1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
24#include <linux/timer.h>
25#include <linux/bug.h>
26#include <linux/delay.h>
27#include <linux/atomic.h>
28#include <linux/prefetch.h>
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31#include <asm/local.h>
32
33#include <linux/percpu.h>
34#include <linux/rculist.h>
35#include <linux/workqueue.h>
36#include <linux/dynamic_queue_limits.h>
37
38#include <net/net_namespace.h>
39#ifdef CONFIG_DCB
40#include <net/dcbnl.h>
41#endif
42#include <net/netprio_cgroup.h>
43
44#include <linux/netdev_features.h>
45#include <linux/neighbour.h>
46#include <uapi/linux/netdevice.h>
47#include <uapi/linux/if_bonding.h>
48#include <uapi/linux/pkt_cls.h>
49#include <uapi/linux/netdev.h>
50#include <linux/hashtable.h>
51#include <linux/rbtree.h>
52#include <net/net_trackers.h>
53#include <net/net_debug.h>
54#include <net/dropreason-core.h>
55
56struct netpoll_info;
57struct device;
58struct ethtool_ops;
59struct kernel_hwtstamp_config;
60struct phy_device;
61struct dsa_port;
62struct ip_tunnel_parm;
63struct macsec_context;
64struct macsec_ops;
65struct netdev_name_node;
66struct sd_flow_limit;
67struct sfp_bus;
68/* 802.11 specific */
69struct wireless_dev;
70/* 802.15.4 specific */
71struct wpan_dev;
72struct mpls_dev;
73/* UDP Tunnel offloads */
74struct udp_tunnel_info;
75struct udp_tunnel_nic_info;
76struct udp_tunnel_nic;
77struct bpf_prog;
78struct xdp_buff;
79struct xdp_frame;
80struct xdp_metadata_ops;
81struct xdp_md;
82
83typedef u32 xdp_features_t;
84
85void synchronize_net(void);
86void netdev_set_default_ethtool_ops(struct net_device *dev,
87 const struct ethtool_ops *ops);
88void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
89
90/* Backlog congestion levels */
91#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
92#define NET_RX_DROP 1 /* packet dropped */
93
94#define MAX_NEST_DEV 8
95
96/*
97 * Transmit return codes: transmit return codes originate from three different
98 * namespaces:
99 *
100 * - qdisc return codes
101 * - driver transmit return codes
102 * - errno values
103 *
104 * Drivers are allowed to return any one of those in their hard_start_xmit()
105 * function. Real network devices commonly used with qdiscs should only return
106 * the driver transmit return codes though - when qdiscs are used, the actual
107 * transmission happens asynchronously, so the value is not propagated to
108 * higher layers. Virtual network devices transmit synchronously; in this case
109 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
110 * others are propagated to higher layers.
111 */
112
113/* qdisc ->enqueue() return codes. */
114#define NET_XMIT_SUCCESS 0x00
115#define NET_XMIT_DROP 0x01 /* skb dropped */
116#define NET_XMIT_CN 0x02 /* congestion notification */
117#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
118
119/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
120 * indicates that the device will soon be dropping packets, or already drops
121 * some packets of the same priority; prompting us to send less aggressively. */
122#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
123#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
124
125/* Driver transmit return codes */
126#define NETDEV_TX_MASK 0xf0
127
128enum netdev_tx {
129 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
130 NETDEV_TX_OK = 0x00, /* driver took care of packet */
131 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
132};
133typedef enum netdev_tx netdev_tx_t;
134
135/*
136 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
137 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
138 */
139static inline bool dev_xmit_complete(int rc)
140{
141 /*
142 * Positive cases with an skb consumed by a driver:
143 * - successful transmission (rc == NETDEV_TX_OK)
144 * - error while transmitting (rc < 0)
145 * - error while queueing to a different device (rc & NET_XMIT_MASK)
146 */
147 if (likely(rc < NET_XMIT_MASK))
148 return true;
149
150 return false;
151}
152
153/*
154 * Compute the worst-case header length according to the protocols
155 * used.
156 */
157
158#if defined(CONFIG_HYPERV_NET)
159# define LL_MAX_HEADER 128
160#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
161# if defined(CONFIG_MAC80211_MESH)
162# define LL_MAX_HEADER 128
163# else
164# define LL_MAX_HEADER 96
165# endif
166#else
167# define LL_MAX_HEADER 32
168#endif
169
170#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
171 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
172#define MAX_HEADER LL_MAX_HEADER
173#else
174#define MAX_HEADER (LL_MAX_HEADER + 48)
175#endif
176
177/*
178 * Old network device statistics. Fields are native words
179 * (unsigned long) so they can be read and written atomically.
180 */
181
182#define NET_DEV_STAT(FIELD) \
183 union { \
184 unsigned long FIELD; \
185 atomic_long_t __##FIELD; \
186 }
187
188struct net_device_stats {
189 NET_DEV_STAT(rx_packets);
190 NET_DEV_STAT(tx_packets);
191 NET_DEV_STAT(rx_bytes);
192 NET_DEV_STAT(tx_bytes);
193 NET_DEV_STAT(rx_errors);
194 NET_DEV_STAT(tx_errors);
195 NET_DEV_STAT(rx_dropped);
196 NET_DEV_STAT(tx_dropped);
197 NET_DEV_STAT(multicast);
198 NET_DEV_STAT(collisions);
199 NET_DEV_STAT(rx_length_errors);
200 NET_DEV_STAT(rx_over_errors);
201 NET_DEV_STAT(rx_crc_errors);
202 NET_DEV_STAT(rx_frame_errors);
203 NET_DEV_STAT(rx_fifo_errors);
204 NET_DEV_STAT(rx_missed_errors);
205 NET_DEV_STAT(tx_aborted_errors);
206 NET_DEV_STAT(tx_carrier_errors);
207 NET_DEV_STAT(tx_fifo_errors);
208 NET_DEV_STAT(tx_heartbeat_errors);
209 NET_DEV_STAT(tx_window_errors);
210 NET_DEV_STAT(rx_compressed);
211 NET_DEV_STAT(tx_compressed);
212};
213#undef NET_DEV_STAT
214
215/* per-cpu stats, allocated on demand.
216 * Try to fit them in a single cache line, for dev_get_stats() sake.
217 */
218struct net_device_core_stats {
219 unsigned long rx_dropped;
220 unsigned long tx_dropped;
221 unsigned long rx_nohandler;
222 unsigned long rx_otherhost_dropped;
223} __aligned(4 * sizeof(unsigned long));
224
225#include <linux/cache.h>
226#include <linux/skbuff.h>
227
228struct neighbour;
229struct neigh_parms;
230struct sk_buff;
231
232struct netdev_hw_addr {
233 struct list_head list;
234 struct rb_node node;
235 unsigned char addr[MAX_ADDR_LEN];
236 unsigned char type;
237#define NETDEV_HW_ADDR_T_LAN 1
238#define NETDEV_HW_ADDR_T_SAN 2
239#define NETDEV_HW_ADDR_T_UNICAST 3
240#define NETDEV_HW_ADDR_T_MULTICAST 4
241 bool global_use;
242 int sync_cnt;
243 int refcount;
244 int synced;
245 struct rcu_head rcu_head;
246};
247
248struct netdev_hw_addr_list {
249 struct list_head list;
250 int count;
251
252 /* Auxiliary tree for faster lookup on addition and deletion */
253 struct rb_root tree;
254};
255
256#define netdev_hw_addr_list_count(l) ((l)->count)
257#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
258#define netdev_hw_addr_list_for_each(ha, l) \
259 list_for_each_entry(ha, &(l)->list, list)
260
261#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
262#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
263#define netdev_for_each_uc_addr(ha, dev) \
264 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
265#define netdev_for_each_synced_uc_addr(_ha, _dev) \
266 netdev_for_each_uc_addr((_ha), (_dev)) \
267 if ((_ha)->sync_cnt)
268
269#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
270#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
271#define netdev_for_each_mc_addr(ha, dev) \
272 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
273#define netdev_for_each_synced_mc_addr(_ha, _dev) \
274 netdev_for_each_mc_addr((_ha), (_dev)) \
275 if ((_ha)->sync_cnt)
276
277struct hh_cache {
278 unsigned int hh_len;
279 seqlock_t hh_lock;
280
281 /* cached hardware header; allow for machine alignment needs. */
282#define HH_DATA_MOD 16
283#define HH_DATA_OFF(__len) \
284 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
285#define HH_DATA_ALIGN(__len) \
286 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
287 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
288};
289
290/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
291 * Alternative is:
292 * dev->hard_header_len ? (dev->hard_header_len +
293 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
294 *
295 * We could use other alignment values, but we must maintain the
296 * relationship HH alignment <= LL alignment.
297 */
298#define LL_RESERVED_SPACE(dev) \
299 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
300 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
301#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
302 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
303 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
304
305struct header_ops {
306 int (*create) (struct sk_buff *skb, struct net_device *dev,
307 unsigned short type, const void *daddr,
308 const void *saddr, unsigned int len);
309 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
310 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
311 void (*cache_update)(struct hh_cache *hh,
312 const struct net_device *dev,
313 const unsigned char *haddr);
314 bool (*validate)(const char *ll_header, unsigned int len);
315 __be16 (*parse_protocol)(const struct sk_buff *skb);
316};
317
318/* These flag bits are private to the generic network queueing
319 * layer; they may not be explicitly referenced by any other
320 * code.
321 */
322
323enum netdev_state_t {
324 __LINK_STATE_START,
325 __LINK_STATE_PRESENT,
326 __LINK_STATE_NOCARRIER,
327 __LINK_STATE_LINKWATCH_PENDING,
328 __LINK_STATE_DORMANT,
329 __LINK_STATE_TESTING,
330};
331
332struct gro_list {
333 struct list_head list;
334 int count;
335};
336
337/*
338 * size of gro hash buckets, must less than bit number of
339 * napi_struct::gro_bitmask
340 */
341#define GRO_HASH_BUCKETS 8
342
343/*
344 * Structure for NAPI scheduling similar to tasklet but with weighting
345 */
346struct napi_struct {
347 /* The poll_list must only be managed by the entity which
348 * changes the state of the NAPI_STATE_SCHED bit. This means
349 * whoever atomically sets that bit can add this napi_struct
350 * to the per-CPU poll_list, and whoever clears that bit
351 * can remove from the list right before clearing the bit.
352 */
353 struct list_head poll_list;
354
355 unsigned long state;
356 int weight;
357 int defer_hard_irqs_count;
358 unsigned long gro_bitmask;
359 int (*poll)(struct napi_struct *, int);
360#ifdef CONFIG_NETPOLL
361 /* CPU actively polling if netpoll is configured */
362 int poll_owner;
363#endif
364 /* CPU on which NAPI has been scheduled for processing */
365 int list_owner;
366 struct net_device *dev;
367 struct gro_list gro_hash[GRO_HASH_BUCKETS];
368 struct sk_buff *skb;
369 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
370 int rx_count; /* length of rx_list */
371 unsigned int napi_id;
372 struct hrtimer timer;
373 struct task_struct *thread;
374 /* control-path-only fields follow */
375 struct list_head dev_list;
376 struct hlist_node napi_hash_node;
377 int irq;
378};
379
380enum {
381 NAPI_STATE_SCHED, /* Poll is scheduled */
382 NAPI_STATE_MISSED, /* reschedule a napi */
383 NAPI_STATE_DISABLE, /* Disable pending */
384 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
385 NAPI_STATE_LISTED, /* NAPI added to system lists */
386 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
387 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
388 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
389 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
390 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
391};
392
393enum {
394 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
395 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
396 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
397 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
398 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
399 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
400 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
401 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
402 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
403 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
404};
405
406enum gro_result {
407 GRO_MERGED,
408 GRO_MERGED_FREE,
409 GRO_HELD,
410 GRO_NORMAL,
411 GRO_CONSUMED,
412};
413typedef enum gro_result gro_result_t;
414
415/*
416 * enum rx_handler_result - Possible return values for rx_handlers.
417 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
418 * further.
419 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
420 * case skb->dev was changed by rx_handler.
421 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
422 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
423 *
424 * rx_handlers are functions called from inside __netif_receive_skb(), to do
425 * special processing of the skb, prior to delivery to protocol handlers.
426 *
427 * Currently, a net_device can only have a single rx_handler registered. Trying
428 * to register a second rx_handler will return -EBUSY.
429 *
430 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
431 * To unregister a rx_handler on a net_device, use
432 * netdev_rx_handler_unregister().
433 *
434 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
435 * do with the skb.
436 *
437 * If the rx_handler consumed the skb in some way, it should return
438 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
439 * the skb to be delivered in some other way.
440 *
441 * If the rx_handler changed skb->dev, to divert the skb to another
442 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
443 * new device will be called if it exists.
444 *
445 * If the rx_handler decides the skb should be ignored, it should return
446 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
447 * are registered on exact device (ptype->dev == skb->dev).
448 *
449 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
450 * delivered, it should return RX_HANDLER_PASS.
451 *
452 * A device without a registered rx_handler will behave as if rx_handler
453 * returned RX_HANDLER_PASS.
454 */
455
456enum rx_handler_result {
457 RX_HANDLER_CONSUMED,
458 RX_HANDLER_ANOTHER,
459 RX_HANDLER_EXACT,
460 RX_HANDLER_PASS,
461};
462typedef enum rx_handler_result rx_handler_result_t;
463typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
464
465void __napi_schedule(struct napi_struct *n);
466void __napi_schedule_irqoff(struct napi_struct *n);
467
468static inline bool napi_disable_pending(struct napi_struct *n)
469{
470 return test_bit(NAPI_STATE_DISABLE, &n->state);
471}
472
473static inline bool napi_prefer_busy_poll(struct napi_struct *n)
474{
475 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
476}
477
478/**
479 * napi_is_scheduled - test if NAPI is scheduled
480 * @n: NAPI context
481 *
482 * This check is "best-effort". With no locking implemented,
483 * a NAPI can be scheduled or terminate right after this check
484 * and produce not precise results.
485 *
486 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
487 * should not be used normally and napi_schedule should be
488 * used instead.
489 *
490 * Use only if the driver really needs to check if a NAPI
491 * is scheduled for example in the context of delayed timer
492 * that can be skipped if a NAPI is already scheduled.
493 *
494 * Return True if NAPI is scheduled, False otherwise.
495 */
496static inline bool napi_is_scheduled(struct napi_struct *n)
497{
498 return test_bit(NAPI_STATE_SCHED, &n->state);
499}
500
501bool napi_schedule_prep(struct napi_struct *n);
502
503/**
504 * napi_schedule - schedule NAPI poll
505 * @n: NAPI context
506 *
507 * Schedule NAPI poll routine to be called if it is not already
508 * running.
509 * Return true if we schedule a NAPI or false if not.
510 * Refer to napi_schedule_prep() for additional reason on why
511 * a NAPI might not be scheduled.
512 */
513static inline bool napi_schedule(struct napi_struct *n)
514{
515 if (napi_schedule_prep(n)) {
516 __napi_schedule(n);
517 return true;
518 }
519
520 return false;
521}
522
523/**
524 * napi_schedule_irqoff - schedule NAPI poll
525 * @n: NAPI context
526 *
527 * Variant of napi_schedule(), assuming hard irqs are masked.
528 */
529static inline void napi_schedule_irqoff(struct napi_struct *n)
530{
531 if (napi_schedule_prep(n))
532 __napi_schedule_irqoff(n);
533}
534
535/**
536 * napi_complete_done - NAPI processing complete
537 * @n: NAPI context
538 * @work_done: number of packets processed
539 *
540 * Mark NAPI processing as complete. Should only be called if poll budget
541 * has not been completely consumed.
542 * Prefer over napi_complete().
543 * Return false if device should avoid rearming interrupts.
544 */
545bool napi_complete_done(struct napi_struct *n, int work_done);
546
547static inline bool napi_complete(struct napi_struct *n)
548{
549 return napi_complete_done(n, work_done: 0);
550}
551
552int dev_set_threaded(struct net_device *dev, bool threaded);
553
554/**
555 * napi_disable - prevent NAPI from scheduling
556 * @n: NAPI context
557 *
558 * Stop NAPI from being scheduled on this context.
559 * Waits till any outstanding processing completes.
560 */
561void napi_disable(struct napi_struct *n);
562
563void napi_enable(struct napi_struct *n);
564
565/**
566 * napi_synchronize - wait until NAPI is not running
567 * @n: NAPI context
568 *
569 * Wait until NAPI is done being scheduled on this context.
570 * Waits till any outstanding processing completes but
571 * does not disable future activations.
572 */
573static inline void napi_synchronize(const struct napi_struct *n)
574{
575 if (IS_ENABLED(CONFIG_SMP))
576 while (test_bit(NAPI_STATE_SCHED, &n->state))
577 msleep(msecs: 1);
578 else
579 barrier();
580}
581
582/**
583 * napi_if_scheduled_mark_missed - if napi is running, set the
584 * NAPIF_STATE_MISSED
585 * @n: NAPI context
586 *
587 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
588 * NAPI is scheduled.
589 **/
590static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
591{
592 unsigned long val, new;
593
594 val = READ_ONCE(n->state);
595 do {
596 if (val & NAPIF_STATE_DISABLE)
597 return true;
598
599 if (!(val & NAPIF_STATE_SCHED))
600 return false;
601
602 new = val | NAPIF_STATE_MISSED;
603 } while (!try_cmpxchg(&n->state, &val, new));
604
605 return true;
606}
607
608enum netdev_queue_state_t {
609 __QUEUE_STATE_DRV_XOFF,
610 __QUEUE_STATE_STACK_XOFF,
611 __QUEUE_STATE_FROZEN,
612};
613
614#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
615#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
616#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
617
618#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
619#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
620 QUEUE_STATE_FROZEN)
621#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
622 QUEUE_STATE_FROZEN)
623
624/*
625 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
626 * netif_tx_* functions below are used to manipulate this flag. The
627 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
628 * queue independently. The netif_xmit_*stopped functions below are called
629 * to check if the queue has been stopped by the driver or stack (either
630 * of the XOFF bits are set in the state). Drivers should not need to call
631 * netif_xmit*stopped functions, they should only be using netif_tx_*.
632 */
633
634struct netdev_queue {
635/*
636 * read-mostly part
637 */
638 struct net_device *dev;
639 netdevice_tracker dev_tracker;
640
641 struct Qdisc __rcu *qdisc;
642 struct Qdisc __rcu *qdisc_sleeping;
643#ifdef CONFIG_SYSFS
644 struct kobject kobj;
645#endif
646#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
647 int numa_node;
648#endif
649 unsigned long tx_maxrate;
650 /*
651 * Number of TX timeouts for this queue
652 * (/sys/class/net/DEV/Q/trans_timeout)
653 */
654 atomic_long_t trans_timeout;
655
656 /* Subordinate device that the queue has been assigned to */
657 struct net_device *sb_dev;
658#ifdef CONFIG_XDP_SOCKETS
659 struct xsk_buff_pool *pool;
660#endif
661 /* NAPI instance for the queue
662 * Readers and writers must hold RTNL
663 */
664 struct napi_struct *napi;
665/*
666 * write-mostly part
667 */
668 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
669 int xmit_lock_owner;
670 /*
671 * Time (in jiffies) of last Tx
672 */
673 unsigned long trans_start;
674
675 unsigned long state;
676
677#ifdef CONFIG_BQL
678 struct dql dql;
679#endif
680} ____cacheline_aligned_in_smp;
681
682extern int sysctl_fb_tunnels_only_for_init_net;
683extern int sysctl_devconf_inherit_init_net;
684
685/*
686 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
687 * == 1 : For initns only
688 * == 2 : For none.
689 */
690static inline bool net_has_fallback_tunnels(const struct net *net)
691{
692#if IS_ENABLED(CONFIG_SYSCTL)
693 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
694
695 return !fb_tunnels_only_for_init_net ||
696 (net_eq(net1: net, net2: &init_net) && fb_tunnels_only_for_init_net == 1);
697#else
698 return true;
699#endif
700}
701
702static inline int net_inherit_devconf(void)
703{
704#if IS_ENABLED(CONFIG_SYSCTL)
705 return READ_ONCE(sysctl_devconf_inherit_init_net);
706#else
707 return 0;
708#endif
709}
710
711static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
712{
713#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
714 return q->numa_node;
715#else
716 return NUMA_NO_NODE;
717#endif
718}
719
720static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
721{
722#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
723 q->numa_node = node;
724#endif
725}
726
727#ifdef CONFIG_RFS_ACCEL
728bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
729 u16 filter_id);
730#endif
731
732/* XPS map type and offset of the xps map within net_device->xps_maps[]. */
733enum xps_map_type {
734 XPS_CPUS = 0,
735 XPS_RXQS,
736 XPS_MAPS_MAX,
737};
738
739#ifdef CONFIG_XPS
740/*
741 * This structure holds an XPS map which can be of variable length. The
742 * map is an array of queues.
743 */
744struct xps_map {
745 unsigned int len;
746 unsigned int alloc_len;
747 struct rcu_head rcu;
748 u16 queues[];
749};
750#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
751#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
752 - sizeof(struct xps_map)) / sizeof(u16))
753
754/*
755 * This structure holds all XPS maps for device. Maps are indexed by CPU.
756 *
757 * We keep track of the number of cpus/rxqs used when the struct is allocated,
758 * in nr_ids. This will help not accessing out-of-bound memory.
759 *
760 * We keep track of the number of traffic classes used when the struct is
761 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
762 * not crossing its upper bound, as the original dev->num_tc can be updated in
763 * the meantime.
764 */
765struct xps_dev_maps {
766 struct rcu_head rcu;
767 unsigned int nr_ids;
768 s16 num_tc;
769 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
770};
771
772#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
773 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
774
775#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
776 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
777
778#endif /* CONFIG_XPS */
779
780#define TC_MAX_QUEUE 16
781#define TC_BITMASK 15
782/* HW offloaded queuing disciplines txq count and offset maps */
783struct netdev_tc_txq {
784 u16 count;
785 u16 offset;
786};
787
788#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
789/*
790 * This structure is to hold information about the device
791 * configured to run FCoE protocol stack.
792 */
793struct netdev_fcoe_hbainfo {
794 char manufacturer[64];
795 char serial_number[64];
796 char hardware_version[64];
797 char driver_version[64];
798 char optionrom_version[64];
799 char firmware_version[64];
800 char model[256];
801 char model_description[256];
802};
803#endif
804
805#define MAX_PHYS_ITEM_ID_LEN 32
806
807/* This structure holds a unique identifier to identify some
808 * physical item (port for example) used by a netdevice.
809 */
810struct netdev_phys_item_id {
811 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
812 unsigned char id_len;
813};
814
815static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
816 struct netdev_phys_item_id *b)
817{
818 return a->id_len == b->id_len &&
819 memcmp(p: a->id, q: b->id, size: a->id_len) == 0;
820}
821
822typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
823 struct sk_buff *skb,
824 struct net_device *sb_dev);
825
826enum net_device_path_type {
827 DEV_PATH_ETHERNET = 0,
828 DEV_PATH_VLAN,
829 DEV_PATH_BRIDGE,
830 DEV_PATH_PPPOE,
831 DEV_PATH_DSA,
832 DEV_PATH_MTK_WDMA,
833};
834
835struct net_device_path {
836 enum net_device_path_type type;
837 const struct net_device *dev;
838 union {
839 struct {
840 u16 id;
841 __be16 proto;
842 u8 h_dest[ETH_ALEN];
843 } encap;
844 struct {
845 enum {
846 DEV_PATH_BR_VLAN_KEEP,
847 DEV_PATH_BR_VLAN_TAG,
848 DEV_PATH_BR_VLAN_UNTAG,
849 DEV_PATH_BR_VLAN_UNTAG_HW,
850 } vlan_mode;
851 u16 vlan_id;
852 __be16 vlan_proto;
853 } bridge;
854 struct {
855 int port;
856 u16 proto;
857 } dsa;
858 struct {
859 u8 wdma_idx;
860 u8 queue;
861 u16 wcid;
862 u8 bss;
863 u8 amsdu;
864 } mtk_wdma;
865 };
866};
867
868#define NET_DEVICE_PATH_STACK_MAX 5
869#define NET_DEVICE_PATH_VLAN_MAX 2
870
871struct net_device_path_stack {
872 int num_paths;
873 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
874};
875
876struct net_device_path_ctx {
877 const struct net_device *dev;
878 u8 daddr[ETH_ALEN];
879
880 int num_vlans;
881 struct {
882 u16 id;
883 __be16 proto;
884 } vlan[NET_DEVICE_PATH_VLAN_MAX];
885};
886
887enum tc_setup_type {
888 TC_QUERY_CAPS,
889 TC_SETUP_QDISC_MQPRIO,
890 TC_SETUP_CLSU32,
891 TC_SETUP_CLSFLOWER,
892 TC_SETUP_CLSMATCHALL,
893 TC_SETUP_CLSBPF,
894 TC_SETUP_BLOCK,
895 TC_SETUP_QDISC_CBS,
896 TC_SETUP_QDISC_RED,
897 TC_SETUP_QDISC_PRIO,
898 TC_SETUP_QDISC_MQ,
899 TC_SETUP_QDISC_ETF,
900 TC_SETUP_ROOT_QDISC,
901 TC_SETUP_QDISC_GRED,
902 TC_SETUP_QDISC_TAPRIO,
903 TC_SETUP_FT,
904 TC_SETUP_QDISC_ETS,
905 TC_SETUP_QDISC_TBF,
906 TC_SETUP_QDISC_FIFO,
907 TC_SETUP_QDISC_HTB,
908 TC_SETUP_ACT,
909};
910
911/* These structures hold the attributes of bpf state that are being passed
912 * to the netdevice through the bpf op.
913 */
914enum bpf_netdev_command {
915 /* Set or clear a bpf program used in the earliest stages of packet
916 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
917 * is responsible for calling bpf_prog_put on any old progs that are
918 * stored. In case of error, the callee need not release the new prog
919 * reference, but on success it takes ownership and must bpf_prog_put
920 * when it is no longer used.
921 */
922 XDP_SETUP_PROG,
923 XDP_SETUP_PROG_HW,
924 /* BPF program for offload callbacks, invoked at program load time. */
925 BPF_OFFLOAD_MAP_ALLOC,
926 BPF_OFFLOAD_MAP_FREE,
927 XDP_SETUP_XSK_POOL,
928};
929
930struct bpf_prog_offload_ops;
931struct netlink_ext_ack;
932struct xdp_umem;
933struct xdp_dev_bulk_queue;
934struct bpf_xdp_link;
935
936enum bpf_xdp_mode {
937 XDP_MODE_SKB = 0,
938 XDP_MODE_DRV = 1,
939 XDP_MODE_HW = 2,
940 __MAX_XDP_MODE
941};
942
943struct bpf_xdp_entity {
944 struct bpf_prog *prog;
945 struct bpf_xdp_link *link;
946};
947
948struct netdev_bpf {
949 enum bpf_netdev_command command;
950 union {
951 /* XDP_SETUP_PROG */
952 struct {
953 u32 flags;
954 struct bpf_prog *prog;
955 struct netlink_ext_ack *extack;
956 };
957 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
958 struct {
959 struct bpf_offloaded_map *offmap;
960 };
961 /* XDP_SETUP_XSK_POOL */
962 struct {
963 struct xsk_buff_pool *pool;
964 u16 queue_id;
965 } xsk;
966 };
967};
968
969/* Flags for ndo_xsk_wakeup. */
970#define XDP_WAKEUP_RX (1 << 0)
971#define XDP_WAKEUP_TX (1 << 1)
972
973#ifdef CONFIG_XFRM_OFFLOAD
974struct xfrmdev_ops {
975 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
976 void (*xdo_dev_state_delete) (struct xfrm_state *x);
977 void (*xdo_dev_state_free) (struct xfrm_state *x);
978 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
979 struct xfrm_state *x);
980 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
981 void (*xdo_dev_state_update_stats) (struct xfrm_state *x);
982 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
983 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
984 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
985};
986#endif
987
988struct dev_ifalias {
989 struct rcu_head rcuhead;
990 char ifalias[];
991};
992
993struct devlink;
994struct tlsdev_ops;
995
996struct netdev_net_notifier {
997 struct list_head list;
998 struct notifier_block *nb;
999};
1000
1001/*
1002 * This structure defines the management hooks for network devices.
1003 * The following hooks can be defined; unless noted otherwise, they are
1004 * optional and can be filled with a null pointer.
1005 *
1006 * int (*ndo_init)(struct net_device *dev);
1007 * This function is called once when a network device is registered.
1008 * The network device can use this for any late stage initialization
1009 * or semantic validation. It can fail with an error code which will
1010 * be propagated back to register_netdev.
1011 *
1012 * void (*ndo_uninit)(struct net_device *dev);
1013 * This function is called when device is unregistered or when registration
1014 * fails. It is not called if init fails.
1015 *
1016 * int (*ndo_open)(struct net_device *dev);
1017 * This function is called when a network device transitions to the up
1018 * state.
1019 *
1020 * int (*ndo_stop)(struct net_device *dev);
1021 * This function is called when a network device transitions to the down
1022 * state.
1023 *
1024 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1025 * struct net_device *dev);
1026 * Called when a packet needs to be transmitted.
1027 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1028 * the queue before that can happen; it's for obsolete devices and weird
1029 * corner cases, but the stack really does a non-trivial amount
1030 * of useless work if you return NETDEV_TX_BUSY.
1031 * Required; cannot be NULL.
1032 *
1033 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1034 * struct net_device *dev
1035 * netdev_features_t features);
1036 * Called by core transmit path to determine if device is capable of
1037 * performing offload operations on a given packet. This is to give
1038 * the device an opportunity to implement any restrictions that cannot
1039 * be otherwise expressed by feature flags. The check is called with
1040 * the set of features that the stack has calculated and it returns
1041 * those the driver believes to be appropriate.
1042 *
1043 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1044 * struct net_device *sb_dev);
1045 * Called to decide which queue to use when device supports multiple
1046 * transmit queues.
1047 *
1048 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1049 * This function is called to allow device receiver to make
1050 * changes to configuration when multicast or promiscuous is enabled.
1051 *
1052 * void (*ndo_set_rx_mode)(struct net_device *dev);
1053 * This function is called device changes address list filtering.
1054 * If driver handles unicast address filtering, it should set
1055 * IFF_UNICAST_FLT in its priv_flags.
1056 *
1057 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1058 * This function is called when the Media Access Control address
1059 * needs to be changed. If this interface is not defined, the
1060 * MAC address can not be changed.
1061 *
1062 * int (*ndo_validate_addr)(struct net_device *dev);
1063 * Test if Media Access Control address is valid for the device.
1064 *
1065 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1066 * Old-style ioctl entry point. This is used internally by the
1067 * appletalk and ieee802154 subsystems but is no longer called by
1068 * the device ioctl handler.
1069 *
1070 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1071 * Used by the bonding driver for its device specific ioctls:
1072 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1073 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1074 *
1075 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1076 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1077 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1078 *
1079 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1080 * Used to set network devices bus interface parameters. This interface
1081 * is retained for legacy reasons; new devices should use the bus
1082 * interface (PCI) for low level management.
1083 *
1084 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1085 * Called when a user wants to change the Maximum Transfer Unit
1086 * of a device.
1087 *
1088 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1089 * Callback used when the transmitter has not made any progress
1090 * for dev->watchdog ticks.
1091 *
1092 * void (*ndo_get_stats64)(struct net_device *dev,
1093 * struct rtnl_link_stats64 *storage);
1094 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1095 * Called when a user wants to get the network device usage
1096 * statistics. Drivers must do one of the following:
1097 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1098 * rtnl_link_stats64 structure passed by the caller.
1099 * 2. Define @ndo_get_stats to update a net_device_stats structure
1100 * (which should normally be dev->stats) and return a pointer to
1101 * it. The structure may be changed asynchronously only if each
1102 * field is written atomically.
1103 * 3. Update dev->stats asynchronously and atomically, and define
1104 * neither operation.
1105 *
1106 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1107 * Return true if this device supports offload stats of this attr_id.
1108 *
1109 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1110 * void *attr_data)
1111 * Get statistics for offload operations by attr_id. Write it into the
1112 * attr_data pointer.
1113 *
1114 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1115 * If device supports VLAN filtering this function is called when a
1116 * VLAN id is registered.
1117 *
1118 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1119 * If device supports VLAN filtering this function is called when a
1120 * VLAN id is unregistered.
1121 *
1122 * void (*ndo_poll_controller)(struct net_device *dev);
1123 *
1124 * SR-IOV management functions.
1125 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1126 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1127 * u8 qos, __be16 proto);
1128 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1129 * int max_tx_rate);
1130 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1131 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1132 * int (*ndo_get_vf_config)(struct net_device *dev,
1133 * int vf, struct ifla_vf_info *ivf);
1134 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1135 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1136 * struct nlattr *port[]);
1137 *
1138 * Enable or disable the VF ability to query its RSS Redirection Table and
1139 * Hash Key. This is needed since on some devices VF share this information
1140 * with PF and querying it may introduce a theoretical security risk.
1141 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1142 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1143 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1144 * void *type_data);
1145 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1146 * This is always called from the stack with the rtnl lock held and netif
1147 * tx queues stopped. This allows the netdevice to perform queue
1148 * management safely.
1149 *
1150 * Fiber Channel over Ethernet (FCoE) offload functions.
1151 * int (*ndo_fcoe_enable)(struct net_device *dev);
1152 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1153 * so the underlying device can perform whatever needed configuration or
1154 * initialization to support acceleration of FCoE traffic.
1155 *
1156 * int (*ndo_fcoe_disable)(struct net_device *dev);
1157 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1158 * so the underlying device can perform whatever needed clean-ups to
1159 * stop supporting acceleration of FCoE traffic.
1160 *
1161 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1162 * struct scatterlist *sgl, unsigned int sgc);
1163 * Called when the FCoE Initiator wants to initialize an I/O that
1164 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1165 * perform necessary setup and returns 1 to indicate the device is set up
1166 * successfully to perform DDP on this I/O, otherwise this returns 0.
1167 *
1168 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1169 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1170 * indicated by the FC exchange id 'xid', so the underlying device can
1171 * clean up and reuse resources for later DDP requests.
1172 *
1173 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1174 * struct scatterlist *sgl, unsigned int sgc);
1175 * Called when the FCoE Target wants to initialize an I/O that
1176 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1177 * perform necessary setup and returns 1 to indicate the device is set up
1178 * successfully to perform DDP on this I/O, otherwise this returns 0.
1179 *
1180 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1181 * struct netdev_fcoe_hbainfo *hbainfo);
1182 * Called when the FCoE Protocol stack wants information on the underlying
1183 * device. This information is utilized by the FCoE protocol stack to
1184 * register attributes with Fiber Channel management service as per the
1185 * FC-GS Fabric Device Management Information(FDMI) specification.
1186 *
1187 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1188 * Called when the underlying device wants to override default World Wide
1189 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1190 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1191 * protocol stack to use.
1192 *
1193 * RFS acceleration.
1194 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1195 * u16 rxq_index, u32 flow_id);
1196 * Set hardware filter for RFS. rxq_index is the target queue index;
1197 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1198 * Return the filter ID on success, or a negative error code.
1199 *
1200 * Slave management functions (for bridge, bonding, etc).
1201 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1202 * Called to make another netdev an underling.
1203 *
1204 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1205 * Called to release previously enslaved netdev.
1206 *
1207 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1208 * struct sk_buff *skb,
1209 * bool all_slaves);
1210 * Get the xmit slave of master device. If all_slaves is true, function
1211 * assume all the slaves can transmit.
1212 *
1213 * Feature/offload setting functions.
1214 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1215 * netdev_features_t features);
1216 * Adjusts the requested feature flags according to device-specific
1217 * constraints, and returns the resulting flags. Must not modify
1218 * the device state.
1219 *
1220 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1221 * Called to update device configuration to new features. Passed
1222 * feature set might be less than what was returned by ndo_fix_features()).
1223 * Must return >0 or -errno if it changed dev->features itself.
1224 *
1225 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1226 * struct net_device *dev,
1227 * const unsigned char *addr, u16 vid, u16 flags,
1228 * struct netlink_ext_ack *extack);
1229 * Adds an FDB entry to dev for addr.
1230 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1231 * struct net_device *dev,
1232 * const unsigned char *addr, u16 vid)
1233 * Deletes the FDB entry from dev coresponding to addr.
1234 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1235 * struct netlink_ext_ack *extack);
1236 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1237 * struct net_device *dev, struct net_device *filter_dev,
1238 * int *idx)
1239 * Used to add FDB entries to dump requests. Implementers should add
1240 * entries to skb and update idx with the number of entries.
1241 *
1242 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1243 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1244 * Adds an MDB entry to dev.
1245 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1246 * struct netlink_ext_ack *extack);
1247 * Deletes the MDB entry from dev.
1248 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1249 * struct netlink_ext_ack *extack);
1250 * Bulk deletes MDB entries from dev.
1251 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1252 * struct netlink_callback *cb);
1253 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1254 * callback is used by core rtnetlink code.
1255 *
1256 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1257 * u16 flags, struct netlink_ext_ack *extack)
1258 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1259 * struct net_device *dev, u32 filter_mask,
1260 * int nlflags)
1261 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1262 * u16 flags);
1263 *
1264 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1265 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1266 * which do not represent real hardware may define this to allow their
1267 * userspace components to manage their virtual carrier state. Devices
1268 * that determine carrier state from physical hardware properties (eg
1269 * network cables) or protocol-dependent mechanisms (eg
1270 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1271 *
1272 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1273 * struct netdev_phys_item_id *ppid);
1274 * Called to get ID of physical port of this device. If driver does
1275 * not implement this, it is assumed that the hw is not able to have
1276 * multiple net devices on single physical port.
1277 *
1278 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1279 * struct netdev_phys_item_id *ppid)
1280 * Called to get the parent ID of the physical port of this device.
1281 *
1282 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1283 * struct net_device *dev)
1284 * Called by upper layer devices to accelerate switching or other
1285 * station functionality into hardware. 'pdev is the lowerdev
1286 * to use for the offload and 'dev' is the net device that will
1287 * back the offload. Returns a pointer to the private structure
1288 * the upper layer will maintain.
1289 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1290 * Called by upper layer device to delete the station created
1291 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1292 * the station and priv is the structure returned by the add
1293 * operation.
1294 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1295 * int queue_index, u32 maxrate);
1296 * Called when a user wants to set a max-rate limitation of specific
1297 * TX queue.
1298 * int (*ndo_get_iflink)(const struct net_device *dev);
1299 * Called to get the iflink value of this device.
1300 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1301 * This function is used to get egress tunnel information for given skb.
1302 * This is useful for retrieving outer tunnel header parameters while
1303 * sampling packet.
1304 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1305 * This function is used to specify the headroom that the skb must
1306 * consider when allocation skb during packet reception. Setting
1307 * appropriate rx headroom value allows avoiding skb head copy on
1308 * forward. Setting a negative value resets the rx headroom to the
1309 * default value.
1310 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1311 * This function is used to set or query state related to XDP on the
1312 * netdevice and manage BPF offload. See definition of
1313 * enum bpf_netdev_command for details.
1314 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1315 * u32 flags);
1316 * This function is used to submit @n XDP packets for transmit on a
1317 * netdevice. Returns number of frames successfully transmitted, frames
1318 * that got dropped are freed/returned via xdp_return_frame().
1319 * Returns negative number, means general error invoking ndo, meaning
1320 * no frames were xmit'ed and core-caller will free all frames.
1321 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1322 * struct xdp_buff *xdp);
1323 * Get the xmit slave of master device based on the xdp_buff.
1324 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1325 * This function is used to wake up the softirq, ksoftirqd or kthread
1326 * responsible for sending and/or receiving packets on a specific
1327 * queue id bound to an AF_XDP socket. The flags field specifies if
1328 * only RX, only Tx, or both should be woken up using the flags
1329 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1330 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1331 * int cmd);
1332 * Add, change, delete or get information on an IPv4 tunnel.
1333 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1334 * If a device is paired with a peer device, return the peer instance.
1335 * The caller must be under RCU read context.
1336 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1337 * Get the forwarding path to reach the real device from the HW destination address
1338 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1339 * const struct skb_shared_hwtstamps *hwtstamps,
1340 * bool cycles);
1341 * Get hardware timestamp based on normal/adjustable time or free running
1342 * cycle counter. This function is required if physical clock supports a
1343 * free running cycle counter.
1344 *
1345 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1346 * struct kernel_hwtstamp_config *kernel_config);
1347 * Get the currently configured hardware timestamping parameters for the
1348 * NIC device.
1349 *
1350 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1351 * struct kernel_hwtstamp_config *kernel_config,
1352 * struct netlink_ext_ack *extack);
1353 * Change the hardware timestamping parameters for NIC device.
1354 */
1355struct net_device_ops {
1356 int (*ndo_init)(struct net_device *dev);
1357 void (*ndo_uninit)(struct net_device *dev);
1358 int (*ndo_open)(struct net_device *dev);
1359 int (*ndo_stop)(struct net_device *dev);
1360 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1361 struct net_device *dev);
1362 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1363 struct net_device *dev,
1364 netdev_features_t features);
1365 u16 (*ndo_select_queue)(struct net_device *dev,
1366 struct sk_buff *skb,
1367 struct net_device *sb_dev);
1368 void (*ndo_change_rx_flags)(struct net_device *dev,
1369 int flags);
1370 void (*ndo_set_rx_mode)(struct net_device *dev);
1371 int (*ndo_set_mac_address)(struct net_device *dev,
1372 void *addr);
1373 int (*ndo_validate_addr)(struct net_device *dev);
1374 int (*ndo_do_ioctl)(struct net_device *dev,
1375 struct ifreq *ifr, int cmd);
1376 int (*ndo_eth_ioctl)(struct net_device *dev,
1377 struct ifreq *ifr, int cmd);
1378 int (*ndo_siocbond)(struct net_device *dev,
1379 struct ifreq *ifr, int cmd);
1380 int (*ndo_siocwandev)(struct net_device *dev,
1381 struct if_settings *ifs);
1382 int (*ndo_siocdevprivate)(struct net_device *dev,
1383 struct ifreq *ifr,
1384 void __user *data, int cmd);
1385 int (*ndo_set_config)(struct net_device *dev,
1386 struct ifmap *map);
1387 int (*ndo_change_mtu)(struct net_device *dev,
1388 int new_mtu);
1389 int (*ndo_neigh_setup)(struct net_device *dev,
1390 struct neigh_parms *);
1391 void (*ndo_tx_timeout) (struct net_device *dev,
1392 unsigned int txqueue);
1393
1394 void (*ndo_get_stats64)(struct net_device *dev,
1395 struct rtnl_link_stats64 *storage);
1396 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1397 int (*ndo_get_offload_stats)(int attr_id,
1398 const struct net_device *dev,
1399 void *attr_data);
1400 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1401
1402 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1403 __be16 proto, u16 vid);
1404 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1405 __be16 proto, u16 vid);
1406#ifdef CONFIG_NET_POLL_CONTROLLER
1407 void (*ndo_poll_controller)(struct net_device *dev);
1408 int (*ndo_netpoll_setup)(struct net_device *dev,
1409 struct netpoll_info *info);
1410 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1411#endif
1412 int (*ndo_set_vf_mac)(struct net_device *dev,
1413 int queue, u8 *mac);
1414 int (*ndo_set_vf_vlan)(struct net_device *dev,
1415 int queue, u16 vlan,
1416 u8 qos, __be16 proto);
1417 int (*ndo_set_vf_rate)(struct net_device *dev,
1418 int vf, int min_tx_rate,
1419 int max_tx_rate);
1420 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1421 int vf, bool setting);
1422 int (*ndo_set_vf_trust)(struct net_device *dev,
1423 int vf, bool setting);
1424 int (*ndo_get_vf_config)(struct net_device *dev,
1425 int vf,
1426 struct ifla_vf_info *ivf);
1427 int (*ndo_set_vf_link_state)(struct net_device *dev,
1428 int vf, int link_state);
1429 int (*ndo_get_vf_stats)(struct net_device *dev,
1430 int vf,
1431 struct ifla_vf_stats
1432 *vf_stats);
1433 int (*ndo_set_vf_port)(struct net_device *dev,
1434 int vf,
1435 struct nlattr *port[]);
1436 int (*ndo_get_vf_port)(struct net_device *dev,
1437 int vf, struct sk_buff *skb);
1438 int (*ndo_get_vf_guid)(struct net_device *dev,
1439 int vf,
1440 struct ifla_vf_guid *node_guid,
1441 struct ifla_vf_guid *port_guid);
1442 int (*ndo_set_vf_guid)(struct net_device *dev,
1443 int vf, u64 guid,
1444 int guid_type);
1445 int (*ndo_set_vf_rss_query_en)(
1446 struct net_device *dev,
1447 int vf, bool setting);
1448 int (*ndo_setup_tc)(struct net_device *dev,
1449 enum tc_setup_type type,
1450 void *type_data);
1451#if IS_ENABLED(CONFIG_FCOE)
1452 int (*ndo_fcoe_enable)(struct net_device *dev);
1453 int (*ndo_fcoe_disable)(struct net_device *dev);
1454 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1455 u16 xid,
1456 struct scatterlist *sgl,
1457 unsigned int sgc);
1458 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1459 u16 xid);
1460 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1461 u16 xid,
1462 struct scatterlist *sgl,
1463 unsigned int sgc);
1464 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1465 struct netdev_fcoe_hbainfo *hbainfo);
1466#endif
1467
1468#if IS_ENABLED(CONFIG_LIBFCOE)
1469#define NETDEV_FCOE_WWNN 0
1470#define NETDEV_FCOE_WWPN 1
1471 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1472 u64 *wwn, int type);
1473#endif
1474
1475#ifdef CONFIG_RFS_ACCEL
1476 int (*ndo_rx_flow_steer)(struct net_device *dev,
1477 const struct sk_buff *skb,
1478 u16 rxq_index,
1479 u32 flow_id);
1480#endif
1481 int (*ndo_add_slave)(struct net_device *dev,
1482 struct net_device *slave_dev,
1483 struct netlink_ext_ack *extack);
1484 int (*ndo_del_slave)(struct net_device *dev,
1485 struct net_device *slave_dev);
1486 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1487 struct sk_buff *skb,
1488 bool all_slaves);
1489 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1490 struct sock *sk);
1491 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1492 netdev_features_t features);
1493 int (*ndo_set_features)(struct net_device *dev,
1494 netdev_features_t features);
1495 int (*ndo_neigh_construct)(struct net_device *dev,
1496 struct neighbour *n);
1497 void (*ndo_neigh_destroy)(struct net_device *dev,
1498 struct neighbour *n);
1499
1500 int (*ndo_fdb_add)(struct ndmsg *ndm,
1501 struct nlattr *tb[],
1502 struct net_device *dev,
1503 const unsigned char *addr,
1504 u16 vid,
1505 u16 flags,
1506 struct netlink_ext_ack *extack);
1507 int (*ndo_fdb_del)(struct ndmsg *ndm,
1508 struct nlattr *tb[],
1509 struct net_device *dev,
1510 const unsigned char *addr,
1511 u16 vid, struct netlink_ext_ack *extack);
1512 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1513 struct net_device *dev,
1514 struct netlink_ext_ack *extack);
1515 int (*ndo_fdb_dump)(struct sk_buff *skb,
1516 struct netlink_callback *cb,
1517 struct net_device *dev,
1518 struct net_device *filter_dev,
1519 int *idx);
1520 int (*ndo_fdb_get)(struct sk_buff *skb,
1521 struct nlattr *tb[],
1522 struct net_device *dev,
1523 const unsigned char *addr,
1524 u16 vid, u32 portid, u32 seq,
1525 struct netlink_ext_ack *extack);
1526 int (*ndo_mdb_add)(struct net_device *dev,
1527 struct nlattr *tb[],
1528 u16 nlmsg_flags,
1529 struct netlink_ext_ack *extack);
1530 int (*ndo_mdb_del)(struct net_device *dev,
1531 struct nlattr *tb[],
1532 struct netlink_ext_ack *extack);
1533 int (*ndo_mdb_del_bulk)(struct net_device *dev,
1534 struct nlattr *tb[],
1535 struct netlink_ext_ack *extack);
1536 int (*ndo_mdb_dump)(struct net_device *dev,
1537 struct sk_buff *skb,
1538 struct netlink_callback *cb);
1539 int (*ndo_mdb_get)(struct net_device *dev,
1540 struct nlattr *tb[], u32 portid,
1541 u32 seq,
1542 struct netlink_ext_ack *extack);
1543 int (*ndo_bridge_setlink)(struct net_device *dev,
1544 struct nlmsghdr *nlh,
1545 u16 flags,
1546 struct netlink_ext_ack *extack);
1547 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1548 u32 pid, u32 seq,
1549 struct net_device *dev,
1550 u32 filter_mask,
1551 int nlflags);
1552 int (*ndo_bridge_dellink)(struct net_device *dev,
1553 struct nlmsghdr *nlh,
1554 u16 flags);
1555 int (*ndo_change_carrier)(struct net_device *dev,
1556 bool new_carrier);
1557 int (*ndo_get_phys_port_id)(struct net_device *dev,
1558 struct netdev_phys_item_id *ppid);
1559 int (*ndo_get_port_parent_id)(struct net_device *dev,
1560 struct netdev_phys_item_id *ppid);
1561 int (*ndo_get_phys_port_name)(struct net_device *dev,
1562 char *name, size_t len);
1563 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1564 struct net_device *dev);
1565 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1566 void *priv);
1567
1568 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1569 int queue_index,
1570 u32 maxrate);
1571 int (*ndo_get_iflink)(const struct net_device *dev);
1572 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1573 struct sk_buff *skb);
1574 void (*ndo_set_rx_headroom)(struct net_device *dev,
1575 int needed_headroom);
1576 int (*ndo_bpf)(struct net_device *dev,
1577 struct netdev_bpf *bpf);
1578 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1579 struct xdp_frame **xdp,
1580 u32 flags);
1581 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1582 struct xdp_buff *xdp);
1583 int (*ndo_xsk_wakeup)(struct net_device *dev,
1584 u32 queue_id, u32 flags);
1585 int (*ndo_tunnel_ctl)(struct net_device *dev,
1586 struct ip_tunnel_parm *p, int cmd);
1587 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1588 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1589 struct net_device_path *path);
1590 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1591 const struct skb_shared_hwtstamps *hwtstamps,
1592 bool cycles);
1593 int (*ndo_hwtstamp_get)(struct net_device *dev,
1594 struct kernel_hwtstamp_config *kernel_config);
1595 int (*ndo_hwtstamp_set)(struct net_device *dev,
1596 struct kernel_hwtstamp_config *kernel_config,
1597 struct netlink_ext_ack *extack);
1598};
1599
1600/**
1601 * enum netdev_priv_flags - &struct net_device priv_flags
1602 *
1603 * These are the &struct net_device, they are only set internally
1604 * by drivers and used in the kernel. These flags are invisible to
1605 * userspace; this means that the order of these flags can change
1606 * during any kernel release.
1607 *
1608 * You should have a pretty good reason to be extending these flags.
1609 *
1610 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1611 * @IFF_EBRIDGE: Ethernet bridging device
1612 * @IFF_BONDING: bonding master or slave
1613 * @IFF_ISATAP: ISATAP interface (RFC4214)
1614 * @IFF_WAN_HDLC: WAN HDLC device
1615 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1616 * release skb->dst
1617 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1618 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1619 * @IFF_MACVLAN_PORT: device used as macvlan port
1620 * @IFF_BRIDGE_PORT: device used as bridge port
1621 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1622 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1623 * @IFF_UNICAST_FLT: Supports unicast filtering
1624 * @IFF_TEAM_PORT: device used as team port
1625 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1626 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1627 * change when it's running
1628 * @IFF_MACVLAN: Macvlan device
1629 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1630 * underlying stacked devices
1631 * @IFF_L3MDEV_MASTER: device is an L3 master device
1632 * @IFF_NO_QUEUE: device can run without qdisc attached
1633 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1634 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1635 * @IFF_TEAM: device is a team device
1636 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1637 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1638 * entity (i.e. the master device for bridged veth)
1639 * @IFF_MACSEC: device is a MACsec device
1640 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1641 * @IFF_FAILOVER: device is a failover master device
1642 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1643 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1644 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1645 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1646 * skb_headlen(skb) == 0 (data starts from frag0)
1647 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1648 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1649 * ndo_hwtstamp_set() for all timestamp requests regardless of source,
1650 * even if those aren't HWTSTAMP_SOURCE_NETDEV.
1651 */
1652enum netdev_priv_flags {
1653 IFF_802_1Q_VLAN = 1<<0,
1654 IFF_EBRIDGE = 1<<1,
1655 IFF_BONDING = 1<<2,
1656 IFF_ISATAP = 1<<3,
1657 IFF_WAN_HDLC = 1<<4,
1658 IFF_XMIT_DST_RELEASE = 1<<5,
1659 IFF_DONT_BRIDGE = 1<<6,
1660 IFF_DISABLE_NETPOLL = 1<<7,
1661 IFF_MACVLAN_PORT = 1<<8,
1662 IFF_BRIDGE_PORT = 1<<9,
1663 IFF_OVS_DATAPATH = 1<<10,
1664 IFF_TX_SKB_SHARING = 1<<11,
1665 IFF_UNICAST_FLT = 1<<12,
1666 IFF_TEAM_PORT = 1<<13,
1667 IFF_SUPP_NOFCS = 1<<14,
1668 IFF_LIVE_ADDR_CHANGE = 1<<15,
1669 IFF_MACVLAN = 1<<16,
1670 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1671 IFF_L3MDEV_MASTER = 1<<18,
1672 IFF_NO_QUEUE = 1<<19,
1673 IFF_OPENVSWITCH = 1<<20,
1674 IFF_L3MDEV_SLAVE = 1<<21,
1675 IFF_TEAM = 1<<22,
1676 IFF_RXFH_CONFIGURED = 1<<23,
1677 IFF_PHONY_HEADROOM = 1<<24,
1678 IFF_MACSEC = 1<<25,
1679 IFF_NO_RX_HANDLER = 1<<26,
1680 IFF_FAILOVER = 1<<27,
1681 IFF_FAILOVER_SLAVE = 1<<28,
1682 IFF_L3MDEV_RX_HANDLER = 1<<29,
1683 IFF_NO_ADDRCONF = BIT_ULL(30),
1684 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1685 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1686 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33),
1687};
1688
1689#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1690#define IFF_EBRIDGE IFF_EBRIDGE
1691#define IFF_BONDING IFF_BONDING
1692#define IFF_ISATAP IFF_ISATAP
1693#define IFF_WAN_HDLC IFF_WAN_HDLC
1694#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1695#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1696#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1697#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1698#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1699#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1700#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1701#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1702#define IFF_TEAM_PORT IFF_TEAM_PORT
1703#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1704#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1705#define IFF_MACVLAN IFF_MACVLAN
1706#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1707#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1708#define IFF_NO_QUEUE IFF_NO_QUEUE
1709#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1710#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1711#define IFF_TEAM IFF_TEAM
1712#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1713#define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1714#define IFF_MACSEC IFF_MACSEC
1715#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1716#define IFF_FAILOVER IFF_FAILOVER
1717#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1718#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1719#define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1720
1721/* Specifies the type of the struct net_device::ml_priv pointer */
1722enum netdev_ml_priv_type {
1723 ML_PRIV_NONE,
1724 ML_PRIV_CAN,
1725};
1726
1727enum netdev_stat_type {
1728 NETDEV_PCPU_STAT_NONE,
1729 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1730 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1731 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1732};
1733
1734enum netdev_reg_state {
1735 NETREG_UNINITIALIZED = 0,
1736 NETREG_REGISTERED, /* completed register_netdevice */
1737 NETREG_UNREGISTERING, /* called unregister_netdevice */
1738 NETREG_UNREGISTERED, /* completed unregister todo */
1739 NETREG_RELEASED, /* called free_netdev */
1740 NETREG_DUMMY, /* dummy device for NAPI poll */
1741};
1742
1743/**
1744 * struct net_device - The DEVICE structure.
1745 *
1746 * Actually, this whole structure is a big mistake. It mixes I/O
1747 * data with strictly "high-level" data, and it has to know about
1748 * almost every data structure used in the INET module.
1749 *
1750 * @name: This is the first field of the "visible" part of this structure
1751 * (i.e. as seen by users in the "Space.c" file). It is the name
1752 * of the interface.
1753 *
1754 * @name_node: Name hashlist node
1755 * @ifalias: SNMP alias
1756 * @mem_end: Shared memory end
1757 * @mem_start: Shared memory start
1758 * @base_addr: Device I/O address
1759 * @irq: Device IRQ number
1760 *
1761 * @state: Generic network queuing layer state, see netdev_state_t
1762 * @dev_list: The global list of network devices
1763 * @napi_list: List entry used for polling NAPI devices
1764 * @unreg_list: List entry when we are unregistering the
1765 * device; see the function unregister_netdev
1766 * @close_list: List entry used when we are closing the device
1767 * @ptype_all: Device-specific packet handlers for all protocols
1768 * @ptype_specific: Device-specific, protocol-specific packet handlers
1769 *
1770 * @adj_list: Directly linked devices, like slaves for bonding
1771 * @features: Currently active device features
1772 * @hw_features: User-changeable features
1773 *
1774 * @wanted_features: User-requested features
1775 * @vlan_features: Mask of features inheritable by VLAN devices
1776 *
1777 * @hw_enc_features: Mask of features inherited by encapsulating devices
1778 * This field indicates what encapsulation
1779 * offloads the hardware is capable of doing,
1780 * and drivers will need to set them appropriately.
1781 *
1782 * @mpls_features: Mask of features inheritable by MPLS
1783 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1784 *
1785 * @ifindex: interface index
1786 * @group: The group the device belongs to
1787 *
1788 * @stats: Statistics struct, which was left as a legacy, use
1789 * rtnl_link_stats64 instead
1790 *
1791 * @core_stats: core networking counters,
1792 * do not use this in drivers
1793 * @carrier_up_count: Number of times the carrier has been up
1794 * @carrier_down_count: Number of times the carrier has been down
1795 *
1796 * @wireless_handlers: List of functions to handle Wireless Extensions,
1797 * instead of ioctl,
1798 * see <net/iw_handler.h> for details.
1799 * @wireless_data: Instance data managed by the core of wireless extensions
1800 *
1801 * @netdev_ops: Includes several pointers to callbacks,
1802 * if one wants to override the ndo_*() functions
1803 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1804 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks.
1805 * @ethtool_ops: Management operations
1806 * @l3mdev_ops: Layer 3 master device operations
1807 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1808 * discovery handling. Necessary for e.g. 6LoWPAN.
1809 * @xfrmdev_ops: Transformation offload operations
1810 * @tlsdev_ops: Transport Layer Security offload operations
1811 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1812 * of Layer 2 headers.
1813 *
1814 * @flags: Interface flags (a la BSD)
1815 * @xdp_features: XDP capability supported by the device
1816 * @priv_flags: Like 'flags' but invisible to userspace,
1817 * see if.h for the definitions
1818 * @gflags: Global flags ( kept as legacy )
1819 * @padded: How much padding added by alloc_netdev()
1820 * @operstate: RFC2863 operstate
1821 * @link_mode: Mapping policy to operstate
1822 * @if_port: Selectable AUI, TP, ...
1823 * @dma: DMA channel
1824 * @mtu: Interface MTU value
1825 * @min_mtu: Interface Minimum MTU value
1826 * @max_mtu: Interface Maximum MTU value
1827 * @type: Interface hardware type
1828 * @hard_header_len: Maximum hardware header length.
1829 * @min_header_len: Minimum hardware header length
1830 *
1831 * @needed_headroom: Extra headroom the hardware may need, but not in all
1832 * cases can this be guaranteed
1833 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1834 * cases can this be guaranteed. Some cases also use
1835 * LL_MAX_HEADER instead to allocate the skb
1836 *
1837 * interface address info:
1838 *
1839 * @perm_addr: Permanent hw address
1840 * @addr_assign_type: Hw address assignment type
1841 * @addr_len: Hardware address length
1842 * @upper_level: Maximum depth level of upper devices.
1843 * @lower_level: Maximum depth level of lower devices.
1844 * @neigh_priv_len: Used in neigh_alloc()
1845 * @dev_id: Used to differentiate devices that share
1846 * the same link layer address
1847 * @dev_port: Used to differentiate devices that share
1848 * the same function
1849 * @addr_list_lock: XXX: need comments on this one
1850 * @name_assign_type: network interface name assignment type
1851 * @uc_promisc: Counter that indicates promiscuous mode
1852 * has been enabled due to the need to listen to
1853 * additional unicast addresses in a device that
1854 * does not implement ndo_set_rx_mode()
1855 * @uc: unicast mac addresses
1856 * @mc: multicast mac addresses
1857 * @dev_addrs: list of device hw addresses
1858 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1859 * @promiscuity: Number of times the NIC is told to work in
1860 * promiscuous mode; if it becomes 0 the NIC will
1861 * exit promiscuous mode
1862 * @allmulti: Counter, enables or disables allmulticast mode
1863 *
1864 * @vlan_info: VLAN info
1865 * @dsa_ptr: dsa specific data
1866 * @tipc_ptr: TIPC specific data
1867 * @atalk_ptr: AppleTalk link
1868 * @ip_ptr: IPv4 specific data
1869 * @ip6_ptr: IPv6 specific data
1870 * @ax25_ptr: AX.25 specific data
1871 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1872 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1873 * device struct
1874 * @mpls_ptr: mpls_dev struct pointer
1875 * @mctp_ptr: MCTP specific data
1876 *
1877 * @dev_addr: Hw address (before bcast,
1878 * because most packets are unicast)
1879 *
1880 * @_rx: Array of RX queues
1881 * @num_rx_queues: Number of RX queues
1882 * allocated at register_netdev() time
1883 * @real_num_rx_queues: Number of RX queues currently active in device
1884 * @xdp_prog: XDP sockets filter program pointer
1885 * @gro_flush_timeout: timeout for GRO layer in NAPI
1886 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1887 * allow to avoid NIC hard IRQ, on busy queues.
1888 *
1889 * @rx_handler: handler for received packets
1890 * @rx_handler_data: XXX: need comments on this one
1891 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1892 * @ingress_queue: XXX: need comments on this one
1893 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1894 * @broadcast: hw bcast address
1895 *
1896 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1897 * indexed by RX queue number. Assigned by driver.
1898 * This must only be set if the ndo_rx_flow_steer
1899 * operation is defined
1900 * @index_hlist: Device index hash chain
1901 *
1902 * @_tx: Array of TX queues
1903 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1904 * @real_num_tx_queues: Number of TX queues currently active in device
1905 * @qdisc: Root qdisc from userspace point of view
1906 * @tx_queue_len: Max frames per queue allowed
1907 * @tx_global_lock: XXX: need comments on this one
1908 * @xdp_bulkq: XDP device bulk queue
1909 * @xps_maps: all CPUs/RXQs maps for XPS device
1910 *
1911 * @xps_maps: XXX: need comments on this one
1912 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1913 * @nf_hooks_egress: netfilter hooks executed for egress packets
1914 * @qdisc_hash: qdisc hash table
1915 * @watchdog_timeo: Represents the timeout that is used by
1916 * the watchdog (see dev_watchdog())
1917 * @watchdog_timer: List of timers
1918 *
1919 * @proto_down_reason: reason a netdev interface is held down
1920 * @pcpu_refcnt: Number of references to this device
1921 * @dev_refcnt: Number of references to this device
1922 * @refcnt_tracker: Tracker directory for tracked references to this device
1923 * @todo_list: Delayed register/unregister
1924 * @link_watch_list: XXX: need comments on this one
1925 *
1926 * @reg_state: Register/unregister state machine
1927 * @dismantle: Device is going to be freed
1928 * @rtnl_link_state: This enum represents the phases of creating
1929 * a new link
1930 *
1931 * @needs_free_netdev: Should unregister perform free_netdev?
1932 * @priv_destructor: Called from unregister
1933 * @npinfo: XXX: need comments on this one
1934 * @nd_net: Network namespace this network device is inside
1935 *
1936 * @ml_priv: Mid-layer private
1937 * @ml_priv_type: Mid-layer private type
1938 *
1939 * @pcpu_stat_type: Type of device statistics which the core should
1940 * allocate/free: none, lstats, tstats, dstats. none
1941 * means the driver is handling statistics allocation/
1942 * freeing internally.
1943 * @lstats: Loopback statistics: packets, bytes
1944 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes
1945 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes
1946 *
1947 * @garp_port: GARP
1948 * @mrp_port: MRP
1949 *
1950 * @dm_private: Drop monitor private
1951 *
1952 * @dev: Class/net/name entry
1953 * @sysfs_groups: Space for optional device, statistics and wireless
1954 * sysfs groups
1955 *
1956 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1957 * @rtnl_link_ops: Rtnl_link_ops
1958 * @stat_ops: Optional ops for queue-aware statistics
1959 *
1960 * @gso_max_size: Maximum size of generic segmentation offload
1961 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1962 * @gso_max_segs: Maximum number of segments that can be passed to the
1963 * NIC for GSO
1964 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1965 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
1966 * for IPv4.
1967 *
1968 * @dcbnl_ops: Data Center Bridging netlink ops
1969 * @num_tc: Number of traffic classes in the net device
1970 * @tc_to_txq: XXX: need comments on this one
1971 * @prio_tc_map: XXX: need comments on this one
1972 *
1973 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1974 *
1975 * @priomap: XXX: need comments on this one
1976 * @phydev: Physical device may attach itself
1977 * for hardware timestamping
1978 * @sfp_bus: attached &struct sfp_bus structure.
1979 *
1980 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1981 *
1982 * @proto_down: protocol port state information can be sent to the
1983 * switch driver and used to set the phys state of the
1984 * switch port.
1985 *
1986 * @wol_enabled: Wake-on-LAN is enabled
1987 *
1988 * @threaded: napi threaded mode is enabled
1989 *
1990 * @net_notifier_list: List of per-net netdev notifier block
1991 * that follow this device when it is moved
1992 * to another network namespace.
1993 *
1994 * @macsec_ops: MACsec offloading ops
1995 *
1996 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1997 * offload capabilities of the device
1998 * @udp_tunnel_nic: UDP tunnel offload state
1999 * @xdp_state: stores info on attached XDP BPF programs
2000 *
2001 * @nested_level: Used as a parameter of spin_lock_nested() of
2002 * dev->addr_list_lock.
2003 * @unlink_list: As netif_addr_lock() can be called recursively,
2004 * keep a list of interfaces to be deleted.
2005 * @gro_max_size: Maximum size of aggregated packet in generic
2006 * receive offload (GRO)
2007 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2008 * receive offload (GRO), for IPv4.
2009 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2010 * zero copy driver
2011 *
2012 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2013 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2014 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2015 * @dev_registered_tracker: tracker for reference held while
2016 * registered
2017 * @offload_xstats_l3: L3 HW stats for this netdevice.
2018 *
2019 * @devlink_port: Pointer to related devlink port structure.
2020 * Assigned by a driver before netdev registration using
2021 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2022 * during the time netdevice is registered.
2023 *
2024 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2025 * where the clock is recovered.
2026 *
2027 * FIXME: cleanup struct net_device such that network protocol info
2028 * moves out.
2029 */
2030
2031struct net_device {
2032 /* Cacheline organization can be found documented in
2033 * Documentation/networking/net_cachelines/net_device.rst.
2034 * Please update the document when adding new fields.
2035 */
2036
2037 /* TX read-mostly hotpath */
2038 __cacheline_group_begin(net_device_read_tx);
2039 unsigned long long priv_flags;
2040 const struct net_device_ops *netdev_ops;
2041 const struct header_ops *header_ops;
2042 struct netdev_queue *_tx;
2043 netdev_features_t gso_partial_features;
2044 unsigned int real_num_tx_queues;
2045 unsigned int gso_max_size;
2046 unsigned int gso_ipv4_max_size;
2047 u16 gso_max_segs;
2048 s16 num_tc;
2049 /* Note : dev->mtu is often read without holding a lock.
2050 * Writers usually hold RTNL.
2051 * It is recommended to use READ_ONCE() to annotate the reads,
2052 * and to use WRITE_ONCE() to annotate the writes.
2053 */
2054 unsigned int mtu;
2055 unsigned short needed_headroom;
2056 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2057#ifdef CONFIG_XPS
2058 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2059#endif
2060#ifdef CONFIG_NETFILTER_EGRESS
2061 struct nf_hook_entries __rcu *nf_hooks_egress;
2062#endif
2063#ifdef CONFIG_NET_XGRESS
2064 struct bpf_mprog_entry __rcu *tcx_egress;
2065#endif
2066 __cacheline_group_end(net_device_read_tx);
2067
2068 /* TXRX read-mostly hotpath */
2069 __cacheline_group_begin(net_device_read_txrx);
2070 union {
2071 struct pcpu_lstats __percpu *lstats;
2072 struct pcpu_sw_netstats __percpu *tstats;
2073 struct pcpu_dstats __percpu *dstats;
2074 };
2075 unsigned long state;
2076 unsigned int flags;
2077 unsigned short hard_header_len;
2078 netdev_features_t features;
2079 struct inet6_dev __rcu *ip6_ptr;
2080 __cacheline_group_end(net_device_read_txrx);
2081
2082 /* RX read-mostly hotpath */
2083 __cacheline_group_begin(net_device_read_rx);
2084 struct bpf_prog __rcu *xdp_prog;
2085 struct list_head ptype_specific;
2086 int ifindex;
2087 unsigned int real_num_rx_queues;
2088 struct netdev_rx_queue *_rx;
2089 unsigned long gro_flush_timeout;
2090 int napi_defer_hard_irqs;
2091 unsigned int gro_max_size;
2092 unsigned int gro_ipv4_max_size;
2093 rx_handler_func_t __rcu *rx_handler;
2094 void __rcu *rx_handler_data;
2095 possible_net_t nd_net;
2096#ifdef CONFIG_NETPOLL
2097 struct netpoll_info __rcu *npinfo;
2098#endif
2099#ifdef CONFIG_NET_XGRESS
2100 struct bpf_mprog_entry __rcu *tcx_ingress;
2101#endif
2102 __cacheline_group_end(net_device_read_rx);
2103
2104 char name[IFNAMSIZ];
2105 struct netdev_name_node *name_node;
2106 struct dev_ifalias __rcu *ifalias;
2107 /*
2108 * I/O specific fields
2109 * FIXME: Merge these and struct ifmap into one
2110 */
2111 unsigned long mem_end;
2112 unsigned long mem_start;
2113 unsigned long base_addr;
2114
2115 /*
2116 * Some hardware also needs these fields (state,dev_list,
2117 * napi_list,unreg_list,close_list) but they are not
2118 * part of the usual set specified in Space.c.
2119 */
2120
2121
2122 struct list_head dev_list;
2123 struct list_head napi_list;
2124 struct list_head unreg_list;
2125 struct list_head close_list;
2126 struct list_head ptype_all;
2127
2128 struct {
2129 struct list_head upper;
2130 struct list_head lower;
2131 } adj_list;
2132
2133 /* Read-mostly cache-line for fast-path access */
2134 xdp_features_t xdp_features;
2135 const struct xdp_metadata_ops *xdp_metadata_ops;
2136 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2137 unsigned short gflags;
2138
2139 unsigned short needed_tailroom;
2140
2141 netdev_features_t hw_features;
2142 netdev_features_t wanted_features;
2143 netdev_features_t vlan_features;
2144 netdev_features_t hw_enc_features;
2145 netdev_features_t mpls_features;
2146
2147 unsigned int min_mtu;
2148 unsigned int max_mtu;
2149 unsigned short type;
2150 unsigned char min_header_len;
2151 unsigned char name_assign_type;
2152
2153 int group;
2154
2155 struct net_device_stats stats; /* not used by modern drivers */
2156
2157 struct net_device_core_stats __percpu *core_stats;
2158
2159 /* Stats to monitor link on/off, flapping */
2160 atomic_t carrier_up_count;
2161 atomic_t carrier_down_count;
2162
2163#ifdef CONFIG_WIRELESS_EXT
2164 const struct iw_handler_def *wireless_handlers;
2165 struct iw_public_data *wireless_data;
2166#endif
2167 const struct ethtool_ops *ethtool_ops;
2168#ifdef CONFIG_NET_L3_MASTER_DEV
2169 const struct l3mdev_ops *l3mdev_ops;
2170#endif
2171#if IS_ENABLED(CONFIG_IPV6)
2172 const struct ndisc_ops *ndisc_ops;
2173#endif
2174
2175#ifdef CONFIG_XFRM_OFFLOAD
2176 const struct xfrmdev_ops *xfrmdev_ops;
2177#endif
2178
2179#if IS_ENABLED(CONFIG_TLS_DEVICE)
2180 const struct tlsdev_ops *tlsdev_ops;
2181#endif
2182
2183 unsigned int operstate;
2184 unsigned char link_mode;
2185
2186 unsigned char if_port;
2187 unsigned char dma;
2188
2189 /* Interface address info. */
2190 unsigned char perm_addr[MAX_ADDR_LEN];
2191 unsigned char addr_assign_type;
2192 unsigned char addr_len;
2193 unsigned char upper_level;
2194 unsigned char lower_level;
2195
2196 unsigned short neigh_priv_len;
2197 unsigned short dev_id;
2198 unsigned short dev_port;
2199 unsigned short padded;
2200
2201 spinlock_t addr_list_lock;
2202 int irq;
2203
2204 struct netdev_hw_addr_list uc;
2205 struct netdev_hw_addr_list mc;
2206 struct netdev_hw_addr_list dev_addrs;
2207
2208#ifdef CONFIG_SYSFS
2209 struct kset *queues_kset;
2210#endif
2211#ifdef CONFIG_LOCKDEP
2212 struct list_head unlink_list;
2213#endif
2214 unsigned int promiscuity;
2215 unsigned int allmulti;
2216 bool uc_promisc;
2217#ifdef CONFIG_LOCKDEP
2218 unsigned char nested_level;
2219#endif
2220
2221
2222 /* Protocol-specific pointers */
2223 struct in_device __rcu *ip_ptr;
2224#if IS_ENABLED(CONFIG_VLAN_8021Q)
2225 struct vlan_info __rcu *vlan_info;
2226#endif
2227#if IS_ENABLED(CONFIG_NET_DSA)
2228 struct dsa_port *dsa_ptr;
2229#endif
2230#if IS_ENABLED(CONFIG_TIPC)
2231 struct tipc_bearer __rcu *tipc_ptr;
2232#endif
2233#if IS_ENABLED(CONFIG_ATALK)
2234 void *atalk_ptr;
2235#endif
2236#if IS_ENABLED(CONFIG_AX25)
2237 void *ax25_ptr;
2238#endif
2239#if IS_ENABLED(CONFIG_CFG80211)
2240 struct wireless_dev *ieee80211_ptr;
2241#endif
2242#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2243 struct wpan_dev *ieee802154_ptr;
2244#endif
2245#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2246 struct mpls_dev __rcu *mpls_ptr;
2247#endif
2248#if IS_ENABLED(CONFIG_MCTP)
2249 struct mctp_dev __rcu *mctp_ptr;
2250#endif
2251
2252/*
2253 * Cache lines mostly used on receive path (including eth_type_trans())
2254 */
2255 /* Interface address info used in eth_type_trans() */
2256 const unsigned char *dev_addr;
2257
2258 unsigned int num_rx_queues;
2259#define GRO_LEGACY_MAX_SIZE 65536u
2260/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2261 * and shinfo->gso_segs is a 16bit field.
2262 */
2263#define GRO_MAX_SIZE (8 * 65535u)
2264 unsigned int xdp_zc_max_segs;
2265 struct netdev_queue __rcu *ingress_queue;
2266#ifdef CONFIG_NETFILTER_INGRESS
2267 struct nf_hook_entries __rcu *nf_hooks_ingress;
2268#endif
2269
2270 unsigned char broadcast[MAX_ADDR_LEN];
2271#ifdef CONFIG_RFS_ACCEL
2272 struct cpu_rmap *rx_cpu_rmap;
2273#endif
2274 struct hlist_node index_hlist;
2275
2276/*
2277 * Cache lines mostly used on transmit path
2278 */
2279 unsigned int num_tx_queues;
2280 struct Qdisc __rcu *qdisc;
2281 unsigned int tx_queue_len;
2282 spinlock_t tx_global_lock;
2283
2284 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2285
2286#ifdef CONFIG_NET_SCHED
2287 DECLARE_HASHTABLE (qdisc_hash, 4);
2288#endif
2289 /* These may be needed for future network-power-down code. */
2290 struct timer_list watchdog_timer;
2291 int watchdog_timeo;
2292
2293 u32 proto_down_reason;
2294
2295 struct list_head todo_list;
2296
2297#ifdef CONFIG_PCPU_DEV_REFCNT
2298 int __percpu *pcpu_refcnt;
2299#else
2300 refcount_t dev_refcnt;
2301#endif
2302 struct ref_tracker_dir refcnt_tracker;
2303
2304 struct list_head link_watch_list;
2305
2306 u8 reg_state;
2307
2308 bool dismantle;
2309
2310 enum {
2311 RTNL_LINK_INITIALIZED,
2312 RTNL_LINK_INITIALIZING,
2313 } rtnl_link_state:16;
2314
2315 bool needs_free_netdev;
2316 void (*priv_destructor)(struct net_device *dev);
2317
2318 /* mid-layer private */
2319 void *ml_priv;
2320 enum netdev_ml_priv_type ml_priv_type;
2321
2322 enum netdev_stat_type pcpu_stat_type:8;
2323
2324#if IS_ENABLED(CONFIG_GARP)
2325 struct garp_port __rcu *garp_port;
2326#endif
2327#if IS_ENABLED(CONFIG_MRP)
2328 struct mrp_port __rcu *mrp_port;
2329#endif
2330#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2331 struct dm_hw_stat_delta __rcu *dm_private;
2332#endif
2333 struct device dev;
2334 const struct attribute_group *sysfs_groups[4];
2335 const struct attribute_group *sysfs_rx_queue_group;
2336
2337 const struct rtnl_link_ops *rtnl_link_ops;
2338
2339 const struct netdev_stat_ops *stat_ops;
2340
2341 /* for setting kernel sock attribute on TCP connection setup */
2342#define GSO_MAX_SEGS 65535u
2343#define GSO_LEGACY_MAX_SIZE 65536u
2344/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2345 * and shinfo->gso_segs is a 16bit field.
2346 */
2347#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2348
2349#define TSO_LEGACY_MAX_SIZE 65536
2350#define TSO_MAX_SIZE UINT_MAX
2351 unsigned int tso_max_size;
2352#define TSO_MAX_SEGS U16_MAX
2353 u16 tso_max_segs;
2354
2355#ifdef CONFIG_DCB
2356 const struct dcbnl_rtnl_ops *dcbnl_ops;
2357#endif
2358 u8 prio_tc_map[TC_BITMASK + 1];
2359
2360#if IS_ENABLED(CONFIG_FCOE)
2361 unsigned int fcoe_ddp_xid;
2362#endif
2363#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2364 struct netprio_map __rcu *priomap;
2365#endif
2366 struct phy_device *phydev;
2367 struct sfp_bus *sfp_bus;
2368 struct lock_class_key *qdisc_tx_busylock;
2369 bool proto_down;
2370 unsigned wol_enabled:1;
2371 unsigned threaded:1;
2372
2373 struct list_head net_notifier_list;
2374
2375#if IS_ENABLED(CONFIG_MACSEC)
2376 /* MACsec management functions */
2377 const struct macsec_ops *macsec_ops;
2378#endif
2379 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2380 struct udp_tunnel_nic *udp_tunnel_nic;
2381
2382 /* protected by rtnl_lock */
2383 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2384
2385 u8 dev_addr_shadow[MAX_ADDR_LEN];
2386 netdevice_tracker linkwatch_dev_tracker;
2387 netdevice_tracker watchdog_dev_tracker;
2388 netdevice_tracker dev_registered_tracker;
2389 struct rtnl_hw_stats64 *offload_xstats_l3;
2390
2391 struct devlink_port *devlink_port;
2392
2393#if IS_ENABLED(CONFIG_DPLL)
2394 struct dpll_pin __rcu *dpll_pin;
2395#endif
2396#if IS_ENABLED(CONFIG_PAGE_POOL)
2397 /** @page_pools: page pools created for this netdevice */
2398 struct hlist_head page_pools;
2399#endif
2400};
2401#define to_net_dev(d) container_of(d, struct net_device, dev)
2402
2403/*
2404 * Driver should use this to assign devlink port instance to a netdevice
2405 * before it registers the netdevice. Therefore devlink_port is static
2406 * during the netdev lifetime after it is registered.
2407 */
2408#define SET_NETDEV_DEVLINK_PORT(dev, port) \
2409({ \
2410 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2411 ((dev)->devlink_port = (port)); \
2412})
2413
2414static inline bool netif_elide_gro(const struct net_device *dev)
2415{
2416 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2417 return true;
2418 return false;
2419}
2420
2421#define NETDEV_ALIGN 32
2422
2423static inline
2424int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2425{
2426 return dev->prio_tc_map[prio & TC_BITMASK];
2427}
2428
2429static inline
2430int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2431{
2432 if (tc >= dev->num_tc)
2433 return -EINVAL;
2434
2435 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2436 return 0;
2437}
2438
2439int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2440void netdev_reset_tc(struct net_device *dev);
2441int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2442int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2443
2444static inline
2445int netdev_get_num_tc(struct net_device *dev)
2446{
2447 return dev->num_tc;
2448}
2449
2450static inline void net_prefetch(void *p)
2451{
2452 prefetch(p);
2453#if L1_CACHE_BYTES < 128
2454 prefetch((u8 *)p + L1_CACHE_BYTES);
2455#endif
2456}
2457
2458static inline void net_prefetchw(void *p)
2459{
2460 prefetchw(x: p);
2461#if L1_CACHE_BYTES < 128
2462 prefetchw(x: (u8 *)p + L1_CACHE_BYTES);
2463#endif
2464}
2465
2466void netdev_unbind_sb_channel(struct net_device *dev,
2467 struct net_device *sb_dev);
2468int netdev_bind_sb_channel_queue(struct net_device *dev,
2469 struct net_device *sb_dev,
2470 u8 tc, u16 count, u16 offset);
2471int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2472static inline int netdev_get_sb_channel(struct net_device *dev)
2473{
2474 return max_t(int, -dev->num_tc, 0);
2475}
2476
2477static inline
2478struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2479 unsigned int index)
2480{
2481 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2482 return &dev->_tx[index];
2483}
2484
2485static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2486 const struct sk_buff *skb)
2487{
2488 return netdev_get_tx_queue(dev, index: skb_get_queue_mapping(skb));
2489}
2490
2491static inline void netdev_for_each_tx_queue(struct net_device *dev,
2492 void (*f)(struct net_device *,
2493 struct netdev_queue *,
2494 void *),
2495 void *arg)
2496{
2497 unsigned int i;
2498
2499 for (i = 0; i < dev->num_tx_queues; i++)
2500 f(dev, &dev->_tx[i], arg);
2501}
2502
2503#define netdev_lockdep_set_classes(dev) \
2504{ \
2505 static struct lock_class_key qdisc_tx_busylock_key; \
2506 static struct lock_class_key qdisc_xmit_lock_key; \
2507 static struct lock_class_key dev_addr_list_lock_key; \
2508 unsigned int i; \
2509 \
2510 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2511 lockdep_set_class(&(dev)->addr_list_lock, \
2512 &dev_addr_list_lock_key); \
2513 for (i = 0; i < (dev)->num_tx_queues; i++) \
2514 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2515 &qdisc_xmit_lock_key); \
2516}
2517
2518u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2519 struct net_device *sb_dev);
2520struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2521 struct sk_buff *skb,
2522 struct net_device *sb_dev);
2523
2524/* returns the headroom that the master device needs to take in account
2525 * when forwarding to this dev
2526 */
2527static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2528{
2529 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2530}
2531
2532static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2533{
2534 if (dev->netdev_ops->ndo_set_rx_headroom)
2535 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2536}
2537
2538/* set the device rx headroom to the dev's default */
2539static inline void netdev_reset_rx_headroom(struct net_device *dev)
2540{
2541 netdev_set_rx_headroom(dev, new_hr: -1);
2542}
2543
2544static inline void *netdev_get_ml_priv(struct net_device *dev,
2545 enum netdev_ml_priv_type type)
2546{
2547 if (dev->ml_priv_type != type)
2548 return NULL;
2549
2550 return dev->ml_priv;
2551}
2552
2553static inline void netdev_set_ml_priv(struct net_device *dev,
2554 void *ml_priv,
2555 enum netdev_ml_priv_type type)
2556{
2557 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2558 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2559 dev->ml_priv_type, type);
2560 WARN(!dev->ml_priv_type && dev->ml_priv,
2561 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2562
2563 dev->ml_priv = ml_priv;
2564 dev->ml_priv_type = type;
2565}
2566
2567/*
2568 * Net namespace inlines
2569 */
2570static inline
2571struct net *dev_net(const struct net_device *dev)
2572{
2573 return read_pnet(pnet: &dev->nd_net);
2574}
2575
2576static inline
2577void dev_net_set(struct net_device *dev, struct net *net)
2578{
2579 write_pnet(pnet: &dev->nd_net, net);
2580}
2581
2582/**
2583 * netdev_priv - access network device private data
2584 * @dev: network device
2585 *
2586 * Get network device private data
2587 */
2588static inline void *netdev_priv(const struct net_device *dev)
2589{
2590 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2591}
2592
2593/* Set the sysfs physical device reference for the network logical device
2594 * if set prior to registration will cause a symlink during initialization.
2595 */
2596#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2597
2598/* Set the sysfs device type for the network logical device to allow
2599 * fine-grained identification of different network device types. For
2600 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2601 */
2602#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2603
2604void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2605 enum netdev_queue_type type,
2606 struct napi_struct *napi);
2607
2608static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2609{
2610 napi->irq = irq;
2611}
2612
2613/* Default NAPI poll() weight
2614 * Device drivers are strongly advised to not use bigger value
2615 */
2616#define NAPI_POLL_WEIGHT 64
2617
2618void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2619 int (*poll)(struct napi_struct *, int), int weight);
2620
2621/**
2622 * netif_napi_add() - initialize a NAPI context
2623 * @dev: network device
2624 * @napi: NAPI context
2625 * @poll: polling function
2626 *
2627 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2628 * *any* of the other NAPI-related functions.
2629 */
2630static inline void
2631netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2632 int (*poll)(struct napi_struct *, int))
2633{
2634 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2635}
2636
2637static inline void
2638netif_napi_add_tx_weight(struct net_device *dev,
2639 struct napi_struct *napi,
2640 int (*poll)(struct napi_struct *, int),
2641 int weight)
2642{
2643 set_bit(nr: NAPI_STATE_NO_BUSY_POLL, addr: &napi->state);
2644 netif_napi_add_weight(dev, napi, poll, weight);
2645}
2646
2647/**
2648 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2649 * @dev: network device
2650 * @napi: NAPI context
2651 * @poll: polling function
2652 *
2653 * This variant of netif_napi_add() should be used from drivers using NAPI
2654 * to exclusively poll a TX queue.
2655 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2656 */
2657static inline void netif_napi_add_tx(struct net_device *dev,
2658 struct napi_struct *napi,
2659 int (*poll)(struct napi_struct *, int))
2660{
2661 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2662}
2663
2664/**
2665 * __netif_napi_del - remove a NAPI context
2666 * @napi: NAPI context
2667 *
2668 * Warning: caller must observe RCU grace period before freeing memory
2669 * containing @napi. Drivers might want to call this helper to combine
2670 * all the needed RCU grace periods into a single one.
2671 */
2672void __netif_napi_del(struct napi_struct *napi);
2673
2674/**
2675 * netif_napi_del - remove a NAPI context
2676 * @napi: NAPI context
2677 *
2678 * netif_napi_del() removes a NAPI context from the network device NAPI list
2679 */
2680static inline void netif_napi_del(struct napi_struct *napi)
2681{
2682 __netif_napi_del(napi);
2683 synchronize_net();
2684}
2685
2686struct packet_type {
2687 __be16 type; /* This is really htons(ether_type). */
2688 bool ignore_outgoing;
2689 struct net_device *dev; /* NULL is wildcarded here */
2690 netdevice_tracker dev_tracker;
2691 int (*func) (struct sk_buff *,
2692 struct net_device *,
2693 struct packet_type *,
2694 struct net_device *);
2695 void (*list_func) (struct list_head *,
2696 struct packet_type *,
2697 struct net_device *);
2698 bool (*id_match)(struct packet_type *ptype,
2699 struct sock *sk);
2700 struct net *af_packet_net;
2701 void *af_packet_priv;
2702 struct list_head list;
2703};
2704
2705struct offload_callbacks {
2706 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2707 netdev_features_t features);
2708 struct sk_buff *(*gro_receive)(struct list_head *head,
2709 struct sk_buff *skb);
2710 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2711};
2712
2713struct packet_offload {
2714 __be16 type; /* This is really htons(ether_type). */
2715 u16 priority;
2716 struct offload_callbacks callbacks;
2717 struct list_head list;
2718};
2719
2720/* often modified stats are per-CPU, other are shared (netdev->stats) */
2721struct pcpu_sw_netstats {
2722 u64_stats_t rx_packets;
2723 u64_stats_t rx_bytes;
2724 u64_stats_t tx_packets;
2725 u64_stats_t tx_bytes;
2726 struct u64_stats_sync syncp;
2727} __aligned(4 * sizeof(u64));
2728
2729struct pcpu_dstats {
2730 u64 rx_packets;
2731 u64 rx_bytes;
2732 u64 rx_drops;
2733 u64 tx_packets;
2734 u64 tx_bytes;
2735 u64 tx_drops;
2736 struct u64_stats_sync syncp;
2737} __aligned(8 * sizeof(u64));
2738
2739struct pcpu_lstats {
2740 u64_stats_t packets;
2741 u64_stats_t bytes;
2742 struct u64_stats_sync syncp;
2743} __aligned(2 * sizeof(u64));
2744
2745void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2746
2747static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2748{
2749 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2750
2751 u64_stats_update_begin(syncp: &tstats->syncp);
2752 u64_stats_add(p: &tstats->rx_bytes, val: len);
2753 u64_stats_inc(p: &tstats->rx_packets);
2754 u64_stats_update_end(syncp: &tstats->syncp);
2755}
2756
2757static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2758 unsigned int packets,
2759 unsigned int len)
2760{
2761 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2762
2763 u64_stats_update_begin(syncp: &tstats->syncp);
2764 u64_stats_add(p: &tstats->tx_bytes, val: len);
2765 u64_stats_add(p: &tstats->tx_packets, val: packets);
2766 u64_stats_update_end(syncp: &tstats->syncp);
2767}
2768
2769static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2770{
2771 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2772
2773 u64_stats_update_begin(syncp: &lstats->syncp);
2774 u64_stats_add(p: &lstats->bytes, val: len);
2775 u64_stats_inc(p: &lstats->packets);
2776 u64_stats_update_end(syncp: &lstats->syncp);
2777}
2778
2779#define __netdev_alloc_pcpu_stats(type, gfp) \
2780({ \
2781 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2782 if (pcpu_stats) { \
2783 int __cpu; \
2784 for_each_possible_cpu(__cpu) { \
2785 typeof(type) *stat; \
2786 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2787 u64_stats_init(&stat->syncp); \
2788 } \
2789 } \
2790 pcpu_stats; \
2791})
2792
2793#define netdev_alloc_pcpu_stats(type) \
2794 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2795
2796#define devm_netdev_alloc_pcpu_stats(dev, type) \
2797({ \
2798 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2799 if (pcpu_stats) { \
2800 int __cpu; \
2801 for_each_possible_cpu(__cpu) { \
2802 typeof(type) *stat; \
2803 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2804 u64_stats_init(&stat->syncp); \
2805 } \
2806 } \
2807 pcpu_stats; \
2808})
2809
2810enum netdev_lag_tx_type {
2811 NETDEV_LAG_TX_TYPE_UNKNOWN,
2812 NETDEV_LAG_TX_TYPE_RANDOM,
2813 NETDEV_LAG_TX_TYPE_BROADCAST,
2814 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2815 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2816 NETDEV_LAG_TX_TYPE_HASH,
2817};
2818
2819enum netdev_lag_hash {
2820 NETDEV_LAG_HASH_NONE,
2821 NETDEV_LAG_HASH_L2,
2822 NETDEV_LAG_HASH_L34,
2823 NETDEV_LAG_HASH_L23,
2824 NETDEV_LAG_HASH_E23,
2825 NETDEV_LAG_HASH_E34,
2826 NETDEV_LAG_HASH_VLAN_SRCMAC,
2827 NETDEV_LAG_HASH_UNKNOWN,
2828};
2829
2830struct netdev_lag_upper_info {
2831 enum netdev_lag_tx_type tx_type;
2832 enum netdev_lag_hash hash_type;
2833};
2834
2835struct netdev_lag_lower_state_info {
2836 u8 link_up : 1,
2837 tx_enabled : 1;
2838};
2839
2840#include <linux/notifier.h>
2841
2842/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2843 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2844 * adding new types.
2845 */
2846enum netdev_cmd {
2847 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2848 NETDEV_DOWN,
2849 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2850 detected a hardware crash and restarted
2851 - we can use this eg to kick tcp sessions
2852 once done */
2853 NETDEV_CHANGE, /* Notify device state change */
2854 NETDEV_REGISTER,
2855 NETDEV_UNREGISTER,
2856 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2857 NETDEV_CHANGEADDR, /* notify after the address change */
2858 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2859 NETDEV_GOING_DOWN,
2860 NETDEV_CHANGENAME,
2861 NETDEV_FEAT_CHANGE,
2862 NETDEV_BONDING_FAILOVER,
2863 NETDEV_PRE_UP,
2864 NETDEV_PRE_TYPE_CHANGE,
2865 NETDEV_POST_TYPE_CHANGE,
2866 NETDEV_POST_INIT,
2867 NETDEV_PRE_UNINIT,
2868 NETDEV_RELEASE,
2869 NETDEV_NOTIFY_PEERS,
2870 NETDEV_JOIN,
2871 NETDEV_CHANGEUPPER,
2872 NETDEV_RESEND_IGMP,
2873 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2874 NETDEV_CHANGEINFODATA,
2875 NETDEV_BONDING_INFO,
2876 NETDEV_PRECHANGEUPPER,
2877 NETDEV_CHANGELOWERSTATE,
2878 NETDEV_UDP_TUNNEL_PUSH_INFO,
2879 NETDEV_UDP_TUNNEL_DROP_INFO,
2880 NETDEV_CHANGE_TX_QUEUE_LEN,
2881 NETDEV_CVLAN_FILTER_PUSH_INFO,
2882 NETDEV_CVLAN_FILTER_DROP_INFO,
2883 NETDEV_SVLAN_FILTER_PUSH_INFO,
2884 NETDEV_SVLAN_FILTER_DROP_INFO,
2885 NETDEV_OFFLOAD_XSTATS_ENABLE,
2886 NETDEV_OFFLOAD_XSTATS_DISABLE,
2887 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2888 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2889 NETDEV_XDP_FEAT_CHANGE,
2890};
2891const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2892
2893int register_netdevice_notifier(struct notifier_block *nb);
2894int unregister_netdevice_notifier(struct notifier_block *nb);
2895int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2896int unregister_netdevice_notifier_net(struct net *net,
2897 struct notifier_block *nb);
2898int register_netdevice_notifier_dev_net(struct net_device *dev,
2899 struct notifier_block *nb,
2900 struct netdev_net_notifier *nn);
2901int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2902 struct notifier_block *nb,
2903 struct netdev_net_notifier *nn);
2904
2905struct netdev_notifier_info {
2906 struct net_device *dev;
2907 struct netlink_ext_ack *extack;
2908};
2909
2910struct netdev_notifier_info_ext {
2911 struct netdev_notifier_info info; /* must be first */
2912 union {
2913 u32 mtu;
2914 } ext;
2915};
2916
2917struct netdev_notifier_change_info {
2918 struct netdev_notifier_info info; /* must be first */
2919 unsigned int flags_changed;
2920};
2921
2922struct netdev_notifier_changeupper_info {
2923 struct netdev_notifier_info info; /* must be first */
2924 struct net_device *upper_dev; /* new upper dev */
2925 bool master; /* is upper dev master */
2926 bool linking; /* is the notification for link or unlink */
2927 void *upper_info; /* upper dev info */
2928};
2929
2930struct netdev_notifier_changelowerstate_info {
2931 struct netdev_notifier_info info; /* must be first */
2932 void *lower_state_info; /* is lower dev state */
2933};
2934
2935struct netdev_notifier_pre_changeaddr_info {
2936 struct netdev_notifier_info info; /* must be first */
2937 const unsigned char *dev_addr;
2938};
2939
2940enum netdev_offload_xstats_type {
2941 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2942};
2943
2944struct netdev_notifier_offload_xstats_info {
2945 struct netdev_notifier_info info; /* must be first */
2946 enum netdev_offload_xstats_type type;
2947
2948 union {
2949 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2950 struct netdev_notifier_offload_xstats_rd *report_delta;
2951 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2952 struct netdev_notifier_offload_xstats_ru *report_used;
2953 };
2954};
2955
2956int netdev_offload_xstats_enable(struct net_device *dev,
2957 enum netdev_offload_xstats_type type,
2958 struct netlink_ext_ack *extack);
2959int netdev_offload_xstats_disable(struct net_device *dev,
2960 enum netdev_offload_xstats_type type);
2961bool netdev_offload_xstats_enabled(const struct net_device *dev,
2962 enum netdev_offload_xstats_type type);
2963int netdev_offload_xstats_get(struct net_device *dev,
2964 enum netdev_offload_xstats_type type,
2965 struct rtnl_hw_stats64 *stats, bool *used,
2966 struct netlink_ext_ack *extack);
2967void
2968netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2969 const struct rtnl_hw_stats64 *stats);
2970void
2971netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2972void netdev_offload_xstats_push_delta(struct net_device *dev,
2973 enum netdev_offload_xstats_type type,
2974 const struct rtnl_hw_stats64 *stats);
2975
2976static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2977 struct net_device *dev)
2978{
2979 info->dev = dev;
2980 info->extack = NULL;
2981}
2982
2983static inline struct net_device *
2984netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2985{
2986 return info->dev;
2987}
2988
2989static inline struct netlink_ext_ack *
2990netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2991{
2992 return info->extack;
2993}
2994
2995int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2996int call_netdevice_notifiers_info(unsigned long val,
2997 struct netdev_notifier_info *info);
2998
2999#define for_each_netdev(net, d) \
3000 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3001#define for_each_netdev_reverse(net, d) \
3002 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3003#define for_each_netdev_rcu(net, d) \
3004 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3005#define for_each_netdev_safe(net, d, n) \
3006 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3007#define for_each_netdev_continue(net, d) \
3008 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3009#define for_each_netdev_continue_reverse(net, d) \
3010 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3011 dev_list)
3012#define for_each_netdev_continue_rcu(net, d) \
3013 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3014#define for_each_netdev_in_bond_rcu(bond, slave) \
3015 for_each_netdev_rcu(&init_net, slave) \
3016 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3017#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3018
3019#define for_each_netdev_dump(net, d, ifindex) \
3020 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3021
3022static inline struct net_device *next_net_device(struct net_device *dev)
3023{
3024 struct list_head *lh;
3025 struct net *net;
3026
3027 net = dev_net(dev);
3028 lh = dev->dev_list.next;
3029 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3030}
3031
3032static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3033{
3034 struct list_head *lh;
3035 struct net *net;
3036
3037 net = dev_net(dev);
3038 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3039 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3040}
3041
3042static inline struct net_device *first_net_device(struct net *net)
3043{
3044 return list_empty(head: &net->dev_base_head) ? NULL :
3045 net_device_entry(net->dev_base_head.next);
3046}
3047
3048static inline struct net_device *first_net_device_rcu(struct net *net)
3049{
3050 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3051
3052 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3053}
3054
3055int netdev_boot_setup_check(struct net_device *dev);
3056struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3057 const char *hwaddr);
3058struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3059void dev_add_pack(struct packet_type *pt);
3060void dev_remove_pack(struct packet_type *pt);
3061void __dev_remove_pack(struct packet_type *pt);
3062void dev_add_offload(struct packet_offload *po);
3063void dev_remove_offload(struct packet_offload *po);
3064
3065int dev_get_iflink(const struct net_device *dev);
3066int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3067int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3068 struct net_device_path_stack *stack);
3069struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3070 unsigned short mask);
3071struct net_device *dev_get_by_name(struct net *net, const char *name);
3072struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3073struct net_device *__dev_get_by_name(struct net *net, const char *name);
3074bool netdev_name_in_use(struct net *net, const char *name);
3075int dev_alloc_name(struct net_device *dev, const char *name);
3076int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3077void dev_close(struct net_device *dev);
3078void dev_close_many(struct list_head *head, bool unlink);
3079void dev_disable_lro(struct net_device *dev);
3080int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3081u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3082 struct net_device *sb_dev);
3083u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3084 struct net_device *sb_dev);
3085
3086int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3087int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3088
3089static inline int dev_queue_xmit(struct sk_buff *skb)
3090{
3091 return __dev_queue_xmit(skb, NULL);
3092}
3093
3094static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3095 struct net_device *sb_dev)
3096{
3097 return __dev_queue_xmit(skb, sb_dev);
3098}
3099
3100static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3101{
3102 int ret;
3103
3104 ret = __dev_direct_xmit(skb, queue_id);
3105 if (!dev_xmit_complete(rc: ret))
3106 kfree_skb(skb);
3107 return ret;
3108}
3109
3110int register_netdevice(struct net_device *dev);
3111void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3112void unregister_netdevice_many(struct list_head *head);
3113static inline void unregister_netdevice(struct net_device *dev)
3114{
3115 unregister_netdevice_queue(dev, NULL);
3116}
3117
3118int netdev_refcnt_read(const struct net_device *dev);
3119void free_netdev(struct net_device *dev);
3120void netdev_freemem(struct net_device *dev);
3121void init_dummy_netdev(struct net_device *dev);
3122
3123struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3124 struct sk_buff *skb,
3125 bool all_slaves);
3126struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3127 struct sock *sk);
3128struct net_device *dev_get_by_index(struct net *net, int ifindex);
3129struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3130struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3131 netdevice_tracker *tracker, gfp_t gfp);
3132struct net_device *netdev_get_by_name(struct net *net, const char *name,
3133 netdevice_tracker *tracker, gfp_t gfp);
3134struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3135struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3136
3137static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3138 unsigned short type,
3139 const void *daddr, const void *saddr,
3140 unsigned int len)
3141{
3142 if (!dev->header_ops || !dev->header_ops->create)
3143 return 0;
3144
3145 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3146}
3147
3148static inline int dev_parse_header(const struct sk_buff *skb,
3149 unsigned char *haddr)
3150{
3151 const struct net_device *dev = skb->dev;
3152
3153 if (!dev->header_ops || !dev->header_ops->parse)
3154 return 0;
3155 return dev->header_ops->parse(skb, haddr);
3156}
3157
3158static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3159{
3160 const struct net_device *dev = skb->dev;
3161
3162 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3163 return 0;
3164 return dev->header_ops->parse_protocol(skb);
3165}
3166
3167/* ll_header must have at least hard_header_len allocated */
3168static inline bool dev_validate_header(const struct net_device *dev,
3169 char *ll_header, int len)
3170{
3171 if (likely(len >= dev->hard_header_len))
3172 return true;
3173 if (len < dev->min_header_len)
3174 return false;
3175
3176 if (capable(CAP_SYS_RAWIO)) {
3177 memset(ll_header + len, 0, dev->hard_header_len - len);
3178 return true;
3179 }
3180
3181 if (dev->header_ops && dev->header_ops->validate)
3182 return dev->header_ops->validate(ll_header, len);
3183
3184 return false;
3185}
3186
3187static inline bool dev_has_header(const struct net_device *dev)
3188{
3189 return dev->header_ops && dev->header_ops->create;
3190}
3191
3192/*
3193 * Incoming packets are placed on per-CPU queues
3194 */
3195struct softnet_data {
3196 struct list_head poll_list;
3197 struct sk_buff_head process_queue;
3198
3199 /* stats */
3200 unsigned int processed;
3201 unsigned int time_squeeze;
3202#ifdef CONFIG_RPS
3203 struct softnet_data *rps_ipi_list;
3204#endif
3205
3206 bool in_net_rx_action;
3207 bool in_napi_threaded_poll;
3208
3209#ifdef CONFIG_NET_FLOW_LIMIT
3210 struct sd_flow_limit __rcu *flow_limit;
3211#endif
3212 struct Qdisc *output_queue;
3213 struct Qdisc **output_queue_tailp;
3214 struct sk_buff *completion_queue;
3215#ifdef CONFIG_XFRM_OFFLOAD
3216 struct sk_buff_head xfrm_backlog;
3217#endif
3218 /* written and read only by owning cpu: */
3219 struct {
3220 u16 recursion;
3221 u8 more;
3222#ifdef CONFIG_NET_EGRESS
3223 u8 skip_txqueue;
3224#endif
3225 } xmit;
3226#ifdef CONFIG_RPS
3227 /* input_queue_head should be written by cpu owning this struct,
3228 * and only read by other cpus. Worth using a cache line.
3229 */
3230 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3231
3232 /* Elements below can be accessed between CPUs for RPS/RFS */
3233 call_single_data_t csd ____cacheline_aligned_in_smp;
3234 struct softnet_data *rps_ipi_next;
3235 unsigned int cpu;
3236 unsigned int input_queue_tail;
3237#endif
3238 unsigned int received_rps;
3239 unsigned int dropped;
3240 struct sk_buff_head input_pkt_queue;
3241 struct napi_struct backlog;
3242
3243 /* Another possibly contended cache line */
3244 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3245 int defer_count;
3246 int defer_ipi_scheduled;
3247 struct sk_buff *defer_list;
3248 call_single_data_t defer_csd;
3249};
3250
3251static inline void input_queue_head_incr(struct softnet_data *sd)
3252{
3253#ifdef CONFIG_RPS
3254 sd->input_queue_head++;
3255#endif
3256}
3257
3258static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3259 unsigned int *qtail)
3260{
3261#ifdef CONFIG_RPS
3262 *qtail = ++sd->input_queue_tail;
3263#endif
3264}
3265
3266DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3267
3268static inline int dev_recursion_level(void)
3269{
3270 return this_cpu_read(softnet_data.xmit.recursion);
3271}
3272
3273#define XMIT_RECURSION_LIMIT 8
3274static inline bool dev_xmit_recursion(void)
3275{
3276 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3277 XMIT_RECURSION_LIMIT);
3278}
3279
3280static inline void dev_xmit_recursion_inc(void)
3281{
3282 __this_cpu_inc(softnet_data.xmit.recursion);
3283}
3284
3285static inline void dev_xmit_recursion_dec(void)
3286{
3287 __this_cpu_dec(softnet_data.xmit.recursion);
3288}
3289
3290void __netif_schedule(struct Qdisc *q);
3291void netif_schedule_queue(struct netdev_queue *txq);
3292
3293static inline void netif_tx_schedule_all(struct net_device *dev)
3294{
3295 unsigned int i;
3296
3297 for (i = 0; i < dev->num_tx_queues; i++)
3298 netif_schedule_queue(txq: netdev_get_tx_queue(dev, index: i));
3299}
3300
3301static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3302{
3303 clear_bit(nr: __QUEUE_STATE_DRV_XOFF, addr: &dev_queue->state);
3304}
3305
3306/**
3307 * netif_start_queue - allow transmit
3308 * @dev: network device
3309 *
3310 * Allow upper layers to call the device hard_start_xmit routine.
3311 */
3312static inline void netif_start_queue(struct net_device *dev)
3313{
3314 netif_tx_start_queue(dev_queue: netdev_get_tx_queue(dev, index: 0));
3315}
3316
3317static inline void netif_tx_start_all_queues(struct net_device *dev)
3318{
3319 unsigned int i;
3320
3321 for (i = 0; i < dev->num_tx_queues; i++) {
3322 struct netdev_queue *txq = netdev_get_tx_queue(dev, index: i);
3323 netif_tx_start_queue(dev_queue: txq);
3324 }
3325}
3326
3327void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3328
3329/**
3330 * netif_wake_queue - restart transmit
3331 * @dev: network device
3332 *
3333 * Allow upper layers to call the device hard_start_xmit routine.
3334 * Used for flow control when transmit resources are available.
3335 */
3336static inline void netif_wake_queue(struct net_device *dev)
3337{
3338 netif_tx_wake_queue(dev_queue: netdev_get_tx_queue(dev, index: 0));
3339}
3340
3341static inline void netif_tx_wake_all_queues(struct net_device *dev)
3342{
3343 unsigned int i;
3344
3345 for (i = 0; i < dev->num_tx_queues; i++) {
3346 struct netdev_queue *txq = netdev_get_tx_queue(dev, index: i);
3347 netif_tx_wake_queue(dev_queue: txq);
3348 }
3349}
3350
3351static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3352{
3353 /* Must be an atomic op see netif_txq_try_stop() */
3354 set_bit(nr: __QUEUE_STATE_DRV_XOFF, addr: &dev_queue->state);
3355}
3356
3357/**
3358 * netif_stop_queue - stop transmitted packets
3359 * @dev: network device
3360 *
3361 * Stop upper layers calling the device hard_start_xmit routine.
3362 * Used for flow control when transmit resources are unavailable.
3363 */
3364static inline void netif_stop_queue(struct net_device *dev)
3365{
3366 netif_tx_stop_queue(dev_queue: netdev_get_tx_queue(dev, index: 0));
3367}
3368
3369void netif_tx_stop_all_queues(struct net_device *dev);
3370
3371static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3372{
3373 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3374}
3375
3376/**
3377 * netif_queue_stopped - test if transmit queue is flowblocked
3378 * @dev: network device
3379 *
3380 * Test if transmit queue on device is currently unable to send.
3381 */
3382static inline bool netif_queue_stopped(const struct net_device *dev)
3383{
3384 return netif_tx_queue_stopped(dev_queue: netdev_get_tx_queue(dev, index: 0));
3385}
3386
3387static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3388{
3389 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3390}
3391
3392static inline bool
3393netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3394{
3395 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3396}
3397
3398static inline bool
3399netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3400{
3401 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3402}
3403
3404/**
3405 * netdev_queue_set_dql_min_limit - set dql minimum limit
3406 * @dev_queue: pointer to transmit queue
3407 * @min_limit: dql minimum limit
3408 *
3409 * Forces xmit_more() to return true until the minimum threshold
3410 * defined by @min_limit is reached (or until the tx queue is
3411 * empty). Warning: to be use with care, misuse will impact the
3412 * latency.
3413 */
3414static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3415 unsigned int min_limit)
3416{
3417#ifdef CONFIG_BQL
3418 dev_queue->dql.min_limit = min_limit;
3419#endif
3420}
3421
3422static inline int netdev_queue_dql_avail(const struct netdev_queue *txq)
3423{
3424#ifdef CONFIG_BQL
3425 /* Non-BQL migrated drivers will return 0, too. */
3426 return dql_avail(dql: &txq->dql);
3427#else
3428 return 0;
3429#endif
3430}
3431
3432/**
3433 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3434 * @dev_queue: pointer to transmit queue
3435 *
3436 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3437 * to give appropriate hint to the CPU.
3438 */
3439static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3440{
3441#ifdef CONFIG_BQL
3442 prefetchw(x: &dev_queue->dql.num_queued);
3443#endif
3444}
3445
3446/**
3447 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3448 * @dev_queue: pointer to transmit queue
3449 *
3450 * BQL enabled drivers might use this helper in their TX completion path,
3451 * to give appropriate hint to the CPU.
3452 */
3453static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3454{
3455#ifdef CONFIG_BQL
3456 prefetchw(x: &dev_queue->dql.limit);
3457#endif
3458}
3459
3460/**
3461 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3462 * @dev_queue: network device queue
3463 * @bytes: number of bytes queued to the device queue
3464 *
3465 * Report the number of bytes queued for sending/completion to the network
3466 * device hardware queue. @bytes should be a good approximation and should
3467 * exactly match netdev_completed_queue() @bytes.
3468 * This is typically called once per packet, from ndo_start_xmit().
3469 */
3470static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3471 unsigned int bytes)
3472{
3473#ifdef CONFIG_BQL
3474 dql_queued(dql: &dev_queue->dql, count: bytes);
3475
3476 if (likely(dql_avail(&dev_queue->dql) >= 0))
3477 return;
3478
3479 set_bit(nr: __QUEUE_STATE_STACK_XOFF, addr: &dev_queue->state);
3480
3481 /*
3482 * The XOFF flag must be set before checking the dql_avail below,
3483 * because in netdev_tx_completed_queue we update the dql_completed
3484 * before checking the XOFF flag.
3485 */
3486 smp_mb();
3487
3488 /* check again in case another CPU has just made room avail */
3489 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3490 clear_bit(nr: __QUEUE_STATE_STACK_XOFF, addr: &dev_queue->state);
3491#endif
3492}
3493
3494/* Variant of netdev_tx_sent_queue() for drivers that are aware
3495 * that they should not test BQL status themselves.
3496 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3497 * skb of a batch.
3498 * Returns true if the doorbell must be used to kick the NIC.
3499 */
3500static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3501 unsigned int bytes,
3502 bool xmit_more)
3503{
3504 if (xmit_more) {
3505#ifdef CONFIG_BQL
3506 dql_queued(dql: &dev_queue->dql, count: bytes);
3507#endif
3508 return netif_tx_queue_stopped(dev_queue);
3509 }
3510 netdev_tx_sent_queue(dev_queue, bytes);
3511 return true;
3512}
3513
3514/**
3515 * netdev_sent_queue - report the number of bytes queued to hardware
3516 * @dev: network device
3517 * @bytes: number of bytes queued to the hardware device queue
3518 *
3519 * Report the number of bytes queued for sending/completion to the network
3520 * device hardware queue#0. @bytes should be a good approximation and should
3521 * exactly match netdev_completed_queue() @bytes.
3522 * This is typically called once per packet, from ndo_start_xmit().
3523 */
3524static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3525{
3526 netdev_tx_sent_queue(dev_queue: netdev_get_tx_queue(dev, index: 0), bytes);
3527}
3528
3529static inline bool __netdev_sent_queue(struct net_device *dev,
3530 unsigned int bytes,
3531 bool xmit_more)
3532{
3533 return __netdev_tx_sent_queue(dev_queue: netdev_get_tx_queue(dev, index: 0), bytes,
3534 xmit_more);
3535}
3536
3537/**
3538 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3539 * @dev_queue: network device queue
3540 * @pkts: number of packets (currently ignored)
3541 * @bytes: number of bytes dequeued from the device queue
3542 *
3543 * Must be called at most once per TX completion round (and not per
3544 * individual packet), so that BQL can adjust its limits appropriately.
3545 */
3546static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3547 unsigned int pkts, unsigned int bytes)
3548{
3549#ifdef CONFIG_BQL
3550 if (unlikely(!bytes))
3551 return;
3552
3553 dql_completed(dql: &dev_queue->dql, count: bytes);
3554
3555 /*
3556 * Without the memory barrier there is a small possiblity that
3557 * netdev_tx_sent_queue will miss the update and cause the queue to
3558 * be stopped forever
3559 */
3560 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3561
3562 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3563 return;
3564
3565 if (test_and_clear_bit(nr: __QUEUE_STATE_STACK_XOFF, addr: &dev_queue->state))
3566 netif_schedule_queue(txq: dev_queue);
3567#endif
3568}
3569
3570/**
3571 * netdev_completed_queue - report bytes and packets completed by device
3572 * @dev: network device
3573 * @pkts: actual number of packets sent over the medium
3574 * @bytes: actual number of bytes sent over the medium
3575 *
3576 * Report the number of bytes and packets transmitted by the network device
3577 * hardware queue over the physical medium, @bytes must exactly match the
3578 * @bytes amount passed to netdev_sent_queue()
3579 */
3580static inline void netdev_completed_queue(struct net_device *dev,
3581 unsigned int pkts, unsigned int bytes)
3582{
3583 netdev_tx_completed_queue(dev_queue: netdev_get_tx_queue(dev, index: 0), pkts, bytes);
3584}
3585
3586static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3587{
3588#ifdef CONFIG_BQL
3589 clear_bit(nr: __QUEUE_STATE_STACK_XOFF, addr: &q->state);
3590 dql_reset(dql: &q->dql);
3591#endif
3592}
3593
3594/**
3595 * netdev_reset_queue - reset the packets and bytes count of a network device
3596 * @dev_queue: network device
3597 *
3598 * Reset the bytes and packet count of a network device and clear the
3599 * software flow control OFF bit for this network device
3600 */
3601static inline void netdev_reset_queue(struct net_device *dev_queue)
3602{
3603 netdev_tx_reset_queue(q: netdev_get_tx_queue(dev: dev_queue, index: 0));
3604}
3605
3606/**
3607 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3608 * @dev: network device
3609 * @queue_index: given tx queue index
3610 *
3611 * Returns 0 if given tx queue index >= number of device tx queues,
3612 * otherwise returns the originally passed tx queue index.
3613 */
3614static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3615{
3616 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3617 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3618 dev->name, queue_index,
3619 dev->real_num_tx_queues);
3620 return 0;
3621 }
3622
3623 return queue_index;
3624}
3625
3626/**
3627 * netif_running - test if up
3628 * @dev: network device
3629 *
3630 * Test if the device has been brought up.
3631 */
3632static inline bool netif_running(const struct net_device *dev)
3633{
3634 return test_bit(__LINK_STATE_START, &dev->state);
3635}
3636
3637/*
3638 * Routines to manage the subqueues on a device. We only need start,
3639 * stop, and a check if it's stopped. All other device management is
3640 * done at the overall netdevice level.
3641 * Also test the device if we're multiqueue.
3642 */
3643
3644/**
3645 * netif_start_subqueue - allow sending packets on subqueue
3646 * @dev: network device
3647 * @queue_index: sub queue index
3648 *
3649 * Start individual transmit queue of a device with multiple transmit queues.
3650 */
3651static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3652{
3653 struct netdev_queue *txq = netdev_get_tx_queue(dev, index: queue_index);
3654
3655 netif_tx_start_queue(dev_queue: txq);
3656}
3657
3658/**
3659 * netif_stop_subqueue - stop sending packets on subqueue
3660 * @dev: network device
3661 * @queue_index: sub queue index
3662 *
3663 * Stop individual transmit queue of a device with multiple transmit queues.
3664 */
3665static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3666{
3667 struct netdev_queue *txq = netdev_get_tx_queue(dev, index: queue_index);
3668 netif_tx_stop_queue(dev_queue: txq);
3669}
3670
3671/**
3672 * __netif_subqueue_stopped - test status of subqueue
3673 * @dev: network device
3674 * @queue_index: sub queue index
3675 *
3676 * Check individual transmit queue of a device with multiple transmit queues.
3677 */
3678static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3679 u16 queue_index)
3680{
3681 struct netdev_queue *txq = netdev_get_tx_queue(dev, index: queue_index);
3682
3683 return netif_tx_queue_stopped(dev_queue: txq);
3684}
3685
3686/**
3687 * netif_subqueue_stopped - test status of subqueue
3688 * @dev: network device
3689 * @skb: sub queue buffer pointer
3690 *
3691 * Check individual transmit queue of a device with multiple transmit queues.
3692 */
3693static inline bool netif_subqueue_stopped(const struct net_device *dev,
3694 struct sk_buff *skb)
3695{
3696 return __netif_subqueue_stopped(dev, queue_index: skb_get_queue_mapping(skb));
3697}
3698
3699/**
3700 * netif_wake_subqueue - allow sending packets on subqueue
3701 * @dev: network device
3702 * @queue_index: sub queue index
3703 *
3704 * Resume individual transmit queue of a device with multiple transmit queues.
3705 */
3706static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3707{
3708 struct netdev_queue *txq = netdev_get_tx_queue(dev, index: queue_index);
3709
3710 netif_tx_wake_queue(dev_queue: txq);
3711}
3712
3713#ifdef CONFIG_XPS
3714int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3715 u16 index);
3716int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3717 u16 index, enum xps_map_type type);
3718
3719/**
3720 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3721 * @j: CPU/Rx queue index
3722 * @mask: bitmask of all cpus/rx queues
3723 * @nr_bits: number of bits in the bitmask
3724 *
3725 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3726 */
3727static inline bool netif_attr_test_mask(unsigned long j,
3728 const unsigned long *mask,
3729 unsigned int nr_bits)
3730{
3731 cpu_max_bits_warn(cpu: j, bits: nr_bits);
3732 return test_bit(j, mask);
3733}
3734
3735/**
3736 * netif_attr_test_online - Test for online CPU/Rx queue
3737 * @j: CPU/Rx queue index
3738 * @online_mask: bitmask for CPUs/Rx queues that are online
3739 * @nr_bits: number of bits in the bitmask
3740 *
3741 * Returns true if a CPU/Rx queue is online.
3742 */
3743static inline bool netif_attr_test_online(unsigned long j,
3744 const unsigned long *online_mask,
3745 unsigned int nr_bits)
3746{
3747 cpu_max_bits_warn(cpu: j, bits: nr_bits);
3748
3749 if (online_mask)
3750 return test_bit(j, online_mask);
3751
3752 return (j < nr_bits);
3753}
3754
3755/**
3756 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3757 * @n: CPU/Rx queue index
3758 * @srcp: the cpumask/Rx queue mask pointer
3759 * @nr_bits: number of bits in the bitmask
3760 *
3761 * Returns >= nr_bits if no further CPUs/Rx queues set.
3762 */
3763static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3764 unsigned int nr_bits)
3765{
3766 /* -1 is a legal arg here. */
3767 if (n != -1)
3768 cpu_max_bits_warn(cpu: n, bits: nr_bits);
3769
3770 if (srcp)
3771 return find_next_bit(addr: srcp, size: nr_bits, offset: n + 1);
3772
3773 return n + 1;
3774}
3775
3776/**
3777 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3778 * @n: CPU/Rx queue index
3779 * @src1p: the first CPUs/Rx queues mask pointer
3780 * @src2p: the second CPUs/Rx queues mask pointer
3781 * @nr_bits: number of bits in the bitmask
3782 *
3783 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3784 */
3785static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3786 const unsigned long *src2p,
3787 unsigned int nr_bits)
3788{
3789 /* -1 is a legal arg here. */
3790 if (n != -1)
3791 cpu_max_bits_warn(cpu: n, bits: nr_bits);
3792
3793 if (src1p && src2p)
3794 return find_next_and_bit(addr1: src1p, addr2: src2p, size: nr_bits, offset: n + 1);
3795 else if (src1p)
3796 return find_next_bit(addr: src1p, size: nr_bits, offset: n + 1);
3797 else if (src2p)
3798 return find_next_bit(addr: src2p, size: nr_bits, offset: n + 1);
3799
3800 return n + 1;
3801}
3802#else
3803static inline int netif_set_xps_queue(struct net_device *dev,
3804 const struct cpumask *mask,
3805 u16 index)
3806{
3807 return 0;
3808}
3809
3810static inline int __netif_set_xps_queue(struct net_device *dev,
3811 const unsigned long *mask,
3812 u16 index, enum xps_map_type type)
3813{
3814 return 0;
3815}
3816#endif
3817
3818/**
3819 * netif_is_multiqueue - test if device has multiple transmit queues
3820 * @dev: network device
3821 *
3822 * Check if device has multiple transmit queues
3823 */
3824static inline bool netif_is_multiqueue(const struct net_device *dev)
3825{
3826 return dev->num_tx_queues > 1;
3827}
3828
3829int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3830
3831#ifdef CONFIG_SYSFS
3832int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3833#else
3834static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3835 unsigned int rxqs)
3836{
3837 dev->real_num_rx_queues = rxqs;
3838 return 0;
3839}
3840#endif
3841int netif_set_real_num_queues(struct net_device *dev,
3842 unsigned int txq, unsigned int rxq);
3843
3844int netif_get_num_default_rss_queues(void);
3845
3846void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3847void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3848
3849/*
3850 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3851 * interrupt context or with hardware interrupts being disabled.
3852 * (in_hardirq() || irqs_disabled())
3853 *
3854 * We provide four helpers that can be used in following contexts :
3855 *
3856 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3857 * replacing kfree_skb(skb)
3858 *
3859 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3860 * Typically used in place of consume_skb(skb) in TX completion path
3861 *
3862 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3863 * replacing kfree_skb(skb)
3864 *
3865 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3866 * and consumed a packet. Used in place of consume_skb(skb)
3867 */
3868static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3869{
3870 dev_kfree_skb_irq_reason(skb, reason: SKB_DROP_REASON_NOT_SPECIFIED);
3871}
3872
3873static inline void dev_consume_skb_irq(struct sk_buff *skb)
3874{
3875 dev_kfree_skb_irq_reason(skb, reason: SKB_CONSUMED);
3876}
3877
3878static inline void dev_kfree_skb_any(struct sk_buff *skb)
3879{
3880 dev_kfree_skb_any_reason(skb, reason: SKB_DROP_REASON_NOT_SPECIFIED);
3881}
3882
3883static inline void dev_consume_skb_any(struct sk_buff *skb)
3884{
3885 dev_kfree_skb_any_reason(skb, reason: SKB_CONSUMED);
3886}
3887
3888u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3889 struct bpf_prog *xdp_prog);
3890void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3891int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb);
3892int netif_rx(struct sk_buff *skb);
3893int __netif_rx(struct sk_buff *skb);
3894
3895int netif_receive_skb(struct sk_buff *skb);
3896int netif_receive_skb_core(struct sk_buff *skb);
3897void netif_receive_skb_list_internal(struct list_head *head);
3898void netif_receive_skb_list(struct list_head *head);
3899gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3900void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3901struct sk_buff *napi_get_frags(struct napi_struct *napi);
3902void napi_get_frags_check(struct napi_struct *napi);
3903gro_result_t napi_gro_frags(struct napi_struct *napi);
3904
3905static inline void napi_free_frags(struct napi_struct *napi)
3906{
3907 kfree_skb(skb: napi->skb);
3908 napi->skb = NULL;
3909}
3910
3911bool netdev_is_rx_handler_busy(struct net_device *dev);
3912int netdev_rx_handler_register(struct net_device *dev,
3913 rx_handler_func_t *rx_handler,
3914 void *rx_handler_data);
3915void netdev_rx_handler_unregister(struct net_device *dev);
3916
3917bool dev_valid_name(const char *name);
3918static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3919{
3920 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3921}
3922int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3923int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3924int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3925 void __user *data, bool *need_copyout);
3926int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3927int generic_hwtstamp_get_lower(struct net_device *dev,
3928 struct kernel_hwtstamp_config *kernel_cfg);
3929int generic_hwtstamp_set_lower(struct net_device *dev,
3930 struct kernel_hwtstamp_config *kernel_cfg,
3931 struct netlink_ext_ack *extack);
3932int dev_set_hwtstamp_phylib(struct net_device *dev,
3933 struct kernel_hwtstamp_config *cfg,
3934 struct netlink_ext_ack *extack);
3935int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3936unsigned int dev_get_flags(const struct net_device *);
3937int __dev_change_flags(struct net_device *dev, unsigned int flags,
3938 struct netlink_ext_ack *extack);
3939int dev_change_flags(struct net_device *dev, unsigned int flags,
3940 struct netlink_ext_ack *extack);
3941int dev_set_alias(struct net_device *, const char *, size_t);
3942int dev_get_alias(const struct net_device *, char *, size_t);
3943int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3944 const char *pat, int new_ifindex);
3945static inline
3946int dev_change_net_namespace(struct net_device *dev, struct net *net,
3947 const char *pat)
3948{
3949 return __dev_change_net_namespace(dev, net, pat, new_ifindex: 0);
3950}
3951int __dev_set_mtu(struct net_device *, int);
3952int dev_set_mtu(struct net_device *, int);
3953int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3954 struct netlink_ext_ack *extack);
3955int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3956 struct netlink_ext_ack *extack);
3957int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3958 struct netlink_ext_ack *extack);
3959int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3960int dev_get_port_parent_id(struct net_device *dev,
3961 struct netdev_phys_item_id *ppid, bool recurse);
3962bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3963
3964struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3965struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3966 struct netdev_queue *txq, int *ret);
3967
3968int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3969u8 dev_xdp_prog_count(struct net_device *dev);
3970u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3971
3972int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3973int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3974int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3975bool is_skb_forwardable(const struct net_device *dev,
3976 const struct sk_buff *skb);
3977
3978static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3979 const struct sk_buff *skb,
3980 const bool check_mtu)
3981{
3982 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3983 unsigned int len;
3984
3985 if (!(dev->flags & IFF_UP))
3986 return false;
3987
3988 if (!check_mtu)
3989 return true;
3990
3991 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3992 if (skb->len <= len)
3993 return true;
3994
3995 /* if TSO is enabled, we don't care about the length as the packet
3996 * could be forwarded without being segmented before
3997 */
3998 if (skb_is_gso(skb))
3999 return true;
4000
4001 return false;
4002}
4003
4004void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4005
4006#define DEV_CORE_STATS_INC(FIELD) \
4007static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4008{ \
4009 netdev_core_stats_inc(dev, \
4010 offsetof(struct net_device_core_stats, FIELD)); \
4011}
4012DEV_CORE_STATS_INC(rx_dropped)
4013DEV_CORE_STATS_INC(tx_dropped)
4014DEV_CORE_STATS_INC(rx_nohandler)
4015DEV_CORE_STATS_INC(rx_otherhost_dropped)
4016#undef DEV_CORE_STATS_INC
4017
4018static __always_inline int ____dev_forward_skb(struct net_device *dev,
4019 struct sk_buff *skb,
4020 const bool check_mtu)
4021{
4022 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4023 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4024 dev_core_stats_rx_dropped_inc(dev);
4025 kfree_skb(skb);
4026 return NET_RX_DROP;
4027 }
4028
4029 skb_scrub_packet(skb, xnet: !net_eq(net1: dev_net(dev), net2: dev_net(dev: skb->dev)));
4030 skb->priority = 0;
4031 return 0;
4032}
4033
4034bool dev_nit_active(struct net_device *dev);
4035void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4036
4037static inline void __dev_put(struct net_device *dev)
4038{
4039 if (dev) {
4040#ifdef CONFIG_PCPU_DEV_REFCNT
4041 this_cpu_dec(*dev->pcpu_refcnt);
4042#else
4043 refcount_dec(&dev->dev_refcnt);
4044#endif
4045 }
4046}
4047
4048static inline void __dev_hold(struct net_device *dev)
4049{
4050 if (dev) {
4051#ifdef CONFIG_PCPU_DEV_REFCNT
4052 this_cpu_inc(*dev->pcpu_refcnt);
4053#else
4054 refcount_inc(&dev->dev_refcnt);
4055#endif
4056 }
4057}
4058
4059static inline void __netdev_tracker_alloc(struct net_device *dev,
4060 netdevice_tracker *tracker,
4061 gfp_t gfp)
4062{
4063#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4064 ref_tracker_alloc(dir: &dev->refcnt_tracker, trackerp: tracker, gfp);
4065#endif
4066}
4067
4068/* netdev_tracker_alloc() can upgrade a prior untracked reference
4069 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4070 */
4071static inline void netdev_tracker_alloc(struct net_device *dev,
4072 netdevice_tracker *tracker, gfp_t gfp)
4073{
4074#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4075 refcount_dec(r: &dev->refcnt_tracker.no_tracker);
4076 __netdev_tracker_alloc(dev, tracker, gfp);
4077#endif
4078}
4079
4080static inline void netdev_tracker_free(struct net_device *dev,
4081 netdevice_tracker *tracker)
4082{
4083#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4084 ref_tracker_free(dir: &dev->refcnt_tracker, trackerp: tracker);
4085#endif
4086}
4087
4088static inline void netdev_hold(struct net_device *dev,
4089 netdevice_tracker *tracker, gfp_t gfp)
4090{
4091 if (dev) {
4092 __dev_hold(dev);
4093 __netdev_tracker_alloc(dev, tracker, gfp);
4094 }
4095}
4096
4097static inline void netdev_put(struct net_device *dev,
4098 netdevice_tracker *tracker)
4099{
4100 if (dev) {
4101 netdev_tracker_free(dev, tracker);
4102 __dev_put(dev);
4103 }
4104}
4105
4106/**
4107 * dev_hold - get reference to device
4108 * @dev: network device
4109 *
4110 * Hold reference to device to keep it from being freed.
4111 * Try using netdev_hold() instead.
4112 */
4113static inline void dev_hold(struct net_device *dev)
4114{
4115 netdev_hold(dev, NULL, GFP_ATOMIC);
4116}
4117
4118/**
4119 * dev_put - release reference to device
4120 * @dev: network device
4121 *
4122 * Release reference to device to allow it to be freed.
4123 * Try using netdev_put() instead.
4124 */
4125static inline void dev_put(struct net_device *dev)
4126{
4127 netdev_put(dev, NULL);
4128}
4129
4130static inline void netdev_ref_replace(struct net_device *odev,
4131 struct net_device *ndev,
4132 netdevice_tracker *tracker,
4133 gfp_t gfp)
4134{
4135 if (odev)
4136 netdev_tracker_free(dev: odev, tracker);
4137
4138 __dev_hold(dev: ndev);
4139 __dev_put(dev: odev);
4140
4141 if (ndev)
4142 __netdev_tracker_alloc(dev: ndev, tracker, gfp);
4143}
4144
4145/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4146 * and _off may be called from IRQ context, but it is caller
4147 * who is responsible for serialization of these calls.
4148 *
4149 * The name carrier is inappropriate, these functions should really be
4150 * called netif_lowerlayer_*() because they represent the state of any
4151 * kind of lower layer not just hardware media.
4152 */
4153void linkwatch_fire_event(struct net_device *dev);
4154
4155/**
4156 * linkwatch_sync_dev - sync linkwatch for the given device
4157 * @dev: network device to sync linkwatch for
4158 *
4159 * Sync linkwatch for the given device, removing it from the
4160 * pending work list (if queued).
4161 */
4162void linkwatch_sync_dev(struct net_device *dev);
4163
4164/**
4165 * netif_carrier_ok - test if carrier present
4166 * @dev: network device
4167 *
4168 * Check if carrier is present on device
4169 */
4170static inline bool netif_carrier_ok(const struct net_device *dev)
4171{
4172 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4173}
4174
4175unsigned long dev_trans_start(struct net_device *dev);
4176
4177void __netdev_watchdog_up(struct net_device *dev);
4178
4179void netif_carrier_on(struct net_device *dev);
4180void netif_carrier_off(struct net_device *dev);
4181void netif_carrier_event(struct net_device *dev);
4182
4183/**
4184 * netif_dormant_on - mark device as dormant.
4185 * @dev: network device
4186 *
4187 * Mark device as dormant (as per RFC2863).
4188 *
4189 * The dormant state indicates that the relevant interface is not
4190 * actually in a condition to pass packets (i.e., it is not 'up') but is
4191 * in a "pending" state, waiting for some external event. For "on-
4192 * demand" interfaces, this new state identifies the situation where the
4193 * interface is waiting for events to place it in the up state.
4194 */
4195static inline void netif_dormant_on(struct net_device *dev)
4196{
4197 if (!test_and_set_bit(nr: __LINK_STATE_DORMANT, addr: &dev->state))
4198 linkwatch_fire_event(dev);
4199}
4200
4201/**
4202 * netif_dormant_off - set device as not dormant.
4203 * @dev: network device
4204 *
4205 * Device is not in dormant state.
4206 */
4207static inline void netif_dormant_off(struct net_device *dev)
4208{
4209 if (test_and_clear_bit(nr: __LINK_STATE_DORMANT, addr: &dev->state))
4210 linkwatch_fire_event(dev);
4211}
4212
4213/**
4214 * netif_dormant - test if device is dormant
4215 * @dev: network device
4216 *
4217 * Check if device is dormant.
4218 */
4219static inline bool netif_dormant(const struct net_device *dev)
4220{
4221 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4222}
4223
4224
4225/**
4226 * netif_testing_on - mark device as under test.
4227 * @dev: network device
4228 *
4229 * Mark device as under test (as per RFC2863).
4230 *
4231 * The testing state indicates that some test(s) must be performed on
4232 * the interface. After completion, of the test, the interface state
4233 * will change to up, dormant, or down, as appropriate.
4234 */
4235static inline void netif_testing_on(struct net_device *dev)
4236{
4237 if (!test_and_set_bit(nr: __LINK_STATE_TESTING, addr: &dev->state))
4238 linkwatch_fire_event(dev);
4239}
4240
4241/**
4242 * netif_testing_off - set device as not under test.
4243 * @dev: network device
4244 *
4245 * Device is not in testing state.
4246 */
4247static inline void netif_testing_off(struct net_device *dev)
4248{
4249 if (test_and_clear_bit(nr: __LINK_STATE_TESTING, addr: &dev->state))
4250 linkwatch_fire_event(dev);
4251}
4252
4253/**
4254 * netif_testing - test if device is under test
4255 * @dev: network device
4256 *
4257 * Check if device is under test
4258 */
4259static inline bool netif_testing(const struct net_device *dev)
4260{
4261 return test_bit(__LINK_STATE_TESTING, &dev->state);
4262}
4263
4264
4265/**
4266 * netif_oper_up - test if device is operational
4267 * @dev: network device
4268 *
4269 * Check if carrier is operational
4270 */
4271static inline bool netif_oper_up(const struct net_device *dev)
4272{
4273 unsigned int operstate = READ_ONCE(dev->operstate);
4274
4275 return operstate == IF_OPER_UP ||
4276 operstate == IF_OPER_UNKNOWN /* backward compat */;
4277}
4278
4279/**
4280 * netif_device_present - is device available or removed
4281 * @dev: network device
4282 *
4283 * Check if device has not been removed from system.
4284 */
4285static inline bool netif_device_present(const struct net_device *dev)
4286{
4287 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4288}
4289
4290void netif_device_detach(struct net_device *dev);
4291
4292void netif_device_attach(struct net_device *dev);
4293
4294/*
4295 * Network interface message level settings
4296 */
4297
4298enum {
4299 NETIF_MSG_DRV_BIT,
4300 NETIF_MSG_PROBE_BIT,
4301 NETIF_MSG_LINK_BIT,
4302 NETIF_MSG_TIMER_BIT,
4303 NETIF_MSG_IFDOWN_BIT,
4304 NETIF_MSG_IFUP_BIT,
4305 NETIF_MSG_RX_ERR_BIT,
4306 NETIF_MSG_TX_ERR_BIT,
4307 NETIF_MSG_TX_QUEUED_BIT,
4308 NETIF_MSG_INTR_BIT,
4309 NETIF_MSG_TX_DONE_BIT,
4310 NETIF_MSG_RX_STATUS_BIT,
4311 NETIF_MSG_PKTDATA_BIT,
4312 NETIF_MSG_HW_BIT,
4313 NETIF_MSG_WOL_BIT,
4314
4315 /* When you add a new bit above, update netif_msg_class_names array
4316 * in net/ethtool/common.c
4317 */
4318 NETIF_MSG_CLASS_COUNT,
4319};
4320/* Both ethtool_ops interface and internal driver implementation use u32 */
4321static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4322
4323#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4324#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4325
4326#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4327#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4328#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4329#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4330#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4331#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4332#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4333#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4334#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4335#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4336#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4337#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4338#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4339#define NETIF_MSG_HW __NETIF_MSG(HW)
4340#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4341
4342#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4343#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4344#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4345#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4346#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4347#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4348#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4349#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4350#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4351#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4352#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4353#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4354#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4355#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4356#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4357
4358static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4359{
4360 /* use default */
4361 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4362 return default_msg_enable_bits;
4363 if (debug_value == 0) /* no output */
4364 return 0;
4365 /* set low N bits */
4366 return (1U << debug_value) - 1;
4367}
4368
4369static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4370{
4371 spin_lock(lock: &txq->_xmit_lock);
4372 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4373 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4374}
4375
4376static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4377{
4378 __acquire(&txq->_xmit_lock);
4379 return true;
4380}
4381
4382static inline void __netif_tx_release(struct netdev_queue *txq)
4383{
4384 __release(&txq->_xmit_lock);
4385}
4386
4387static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4388{
4389 spin_lock_bh(lock: &txq->_xmit_lock);
4390 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4391 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4392}
4393
4394static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4395{
4396 bool ok = spin_trylock(lock: &txq->_xmit_lock);
4397
4398 if (likely(ok)) {
4399 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4400 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4401 }
4402 return ok;
4403}
4404
4405static inline void __netif_tx_unlock(struct netdev_queue *txq)
4406{
4407 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4408 WRITE_ONCE(txq->xmit_lock_owner, -1);
4409 spin_unlock(lock: &txq->_xmit_lock);
4410}
4411
4412static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4413{
4414 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4415 WRITE_ONCE(txq->xmit_lock_owner, -1);
4416 spin_unlock_bh(lock: &txq->_xmit_lock);
4417}
4418
4419/*
4420 * txq->trans_start can be read locklessly from dev_watchdog()
4421 */
4422static inline void txq_trans_update(struct netdev_queue *txq)
4423{
4424 if (txq->xmit_lock_owner != -1)
4425 WRITE_ONCE(txq->trans_start, jiffies);
4426}
4427
4428static inline void txq_trans_cond_update(struct netdev_queue *txq)
4429{
4430 unsigned long now = jiffies;
4431
4432 if (READ_ONCE(txq->trans_start) != now)
4433 WRITE_ONCE(txq->trans_start, now);
4434}
4435
4436/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4437static inline void netif_trans_update(struct net_device *dev)
4438{
4439 struct netdev_queue *txq = netdev_get_tx_queue(dev, index: 0);
4440
4441 txq_trans_cond_update(txq);
4442}
4443
4444/**
4445 * netif_tx_lock - grab network device transmit lock
4446 * @dev: network device
4447 *
4448 * Get network device transmit lock
4449 */
4450void netif_tx_lock(struct net_device *dev);
4451
4452static inline void netif_tx_lock_bh(struct net_device *dev)
4453{
4454 local_bh_disable();
4455 netif_tx_lock(dev);
4456}
4457
4458void netif_tx_unlock(struct net_device *dev);
4459
4460static inline void netif_tx_unlock_bh(struct net_device *dev)
4461{
4462 netif_tx_unlock(dev);
4463 local_bh_enable();
4464}
4465
4466#define HARD_TX_LOCK(dev, txq, cpu) { \
4467 if ((dev->features & NETIF_F_LLTX) == 0) { \
4468 __netif_tx_lock(txq, cpu); \
4469 } else { \
4470 __netif_tx_acquire(txq); \
4471 } \
4472}
4473
4474#define HARD_TX_TRYLOCK(dev, txq) \
4475 (((dev->features & NETIF_F_LLTX) == 0) ? \
4476 __netif_tx_trylock(txq) : \
4477 __netif_tx_acquire(txq))
4478
4479#define HARD_TX_UNLOCK(dev, txq) { \
4480 if ((dev->features & NETIF_F_LLTX) == 0) { \
4481 __netif_tx_unlock(txq); \
4482 } else { \
4483 __netif_tx_release(txq); \
4484 } \
4485}
4486
4487static inline void netif_tx_disable(struct net_device *dev)
4488{
4489 unsigned int i;
4490 int cpu;
4491
4492 local_bh_disable();
4493 cpu = smp_processor_id();
4494 spin_lock(lock: &dev->tx_global_lock);
4495 for (i = 0; i < dev->num_tx_queues; i++) {
4496 struct netdev_queue *txq = netdev_get_tx_queue(dev, index: i);
4497
4498 __netif_tx_lock(txq, cpu);
4499 netif_tx_stop_queue(dev_queue: txq);
4500 __netif_tx_unlock(txq);
4501 }
4502 spin_unlock(lock: &dev->tx_global_lock);
4503 local_bh_enable();
4504}
4505
4506static inline void netif_addr_lock(struct net_device *dev)
4507{
4508 unsigned char nest_level = 0;
4509
4510#ifdef CONFIG_LOCKDEP
4511 nest_level = dev->nested_level;
4512#endif
4513 spin_lock_nested(&dev->addr_list_lock, nest_level);
4514}
4515
4516static inline void netif_addr_lock_bh(struct net_device *dev)
4517{
4518 unsigned char nest_level = 0;
4519
4520#ifdef CONFIG_LOCKDEP
4521 nest_level = dev->nested_level;
4522#endif
4523 local_bh_disable();
4524 spin_lock_nested(&dev->addr_list_lock, nest_level);
4525}
4526
4527static inline void netif_addr_unlock(struct net_device *dev)
4528{
4529 spin_unlock(lock: &dev->addr_list_lock);
4530}
4531
4532static inline void netif_addr_unlock_bh(struct net_device *dev)
4533{
4534 spin_unlock_bh(lock: &dev->addr_list_lock);
4535}
4536
4537/*
4538 * dev_addrs walker. Should be used only for read access. Call with
4539 * rcu_read_lock held.
4540 */
4541#define for_each_dev_addr(dev, ha) \
4542 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4543
4544/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4545
4546void ether_setup(struct net_device *dev);
4547
4548/* Support for loadable net-drivers */
4549struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4550 unsigned char name_assign_type,
4551 void (*setup)(struct net_device *),
4552 unsigned int txqs, unsigned int rxqs);
4553#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4554 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4555
4556#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4557 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4558 count)
4559
4560int register_netdev(struct net_device *dev);
4561void unregister_netdev(struct net_device *dev);
4562
4563int devm_register_netdev(struct device *dev, struct net_device *ndev);
4564
4565/* General hardware address lists handling functions */
4566int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4567 struct netdev_hw_addr_list *from_list, int addr_len);
4568void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4569 struct netdev_hw_addr_list *from_list, int addr_len);
4570int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4571 struct net_device *dev,
4572 int (*sync)(struct net_device *, const unsigned char *),
4573 int (*unsync)(struct net_device *,
4574 const unsigned char *));
4575int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4576 struct net_device *dev,
4577 int (*sync)(struct net_device *,
4578 const unsigned char *, int),
4579 int (*unsync)(struct net_device *,
4580 const unsigned char *, int));
4581void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4582 struct net_device *dev,
4583 int (*unsync)(struct net_device *,
4584 const unsigned char *, int));
4585void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4586 struct net_device *dev,
4587 int (*unsync)(struct net_device *,
4588 const unsigned char *));
4589void __hw_addr_init(struct netdev_hw_addr_list *list);
4590
4591/* Functions used for device addresses handling */
4592void dev_addr_mod(struct net_device *dev, unsigned int offset,
4593 const void *addr, size_t len);
4594
4595static inline void
4596__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4597{
4598 dev_addr_mod(dev, offset: 0, addr, len);
4599}
4600
4601static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4602{
4603 __dev_addr_set(dev, addr, len: dev->addr_len);
4604}
4605
4606int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4607 unsigned char addr_type);
4608int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4609 unsigned char addr_type);
4610
4611/* Functions used for unicast addresses handling */
4612int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4613int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4614int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4615int dev_uc_sync(struct net_device *to, struct net_device *from);
4616int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4617void dev_uc_unsync(struct net_device *to, struct net_device *from);
4618void dev_uc_flush(struct net_device *dev);
4619void dev_uc_init(struct net_device *dev);
4620
4621/**
4622 * __dev_uc_sync - Synchonize device's unicast list
4623 * @dev: device to sync
4624 * @sync: function to call if address should be added
4625 * @unsync: function to call if address should be removed
4626 *
4627 * Add newly added addresses to the interface, and release
4628 * addresses that have been deleted.
4629 */
4630static inline int __dev_uc_sync(struct net_device *dev,
4631 int (*sync)(struct net_device *,
4632 const unsigned char *),
4633 int (*unsync)(struct net_device *,
4634 const unsigned char *))
4635{
4636 return __hw_addr_sync_dev(list: &dev->uc, dev, sync, unsync);
4637}
4638
4639/**
4640 * __dev_uc_unsync - Remove synchronized addresses from device
4641 * @dev: device to sync
4642 * @unsync: function to call if address should be removed
4643 *
4644 * Remove all addresses that were added to the device by dev_uc_sync().
4645 */
4646static inline void __dev_uc_unsync(struct net_device *dev,
4647 int (*unsync)(struct net_device *,
4648 const unsigned char *))
4649{
4650 __hw_addr_unsync_dev(list: &dev->uc, dev, unsync);
4651}
4652
4653/* Functions used for multicast addresses handling */
4654int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4655int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4656int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4657int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4658int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4659int dev_mc_sync(struct net_device *to, struct net_device *from);
4660int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4661void dev_mc_unsync(struct net_device *to, struct net_device *from);
4662void dev_mc_flush(struct net_device *dev);
4663void dev_mc_init(struct net_device *dev);
4664
4665/**
4666 * __dev_mc_sync - Synchonize device's multicast list
4667 * @dev: device to sync
4668 * @sync: function to call if address should be added
4669 * @unsync: function to call if address should be removed
4670 *
4671 * Add newly added addresses to the interface, and release
4672 * addresses that have been deleted.
4673 */
4674static inline int __dev_mc_sync(struct net_device *dev,
4675 int (*sync)(struct net_device *,
4676 const unsigned char *),
4677 int (*unsync)(struct net_device *,
4678 const unsigned char *))
4679{
4680 return __hw_addr_sync_dev(list: &dev->mc, dev, sync, unsync);
4681}
4682
4683/**
4684 * __dev_mc_unsync - Remove synchronized addresses from device
4685 * @dev: device to sync
4686 * @unsync: function to call if address should be removed
4687 *
4688 * Remove all addresses that were added to the device by dev_mc_sync().
4689 */
4690static inline void __dev_mc_unsync(struct net_device *dev,
4691 int (*unsync)(struct net_device *,
4692 const unsigned char *))
4693{
4694 __hw_addr_unsync_dev(list: &dev->mc, dev, unsync);
4695}
4696
4697/* Functions used for secondary unicast and multicast support */
4698void dev_set_rx_mode(struct net_device *dev);
4699int dev_set_promiscuity(struct net_device *dev, int inc);
4700int dev_set_allmulti(struct net_device *dev, int inc);
4701void netdev_state_change(struct net_device *dev);
4702void __netdev_notify_peers(struct net_device *dev);
4703void netdev_notify_peers(struct net_device *dev);
4704void netdev_features_change(struct net_device *dev);
4705/* Load a device via the kmod */
4706void dev_load(struct net *net, const char *name);
4707struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4708 struct rtnl_link_stats64 *storage);
4709void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4710 const struct net_device_stats *netdev_stats);
4711void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4712 const struct pcpu_sw_netstats __percpu *netstats);
4713void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4714
4715enum {
4716 NESTED_SYNC_IMM_BIT,
4717 NESTED_SYNC_TODO_BIT,
4718};
4719
4720#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4721#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4722
4723#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4724#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4725
4726struct netdev_nested_priv {
4727 unsigned char flags;
4728 void *data;
4729};
4730
4731bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4732struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4733 struct list_head **iter);
4734
4735/* iterate through upper list, must be called under RCU read lock */
4736#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4737 for (iter = &(dev)->adj_list.upper, \
4738 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4739 updev; \
4740 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4741
4742int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4743 int (*fn)(struct net_device *upper_dev,
4744 struct netdev_nested_priv *priv),
4745 struct netdev_nested_priv *priv);
4746
4747bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4748 struct net_device *upper_dev);
4749
4750bool netdev_has_any_upper_dev(struct net_device *dev);
4751
4752void *netdev_lower_get_next_private(struct net_device *dev,
4753 struct list_head **iter);
4754void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4755 struct list_head **iter);
4756
4757#define netdev_for_each_lower_private(dev, priv, iter) \
4758 for (iter = (dev)->adj_list.lower.next, \
4759 priv = netdev_lower_get_next_private(dev, &(iter)); \
4760 priv; \
4761 priv = netdev_lower_get_next_private(dev, &(iter)))
4762
4763#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4764 for (iter = &(dev)->adj_list.lower, \
4765 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4766 priv; \
4767 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4768
4769void *netdev_lower_get_next(struct net_device *dev,
4770 struct list_head **iter);
4771
4772#define netdev_for_each_lower_dev(dev, ldev, iter) \
4773 for (iter = (dev)->adj_list.lower.next, \
4774 ldev = netdev_lower_get_next(dev, &(iter)); \
4775 ldev; \
4776 ldev = netdev_lower_get_next(dev, &(iter)))
4777
4778struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4779 struct list_head **iter);
4780int netdev_walk_all_lower_dev(struct net_device *dev,
4781 int (*fn)(struct net_device *lower_dev,
4782 struct netdev_nested_priv *priv),
4783 struct netdev_nested_priv *priv);
4784int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4785 int (*fn)(struct net_device *lower_dev,
4786 struct netdev_nested_priv *priv),
4787 struct netdev_nested_priv *priv);
4788
4789void *netdev_adjacent_get_private(struct list_head *adj_list);
4790void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4791struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4792struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4793int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4794 struct netlink_ext_ack *extack);
4795int netdev_master_upper_dev_link(struct net_device *dev,
4796 struct net_device *upper_dev,
4797 void *upper_priv, void *upper_info,
4798 struct netlink_ext_ack *extack);
4799void netdev_upper_dev_unlink(struct net_device *dev,
4800 struct net_device *upper_dev);
4801int netdev_adjacent_change_prepare(struct net_device *old_dev,
4802 struct net_device *new_dev,
4803 struct net_device *dev,
4804 struct netlink_ext_ack *extack);
4805void netdev_adjacent_change_commit(struct net_device *old_dev,
4806 struct net_device *new_dev,
4807 struct net_device *dev);
4808void netdev_adjacent_change_abort(struct net_device *old_dev,
4809 struct net_device *new_dev,
4810 struct net_device *dev);
4811void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4812void *netdev_lower_dev_get_private(struct net_device *dev,
4813 struct net_device *lower_dev);
4814void netdev_lower_state_changed(struct net_device *lower_dev,
4815 void *lower_state_info);
4816
4817/* RSS keys are 40 or 52 bytes long */
4818#define NETDEV_RSS_KEY_LEN 52
4819extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4820void netdev_rss_key_fill(void *buffer, size_t len);
4821
4822int skb_checksum_help(struct sk_buff *skb);
4823int skb_crc32c_csum_help(struct sk_buff *skb);
4824int skb_csum_hwoffload_help(struct sk_buff *skb,
4825 const netdev_features_t features);
4826
4827struct netdev_bonding_info {
4828 ifslave slave;
4829 ifbond master;
4830};
4831
4832struct netdev_notifier_bonding_info {
4833 struct netdev_notifier_info info; /* must be first */
4834 struct netdev_bonding_info bonding_info;
4835};
4836
4837void netdev_bonding_info_change(struct net_device *dev,
4838 struct netdev_bonding_info *bonding_info);
4839
4840#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4841void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4842#else
4843static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4844 const void *data)
4845{
4846}
4847#endif
4848
4849__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4850
4851static inline bool can_checksum_protocol(netdev_features_t features,
4852 __be16 protocol)
4853{
4854 if (protocol == htons(ETH_P_FCOE))
4855 return !!(features & NETIF_F_FCOE_CRC);
4856
4857 /* Assume this is an IP checksum (not SCTP CRC) */
4858
4859 if (features & NETIF_F_HW_CSUM) {
4860 /* Can checksum everything */
4861 return true;
4862 }
4863
4864 switch (protocol) {
4865 case htons(ETH_P_IP):
4866 return !!(features & NETIF_F_IP_CSUM);
4867 case htons(ETH_P_IPV6):
4868 return !!(features & NETIF_F_IPV6_CSUM);
4869 default:
4870 return false;
4871 }
4872}
4873
4874#ifdef CONFIG_BUG
4875void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4876#else
4877static inline void netdev_rx_csum_fault(struct net_device *dev,
4878 struct sk_buff *skb)
4879{
4880}
4881#endif
4882/* rx skb timestamps */
4883void net_enable_timestamp(void);
4884void net_disable_timestamp(void);
4885
4886static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4887 const struct skb_shared_hwtstamps *hwtstamps,
4888 bool cycles)
4889{
4890 const struct net_device_ops *ops = dev->netdev_ops;
4891
4892 if (ops->ndo_get_tstamp)
4893 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4894
4895 return hwtstamps->hwtstamp;
4896}
4897
4898static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4899 struct sk_buff *skb, struct net_device *dev,
4900 bool more)
4901{
4902 __this_cpu_write(softnet_data.xmit.more, more);
4903 return ops->ndo_start_xmit(skb, dev);
4904}
4905
4906static inline bool netdev_xmit_more(void)
4907{
4908 return __this_cpu_read(softnet_data.xmit.more);
4909}
4910
4911static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4912 struct netdev_queue *txq, bool more)
4913{
4914 const struct net_device_ops *ops = dev->netdev_ops;
4915 netdev_tx_t rc;
4916
4917 rc = __netdev_start_xmit(ops, skb, dev, more);
4918 if (rc == NETDEV_TX_OK)
4919 txq_trans_update(txq);
4920
4921 return rc;
4922}
4923
4924int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4925 const void *ns);
4926void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4927 const void *ns);
4928
4929extern const struct kobj_ns_type_operations net_ns_type_operations;
4930
4931const char *netdev_drivername(const struct net_device *dev);
4932
4933static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4934 netdev_features_t f2)
4935{
4936 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4937 if (f1 & NETIF_F_HW_CSUM)
4938 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4939 else
4940 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4941 }
4942
4943 return f1 & f2;
4944}
4945
4946static inline netdev_features_t netdev_get_wanted_features(
4947 struct net_device *dev)
4948{
4949 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4950}
4951netdev_features_t netdev_increment_features(netdev_features_t all,
4952 netdev_features_t one, netdev_features_t mask);
4953
4954/* Allow TSO being used on stacked device :
4955 * Performing the GSO segmentation before last device
4956 * is a performance improvement.
4957 */
4958static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4959 netdev_features_t mask)
4960{
4961 return netdev_increment_features(all: features, NETIF_F_ALL_TSO, mask);
4962}
4963
4964int __netdev_update_features(struct net_device *dev);
4965void netdev_update_features(struct net_device *dev);
4966void netdev_change_features(struct net_device *dev);
4967
4968void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4969 struct net_device *dev);
4970
4971netdev_features_t passthru_features_check(struct sk_buff *skb,
4972 struct net_device *dev,
4973 netdev_features_t features);
4974netdev_features_t netif_skb_features(struct sk_buff *skb);
4975void skb_warn_bad_offload(const struct sk_buff *skb);
4976
4977static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4978{
4979 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4980
4981 /* check flags correspondence */
4982 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4983 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4984 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4985 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4986 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4987 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4988 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4989 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4990 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4991 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4992 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4993 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4994 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4995 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4996 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4997 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4998 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4999 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5000 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5001
5002 return (features & feature) == feature;
5003}
5004
5005static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5006{
5007 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5008 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5009}
5010
5011static inline bool netif_needs_gso(struct sk_buff *skb,
5012 netdev_features_t features)
5013{
5014 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5015 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5016 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5017}
5018
5019void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5020void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5021void netif_inherit_tso_max(struct net_device *to,
5022 const struct net_device *from);
5023
5024static inline bool netif_is_macsec(const struct net_device *dev)
5025{
5026 return dev->priv_flags & IFF_MACSEC;
5027}
5028
5029static inline bool netif_is_macvlan(const struct net_device *dev)
5030{
5031 return dev->priv_flags & IFF_MACVLAN;
5032}
5033
5034static inline bool netif_is_macvlan_port(const struct net_device *dev)
5035{
5036 return dev->priv_flags & IFF_MACVLAN_PORT;
5037}
5038
5039static inline bool netif_is_bond_master(const struct net_device *dev)
5040{
5041 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5042}
5043
5044static inline bool netif_is_bond_slave(const struct net_device *dev)
5045{
5046 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5047}
5048
5049static inline bool netif_supports_nofcs(struct net_device *dev)
5050{
5051 return dev->priv_flags & IFF_SUPP_NOFCS;
5052}
5053
5054static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5055{
5056 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5057}
5058
5059static inline bool netif_is_l3_master(const struct net_device *dev)
5060{
5061 return dev->priv_flags & IFF_L3MDEV_MASTER;
5062}
5063
5064static inline bool netif_is_l3_slave(const struct net_device *dev)
5065{
5066 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5067}
5068
5069static inline int dev_sdif(const struct net_device *dev)
5070{
5071#ifdef CONFIG_NET_L3_MASTER_DEV
5072 if (netif_is_l3_slave(dev))
5073 return dev->ifindex;
5074#endif
5075 return 0;
5076}
5077
5078static inline bool netif_is_bridge_master(const struct net_device *dev)
5079{
5080 return dev->priv_flags & IFF_EBRIDGE;
5081}
5082
5083static inline bool netif_is_bridge_port(const struct net_device *dev)
5084{
5085 return dev->priv_flags & IFF_BRIDGE_PORT;
5086}
5087
5088static inline bool netif_is_ovs_master(const struct net_device *dev)
5089{
5090 return dev->priv_flags & IFF_OPENVSWITCH;
5091}
5092
5093static inline bool netif_is_ovs_port(const struct net_device *dev)
5094{
5095 return dev->priv_flags & IFF_OVS_DATAPATH;
5096}
5097
5098static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5099{
5100 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5101}
5102
5103static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5104{
5105 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5106}
5107
5108static inline bool netif_is_team_master(const struct net_device *dev)
5109{
5110 return dev->priv_flags & IFF_TEAM;
5111}
5112
5113static inline bool netif_is_team_port(const struct net_device *dev)
5114{
5115 return dev->priv_flags & IFF_TEAM_PORT;
5116}
5117
5118static inline bool netif_is_lag_master(const struct net_device *dev)
5119{
5120 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5121}
5122
5123static inline bool netif_is_lag_port(const struct net_device *dev)
5124{
5125 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5126}
5127
5128static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5129{
5130 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5131}
5132
5133static inline bool netif_is_failover(const struct net_device *dev)
5134{
5135 return dev->priv_flags & IFF_FAILOVER;
5136}
5137
5138static inline bool netif_is_failover_slave(const struct net_device *dev)
5139{
5140 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5141}
5142
5143/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5144static inline void netif_keep_dst(struct net_device *dev)
5145{
5146 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5147}
5148
5149/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5150static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5151{
5152 /* TODO: reserve and use an additional IFF bit, if we get more users */
5153 return netif_is_macsec(dev);
5154}
5155
5156extern struct pernet_operations __net_initdata loopback_net_ops;
5157
5158/* Logging, debugging and troubleshooting/diagnostic helpers. */
5159
5160/* netdev_printk helpers, similar to dev_printk */
5161
5162static inline const char *netdev_name(const struct net_device *dev)
5163{
5164 if (!dev->name[0] || strchr(dev->name, '%'))
5165 return "(unnamed net_device)";
5166 return dev->name;
5167}
5168
5169static inline const char *netdev_reg_state(const struct net_device *dev)
5170{
5171 u8 reg_state = READ_ONCE(dev->reg_state);
5172
5173 switch (reg_state) {
5174 case NETREG_UNINITIALIZED: return " (uninitialized)";
5175 case NETREG_REGISTERED: return "";
5176 case NETREG_UNREGISTERING: return " (unregistering)";
5177 case NETREG_UNREGISTERED: return " (unregistered)";
5178 case NETREG_RELEASED: return " (released)";
5179 case NETREG_DUMMY: return " (dummy)";
5180 }
5181
5182 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state);
5183 return " (unknown)";
5184}
5185
5186#define MODULE_ALIAS_NETDEV(device) \
5187 MODULE_ALIAS("netdev-" device)
5188
5189/*
5190 * netdev_WARN() acts like dev_printk(), but with the key difference
5191 * of using a WARN/WARN_ON to get the message out, including the
5192 * file/line information and a backtrace.
5193 */
5194#define netdev_WARN(dev, format, args...) \
5195 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5196 netdev_reg_state(dev), ##args)
5197
5198#define netdev_WARN_ONCE(dev, format, args...) \
5199 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5200 netdev_reg_state(dev), ##args)
5201
5202/*
5203 * The list of packet types we will receive (as opposed to discard)
5204 * and the routines to invoke.
5205 *
5206 * Why 16. Because with 16 the only overlap we get on a hash of the
5207 * low nibble of the protocol value is RARP/SNAP/X.25.
5208 *
5209 * 0800 IP
5210 * 0001 802.3
5211 * 0002 AX.25
5212 * 0004 802.2
5213 * 8035 RARP
5214 * 0005 SNAP
5215 * 0805 X.25
5216 * 0806 ARP
5217 * 8137 IPX
5218 * 0009 Localtalk
5219 * 86DD IPv6
5220 */
5221#define PTYPE_HASH_SIZE (16)
5222#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5223
5224extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5225
5226extern struct net_device *blackhole_netdev;
5227
5228/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5229#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5230#define DEV_STATS_ADD(DEV, FIELD, VAL) \
5231 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5232#define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5233
5234#endif /* _LINUX_NETDEVICE_H */
5235

source code of linux/include/linux/netdevice.h