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 <net/neighbour_tables.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
321 /* These flag bits are private to the generic network queueing
322 * layer; they may not be explicitly referenced by any other
323 * code.
324 */
325
326 enum netdev_state_t {
327 __LINK_STATE_START,
328 __LINK_STATE_PRESENT,
329 __LINK_STATE_NOCARRIER,
330 __LINK_STATE_LINKWATCH_PENDING,
331 __LINK_STATE_DORMANT,
332 __LINK_STATE_TESTING,
333 };
334
335 struct gro_list {
336 struct list_head list;
337 int count;
338 };
339
340 /*
341 * size of gro hash buckets, must less than bit number of
342 * napi_struct::gro_bitmask
343 */
344 #define GRO_HASH_BUCKETS 8
345
346 /*
347 * Structure for per-NAPI config
348 */
349 struct napi_config {
350 u64 gro_flush_timeout;
351 u64 irq_suspend_timeout;
352 u32 defer_hard_irqs;
353 unsigned int napi_id;
354 };
355
356 /*
357 * Structure for NAPI scheduling similar to tasklet but with weighting
358 */
359 struct napi_struct {
360 /* The poll_list must only be managed by the entity which
361 * changes the state of the NAPI_STATE_SCHED bit. This means
362 * whoever atomically sets that bit can add this napi_struct
363 * to the per-CPU poll_list, and whoever clears that bit
364 * can remove from the list right before clearing the bit.
365 */
366 struct list_head poll_list;
367
368 unsigned long state;
369 int weight;
370 u32 defer_hard_irqs_count;
371 unsigned long gro_bitmask;
372 int (*poll)(struct napi_struct *, int);
373 #ifdef CONFIG_NETPOLL
374 /* CPU actively polling if netpoll is configured */
375 int poll_owner;
376 #endif
377 /* CPU on which NAPI has been scheduled for processing */
378 int list_owner;
379 struct net_device *dev;
380 struct gro_list gro_hash[GRO_HASH_BUCKETS];
381 struct sk_buff *skb;
382 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
383 int rx_count; /* length of rx_list */
384 unsigned int napi_id;
385 struct hrtimer timer;
386 struct task_struct *thread;
387 unsigned long gro_flush_timeout;
388 unsigned long irq_suspend_timeout;
389 u32 defer_hard_irqs;
390 /* control-path-only fields follow */
391 struct list_head dev_list;
392 struct hlist_node napi_hash_node;
393 int irq;
394 int index;
395 struct napi_config *config;
396 };
397
398 enum {
399 NAPI_STATE_SCHED, /* Poll is scheduled */
400 NAPI_STATE_MISSED, /* reschedule a napi */
401 NAPI_STATE_DISABLE, /* Disable pending */
402 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
403 NAPI_STATE_LISTED, /* NAPI added to system lists */
404 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
405 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
406 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
407 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
408 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
409 };
410
411 enum {
412 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
413 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
414 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
415 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
416 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
417 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
418 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
419 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
420 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
421 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
422 };
423
424 enum gro_result {
425 GRO_MERGED,
426 GRO_MERGED_FREE,
427 GRO_HELD,
428 GRO_NORMAL,
429 GRO_CONSUMED,
430 };
431 typedef enum gro_result gro_result_t;
432
433 /*
434 * enum rx_handler_result - Possible return values for rx_handlers.
435 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
436 * further.
437 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
438 * case skb->dev was changed by rx_handler.
439 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
440 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
441 *
442 * rx_handlers are functions called from inside __netif_receive_skb(), to do
443 * special processing of the skb, prior to delivery to protocol handlers.
444 *
445 * Currently, a net_device can only have a single rx_handler registered. Trying
446 * to register a second rx_handler will return -EBUSY.
447 *
448 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
449 * To unregister a rx_handler on a net_device, use
450 * netdev_rx_handler_unregister().
451 *
452 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
453 * do with the skb.
454 *
455 * If the rx_handler consumed the skb in some way, it should return
456 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
457 * the skb to be delivered in some other way.
458 *
459 * If the rx_handler changed skb->dev, to divert the skb to another
460 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
461 * new device will be called if it exists.
462 *
463 * If the rx_handler decides the skb should be ignored, it should return
464 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
465 * are registered on exact device (ptype->dev == skb->dev).
466 *
467 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
468 * delivered, it should return RX_HANDLER_PASS.
469 *
470 * A device without a registered rx_handler will behave as if rx_handler
471 * returned RX_HANDLER_PASS.
472 */
473
474 enum rx_handler_result {
475 RX_HANDLER_CONSUMED,
476 RX_HANDLER_ANOTHER,
477 RX_HANDLER_EXACT,
478 RX_HANDLER_PASS,
479 };
480 typedef enum rx_handler_result rx_handler_result_t;
481 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
482
483 void __napi_schedule(struct napi_struct *n);
484 void __napi_schedule_irqoff(struct napi_struct *n);
485
napi_disable_pending(struct napi_struct * n)486 static inline bool napi_disable_pending(struct napi_struct *n)
487 {
488 return test_bit(NAPI_STATE_DISABLE, &n->state);
489 }
490
napi_prefer_busy_poll(struct napi_struct * n)491 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
492 {
493 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
494 }
495
496 /**
497 * napi_is_scheduled - test if NAPI is scheduled
498 * @n: NAPI context
499 *
500 * This check is "best-effort". With no locking implemented,
501 * a NAPI can be scheduled or terminate right after this check
502 * and produce not precise results.
503 *
504 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
505 * should not be used normally and napi_schedule should be
506 * used instead.
507 *
508 * Use only if the driver really needs to check if a NAPI
509 * is scheduled for example in the context of delayed timer
510 * that can be skipped if a NAPI is already scheduled.
511 *
512 * Return True if NAPI is scheduled, False otherwise.
513 */
napi_is_scheduled(struct napi_struct * n)514 static inline bool napi_is_scheduled(struct napi_struct *n)
515 {
516 return test_bit(NAPI_STATE_SCHED, &n->state);
517 }
518
519 bool napi_schedule_prep(struct napi_struct *n);
520
521 /**
522 * napi_schedule - schedule NAPI poll
523 * @n: NAPI context
524 *
525 * Schedule NAPI poll routine to be called if it is not already
526 * running.
527 * Return true if we schedule a NAPI or false if not.
528 * Refer to napi_schedule_prep() for additional reason on why
529 * a NAPI might not be scheduled.
530 */
napi_schedule(struct napi_struct * n)531 static inline bool napi_schedule(struct napi_struct *n)
532 {
533 if (napi_schedule_prep(n)) {
534 __napi_schedule(n);
535 return true;
536 }
537
538 return false;
539 }
540
541 /**
542 * napi_schedule_irqoff - schedule NAPI poll
543 * @n: NAPI context
544 *
545 * Variant of napi_schedule(), assuming hard irqs are masked.
546 */
napi_schedule_irqoff(struct napi_struct * n)547 static inline void napi_schedule_irqoff(struct napi_struct *n)
548 {
549 if (napi_schedule_prep(n))
550 __napi_schedule_irqoff(n);
551 }
552
553 /**
554 * napi_complete_done - NAPI processing complete
555 * @n: NAPI context
556 * @work_done: number of packets processed
557 *
558 * Mark NAPI processing as complete. Should only be called if poll budget
559 * has not been completely consumed.
560 * Prefer over napi_complete().
561 * Return false if device should avoid rearming interrupts.
562 */
563 bool napi_complete_done(struct napi_struct *n, int work_done);
564
napi_complete(struct napi_struct * n)565 static inline bool napi_complete(struct napi_struct *n)
566 {
567 return napi_complete_done(n, 0);
568 }
569
570 int dev_set_threaded(struct net_device *dev, bool threaded);
571
572 /**
573 * napi_disable - prevent NAPI from scheduling
574 * @n: NAPI context
575 *
576 * Stop NAPI from being scheduled on this context.
577 * Waits till any outstanding processing completes.
578 */
579 void napi_disable(struct napi_struct *n);
580
581 void napi_enable(struct napi_struct *n);
582
583 /**
584 * napi_synchronize - wait until NAPI is not running
585 * @n: NAPI context
586 *
587 * Wait until NAPI is done being scheduled on this context.
588 * Waits till any outstanding processing completes but
589 * does not disable future activations.
590 */
napi_synchronize(const struct napi_struct * n)591 static inline void napi_synchronize(const struct napi_struct *n)
592 {
593 if (IS_ENABLED(CONFIG_SMP))
594 while (test_bit(NAPI_STATE_SCHED, &n->state))
595 msleep(1);
596 else
597 barrier();
598 }
599
600 /**
601 * napi_if_scheduled_mark_missed - if napi is running, set the
602 * NAPIF_STATE_MISSED
603 * @n: NAPI context
604 *
605 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
606 * NAPI is scheduled.
607 **/
napi_if_scheduled_mark_missed(struct napi_struct * n)608 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
609 {
610 unsigned long val, new;
611
612 val = READ_ONCE(n->state);
613 do {
614 if (val & NAPIF_STATE_DISABLE)
615 return true;
616
617 if (!(val & NAPIF_STATE_SCHED))
618 return false;
619
620 new = val | NAPIF_STATE_MISSED;
621 } while (!try_cmpxchg(&n->state, &val, new));
622
623 return true;
624 }
625
626 enum netdev_queue_state_t {
627 __QUEUE_STATE_DRV_XOFF,
628 __QUEUE_STATE_STACK_XOFF,
629 __QUEUE_STATE_FROZEN,
630 };
631
632 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
633 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
634 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
635
636 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
637 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
638 QUEUE_STATE_FROZEN)
639 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
640 QUEUE_STATE_FROZEN)
641
642 /*
643 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
644 * netif_tx_* functions below are used to manipulate this flag. The
645 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
646 * queue independently. The netif_xmit_*stopped functions below are called
647 * to check if the queue has been stopped by the driver or stack (either
648 * of the XOFF bits are set in the state). Drivers should not need to call
649 * netif_xmit*stopped functions, they should only be using netif_tx_*.
650 */
651
652 struct netdev_queue {
653 /*
654 * read-mostly part
655 */
656 struct net_device *dev;
657 netdevice_tracker dev_tracker;
658
659 struct Qdisc __rcu *qdisc;
660 struct Qdisc __rcu *qdisc_sleeping;
661 #ifdef CONFIG_SYSFS
662 struct kobject kobj;
663 #endif
664 unsigned long tx_maxrate;
665 /*
666 * Number of TX timeouts for this queue
667 * (/sys/class/net/DEV/Q/trans_timeout)
668 */
669 atomic_long_t trans_timeout;
670
671 /* Subordinate device that the queue has been assigned to */
672 struct net_device *sb_dev;
673 #ifdef CONFIG_XDP_SOCKETS
674 struct xsk_buff_pool *pool;
675 #endif
676
677 /*
678 * write-mostly part
679 */
680 #ifdef CONFIG_BQL
681 struct dql dql;
682 #endif
683 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
684 int xmit_lock_owner;
685 /*
686 * Time (in jiffies) of last Tx
687 */
688 unsigned long trans_start;
689
690 unsigned long state;
691
692 /*
693 * slow- / control-path part
694 */
695 /* NAPI instance for the queue
696 * Readers and writers must hold RTNL
697 */
698 struct napi_struct *napi;
699
700 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
701 int numa_node;
702 #endif
703 } ____cacheline_aligned_in_smp;
704
705 extern int sysctl_fb_tunnels_only_for_init_net;
706 extern int sysctl_devconf_inherit_init_net;
707
708 /*
709 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
710 * == 1 : For initns only
711 * == 2 : For none.
712 */
net_has_fallback_tunnels(const struct net * net)713 static inline bool net_has_fallback_tunnels(const struct net *net)
714 {
715 #if IS_ENABLED(CONFIG_SYSCTL)
716 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
717
718 return !fb_tunnels_only_for_init_net ||
719 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
720 #else
721 return true;
722 #endif
723 }
724
net_inherit_devconf(void)725 static inline int net_inherit_devconf(void)
726 {
727 #if IS_ENABLED(CONFIG_SYSCTL)
728 return READ_ONCE(sysctl_devconf_inherit_init_net);
729 #else
730 return 0;
731 #endif
732 }
733
netdev_queue_numa_node_read(const struct netdev_queue * q)734 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
735 {
736 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
737 return q->numa_node;
738 #else
739 return NUMA_NO_NODE;
740 #endif
741 }
742
netdev_queue_numa_node_write(struct netdev_queue * q,int node)743 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
744 {
745 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
746 q->numa_node = node;
747 #endif
748 }
749
750 #ifdef CONFIG_RFS_ACCEL
751 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
752 u16 filter_id);
753 #endif
754
755 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
756 enum xps_map_type {
757 XPS_CPUS = 0,
758 XPS_RXQS,
759 XPS_MAPS_MAX,
760 };
761
762 #ifdef CONFIG_XPS
763 /*
764 * This structure holds an XPS map which can be of variable length. The
765 * map is an array of queues.
766 */
767 struct xps_map {
768 unsigned int len;
769 unsigned int alloc_len;
770 struct rcu_head rcu;
771 u16 queues[];
772 };
773 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
774 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
775 - sizeof(struct xps_map)) / sizeof(u16))
776
777 /*
778 * This structure holds all XPS maps for device. Maps are indexed by CPU.
779 *
780 * We keep track of the number of cpus/rxqs used when the struct is allocated,
781 * in nr_ids. This will help not accessing out-of-bound memory.
782 *
783 * We keep track of the number of traffic classes used when the struct is
784 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
785 * not crossing its upper bound, as the original dev->num_tc can be updated in
786 * the meantime.
787 */
788 struct xps_dev_maps {
789 struct rcu_head rcu;
790 unsigned int nr_ids;
791 s16 num_tc;
792 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
793 };
794
795 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
796 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
797
798 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
799 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
800
801 #endif /* CONFIG_XPS */
802
803 #define TC_MAX_QUEUE 16
804 #define TC_BITMASK 15
805 /* HW offloaded queuing disciplines txq count and offset maps */
806 struct netdev_tc_txq {
807 u16 count;
808 u16 offset;
809 };
810
811 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
812 /*
813 * This structure is to hold information about the device
814 * configured to run FCoE protocol stack.
815 */
816 struct netdev_fcoe_hbainfo {
817 char manufacturer[64];
818 char serial_number[64];
819 char hardware_version[64];
820 char driver_version[64];
821 char optionrom_version[64];
822 char firmware_version[64];
823 char model[256];
824 char model_description[256];
825 };
826 #endif
827
828 #define MAX_PHYS_ITEM_ID_LEN 32
829
830 /* This structure holds a unique identifier to identify some
831 * physical item (port for example) used by a netdevice.
832 */
833 struct netdev_phys_item_id {
834 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
835 unsigned char id_len;
836 };
837
netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)838 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
839 struct netdev_phys_item_id *b)
840 {
841 return a->id_len == b->id_len &&
842 memcmp(a->id, b->id, a->id_len) == 0;
843 }
844
845 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
846 struct sk_buff *skb,
847 struct net_device *sb_dev);
848
849 enum net_device_path_type {
850 DEV_PATH_ETHERNET = 0,
851 DEV_PATH_VLAN,
852 DEV_PATH_BRIDGE,
853 DEV_PATH_PPPOE,
854 DEV_PATH_DSA,
855 DEV_PATH_MTK_WDMA,
856 };
857
858 struct net_device_path {
859 enum net_device_path_type type;
860 const struct net_device *dev;
861 union {
862 struct {
863 u16 id;
864 __be16 proto;
865 u8 h_dest[ETH_ALEN];
866 } encap;
867 struct {
868 enum {
869 DEV_PATH_BR_VLAN_KEEP,
870 DEV_PATH_BR_VLAN_TAG,
871 DEV_PATH_BR_VLAN_UNTAG,
872 DEV_PATH_BR_VLAN_UNTAG_HW,
873 } vlan_mode;
874 u16 vlan_id;
875 __be16 vlan_proto;
876 } bridge;
877 struct {
878 int port;
879 u16 proto;
880 } dsa;
881 struct {
882 u8 wdma_idx;
883 u8 queue;
884 u16 wcid;
885 u8 bss;
886 u8 amsdu;
887 } mtk_wdma;
888 };
889 };
890
891 #define NET_DEVICE_PATH_STACK_MAX 5
892 #define NET_DEVICE_PATH_VLAN_MAX 2
893
894 struct net_device_path_stack {
895 int num_paths;
896 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
897 };
898
899 struct net_device_path_ctx {
900 const struct net_device *dev;
901 u8 daddr[ETH_ALEN];
902
903 int num_vlans;
904 struct {
905 u16 id;
906 __be16 proto;
907 } vlan[NET_DEVICE_PATH_VLAN_MAX];
908 };
909
910 enum tc_setup_type {
911 TC_QUERY_CAPS,
912 TC_SETUP_QDISC_MQPRIO,
913 TC_SETUP_CLSU32,
914 TC_SETUP_CLSFLOWER,
915 TC_SETUP_CLSMATCHALL,
916 TC_SETUP_CLSBPF,
917 TC_SETUP_BLOCK,
918 TC_SETUP_QDISC_CBS,
919 TC_SETUP_QDISC_RED,
920 TC_SETUP_QDISC_PRIO,
921 TC_SETUP_QDISC_MQ,
922 TC_SETUP_QDISC_ETF,
923 TC_SETUP_ROOT_QDISC,
924 TC_SETUP_QDISC_GRED,
925 TC_SETUP_QDISC_TAPRIO,
926 TC_SETUP_FT,
927 TC_SETUP_QDISC_ETS,
928 TC_SETUP_QDISC_TBF,
929 TC_SETUP_QDISC_FIFO,
930 TC_SETUP_QDISC_HTB,
931 TC_SETUP_ACT,
932 };
933
934 /* These structures hold the attributes of bpf state that are being passed
935 * to the netdevice through the bpf op.
936 */
937 enum bpf_netdev_command {
938 /* Set or clear a bpf program used in the earliest stages of packet
939 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
940 * is responsible for calling bpf_prog_put on any old progs that are
941 * stored. In case of error, the callee need not release the new prog
942 * reference, but on success it takes ownership and must bpf_prog_put
943 * when it is no longer used.
944 */
945 XDP_SETUP_PROG,
946 XDP_SETUP_PROG_HW,
947 /* BPF program for offload callbacks, invoked at program load time. */
948 BPF_OFFLOAD_MAP_ALLOC,
949 BPF_OFFLOAD_MAP_FREE,
950 XDP_SETUP_XSK_POOL,
951 };
952
953 struct bpf_prog_offload_ops;
954 struct netlink_ext_ack;
955 struct xdp_umem;
956 struct xdp_dev_bulk_queue;
957 struct bpf_xdp_link;
958
959 enum bpf_xdp_mode {
960 XDP_MODE_SKB = 0,
961 XDP_MODE_DRV = 1,
962 XDP_MODE_HW = 2,
963 __MAX_XDP_MODE
964 };
965
966 struct bpf_xdp_entity {
967 struct bpf_prog *prog;
968 struct bpf_xdp_link *link;
969 };
970
971 struct netdev_bpf {
972 enum bpf_netdev_command command;
973 union {
974 /* XDP_SETUP_PROG */
975 struct {
976 u32 flags;
977 struct bpf_prog *prog;
978 struct netlink_ext_ack *extack;
979 };
980 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
981 struct {
982 struct bpf_offloaded_map *offmap;
983 };
984 /* XDP_SETUP_XSK_POOL */
985 struct {
986 struct xsk_buff_pool *pool;
987 u16 queue_id;
988 } xsk;
989 };
990 };
991
992 /* Flags for ndo_xsk_wakeup. */
993 #define XDP_WAKEUP_RX (1 << 0)
994 #define XDP_WAKEUP_TX (1 << 1)
995
996 #ifdef CONFIG_XFRM_OFFLOAD
997 struct xfrmdev_ops {
998 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
999 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1000 void (*xdo_dev_state_free) (struct xfrm_state *x);
1001 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1002 struct xfrm_state *x);
1003 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1004 void (*xdo_dev_state_update_stats) (struct xfrm_state *x);
1005 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1006 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1007 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1008 };
1009 #endif
1010
1011 struct dev_ifalias {
1012 struct rcu_head rcuhead;
1013 char ifalias[];
1014 };
1015
1016 struct devlink;
1017 struct tlsdev_ops;
1018
1019 struct netdev_net_notifier {
1020 struct list_head list;
1021 struct notifier_block *nb;
1022 };
1023
1024 /*
1025 * This structure defines the management hooks for network devices.
1026 * The following hooks can be defined; unless noted otherwise, they are
1027 * optional and can be filled with a null pointer.
1028 *
1029 * int (*ndo_init)(struct net_device *dev);
1030 * This function is called once when a network device is registered.
1031 * The network device can use this for any late stage initialization
1032 * or semantic validation. It can fail with an error code which will
1033 * be propagated back to register_netdev.
1034 *
1035 * void (*ndo_uninit)(struct net_device *dev);
1036 * This function is called when device is unregistered or when registration
1037 * fails. It is not called if init fails.
1038 *
1039 * int (*ndo_open)(struct net_device *dev);
1040 * This function is called when a network device transitions to the up
1041 * state.
1042 *
1043 * int (*ndo_stop)(struct net_device *dev);
1044 * This function is called when a network device transitions to the down
1045 * state.
1046 *
1047 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1048 * struct net_device *dev);
1049 * Called when a packet needs to be transmitted.
1050 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1051 * the queue before that can happen; it's for obsolete devices and weird
1052 * corner cases, but the stack really does a non-trivial amount
1053 * of useless work if you return NETDEV_TX_BUSY.
1054 * Required; cannot be NULL.
1055 *
1056 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1057 * struct net_device *dev
1058 * netdev_features_t features);
1059 * Called by core transmit path to determine if device is capable of
1060 * performing offload operations on a given packet. This is to give
1061 * the device an opportunity to implement any restrictions that cannot
1062 * be otherwise expressed by feature flags. The check is called with
1063 * the set of features that the stack has calculated and it returns
1064 * those the driver believes to be appropriate.
1065 *
1066 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1067 * struct net_device *sb_dev);
1068 * Called to decide which queue to use when device supports multiple
1069 * transmit queues.
1070 *
1071 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1072 * This function is called to allow device receiver to make
1073 * changes to configuration when multicast or promiscuous is enabled.
1074 *
1075 * void (*ndo_set_rx_mode)(struct net_device *dev);
1076 * This function is called device changes address list filtering.
1077 * If driver handles unicast address filtering, it should set
1078 * IFF_UNICAST_FLT in its priv_flags.
1079 *
1080 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1081 * This function is called when the Media Access Control address
1082 * needs to be changed. If this interface is not defined, the
1083 * MAC address can not be changed.
1084 *
1085 * int (*ndo_validate_addr)(struct net_device *dev);
1086 * Test if Media Access Control address is valid for the device.
1087 *
1088 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1089 * Old-style ioctl entry point. This is used internally by the
1090 * appletalk and ieee802154 subsystems but is no longer called by
1091 * the device ioctl handler.
1092 *
1093 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1094 * Used by the bonding driver for its device specific ioctls:
1095 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1096 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1097 *
1098 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1099 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1100 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1101 *
1102 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1103 * Used to set network devices bus interface parameters. This interface
1104 * is retained for legacy reasons; new devices should use the bus
1105 * interface (PCI) for low level management.
1106 *
1107 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1108 * Called when a user wants to change the Maximum Transfer Unit
1109 * of a device.
1110 *
1111 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1112 * Callback used when the transmitter has not made any progress
1113 * for dev->watchdog ticks.
1114 *
1115 * void (*ndo_get_stats64)(struct net_device *dev,
1116 * struct rtnl_link_stats64 *storage);
1117 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1118 * Called when a user wants to get the network device usage
1119 * statistics. Drivers must do one of the following:
1120 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1121 * rtnl_link_stats64 structure passed by the caller.
1122 * 2. Define @ndo_get_stats to update a net_device_stats structure
1123 * (which should normally be dev->stats) and return a pointer to
1124 * it. The structure may be changed asynchronously only if each
1125 * field is written atomically.
1126 * 3. Update dev->stats asynchronously and atomically, and define
1127 * neither operation.
1128 *
1129 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1130 * Return true if this device supports offload stats of this attr_id.
1131 *
1132 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1133 * void *attr_data)
1134 * Get statistics for offload operations by attr_id. Write it into the
1135 * attr_data pointer.
1136 *
1137 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1138 * If device supports VLAN filtering this function is called when a
1139 * VLAN id is registered.
1140 *
1141 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1142 * If device supports VLAN filtering this function is called when a
1143 * VLAN id is unregistered.
1144 *
1145 * void (*ndo_poll_controller)(struct net_device *dev);
1146 *
1147 * SR-IOV management functions.
1148 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1149 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1150 * u8 qos, __be16 proto);
1151 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1152 * int max_tx_rate);
1153 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1154 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1155 * int (*ndo_get_vf_config)(struct net_device *dev,
1156 * int vf, struct ifla_vf_info *ivf);
1157 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1158 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1159 * struct nlattr *port[]);
1160 *
1161 * Enable or disable the VF ability to query its RSS Redirection Table and
1162 * Hash Key. This is needed since on some devices VF share this information
1163 * with PF and querying it may introduce a theoretical security risk.
1164 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1165 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1166 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1167 * void *type_data);
1168 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1169 * This is always called from the stack with the rtnl lock held and netif
1170 * tx queues stopped. This allows the netdevice to perform queue
1171 * management safely.
1172 *
1173 * Fiber Channel over Ethernet (FCoE) offload functions.
1174 * int (*ndo_fcoe_enable)(struct net_device *dev);
1175 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1176 * so the underlying device can perform whatever needed configuration or
1177 * initialization to support acceleration of FCoE traffic.
1178 *
1179 * int (*ndo_fcoe_disable)(struct net_device *dev);
1180 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1181 * so the underlying device can perform whatever needed clean-ups to
1182 * stop supporting acceleration of FCoE traffic.
1183 *
1184 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1185 * struct scatterlist *sgl, unsigned int sgc);
1186 * Called when the FCoE Initiator wants to initialize an I/O that
1187 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1188 * perform necessary setup and returns 1 to indicate the device is set up
1189 * successfully to perform DDP on this I/O, otherwise this returns 0.
1190 *
1191 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1192 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1193 * indicated by the FC exchange id 'xid', so the underlying device can
1194 * clean up and reuse resources for later DDP requests.
1195 *
1196 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1197 * struct scatterlist *sgl, unsigned int sgc);
1198 * Called when the FCoE Target wants to initialize an I/O that
1199 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1200 * perform necessary setup and returns 1 to indicate the device is set up
1201 * successfully to perform DDP on this I/O, otherwise this returns 0.
1202 *
1203 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1204 * struct netdev_fcoe_hbainfo *hbainfo);
1205 * Called when the FCoE Protocol stack wants information on the underlying
1206 * device. This information is utilized by the FCoE protocol stack to
1207 * register attributes with Fiber Channel management service as per the
1208 * FC-GS Fabric Device Management Information(FDMI) specification.
1209 *
1210 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1211 * Called when the underlying device wants to override default World Wide
1212 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1213 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1214 * protocol stack to use.
1215 *
1216 * RFS acceleration.
1217 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1218 * u16 rxq_index, u32 flow_id);
1219 * Set hardware filter for RFS. rxq_index is the target queue index;
1220 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1221 * Return the filter ID on success, or a negative error code.
1222 *
1223 * Slave management functions (for bridge, bonding, etc).
1224 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1225 * Called to make another netdev an underling.
1226 *
1227 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1228 * Called to release previously enslaved netdev.
1229 *
1230 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1231 * struct sk_buff *skb,
1232 * bool all_slaves);
1233 * Get the xmit slave of master device. If all_slaves is true, function
1234 * assume all the slaves can transmit.
1235 *
1236 * Feature/offload setting functions.
1237 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1238 * netdev_features_t features);
1239 * Adjusts the requested feature flags according to device-specific
1240 * constraints, and returns the resulting flags. Must not modify
1241 * the device state.
1242 *
1243 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1244 * Called to update device configuration to new features. Passed
1245 * feature set might be less than what was returned by ndo_fix_features()).
1246 * Must return >0 or -errno if it changed dev->features itself.
1247 *
1248 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1249 * struct net_device *dev,
1250 * const unsigned char *addr, u16 vid, u16 flags,
1251 * bool *notified, struct netlink_ext_ack *extack);
1252 * Adds an FDB entry to dev for addr.
1253 * Callee shall set *notified to true if it sent any appropriate
1254 * notification(s). Otherwise core will send a generic one.
1255 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1256 * struct net_device *dev,
1257 * const unsigned char *addr, u16 vid
1258 * bool *notified, struct netlink_ext_ack *extack);
1259 * Deletes the FDB entry from dev corresponding to addr.
1260 * Callee shall set *notified to true if it sent any appropriate
1261 * notification(s). Otherwise core will send a generic one.
1262 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1263 * struct netlink_ext_ack *extack);
1264 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1265 * struct net_device *dev, struct net_device *filter_dev,
1266 * int *idx)
1267 * Used to add FDB entries to dump requests. Implementers should add
1268 * entries to skb and update idx with the number of entries.
1269 *
1270 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1271 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1272 * Adds an MDB entry to dev.
1273 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1274 * struct netlink_ext_ack *extack);
1275 * Deletes the MDB entry from dev.
1276 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1277 * struct netlink_ext_ack *extack);
1278 * Bulk deletes MDB entries from dev.
1279 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1280 * struct netlink_callback *cb);
1281 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1282 * callback is used by core rtnetlink code.
1283 *
1284 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1285 * u16 flags, struct netlink_ext_ack *extack)
1286 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1287 * struct net_device *dev, u32 filter_mask,
1288 * int nlflags)
1289 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1290 * u16 flags);
1291 *
1292 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1293 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1294 * which do not represent real hardware may define this to allow their
1295 * userspace components to manage their virtual carrier state. Devices
1296 * that determine carrier state from physical hardware properties (eg
1297 * network cables) or protocol-dependent mechanisms (eg
1298 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1299 *
1300 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1301 * struct netdev_phys_item_id *ppid);
1302 * Called to get ID of physical port of this device. If driver does
1303 * not implement this, it is assumed that the hw is not able to have
1304 * multiple net devices on single physical port.
1305 *
1306 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1307 * struct netdev_phys_item_id *ppid)
1308 * Called to get the parent ID of the physical port of this device.
1309 *
1310 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1311 * struct net_device *dev)
1312 * Called by upper layer devices to accelerate switching or other
1313 * station functionality into hardware. 'pdev is the lowerdev
1314 * to use for the offload and 'dev' is the net device that will
1315 * back the offload. Returns a pointer to the private structure
1316 * the upper layer will maintain.
1317 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1318 * Called by upper layer device to delete the station created
1319 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1320 * the station and priv is the structure returned by the add
1321 * operation.
1322 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1323 * int queue_index, u32 maxrate);
1324 * Called when a user wants to set a max-rate limitation of specific
1325 * TX queue.
1326 * int (*ndo_get_iflink)(const struct net_device *dev);
1327 * Called to get the iflink value of this device.
1328 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1329 * This function is used to get egress tunnel information for given skb.
1330 * This is useful for retrieving outer tunnel header parameters while
1331 * sampling packet.
1332 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1333 * This function is used to specify the headroom that the skb must
1334 * consider when allocation skb during packet reception. Setting
1335 * appropriate rx headroom value allows avoiding skb head copy on
1336 * forward. Setting a negative value resets the rx headroom to the
1337 * default value.
1338 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1339 * This function is used to set or query state related to XDP on the
1340 * netdevice and manage BPF offload. See definition of
1341 * enum bpf_netdev_command for details.
1342 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1343 * u32 flags);
1344 * This function is used to submit @n XDP packets for transmit on a
1345 * netdevice. Returns number of frames successfully transmitted, frames
1346 * that got dropped are freed/returned via xdp_return_frame().
1347 * Returns negative number, means general error invoking ndo, meaning
1348 * no frames were xmit'ed and core-caller will free all frames.
1349 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1350 * struct xdp_buff *xdp);
1351 * Get the xmit slave of master device based on the xdp_buff.
1352 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1353 * This function is used to wake up the softirq, ksoftirqd or kthread
1354 * responsible for sending and/or receiving packets on a specific
1355 * queue id bound to an AF_XDP socket. The flags field specifies if
1356 * only RX, only Tx, or both should be woken up using the flags
1357 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1358 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1359 * int cmd);
1360 * Add, change, delete or get information on an IPv4 tunnel.
1361 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1362 * If a device is paired with a peer device, return the peer instance.
1363 * The caller must be under RCU read context.
1364 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1365 * Get the forwarding path to reach the real device from the HW destination address
1366 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1367 * const struct skb_shared_hwtstamps *hwtstamps,
1368 * bool cycles);
1369 * Get hardware timestamp based on normal/adjustable time or free running
1370 * cycle counter. This function is required if physical clock supports a
1371 * free running cycle counter.
1372 *
1373 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1374 * struct kernel_hwtstamp_config *kernel_config);
1375 * Get the currently configured hardware timestamping parameters for the
1376 * NIC device.
1377 *
1378 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1379 * struct kernel_hwtstamp_config *kernel_config,
1380 * struct netlink_ext_ack *extack);
1381 * Change the hardware timestamping parameters for NIC device.
1382 */
1383 struct net_device_ops {
1384 int (*ndo_init)(struct net_device *dev);
1385 void (*ndo_uninit)(struct net_device *dev);
1386 int (*ndo_open)(struct net_device *dev);
1387 int (*ndo_stop)(struct net_device *dev);
1388 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1389 struct net_device *dev);
1390 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1391 struct net_device *dev,
1392 netdev_features_t features);
1393 u16 (*ndo_select_queue)(struct net_device *dev,
1394 struct sk_buff *skb,
1395 struct net_device *sb_dev);
1396 void (*ndo_change_rx_flags)(struct net_device *dev,
1397 int flags);
1398 void (*ndo_set_rx_mode)(struct net_device *dev);
1399 int (*ndo_set_mac_address)(struct net_device *dev,
1400 void *addr);
1401 int (*ndo_validate_addr)(struct net_device *dev);
1402 int (*ndo_do_ioctl)(struct net_device *dev,
1403 struct ifreq *ifr, int cmd);
1404 int (*ndo_eth_ioctl)(struct net_device *dev,
1405 struct ifreq *ifr, int cmd);
1406 int (*ndo_siocbond)(struct net_device *dev,
1407 struct ifreq *ifr, int cmd);
1408 int (*ndo_siocwandev)(struct net_device *dev,
1409 struct if_settings *ifs);
1410 int (*ndo_siocdevprivate)(struct net_device *dev,
1411 struct ifreq *ifr,
1412 void __user *data, int cmd);
1413 int (*ndo_set_config)(struct net_device *dev,
1414 struct ifmap *map);
1415 int (*ndo_change_mtu)(struct net_device *dev,
1416 int new_mtu);
1417 int (*ndo_neigh_setup)(struct net_device *dev,
1418 struct neigh_parms *);
1419 void (*ndo_tx_timeout) (struct net_device *dev,
1420 unsigned int txqueue);
1421
1422 void (*ndo_get_stats64)(struct net_device *dev,
1423 struct rtnl_link_stats64 *storage);
1424 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1425 int (*ndo_get_offload_stats)(int attr_id,
1426 const struct net_device *dev,
1427 void *attr_data);
1428 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1429
1430 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1431 __be16 proto, u16 vid);
1432 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1433 __be16 proto, u16 vid);
1434 #ifdef CONFIG_NET_POLL_CONTROLLER
1435 void (*ndo_poll_controller)(struct net_device *dev);
1436 int (*ndo_netpoll_setup)(struct net_device *dev);
1437 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1438 #endif
1439 int (*ndo_set_vf_mac)(struct net_device *dev,
1440 int queue, u8 *mac);
1441 int (*ndo_set_vf_vlan)(struct net_device *dev,
1442 int queue, u16 vlan,
1443 u8 qos, __be16 proto);
1444 int (*ndo_set_vf_rate)(struct net_device *dev,
1445 int vf, int min_tx_rate,
1446 int max_tx_rate);
1447 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1448 int vf, bool setting);
1449 int (*ndo_set_vf_trust)(struct net_device *dev,
1450 int vf, bool setting);
1451 int (*ndo_get_vf_config)(struct net_device *dev,
1452 int vf,
1453 struct ifla_vf_info *ivf);
1454 int (*ndo_set_vf_link_state)(struct net_device *dev,
1455 int vf, int link_state);
1456 int (*ndo_get_vf_stats)(struct net_device *dev,
1457 int vf,
1458 struct ifla_vf_stats
1459 *vf_stats);
1460 int (*ndo_set_vf_port)(struct net_device *dev,
1461 int vf,
1462 struct nlattr *port[]);
1463 int (*ndo_get_vf_port)(struct net_device *dev,
1464 int vf, struct sk_buff *skb);
1465 int (*ndo_get_vf_guid)(struct net_device *dev,
1466 int vf,
1467 struct ifla_vf_guid *node_guid,
1468 struct ifla_vf_guid *port_guid);
1469 int (*ndo_set_vf_guid)(struct net_device *dev,
1470 int vf, u64 guid,
1471 int guid_type);
1472 int (*ndo_set_vf_rss_query_en)(
1473 struct net_device *dev,
1474 int vf, bool setting);
1475 int (*ndo_setup_tc)(struct net_device *dev,
1476 enum tc_setup_type type,
1477 void *type_data);
1478 #if IS_ENABLED(CONFIG_FCOE)
1479 int (*ndo_fcoe_enable)(struct net_device *dev);
1480 int (*ndo_fcoe_disable)(struct net_device *dev);
1481 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1482 u16 xid,
1483 struct scatterlist *sgl,
1484 unsigned int sgc);
1485 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1486 u16 xid);
1487 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1488 u16 xid,
1489 struct scatterlist *sgl,
1490 unsigned int sgc);
1491 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1492 struct netdev_fcoe_hbainfo *hbainfo);
1493 #endif
1494
1495 #if IS_ENABLED(CONFIG_LIBFCOE)
1496 #define NETDEV_FCOE_WWNN 0
1497 #define NETDEV_FCOE_WWPN 1
1498 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1499 u64 *wwn, int type);
1500 #endif
1501
1502 #ifdef CONFIG_RFS_ACCEL
1503 int (*ndo_rx_flow_steer)(struct net_device *dev,
1504 const struct sk_buff *skb,
1505 u16 rxq_index,
1506 u32 flow_id);
1507 #endif
1508 int (*ndo_add_slave)(struct net_device *dev,
1509 struct net_device *slave_dev,
1510 struct netlink_ext_ack *extack);
1511 int (*ndo_del_slave)(struct net_device *dev,
1512 struct net_device *slave_dev);
1513 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1514 struct sk_buff *skb,
1515 bool all_slaves);
1516 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1517 struct sock *sk);
1518 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1519 netdev_features_t features);
1520 int (*ndo_set_features)(struct net_device *dev,
1521 netdev_features_t features);
1522 int (*ndo_neigh_construct)(struct net_device *dev,
1523 struct neighbour *n);
1524 void (*ndo_neigh_destroy)(struct net_device *dev,
1525 struct neighbour *n);
1526
1527 int (*ndo_fdb_add)(struct ndmsg *ndm,
1528 struct nlattr *tb[],
1529 struct net_device *dev,
1530 const unsigned char *addr,
1531 u16 vid,
1532 u16 flags,
1533 bool *notified,
1534 struct netlink_ext_ack *extack);
1535 int (*ndo_fdb_del)(struct ndmsg *ndm,
1536 struct nlattr *tb[],
1537 struct net_device *dev,
1538 const unsigned char *addr,
1539 u16 vid,
1540 bool *notified,
1541 struct netlink_ext_ack *extack);
1542 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1543 struct net_device *dev,
1544 struct netlink_ext_ack *extack);
1545 int (*ndo_fdb_dump)(struct sk_buff *skb,
1546 struct netlink_callback *cb,
1547 struct net_device *dev,
1548 struct net_device *filter_dev,
1549 int *idx);
1550 int (*ndo_fdb_get)(struct sk_buff *skb,
1551 struct nlattr *tb[],
1552 struct net_device *dev,
1553 const unsigned char *addr,
1554 u16 vid, u32 portid, u32 seq,
1555 struct netlink_ext_ack *extack);
1556 int (*ndo_mdb_add)(struct net_device *dev,
1557 struct nlattr *tb[],
1558 u16 nlmsg_flags,
1559 struct netlink_ext_ack *extack);
1560 int (*ndo_mdb_del)(struct net_device *dev,
1561 struct nlattr *tb[],
1562 struct netlink_ext_ack *extack);
1563 int (*ndo_mdb_del_bulk)(struct net_device *dev,
1564 struct nlattr *tb[],
1565 struct netlink_ext_ack *extack);
1566 int (*ndo_mdb_dump)(struct net_device *dev,
1567 struct sk_buff *skb,
1568 struct netlink_callback *cb);
1569 int (*ndo_mdb_get)(struct net_device *dev,
1570 struct nlattr *tb[], u32 portid,
1571 u32 seq,
1572 struct netlink_ext_ack *extack);
1573 int (*ndo_bridge_setlink)(struct net_device *dev,
1574 struct nlmsghdr *nlh,
1575 u16 flags,
1576 struct netlink_ext_ack *extack);
1577 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1578 u32 pid, u32 seq,
1579 struct net_device *dev,
1580 u32 filter_mask,
1581 int nlflags);
1582 int (*ndo_bridge_dellink)(struct net_device *dev,
1583 struct nlmsghdr *nlh,
1584 u16 flags);
1585 int (*ndo_change_carrier)(struct net_device *dev,
1586 bool new_carrier);
1587 int (*ndo_get_phys_port_id)(struct net_device *dev,
1588 struct netdev_phys_item_id *ppid);
1589 int (*ndo_get_port_parent_id)(struct net_device *dev,
1590 struct netdev_phys_item_id *ppid);
1591 int (*ndo_get_phys_port_name)(struct net_device *dev,
1592 char *name, size_t len);
1593 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1594 struct net_device *dev);
1595 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1596 void *priv);
1597
1598 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1599 int queue_index,
1600 u32 maxrate);
1601 int (*ndo_get_iflink)(const struct net_device *dev);
1602 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1603 struct sk_buff *skb);
1604 void (*ndo_set_rx_headroom)(struct net_device *dev,
1605 int needed_headroom);
1606 int (*ndo_bpf)(struct net_device *dev,
1607 struct netdev_bpf *bpf);
1608 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1609 struct xdp_frame **xdp,
1610 u32 flags);
1611 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1612 struct xdp_buff *xdp);
1613 int (*ndo_xsk_wakeup)(struct net_device *dev,
1614 u32 queue_id, u32 flags);
1615 int (*ndo_tunnel_ctl)(struct net_device *dev,
1616 struct ip_tunnel_parm_kern *p,
1617 int cmd);
1618 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1619 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1620 struct net_device_path *path);
1621 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1622 const struct skb_shared_hwtstamps *hwtstamps,
1623 bool cycles);
1624 int (*ndo_hwtstamp_get)(struct net_device *dev,
1625 struct kernel_hwtstamp_config *kernel_config);
1626 int (*ndo_hwtstamp_set)(struct net_device *dev,
1627 struct kernel_hwtstamp_config *kernel_config,
1628 struct netlink_ext_ack *extack);
1629
1630 #if IS_ENABLED(CONFIG_NET_SHAPER)
1631 /**
1632 * @net_shaper_ops: Device shaping offload operations
1633 * see include/net/net_shapers.h
1634 */
1635 const struct net_shaper_ops *net_shaper_ops;
1636 #endif
1637 };
1638
1639 /**
1640 * enum netdev_priv_flags - &struct net_device priv_flags
1641 *
1642 * These are the &struct net_device, they are only set internally
1643 * by drivers and used in the kernel. These flags are invisible to
1644 * userspace; this means that the order of these flags can change
1645 * during any kernel release.
1646 *
1647 * You should add bitfield booleans after either net_device::priv_flags
1648 * (hotpath) or ::threaded (slowpath) instead of extending these flags.
1649 *
1650 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1651 * @IFF_EBRIDGE: Ethernet bridging device
1652 * @IFF_BONDING: bonding master or slave
1653 * @IFF_ISATAP: ISATAP interface (RFC4214)
1654 * @IFF_WAN_HDLC: WAN HDLC device
1655 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1656 * release skb->dst
1657 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1658 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1659 * @IFF_MACVLAN_PORT: device used as macvlan port
1660 * @IFF_BRIDGE_PORT: device used as bridge port
1661 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1662 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1663 * @IFF_UNICAST_FLT: Supports unicast filtering
1664 * @IFF_TEAM_PORT: device used as team port
1665 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1666 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1667 * change when it's running
1668 * @IFF_MACVLAN: Macvlan device
1669 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1670 * underlying stacked devices
1671 * @IFF_L3MDEV_MASTER: device is an L3 master device
1672 * @IFF_NO_QUEUE: device can run without qdisc attached
1673 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1674 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1675 * @IFF_TEAM: device is a team device
1676 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1677 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1678 * entity (i.e. the master device for bridged veth)
1679 * @IFF_MACSEC: device is a MACsec device
1680 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1681 * @IFF_FAILOVER: device is a failover master device
1682 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1683 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1684 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1685 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1686 * skb_headlen(skb) == 0 (data starts from frag0)
1687 */
1688 enum netdev_priv_flags {
1689 IFF_802_1Q_VLAN = 1<<0,
1690 IFF_EBRIDGE = 1<<1,
1691 IFF_BONDING = 1<<2,
1692 IFF_ISATAP = 1<<3,
1693 IFF_WAN_HDLC = 1<<4,
1694 IFF_XMIT_DST_RELEASE = 1<<5,
1695 IFF_DONT_BRIDGE = 1<<6,
1696 IFF_DISABLE_NETPOLL = 1<<7,
1697 IFF_MACVLAN_PORT = 1<<8,
1698 IFF_BRIDGE_PORT = 1<<9,
1699 IFF_OVS_DATAPATH = 1<<10,
1700 IFF_TX_SKB_SHARING = 1<<11,
1701 IFF_UNICAST_FLT = 1<<12,
1702 IFF_TEAM_PORT = 1<<13,
1703 IFF_SUPP_NOFCS = 1<<14,
1704 IFF_LIVE_ADDR_CHANGE = 1<<15,
1705 IFF_MACVLAN = 1<<16,
1706 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1707 IFF_L3MDEV_MASTER = 1<<18,
1708 IFF_NO_QUEUE = 1<<19,
1709 IFF_OPENVSWITCH = 1<<20,
1710 IFF_L3MDEV_SLAVE = 1<<21,
1711 IFF_TEAM = 1<<22,
1712 IFF_RXFH_CONFIGURED = 1<<23,
1713 IFF_PHONY_HEADROOM = 1<<24,
1714 IFF_MACSEC = 1<<25,
1715 IFF_NO_RX_HANDLER = 1<<26,
1716 IFF_FAILOVER = 1<<27,
1717 IFF_FAILOVER_SLAVE = 1<<28,
1718 IFF_L3MDEV_RX_HANDLER = 1<<29,
1719 IFF_NO_ADDRCONF = BIT_ULL(30),
1720 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1721 };
1722
1723 /* Specifies the type of the struct net_device::ml_priv pointer */
1724 enum netdev_ml_priv_type {
1725 ML_PRIV_NONE,
1726 ML_PRIV_CAN,
1727 };
1728
1729 enum netdev_stat_type {
1730 NETDEV_PCPU_STAT_NONE,
1731 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1732 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1733 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1734 };
1735
1736 enum netdev_reg_state {
1737 NETREG_UNINITIALIZED = 0,
1738 NETREG_REGISTERED, /* completed register_netdevice */
1739 NETREG_UNREGISTERING, /* called unregister_netdevice */
1740 NETREG_UNREGISTERED, /* completed unregister todo */
1741 NETREG_RELEASED, /* called free_netdev */
1742 NETREG_DUMMY, /* dummy device for NAPI poll */
1743 };
1744
1745 /**
1746 * struct net_device - The DEVICE structure.
1747 *
1748 * Actually, this whole structure is a big mistake. It mixes I/O
1749 * data with strictly "high-level" data, and it has to know about
1750 * almost every data structure used in the INET module.
1751 *
1752 * @priv_flags: flags invisible to userspace defined as bits, see
1753 * enum netdev_priv_flags for the definitions
1754 * @lltx: device supports lockless Tx. Deprecated for real HW
1755 * drivers. Mainly used by logical interfaces, such as
1756 * bonding and tunnels
1757 *
1758 * @name: This is the first field of the "visible" part of this structure
1759 * (i.e. as seen by users in the "Space.c" file). It is the name
1760 * of the interface.
1761 *
1762 * @name_node: Name hashlist node
1763 * @ifalias: SNMP alias
1764 * @mem_end: Shared memory end
1765 * @mem_start: Shared memory start
1766 * @base_addr: Device I/O address
1767 * @irq: Device IRQ number
1768 *
1769 * @state: Generic network queuing layer state, see netdev_state_t
1770 * @dev_list: The global list of network devices
1771 * @napi_list: List entry used for polling NAPI devices
1772 * @unreg_list: List entry when we are unregistering the
1773 * device; see the function unregister_netdev
1774 * @close_list: List entry used when we are closing the device
1775 * @ptype_all: Device-specific packet handlers for all protocols
1776 * @ptype_specific: Device-specific, protocol-specific packet handlers
1777 *
1778 * @adj_list: Directly linked devices, like slaves for bonding
1779 * @features: Currently active device features
1780 * @hw_features: User-changeable features
1781 *
1782 * @wanted_features: User-requested features
1783 * @vlan_features: Mask of features inheritable by VLAN devices
1784 *
1785 * @hw_enc_features: Mask of features inherited by encapsulating devices
1786 * This field indicates what encapsulation
1787 * offloads the hardware is capable of doing,
1788 * and drivers will need to set them appropriately.
1789 *
1790 * @mpls_features: Mask of features inheritable by MPLS
1791 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1792 *
1793 * @ifindex: interface index
1794 * @group: The group the device belongs to
1795 *
1796 * @stats: Statistics struct, which was left as a legacy, use
1797 * rtnl_link_stats64 instead
1798 *
1799 * @core_stats: core networking counters,
1800 * do not use this in drivers
1801 * @carrier_up_count: Number of times the carrier has been up
1802 * @carrier_down_count: Number of times the carrier has been down
1803 *
1804 * @wireless_handlers: List of functions to handle Wireless Extensions,
1805 * instead of ioctl,
1806 * see <net/iw_handler.h> for details.
1807 *
1808 * @netdev_ops: Includes several pointers to callbacks,
1809 * if one wants to override the ndo_*() functions
1810 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1811 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks.
1812 * @ethtool_ops: Management operations
1813 * @l3mdev_ops: Layer 3 master device operations
1814 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1815 * discovery handling. Necessary for e.g. 6LoWPAN.
1816 * @xfrmdev_ops: Transformation offload operations
1817 * @tlsdev_ops: Transport Layer Security offload operations
1818 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1819 * of Layer 2 headers.
1820 *
1821 * @flags: Interface flags (a la BSD)
1822 * @xdp_features: XDP capability supported by the device
1823 * @gflags: Global flags ( kept as legacy )
1824 * @priv_len: Size of the ->priv flexible array
1825 * @priv: Flexible array containing private data
1826 * @operstate: RFC2863 operstate
1827 * @link_mode: Mapping policy to operstate
1828 * @if_port: Selectable AUI, TP, ...
1829 * @dma: DMA channel
1830 * @mtu: Interface MTU value
1831 * @min_mtu: Interface Minimum MTU value
1832 * @max_mtu: Interface Maximum MTU value
1833 * @type: Interface hardware type
1834 * @hard_header_len: Maximum hardware header length.
1835 * @min_header_len: Minimum hardware header length
1836 *
1837 * @needed_headroom: Extra headroom the hardware may need, but not in all
1838 * cases can this be guaranteed
1839 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1840 * cases can this be guaranteed. Some cases also use
1841 * LL_MAX_HEADER instead to allocate the skb
1842 *
1843 * interface address info:
1844 *
1845 * @perm_addr: Permanent hw address
1846 * @addr_assign_type: Hw address assignment type
1847 * @addr_len: Hardware address length
1848 * @upper_level: Maximum depth level of upper devices.
1849 * @lower_level: Maximum depth level of lower devices.
1850 * @neigh_priv_len: Used in neigh_alloc()
1851 * @dev_id: Used to differentiate devices that share
1852 * the same link layer address
1853 * @dev_port: Used to differentiate devices that share
1854 * the same function
1855 * @addr_list_lock: XXX: need comments on this one
1856 * @name_assign_type: network interface name assignment type
1857 * @uc_promisc: Counter that indicates promiscuous mode
1858 * has been enabled due to the need to listen to
1859 * additional unicast addresses in a device that
1860 * does not implement ndo_set_rx_mode()
1861 * @uc: unicast mac addresses
1862 * @mc: multicast mac addresses
1863 * @dev_addrs: list of device hw addresses
1864 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1865 * @promiscuity: Number of times the NIC is told to work in
1866 * promiscuous mode; if it becomes 0 the NIC will
1867 * exit promiscuous mode
1868 * @allmulti: Counter, enables or disables allmulticast mode
1869 *
1870 * @vlan_info: VLAN info
1871 * @dsa_ptr: dsa specific data
1872 * @tipc_ptr: TIPC specific data
1873 * @atalk_ptr: AppleTalk link
1874 * @ip_ptr: IPv4 specific data
1875 * @ip6_ptr: IPv6 specific data
1876 * @ax25_ptr: AX.25 specific data
1877 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1878 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1879 * device struct
1880 * @mpls_ptr: mpls_dev struct pointer
1881 * @mctp_ptr: MCTP specific data
1882 *
1883 * @dev_addr: Hw address (before bcast,
1884 * because most packets are unicast)
1885 *
1886 * @_rx: Array of RX queues
1887 * @num_rx_queues: Number of RX queues
1888 * allocated at register_netdev() time
1889 * @real_num_rx_queues: Number of RX queues currently active in device
1890 * @xdp_prog: XDP sockets filter program pointer
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 * @max_pacing_offload_horizon: max EDT offload horizon in nsec.
2040 * @napi_config: An array of napi_config structures containing per-NAPI
2041 * settings.
2042 * @gro_flush_timeout: timeout for GRO layer in NAPI
2043 * @napi_defer_hard_irqs: If not zero, provides a counter that would
2044 * allow to avoid NIC hard IRQ, on busy queues.
2045 *
2046 * @neighbours: List heads pointing to this device's neighbours'
2047 * dev_list, one per address-family.
2048 *
2049 * FIXME: cleanup struct net_device such that network protocol info
2050 * moves out.
2051 */
2052
2053 struct net_device {
2054 /* Cacheline organization can be found documented in
2055 * Documentation/networking/net_cachelines/net_device.rst.
2056 * Please update the document when adding new fields.
2057 */
2058
2059 /* TX read-mostly hotpath */
2060 __cacheline_group_begin(net_device_read_tx);
2061 struct_group(priv_flags_fast,
2062 unsigned long priv_flags:32;
2063 unsigned long lltx:1;
2064 );
2065 const struct net_device_ops *netdev_ops;
2066 const struct header_ops *header_ops;
2067 struct netdev_queue *_tx;
2068 netdev_features_t gso_partial_features;
2069 unsigned int real_num_tx_queues;
2070 unsigned int gso_max_size;
2071 unsigned int gso_ipv4_max_size;
2072 u16 gso_max_segs;
2073 s16 num_tc;
2074 /* Note : dev->mtu is often read without holding a lock.
2075 * Writers usually hold RTNL.
2076 * It is recommended to use READ_ONCE() to annotate the reads,
2077 * and to use WRITE_ONCE() to annotate the writes.
2078 */
2079 unsigned int mtu;
2080 unsigned short needed_headroom;
2081 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2082 #ifdef CONFIG_XPS
2083 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2084 #endif
2085 #ifdef CONFIG_NETFILTER_EGRESS
2086 struct nf_hook_entries __rcu *nf_hooks_egress;
2087 #endif
2088 #ifdef CONFIG_NET_XGRESS
2089 struct bpf_mprog_entry __rcu *tcx_egress;
2090 #endif
2091 __cacheline_group_end(net_device_read_tx);
2092
2093 /* TXRX read-mostly hotpath */
2094 __cacheline_group_begin(net_device_read_txrx);
2095 union {
2096 struct pcpu_lstats __percpu *lstats;
2097 struct pcpu_sw_netstats __percpu *tstats;
2098 struct pcpu_dstats __percpu *dstats;
2099 };
2100 unsigned long state;
2101 unsigned int flags;
2102 unsigned short hard_header_len;
2103 netdev_features_t features;
2104 struct inet6_dev __rcu *ip6_ptr;
2105 __cacheline_group_end(net_device_read_txrx);
2106
2107 /* RX read-mostly hotpath */
2108 __cacheline_group_begin(net_device_read_rx);
2109 struct bpf_prog __rcu *xdp_prog;
2110 struct list_head ptype_specific;
2111 int ifindex;
2112 unsigned int real_num_rx_queues;
2113 struct netdev_rx_queue *_rx;
2114 unsigned int gro_max_size;
2115 unsigned int gro_ipv4_max_size;
2116 rx_handler_func_t __rcu *rx_handler;
2117 void __rcu *rx_handler_data;
2118 possible_net_t nd_net;
2119 #ifdef CONFIG_NETPOLL
2120 struct netpoll_info __rcu *npinfo;
2121 #endif
2122 #ifdef CONFIG_NET_XGRESS
2123 struct bpf_mprog_entry __rcu *tcx_ingress;
2124 #endif
2125 __cacheline_group_end(net_device_read_rx);
2126
2127 char name[IFNAMSIZ];
2128 struct netdev_name_node *name_node;
2129 struct dev_ifalias __rcu *ifalias;
2130 /*
2131 * I/O specific fields
2132 * FIXME: Merge these and struct ifmap into one
2133 */
2134 unsigned long mem_end;
2135 unsigned long mem_start;
2136 unsigned long base_addr;
2137
2138 /*
2139 * Some hardware also needs these fields (state,dev_list,
2140 * napi_list,unreg_list,close_list) but they are not
2141 * part of the usual set specified in Space.c.
2142 */
2143
2144
2145 struct list_head dev_list;
2146 struct list_head napi_list;
2147 struct list_head unreg_list;
2148 struct list_head close_list;
2149 struct list_head ptype_all;
2150
2151 struct {
2152 struct list_head upper;
2153 struct list_head lower;
2154 } adj_list;
2155
2156 /* Read-mostly cache-line for fast-path access */
2157 xdp_features_t xdp_features;
2158 const struct xdp_metadata_ops *xdp_metadata_ops;
2159 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2160 unsigned short gflags;
2161
2162 unsigned short needed_tailroom;
2163
2164 netdev_features_t hw_features;
2165 netdev_features_t wanted_features;
2166 netdev_features_t vlan_features;
2167 netdev_features_t hw_enc_features;
2168 netdev_features_t mpls_features;
2169
2170 unsigned int min_mtu;
2171 unsigned int max_mtu;
2172 unsigned short type;
2173 unsigned char min_header_len;
2174 unsigned char name_assign_type;
2175
2176 int group;
2177
2178 struct net_device_stats stats; /* not used by modern drivers */
2179
2180 struct net_device_core_stats __percpu *core_stats;
2181
2182 /* Stats to monitor link on/off, flapping */
2183 atomic_t carrier_up_count;
2184 atomic_t carrier_down_count;
2185
2186 #ifdef CONFIG_WIRELESS_EXT
2187 const struct iw_handler_def *wireless_handlers;
2188 #endif
2189 const struct ethtool_ops *ethtool_ops;
2190 #ifdef CONFIG_NET_L3_MASTER_DEV
2191 const struct l3mdev_ops *l3mdev_ops;
2192 #endif
2193 #if IS_ENABLED(CONFIG_IPV6)
2194 const struct ndisc_ops *ndisc_ops;
2195 #endif
2196
2197 #ifdef CONFIG_XFRM_OFFLOAD
2198 const struct xfrmdev_ops *xfrmdev_ops;
2199 #endif
2200
2201 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2202 const struct tlsdev_ops *tlsdev_ops;
2203 #endif
2204
2205 unsigned int operstate;
2206 unsigned char link_mode;
2207
2208 unsigned char if_port;
2209 unsigned char dma;
2210
2211 /* Interface address info. */
2212 unsigned char perm_addr[MAX_ADDR_LEN];
2213 unsigned char addr_assign_type;
2214 unsigned char addr_len;
2215 unsigned char upper_level;
2216 unsigned char lower_level;
2217
2218 unsigned short neigh_priv_len;
2219 unsigned short dev_id;
2220 unsigned short dev_port;
2221 int irq;
2222 u32 priv_len;
2223
2224 spinlock_t addr_list_lock;
2225
2226 struct netdev_hw_addr_list uc;
2227 struct netdev_hw_addr_list mc;
2228 struct netdev_hw_addr_list dev_addrs;
2229
2230 #ifdef CONFIG_SYSFS
2231 struct kset *queues_kset;
2232 #endif
2233 #ifdef CONFIG_LOCKDEP
2234 struct list_head unlink_list;
2235 #endif
2236 unsigned int promiscuity;
2237 unsigned int allmulti;
2238 bool uc_promisc;
2239 #ifdef CONFIG_LOCKDEP
2240 unsigned char nested_level;
2241 #endif
2242
2243
2244 /* Protocol-specific pointers */
2245 struct in_device __rcu *ip_ptr;
2246 /** @fib_nh_head: nexthops associated with this netdev */
2247 struct hlist_head fib_nh_head;
2248
2249 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2250 struct vlan_info __rcu *vlan_info;
2251 #endif
2252 #if IS_ENABLED(CONFIG_NET_DSA)
2253 struct dsa_port *dsa_ptr;
2254 #endif
2255 #if IS_ENABLED(CONFIG_TIPC)
2256 struct tipc_bearer __rcu *tipc_ptr;
2257 #endif
2258 #if IS_ENABLED(CONFIG_ATALK)
2259 void *atalk_ptr;
2260 #endif
2261 #if IS_ENABLED(CONFIG_AX25)
2262 void *ax25_ptr;
2263 #endif
2264 #if IS_ENABLED(CONFIG_CFG80211)
2265 struct wireless_dev *ieee80211_ptr;
2266 #endif
2267 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2268 struct wpan_dev *ieee802154_ptr;
2269 #endif
2270 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2271 struct mpls_dev __rcu *mpls_ptr;
2272 #endif
2273 #if IS_ENABLED(CONFIG_MCTP)
2274 struct mctp_dev __rcu *mctp_ptr;
2275 #endif
2276
2277 /*
2278 * Cache lines mostly used on receive path (including eth_type_trans())
2279 */
2280 /* Interface address info used in eth_type_trans() */
2281 const unsigned char *dev_addr;
2282
2283 unsigned int num_rx_queues;
2284 #define GRO_LEGACY_MAX_SIZE 65536u
2285 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2286 * and shinfo->gso_segs is a 16bit field.
2287 */
2288 #define GRO_MAX_SIZE (8 * 65535u)
2289 unsigned int xdp_zc_max_segs;
2290 struct netdev_queue __rcu *ingress_queue;
2291 #ifdef CONFIG_NETFILTER_INGRESS
2292 struct nf_hook_entries __rcu *nf_hooks_ingress;
2293 #endif
2294
2295 unsigned char broadcast[MAX_ADDR_LEN];
2296 #ifdef CONFIG_RFS_ACCEL
2297 struct cpu_rmap *rx_cpu_rmap;
2298 #endif
2299 struct hlist_node index_hlist;
2300
2301 /*
2302 * Cache lines mostly used on transmit path
2303 */
2304 unsigned int num_tx_queues;
2305 struct Qdisc __rcu *qdisc;
2306 unsigned int tx_queue_len;
2307 spinlock_t tx_global_lock;
2308
2309 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2310
2311 #ifdef CONFIG_NET_SCHED
2312 DECLARE_HASHTABLE (qdisc_hash, 4);
2313 #endif
2314 /* These may be needed for future network-power-down code. */
2315 struct timer_list watchdog_timer;
2316 int watchdog_timeo;
2317
2318 u32 proto_down_reason;
2319
2320 struct list_head todo_list;
2321
2322 #ifdef CONFIG_PCPU_DEV_REFCNT
2323 int __percpu *pcpu_refcnt;
2324 #else
2325 refcount_t dev_refcnt;
2326 #endif
2327 struct ref_tracker_dir refcnt_tracker;
2328
2329 struct list_head link_watch_list;
2330
2331 u8 reg_state;
2332
2333 bool dismantle;
2334
2335 enum {
2336 RTNL_LINK_INITIALIZED,
2337 RTNL_LINK_INITIALIZING,
2338 } rtnl_link_state:16;
2339
2340 bool needs_free_netdev;
2341 void (*priv_destructor)(struct net_device *dev);
2342
2343 /* mid-layer private */
2344 void *ml_priv;
2345 enum netdev_ml_priv_type ml_priv_type;
2346
2347 enum netdev_stat_type pcpu_stat_type:8;
2348
2349 #if IS_ENABLED(CONFIG_GARP)
2350 struct garp_port __rcu *garp_port;
2351 #endif
2352 #if IS_ENABLED(CONFIG_MRP)
2353 struct mrp_port __rcu *mrp_port;
2354 #endif
2355 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2356 struct dm_hw_stat_delta __rcu *dm_private;
2357 #endif
2358 struct device dev;
2359 const struct attribute_group *sysfs_groups[4];
2360 const struct attribute_group *sysfs_rx_queue_group;
2361
2362 const struct rtnl_link_ops *rtnl_link_ops;
2363
2364 const struct netdev_stat_ops *stat_ops;
2365
2366 const struct netdev_queue_mgmt_ops *queue_mgmt_ops;
2367
2368 /* for setting kernel sock attribute on TCP connection setup */
2369 #define GSO_MAX_SEGS 65535u
2370 #define GSO_LEGACY_MAX_SIZE 65536u
2371 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2372 * and shinfo->gso_segs is a 16bit field.
2373 */
2374 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2375
2376 #define TSO_LEGACY_MAX_SIZE 65536
2377 #define TSO_MAX_SIZE UINT_MAX
2378 unsigned int tso_max_size;
2379 #define TSO_MAX_SEGS U16_MAX
2380 u16 tso_max_segs;
2381
2382 #ifdef CONFIG_DCB
2383 const struct dcbnl_rtnl_ops *dcbnl_ops;
2384 #endif
2385 u8 prio_tc_map[TC_BITMASK + 1];
2386
2387 #if IS_ENABLED(CONFIG_FCOE)
2388 unsigned int fcoe_ddp_xid;
2389 #endif
2390 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2391 struct netprio_map __rcu *priomap;
2392 #endif
2393 struct phy_link_topology *link_topo;
2394 struct phy_device *phydev;
2395 struct sfp_bus *sfp_bus;
2396 struct lock_class_key *qdisc_tx_busylock;
2397 bool proto_down;
2398 bool threaded;
2399
2400 /* priv_flags_slow, ungrouped to save space */
2401 unsigned long see_all_hwtstamp_requests:1;
2402 unsigned long change_proto_down:1;
2403 unsigned long netns_local:1;
2404 unsigned long fcoe_mtu:1;
2405
2406 struct list_head net_notifier_list;
2407
2408 #if IS_ENABLED(CONFIG_MACSEC)
2409 /* MACsec management functions */
2410 const struct macsec_ops *macsec_ops;
2411 #endif
2412 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2413 struct udp_tunnel_nic *udp_tunnel_nic;
2414
2415 struct ethtool_netdev_state *ethtool;
2416
2417 /* protected by rtnl_lock */
2418 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2419
2420 u8 dev_addr_shadow[MAX_ADDR_LEN];
2421 netdevice_tracker linkwatch_dev_tracker;
2422 netdevice_tracker watchdog_dev_tracker;
2423 netdevice_tracker dev_registered_tracker;
2424 struct rtnl_hw_stats64 *offload_xstats_l3;
2425
2426 struct devlink_port *devlink_port;
2427
2428 #if IS_ENABLED(CONFIG_DPLL)
2429 struct dpll_pin __rcu *dpll_pin;
2430 #endif
2431 #if IS_ENABLED(CONFIG_PAGE_POOL)
2432 /** @page_pools: page pools created for this netdevice */
2433 struct hlist_head page_pools;
2434 #endif
2435
2436 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */
2437 struct dim_irq_moder *irq_moder;
2438
2439 u64 max_pacing_offload_horizon;
2440 struct napi_config *napi_config;
2441 unsigned long gro_flush_timeout;
2442 u32 napi_defer_hard_irqs;
2443
2444 /**
2445 * @lock: protects @net_shaper_hierarchy, feel free to use for other
2446 * netdev-scope protection. Ordering: take after rtnl_lock.
2447 */
2448 struct mutex lock;
2449
2450 #if IS_ENABLED(CONFIG_NET_SHAPER)
2451 /**
2452 * @net_shaper_hierarchy: data tracking the current shaper status
2453 * see include/net/net_shapers.h
2454 */
2455 struct net_shaper_hierarchy *net_shaper_hierarchy;
2456 #endif
2457
2458 struct hlist_head neighbours[NEIGH_NR_TABLES];
2459
2460 u8 priv[] ____cacheline_aligned
2461 __counted_by(priv_len);
2462 } ____cacheline_aligned;
2463 #define to_net_dev(d) container_of(d, struct net_device, dev)
2464
2465 /*
2466 * Driver should use this to assign devlink port instance to a netdevice
2467 * before it registers the netdevice. Therefore devlink_port is static
2468 * during the netdev lifetime after it is registered.
2469 */
2470 #define SET_NETDEV_DEVLINK_PORT(dev, port) \
2471 ({ \
2472 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2473 ((dev)->devlink_port = (port)); \
2474 })
2475
netif_elide_gro(const struct net_device * dev)2476 static inline bool netif_elide_gro(const struct net_device *dev)
2477 {
2478 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2479 return true;
2480 return false;
2481 }
2482
2483 #define NETDEV_ALIGN 32
2484
2485 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)2486 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2487 {
2488 return dev->prio_tc_map[prio & TC_BITMASK];
2489 }
2490
2491 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)2492 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2493 {
2494 if (tc >= dev->num_tc)
2495 return -EINVAL;
2496
2497 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2498 return 0;
2499 }
2500
2501 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2502 void netdev_reset_tc(struct net_device *dev);
2503 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2504 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2505
2506 static inline
netdev_get_num_tc(struct net_device * dev)2507 int netdev_get_num_tc(struct net_device *dev)
2508 {
2509 return dev->num_tc;
2510 }
2511
net_prefetch(void * p)2512 static inline void net_prefetch(void *p)
2513 {
2514 prefetch(p);
2515 #if L1_CACHE_BYTES < 128
2516 prefetch((u8 *)p + L1_CACHE_BYTES);
2517 #endif
2518 }
2519
net_prefetchw(void * p)2520 static inline void net_prefetchw(void *p)
2521 {
2522 prefetchw(p);
2523 #if L1_CACHE_BYTES < 128
2524 prefetchw((u8 *)p + L1_CACHE_BYTES);
2525 #endif
2526 }
2527
2528 void netdev_unbind_sb_channel(struct net_device *dev,
2529 struct net_device *sb_dev);
2530 int netdev_bind_sb_channel_queue(struct net_device *dev,
2531 struct net_device *sb_dev,
2532 u8 tc, u16 count, u16 offset);
2533 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
netdev_get_sb_channel(struct net_device * dev)2534 static inline int netdev_get_sb_channel(struct net_device *dev)
2535 {
2536 return max_t(int, -dev->num_tc, 0);
2537 }
2538
2539 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)2540 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2541 unsigned int index)
2542 {
2543 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2544 return &dev->_tx[index];
2545 }
2546
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)2547 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2548 const struct sk_buff *skb)
2549 {
2550 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2551 }
2552
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)2553 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2554 void (*f)(struct net_device *,
2555 struct netdev_queue *,
2556 void *),
2557 void *arg)
2558 {
2559 unsigned int i;
2560
2561 for (i = 0; i < dev->num_tx_queues; i++)
2562 f(dev, &dev->_tx[i], arg);
2563 }
2564
2565 #define netdev_lockdep_set_classes(dev) \
2566 { \
2567 static struct lock_class_key qdisc_tx_busylock_key; \
2568 static struct lock_class_key qdisc_xmit_lock_key; \
2569 static struct lock_class_key dev_addr_list_lock_key; \
2570 unsigned int i; \
2571 \
2572 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2573 lockdep_set_class(&(dev)->addr_list_lock, \
2574 &dev_addr_list_lock_key); \
2575 for (i = 0; i < (dev)->num_tx_queues; i++) \
2576 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2577 &qdisc_xmit_lock_key); \
2578 }
2579
2580 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2581 struct net_device *sb_dev);
2582 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2583 struct sk_buff *skb,
2584 struct net_device *sb_dev);
2585
2586 /* returns the headroom that the master device needs to take in account
2587 * when forwarding to this dev
2588 */
netdev_get_fwd_headroom(struct net_device * dev)2589 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2590 {
2591 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2592 }
2593
netdev_set_rx_headroom(struct net_device * dev,int new_hr)2594 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2595 {
2596 if (dev->netdev_ops->ndo_set_rx_headroom)
2597 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2598 }
2599
2600 /* set the device rx headroom to the dev's default */
netdev_reset_rx_headroom(struct net_device * dev)2601 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2602 {
2603 netdev_set_rx_headroom(dev, -1);
2604 }
2605
netdev_get_ml_priv(struct net_device * dev,enum netdev_ml_priv_type type)2606 static inline void *netdev_get_ml_priv(struct net_device *dev,
2607 enum netdev_ml_priv_type type)
2608 {
2609 if (dev->ml_priv_type != type)
2610 return NULL;
2611
2612 return dev->ml_priv;
2613 }
2614
netdev_set_ml_priv(struct net_device * dev,void * ml_priv,enum netdev_ml_priv_type type)2615 static inline void netdev_set_ml_priv(struct net_device *dev,
2616 void *ml_priv,
2617 enum netdev_ml_priv_type type)
2618 {
2619 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2620 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2621 dev->ml_priv_type, type);
2622 WARN(!dev->ml_priv_type && dev->ml_priv,
2623 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2624
2625 dev->ml_priv = ml_priv;
2626 dev->ml_priv_type = type;
2627 }
2628
2629 /*
2630 * Net namespace inlines
2631 */
2632 static inline
dev_net(const struct net_device * dev)2633 struct net *dev_net(const struct net_device *dev)
2634 {
2635 return read_pnet(&dev->nd_net);
2636 }
2637
2638 static inline
dev_net_set(struct net_device * dev,struct net * net)2639 void dev_net_set(struct net_device *dev, struct net *net)
2640 {
2641 write_pnet(&dev->nd_net, net);
2642 }
2643
2644 /**
2645 * netdev_priv - access network device private data
2646 * @dev: network device
2647 *
2648 * Get network device private data
2649 */
netdev_priv(const struct net_device * dev)2650 static inline void *netdev_priv(const struct net_device *dev)
2651 {
2652 return (void *)dev->priv;
2653 }
2654
2655 /* Set the sysfs physical device reference for the network logical device
2656 * if set prior to registration will cause a symlink during initialization.
2657 */
2658 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2659
2660 /* Set the sysfs device type for the network logical device to allow
2661 * fine-grained identification of different network device types. For
2662 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2663 */
2664 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2665
2666 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2667 enum netdev_queue_type type,
2668 struct napi_struct *napi);
2669
netif_napi_set_irq(struct napi_struct * napi,int irq)2670 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2671 {
2672 napi->irq = irq;
2673 }
2674
2675 /* Default NAPI poll() weight
2676 * Device drivers are strongly advised to not use bigger value
2677 */
2678 #define NAPI_POLL_WEIGHT 64
2679
2680 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2681 int (*poll)(struct napi_struct *, int), int weight);
2682
2683 /**
2684 * netif_napi_add() - initialize a NAPI context
2685 * @dev: network device
2686 * @napi: NAPI context
2687 * @poll: polling function
2688 *
2689 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2690 * *any* of the other NAPI-related functions.
2691 */
2692 static inline void
netif_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int))2693 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2694 int (*poll)(struct napi_struct *, int))
2695 {
2696 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2697 }
2698
2699 static inline void
netif_napi_add_tx_weight(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2700 netif_napi_add_tx_weight(struct net_device *dev,
2701 struct napi_struct *napi,
2702 int (*poll)(struct napi_struct *, int),
2703 int weight)
2704 {
2705 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2706 netif_napi_add_weight(dev, napi, poll, weight);
2707 }
2708
2709 /**
2710 * netif_napi_add_config - initialize a NAPI context with persistent config
2711 * @dev: network device
2712 * @napi: NAPI context
2713 * @poll: polling function
2714 * @index: the NAPI index
2715 */
2716 static inline void
netif_napi_add_config(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int index)2717 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi,
2718 int (*poll)(struct napi_struct *, int), int index)
2719 {
2720 napi->index = index;
2721 napi->config = &dev->napi_config[index];
2722 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2723 }
2724
2725 /**
2726 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2727 * @dev: network device
2728 * @napi: NAPI context
2729 * @poll: polling function
2730 *
2731 * This variant of netif_napi_add() should be used from drivers using NAPI
2732 * to exclusively poll a TX queue.
2733 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2734 */
netif_napi_add_tx(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int))2735 static inline void netif_napi_add_tx(struct net_device *dev,
2736 struct napi_struct *napi,
2737 int (*poll)(struct napi_struct *, int))
2738 {
2739 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2740 }
2741
2742 /**
2743 * __netif_napi_del - remove a NAPI context
2744 * @napi: NAPI context
2745 *
2746 * Warning: caller must observe RCU grace period before freeing memory
2747 * containing @napi. Drivers might want to call this helper to combine
2748 * all the needed RCU grace periods into a single one.
2749 */
2750 void __netif_napi_del(struct napi_struct *napi);
2751
2752 /**
2753 * netif_napi_del - remove a NAPI context
2754 * @napi: NAPI context
2755 *
2756 * netif_napi_del() removes a NAPI context from the network device NAPI list
2757 */
netif_napi_del(struct napi_struct * napi)2758 static inline void netif_napi_del(struct napi_struct *napi)
2759 {
2760 __netif_napi_del(napi);
2761 synchronize_net();
2762 }
2763
2764 struct packet_type {
2765 __be16 type; /* This is really htons(ether_type). */
2766 bool ignore_outgoing;
2767 struct net_device *dev; /* NULL is wildcarded here */
2768 netdevice_tracker dev_tracker;
2769 int (*func) (struct sk_buff *,
2770 struct net_device *,
2771 struct packet_type *,
2772 struct net_device *);
2773 void (*list_func) (struct list_head *,
2774 struct packet_type *,
2775 struct net_device *);
2776 bool (*id_match)(struct packet_type *ptype,
2777 struct sock *sk);
2778 struct net *af_packet_net;
2779 void *af_packet_priv;
2780 struct list_head list;
2781 };
2782
2783 struct offload_callbacks {
2784 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2785 netdev_features_t features);
2786 struct sk_buff *(*gro_receive)(struct list_head *head,
2787 struct sk_buff *skb);
2788 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2789 };
2790
2791 struct packet_offload {
2792 __be16 type; /* This is really htons(ether_type). */
2793 u16 priority;
2794 struct offload_callbacks callbacks;
2795 struct list_head list;
2796 };
2797
2798 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2799 struct pcpu_sw_netstats {
2800 u64_stats_t rx_packets;
2801 u64_stats_t rx_bytes;
2802 u64_stats_t tx_packets;
2803 u64_stats_t tx_bytes;
2804 struct u64_stats_sync syncp;
2805 } __aligned(4 * sizeof(u64));
2806
2807 struct pcpu_dstats {
2808 u64_stats_t rx_packets;
2809 u64_stats_t rx_bytes;
2810 u64_stats_t rx_drops;
2811 u64_stats_t tx_packets;
2812 u64_stats_t tx_bytes;
2813 u64_stats_t tx_drops;
2814 struct u64_stats_sync syncp;
2815 } __aligned(8 * sizeof(u64));
2816
2817 struct pcpu_lstats {
2818 u64_stats_t packets;
2819 u64_stats_t bytes;
2820 struct u64_stats_sync syncp;
2821 } __aligned(2 * sizeof(u64));
2822
2823 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2824
dev_sw_netstats_rx_add(struct net_device * dev,unsigned int len)2825 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2826 {
2827 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2828
2829 u64_stats_update_begin(&tstats->syncp);
2830 u64_stats_add(&tstats->rx_bytes, len);
2831 u64_stats_inc(&tstats->rx_packets);
2832 u64_stats_update_end(&tstats->syncp);
2833 }
2834
dev_sw_netstats_tx_add(struct net_device * dev,unsigned int packets,unsigned int len)2835 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2836 unsigned int packets,
2837 unsigned int len)
2838 {
2839 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2840
2841 u64_stats_update_begin(&tstats->syncp);
2842 u64_stats_add(&tstats->tx_bytes, len);
2843 u64_stats_add(&tstats->tx_packets, packets);
2844 u64_stats_update_end(&tstats->syncp);
2845 }
2846
dev_lstats_add(struct net_device * dev,unsigned int len)2847 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2848 {
2849 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2850
2851 u64_stats_update_begin(&lstats->syncp);
2852 u64_stats_add(&lstats->bytes, len);
2853 u64_stats_inc(&lstats->packets);
2854 u64_stats_update_end(&lstats->syncp);
2855 }
2856
2857 #define __netdev_alloc_pcpu_stats(type, gfp) \
2858 ({ \
2859 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2860 if (pcpu_stats) { \
2861 int __cpu; \
2862 for_each_possible_cpu(__cpu) { \
2863 typeof(type) *stat; \
2864 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2865 u64_stats_init(&stat->syncp); \
2866 } \
2867 } \
2868 pcpu_stats; \
2869 })
2870
2871 #define netdev_alloc_pcpu_stats(type) \
2872 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2873
2874 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2875 ({ \
2876 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2877 if (pcpu_stats) { \
2878 int __cpu; \
2879 for_each_possible_cpu(__cpu) { \
2880 typeof(type) *stat; \
2881 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2882 u64_stats_init(&stat->syncp); \
2883 } \
2884 } \
2885 pcpu_stats; \
2886 })
2887
2888 enum netdev_lag_tx_type {
2889 NETDEV_LAG_TX_TYPE_UNKNOWN,
2890 NETDEV_LAG_TX_TYPE_RANDOM,
2891 NETDEV_LAG_TX_TYPE_BROADCAST,
2892 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2893 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2894 NETDEV_LAG_TX_TYPE_HASH,
2895 };
2896
2897 enum netdev_lag_hash {
2898 NETDEV_LAG_HASH_NONE,
2899 NETDEV_LAG_HASH_L2,
2900 NETDEV_LAG_HASH_L34,
2901 NETDEV_LAG_HASH_L23,
2902 NETDEV_LAG_HASH_E23,
2903 NETDEV_LAG_HASH_E34,
2904 NETDEV_LAG_HASH_VLAN_SRCMAC,
2905 NETDEV_LAG_HASH_UNKNOWN,
2906 };
2907
2908 struct netdev_lag_upper_info {
2909 enum netdev_lag_tx_type tx_type;
2910 enum netdev_lag_hash hash_type;
2911 };
2912
2913 struct netdev_lag_lower_state_info {
2914 u8 link_up : 1,
2915 tx_enabled : 1;
2916 };
2917
2918 #include <linux/notifier.h>
2919
2920 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2921 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2922 * adding new types.
2923 */
2924 enum netdev_cmd {
2925 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2926 NETDEV_DOWN,
2927 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2928 detected a hardware crash and restarted
2929 - we can use this eg to kick tcp sessions
2930 once done */
2931 NETDEV_CHANGE, /* Notify device state change */
2932 NETDEV_REGISTER,
2933 NETDEV_UNREGISTER,
2934 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2935 NETDEV_CHANGEADDR, /* notify after the address change */
2936 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2937 NETDEV_GOING_DOWN,
2938 NETDEV_CHANGENAME,
2939 NETDEV_FEAT_CHANGE,
2940 NETDEV_BONDING_FAILOVER,
2941 NETDEV_PRE_UP,
2942 NETDEV_PRE_TYPE_CHANGE,
2943 NETDEV_POST_TYPE_CHANGE,
2944 NETDEV_POST_INIT,
2945 NETDEV_PRE_UNINIT,
2946 NETDEV_RELEASE,
2947 NETDEV_NOTIFY_PEERS,
2948 NETDEV_JOIN,
2949 NETDEV_CHANGEUPPER,
2950 NETDEV_RESEND_IGMP,
2951 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2952 NETDEV_CHANGEINFODATA,
2953 NETDEV_BONDING_INFO,
2954 NETDEV_PRECHANGEUPPER,
2955 NETDEV_CHANGELOWERSTATE,
2956 NETDEV_UDP_TUNNEL_PUSH_INFO,
2957 NETDEV_UDP_TUNNEL_DROP_INFO,
2958 NETDEV_CHANGE_TX_QUEUE_LEN,
2959 NETDEV_CVLAN_FILTER_PUSH_INFO,
2960 NETDEV_CVLAN_FILTER_DROP_INFO,
2961 NETDEV_SVLAN_FILTER_PUSH_INFO,
2962 NETDEV_SVLAN_FILTER_DROP_INFO,
2963 NETDEV_OFFLOAD_XSTATS_ENABLE,
2964 NETDEV_OFFLOAD_XSTATS_DISABLE,
2965 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2966 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2967 NETDEV_XDP_FEAT_CHANGE,
2968 };
2969 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2970
2971 int register_netdevice_notifier(struct notifier_block *nb);
2972 int unregister_netdevice_notifier(struct notifier_block *nb);
2973 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2974 int unregister_netdevice_notifier_net(struct net *net,
2975 struct notifier_block *nb);
2976 int register_netdevice_notifier_dev_net(struct net_device *dev,
2977 struct notifier_block *nb,
2978 struct netdev_net_notifier *nn);
2979 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2980 struct notifier_block *nb,
2981 struct netdev_net_notifier *nn);
2982
2983 struct netdev_notifier_info {
2984 struct net_device *dev;
2985 struct netlink_ext_ack *extack;
2986 };
2987
2988 struct netdev_notifier_info_ext {
2989 struct netdev_notifier_info info; /* must be first */
2990 union {
2991 u32 mtu;
2992 } ext;
2993 };
2994
2995 struct netdev_notifier_change_info {
2996 struct netdev_notifier_info info; /* must be first */
2997 unsigned int flags_changed;
2998 };
2999
3000 struct netdev_notifier_changeupper_info {
3001 struct netdev_notifier_info info; /* must be first */
3002 struct net_device *upper_dev; /* new upper dev */
3003 bool master; /* is upper dev master */
3004 bool linking; /* is the notification for link or unlink */
3005 void *upper_info; /* upper dev info */
3006 };
3007
3008 struct netdev_notifier_changelowerstate_info {
3009 struct netdev_notifier_info info; /* must be first */
3010 void *lower_state_info; /* is lower dev state */
3011 };
3012
3013 struct netdev_notifier_pre_changeaddr_info {
3014 struct netdev_notifier_info info; /* must be first */
3015 const unsigned char *dev_addr;
3016 };
3017
3018 enum netdev_offload_xstats_type {
3019 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
3020 };
3021
3022 struct netdev_notifier_offload_xstats_info {
3023 struct netdev_notifier_info info; /* must be first */
3024 enum netdev_offload_xstats_type type;
3025
3026 union {
3027 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
3028 struct netdev_notifier_offload_xstats_rd *report_delta;
3029 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
3030 struct netdev_notifier_offload_xstats_ru *report_used;
3031 };
3032 };
3033
3034 int netdev_offload_xstats_enable(struct net_device *dev,
3035 enum netdev_offload_xstats_type type,
3036 struct netlink_ext_ack *extack);
3037 int netdev_offload_xstats_disable(struct net_device *dev,
3038 enum netdev_offload_xstats_type type);
3039 bool netdev_offload_xstats_enabled(const struct net_device *dev,
3040 enum netdev_offload_xstats_type type);
3041 int netdev_offload_xstats_get(struct net_device *dev,
3042 enum netdev_offload_xstats_type type,
3043 struct rtnl_hw_stats64 *stats, bool *used,
3044 struct netlink_ext_ack *extack);
3045 void
3046 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3047 const struct rtnl_hw_stats64 *stats);
3048 void
3049 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3050 void netdev_offload_xstats_push_delta(struct net_device *dev,
3051 enum netdev_offload_xstats_type type,
3052 const struct rtnl_hw_stats64 *stats);
3053
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)3054 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3055 struct net_device *dev)
3056 {
3057 info->dev = dev;
3058 info->extack = NULL;
3059 }
3060
3061 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)3062 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3063 {
3064 return info->dev;
3065 }
3066
3067 static inline struct netlink_ext_ack *
netdev_notifier_info_to_extack(const struct netdev_notifier_info * info)3068 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3069 {
3070 return info->extack;
3071 }
3072
3073 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3074 int call_netdevice_notifiers_info(unsigned long val,
3075 struct netdev_notifier_info *info);
3076
3077 #define for_each_netdev(net, d) \
3078 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3079 #define for_each_netdev_reverse(net, d) \
3080 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3081 #define for_each_netdev_rcu(net, d) \
3082 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3083 #define for_each_netdev_safe(net, d, n) \
3084 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3085 #define for_each_netdev_continue(net, d) \
3086 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3087 #define for_each_netdev_continue_reverse(net, d) \
3088 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3089 dev_list)
3090 #define for_each_netdev_continue_rcu(net, d) \
3091 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3092 #define for_each_netdev_in_bond_rcu(bond, slave) \
3093 for_each_netdev_rcu(&init_net, slave) \
3094 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3095 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3096
3097 #define for_each_netdev_dump(net, d, ifindex) \
3098 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \
3099 ULONG_MAX, XA_PRESENT)); ifindex++)
3100
next_net_device(struct net_device * dev)3101 static inline struct net_device *next_net_device(struct net_device *dev)
3102 {
3103 struct list_head *lh;
3104 struct net *net;
3105
3106 net = dev_net(dev);
3107 lh = dev->dev_list.next;
3108 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3109 }
3110
next_net_device_rcu(struct net_device * dev)3111 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3112 {
3113 struct list_head *lh;
3114 struct net *net;
3115
3116 net = dev_net(dev);
3117 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3118 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3119 }
3120
first_net_device(struct net * net)3121 static inline struct net_device *first_net_device(struct net *net)
3122 {
3123 return list_empty(&net->dev_base_head) ? NULL :
3124 net_device_entry(net->dev_base_head.next);
3125 }
3126
first_net_device_rcu(struct net * net)3127 static inline struct net_device *first_net_device_rcu(struct net *net)
3128 {
3129 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3130
3131 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3132 }
3133
3134 int netdev_boot_setup_check(struct net_device *dev);
3135 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3136 const char *hwaddr);
3137 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3138 void dev_add_pack(struct packet_type *pt);
3139 void dev_remove_pack(struct packet_type *pt);
3140 void __dev_remove_pack(struct packet_type *pt);
3141 void dev_add_offload(struct packet_offload *po);
3142 void dev_remove_offload(struct packet_offload *po);
3143
3144 int dev_get_iflink(const struct net_device *dev);
3145 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3146 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3147 struct net_device_path_stack *stack);
3148 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3149 unsigned short mask);
3150 struct net_device *dev_get_by_name(struct net *net, const char *name);
3151 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3152 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3153 bool netdev_name_in_use(struct net *net, const char *name);
3154 int dev_alloc_name(struct net_device *dev, const char *name);
3155 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3156 void dev_close(struct net_device *dev);
3157 void dev_close_many(struct list_head *head, bool unlink);
3158 void dev_disable_lro(struct net_device *dev);
3159 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3160 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3161 struct net_device *sb_dev);
3162
3163 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3164 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3165
dev_queue_xmit(struct sk_buff * skb)3166 static inline int dev_queue_xmit(struct sk_buff *skb)
3167 {
3168 return __dev_queue_xmit(skb, NULL);
3169 }
3170
dev_queue_xmit_accel(struct sk_buff * skb,struct net_device * sb_dev)3171 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3172 struct net_device *sb_dev)
3173 {
3174 return __dev_queue_xmit(skb, sb_dev);
3175 }
3176
dev_direct_xmit(struct sk_buff * skb,u16 queue_id)3177 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3178 {
3179 int ret;
3180
3181 ret = __dev_direct_xmit(skb, queue_id);
3182 if (!dev_xmit_complete(ret))
3183 kfree_skb(skb);
3184 return ret;
3185 }
3186
3187 int register_netdevice(struct net_device *dev);
3188 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3189 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)3190 static inline void unregister_netdevice(struct net_device *dev)
3191 {
3192 unregister_netdevice_queue(dev, NULL);
3193 }
3194
3195 int netdev_refcnt_read(const struct net_device *dev);
3196 void free_netdev(struct net_device *dev);
3197 void init_dummy_netdev(struct net_device *dev);
3198
3199 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3200 struct sk_buff *skb,
3201 bool all_slaves);
3202 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3203 struct sock *sk);
3204 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3205 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3206 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3207 netdevice_tracker *tracker, gfp_t gfp);
3208 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3209 netdevice_tracker *tracker, gfp_t gfp);
3210 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3211 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3212 void netdev_copy_name(struct net_device *dev, char *name);
3213
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)3214 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3215 unsigned short type,
3216 const void *daddr, const void *saddr,
3217 unsigned int len)
3218 {
3219 if (!dev->header_ops || !dev->header_ops->create)
3220 return 0;
3221
3222 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3223 }
3224
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)3225 static inline int dev_parse_header(const struct sk_buff *skb,
3226 unsigned char *haddr)
3227 {
3228 const struct net_device *dev = skb->dev;
3229
3230 if (!dev->header_ops || !dev->header_ops->parse)
3231 return 0;
3232 return dev->header_ops->parse(skb, haddr);
3233 }
3234
dev_parse_header_protocol(const struct sk_buff * skb)3235 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3236 {
3237 const struct net_device *dev = skb->dev;
3238
3239 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3240 return 0;
3241 return dev->header_ops->parse_protocol(skb);
3242 }
3243
3244 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)3245 static inline bool dev_validate_header(const struct net_device *dev,
3246 char *ll_header, int len)
3247 {
3248 if (likely(len >= dev->hard_header_len))
3249 return true;
3250 if (len < dev->min_header_len)
3251 return false;
3252
3253 if (capable(CAP_SYS_RAWIO)) {
3254 memset(ll_header + len, 0, dev->hard_header_len - len);
3255 return true;
3256 }
3257
3258 if (dev->header_ops && dev->header_ops->validate)
3259 return dev->header_ops->validate(ll_header, len);
3260
3261 return false;
3262 }
3263
dev_has_header(const struct net_device * dev)3264 static inline bool dev_has_header(const struct net_device *dev)
3265 {
3266 return dev->header_ops && dev->header_ops->create;
3267 }
3268
3269 /*
3270 * Incoming packets are placed on per-CPU queues
3271 */
3272 struct softnet_data {
3273 struct list_head poll_list;
3274 struct sk_buff_head process_queue;
3275 local_lock_t process_queue_bh_lock;
3276
3277 /* stats */
3278 unsigned int processed;
3279 unsigned int time_squeeze;
3280 #ifdef CONFIG_RPS
3281 struct softnet_data *rps_ipi_list;
3282 #endif
3283
3284 unsigned int received_rps;
3285 bool in_net_rx_action;
3286 bool in_napi_threaded_poll;
3287
3288 #ifdef CONFIG_NET_FLOW_LIMIT
3289 struct sd_flow_limit __rcu *flow_limit;
3290 #endif
3291 struct Qdisc *output_queue;
3292 struct Qdisc **output_queue_tailp;
3293 struct sk_buff *completion_queue;
3294 #ifdef CONFIG_XFRM_OFFLOAD
3295 struct sk_buff_head xfrm_backlog;
3296 #endif
3297 /* written and read only by owning cpu: */
3298 struct netdev_xmit xmit;
3299 #ifdef CONFIG_RPS
3300 /* input_queue_head should be written by cpu owning this struct,
3301 * and only read by other cpus. Worth using a cache line.
3302 */
3303 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3304
3305 /* Elements below can be accessed between CPUs for RPS/RFS */
3306 call_single_data_t csd ____cacheline_aligned_in_smp;
3307 struct softnet_data *rps_ipi_next;
3308 unsigned int cpu;
3309 unsigned int input_queue_tail;
3310 #endif
3311 struct sk_buff_head input_pkt_queue;
3312 struct napi_struct backlog;
3313
3314 atomic_t dropped ____cacheline_aligned_in_smp;
3315
3316 /* Another possibly contended cache line */
3317 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3318 int defer_count;
3319 int defer_ipi_scheduled;
3320 struct sk_buff *defer_list;
3321 call_single_data_t defer_csd;
3322 };
3323
3324 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3325
3326 #ifndef CONFIG_PREEMPT_RT
dev_recursion_level(void)3327 static inline int dev_recursion_level(void)
3328 {
3329 return this_cpu_read(softnet_data.xmit.recursion);
3330 }
3331 #else
dev_recursion_level(void)3332 static inline int dev_recursion_level(void)
3333 {
3334 return current->net_xmit.recursion;
3335 }
3336
3337 #endif
3338
3339 void __netif_schedule(struct Qdisc *q);
3340 void netif_schedule_queue(struct netdev_queue *txq);
3341
netif_tx_schedule_all(struct net_device * dev)3342 static inline void netif_tx_schedule_all(struct net_device *dev)
3343 {
3344 unsigned int i;
3345
3346 for (i = 0; i < dev->num_tx_queues; i++)
3347 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3348 }
3349
netif_tx_start_queue(struct netdev_queue * dev_queue)3350 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3351 {
3352 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3353 }
3354
3355 /**
3356 * netif_start_queue - allow transmit
3357 * @dev: network device
3358 *
3359 * Allow upper layers to call the device hard_start_xmit routine.
3360 */
netif_start_queue(struct net_device * dev)3361 static inline void netif_start_queue(struct net_device *dev)
3362 {
3363 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3364 }
3365
netif_tx_start_all_queues(struct net_device * dev)3366 static inline void netif_tx_start_all_queues(struct net_device *dev)
3367 {
3368 unsigned int i;
3369
3370 for (i = 0; i < dev->num_tx_queues; i++) {
3371 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3372 netif_tx_start_queue(txq);
3373 }
3374 }
3375
3376 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3377
3378 /**
3379 * netif_wake_queue - restart transmit
3380 * @dev: network device
3381 *
3382 * Allow upper layers to call the device hard_start_xmit routine.
3383 * Used for flow control when transmit resources are available.
3384 */
netif_wake_queue(struct net_device * dev)3385 static inline void netif_wake_queue(struct net_device *dev)
3386 {
3387 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3388 }
3389
netif_tx_wake_all_queues(struct net_device * dev)3390 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3391 {
3392 unsigned int i;
3393
3394 for (i = 0; i < dev->num_tx_queues; i++) {
3395 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3396 netif_tx_wake_queue(txq);
3397 }
3398 }
3399
netif_tx_stop_queue(struct netdev_queue * dev_queue)3400 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3401 {
3402 /* Paired with READ_ONCE() from dev_watchdog() */
3403 WRITE_ONCE(dev_queue->trans_start, jiffies);
3404
3405 /* This barrier is paired with smp_mb() from dev_watchdog() */
3406 smp_mb__before_atomic();
3407
3408 /* Must be an atomic op see netif_txq_try_stop() */
3409 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3410 }
3411
3412 /**
3413 * netif_stop_queue - stop transmitted packets
3414 * @dev: network device
3415 *
3416 * Stop upper layers calling the device hard_start_xmit routine.
3417 * Used for flow control when transmit resources are unavailable.
3418 */
netif_stop_queue(struct net_device * dev)3419 static inline void netif_stop_queue(struct net_device *dev)
3420 {
3421 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3422 }
3423
3424 void netif_tx_stop_all_queues(struct net_device *dev);
3425
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)3426 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3427 {
3428 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3429 }
3430
3431 /**
3432 * netif_queue_stopped - test if transmit queue is flowblocked
3433 * @dev: network device
3434 *
3435 * Test if transmit queue on device is currently unable to send.
3436 */
netif_queue_stopped(const struct net_device * dev)3437 static inline bool netif_queue_stopped(const struct net_device *dev)
3438 {
3439 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3440 }
3441
netif_xmit_stopped(const struct netdev_queue * dev_queue)3442 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3443 {
3444 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3445 }
3446
3447 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)3448 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3449 {
3450 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3451 }
3452
3453 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)3454 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3455 {
3456 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3457 }
3458
3459 /**
3460 * netdev_queue_set_dql_min_limit - set dql minimum limit
3461 * @dev_queue: pointer to transmit queue
3462 * @min_limit: dql minimum limit
3463 *
3464 * Forces xmit_more() to return true until the minimum threshold
3465 * defined by @min_limit is reached (or until the tx queue is
3466 * empty). Warning: to be use with care, misuse will impact the
3467 * latency.
3468 */
netdev_queue_set_dql_min_limit(struct netdev_queue * dev_queue,unsigned int min_limit)3469 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3470 unsigned int min_limit)
3471 {
3472 #ifdef CONFIG_BQL
3473 dev_queue->dql.min_limit = min_limit;
3474 #endif
3475 }
3476
netdev_queue_dql_avail(const struct netdev_queue * txq)3477 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq)
3478 {
3479 #ifdef CONFIG_BQL
3480 /* Non-BQL migrated drivers will return 0, too. */
3481 return dql_avail(&txq->dql);
3482 #else
3483 return 0;
3484 #endif
3485 }
3486
3487 /**
3488 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3489 * @dev_queue: pointer to transmit queue
3490 *
3491 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3492 * to give appropriate hint to the CPU.
3493 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)3494 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3495 {
3496 #ifdef CONFIG_BQL
3497 prefetchw(&dev_queue->dql.num_queued);
3498 #endif
3499 }
3500
3501 /**
3502 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3503 * @dev_queue: pointer to transmit queue
3504 *
3505 * BQL enabled drivers might use this helper in their TX completion path,
3506 * to give appropriate hint to the CPU.
3507 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3508 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3509 {
3510 #ifdef CONFIG_BQL
3511 prefetchw(&dev_queue->dql.limit);
3512 #endif
3513 }
3514
3515 /**
3516 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3517 * @dev_queue: network device queue
3518 * @bytes: number of bytes queued to the device queue
3519 *
3520 * Report the number of bytes queued for sending/completion to the network
3521 * device hardware queue. @bytes should be a good approximation and should
3522 * exactly match netdev_completed_queue() @bytes.
3523 * This is typically called once per packet, from ndo_start_xmit().
3524 */
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3525 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3526 unsigned int bytes)
3527 {
3528 #ifdef CONFIG_BQL
3529 dql_queued(&dev_queue->dql, bytes);
3530
3531 if (likely(dql_avail(&dev_queue->dql) >= 0))
3532 return;
3533
3534 /* Paired with READ_ONCE() from dev_watchdog() */
3535 WRITE_ONCE(dev_queue->trans_start, jiffies);
3536
3537 /* This barrier is paired with smp_mb() from dev_watchdog() */
3538 smp_mb__before_atomic();
3539
3540 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3541
3542 /*
3543 * The XOFF flag must be set before checking the dql_avail below,
3544 * because in netdev_tx_completed_queue we update the dql_completed
3545 * before checking the XOFF flag.
3546 */
3547 smp_mb__after_atomic();
3548
3549 /* check again in case another CPU has just made room avail */
3550 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3551 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3552 #endif
3553 }
3554
3555 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3556 * that they should not test BQL status themselves.
3557 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3558 * skb of a batch.
3559 * Returns true if the doorbell must be used to kick the NIC.
3560 */
__netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes,bool xmit_more)3561 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3562 unsigned int bytes,
3563 bool xmit_more)
3564 {
3565 if (xmit_more) {
3566 #ifdef CONFIG_BQL
3567 dql_queued(&dev_queue->dql, bytes);
3568 #endif
3569 return netif_tx_queue_stopped(dev_queue);
3570 }
3571 netdev_tx_sent_queue(dev_queue, bytes);
3572 return true;
3573 }
3574
3575 /**
3576 * netdev_sent_queue - report the number of bytes queued to hardware
3577 * @dev: network device
3578 * @bytes: number of bytes queued to the hardware device queue
3579 *
3580 * Report the number of bytes queued for sending/completion to the network
3581 * device hardware queue#0. @bytes should be a good approximation and should
3582 * exactly match netdev_completed_queue() @bytes.
3583 * This is typically called once per packet, from ndo_start_xmit().
3584 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)3585 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3586 {
3587 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3588 }
3589
__netdev_sent_queue(struct net_device * dev,unsigned int bytes,bool xmit_more)3590 static inline bool __netdev_sent_queue(struct net_device *dev,
3591 unsigned int bytes,
3592 bool xmit_more)
3593 {
3594 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3595 xmit_more);
3596 }
3597
3598 /**
3599 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3600 * @dev_queue: network device queue
3601 * @pkts: number of packets (currently ignored)
3602 * @bytes: number of bytes dequeued from the device queue
3603 *
3604 * Must be called at most once per TX completion round (and not per
3605 * individual packet), so that BQL can adjust its limits appropriately.
3606 */
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3607 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3608 unsigned int pkts, unsigned int bytes)
3609 {
3610 #ifdef CONFIG_BQL
3611 if (unlikely(!bytes))
3612 return;
3613
3614 dql_completed(&dev_queue->dql, bytes);
3615
3616 /*
3617 * Without the memory barrier there is a small possibility that
3618 * netdev_tx_sent_queue will miss the update and cause the queue to
3619 * be stopped forever
3620 */
3621 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3622
3623 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3624 return;
3625
3626 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3627 netif_schedule_queue(dev_queue);
3628 #endif
3629 }
3630
3631 /**
3632 * netdev_completed_queue - report bytes and packets completed by device
3633 * @dev: network device
3634 * @pkts: actual number of packets sent over the medium
3635 * @bytes: actual number of bytes sent over the medium
3636 *
3637 * Report the number of bytes and packets transmitted by the network device
3638 * hardware queue over the physical medium, @bytes must exactly match the
3639 * @bytes amount passed to netdev_sent_queue()
3640 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3641 static inline void netdev_completed_queue(struct net_device *dev,
3642 unsigned int pkts, unsigned int bytes)
3643 {
3644 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3645 }
3646
netdev_tx_reset_queue(struct netdev_queue * q)3647 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3648 {
3649 #ifdef CONFIG_BQL
3650 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3651 dql_reset(&q->dql);
3652 #endif
3653 }
3654
3655 /**
3656 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue
3657 * @dev: network device
3658 * @qid: stack index of the queue to reset
3659 */
netdev_tx_reset_subqueue(const struct net_device * dev,u32 qid)3660 static inline void netdev_tx_reset_subqueue(const struct net_device *dev,
3661 u32 qid)
3662 {
3663 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid));
3664 }
3665
3666 /**
3667 * netdev_reset_queue - reset the packets and bytes count of a network device
3668 * @dev_queue: network device
3669 *
3670 * Reset the bytes and packet count of a network device and clear the
3671 * software flow control OFF bit for this network device
3672 */
netdev_reset_queue(struct net_device * dev_queue)3673 static inline void netdev_reset_queue(struct net_device *dev_queue)
3674 {
3675 netdev_tx_reset_subqueue(dev_queue, 0);
3676 }
3677
3678 /**
3679 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3680 * @dev: network device
3681 * @queue_index: given tx queue index
3682 *
3683 * Returns 0 if given tx queue index >= number of device tx queues,
3684 * otherwise returns the originally passed tx queue index.
3685 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3686 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3687 {
3688 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3689 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3690 dev->name, queue_index,
3691 dev->real_num_tx_queues);
3692 return 0;
3693 }
3694
3695 return queue_index;
3696 }
3697
3698 /**
3699 * netif_running - test if up
3700 * @dev: network device
3701 *
3702 * Test if the device has been brought up.
3703 */
netif_running(const struct net_device * dev)3704 static inline bool netif_running(const struct net_device *dev)
3705 {
3706 return test_bit(__LINK_STATE_START, &dev->state);
3707 }
3708
3709 /*
3710 * Routines to manage the subqueues on a device. We only need start,
3711 * stop, and a check if it's stopped. All other device management is
3712 * done at the overall netdevice level.
3713 * Also test the device if we're multiqueue.
3714 */
3715
3716 /**
3717 * netif_start_subqueue - allow sending packets on subqueue
3718 * @dev: network device
3719 * @queue_index: sub queue index
3720 *
3721 * Start individual transmit queue of a device with multiple transmit queues.
3722 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)3723 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3724 {
3725 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3726
3727 netif_tx_start_queue(txq);
3728 }
3729
3730 /**
3731 * netif_stop_subqueue - stop sending packets on subqueue
3732 * @dev: network device
3733 * @queue_index: sub queue index
3734 *
3735 * Stop individual transmit queue of a device with multiple transmit queues.
3736 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)3737 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3738 {
3739 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3740 netif_tx_stop_queue(txq);
3741 }
3742
3743 /**
3744 * __netif_subqueue_stopped - test status of subqueue
3745 * @dev: network device
3746 * @queue_index: sub queue index
3747 *
3748 * Check individual transmit queue of a device with multiple transmit queues.
3749 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3750 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3751 u16 queue_index)
3752 {
3753 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3754
3755 return netif_tx_queue_stopped(txq);
3756 }
3757
3758 /**
3759 * netif_subqueue_stopped - test status of subqueue
3760 * @dev: network device
3761 * @skb: sub queue buffer pointer
3762 *
3763 * Check individual transmit queue of a device with multiple transmit queues.
3764 */
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3765 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3766 struct sk_buff *skb)
3767 {
3768 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3769 }
3770
3771 /**
3772 * netif_wake_subqueue - allow sending packets on subqueue
3773 * @dev: network device
3774 * @queue_index: sub queue index
3775 *
3776 * Resume individual transmit queue of a device with multiple transmit queues.
3777 */
netif_wake_subqueue(struct net_device * dev,u16 queue_index)3778 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3779 {
3780 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3781
3782 netif_tx_wake_queue(txq);
3783 }
3784
3785 #ifdef CONFIG_XPS
3786 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3787 u16 index);
3788 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3789 u16 index, enum xps_map_type type);
3790
3791 /**
3792 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3793 * @j: CPU/Rx queue index
3794 * @mask: bitmask of all cpus/rx queues
3795 * @nr_bits: number of bits in the bitmask
3796 *
3797 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3798 */
netif_attr_test_mask(unsigned long j,const unsigned long * mask,unsigned int nr_bits)3799 static inline bool netif_attr_test_mask(unsigned long j,
3800 const unsigned long *mask,
3801 unsigned int nr_bits)
3802 {
3803 cpu_max_bits_warn(j, nr_bits);
3804 return test_bit(j, mask);
3805 }
3806
3807 /**
3808 * netif_attr_test_online - Test for online CPU/Rx queue
3809 * @j: CPU/Rx queue index
3810 * @online_mask: bitmask for CPUs/Rx queues that are online
3811 * @nr_bits: number of bits in the bitmask
3812 *
3813 * Returns true if a CPU/Rx queue is online.
3814 */
netif_attr_test_online(unsigned long j,const unsigned long * online_mask,unsigned int nr_bits)3815 static inline bool netif_attr_test_online(unsigned long j,
3816 const unsigned long *online_mask,
3817 unsigned int nr_bits)
3818 {
3819 cpu_max_bits_warn(j, nr_bits);
3820
3821 if (online_mask)
3822 return test_bit(j, online_mask);
3823
3824 return (j < nr_bits);
3825 }
3826
3827 /**
3828 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3829 * @n: CPU/Rx queue index
3830 * @srcp: the cpumask/Rx queue mask pointer
3831 * @nr_bits: number of bits in the bitmask
3832 *
3833 * Returns >= nr_bits if no further CPUs/Rx queues set.
3834 */
netif_attrmask_next(int n,const unsigned long * srcp,unsigned int nr_bits)3835 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3836 unsigned int nr_bits)
3837 {
3838 /* -1 is a legal arg here. */
3839 if (n != -1)
3840 cpu_max_bits_warn(n, nr_bits);
3841
3842 if (srcp)
3843 return find_next_bit(srcp, nr_bits, n + 1);
3844
3845 return n + 1;
3846 }
3847
3848 /**
3849 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3850 * @n: CPU/Rx queue index
3851 * @src1p: the first CPUs/Rx queues mask pointer
3852 * @src2p: the second CPUs/Rx queues mask pointer
3853 * @nr_bits: number of bits in the bitmask
3854 *
3855 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3856 */
netif_attrmask_next_and(int n,const unsigned long * src1p,const unsigned long * src2p,unsigned int nr_bits)3857 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3858 const unsigned long *src2p,
3859 unsigned int nr_bits)
3860 {
3861 /* -1 is a legal arg here. */
3862 if (n != -1)
3863 cpu_max_bits_warn(n, nr_bits);
3864
3865 if (src1p && src2p)
3866 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3867 else if (src1p)
3868 return find_next_bit(src1p, nr_bits, n + 1);
3869 else if (src2p)
3870 return find_next_bit(src2p, nr_bits, n + 1);
3871
3872 return n + 1;
3873 }
3874 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3875 static inline int netif_set_xps_queue(struct net_device *dev,
3876 const struct cpumask *mask,
3877 u16 index)
3878 {
3879 return 0;
3880 }
3881
__netif_set_xps_queue(struct net_device * dev,const unsigned long * mask,u16 index,enum xps_map_type type)3882 static inline int __netif_set_xps_queue(struct net_device *dev,
3883 const unsigned long *mask,
3884 u16 index, enum xps_map_type type)
3885 {
3886 return 0;
3887 }
3888 #endif
3889
3890 /**
3891 * netif_is_multiqueue - test if device has multiple transmit queues
3892 * @dev: network device
3893 *
3894 * Check if device has multiple transmit queues
3895 */
netif_is_multiqueue(const struct net_device * dev)3896 static inline bool netif_is_multiqueue(const struct net_device *dev)
3897 {
3898 return dev->num_tx_queues > 1;
3899 }
3900
3901 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3902
3903 #ifdef CONFIG_SYSFS
3904 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3905 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxqs)3906 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3907 unsigned int rxqs)
3908 {
3909 dev->real_num_rx_queues = rxqs;
3910 return 0;
3911 }
3912 #endif
3913 int netif_set_real_num_queues(struct net_device *dev,
3914 unsigned int txq, unsigned int rxq);
3915
3916 int netif_get_num_default_rss_queues(void);
3917
3918 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3919 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3920
3921 /*
3922 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3923 * interrupt context or with hardware interrupts being disabled.
3924 * (in_hardirq() || irqs_disabled())
3925 *
3926 * We provide four helpers that can be used in following contexts :
3927 *
3928 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3929 * replacing kfree_skb(skb)
3930 *
3931 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3932 * Typically used in place of consume_skb(skb) in TX completion path
3933 *
3934 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3935 * replacing kfree_skb(skb)
3936 *
3937 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3938 * and consumed a packet. Used in place of consume_skb(skb)
3939 */
dev_kfree_skb_irq(struct sk_buff * skb)3940 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3941 {
3942 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3943 }
3944
dev_consume_skb_irq(struct sk_buff * skb)3945 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3946 {
3947 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3948 }
3949
dev_kfree_skb_any(struct sk_buff * skb)3950 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3951 {
3952 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3953 }
3954
dev_consume_skb_any(struct sk_buff * skb)3955 static inline void dev_consume_skb_any(struct sk_buff *skb)
3956 {
3957 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3958 }
3959
3960 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3961 struct bpf_prog *xdp_prog);
3962 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3963 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb);
3964 int netif_rx(struct sk_buff *skb);
3965 int __netif_rx(struct sk_buff *skb);
3966
3967 int netif_receive_skb(struct sk_buff *skb);
3968 int netif_receive_skb_core(struct sk_buff *skb);
3969 void netif_receive_skb_list_internal(struct list_head *head);
3970 void netif_receive_skb_list(struct list_head *head);
3971 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3972 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3973 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3974 void napi_get_frags_check(struct napi_struct *napi);
3975 gro_result_t napi_gro_frags(struct napi_struct *napi);
3976
napi_free_frags(struct napi_struct * napi)3977 static inline void napi_free_frags(struct napi_struct *napi)
3978 {
3979 kfree_skb(napi->skb);
3980 napi->skb = NULL;
3981 }
3982
3983 bool netdev_is_rx_handler_busy(struct net_device *dev);
3984 int netdev_rx_handler_register(struct net_device *dev,
3985 rx_handler_func_t *rx_handler,
3986 void *rx_handler_data);
3987 void netdev_rx_handler_unregister(struct net_device *dev);
3988
3989 bool dev_valid_name(const char *name);
is_socket_ioctl_cmd(unsigned int cmd)3990 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3991 {
3992 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3993 }
3994 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3995 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3996 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3997 void __user *data, bool *need_copyout);
3998 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3999 int generic_hwtstamp_get_lower(struct net_device *dev,
4000 struct kernel_hwtstamp_config *kernel_cfg);
4001 int generic_hwtstamp_set_lower(struct net_device *dev,
4002 struct kernel_hwtstamp_config *kernel_cfg,
4003 struct netlink_ext_ack *extack);
4004 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4005 unsigned int dev_get_flags(const struct net_device *);
4006 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4007 struct netlink_ext_ack *extack);
4008 int dev_change_flags(struct net_device *dev, unsigned int flags,
4009 struct netlink_ext_ack *extack);
4010 int dev_set_alias(struct net_device *, const char *, size_t);
4011 int dev_get_alias(const struct net_device *, char *, size_t);
4012 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4013 const char *pat, int new_ifindex);
4014 static inline
dev_change_net_namespace(struct net_device * dev,struct net * net,const char * pat)4015 int dev_change_net_namespace(struct net_device *dev, struct net *net,
4016 const char *pat)
4017 {
4018 return __dev_change_net_namespace(dev, net, pat, 0);
4019 }
4020 int __dev_set_mtu(struct net_device *, int);
4021 int dev_set_mtu(struct net_device *, int);
4022 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4023 struct netlink_ext_ack *extack);
4024 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4025 struct netlink_ext_ack *extack);
4026 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4027 struct netlink_ext_ack *extack);
4028 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4029 int dev_get_port_parent_id(struct net_device *dev,
4030 struct netdev_phys_item_id *ppid, bool recurse);
4031 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4032
4033 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4034 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4035 struct netdev_queue *txq, int *ret);
4036
4037 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4038 u8 dev_xdp_prog_count(struct net_device *dev);
4039 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf);
4040 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4041
4042 u32 dev_get_min_mp_channel_count(const struct net_device *dev);
4043
4044 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4045 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4046 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4047 bool is_skb_forwardable(const struct net_device *dev,
4048 const struct sk_buff *skb);
4049
__is_skb_forwardable(const struct net_device * dev,const struct sk_buff * skb,const bool check_mtu)4050 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4051 const struct sk_buff *skb,
4052 const bool check_mtu)
4053 {
4054 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4055 unsigned int len;
4056
4057 if (!(dev->flags & IFF_UP))
4058 return false;
4059
4060 if (!check_mtu)
4061 return true;
4062
4063 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4064 if (skb->len <= len)
4065 return true;
4066
4067 /* if TSO is enabled, we don't care about the length as the packet
4068 * could be forwarded without being segmented before
4069 */
4070 if (skb_is_gso(skb))
4071 return true;
4072
4073 return false;
4074 }
4075
4076 void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4077
4078 #define DEV_CORE_STATS_INC(FIELD) \
4079 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4080 { \
4081 netdev_core_stats_inc(dev, \
4082 offsetof(struct net_device_core_stats, FIELD)); \
4083 }
4084 DEV_CORE_STATS_INC(rx_dropped)
DEV_CORE_STATS_INC(tx_dropped)4085 DEV_CORE_STATS_INC(tx_dropped)
4086 DEV_CORE_STATS_INC(rx_nohandler)
4087 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4088 #undef DEV_CORE_STATS_INC
4089
4090 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4091 struct sk_buff *skb,
4092 const bool check_mtu)
4093 {
4094 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4095 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4096 dev_core_stats_rx_dropped_inc(dev);
4097 kfree_skb(skb);
4098 return NET_RX_DROP;
4099 }
4100
4101 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4102 skb->priority = 0;
4103 return 0;
4104 }
4105
4106 bool dev_nit_active(struct net_device *dev);
4107 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4108
__dev_put(struct net_device * dev)4109 static inline void __dev_put(struct net_device *dev)
4110 {
4111 if (dev) {
4112 #ifdef CONFIG_PCPU_DEV_REFCNT
4113 this_cpu_dec(*dev->pcpu_refcnt);
4114 #else
4115 refcount_dec(&dev->dev_refcnt);
4116 #endif
4117 }
4118 }
4119
__dev_hold(struct net_device * dev)4120 static inline void __dev_hold(struct net_device *dev)
4121 {
4122 if (dev) {
4123 #ifdef CONFIG_PCPU_DEV_REFCNT
4124 this_cpu_inc(*dev->pcpu_refcnt);
4125 #else
4126 refcount_inc(&dev->dev_refcnt);
4127 #endif
4128 }
4129 }
4130
__netdev_tracker_alloc(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4131 static inline void __netdev_tracker_alloc(struct net_device *dev,
4132 netdevice_tracker *tracker,
4133 gfp_t gfp)
4134 {
4135 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4136 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4137 #endif
4138 }
4139
4140 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4141 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4142 */
netdev_tracker_alloc(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4143 static inline void netdev_tracker_alloc(struct net_device *dev,
4144 netdevice_tracker *tracker, gfp_t gfp)
4145 {
4146 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4147 refcount_dec(&dev->refcnt_tracker.no_tracker);
4148 __netdev_tracker_alloc(dev, tracker, gfp);
4149 #endif
4150 }
4151
netdev_tracker_free(struct net_device * dev,netdevice_tracker * tracker)4152 static inline void netdev_tracker_free(struct net_device *dev,
4153 netdevice_tracker *tracker)
4154 {
4155 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4156 ref_tracker_free(&dev->refcnt_tracker, tracker);
4157 #endif
4158 }
4159
netdev_hold(struct net_device * dev,netdevice_tracker * tracker,gfp_t gfp)4160 static inline void netdev_hold(struct net_device *dev,
4161 netdevice_tracker *tracker, gfp_t gfp)
4162 {
4163 if (dev) {
4164 __dev_hold(dev);
4165 __netdev_tracker_alloc(dev, tracker, gfp);
4166 }
4167 }
4168
netdev_put(struct net_device * dev,netdevice_tracker * tracker)4169 static inline void netdev_put(struct net_device *dev,
4170 netdevice_tracker *tracker)
4171 {
4172 if (dev) {
4173 netdev_tracker_free(dev, tracker);
4174 __dev_put(dev);
4175 }
4176 }
4177
4178 /**
4179 * dev_hold - get reference to device
4180 * @dev: network device
4181 *
4182 * Hold reference to device to keep it from being freed.
4183 * Try using netdev_hold() instead.
4184 */
dev_hold(struct net_device * dev)4185 static inline void dev_hold(struct net_device *dev)
4186 {
4187 netdev_hold(dev, NULL, GFP_ATOMIC);
4188 }
4189
4190 /**
4191 * dev_put - release reference to device
4192 * @dev: network device
4193 *
4194 * Release reference to device to allow it to be freed.
4195 * Try using netdev_put() instead.
4196 */
dev_put(struct net_device * dev)4197 static inline void dev_put(struct net_device *dev)
4198 {
4199 netdev_put(dev, NULL);
4200 }
4201
DEFINE_FREE(dev_put,struct net_device *,if (_T)dev_put (_T))4202 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T))
4203
4204 static inline void netdev_ref_replace(struct net_device *odev,
4205 struct net_device *ndev,
4206 netdevice_tracker *tracker,
4207 gfp_t gfp)
4208 {
4209 if (odev)
4210 netdev_tracker_free(odev, tracker);
4211
4212 __dev_hold(ndev);
4213 __dev_put(odev);
4214
4215 if (ndev)
4216 __netdev_tracker_alloc(ndev, tracker, gfp);
4217 }
4218
4219 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4220 * and _off may be called from IRQ context, but it is caller
4221 * who is responsible for serialization of these calls.
4222 *
4223 * The name carrier is inappropriate, these functions should really be
4224 * called netif_lowerlayer_*() because they represent the state of any
4225 * kind of lower layer not just hardware media.
4226 */
4227 void linkwatch_fire_event(struct net_device *dev);
4228
4229 /**
4230 * linkwatch_sync_dev - sync linkwatch for the given device
4231 * @dev: network device to sync linkwatch for
4232 *
4233 * Sync linkwatch for the given device, removing it from the
4234 * pending work list (if queued).
4235 */
4236 void linkwatch_sync_dev(struct net_device *dev);
4237
4238 /**
4239 * netif_carrier_ok - test if carrier present
4240 * @dev: network device
4241 *
4242 * Check if carrier is present on device
4243 */
netif_carrier_ok(const struct net_device * dev)4244 static inline bool netif_carrier_ok(const struct net_device *dev)
4245 {
4246 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4247 }
4248
4249 unsigned long dev_trans_start(struct net_device *dev);
4250
4251 void __netdev_watchdog_up(struct net_device *dev);
4252
4253 void netif_carrier_on(struct net_device *dev);
4254 void netif_carrier_off(struct net_device *dev);
4255 void netif_carrier_event(struct net_device *dev);
4256
4257 /**
4258 * netif_dormant_on - mark device as dormant.
4259 * @dev: network device
4260 *
4261 * Mark device as dormant (as per RFC2863).
4262 *
4263 * The dormant state indicates that the relevant interface is not
4264 * actually in a condition to pass packets (i.e., it is not 'up') but is
4265 * in a "pending" state, waiting for some external event. For "on-
4266 * demand" interfaces, this new state identifies the situation where the
4267 * interface is waiting for events to place it in the up state.
4268 */
netif_dormant_on(struct net_device * dev)4269 static inline void netif_dormant_on(struct net_device *dev)
4270 {
4271 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4272 linkwatch_fire_event(dev);
4273 }
4274
4275 /**
4276 * netif_dormant_off - set device as not dormant.
4277 * @dev: network device
4278 *
4279 * Device is not in dormant state.
4280 */
netif_dormant_off(struct net_device * dev)4281 static inline void netif_dormant_off(struct net_device *dev)
4282 {
4283 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4284 linkwatch_fire_event(dev);
4285 }
4286
4287 /**
4288 * netif_dormant - test if device is dormant
4289 * @dev: network device
4290 *
4291 * Check if device is dormant.
4292 */
netif_dormant(const struct net_device * dev)4293 static inline bool netif_dormant(const struct net_device *dev)
4294 {
4295 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4296 }
4297
4298
4299 /**
4300 * netif_testing_on - mark device as under test.
4301 * @dev: network device
4302 *
4303 * Mark device as under test (as per RFC2863).
4304 *
4305 * The testing state indicates that some test(s) must be performed on
4306 * the interface. After completion, of the test, the interface state
4307 * will change to up, dormant, or down, as appropriate.
4308 */
netif_testing_on(struct net_device * dev)4309 static inline void netif_testing_on(struct net_device *dev)
4310 {
4311 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4312 linkwatch_fire_event(dev);
4313 }
4314
4315 /**
4316 * netif_testing_off - set device as not under test.
4317 * @dev: network device
4318 *
4319 * Device is not in testing state.
4320 */
netif_testing_off(struct net_device * dev)4321 static inline void netif_testing_off(struct net_device *dev)
4322 {
4323 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4324 linkwatch_fire_event(dev);
4325 }
4326
4327 /**
4328 * netif_testing - test if device is under test
4329 * @dev: network device
4330 *
4331 * Check if device is under test
4332 */
netif_testing(const struct net_device * dev)4333 static inline bool netif_testing(const struct net_device *dev)
4334 {
4335 return test_bit(__LINK_STATE_TESTING, &dev->state);
4336 }
4337
4338
4339 /**
4340 * netif_oper_up - test if device is operational
4341 * @dev: network device
4342 *
4343 * Check if carrier is operational
4344 */
netif_oper_up(const struct net_device * dev)4345 static inline bool netif_oper_up(const struct net_device *dev)
4346 {
4347 unsigned int operstate = READ_ONCE(dev->operstate);
4348
4349 return operstate == IF_OPER_UP ||
4350 operstate == IF_OPER_UNKNOWN /* backward compat */;
4351 }
4352
4353 /**
4354 * netif_device_present - is device available or removed
4355 * @dev: network device
4356 *
4357 * Check if device has not been removed from system.
4358 */
netif_device_present(const struct net_device * dev)4359 static inline bool netif_device_present(const struct net_device *dev)
4360 {
4361 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4362 }
4363
4364 void netif_device_detach(struct net_device *dev);
4365
4366 void netif_device_attach(struct net_device *dev);
4367
4368 /*
4369 * Network interface message level settings
4370 */
4371
4372 enum {
4373 NETIF_MSG_DRV_BIT,
4374 NETIF_MSG_PROBE_BIT,
4375 NETIF_MSG_LINK_BIT,
4376 NETIF_MSG_TIMER_BIT,
4377 NETIF_MSG_IFDOWN_BIT,
4378 NETIF_MSG_IFUP_BIT,
4379 NETIF_MSG_RX_ERR_BIT,
4380 NETIF_MSG_TX_ERR_BIT,
4381 NETIF_MSG_TX_QUEUED_BIT,
4382 NETIF_MSG_INTR_BIT,
4383 NETIF_MSG_TX_DONE_BIT,
4384 NETIF_MSG_RX_STATUS_BIT,
4385 NETIF_MSG_PKTDATA_BIT,
4386 NETIF_MSG_HW_BIT,
4387 NETIF_MSG_WOL_BIT,
4388
4389 /* When you add a new bit above, update netif_msg_class_names array
4390 * in net/ethtool/common.c
4391 */
4392 NETIF_MSG_CLASS_COUNT,
4393 };
4394 /* Both ethtool_ops interface and internal driver implementation use u32 */
4395 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4396
4397 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4398 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4399
4400 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4401 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4402 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4403 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4404 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4405 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4406 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4407 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4408 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4409 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4410 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4411 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4412 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4413 #define NETIF_MSG_HW __NETIF_MSG(HW)
4414 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4415
4416 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4417 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4418 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4419 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4420 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4421 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4422 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4423 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4424 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4425 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4426 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4427 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4428 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4429 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4430 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4431
netif_msg_init(int debug_value,int default_msg_enable_bits)4432 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4433 {
4434 /* use default */
4435 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4436 return default_msg_enable_bits;
4437 if (debug_value == 0) /* no output */
4438 return 0;
4439 /* set low N bits */
4440 return (1U << debug_value) - 1;
4441 }
4442
__netif_tx_lock(struct netdev_queue * txq,int cpu)4443 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4444 {
4445 spin_lock(&txq->_xmit_lock);
4446 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4447 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4448 }
4449
__netif_tx_acquire(struct netdev_queue * txq)4450 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4451 {
4452 __acquire(&txq->_xmit_lock);
4453 return true;
4454 }
4455
__netif_tx_release(struct netdev_queue * txq)4456 static inline void __netif_tx_release(struct netdev_queue *txq)
4457 {
4458 __release(&txq->_xmit_lock);
4459 }
4460
__netif_tx_lock_bh(struct netdev_queue * txq)4461 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4462 {
4463 spin_lock_bh(&txq->_xmit_lock);
4464 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4465 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4466 }
4467
__netif_tx_trylock(struct netdev_queue * txq)4468 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4469 {
4470 bool ok = spin_trylock(&txq->_xmit_lock);
4471
4472 if (likely(ok)) {
4473 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4474 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4475 }
4476 return ok;
4477 }
4478
__netif_tx_unlock(struct netdev_queue * txq)4479 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4480 {
4481 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4482 WRITE_ONCE(txq->xmit_lock_owner, -1);
4483 spin_unlock(&txq->_xmit_lock);
4484 }
4485
__netif_tx_unlock_bh(struct netdev_queue * txq)4486 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4487 {
4488 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4489 WRITE_ONCE(txq->xmit_lock_owner, -1);
4490 spin_unlock_bh(&txq->_xmit_lock);
4491 }
4492
4493 /*
4494 * txq->trans_start can be read locklessly from dev_watchdog()
4495 */
txq_trans_update(struct netdev_queue * txq)4496 static inline void txq_trans_update(struct netdev_queue *txq)
4497 {
4498 if (txq->xmit_lock_owner != -1)
4499 WRITE_ONCE(txq->trans_start, jiffies);
4500 }
4501
txq_trans_cond_update(struct netdev_queue * txq)4502 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4503 {
4504 unsigned long now = jiffies;
4505
4506 if (READ_ONCE(txq->trans_start) != now)
4507 WRITE_ONCE(txq->trans_start, now);
4508 }
4509
4510 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
netif_trans_update(struct net_device * dev)4511 static inline void netif_trans_update(struct net_device *dev)
4512 {
4513 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4514
4515 txq_trans_cond_update(txq);
4516 }
4517
4518 /**
4519 * netif_tx_lock - grab network device transmit lock
4520 * @dev: network device
4521 *
4522 * Get network device transmit lock
4523 */
4524 void netif_tx_lock(struct net_device *dev);
4525
netif_tx_lock_bh(struct net_device * dev)4526 static inline void netif_tx_lock_bh(struct net_device *dev)
4527 {
4528 local_bh_disable();
4529 netif_tx_lock(dev);
4530 }
4531
4532 void netif_tx_unlock(struct net_device *dev);
4533
netif_tx_unlock_bh(struct net_device * dev)4534 static inline void netif_tx_unlock_bh(struct net_device *dev)
4535 {
4536 netif_tx_unlock(dev);
4537 local_bh_enable();
4538 }
4539
4540 #define HARD_TX_LOCK(dev, txq, cpu) { \
4541 if (!(dev)->lltx) { \
4542 __netif_tx_lock(txq, cpu); \
4543 } else { \
4544 __netif_tx_acquire(txq); \
4545 } \
4546 }
4547
4548 #define HARD_TX_TRYLOCK(dev, txq) \
4549 (!(dev)->lltx ? \
4550 __netif_tx_trylock(txq) : \
4551 __netif_tx_acquire(txq))
4552
4553 #define HARD_TX_UNLOCK(dev, txq) { \
4554 if (!(dev)->lltx) { \
4555 __netif_tx_unlock(txq); \
4556 } else { \
4557 __netif_tx_release(txq); \
4558 } \
4559 }
4560
netif_tx_disable(struct net_device * dev)4561 static inline void netif_tx_disable(struct net_device *dev)
4562 {
4563 unsigned int i;
4564 int cpu;
4565
4566 local_bh_disable();
4567 cpu = smp_processor_id();
4568 spin_lock(&dev->tx_global_lock);
4569 for (i = 0; i < dev->num_tx_queues; i++) {
4570 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4571
4572 __netif_tx_lock(txq, cpu);
4573 netif_tx_stop_queue(txq);
4574 __netif_tx_unlock(txq);
4575 }
4576 spin_unlock(&dev->tx_global_lock);
4577 local_bh_enable();
4578 }
4579
netif_addr_lock(struct net_device * dev)4580 static inline void netif_addr_lock(struct net_device *dev)
4581 {
4582 unsigned char nest_level = 0;
4583
4584 #ifdef CONFIG_LOCKDEP
4585 nest_level = dev->nested_level;
4586 #endif
4587 spin_lock_nested(&dev->addr_list_lock, nest_level);
4588 }
4589
netif_addr_lock_bh(struct net_device * dev)4590 static inline void netif_addr_lock_bh(struct net_device *dev)
4591 {
4592 unsigned char nest_level = 0;
4593
4594 #ifdef CONFIG_LOCKDEP
4595 nest_level = dev->nested_level;
4596 #endif
4597 local_bh_disable();
4598 spin_lock_nested(&dev->addr_list_lock, nest_level);
4599 }
4600
netif_addr_unlock(struct net_device * dev)4601 static inline void netif_addr_unlock(struct net_device *dev)
4602 {
4603 spin_unlock(&dev->addr_list_lock);
4604 }
4605
netif_addr_unlock_bh(struct net_device * dev)4606 static inline void netif_addr_unlock_bh(struct net_device *dev)
4607 {
4608 spin_unlock_bh(&dev->addr_list_lock);
4609 }
4610
4611 /*
4612 * dev_addrs walker. Should be used only for read access. Call with
4613 * rcu_read_lock held.
4614 */
4615 #define for_each_dev_addr(dev, ha) \
4616 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4617
4618 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4619
4620 void ether_setup(struct net_device *dev);
4621
4622 /* Allocate dummy net_device */
4623 struct net_device *alloc_netdev_dummy(int sizeof_priv);
4624
4625 /* Support for loadable net-drivers */
4626 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4627 unsigned char name_assign_type,
4628 void (*setup)(struct net_device *),
4629 unsigned int txqs, unsigned int rxqs);
4630 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4631 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4632
4633 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4634 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4635 count)
4636
4637 int register_netdev(struct net_device *dev);
4638 void unregister_netdev(struct net_device *dev);
4639
4640 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4641
4642 /* General hardware address lists handling functions */
4643 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4644 struct netdev_hw_addr_list *from_list, int addr_len);
4645 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4646 struct netdev_hw_addr_list *from_list, int addr_len);
4647 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4648 struct net_device *dev,
4649 int (*sync)(struct net_device *, const unsigned char *),
4650 int (*unsync)(struct net_device *,
4651 const unsigned char *));
4652 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4653 struct net_device *dev,
4654 int (*sync)(struct net_device *,
4655 const unsigned char *, int),
4656 int (*unsync)(struct net_device *,
4657 const unsigned char *, int));
4658 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4659 struct net_device *dev,
4660 int (*unsync)(struct net_device *,
4661 const unsigned char *, int));
4662 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4663 struct net_device *dev,
4664 int (*unsync)(struct net_device *,
4665 const unsigned char *));
4666 void __hw_addr_init(struct netdev_hw_addr_list *list);
4667
4668 /* Functions used for device addresses handling */
4669 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4670 const void *addr, size_t len);
4671
4672 static inline void
__dev_addr_set(struct net_device * dev,const void * addr,size_t len)4673 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4674 {
4675 dev_addr_mod(dev, 0, addr, len);
4676 }
4677
dev_addr_set(struct net_device * dev,const u8 * addr)4678 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4679 {
4680 __dev_addr_set(dev, addr, dev->addr_len);
4681 }
4682
4683 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4684 unsigned char addr_type);
4685 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4686 unsigned char addr_type);
4687
4688 /* Functions used for unicast addresses handling */
4689 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4690 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4691 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4692 int dev_uc_sync(struct net_device *to, struct net_device *from);
4693 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4694 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4695 void dev_uc_flush(struct net_device *dev);
4696 void dev_uc_init(struct net_device *dev);
4697
4698 /**
4699 * __dev_uc_sync - Synchronize device's unicast list
4700 * @dev: device to sync
4701 * @sync: function to call if address should be added
4702 * @unsync: function to call if address should be removed
4703 *
4704 * Add newly added addresses to the interface, and release
4705 * addresses that have been deleted.
4706 */
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4707 static inline int __dev_uc_sync(struct net_device *dev,
4708 int (*sync)(struct net_device *,
4709 const unsigned char *),
4710 int (*unsync)(struct net_device *,
4711 const unsigned char *))
4712 {
4713 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4714 }
4715
4716 /**
4717 * __dev_uc_unsync - Remove synchronized addresses from device
4718 * @dev: device to sync
4719 * @unsync: function to call if address should be removed
4720 *
4721 * Remove all addresses that were added to the device by dev_uc_sync().
4722 */
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4723 static inline void __dev_uc_unsync(struct net_device *dev,
4724 int (*unsync)(struct net_device *,
4725 const unsigned char *))
4726 {
4727 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4728 }
4729
4730 /* Functions used for multicast addresses handling */
4731 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4732 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4733 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4734 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4735 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4736 int dev_mc_sync(struct net_device *to, struct net_device *from);
4737 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4738 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4739 void dev_mc_flush(struct net_device *dev);
4740 void dev_mc_init(struct net_device *dev);
4741
4742 /**
4743 * __dev_mc_sync - Synchronize device's multicast list
4744 * @dev: device to sync
4745 * @sync: function to call if address should be added
4746 * @unsync: function to call if address should be removed
4747 *
4748 * Add newly added addresses to the interface, and release
4749 * addresses that have been deleted.
4750 */
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4751 static inline int __dev_mc_sync(struct net_device *dev,
4752 int (*sync)(struct net_device *,
4753 const unsigned char *),
4754 int (*unsync)(struct net_device *,
4755 const unsigned char *))
4756 {
4757 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4758 }
4759
4760 /**
4761 * __dev_mc_unsync - Remove synchronized addresses from device
4762 * @dev: device to sync
4763 * @unsync: function to call if address should be removed
4764 *
4765 * Remove all addresses that were added to the device by dev_mc_sync().
4766 */
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4767 static inline void __dev_mc_unsync(struct net_device *dev,
4768 int (*unsync)(struct net_device *,
4769 const unsigned char *))
4770 {
4771 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4772 }
4773
4774 /* Functions used for secondary unicast and multicast support */
4775 void dev_set_rx_mode(struct net_device *dev);
4776 int dev_set_promiscuity(struct net_device *dev, int inc);
4777 int dev_set_allmulti(struct net_device *dev, int inc);
4778 void netdev_state_change(struct net_device *dev);
4779 void __netdev_notify_peers(struct net_device *dev);
4780 void netdev_notify_peers(struct net_device *dev);
4781 void netdev_features_change(struct net_device *dev);
4782 /* Load a device via the kmod */
4783 void dev_load(struct net *net, const char *name);
4784 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4785 struct rtnl_link_stats64 *storage);
4786 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4787 const struct net_device_stats *netdev_stats);
4788 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4789 const struct pcpu_sw_netstats __percpu *netstats);
4790 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4791
4792 enum {
4793 NESTED_SYNC_IMM_BIT,
4794 NESTED_SYNC_TODO_BIT,
4795 };
4796
4797 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4798 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4799
4800 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4801 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4802
4803 struct netdev_nested_priv {
4804 unsigned char flags;
4805 void *data;
4806 };
4807
4808 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4809 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4810 struct list_head **iter);
4811
4812 /* iterate through upper list, must be called under RCU read lock */
4813 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4814 for (iter = &(dev)->adj_list.upper, \
4815 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4816 updev; \
4817 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4818
4819 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4820 int (*fn)(struct net_device *upper_dev,
4821 struct netdev_nested_priv *priv),
4822 struct netdev_nested_priv *priv);
4823
4824 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4825 struct net_device *upper_dev);
4826
4827 bool netdev_has_any_upper_dev(struct net_device *dev);
4828
4829 void *netdev_lower_get_next_private(struct net_device *dev,
4830 struct list_head **iter);
4831 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4832 struct list_head **iter);
4833
4834 #define netdev_for_each_lower_private(dev, priv, iter) \
4835 for (iter = (dev)->adj_list.lower.next, \
4836 priv = netdev_lower_get_next_private(dev, &(iter)); \
4837 priv; \
4838 priv = netdev_lower_get_next_private(dev, &(iter)))
4839
4840 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4841 for (iter = &(dev)->adj_list.lower, \
4842 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4843 priv; \
4844 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4845
4846 void *netdev_lower_get_next(struct net_device *dev,
4847 struct list_head **iter);
4848
4849 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4850 for (iter = (dev)->adj_list.lower.next, \
4851 ldev = netdev_lower_get_next(dev, &(iter)); \
4852 ldev; \
4853 ldev = netdev_lower_get_next(dev, &(iter)))
4854
4855 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4856 struct list_head **iter);
4857 int netdev_walk_all_lower_dev(struct net_device *dev,
4858 int (*fn)(struct net_device *lower_dev,
4859 struct netdev_nested_priv *priv),
4860 struct netdev_nested_priv *priv);
4861 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4862 int (*fn)(struct net_device *lower_dev,
4863 struct netdev_nested_priv *priv),
4864 struct netdev_nested_priv *priv);
4865
4866 void *netdev_adjacent_get_private(struct list_head *adj_list);
4867 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4868 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4869 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4870 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4871 struct netlink_ext_ack *extack);
4872 int netdev_master_upper_dev_link(struct net_device *dev,
4873 struct net_device *upper_dev,
4874 void *upper_priv, void *upper_info,
4875 struct netlink_ext_ack *extack);
4876 void netdev_upper_dev_unlink(struct net_device *dev,
4877 struct net_device *upper_dev);
4878 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4879 struct net_device *new_dev,
4880 struct net_device *dev,
4881 struct netlink_ext_ack *extack);
4882 void netdev_adjacent_change_commit(struct net_device *old_dev,
4883 struct net_device *new_dev,
4884 struct net_device *dev);
4885 void netdev_adjacent_change_abort(struct net_device *old_dev,
4886 struct net_device *new_dev,
4887 struct net_device *dev);
4888 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4889 void *netdev_lower_dev_get_private(struct net_device *dev,
4890 struct net_device *lower_dev);
4891 void netdev_lower_state_changed(struct net_device *lower_dev,
4892 void *lower_state_info);
4893
4894 /* RSS keys are 40 or 52 bytes long */
4895 #define NETDEV_RSS_KEY_LEN 52
4896 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4897 void netdev_rss_key_fill(void *buffer, size_t len);
4898
4899 int skb_checksum_help(struct sk_buff *skb);
4900 int skb_crc32c_csum_help(struct sk_buff *skb);
4901 int skb_csum_hwoffload_help(struct sk_buff *skb,
4902 const netdev_features_t features);
4903
4904 struct netdev_bonding_info {
4905 ifslave slave;
4906 ifbond master;
4907 };
4908
4909 struct netdev_notifier_bonding_info {
4910 struct netdev_notifier_info info; /* must be first */
4911 struct netdev_bonding_info bonding_info;
4912 };
4913
4914 void netdev_bonding_info_change(struct net_device *dev,
4915 struct netdev_bonding_info *bonding_info);
4916
4917 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4918 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4919 #else
ethtool_notify(struct net_device * dev,unsigned int cmd,const void * data)4920 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4921 const void *data)
4922 {
4923 }
4924 #endif
4925
4926 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4927
can_checksum_protocol(netdev_features_t features,__be16 protocol)4928 static inline bool can_checksum_protocol(netdev_features_t features,
4929 __be16 protocol)
4930 {
4931 if (protocol == htons(ETH_P_FCOE))
4932 return !!(features & NETIF_F_FCOE_CRC);
4933
4934 /* Assume this is an IP checksum (not SCTP CRC) */
4935
4936 if (features & NETIF_F_HW_CSUM) {
4937 /* Can checksum everything */
4938 return true;
4939 }
4940
4941 switch (protocol) {
4942 case htons(ETH_P_IP):
4943 return !!(features & NETIF_F_IP_CSUM);
4944 case htons(ETH_P_IPV6):
4945 return !!(features & NETIF_F_IPV6_CSUM);
4946 default:
4947 return false;
4948 }
4949 }
4950
4951 #ifdef CONFIG_BUG
4952 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4953 #else
netdev_rx_csum_fault(struct net_device * dev,struct sk_buff * skb)4954 static inline void netdev_rx_csum_fault(struct net_device *dev,
4955 struct sk_buff *skb)
4956 {
4957 }
4958 #endif
4959 /* rx skb timestamps */
4960 void net_enable_timestamp(void);
4961 void net_disable_timestamp(void);
4962
netdev_get_tstamp(struct net_device * dev,const struct skb_shared_hwtstamps * hwtstamps,bool cycles)4963 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4964 const struct skb_shared_hwtstamps *hwtstamps,
4965 bool cycles)
4966 {
4967 const struct net_device_ops *ops = dev->netdev_ops;
4968
4969 if (ops->ndo_get_tstamp)
4970 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4971
4972 return hwtstamps->hwtstamp;
4973 }
4974
4975 #ifndef CONFIG_PREEMPT_RT
netdev_xmit_set_more(bool more)4976 static inline void netdev_xmit_set_more(bool more)
4977 {
4978 __this_cpu_write(softnet_data.xmit.more, more);
4979 }
4980
netdev_xmit_more(void)4981 static inline bool netdev_xmit_more(void)
4982 {
4983 return __this_cpu_read(softnet_data.xmit.more);
4984 }
4985 #else
netdev_xmit_set_more(bool more)4986 static inline void netdev_xmit_set_more(bool more)
4987 {
4988 current->net_xmit.more = more;
4989 }
4990
netdev_xmit_more(void)4991 static inline bool netdev_xmit_more(void)
4992 {
4993 return current->net_xmit.more;
4994 }
4995 #endif
4996
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)4997 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4998 struct sk_buff *skb, struct net_device *dev,
4999 bool more)
5000 {
5001 netdev_xmit_set_more(more);
5002 return ops->ndo_start_xmit(skb, dev);
5003 }
5004
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)5005 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
5006 struct netdev_queue *txq, bool more)
5007 {
5008 const struct net_device_ops *ops = dev->netdev_ops;
5009 netdev_tx_t rc;
5010
5011 rc = __netdev_start_xmit(ops, skb, dev, more);
5012 if (rc == NETDEV_TX_OK)
5013 txq_trans_update(txq);
5014
5015 return rc;
5016 }
5017
5018 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
5019 const void *ns);
5020 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
5021 const void *ns);
5022
5023 extern const struct kobj_ns_type_operations net_ns_type_operations;
5024
5025 const char *netdev_drivername(const struct net_device *dev);
5026
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)5027 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
5028 netdev_features_t f2)
5029 {
5030 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
5031 if (f1 & NETIF_F_HW_CSUM)
5032 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5033 else
5034 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5035 }
5036
5037 return f1 & f2;
5038 }
5039
netdev_get_wanted_features(struct net_device * dev)5040 static inline netdev_features_t netdev_get_wanted_features(
5041 struct net_device *dev)
5042 {
5043 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5044 }
5045 netdev_features_t netdev_increment_features(netdev_features_t all,
5046 netdev_features_t one, netdev_features_t mask);
5047
5048 /* Allow TSO being used on stacked device :
5049 * Performing the GSO segmentation before last device
5050 * is a performance improvement.
5051 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)5052 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5053 netdev_features_t mask)
5054 {
5055 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5056 }
5057
5058 int __netdev_update_features(struct net_device *dev);
5059 void netdev_update_features(struct net_device *dev);
5060 void netdev_change_features(struct net_device *dev);
5061
5062 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5063 struct net_device *dev);
5064
5065 netdev_features_t passthru_features_check(struct sk_buff *skb,
5066 struct net_device *dev,
5067 netdev_features_t features);
5068 netdev_features_t netif_skb_features(struct sk_buff *skb);
5069 void skb_warn_bad_offload(const struct sk_buff *skb);
5070
net_gso_ok(netdev_features_t features,int gso_type)5071 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5072 {
5073 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5074
5075 /* check flags correspondence */
5076 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5077 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5078 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5079 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5080 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5081 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5082 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5083 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5084 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5085 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5086 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5087 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5088 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5089 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5090 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5091 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5092 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5093 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5094 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5095
5096 return (features & feature) == feature;
5097 }
5098
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)5099 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5100 {
5101 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5102 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5103 }
5104
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)5105 static inline bool netif_needs_gso(struct sk_buff *skb,
5106 netdev_features_t features)
5107 {
5108 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5109 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5110 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5111 }
5112
5113 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5114 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5115 void netif_inherit_tso_max(struct net_device *to,
5116 const struct net_device *from);
5117
5118 static inline unsigned int
netif_get_gro_max_size(const struct net_device * dev,const struct sk_buff * skb)5119 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb)
5120 {
5121 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */
5122 return skb->protocol == htons(ETH_P_IPV6) ?
5123 READ_ONCE(dev->gro_max_size) :
5124 READ_ONCE(dev->gro_ipv4_max_size);
5125 }
5126
5127 static inline unsigned int
netif_get_gso_max_size(const struct net_device * dev,const struct sk_buff * skb)5128 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb)
5129 {
5130 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */
5131 return skb->protocol == htons(ETH_P_IPV6) ?
5132 READ_ONCE(dev->gso_max_size) :
5133 READ_ONCE(dev->gso_ipv4_max_size);
5134 }
5135
netif_is_macsec(const struct net_device * dev)5136 static inline bool netif_is_macsec(const struct net_device *dev)
5137 {
5138 return dev->priv_flags & IFF_MACSEC;
5139 }
5140
netif_is_macvlan(const struct net_device * dev)5141 static inline bool netif_is_macvlan(const struct net_device *dev)
5142 {
5143 return dev->priv_flags & IFF_MACVLAN;
5144 }
5145
netif_is_macvlan_port(const struct net_device * dev)5146 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5147 {
5148 return dev->priv_flags & IFF_MACVLAN_PORT;
5149 }
5150
netif_is_bond_master(const struct net_device * dev)5151 static inline bool netif_is_bond_master(const struct net_device *dev)
5152 {
5153 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5154 }
5155
netif_is_bond_slave(const struct net_device * dev)5156 static inline bool netif_is_bond_slave(const struct net_device *dev)
5157 {
5158 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5159 }
5160
netif_supports_nofcs(struct net_device * dev)5161 static inline bool netif_supports_nofcs(struct net_device *dev)
5162 {
5163 return dev->priv_flags & IFF_SUPP_NOFCS;
5164 }
5165
netif_has_l3_rx_handler(const struct net_device * dev)5166 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5167 {
5168 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5169 }
5170
netif_is_l3_master(const struct net_device * dev)5171 static inline bool netif_is_l3_master(const struct net_device *dev)
5172 {
5173 return dev->priv_flags & IFF_L3MDEV_MASTER;
5174 }
5175
netif_is_l3_slave(const struct net_device * dev)5176 static inline bool netif_is_l3_slave(const struct net_device *dev)
5177 {
5178 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5179 }
5180
dev_sdif(const struct net_device * dev)5181 static inline int dev_sdif(const struct net_device *dev)
5182 {
5183 #ifdef CONFIG_NET_L3_MASTER_DEV
5184 if (netif_is_l3_slave(dev))
5185 return dev->ifindex;
5186 #endif
5187 return 0;
5188 }
5189
netif_is_bridge_master(const struct net_device * dev)5190 static inline bool netif_is_bridge_master(const struct net_device *dev)
5191 {
5192 return dev->priv_flags & IFF_EBRIDGE;
5193 }
5194
netif_is_bridge_port(const struct net_device * dev)5195 static inline bool netif_is_bridge_port(const struct net_device *dev)
5196 {
5197 return dev->priv_flags & IFF_BRIDGE_PORT;
5198 }
5199
netif_is_ovs_master(const struct net_device * dev)5200 static inline bool netif_is_ovs_master(const struct net_device *dev)
5201 {
5202 return dev->priv_flags & IFF_OPENVSWITCH;
5203 }
5204
netif_is_ovs_port(const struct net_device * dev)5205 static inline bool netif_is_ovs_port(const struct net_device *dev)
5206 {
5207 return dev->priv_flags & IFF_OVS_DATAPATH;
5208 }
5209
netif_is_any_bridge_master(const struct net_device * dev)5210 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5211 {
5212 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5213 }
5214
netif_is_any_bridge_port(const struct net_device * dev)5215 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5216 {
5217 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5218 }
5219
netif_is_team_master(const struct net_device * dev)5220 static inline bool netif_is_team_master(const struct net_device *dev)
5221 {
5222 return dev->priv_flags & IFF_TEAM;
5223 }
5224
netif_is_team_port(const struct net_device * dev)5225 static inline bool netif_is_team_port(const struct net_device *dev)
5226 {
5227 return dev->priv_flags & IFF_TEAM_PORT;
5228 }
5229
netif_is_lag_master(const struct net_device * dev)5230 static inline bool netif_is_lag_master(const struct net_device *dev)
5231 {
5232 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5233 }
5234
netif_is_lag_port(const struct net_device * dev)5235 static inline bool netif_is_lag_port(const struct net_device *dev)
5236 {
5237 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5238 }
5239
netif_is_rxfh_configured(const struct net_device * dev)5240 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5241 {
5242 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5243 }
5244
netif_is_failover(const struct net_device * dev)5245 static inline bool netif_is_failover(const struct net_device *dev)
5246 {
5247 return dev->priv_flags & IFF_FAILOVER;
5248 }
5249
netif_is_failover_slave(const struct net_device * dev)5250 static inline bool netif_is_failover_slave(const struct net_device *dev)
5251 {
5252 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5253 }
5254
5255 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)5256 static inline void netif_keep_dst(struct net_device *dev)
5257 {
5258 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5259 }
5260
5261 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
netif_reduces_vlan_mtu(struct net_device * dev)5262 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5263 {
5264 /* TODO: reserve and use an additional IFF bit, if we get more users */
5265 return netif_is_macsec(dev);
5266 }
5267
5268 extern struct pernet_operations __net_initdata loopback_net_ops;
5269
5270 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5271
5272 /* netdev_printk helpers, similar to dev_printk */
5273
netdev_name(const struct net_device * dev)5274 static inline const char *netdev_name(const struct net_device *dev)
5275 {
5276 if (!dev->name[0] || strchr(dev->name, '%'))
5277 return "(unnamed net_device)";
5278 return dev->name;
5279 }
5280
netdev_reg_state(const struct net_device * dev)5281 static inline const char *netdev_reg_state(const struct net_device *dev)
5282 {
5283 u8 reg_state = READ_ONCE(dev->reg_state);
5284
5285 switch (reg_state) {
5286 case NETREG_UNINITIALIZED: return " (uninitialized)";
5287 case NETREG_REGISTERED: return "";
5288 case NETREG_UNREGISTERING: return " (unregistering)";
5289 case NETREG_UNREGISTERED: return " (unregistered)";
5290 case NETREG_RELEASED: return " (released)";
5291 case NETREG_DUMMY: return " (dummy)";
5292 }
5293
5294 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state);
5295 return " (unknown)";
5296 }
5297
5298 #define MODULE_ALIAS_NETDEV(device) \
5299 MODULE_ALIAS("netdev-" device)
5300
5301 /*
5302 * netdev_WARN() acts like dev_printk(), but with the key difference
5303 * of using a WARN/WARN_ON to get the message out, including the
5304 * file/line information and a backtrace.
5305 */
5306 #define netdev_WARN(dev, format, args...) \
5307 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5308 netdev_reg_state(dev), ##args)
5309
5310 #define netdev_WARN_ONCE(dev, format, args...) \
5311 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5312 netdev_reg_state(dev), ##args)
5313
5314 /*
5315 * The list of packet types we will receive (as opposed to discard)
5316 * and the routines to invoke.
5317 *
5318 * Why 16. Because with 16 the only overlap we get on a hash of the
5319 * low nibble of the protocol value is RARP/SNAP/X.25.
5320 *
5321 * 0800 IP
5322 * 0001 802.3
5323 * 0002 AX.25
5324 * 0004 802.2
5325 * 8035 RARP
5326 * 0005 SNAP
5327 * 0805 X.25
5328 * 0806 ARP
5329 * 8137 IPX
5330 * 0009 Localtalk
5331 * 86DD IPv6
5332 */
5333 #define PTYPE_HASH_SIZE (16)
5334 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5335
5336 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5337
5338 extern struct net_device *blackhole_netdev;
5339
5340 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5341 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5342 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5343 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5344 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5345
5346 #endif /* _LINUX_NETDEVICE_H */
5347