net_device */
4596 struct net_device *alloc_netdev_dummy(int sizeof_priv);
4597
4598 /* Support for loadable net-drivers */
4599 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4600 unsigned char name_assign_type,
4601 void (*setup)(struct net_device *),
4602 unsigned int txqs, unsigned int rxqs);
4603 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4604 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4605
4606 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4607 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4608 count)
4609
4610 int register_netdev(struct net_device *dev);
4611 void unregister_netdev(struct net_device *dev);
4612
4613 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4614
4615 /* General hardware address lists handling functions */
4616 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4617 struct netdev_hw_addr_list *from_list, int addr_len);
4618 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4619 struct netdev_hw_addr_list *from_list, int addr_len);
4620 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4621 struct net_device *dev,
4622 int (*sync)(struct net_device *, const unsigned char *),
4623 int (*unsync)(struct net_device *,
4624 const unsigned char *));
4625 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4626 struct net_device *dev,
4627 int (*sync)(struct net_device *,
4628 const unsigned char *, int),
4629 int (*unsync)(struct net_device *,
4630 const unsigned char *, int));
4631 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4632 struct net_device *dev,
4633 int (*unsync)(struct net_device *,
4634 const unsigned char *, int));
4635 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4636 struct net_device *dev,
4637 int (*unsync)(struct net_device *,
4638 const unsigned char *));
4639 void __hw_addr_init(struct netdev_hw_addr_list *list);
4640
4641 /* Functions used for device addresses handling */
4642 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4643 const void *addr, size_t len);
4644
4645 static inline void
__dev_addr_set(struct net_device * dev,const void * addr,size_t len)4646 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4647 {
4648 dev_addr_mod(dev, 0, addr, len);
4649 }
4650
dev_addr_set(struct net_device * dev,const u8 * addr)4651 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4652 {
4653 __dev_addr_set(dev, addr, dev->addr_len);
4654 }
4655
4656 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4657 unsigned char addr_type);
4658 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4659 unsigned char addr_type);
4660
4661 /* Functions used for unicast addresses handling */
4662 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4663 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4664 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4665 int dev_uc_sync(struct net_device *to, struct net_device *from);
4666 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4667 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4668 void dev_uc_flush(struct net_device *dev);
4669 void dev_uc_init(struct net_device *dev);
4670
4671 /**
4672 * __dev_uc_sync - Synchronize device's unicast list
4673 * @dev: device to sync
4674 * @sync: function to call if address should be added
4675 * @unsync: function to call if address should be removed
4676 *
4677 * Add newly added addresses to the interface, and release
4678 * addresses that have been deleted.
4679 */
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4680 static inline int __dev_uc_sync(struct net_device *dev,
4681 int (*sync)(struct net_device *,
4682 const unsigned char *),
4683 int (*unsync)(struct net_device *,
4684 const unsigned char *))
4685 {
4686 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4687 }
4688
4689 /**
4690 * __dev_uc_unsync - Remove synchronized addresses from device
4691 * @dev: device to sync
4692 * @unsync: function to call if address should be removed
4693 *
4694 * Remove all addresses that were added to the device by dev_uc_sync().
4695 */
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4696 static inline void __dev_uc_unsync(struct net_device *dev,
4697 int (*unsync)(struct net_device *,
4698 const unsigned char *))
4699 {
4700 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4701 }
4702
4703 /* Functions used for multicast addresses handling */
4704 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4705 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4706 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4707 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4708 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4709 int dev_mc_sync(struct net_device *to, struct net_device *from);
4710 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4711 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4712 void dev_mc_flush(struct net_device *dev);
4713 void dev_mc_init(struct net_device *dev);
4714
4715 /**
4716 * __dev_mc_sync - Synchronize device's multicast list
4717 * @dev: device to sync
4718 * @sync: function to call if address should be added
4719 * @unsync: function to call if address should be removed
4720 *
4721 * Add newly added addresses to the interface, and release
4722 * addresses that have been deleted.
4723 */
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4724 static inline int __dev_mc_sync(struct net_device *dev,
4725 int (*sync)(struct net_device *,
4726 const unsigned char *),
4727 int (*unsync)(struct net_device *,
4728 const unsigned char *))
4729 {
4730 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4731 }
4732
4733 /**
4734 * __dev_mc_unsync - Remove synchronized addresses from device
4735 * @dev: device to sync
4736 * @unsync: function to call if address should be removed
4737 *
4738 * Remove all addresses that were added to the device by dev_mc_sync().
4739 */
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4740 static inline void __dev_mc_unsync(struct net_device *dev,
4741 int (*unsync)(struct net_device *,
4742 const unsigned char *))
4743 {
4744 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4745 }
4746
4747 /* Functions used for secondary unicast and multicast support */
4748 void dev_set_rx_mode(struct net_device *dev);
4749 int dev_set_promiscuity(struct net_device *dev, int inc);
4750 int dev_set_allmulti(struct net_device *dev, int inc);
4751 void netdev_state_change(struct net_device *dev);
4752 void __netdev_notify_peers(struct net_device *dev);
4753 void netdev_notify_peers(struct net_device *dev);
4754 void netdev_features_change(struct net_device *dev);
4755 /* Load a device via the kmod */
4756 void dev_load(struct net *net, const char *name);
4757 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4758 struct rtnl_link_stats64 *storage);
4759 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4760 const struct net_device_stats *netdev_stats);
4761 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4762 const struct pcpu_sw_netstats __percpu *netstats);
4763 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4764
4765 enum {
4766 NESTED_SYNC_IMM_BIT,
4767 NESTED_SYNC_TODO_BIT,
4768 };
4769
4770 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4771 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4772
4773 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4774 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4775
4776 struct netdev_nested_priv {
4777 unsigned char flags;
4778 void *data;
4779 };
4780
4781 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4782 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4783 struct list_head **iter);
4784
4785 /* iterate through upper list, must be called under RCU read lock */
4786 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4787 for (iter = &(dev)->adj_list.upper, \
4788 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4789 updev; \
4790 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4791
4792 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4793 int (*fn)(struct net_device *upper_dev,
4794 struct netdev_nested_priv *priv),
4795 struct netdev_nested_priv *priv);
4796
4797 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4798 struct net_device *upper_dev);
4799
4800 bool netdev_has_any_upper_dev(struct net_device *dev);
4801
4802 void *netdev_lower_get_next_private(struct net_device *dev,
4803 struct list_head **iter);
4804 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4805 struct list_head **iter);
4806
4807 #define netdev_for_each_lower_private(dev, priv, iter) \
4808 for (iter = (dev)->adj_list.lower.next, \
4809 priv = netdev_lower_get_next_private(dev, &(iter)); \
4810 priv; \
4811 priv = netdev_lower_get_next_private(dev, &(iter)))
4812
4813 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4814 for (iter = &(dev)->adj_list.lower, \
4815 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4816 priv; \
4817 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4818
4819 void *netdev_lower_get_next(struct net_device *dev,
4820 struct list_head **iter);
4821
4822 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4823 for (iter = (dev)->adj_list.lower.next, \
4824 ldev = netdev_lower_get_next(dev, &(iter)); \
4825 ldev; \
4826 ldev = netdev_lower_get_next(dev, &(iter)))
4827
4828 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4829 struct list_head **iter);
4830 int netdev_walk_all_lower_dev(struct net_device *dev,
4831 int (*fn)(struct net_device *lower_dev,
4832 struct netdev_nested_priv *priv),
4833 struct netdev_nested_priv *priv);
4834 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4835 int (*fn)(struct net_device *lower_dev,
4836 struct netdev_nested_priv *priv),
4837 struct netdev_nested_priv *priv);
4838
4839 void *netdev_adjacent_get_private(struct list_head *adj_list);
4840 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4841 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4842 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4843 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4844 struct netlink_ext_ack *extack);
4845 int netdev_master_upper_dev_link(struct net_device *dev,
4846 struct net_device *upper_dev,
4847 void *upper_priv, void *upper_info,
4848 struct netlink_ext_ack *extack);
4849 void netdev_upper_dev_unlink(struct net_device *dev,
4850 struct net_device *upper_dev);
4851 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4852 struct net_device *new_dev,
4853 struct net_device *dev,
4854 struct netlink_ext_ack *extack);
4855 void netdev_adjacent_change_commit(struct net_device *old_dev,
4856 struct net_device *new_dev,
4857 struct net_device *dev);
4858 void netdev_adjacent_change_abort(struct net_device *old_dev,
4859 struct net_device *new_dev,
4860 struct net_device *dev);
4861 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4862 void *netdev_lower_dev_get_private(struct net_device *dev,
4863 struct net_device *lower_dev);
4864 void netdev_lower_state_changed(struct net_device *lower_dev,
4865 void *lower_state_info);
4866
4867 /* RSS keys are 40 or 52 bytes long */
4868 #define NETDEV_RSS_KEY_LEN 52
4869 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4870 void netdev_rss_key_fill(void *buffer, size_t len);
4871
4872 int skb_checksum_help(struct sk_buff *skb);
4873 int skb_crc32c_csum_help(struct sk_buff *skb);
4874 int skb_csum_hwoffload_help(struct sk_buff *skb,
4875 const netdev_features_t features);
4876
4877 struct netdev_bonding_info {
4878 ifslave slave;
4879 ifbond master;
4880 };
4881
4882 struct netdev_notifier_bonding_info {
4883 struct netdev_notifier_info info; /* must be first */
4884 struct netdev_bonding_info bonding_info;
4885 };
4886
4887 void netdev_bonding_info_change(struct net_device *dev,
4888 struct netdev_bonding_info *bonding_info);
4889
4890 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4891 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4892 #else
ethtool_notify(struct net_device * dev,unsigned int cmd,const void * data)4893 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4894 const void *data)
4895 {
4896 }
4897 #endif
4898
4899 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4900
can_checksum_protocol(netdev_features_t features,__be16 protocol)4901 static inline bool can_checksum_protocol(netdev_features_t features,
4902 __be16 protocol)
4903 {
4904 if (protocol == htons(ETH_P_FCOE))
4905 return !!(features & NETIF_F_FCOE_CRC);
4906
4907 /* Assume this is an IP checksum (not SCTP CRC) */
4908
4909 if (features & NETIF_F_HW_CSUM) {
4910 /* Can checksum everything */
4911 return true;
4912 }
4913
4914 switch (protocol) {
4915 case htons(ETH_P_IP):
4916 return !!(features & NETIF_F_IP_CSUM);
4917 case htons(ETH_P_IPV6):
4918 return !!(features & NETIF_F_IPV6_CSUM);
4919 default:
4920 return false;
4921 }
4922 }
4923
4924 #ifdef CONFIG_BUG
4925 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4926 #else
netdev_rx_csum_fault(struct net_device * dev,struct sk_buff * skb)4927 static inline void netdev_rx_csum_fault(struct net_device *dev,
4928 struct sk_buff *skb)
4929 {
4930 }
4931 #endif
4932 /* rx skb timestamps */
4933 void net_enable_timestamp(void);
4934 void net_disable_timestamp(void);
4935
netdev_get_tstamp(struct net_device * dev,const struct skb_shared_hwtstamps * hwtstamps,bool cycles)4936 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4937 const struct skb_shared_hwtstamps *hwtstamps,
4938 bool cycles)
4939 {
4940 const struct net_device_ops *ops = dev->netdev_ops;
4941
4942 if (ops->ndo_get_tstamp)
4943 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4944
4945 return hwtstamps->hwtstamp;
4946 }
4947
4948 #ifndef CONFIG_PREEMPT_RT
netdev_xmit_set_more(bool more)4949 static inline void netdev_xmit_set_more(bool more)
4950 {
4951 __this_cpu_write(softnet_data.xmit.more, more);
4952 }
4953
netdev_xmit_more(void)4954 static inline bool netdev_xmit_more(void)
4955 {
4956 return __this_cpu_read(softnet_data.xmit.more);
4957 }
4958 #else
netdev_xmit_set_more(bool more)4959 static inline void netdev_xmit_set_more(bool more)
4960 {
4961 current->net_xmit.more = more;
4962 }
4963
netdev_xmit_more(void)4964 static inline bool netdev_xmit_more(void)
4965 {
4966 return current->net_xmit.more;
4967 }
4968 #endif
4969
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)4970 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4971 struct sk_buff *skb, struct net_device *dev,
4972 bool more)
4973 {
4974 netdev_xmit_set_more(more);
4975 return ops->ndo_start_xmit(skb, dev);
4976 }
4977
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)4978 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4979 struct netdev_queue *txq, bool more)
4980 {
4981 const struct net_device_ops *ops = dev->netdev_ops;
4982 netdev_tx_t rc;
4983
4984 rc = __netdev_start_xmit(ops, skb, dev, more);
4985 if (rc == NETDEV_TX_OK)
4986 txq_trans_update(txq);
4987
4988 return rc;
4989 }
4990
4991 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4992 const void *ns);
4993 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4994 const void *ns);
4995
4996 extern const struct kobj_ns_type_operations net_ns_type_operations;
4997
4998 const char *netdev_drivername(const struct net_device *dev);
4999
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)5000 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
5001 netdev_features_t f2)
5002 {
5003 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
5004 if (f1 & NETIF_F_HW_CSUM)
5005 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5006 else
5007 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5008 }
5009
5010 return f1 & f2;
5011 }
5012
netdev_get_wanted_features(struct net_device * dev)5013 static inline netdev_features_t netdev_get_wanted_features(
5014 struct net_device *dev)
5015 {
5016 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5017 }
5018 netdev_features_t netdev_increment_features(netdev_features_t all,
5019 netdev_features_t one, netdev_features_t mask);
5020
5021 /* Allow TSO being used on stacked device :
5022 * Performing the GSO segmentation before last device
5023 * is a performance improvement.
5024 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)5025 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5026 netdev_features_t mask)
5027 {
5028 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5029 }
5030
5031 int __netdev_update_features(struct net_device *dev);
5032 void netdev_update_features(struct net_device *dev);
5033 void netdev_change_features(struct net_device *dev);
5034
5035 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5036 struct net_device *dev);
5037
5038 netdev_features_t passthru_features_check(struct sk_buff *skb,
5039 struct net_device *dev,
5040 netdev_features_t features);
5041 netdev_features_t netif_skb_features(struct sk_buff *skb);
5042 void skb_warn_bad_offload(const struct sk_buff *skb);
5043
net_gso_ok(netdev_features_t features,int gso_type)5044 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5045 {
5046 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5047
5048 /* check flags correspondence */
5049 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5050 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5051 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5052 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5053 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5054 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5055 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5056 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5057 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5058 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5059 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5060 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5061 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5062 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5063 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5064 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5065 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5066 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5067 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5068
5069 return (features & feature) == feature;
5070 }
5071
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)5072 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5073 {
5074 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5075 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5076 }
5077
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)5078 static inline bool netif_needs_gso(struct sk_buff *skb,
5079 netdev_features_t features)
5080 {
5081 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5082 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5083 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5084 }
5085
5086 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5087 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5088 void netif_inherit_tso_max(struct net_device *to,
5089 const struct net_device *from);
5090
5091 static inline unsigned int
netif_get_gro_max_size(const struct net_device * dev,const struct sk_buff * skb)5092 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb)
5093 {
5094 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */
5095 return skb->protocol == htons(ETH_P_IPV6) ?
5096 READ_ONCE(dev->gro_max_size) :
5097 READ_ONCE(dev->gro_ipv4_max_size);
5098 }
5099
5100 static inline unsigned int
netif_get_gso_max_size(const struct net_device * dev,const struct sk_buff * skb)5101 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb)
5102 {
5103 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */
5104 return skb->protocol == htons(ETH_P_IPV6) ?
5105 READ_ONCE(dev->gso_max_size) :
5106 READ_ONCE(dev->gso_ipv4_max_size);
5107 }
5108
netif_is_macsec(const struct net_device * dev)5109 static inline bool netif_is_macsec(const struct net_device *dev)
5110 {
5111 return dev->priv_flags & IFF_MACSEC;
5112 }
5113
netif_is_macvlan(const struct net_device * dev)5114 static inline bool netif_is_macvlan(const struct net_device *dev)
5115 {
5116 return dev->priv_flags & IFF_MACVLAN;
5117 }
5118
netif_is_macvlan_port(const struct net_device * dev)5119 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5120 {
5121 return dev->priv_flags & IFF_MACVLAN_PORT;
5122 }
5123
netif_is_bond_master(const struct net_device * dev)5124 static inline bool netif_is_bond_master(const struct net_device *dev)
5125 {
5126 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5127 }
5128
netif_is_bond_slave(const struct net_device * dev)5129 static inline bool netif_is_bond_slave(const struct net_device *dev)
5130 {
5131 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5132 }
5133
netif_supports_nofcs(struct net_device * dev)5134 static inline bool netif_supports_nofcs(struct net_device *dev)
5135 {
5136 return dev->priv_flags & IFF_SUPP_NOFCS;
5137 }
5138
netif_has_l3_rx_handler(const struct net_device * dev)5139 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5140 {
5141 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5142 }
5143
netif_is_l3_master(const struct net_device * dev)5144 static inline bool netif_is_l3_master(const struct net_device *dev)
5145 {
5146 return dev->priv_flags & IFF_L3MDEV_MASTER;
5147 }
5148
netif_is_l3_slave(const struct net_device * dev)5149 static inline bool netif_is_l3_slave(const struct net_device *dev)
5150 {
5151 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5152 }
5153
dev_sdif(const struct net_device * dev)5154 static inline int dev_sdif(const struct net_device *dev)
5155 {
5156 #ifdef CONFIG_NET_L3_MASTER_DEV
5157 if (netif_is_l3_slave(dev))
5158 return dev->ifindex;
5159 #endif
5160 return 0;
5161 }
5162
netif_is_bridge_master(const struct net_device * dev)5163 static inline bool netif_is_bridge_master(const struct net_device *dev)
5164 {
5165 return dev->priv_flags & IFF_EBRIDGE;
5166 }
5167
netif_is_bridge_port(const struct net_device * dev)5168 static inline bool netif_is_bridge_port(const struct net_device *dev)
5169 {
5170 return dev->priv_flags & IFF_BRIDGE_PORT;
5171 }
5172
netif_is_ovs_master(const struct net_device * dev)5173 static inline bool netif_is_ovs_master(const struct net_device *dev)
5174 {
5175 return dev->priv_flags & IFF_OPENVSWITCH;
5176 }
5177
netif_is_ovs_port(const struct net_device * dev)5178 static inline bool netif_is_ovs_port(const struct net_device *dev)
5179 {
5180 return dev->priv_flags & IFF_OVS_DATAPATH;
5181 }
5182
netif_is_any_bridge_master(const struct net_device * dev)5183 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5184 {
5185 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5186 }
5187
netif_is_any_bridge_port(const struct net_device * dev)5188 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5189 {
5190 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5191 }
5192
netif_is_team_master(const struct net_device * dev)5193 static inline bool netif_is_team_master(const struct net_device *dev)
5194 {
5195 return dev->priv_flags & IFF_TEAM;
5196 }
5197
netif_is_team_port(const struct net_device * dev)5198 static inline bool netif_is_team_port(const struct net_device *dev)
5199 {
5200 return dev->priv_flags & IFF_TEAM_PORT;
5201 }
5202
netif_is_lag_master(const struct net_device * dev)5203 static inline bool netif_is_lag_master(const struct net_device *dev)
5204 {
5205 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5206 }
5207
netif_is_lag_port(const struct net_device * dev)5208 static inline bool netif_is_lag_port(const struct net_device *dev)
5209 {
5210 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5211 }
5212
netif_is_rxfh_configured(const struct net_device * dev)5213 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5214 {
5215 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5216 }
5217
netif_is_failover(const struct net_device * dev)5218 static inline bool netif_is_failover(const struct net_device *dev)
5219 {
5220 return dev->priv_flags & IFF_FAILOVER;
5221 }
5222
netif_is_failover_slave(const struct net_device * dev)5223 static inline bool netif_is_failover_slave(const struct net_device *dev)
5224 {
5225 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5226 }
5227
5228 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)5229 static inline void netif_keep_dst(struct net_device *dev)
5230 {
5231 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5232 }
5233
5234 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
netif_reduces_vlan_mtu(struct net_device * dev)5235 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5236 {
5237 /* TODO: reserve and use an additional IFF bit, if we get more users */
5238 return netif_is_macsec(dev);
5239 }
5240
5241 extern struct pernet_operations __net_initdata loopback_net_ops;
5242
5243 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5244
5245 /* netdev_printk helpers, similar to dev_printk */
5246
netdev_name(const struct net_device * dev)5247 static inline const char *netdev_name(const struct net_device *dev)
5248 {
5249 if (!dev->name[0] || strchr(dev->name, '%'))
5250 return "(unnamed net_device)";
5251 return dev->name;
5252 }
5253
netdev_reg_state(const struct net_device * dev)5254 static inline const char *netdev_reg_state(const struct net_device *dev)
5255 {
5256 u8 reg_state = READ_ONCE(dev->reg_state);
5257
5258 switch (reg_state) {
5259 case NETREG_UNINITIALIZED: return " (uninitialized)";
5260 case NETREG_REGISTERED: return "";
5261 case NETREG_UNREGISTERING: return " (unregistering)";
5262 case NETREG_UNREGISTERED: return " (unregistered)";
5263 case NETREG_RELEASED: return " (released)";
5264 case NETREG_DUMMY: return " (dummy)";
5265 }
5266
5267 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state);
5268 return " (unknown)";
5269 }
5270
5271 #define MODULE_ALIAS_NETDEV(device) \
5272 MODULE_ALIAS("netdev-" device)
5273
5274 /*
5275 * netdev_WARN() acts like dev_printk(), but with the key difference
5276 * of using a WARN/WARN_ON to get the message out, including the
5277 * file/line information and a backtrace.
5278 */
5279 #define netdev_WARN(dev, format, args...) \
5280 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5281 netdev_reg_state(dev), ##args)
5282
5283 #define netdev_WARN_ONCE(dev, format, args...) \
5284 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5285 netdev_reg_state(dev), ##args)
5286
5287 /*
5288 * The list of packet types we will receive (as opposed to discard)
5289 * and the routines to invoke.
5290 *
5291 * Why 16. Because with 16 the only overlap we get on a hash of the
5292 * low nibble of the protocol value is RARP/SNAP/X.25.
5293 *
5294 * 0800 IP
5295 * 0001 802.3
5296 * 0002 AX.25
5297 * 0004 802.2
5298 * 8035 RARP
5299 * 0005 SNAP
5300 * 0805 X.25
5301 * 0806 ARP
5302 * 8137 IPX
5303 * 0009 Localtalk
5304 * 86DD IPv6
5305 */
5306 #define PTYPE_HASH_SIZE (16)
5307 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5308
5309 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5310
5311 extern struct net_device *blackhole_netdev;
5312
5313 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5314 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5315 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5316 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5317 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5318
5319 #endif /* _LINUX_NETDEVICE_H */
5320