Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
24#include <linux/timer.h>
25#include <linux/bug.h>
26#include <linux/delay.h>
27#include <linux/atomic.h>
28#include <linux/prefetch.h>
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31#include <asm/local.h>
32
33#include <linux/percpu.h>
34#include <linux/rculist.h>
35#include <linux/workqueue.h>
36#include <linux/dynamic_queue_limits.h>
37
38#include <net/net_namespace.h>
39#ifdef CONFIG_DCB
40#include <net/dcbnl.h>
41#endif
42#include <net/netprio_cgroup.h>
43#include <net/xdp.h>
44
45#include <linux/netdev_features.h>
46#include <linux/neighbour.h>
47#include <uapi/linux/netdevice.h>
48#include <uapi/linux/if_bonding.h>
49#include <uapi/linux/pkt_cls.h>
50#include <uapi/linux/netdev.h>
51#include <linux/hashtable.h>
52#include <linux/rbtree.h>
53#include <net/net_trackers.h>
54#include <net/net_debug.h>
55
56struct netpoll_info;
57struct device;
58struct ethtool_ops;
59struct phy_device;
60struct dsa_port;
61struct ip_tunnel_parm;
62struct macsec_context;
63struct macsec_ops;
64struct netdev_name_node;
65struct sd_flow_limit;
66struct sfp_bus;
67/* 802.11 specific */
68struct wireless_dev;
69/* 802.15.4 specific */
70struct wpan_dev;
71struct mpls_dev;
72/* UDP Tunnel offloads */
73struct udp_tunnel_info;
74struct udp_tunnel_nic_info;
75struct udp_tunnel_nic;
76struct bpf_prog;
77struct xdp_buff;
78struct xdp_md;
79
80void synchronize_net(void);
81void netdev_set_default_ethtool_ops(struct net_device *dev,
82 const struct ethtool_ops *ops);
83void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
84
85/* Backlog congestion levels */
86#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
87#define NET_RX_DROP 1 /* packet dropped */
88
89#define MAX_NEST_DEV 8
90
91/*
92 * Transmit return codes: transmit return codes originate from three different
93 * namespaces:
94 *
95 * - qdisc return codes
96 * - driver transmit return codes
97 * - errno values
98 *
99 * Drivers are allowed to return any one of those in their hard_start_xmit()
100 * function. Real network devices commonly used with qdiscs should only return
101 * the driver transmit return codes though - when qdiscs are used, the actual
102 * transmission happens asynchronously, so the value is not propagated to
103 * higher layers. Virtual network devices transmit synchronously; in this case
104 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
105 * others are propagated to higher layers.
106 */
107
108/* qdisc ->enqueue() return codes. */
109#define NET_XMIT_SUCCESS 0x00
110#define NET_XMIT_DROP 0x01 /* skb dropped */
111#define NET_XMIT_CN 0x02 /* congestion notification */
112#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
113
114/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
115 * indicates that the device will soon be dropping packets, or already drops
116 * some packets of the same priority; prompting us to send less aggressively. */
117#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
118#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
119
120/* Driver transmit return codes */
121#define NETDEV_TX_MASK 0xf0
122
123enum netdev_tx {
124 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
125 NETDEV_TX_OK = 0x00, /* driver took care of packet */
126 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
127};
128typedef enum netdev_tx netdev_tx_t;
129
130/*
131 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
132 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
133 */
134static inline bool dev_xmit_complete(int rc)
135{
136 /*
137 * Positive cases with an skb consumed by a driver:
138 * - successful transmission (rc == NETDEV_TX_OK)
139 * - error while transmitting (rc < 0)
140 * - error while queueing to a different device (rc & NET_XMIT_MASK)
141 */
142 if (likely(rc < NET_XMIT_MASK))
143 return true;
144
145 return false;
146}
147
148/*
149 * Compute the worst-case header length according to the protocols
150 * used.
151 */
152
153#if defined(CONFIG_HYPERV_NET)
154# define LL_MAX_HEADER 128
155#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
156# if defined(CONFIG_MAC80211_MESH)
157# define LL_MAX_HEADER 128
158# else
159# define LL_MAX_HEADER 96
160# endif
161#else
162# define LL_MAX_HEADER 32
163#endif
164
165#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
166 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
167#define MAX_HEADER LL_MAX_HEADER
168#else
169#define MAX_HEADER (LL_MAX_HEADER + 48)
170#endif
171
172/*
173 * Old network device statistics. Fields are native words
174 * (unsigned long) so they can be read and written atomically.
175 */
176
177#define NET_DEV_STAT(FIELD) \
178 union { \
179 unsigned long FIELD; \
180 atomic_long_t __##FIELD; \
181 }
182
183struct net_device_stats {
184 NET_DEV_STAT(rx_packets);
185 NET_DEV_STAT(tx_packets);
186 NET_DEV_STAT(rx_bytes);
187 NET_DEV_STAT(tx_bytes);
188 NET_DEV_STAT(rx_errors);
189 NET_DEV_STAT(tx_errors);
190 NET_DEV_STAT(rx_dropped);
191 NET_DEV_STAT(tx_dropped);
192 NET_DEV_STAT(multicast);
193 NET_DEV_STAT(collisions);
194 NET_DEV_STAT(rx_length_errors);
195 NET_DEV_STAT(rx_over_errors);
196 NET_DEV_STAT(rx_crc_errors);
197 NET_DEV_STAT(rx_frame_errors);
198 NET_DEV_STAT(rx_fifo_errors);
199 NET_DEV_STAT(rx_missed_errors);
200 NET_DEV_STAT(tx_aborted_errors);
201 NET_DEV_STAT(tx_carrier_errors);
202 NET_DEV_STAT(tx_fifo_errors);
203 NET_DEV_STAT(tx_heartbeat_errors);
204 NET_DEV_STAT(tx_window_errors);
205 NET_DEV_STAT(rx_compressed);
206 NET_DEV_STAT(tx_compressed);
207};
208#undef NET_DEV_STAT
209
210/* per-cpu stats, allocated on demand.
211 * Try to fit them in a single cache line, for dev_get_stats() sake.
212 */
213struct net_device_core_stats {
214 unsigned long rx_dropped;
215 unsigned long tx_dropped;
216 unsigned long rx_nohandler;
217 unsigned long rx_otherhost_dropped;
218} __aligned(4 * sizeof(unsigned long));
219
220#include <linux/cache.h>
221#include <linux/skbuff.h>
222
223#ifdef CONFIG_RPS
224#include <linux/static_key.h>
225extern struct static_key_false rps_needed;
226extern struct static_key_false rfs_needed;
227#endif
228
229struct neighbour;
230struct neigh_parms;
231struct sk_buff;
232
233struct netdev_hw_addr {
234 struct list_head list;
235 struct rb_node node;
236 unsigned char addr[MAX_ADDR_LEN];
237 unsigned char type;
238#define NETDEV_HW_ADDR_T_LAN 1
239#define NETDEV_HW_ADDR_T_SAN 2
240#define NETDEV_HW_ADDR_T_UNICAST 3
241#define NETDEV_HW_ADDR_T_MULTICAST 4
242 bool global_use;
243 int sync_cnt;
244 int refcount;
245 int synced;
246 struct rcu_head rcu_head;
247};
248
249struct netdev_hw_addr_list {
250 struct list_head list;
251 int count;
252
253 /* Auxiliary tree for faster lookup on addition and deletion */
254 struct rb_root tree;
255};
256
257#define netdev_hw_addr_list_count(l) ((l)->count)
258#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
259#define netdev_hw_addr_list_for_each(ha, l) \
260 list_for_each_entry(ha, &(l)->list, list)
261
262#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
263#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
264#define netdev_for_each_uc_addr(ha, dev) \
265 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
266#define netdev_for_each_synced_uc_addr(_ha, _dev) \
267 netdev_for_each_uc_addr((_ha), (_dev)) \
268 if ((_ha)->sync_cnt)
269
270#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
271#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
272#define netdev_for_each_mc_addr(ha, dev) \
273 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
274#define netdev_for_each_synced_mc_addr(_ha, _dev) \
275 netdev_for_each_mc_addr((_ha), (_dev)) \
276 if ((_ha)->sync_cnt)
277
278struct hh_cache {
279 unsigned int hh_len;
280 seqlock_t hh_lock;
281
282 /* cached hardware header; allow for machine alignment needs. */
283#define HH_DATA_MOD 16
284#define HH_DATA_OFF(__len) \
285 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
286#define HH_DATA_ALIGN(__len) \
287 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
288 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
289};
290
291/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
292 * Alternative is:
293 * dev->hard_header_len ? (dev->hard_header_len +
294 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
295 *
296 * We could use other alignment values, but we must maintain the
297 * relationship HH alignment <= LL alignment.
298 */
299#define LL_RESERVED_SPACE(dev) \
300 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
301 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
302#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
303 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
304 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
305
306struct header_ops {
307 int (*create) (struct sk_buff *skb, struct net_device *dev,
308 unsigned short type, const void *daddr,
309 const void *saddr, unsigned int len);
310 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
311 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
312 void (*cache_update)(struct hh_cache *hh,
313 const struct net_device *dev,
314 const unsigned char *haddr);
315 bool (*validate)(const char *ll_header, unsigned int len);
316 __be16 (*parse_protocol)(const struct sk_buff *skb);
317};
318
319/* These flag bits are private to the generic network queueing
320 * layer; they may not be explicitly referenced by any other
321 * code.
322 */
323
324enum netdev_state_t {
325 __LINK_STATE_START,
326 __LINK_STATE_PRESENT,
327 __LINK_STATE_NOCARRIER,
328 __LINK_STATE_LINKWATCH_PENDING,
329 __LINK_STATE_DORMANT,
330 __LINK_STATE_TESTING,
331};
332
333struct gro_list {
334 struct list_head list;
335 int count;
336};
337
338/*
339 * size of gro hash buckets, must less than bit number of
340 * napi_struct::gro_bitmask
341 */
342#define GRO_HASH_BUCKETS 8
343
344/*
345 * Structure for NAPI scheduling similar to tasklet but with weighting
346 */
347struct napi_struct {
348 /* The poll_list must only be managed by the entity which
349 * changes the state of the NAPI_STATE_SCHED bit. This means
350 * whoever atomically sets that bit can add this napi_struct
351 * to the per-CPU poll_list, and whoever clears that bit
352 * can remove from the list right before clearing the bit.
353 */
354 struct list_head poll_list;
355
356 unsigned long state;
357 int weight;
358 int defer_hard_irqs_count;
359 unsigned long gro_bitmask;
360 int (*poll)(struct napi_struct *, int);
361#ifdef CONFIG_NETPOLL
362 int poll_owner;
363#endif
364 struct net_device *dev;
365 struct gro_list gro_hash[GRO_HASH_BUCKETS];
366 struct sk_buff *skb;
367 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
368 int rx_count; /* length of rx_list */
369 struct hrtimer timer;
370 struct list_head dev_list;
371 struct hlist_node napi_hash_node;
372 unsigned int napi_id;
373 struct task_struct *thread;
374};
375
376enum {
377 NAPI_STATE_SCHED, /* Poll is scheduled */
378 NAPI_STATE_MISSED, /* reschedule a napi */
379 NAPI_STATE_DISABLE, /* Disable pending */
380 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
381 NAPI_STATE_LISTED, /* NAPI added to system lists */
382 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
383 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
384 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
385 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
386 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
387};
388
389enum {
390 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
391 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
392 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
393 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
394 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
395 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
396 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
397 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
398 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
399 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
400};
401
402enum gro_result {
403 GRO_MERGED,
404 GRO_MERGED_FREE,
405 GRO_HELD,
406 GRO_NORMAL,
407 GRO_CONSUMED,
408};
409typedef enum gro_result gro_result_t;
410
411/*
412 * enum rx_handler_result - Possible return values for rx_handlers.
413 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
414 * further.
415 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
416 * case skb->dev was changed by rx_handler.
417 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
418 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
419 *
420 * rx_handlers are functions called from inside __netif_receive_skb(), to do
421 * special processing of the skb, prior to delivery to protocol handlers.
422 *
423 * Currently, a net_device can only have a single rx_handler registered. Trying
424 * to register a second rx_handler will return -EBUSY.
425 *
426 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
427 * To unregister a rx_handler on a net_device, use
428 * netdev_rx_handler_unregister().
429 *
430 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
431 * do with the skb.
432 *
433 * If the rx_handler consumed the skb in some way, it should return
434 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
435 * the skb to be delivered in some other way.
436 *
437 * If the rx_handler changed skb->dev, to divert the skb to another
438 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
439 * new device will be called if it exists.
440 *
441 * If the rx_handler decides the skb should be ignored, it should return
442 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
443 * are registered on exact device (ptype->dev == skb->dev).
444 *
445 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
446 * delivered, it should return RX_HANDLER_PASS.
447 *
448 * A device without a registered rx_handler will behave as if rx_handler
449 * returned RX_HANDLER_PASS.
450 */
451
452enum rx_handler_result {
453 RX_HANDLER_CONSUMED,
454 RX_HANDLER_ANOTHER,
455 RX_HANDLER_EXACT,
456 RX_HANDLER_PASS,
457};
458typedef enum rx_handler_result rx_handler_result_t;
459typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
460
461void __napi_schedule(struct napi_struct *n);
462void __napi_schedule_irqoff(struct napi_struct *n);
463
464static inline bool napi_disable_pending(struct napi_struct *n)
465{
466 return test_bit(NAPI_STATE_DISABLE, &n->state);
467}
468
469static inline bool napi_prefer_busy_poll(struct napi_struct *n)
470{
471 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
472}
473
474bool napi_schedule_prep(struct napi_struct *n);
475
476/**
477 * napi_schedule - schedule NAPI poll
478 * @n: NAPI context
479 *
480 * Schedule NAPI poll routine to be called if it is not already
481 * running.
482 */
483static inline void napi_schedule(struct napi_struct *n)
484{
485 if (napi_schedule_prep(n))
486 __napi_schedule(n);
487}
488
489/**
490 * napi_schedule_irqoff - schedule NAPI poll
491 * @n: NAPI context
492 *
493 * Variant of napi_schedule(), assuming hard irqs are masked.
494 */
495static inline void napi_schedule_irqoff(struct napi_struct *n)
496{
497 if (napi_schedule_prep(n))
498 __napi_schedule_irqoff(n);
499}
500
501/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
502static inline bool napi_reschedule(struct napi_struct *napi)
503{
504 if (napi_schedule_prep(napi)) {
505 __napi_schedule(napi);
506 return true;
507 }
508 return false;
509}
510
511bool napi_complete_done(struct napi_struct *n, int work_done);
512/**
513 * napi_complete - NAPI processing complete
514 * @n: NAPI context
515 *
516 * Mark NAPI processing as complete.
517 * Consider using napi_complete_done() instead.
518 * Return false if device should avoid rearming interrupts.
519 */
520static inline bool napi_complete(struct napi_struct *n)
521{
522 return napi_complete_done(n, 0);
523}
524
525int dev_set_threaded(struct net_device *dev, bool threaded);
526
527/**
528 * napi_disable - prevent NAPI from scheduling
529 * @n: NAPI context
530 *
531 * Stop NAPI from being scheduled on this context.
532 * Waits till any outstanding processing completes.
533 */
534void napi_disable(struct napi_struct *n);
535
536void napi_enable(struct napi_struct *n);
537
538/**
539 * napi_synchronize - wait until NAPI is not running
540 * @n: NAPI context
541 *
542 * Wait until NAPI is done being scheduled on this context.
543 * Waits till any outstanding processing completes but
544 * does not disable future activations.
545 */
546static inline void napi_synchronize(const struct napi_struct *n)
547{
548 if (IS_ENABLED(CONFIG_SMP))
549 while (test_bit(NAPI_STATE_SCHED, &n->state))
550 msleep(1);
551 else
552 barrier();
553}
554
555/**
556 * napi_if_scheduled_mark_missed - if napi is running, set the
557 * NAPIF_STATE_MISSED
558 * @n: NAPI context
559 *
560 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
561 * NAPI is scheduled.
562 **/
563static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
564{
565 unsigned long val, new;
566
567 val = READ_ONCE(n->state);
568 do {
569 if (val & NAPIF_STATE_DISABLE)
570 return true;
571
572 if (!(val & NAPIF_STATE_SCHED))
573 return false;
574
575 new = val | NAPIF_STATE_MISSED;
576 } while (!try_cmpxchg(&n->state, &val, new));
577
578 return true;
579}
580
581enum netdev_queue_state_t {
582 __QUEUE_STATE_DRV_XOFF,
583 __QUEUE_STATE_STACK_XOFF,
584 __QUEUE_STATE_FROZEN,
585};
586
587#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
588#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
589#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
590
591#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
592#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
593 QUEUE_STATE_FROZEN)
594#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
595 QUEUE_STATE_FROZEN)
596
597/*
598 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
599 * netif_tx_* functions below are used to manipulate this flag. The
600 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
601 * queue independently. The netif_xmit_*stopped functions below are called
602 * to check if the queue has been stopped by the driver or stack (either
603 * of the XOFF bits are set in the state). Drivers should not need to call
604 * netif_xmit*stopped functions, they should only be using netif_tx_*.
605 */
606
607struct netdev_queue {
608/*
609 * read-mostly part
610 */
611 struct net_device *dev;
612 netdevice_tracker dev_tracker;
613
614 struct Qdisc __rcu *qdisc;
615 struct Qdisc *qdisc_sleeping;
616#ifdef CONFIG_SYSFS
617 struct kobject kobj;
618#endif
619#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
620 int numa_node;
621#endif
622 unsigned long tx_maxrate;
623 /*
624 * Number of TX timeouts for this queue
625 * (/sys/class/net/DEV/Q/trans_timeout)
626 */
627 atomic_long_t trans_timeout;
628
629 /* Subordinate device that the queue has been assigned to */
630 struct net_device *sb_dev;
631#ifdef CONFIG_XDP_SOCKETS
632 struct xsk_buff_pool *pool;
633#endif
634/*
635 * write-mostly part
636 */
637 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
638 int xmit_lock_owner;
639 /*
640 * Time (in jiffies) of last Tx
641 */
642 unsigned long trans_start;
643
644 unsigned long state;
645
646#ifdef CONFIG_BQL
647 struct dql dql;
648#endif
649} ____cacheline_aligned_in_smp;
650
651extern int sysctl_fb_tunnels_only_for_init_net;
652extern int sysctl_devconf_inherit_init_net;
653
654/*
655 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
656 * == 1 : For initns only
657 * == 2 : For none.
658 */
659static inline bool net_has_fallback_tunnels(const struct net *net)
660{
661#if IS_ENABLED(CONFIG_SYSCTL)
662 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
663
664 return !fb_tunnels_only_for_init_net ||
665 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
666#else
667 return true;
668#endif
669}
670
671static inline int net_inherit_devconf(void)
672{
673#if IS_ENABLED(CONFIG_SYSCTL)
674 return READ_ONCE(sysctl_devconf_inherit_init_net);
675#else
676 return 0;
677#endif
678}
679
680static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
681{
682#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
683 return q->numa_node;
684#else
685 return NUMA_NO_NODE;
686#endif
687}
688
689static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
690{
691#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
692 q->numa_node = node;
693#endif
694}
695
696#ifdef CONFIG_RPS
697/*
698 * This structure holds an RPS map which can be of variable length. The
699 * map is an array of CPUs.
700 */
701struct rps_map {
702 unsigned int len;
703 struct rcu_head rcu;
704 u16 cpus[];
705};
706#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
707
708/*
709 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
710 * tail pointer for that CPU's input queue at the time of last enqueue, and
711 * a hardware filter index.
712 */
713struct rps_dev_flow {
714 u16 cpu;
715 u16 filter;
716 unsigned int last_qtail;
717};
718#define RPS_NO_FILTER 0xffff
719
720/*
721 * The rps_dev_flow_table structure contains a table of flow mappings.
722 */
723struct rps_dev_flow_table {
724 unsigned int mask;
725 struct rcu_head rcu;
726 struct rps_dev_flow flows[];
727};
728#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
729 ((_num) * sizeof(struct rps_dev_flow)))
730
731/*
732 * The rps_sock_flow_table contains mappings of flows to the last CPU
733 * on which they were processed by the application (set in recvmsg).
734 * Each entry is a 32bit value. Upper part is the high-order bits
735 * of flow hash, lower part is CPU number.
736 * rps_cpu_mask is used to partition the space, depending on number of
737 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
738 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
739 * meaning we use 32-6=26 bits for the hash.
740 */
741struct rps_sock_flow_table {
742 u32 mask;
743
744 u32 ents[] ____cacheline_aligned_in_smp;
745};
746#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
747
748#define RPS_NO_CPU 0xffff
749
750extern u32 rps_cpu_mask;
751extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
752
753static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
754 u32 hash)
755{
756 if (table && hash) {
757 unsigned int index = hash & table->mask;
758 u32 val = hash & ~rps_cpu_mask;
759
760 /* We only give a hint, preemption can change CPU under us */
761 val |= raw_smp_processor_id();
762
763 if (table->ents[index] != val)
764 table->ents[index] = val;
765 }
766}
767
768#ifdef CONFIG_RFS_ACCEL
769bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
770 u16 filter_id);
771#endif
772#endif /* CONFIG_RPS */
773
774/* This structure contains an instance of an RX queue. */
775struct netdev_rx_queue {
776 struct xdp_rxq_info xdp_rxq;
777#ifdef CONFIG_RPS
778 struct rps_map __rcu *rps_map;
779 struct rps_dev_flow_table __rcu *rps_flow_table;
780#endif
781 struct kobject kobj;
782 struct net_device *dev;
783 netdevice_tracker dev_tracker;
784
785#ifdef CONFIG_XDP_SOCKETS
786 struct xsk_buff_pool *pool;
787#endif
788} ____cacheline_aligned_in_smp;
789
790/*
791 * RX queue sysfs structures and functions.
792 */
793struct rx_queue_attribute {
794 struct attribute attr;
795 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
796 ssize_t (*store)(struct netdev_rx_queue *queue,
797 const char *buf, size_t len);
798};
799
800/* XPS map type and offset of the xps map within net_device->xps_maps[]. */
801enum xps_map_type {
802 XPS_CPUS = 0,
803 XPS_RXQS,
804 XPS_MAPS_MAX,
805};
806
807#ifdef CONFIG_XPS
808/*
809 * This structure holds an XPS map which can be of variable length. The
810 * map is an array of queues.
811 */
812struct xps_map {
813 unsigned int len;
814 unsigned int alloc_len;
815 struct rcu_head rcu;
816 u16 queues[];
817};
818#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
819#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
820 - sizeof(struct xps_map)) / sizeof(u16))
821
822/*
823 * This structure holds all XPS maps for device. Maps are indexed by CPU.
824 *
825 * We keep track of the number of cpus/rxqs used when the struct is allocated,
826 * in nr_ids. This will help not accessing out-of-bound memory.
827 *
828 * We keep track of the number of traffic classes used when the struct is
829 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
830 * not crossing its upper bound, as the original dev->num_tc can be updated in
831 * the meantime.
832 */
833struct xps_dev_maps {
834 struct rcu_head rcu;
835 unsigned int nr_ids;
836 s16 num_tc;
837 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
838};
839
840#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
841 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
842
843#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
844 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
845
846#endif /* CONFIG_XPS */
847
848#define TC_MAX_QUEUE 16
849#define TC_BITMASK 15
850/* HW offloaded queuing disciplines txq count and offset maps */
851struct netdev_tc_txq {
852 u16 count;
853 u16 offset;
854};
855
856#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
857/*
858 * This structure is to hold information about the device
859 * configured to run FCoE protocol stack.
860 */
861struct netdev_fcoe_hbainfo {
862 char manufacturer[64];
863 char serial_number[64];
864 char hardware_version[64];
865 char driver_version[64];
866 char optionrom_version[64];
867 char firmware_version[64];
868 char model[256];
869 char model_description[256];
870};
871#endif
872
873#define MAX_PHYS_ITEM_ID_LEN 32
874
875/* This structure holds a unique identifier to identify some
876 * physical item (port for example) used by a netdevice.
877 */
878struct netdev_phys_item_id {
879 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
880 unsigned char id_len;
881};
882
883static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
884 struct netdev_phys_item_id *b)
885{
886 return a->id_len == b->id_len &&
887 memcmp(a->id, b->id, a->id_len) == 0;
888}
889
890typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
891 struct sk_buff *skb,
892 struct net_device *sb_dev);
893
894enum net_device_path_type {
895 DEV_PATH_ETHERNET = 0,
896 DEV_PATH_VLAN,
897 DEV_PATH_BRIDGE,
898 DEV_PATH_PPPOE,
899 DEV_PATH_DSA,
900 DEV_PATH_MTK_WDMA,
901};
902
903struct net_device_path {
904 enum net_device_path_type type;
905 const struct net_device *dev;
906 union {
907 struct {
908 u16 id;
909 __be16 proto;
910 u8 h_dest[ETH_ALEN];
911 } encap;
912 struct {
913 enum {
914 DEV_PATH_BR_VLAN_KEEP,
915 DEV_PATH_BR_VLAN_TAG,
916 DEV_PATH_BR_VLAN_UNTAG,
917 DEV_PATH_BR_VLAN_UNTAG_HW,
918 } vlan_mode;
919 u16 vlan_id;
920 __be16 vlan_proto;
921 } bridge;
922 struct {
923 int port;
924 u16 proto;
925 } dsa;
926 struct {
927 u8 wdma_idx;
928 u8 queue;
929 u16 wcid;
930 u8 bss;
931 } mtk_wdma;
932 };
933};
934
935#define NET_DEVICE_PATH_STACK_MAX 5
936#define NET_DEVICE_PATH_VLAN_MAX 2
937
938struct net_device_path_stack {
939 int num_paths;
940 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
941};
942
943struct net_device_path_ctx {
944 const struct net_device *dev;
945 u8 daddr[ETH_ALEN];
946
947 int num_vlans;
948 struct {
949 u16 id;
950 __be16 proto;
951 } vlan[NET_DEVICE_PATH_VLAN_MAX];
952};
953
954enum tc_setup_type {
955 TC_QUERY_CAPS,
956 TC_SETUP_QDISC_MQPRIO,
957 TC_SETUP_CLSU32,
958 TC_SETUP_CLSFLOWER,
959 TC_SETUP_CLSMATCHALL,
960 TC_SETUP_CLSBPF,
961 TC_SETUP_BLOCK,
962 TC_SETUP_QDISC_CBS,
963 TC_SETUP_QDISC_RED,
964 TC_SETUP_QDISC_PRIO,
965 TC_SETUP_QDISC_MQ,
966 TC_SETUP_QDISC_ETF,
967 TC_SETUP_ROOT_QDISC,
968 TC_SETUP_QDISC_GRED,
969 TC_SETUP_QDISC_TAPRIO,
970 TC_SETUP_FT,
971 TC_SETUP_QDISC_ETS,
972 TC_SETUP_QDISC_TBF,
973 TC_SETUP_QDISC_FIFO,
974 TC_SETUP_QDISC_HTB,
975 TC_SETUP_ACT,
976};
977
978/* These structures hold the attributes of bpf state that are being passed
979 * to the netdevice through the bpf op.
980 */
981enum bpf_netdev_command {
982 /* Set or clear a bpf program used in the earliest stages of packet
983 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
984 * is responsible for calling bpf_prog_put on any old progs that are
985 * stored. In case of error, the callee need not release the new prog
986 * reference, but on success it takes ownership and must bpf_prog_put
987 * when it is no longer used.
988 */
989 XDP_SETUP_PROG,
990 XDP_SETUP_PROG_HW,
991 /* BPF program for offload callbacks, invoked at program load time. */
992 BPF_OFFLOAD_MAP_ALLOC,
993 BPF_OFFLOAD_MAP_FREE,
994 XDP_SETUP_XSK_POOL,
995};
996
997struct bpf_prog_offload_ops;
998struct netlink_ext_ack;
999struct xdp_umem;
1000struct xdp_dev_bulk_queue;
1001struct bpf_xdp_link;
1002
1003enum bpf_xdp_mode {
1004 XDP_MODE_SKB = 0,
1005 XDP_MODE_DRV = 1,
1006 XDP_MODE_HW = 2,
1007 __MAX_XDP_MODE
1008};
1009
1010struct bpf_xdp_entity {
1011 struct bpf_prog *prog;
1012 struct bpf_xdp_link *link;
1013};
1014
1015struct netdev_bpf {
1016 enum bpf_netdev_command command;
1017 union {
1018 /* XDP_SETUP_PROG */
1019 struct {
1020 u32 flags;
1021 struct bpf_prog *prog;
1022 struct netlink_ext_ack *extack;
1023 };
1024 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1025 struct {
1026 struct bpf_offloaded_map *offmap;
1027 };
1028 /* XDP_SETUP_XSK_POOL */
1029 struct {
1030 struct xsk_buff_pool *pool;
1031 u16 queue_id;
1032 } xsk;
1033 };
1034};
1035
1036/* Flags for ndo_xsk_wakeup. */
1037#define XDP_WAKEUP_RX (1 << 0)
1038#define XDP_WAKEUP_TX (1 << 1)
1039
1040#ifdef CONFIG_XFRM_OFFLOAD
1041struct xfrmdev_ops {
1042 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1043 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1044 void (*xdo_dev_state_free) (struct xfrm_state *x);
1045 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1046 struct xfrm_state *x);
1047 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1048 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1049 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1050 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1051 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1052};
1053#endif
1054
1055struct dev_ifalias {
1056 struct rcu_head rcuhead;
1057 char ifalias[];
1058};
1059
1060struct devlink;
1061struct tlsdev_ops;
1062
1063struct netdev_net_notifier {
1064 struct list_head list;
1065 struct notifier_block *nb;
1066};
1067
1068/*
1069 * This structure defines the management hooks for network devices.
1070 * The following hooks can be defined; unless noted otherwise, they are
1071 * optional and can be filled with a null pointer.
1072 *
1073 * int (*ndo_init)(struct net_device *dev);
1074 * This function is called once when a network device is registered.
1075 * The network device can use this for any late stage initialization
1076 * or semantic validation. It can fail with an error code which will
1077 * be propagated back to register_netdev.
1078 *
1079 * void (*ndo_uninit)(struct net_device *dev);
1080 * This function is called when device is unregistered or when registration
1081 * fails. It is not called if init fails.
1082 *
1083 * int (*ndo_open)(struct net_device *dev);
1084 * This function is called when a network device transitions to the up
1085 * state.
1086 *
1087 * int (*ndo_stop)(struct net_device *dev);
1088 * This function is called when a network device transitions to the down
1089 * state.
1090 *
1091 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1092 * struct net_device *dev);
1093 * Called when a packet needs to be transmitted.
1094 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1095 * the queue before that can happen; it's for obsolete devices and weird
1096 * corner cases, but the stack really does a non-trivial amount
1097 * of useless work if you return NETDEV_TX_BUSY.
1098 * Required; cannot be NULL.
1099 *
1100 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1101 * struct net_device *dev
1102 * netdev_features_t features);
1103 * Called by core transmit path to determine if device is capable of
1104 * performing offload operations on a given packet. This is to give
1105 * the device an opportunity to implement any restrictions that cannot
1106 * be otherwise expressed by feature flags. The check is called with
1107 * the set of features that the stack has calculated and it returns
1108 * those the driver believes to be appropriate.
1109 *
1110 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1111 * struct net_device *sb_dev);
1112 * Called to decide which queue to use when device supports multiple
1113 * transmit queues.
1114 *
1115 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1116 * This function is called to allow device receiver to make
1117 * changes to configuration when multicast or promiscuous is enabled.
1118 *
1119 * void (*ndo_set_rx_mode)(struct net_device *dev);
1120 * This function is called device changes address list filtering.
1121 * If driver handles unicast address filtering, it should set
1122 * IFF_UNICAST_FLT in its priv_flags.
1123 *
1124 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1125 * This function is called when the Media Access Control address
1126 * needs to be changed. If this interface is not defined, the
1127 * MAC address can not be changed.
1128 *
1129 * int (*ndo_validate_addr)(struct net_device *dev);
1130 * Test if Media Access Control address is valid for the device.
1131 *
1132 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1133 * Old-style ioctl entry point. This is used internally by the
1134 * appletalk and ieee802154 subsystems but is no longer called by
1135 * the device ioctl handler.
1136 *
1137 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1138 * Used by the bonding driver for its device specific ioctls:
1139 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1140 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1141 *
1142 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1143 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1144 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1145 *
1146 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1147 * Used to set network devices bus interface parameters. This interface
1148 * is retained for legacy reasons; new devices should use the bus
1149 * interface (PCI) for low level management.
1150 *
1151 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1152 * Called when a user wants to change the Maximum Transfer Unit
1153 * of a device.
1154 *
1155 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1156 * Callback used when the transmitter has not made any progress
1157 * for dev->watchdog ticks.
1158 *
1159 * void (*ndo_get_stats64)(struct net_device *dev,
1160 * struct rtnl_link_stats64 *storage);
1161 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1162 * Called when a user wants to get the network device usage
1163 * statistics. Drivers must do one of the following:
1164 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1165 * rtnl_link_stats64 structure passed by the caller.
1166 * 2. Define @ndo_get_stats to update a net_device_stats structure
1167 * (which should normally be dev->stats) and return a pointer to
1168 * it. The structure may be changed asynchronously only if each
1169 * field is written atomically.
1170 * 3. Update dev->stats asynchronously and atomically, and define
1171 * neither operation.
1172 *
1173 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1174 * Return true if this device supports offload stats of this attr_id.
1175 *
1176 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1177 * void *attr_data)
1178 * Get statistics for offload operations by attr_id. Write it into the
1179 * attr_data pointer.
1180 *
1181 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1182 * If device supports VLAN filtering this function is called when a
1183 * VLAN id is registered.
1184 *
1185 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1186 * If device supports VLAN filtering this function is called when a
1187 * VLAN id is unregistered.
1188 *
1189 * void (*ndo_poll_controller)(struct net_device *dev);
1190 *
1191 * SR-IOV management functions.
1192 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1193 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1194 * u8 qos, __be16 proto);
1195 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1196 * int max_tx_rate);
1197 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1198 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1199 * int (*ndo_get_vf_config)(struct net_device *dev,
1200 * int vf, struct ifla_vf_info *ivf);
1201 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1202 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1203 * struct nlattr *port[]);
1204 *
1205 * Enable or disable the VF ability to query its RSS Redirection Table and
1206 * Hash Key. This is needed since on some devices VF share this information
1207 * with PF and querying it may introduce a theoretical security risk.
1208 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1209 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1210 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1211 * void *type_data);
1212 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1213 * This is always called from the stack with the rtnl lock held and netif
1214 * tx queues stopped. This allows the netdevice to perform queue
1215 * management safely.
1216 *
1217 * Fiber Channel over Ethernet (FCoE) offload functions.
1218 * int (*ndo_fcoe_enable)(struct net_device *dev);
1219 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1220 * so the underlying device can perform whatever needed configuration or
1221 * initialization to support acceleration of FCoE traffic.
1222 *
1223 * int (*ndo_fcoe_disable)(struct net_device *dev);
1224 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1225 * so the underlying device can perform whatever needed clean-ups to
1226 * stop supporting acceleration of FCoE traffic.
1227 *
1228 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1229 * struct scatterlist *sgl, unsigned int sgc);
1230 * Called when the FCoE Initiator wants to initialize an I/O that
1231 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1232 * perform necessary setup and returns 1 to indicate the device is set up
1233 * successfully to perform DDP on this I/O, otherwise this returns 0.
1234 *
1235 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1236 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1237 * indicated by the FC exchange id 'xid', so the underlying device can
1238 * clean up and reuse resources for later DDP requests.
1239 *
1240 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1241 * struct scatterlist *sgl, unsigned int sgc);
1242 * Called when the FCoE Target wants to initialize an I/O that
1243 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1244 * perform necessary setup and returns 1 to indicate the device is set up
1245 * successfully to perform DDP on this I/O, otherwise this returns 0.
1246 *
1247 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1248 * struct netdev_fcoe_hbainfo *hbainfo);
1249 * Called when the FCoE Protocol stack wants information on the underlying
1250 * device. This information is utilized by the FCoE protocol stack to
1251 * register attributes with Fiber Channel management service as per the
1252 * FC-GS Fabric Device Management Information(FDMI) specification.
1253 *
1254 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1255 * Called when the underlying device wants to override default World Wide
1256 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1257 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1258 * protocol stack to use.
1259 *
1260 * RFS acceleration.
1261 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1262 * u16 rxq_index, u32 flow_id);
1263 * Set hardware filter for RFS. rxq_index is the target queue index;
1264 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1265 * Return the filter ID on success, or a negative error code.
1266 *
1267 * Slave management functions (for bridge, bonding, etc).
1268 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1269 * Called to make another netdev an underling.
1270 *
1271 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1272 * Called to release previously enslaved netdev.
1273 *
1274 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1275 * struct sk_buff *skb,
1276 * bool all_slaves);
1277 * Get the xmit slave of master device. If all_slaves is true, function
1278 * assume all the slaves can transmit.
1279 *
1280 * Feature/offload setting functions.
1281 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1282 * netdev_features_t features);
1283 * Adjusts the requested feature flags according to device-specific
1284 * constraints, and returns the resulting flags. Must not modify
1285 * the device state.
1286 *
1287 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1288 * Called to update device configuration to new features. Passed
1289 * feature set might be less than what was returned by ndo_fix_features()).
1290 * Must return >0 or -errno if it changed dev->features itself.
1291 *
1292 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1293 * struct net_device *dev,
1294 * const unsigned char *addr, u16 vid, u16 flags,
1295 * struct netlink_ext_ack *extack);
1296 * Adds an FDB entry to dev for addr.
1297 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1298 * struct net_device *dev,
1299 * const unsigned char *addr, u16 vid)
1300 * Deletes the FDB entry from dev coresponding to addr.
1301 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1302 * struct net_device *dev,
1303 * u16 vid,
1304 * struct netlink_ext_ack *extack);
1305 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1306 * struct net_device *dev, struct net_device *filter_dev,
1307 * int *idx)
1308 * Used to add FDB entries to dump requests. Implementers should add
1309 * entries to skb and update idx with the number of entries.
1310 *
1311 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1312 * u16 flags, struct netlink_ext_ack *extack)
1313 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1314 * struct net_device *dev, u32 filter_mask,
1315 * int nlflags)
1316 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1317 * u16 flags);
1318 *
1319 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1320 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1321 * which do not represent real hardware may define this to allow their
1322 * userspace components to manage their virtual carrier state. Devices
1323 * that determine carrier state from physical hardware properties (eg
1324 * network cables) or protocol-dependent mechanisms (eg
1325 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1326 *
1327 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1328 * struct netdev_phys_item_id *ppid);
1329 * Called to get ID of physical port of this device. If driver does
1330 * not implement this, it is assumed that the hw is not able to have
1331 * multiple net devices on single physical port.
1332 *
1333 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1334 * struct netdev_phys_item_id *ppid)
1335 * Called to get the parent ID of the physical port of this device.
1336 *
1337 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1338 * struct net_device *dev)
1339 * Called by upper layer devices to accelerate switching or other
1340 * station functionality into hardware. 'pdev is the lowerdev
1341 * to use for the offload and 'dev' is the net device that will
1342 * back the offload. Returns a pointer to the private structure
1343 * the upper layer will maintain.
1344 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1345 * Called by upper layer device to delete the station created
1346 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1347 * the station and priv is the structure returned by the add
1348 * operation.
1349 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1350 * int queue_index, u32 maxrate);
1351 * Called when a user wants to set a max-rate limitation of specific
1352 * TX queue.
1353 * int (*ndo_get_iflink)(const struct net_device *dev);
1354 * Called to get the iflink value of this device.
1355 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1356 * This function is used to get egress tunnel information for given skb.
1357 * This is useful for retrieving outer tunnel header parameters while
1358 * sampling packet.
1359 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1360 * This function is used to specify the headroom that the skb must
1361 * consider when allocation skb during packet reception. Setting
1362 * appropriate rx headroom value allows avoiding skb head copy on
1363 * forward. Setting a negative value resets the rx headroom to the
1364 * default value.
1365 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1366 * This function is used to set or query state related to XDP on the
1367 * netdevice and manage BPF offload. See definition of
1368 * enum bpf_netdev_command for details.
1369 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1370 * u32 flags);
1371 * This function is used to submit @n XDP packets for transmit on a
1372 * netdevice. Returns number of frames successfully transmitted, frames
1373 * that got dropped are freed/returned via xdp_return_frame().
1374 * Returns negative number, means general error invoking ndo, meaning
1375 * no frames were xmit'ed and core-caller will free all frames.
1376 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1377 * struct xdp_buff *xdp);
1378 * Get the xmit slave of master device based on the xdp_buff.
1379 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1380 * This function is used to wake up the softirq, ksoftirqd or kthread
1381 * responsible for sending and/or receiving packets on a specific
1382 * queue id bound to an AF_XDP socket. The flags field specifies if
1383 * only RX, only Tx, or both should be woken up using the flags
1384 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1385 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1386 * int cmd);
1387 * Add, change, delete or get information on an IPv4 tunnel.
1388 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1389 * If a device is paired with a peer device, return the peer instance.
1390 * The caller must be under RCU read context.
1391 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1392 * Get the forwarding path to reach the real device from the HW destination address
1393 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1394 * const struct skb_shared_hwtstamps *hwtstamps,
1395 * bool cycles);
1396 * Get hardware timestamp based on normal/adjustable time or free running
1397 * cycle counter. This function is required if physical clock supports a
1398 * free running cycle counter.
1399 */
1400struct net_device_ops {
1401 int (*ndo_init)(struct net_device *dev);
1402 void (*ndo_uninit)(struct net_device *dev);
1403 int (*ndo_open)(struct net_device *dev);
1404 int (*ndo_stop)(struct net_device *dev);
1405 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1406 struct net_device *dev);
1407 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1408 struct net_device *dev,
1409 netdev_features_t features);
1410 u16 (*ndo_select_queue)(struct net_device *dev,
1411 struct sk_buff *skb,
1412 struct net_device *sb_dev);
1413 void (*ndo_change_rx_flags)(struct net_device *dev,
1414 int flags);
1415 void (*ndo_set_rx_mode)(struct net_device *dev);
1416 int (*ndo_set_mac_address)(struct net_device *dev,
1417 void *addr);
1418 int (*ndo_validate_addr)(struct net_device *dev);
1419 int (*ndo_do_ioctl)(struct net_device *dev,
1420 struct ifreq *ifr, int cmd);
1421 int (*ndo_eth_ioctl)(struct net_device *dev,
1422 struct ifreq *ifr, int cmd);
1423 int (*ndo_siocbond)(struct net_device *dev,
1424 struct ifreq *ifr, int cmd);
1425 int (*ndo_siocwandev)(struct net_device *dev,
1426 struct if_settings *ifs);
1427 int (*ndo_siocdevprivate)(struct net_device *dev,
1428 struct ifreq *ifr,
1429 void __user *data, int cmd);
1430 int (*ndo_set_config)(struct net_device *dev,
1431 struct ifmap *map);
1432 int (*ndo_change_mtu)(struct net_device *dev,
1433 int new_mtu);
1434 int (*ndo_neigh_setup)(struct net_device *dev,
1435 struct neigh_parms *);
1436 void (*ndo_tx_timeout) (struct net_device *dev,
1437 unsigned int txqueue);
1438
1439 void (*ndo_get_stats64)(struct net_device *dev,
1440 struct rtnl_link_stats64 *storage);
1441 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1442 int (*ndo_get_offload_stats)(int attr_id,
1443 const struct net_device *dev,
1444 void *attr_data);
1445 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1446
1447 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1448 __be16 proto, u16 vid);
1449 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1450 __be16 proto, u16 vid);
1451#ifdef CONFIG_NET_POLL_CONTROLLER
1452 void (*ndo_poll_controller)(struct net_device *dev);
1453 int (*ndo_netpoll_setup)(struct net_device *dev,
1454 struct netpoll_info *info);
1455 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1456#endif
1457 int (*ndo_set_vf_mac)(struct net_device *dev,
1458 int queue, u8 *mac);
1459 int (*ndo_set_vf_vlan)(struct net_device *dev,
1460 int queue, u16 vlan,
1461 u8 qos, __be16 proto);
1462 int (*ndo_set_vf_rate)(struct net_device *dev,
1463 int vf, int min_tx_rate,
1464 int max_tx_rate);
1465 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1466 int vf, bool setting);
1467 int (*ndo_set_vf_trust)(struct net_device *dev,
1468 int vf, bool setting);
1469 int (*ndo_get_vf_config)(struct net_device *dev,
1470 int vf,
1471 struct ifla_vf_info *ivf);
1472 int (*ndo_set_vf_link_state)(struct net_device *dev,
1473 int vf, int link_state);
1474 int (*ndo_get_vf_stats)(struct net_device *dev,
1475 int vf,
1476 struct ifla_vf_stats
1477 *vf_stats);
1478 int (*ndo_set_vf_port)(struct net_device *dev,
1479 int vf,
1480 struct nlattr *port[]);
1481 int (*ndo_get_vf_port)(struct net_device *dev,
1482 int vf, struct sk_buff *skb);
1483 int (*ndo_get_vf_guid)(struct net_device *dev,
1484 int vf,
1485 struct ifla_vf_guid *node_guid,
1486 struct ifla_vf_guid *port_guid);
1487 int (*ndo_set_vf_guid)(struct net_device *dev,
1488 int vf, u64 guid,
1489 int guid_type);
1490 int (*ndo_set_vf_rss_query_en)(
1491 struct net_device *dev,
1492 int vf, bool setting);
1493 int (*ndo_setup_tc)(struct net_device *dev,
1494 enum tc_setup_type type,
1495 void *type_data);
1496#if IS_ENABLED(CONFIG_FCOE)
1497 int (*ndo_fcoe_enable)(struct net_device *dev);
1498 int (*ndo_fcoe_disable)(struct net_device *dev);
1499 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1500 u16 xid,
1501 struct scatterlist *sgl,
1502 unsigned int sgc);
1503 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1504 u16 xid);
1505 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1506 u16 xid,
1507 struct scatterlist *sgl,
1508 unsigned int sgc);
1509 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1510 struct netdev_fcoe_hbainfo *hbainfo);
1511#endif
1512
1513#if IS_ENABLED(CONFIG_LIBFCOE)
1514#define NETDEV_FCOE_WWNN 0
1515#define NETDEV_FCOE_WWPN 1
1516 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1517 u64 *wwn, int type);
1518#endif
1519
1520#ifdef CONFIG_RFS_ACCEL
1521 int (*ndo_rx_flow_steer)(struct net_device *dev,
1522 const struct sk_buff *skb,
1523 u16 rxq_index,
1524 u32 flow_id);
1525#endif
1526 int (*ndo_add_slave)(struct net_device *dev,
1527 struct net_device *slave_dev,
1528 struct netlink_ext_ack *extack);
1529 int (*ndo_del_slave)(struct net_device *dev,
1530 struct net_device *slave_dev);
1531 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1532 struct sk_buff *skb,
1533 bool all_slaves);
1534 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1535 struct sock *sk);
1536 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1537 netdev_features_t features);
1538 int (*ndo_set_features)(struct net_device *dev,
1539 netdev_features_t features);
1540 int (*ndo_neigh_construct)(struct net_device *dev,
1541 struct neighbour *n);
1542 void (*ndo_neigh_destroy)(struct net_device *dev,
1543 struct neighbour *n);
1544
1545 int (*ndo_fdb_add)(struct ndmsg *ndm,
1546 struct nlattr *tb[],
1547 struct net_device *dev,
1548 const unsigned char *addr,
1549 u16 vid,
1550 u16 flags,
1551 struct netlink_ext_ack *extack);
1552 int (*ndo_fdb_del)(struct ndmsg *ndm,
1553 struct nlattr *tb[],
1554 struct net_device *dev,
1555 const unsigned char *addr,
1556 u16 vid, struct netlink_ext_ack *extack);
1557 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1558 struct nlattr *tb[],
1559 struct net_device *dev,
1560 u16 vid,
1561 struct netlink_ext_ack *extack);
1562 int (*ndo_fdb_dump)(struct sk_buff *skb,
1563 struct netlink_callback *cb,
1564 struct net_device *dev,
1565 struct net_device *filter_dev,
1566 int *idx);
1567 int (*ndo_fdb_get)(struct sk_buff *skb,
1568 struct nlattr *tb[],
1569 struct net_device *dev,
1570 const unsigned char *addr,
1571 u16 vid, u32 portid, u32 seq,
1572 struct netlink_ext_ack *extack);
1573 int (*ndo_bridge_setlink)(struct net_device *dev,
1574 struct nlmsghdr *nlh,
1575 u16 flags,
1576 struct netlink_ext_ack *extack);
1577 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1578 u32 pid, u32 seq,
1579 struct net_device *dev,
1580 u32 filter_mask,
1581 int nlflags);
1582 int (*ndo_bridge_dellink)(struct net_device *dev,
1583 struct nlmsghdr *nlh,
1584 u16 flags);
1585 int (*ndo_change_carrier)(struct net_device *dev,
1586 bool new_carrier);
1587 int (*ndo_get_phys_port_id)(struct net_device *dev,
1588 struct netdev_phys_item_id *ppid);
1589 int (*ndo_get_port_parent_id)(struct net_device *dev,
1590 struct netdev_phys_item_id *ppid);
1591 int (*ndo_get_phys_port_name)(struct net_device *dev,
1592 char *name, size_t len);
1593 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1594 struct net_device *dev);
1595 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1596 void *priv);
1597
1598 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1599 int queue_index,
1600 u32 maxrate);
1601 int (*ndo_get_iflink)(const struct net_device *dev);
1602 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1603 struct sk_buff *skb);
1604 void (*ndo_set_rx_headroom)(struct net_device *dev,
1605 int needed_headroom);
1606 int (*ndo_bpf)(struct net_device *dev,
1607 struct netdev_bpf *bpf);
1608 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1609 struct xdp_frame **xdp,
1610 u32 flags);
1611 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1612 struct xdp_buff *xdp);
1613 int (*ndo_xsk_wakeup)(struct net_device *dev,
1614 u32 queue_id, u32 flags);
1615 int (*ndo_tunnel_ctl)(struct net_device *dev,
1616 struct ip_tunnel_parm *p, int cmd);
1617 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1618 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1619 struct net_device_path *path);
1620 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1621 const struct skb_shared_hwtstamps *hwtstamps,
1622 bool cycles);
1623};
1624
1625struct xdp_metadata_ops {
1626 int (*xmo_rx_timestamp)(const struct xdp_md *ctx, u64 *timestamp);
1627 int (*xmo_rx_hash)(const struct xdp_md *ctx, u32 *hash);
1628};
1629
1630/**
1631 * enum netdev_priv_flags - &struct net_device priv_flags
1632 *
1633 * These are the &struct net_device, they are only set internally
1634 * by drivers and used in the kernel. These flags are invisible to
1635 * userspace; this means that the order of these flags can change
1636 * during any kernel release.
1637 *
1638 * You should have a pretty good reason to be extending these flags.
1639 *
1640 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1641 * @IFF_EBRIDGE: Ethernet bridging device
1642 * @IFF_BONDING: bonding master or slave
1643 * @IFF_ISATAP: ISATAP interface (RFC4214)
1644 * @IFF_WAN_HDLC: WAN HDLC device
1645 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1646 * release skb->dst
1647 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1648 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1649 * @IFF_MACVLAN_PORT: device used as macvlan port
1650 * @IFF_BRIDGE_PORT: device used as bridge port
1651 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1652 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1653 * @IFF_UNICAST_FLT: Supports unicast filtering
1654 * @IFF_TEAM_PORT: device used as team port
1655 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1656 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1657 * change when it's running
1658 * @IFF_MACVLAN: Macvlan device
1659 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1660 * underlying stacked devices
1661 * @IFF_L3MDEV_MASTER: device is an L3 master device
1662 * @IFF_NO_QUEUE: device can run without qdisc attached
1663 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1664 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1665 * @IFF_TEAM: device is a team device
1666 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1667 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1668 * entity (i.e. the master device for bridged veth)
1669 * @IFF_MACSEC: device is a MACsec device
1670 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1671 * @IFF_FAILOVER: device is a failover master device
1672 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1673 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1674 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1675 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1676 * skb_headlen(skb) == 0 (data starts from frag0)
1677 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1678 */
1679enum netdev_priv_flags {
1680 IFF_802_1Q_VLAN = 1<<0,
1681 IFF_EBRIDGE = 1<<1,
1682 IFF_BONDING = 1<<2,
1683 IFF_ISATAP = 1<<3,
1684 IFF_WAN_HDLC = 1<<4,
1685 IFF_XMIT_DST_RELEASE = 1<<5,
1686 IFF_DONT_BRIDGE = 1<<6,
1687 IFF_DISABLE_NETPOLL = 1<<7,
1688 IFF_MACVLAN_PORT = 1<<8,
1689 IFF_BRIDGE_PORT = 1<<9,
1690 IFF_OVS_DATAPATH = 1<<10,
1691 IFF_TX_SKB_SHARING = 1<<11,
1692 IFF_UNICAST_FLT = 1<<12,
1693 IFF_TEAM_PORT = 1<<13,
1694 IFF_SUPP_NOFCS = 1<<14,
1695 IFF_LIVE_ADDR_CHANGE = 1<<15,
1696 IFF_MACVLAN = 1<<16,
1697 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1698 IFF_L3MDEV_MASTER = 1<<18,
1699 IFF_NO_QUEUE = 1<<19,
1700 IFF_OPENVSWITCH = 1<<20,
1701 IFF_L3MDEV_SLAVE = 1<<21,
1702 IFF_TEAM = 1<<22,
1703 IFF_RXFH_CONFIGURED = 1<<23,
1704 IFF_PHONY_HEADROOM = 1<<24,
1705 IFF_MACSEC = 1<<25,
1706 IFF_NO_RX_HANDLER = 1<<26,
1707 IFF_FAILOVER = 1<<27,
1708 IFF_FAILOVER_SLAVE = 1<<28,
1709 IFF_L3MDEV_RX_HANDLER = 1<<29,
1710 IFF_NO_ADDRCONF = BIT_ULL(30),
1711 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1712 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1713};
1714
1715#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1716#define IFF_EBRIDGE IFF_EBRIDGE
1717#define IFF_BONDING IFF_BONDING
1718#define IFF_ISATAP IFF_ISATAP
1719#define IFF_WAN_HDLC IFF_WAN_HDLC
1720#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1721#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1722#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1723#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1724#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1725#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1726#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1727#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1728#define IFF_TEAM_PORT IFF_TEAM_PORT
1729#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1730#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1731#define IFF_MACVLAN IFF_MACVLAN
1732#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1733#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1734#define IFF_NO_QUEUE IFF_NO_QUEUE
1735#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1736#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1737#define IFF_TEAM IFF_TEAM
1738#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1739#define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1740#define IFF_MACSEC IFF_MACSEC
1741#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1742#define IFF_FAILOVER IFF_FAILOVER
1743#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1744#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1745#define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1746
1747/* Specifies the type of the struct net_device::ml_priv pointer */
1748enum netdev_ml_priv_type {
1749 ML_PRIV_NONE,
1750 ML_PRIV_CAN,
1751};
1752
1753/**
1754 * struct net_device - The DEVICE structure.
1755 *
1756 * Actually, this whole structure is a big mistake. It mixes I/O
1757 * data with strictly "high-level" data, and it has to know about
1758 * almost every data structure used in the INET module.
1759 *
1760 * @name: This is the first field of the "visible" part of this structure
1761 * (i.e. as seen by users in the "Space.c" file). It is the name
1762 * of the interface.
1763 *
1764 * @name_node: Name hashlist node
1765 * @ifalias: SNMP alias
1766 * @mem_end: Shared memory end
1767 * @mem_start: Shared memory start
1768 * @base_addr: Device I/O address
1769 * @irq: Device IRQ number
1770 *
1771 * @state: Generic network queuing layer state, see netdev_state_t
1772 * @dev_list: The global list of network devices
1773 * @napi_list: List entry used for polling NAPI devices
1774 * @unreg_list: List entry when we are unregistering the
1775 * device; see the function unregister_netdev
1776 * @close_list: List entry used when we are closing the device
1777 * @ptype_all: Device-specific packet handlers for all protocols
1778 * @ptype_specific: Device-specific, protocol-specific packet handlers
1779 *
1780 * @adj_list: Directly linked devices, like slaves for bonding
1781 * @features: Currently active device features
1782 * @hw_features: User-changeable features
1783 *
1784 * @wanted_features: User-requested features
1785 * @vlan_features: Mask of features inheritable by VLAN devices
1786 *
1787 * @hw_enc_features: Mask of features inherited by encapsulating devices
1788 * This field indicates what encapsulation
1789 * offloads the hardware is capable of doing,
1790 * and drivers will need to set them appropriately.
1791 *
1792 * @mpls_features: Mask of features inheritable by MPLS
1793 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1794 *
1795 * @ifindex: interface index
1796 * @group: The group the device belongs to
1797 *
1798 * @stats: Statistics struct, which was left as a legacy, use
1799 * rtnl_link_stats64 instead
1800 *
1801 * @core_stats: core networking counters,
1802 * do not use this in drivers
1803 * @carrier_up_count: Number of times the carrier has been up
1804 * @carrier_down_count: Number of times the carrier has been down
1805 *
1806 * @wireless_handlers: List of functions to handle Wireless Extensions,
1807 * instead of ioctl,
1808 * see <net/iw_handler.h> for details.
1809 * @wireless_data: Instance data managed by the core of wireless extensions
1810 *
1811 * @netdev_ops: Includes several pointers to callbacks,
1812 * if one wants to override the ndo_*() functions
1813 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1814 * @ethtool_ops: Management operations
1815 * @l3mdev_ops: Layer 3 master device operations
1816 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1817 * discovery handling. Necessary for e.g. 6LoWPAN.
1818 * @xfrmdev_ops: Transformation offload operations
1819 * @tlsdev_ops: Transport Layer Security offload operations
1820 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1821 * of Layer 2 headers.
1822 *
1823 * @flags: Interface flags (a la BSD)
1824 * @xdp_features: XDP capability supported by the device
1825 * @priv_flags: Like 'flags' but invisible to userspace,
1826 * see if.h for the definitions
1827 * @gflags: Global flags ( kept as legacy )
1828 * @padded: How much padding added by alloc_netdev()
1829 * @operstate: RFC2863 operstate
1830 * @link_mode: Mapping policy to operstate
1831 * @if_port: Selectable AUI, TP, ...
1832 * @dma: DMA channel
1833 * @mtu: Interface MTU value
1834 * @min_mtu: Interface Minimum MTU value
1835 * @max_mtu: Interface Maximum MTU value
1836 * @type: Interface hardware type
1837 * @hard_header_len: Maximum hardware header length.
1838 * @min_header_len: Minimum hardware header length
1839 *
1840 * @needed_headroom: Extra headroom the hardware may need, but not in all
1841 * cases can this be guaranteed
1842 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1843 * cases can this be guaranteed. Some cases also use
1844 * LL_MAX_HEADER instead to allocate the skb
1845 *
1846 * interface address info:
1847 *
1848 * @perm_addr: Permanent hw address
1849 * @addr_assign_type: Hw address assignment type
1850 * @addr_len: Hardware address length
1851 * @upper_level: Maximum depth level of upper devices.
1852 * @lower_level: Maximum depth level of lower devices.
1853 * @neigh_priv_len: Used in neigh_alloc()
1854 * @dev_id: Used to differentiate devices that share
1855 * the same link layer address
1856 * @dev_port: Used to differentiate devices that share
1857 * the same function
1858 * @addr_list_lock: XXX: need comments on this one
1859 * @name_assign_type: network interface name assignment type
1860 * @uc_promisc: Counter that indicates promiscuous mode
1861 * has been enabled due to the need to listen to
1862 * additional unicast addresses in a device that
1863 * does not implement ndo_set_rx_mode()
1864 * @uc: unicast mac addresses
1865 * @mc: multicast mac addresses
1866 * @dev_addrs: list of device hw addresses
1867 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1868 * @promiscuity: Number of times the NIC is told to work in
1869 * promiscuous mode; if it becomes 0 the NIC will
1870 * exit promiscuous mode
1871 * @allmulti: Counter, enables or disables allmulticast mode
1872 *
1873 * @vlan_info: VLAN info
1874 * @dsa_ptr: dsa specific data
1875 * @tipc_ptr: TIPC specific data
1876 * @atalk_ptr: AppleTalk link
1877 * @ip_ptr: IPv4 specific data
1878 * @ip6_ptr: IPv6 specific data
1879 * @ax25_ptr: AX.25 specific data
1880 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1881 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1882 * device struct
1883 * @mpls_ptr: mpls_dev struct pointer
1884 * @mctp_ptr: MCTP specific data
1885 *
1886 * @dev_addr: Hw address (before bcast,
1887 * because most packets are unicast)
1888 *
1889 * @_rx: Array of RX queues
1890 * @num_rx_queues: Number of RX queues
1891 * allocated at register_netdev() time
1892 * @real_num_rx_queues: Number of RX queues currently active in device
1893 * @xdp_prog: XDP sockets filter program pointer
1894 * @gro_flush_timeout: timeout for GRO layer in NAPI
1895 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1896 * allow to avoid NIC hard IRQ, on busy queues.
1897 *
1898 * @rx_handler: handler for received packets
1899 * @rx_handler_data: XXX: need comments on this one
1900 * @miniq_ingress: ingress/clsact qdisc specific data for
1901 * ingress processing
1902 * @ingress_queue: XXX: need comments on this one
1903 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1904 * @broadcast: hw bcast address
1905 *
1906 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1907 * indexed by RX queue number. Assigned by driver.
1908 * This must only be set if the ndo_rx_flow_steer
1909 * operation is defined
1910 * @index_hlist: Device index hash chain
1911 *
1912 * @_tx: Array of TX queues
1913 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1914 * @real_num_tx_queues: Number of TX queues currently active in device
1915 * @qdisc: Root qdisc from userspace point of view
1916 * @tx_queue_len: Max frames per queue allowed
1917 * @tx_global_lock: XXX: need comments on this one
1918 * @xdp_bulkq: XDP device bulk queue
1919 * @xps_maps: all CPUs/RXQs maps for XPS device
1920 *
1921 * @xps_maps: XXX: need comments on this one
1922 * @miniq_egress: clsact qdisc specific data for
1923 * egress processing
1924 * @nf_hooks_egress: netfilter hooks executed for egress packets
1925 * @qdisc_hash: qdisc hash table
1926 * @watchdog_timeo: Represents the timeout that is used by
1927 * the watchdog (see dev_watchdog())
1928 * @watchdog_timer: List of timers
1929 *
1930 * @proto_down_reason: reason a netdev interface is held down
1931 * @pcpu_refcnt: Number of references to this device
1932 * @dev_refcnt: Number of references to this device
1933 * @refcnt_tracker: Tracker directory for tracked references to this device
1934 * @todo_list: Delayed register/unregister
1935 * @link_watch_list: XXX: need comments on this one
1936 *
1937 * @reg_state: Register/unregister state machine
1938 * @dismantle: Device is going to be freed
1939 * @rtnl_link_state: This enum represents the phases of creating
1940 * a new link
1941 *
1942 * @needs_free_netdev: Should unregister perform free_netdev?
1943 * @priv_destructor: Called from unregister
1944 * @npinfo: XXX: need comments on this one
1945 * @nd_net: Network namespace this network device is inside
1946 *
1947 * @ml_priv: Mid-layer private
1948 * @ml_priv_type: Mid-layer private type
1949 * @lstats: Loopback statistics
1950 * @tstats: Tunnel statistics
1951 * @dstats: Dummy statistics
1952 * @vstats: Virtual ethernet statistics
1953 *
1954 * @garp_port: GARP
1955 * @mrp_port: MRP
1956 *
1957 * @dm_private: Drop monitor private
1958 *
1959 * @dev: Class/net/name entry
1960 * @sysfs_groups: Space for optional device, statistics and wireless
1961 * sysfs groups
1962 *
1963 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1964 * @rtnl_link_ops: Rtnl_link_ops
1965 *
1966 * @gso_max_size: Maximum size of generic segmentation offload
1967 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1968 * @gso_max_segs: Maximum number of segments that can be passed to the
1969 * NIC for GSO
1970 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1971 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
1972 * for IPv4.
1973 *
1974 * @dcbnl_ops: Data Center Bridging netlink ops
1975 * @num_tc: Number of traffic classes in the net device
1976 * @tc_to_txq: XXX: need comments on this one
1977 * @prio_tc_map: XXX: need comments on this one
1978 *
1979 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1980 *
1981 * @priomap: XXX: need comments on this one
1982 * @phydev: Physical device may attach itself
1983 * for hardware timestamping
1984 * @sfp_bus: attached &struct sfp_bus structure.
1985 *
1986 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1987 *
1988 * @proto_down: protocol port state information can be sent to the
1989 * switch driver and used to set the phys state of the
1990 * switch port.
1991 *
1992 * @wol_enabled: Wake-on-LAN is enabled
1993 *
1994 * @threaded: napi threaded mode is enabled
1995 *
1996 * @net_notifier_list: List of per-net netdev notifier block
1997 * that follow this device when it is moved
1998 * to another network namespace.
1999 *
2000 * @macsec_ops: MACsec offloading ops
2001 *
2002 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2003 * offload capabilities of the device
2004 * @udp_tunnel_nic: UDP tunnel offload state
2005 * @xdp_state: stores info on attached XDP BPF programs
2006 *
2007 * @nested_level: Used as a parameter of spin_lock_nested() of
2008 * dev->addr_list_lock.
2009 * @unlink_list: As netif_addr_lock() can be called recursively,
2010 * keep a list of interfaces to be deleted.
2011 * @gro_max_size: Maximum size of aggregated packet in generic
2012 * receive offload (GRO)
2013 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2014 * receive offload (GRO), for IPv4.
2015 *
2016 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2017 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2018 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2019 * @dev_registered_tracker: tracker for reference held while
2020 * registered
2021 * @offload_xstats_l3: L3 HW stats for this netdevice.
2022 *
2023 * @devlink_port: Pointer to related devlink port structure.
2024 * Assigned by a driver before netdev registration using
2025 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2026 * during the time netdevice is registered.
2027 *
2028 * FIXME: cleanup struct net_device such that network protocol info
2029 * moves out.
2030 */
2031
2032struct net_device {
2033 char name[IFNAMSIZ];
2034 struct netdev_name_node *name_node;
2035 struct dev_ifalias __rcu *ifalias;
2036 /*
2037 * I/O specific fields
2038 * FIXME: Merge these and struct ifmap into one
2039 */
2040 unsigned long mem_end;
2041 unsigned long mem_start;
2042 unsigned long base_addr;
2043
2044 /*
2045 * Some hardware also needs these fields (state,dev_list,
2046 * napi_list,unreg_list,close_list) but they are not
2047 * part of the usual set specified in Space.c.
2048 */
2049
2050 unsigned long state;
2051
2052 struct list_head dev_list;
2053 struct list_head napi_list;
2054 struct list_head unreg_list;
2055 struct list_head close_list;
2056 struct list_head ptype_all;
2057 struct list_head ptype_specific;
2058
2059 struct {
2060 struct list_head upper;
2061 struct list_head lower;
2062 } adj_list;
2063
2064 /* Read-mostly cache-line for fast-path access */
2065 unsigned int flags;
2066 xdp_features_t xdp_features;
2067 unsigned long long priv_flags;
2068 const struct net_device_ops *netdev_ops;
2069 const struct xdp_metadata_ops *xdp_metadata_ops;
2070 int ifindex;
2071 unsigned short gflags;
2072 unsigned short hard_header_len;
2073
2074 /* Note : dev->mtu is often read without holding a lock.
2075 * Writers usually hold RTNL.
2076 * It is recommended to use READ_ONCE() to annotate the reads,
2077 * and to use WRITE_ONCE() to annotate the writes.
2078 */
2079 unsigned int mtu;
2080 unsigned short needed_headroom;
2081 unsigned short needed_tailroom;
2082
2083 netdev_features_t features;
2084 netdev_features_t hw_features;
2085 netdev_features_t wanted_features;
2086 netdev_features_t vlan_features;
2087 netdev_features_t hw_enc_features;
2088 netdev_features_t mpls_features;
2089 netdev_features_t gso_partial_features;
2090
2091 unsigned int min_mtu;
2092 unsigned int max_mtu;
2093 unsigned short type;
2094 unsigned char min_header_len;
2095 unsigned char name_assign_type;
2096
2097 int group;
2098
2099 struct net_device_stats stats; /* not used by modern drivers */
2100
2101 struct net_device_core_stats __percpu *core_stats;
2102
2103 /* Stats to monitor link on/off, flapping */
2104 atomic_t carrier_up_count;
2105 atomic_t carrier_down_count;
2106
2107#ifdef CONFIG_WIRELESS_EXT
2108 const struct iw_handler_def *wireless_handlers;
2109 struct iw_public_data *wireless_data;
2110#endif
2111 const struct ethtool_ops *ethtool_ops;
2112#ifdef CONFIG_NET_L3_MASTER_DEV
2113 const struct l3mdev_ops *l3mdev_ops;
2114#endif
2115#if IS_ENABLED(CONFIG_IPV6)
2116 const struct ndisc_ops *ndisc_ops;
2117#endif
2118
2119#ifdef CONFIG_XFRM_OFFLOAD
2120 const struct xfrmdev_ops *xfrmdev_ops;
2121#endif
2122
2123#if IS_ENABLED(CONFIG_TLS_DEVICE)
2124 const struct tlsdev_ops *tlsdev_ops;
2125#endif
2126
2127 const struct header_ops *header_ops;
2128
2129 unsigned char operstate;
2130 unsigned char link_mode;
2131
2132 unsigned char if_port;
2133 unsigned char dma;
2134
2135 /* Interface address info. */
2136 unsigned char perm_addr[MAX_ADDR_LEN];
2137 unsigned char addr_assign_type;
2138 unsigned char addr_len;
2139 unsigned char upper_level;
2140 unsigned char lower_level;
2141
2142 unsigned short neigh_priv_len;
2143 unsigned short dev_id;
2144 unsigned short dev_port;
2145 unsigned short padded;
2146
2147 spinlock_t addr_list_lock;
2148 int irq;
2149
2150 struct netdev_hw_addr_list uc;
2151 struct netdev_hw_addr_list mc;
2152 struct netdev_hw_addr_list dev_addrs;
2153
2154#ifdef CONFIG_SYSFS
2155 struct kset *queues_kset;
2156#endif
2157#ifdef CONFIG_LOCKDEP
2158 struct list_head unlink_list;
2159#endif
2160 unsigned int promiscuity;
2161 unsigned int allmulti;
2162 bool uc_promisc;
2163#ifdef CONFIG_LOCKDEP
2164 unsigned char nested_level;
2165#endif
2166
2167
2168 /* Protocol-specific pointers */
2169
2170 struct in_device __rcu *ip_ptr;
2171 struct inet6_dev __rcu *ip6_ptr;
2172#if IS_ENABLED(CONFIG_VLAN_8021Q)
2173 struct vlan_info __rcu *vlan_info;
2174#endif
2175#if IS_ENABLED(CONFIG_NET_DSA)
2176 struct dsa_port *dsa_ptr;
2177#endif
2178#if IS_ENABLED(CONFIG_TIPC)
2179 struct tipc_bearer __rcu *tipc_ptr;
2180#endif
2181#if IS_ENABLED(CONFIG_ATALK)
2182 void *atalk_ptr;
2183#endif
2184#if IS_ENABLED(CONFIG_AX25)
2185 void *ax25_ptr;
2186#endif
2187#if IS_ENABLED(CONFIG_CFG80211)
2188 struct wireless_dev *ieee80211_ptr;
2189#endif
2190#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2191 struct wpan_dev *ieee802154_ptr;
2192#endif
2193#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2194 struct mpls_dev __rcu *mpls_ptr;
2195#endif
2196#if IS_ENABLED(CONFIG_MCTP)
2197 struct mctp_dev __rcu *mctp_ptr;
2198#endif
2199
2200/*
2201 * Cache lines mostly used on receive path (including eth_type_trans())
2202 */
2203 /* Interface address info used in eth_type_trans() */
2204 const unsigned char *dev_addr;
2205
2206 struct netdev_rx_queue *_rx;
2207 unsigned int num_rx_queues;
2208 unsigned int real_num_rx_queues;
2209
2210 struct bpf_prog __rcu *xdp_prog;
2211 unsigned long gro_flush_timeout;
2212 int napi_defer_hard_irqs;
2213#define GRO_LEGACY_MAX_SIZE 65536u
2214/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2215 * and shinfo->gso_segs is a 16bit field.
2216 */
2217#define GRO_MAX_SIZE (8 * 65535u)
2218 unsigned int gro_max_size;
2219 unsigned int gro_ipv4_max_size;
2220 rx_handler_func_t __rcu *rx_handler;
2221 void __rcu *rx_handler_data;
2222
2223#ifdef CONFIG_NET_CLS_ACT
2224 struct mini_Qdisc __rcu *miniq_ingress;
2225#endif
2226 struct netdev_queue __rcu *ingress_queue;
2227#ifdef CONFIG_NETFILTER_INGRESS
2228 struct nf_hook_entries __rcu *nf_hooks_ingress;
2229#endif
2230
2231 unsigned char broadcast[MAX_ADDR_LEN];
2232#ifdef CONFIG_RFS_ACCEL
2233 struct cpu_rmap *rx_cpu_rmap;
2234#endif
2235 struct hlist_node index_hlist;
2236
2237/*
2238 * Cache lines mostly used on transmit path
2239 */
2240 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2241 unsigned int num_tx_queues;
2242 unsigned int real_num_tx_queues;
2243 struct Qdisc __rcu *qdisc;
2244 unsigned int tx_queue_len;
2245 spinlock_t tx_global_lock;
2246
2247 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2248
2249#ifdef CONFIG_XPS
2250 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2251#endif
2252#ifdef CONFIG_NET_CLS_ACT
2253 struct mini_Qdisc __rcu *miniq_egress;
2254#endif
2255#ifdef CONFIG_NETFILTER_EGRESS
2256 struct nf_hook_entries __rcu *nf_hooks_egress;
2257#endif
2258
2259#ifdef CONFIG_NET_SCHED
2260 DECLARE_HASHTABLE (qdisc_hash, 4);
2261#endif
2262 /* These may be needed for future network-power-down code. */
2263 struct timer_list watchdog_timer;
2264 int watchdog_timeo;
2265
2266 u32 proto_down_reason;
2267
2268 struct list_head todo_list;
2269
2270#ifdef CONFIG_PCPU_DEV_REFCNT
2271 int __percpu *pcpu_refcnt;
2272#else
2273 refcount_t dev_refcnt;
2274#endif
2275 struct ref_tracker_dir refcnt_tracker;
2276
2277 struct list_head link_watch_list;
2278
2279 enum { NETREG_UNINITIALIZED=0,
2280 NETREG_REGISTERED, /* completed register_netdevice */
2281 NETREG_UNREGISTERING, /* called unregister_netdevice */
2282 NETREG_UNREGISTERED, /* completed unregister todo */
2283 NETREG_RELEASED, /* called free_netdev */
2284 NETREG_DUMMY, /* dummy device for NAPI poll */
2285 } reg_state:8;
2286
2287 bool dismantle;
2288
2289 enum {
2290 RTNL_LINK_INITIALIZED,
2291 RTNL_LINK_INITIALIZING,
2292 } rtnl_link_state:16;
2293
2294 bool needs_free_netdev;
2295 void (*priv_destructor)(struct net_device *dev);
2296
2297#ifdef CONFIG_NETPOLL
2298 struct netpoll_info __rcu *npinfo;
2299#endif
2300
2301 possible_net_t nd_net;
2302
2303 /* mid-layer private */
2304 void *ml_priv;
2305 enum netdev_ml_priv_type ml_priv_type;
2306
2307 union {
2308 struct pcpu_lstats __percpu *lstats;
2309 struct pcpu_sw_netstats __percpu *tstats;
2310 struct pcpu_dstats __percpu *dstats;
2311 };
2312
2313#if IS_ENABLED(CONFIG_GARP)
2314 struct garp_port __rcu *garp_port;
2315#endif
2316#if IS_ENABLED(CONFIG_MRP)
2317 struct mrp_port __rcu *mrp_port;
2318#endif
2319#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2320 struct dm_hw_stat_delta __rcu *dm_private;
2321#endif
2322 struct device dev;
2323 const struct attribute_group *sysfs_groups[4];
2324 const struct attribute_group *sysfs_rx_queue_group;
2325
2326 const struct rtnl_link_ops *rtnl_link_ops;
2327
2328 /* for setting kernel sock attribute on TCP connection setup */
2329#define GSO_MAX_SEGS 65535u
2330#define GSO_LEGACY_MAX_SIZE 65536u
2331/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2332 * and shinfo->gso_segs is a 16bit field.
2333 */
2334#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2335
2336 unsigned int gso_max_size;
2337#define TSO_LEGACY_MAX_SIZE 65536
2338#define TSO_MAX_SIZE UINT_MAX
2339 unsigned int tso_max_size;
2340 u16 gso_max_segs;
2341#define TSO_MAX_SEGS U16_MAX
2342 u16 tso_max_segs;
2343 unsigned int gso_ipv4_max_size;
2344
2345#ifdef CONFIG_DCB
2346 const struct dcbnl_rtnl_ops *dcbnl_ops;
2347#endif
2348 s16 num_tc;
2349 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2350 u8 prio_tc_map[TC_BITMASK + 1];
2351
2352#if IS_ENABLED(CONFIG_FCOE)
2353 unsigned int fcoe_ddp_xid;
2354#endif
2355#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2356 struct netprio_map __rcu *priomap;
2357#endif
2358 struct phy_device *phydev;
2359 struct sfp_bus *sfp_bus;
2360 struct lock_class_key *qdisc_tx_busylock;
2361 bool proto_down;
2362 unsigned wol_enabled:1;
2363 unsigned threaded:1;
2364
2365 struct list_head net_notifier_list;
2366
2367#if IS_ENABLED(CONFIG_MACSEC)
2368 /* MACsec management functions */
2369 const struct macsec_ops *macsec_ops;
2370#endif
2371 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2372 struct udp_tunnel_nic *udp_tunnel_nic;
2373
2374 /* protected by rtnl_lock */
2375 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2376
2377 u8 dev_addr_shadow[MAX_ADDR_LEN];
2378 netdevice_tracker linkwatch_dev_tracker;
2379 netdevice_tracker watchdog_dev_tracker;
2380 netdevice_tracker dev_registered_tracker;
2381 struct rtnl_hw_stats64 *offload_xstats_l3;
2382
2383 struct devlink_port *devlink_port;
2384};
2385#define to_net_dev(d) container_of(d, struct net_device, dev)
2386
2387/*
2388 * Driver should use this to assign devlink port instance to a netdevice
2389 * before it registers the netdevice. Therefore devlink_port is static
2390 * during the netdev lifetime after it is registered.
2391 */
2392#define SET_NETDEV_DEVLINK_PORT(dev, port) \
2393({ \
2394 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2395 ((dev)->devlink_port = (port)); \
2396})
2397
2398static inline bool netif_elide_gro(const struct net_device *dev)
2399{
2400 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2401 return true;
2402 return false;
2403}
2404
2405#define NETDEV_ALIGN 32
2406
2407static inline
2408int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2409{
2410 return dev->prio_tc_map[prio & TC_BITMASK];
2411}
2412
2413static inline
2414int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2415{
2416 if (tc >= dev->num_tc)
2417 return -EINVAL;
2418
2419 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2420 return 0;
2421}
2422
2423int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2424void netdev_reset_tc(struct net_device *dev);
2425int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2426int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2427
2428static inline
2429int netdev_get_num_tc(struct net_device *dev)
2430{
2431 return dev->num_tc;
2432}
2433
2434static inline void net_prefetch(void *p)
2435{
2436 prefetch(p);
2437#if L1_CACHE_BYTES < 128
2438 prefetch((u8 *)p + L1_CACHE_BYTES);
2439#endif
2440}
2441
2442static inline void net_prefetchw(void *p)
2443{
2444 prefetchw(p);
2445#if L1_CACHE_BYTES < 128
2446 prefetchw((u8 *)p + L1_CACHE_BYTES);
2447#endif
2448}
2449
2450void netdev_unbind_sb_channel(struct net_device *dev,
2451 struct net_device *sb_dev);
2452int netdev_bind_sb_channel_queue(struct net_device *dev,
2453 struct net_device *sb_dev,
2454 u8 tc, u16 count, u16 offset);
2455int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2456static inline int netdev_get_sb_channel(struct net_device *dev)
2457{
2458 return max_t(int, -dev->num_tc, 0);
2459}
2460
2461static inline
2462struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2463 unsigned int index)
2464{
2465 return &dev->_tx[index];
2466}
2467
2468static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2469 const struct sk_buff *skb)
2470{
2471 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2472}
2473
2474static inline void netdev_for_each_tx_queue(struct net_device *dev,
2475 void (*f)(struct net_device *,
2476 struct netdev_queue *,
2477 void *),
2478 void *arg)
2479{
2480 unsigned int i;
2481
2482 for (i = 0; i < dev->num_tx_queues; i++)
2483 f(dev, &dev->_tx[i], arg);
2484}
2485
2486#define netdev_lockdep_set_classes(dev) \
2487{ \
2488 static struct lock_class_key qdisc_tx_busylock_key; \
2489 static struct lock_class_key qdisc_xmit_lock_key; \
2490 static struct lock_class_key dev_addr_list_lock_key; \
2491 unsigned int i; \
2492 \
2493 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2494 lockdep_set_class(&(dev)->addr_list_lock, \
2495 &dev_addr_list_lock_key); \
2496 for (i = 0; i < (dev)->num_tx_queues; i++) \
2497 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2498 &qdisc_xmit_lock_key); \
2499}
2500
2501u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2502 struct net_device *sb_dev);
2503struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2504 struct sk_buff *skb,
2505 struct net_device *sb_dev);
2506
2507/* returns the headroom that the master device needs to take in account
2508 * when forwarding to this dev
2509 */
2510static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2511{
2512 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2513}
2514
2515static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2516{
2517 if (dev->netdev_ops->ndo_set_rx_headroom)
2518 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2519}
2520
2521/* set the device rx headroom to the dev's default */
2522static inline void netdev_reset_rx_headroom(struct net_device *dev)
2523{
2524 netdev_set_rx_headroom(dev, -1);
2525}
2526
2527static inline void *netdev_get_ml_priv(struct net_device *dev,
2528 enum netdev_ml_priv_type type)
2529{
2530 if (dev->ml_priv_type != type)
2531 return NULL;
2532
2533 return dev->ml_priv;
2534}
2535
2536static inline void netdev_set_ml_priv(struct net_device *dev,
2537 void *ml_priv,
2538 enum netdev_ml_priv_type type)
2539{
2540 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2541 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2542 dev->ml_priv_type, type);
2543 WARN(!dev->ml_priv_type && dev->ml_priv,
2544 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2545
2546 dev->ml_priv = ml_priv;
2547 dev->ml_priv_type = type;
2548}
2549
2550/*
2551 * Net namespace inlines
2552 */
2553static inline
2554struct net *dev_net(const struct net_device *dev)
2555{
2556 return read_pnet(&dev->nd_net);
2557}
2558
2559static inline
2560void dev_net_set(struct net_device *dev, struct net *net)
2561{
2562 write_pnet(&dev->nd_net, net);
2563}
2564
2565/**
2566 * netdev_priv - access network device private data
2567 * @dev: network device
2568 *
2569 * Get network device private data
2570 */
2571static inline void *netdev_priv(const struct net_device *dev)
2572{
2573 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2574}
2575
2576/* Set the sysfs physical device reference for the network logical device
2577 * if set prior to registration will cause a symlink during initialization.
2578 */
2579#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2580
2581/* Set the sysfs device type for the network logical device to allow
2582 * fine-grained identification of different network device types. For
2583 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2584 */
2585#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2586
2587/* Default NAPI poll() weight
2588 * Device drivers are strongly advised to not use bigger value
2589 */
2590#define NAPI_POLL_WEIGHT 64
2591
2592void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2593 int (*poll)(struct napi_struct *, int), int weight);
2594
2595/**
2596 * netif_napi_add() - initialize a NAPI context
2597 * @dev: network device
2598 * @napi: NAPI context
2599 * @poll: polling function
2600 *
2601 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2602 * *any* of the other NAPI-related functions.
2603 */
2604static inline void
2605netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2606 int (*poll)(struct napi_struct *, int))
2607{
2608 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2609}
2610
2611static inline void
2612netif_napi_add_tx_weight(struct net_device *dev,
2613 struct napi_struct *napi,
2614 int (*poll)(struct napi_struct *, int),
2615 int weight)
2616{
2617 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2618 netif_napi_add_weight(dev, napi, poll, weight);
2619}
2620
2621/**
2622 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2623 * @dev: network device
2624 * @napi: NAPI context
2625 * @poll: polling function
2626 *
2627 * This variant of netif_napi_add() should be used from drivers using NAPI
2628 * to exclusively poll a TX queue.
2629 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2630 */
2631static inline void netif_napi_add_tx(struct net_device *dev,
2632 struct napi_struct *napi,
2633 int (*poll)(struct napi_struct *, int))
2634{
2635 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2636}
2637
2638/**
2639 * __netif_napi_del - remove a NAPI context
2640 * @napi: NAPI context
2641 *
2642 * Warning: caller must observe RCU grace period before freeing memory
2643 * containing @napi. Drivers might want to call this helper to combine
2644 * all the needed RCU grace periods into a single one.
2645 */
2646void __netif_napi_del(struct napi_struct *napi);
2647
2648/**
2649 * netif_napi_del - remove a NAPI context
2650 * @napi: NAPI context
2651 *
2652 * netif_napi_del() removes a NAPI context from the network device NAPI list
2653 */
2654static inline void netif_napi_del(struct napi_struct *napi)
2655{
2656 __netif_napi_del(napi);
2657 synchronize_net();
2658}
2659
2660struct packet_type {
2661 __be16 type; /* This is really htons(ether_type). */
2662 bool ignore_outgoing;
2663 struct net_device *dev; /* NULL is wildcarded here */
2664 netdevice_tracker dev_tracker;
2665 int (*func) (struct sk_buff *,
2666 struct net_device *,
2667 struct packet_type *,
2668 struct net_device *);
2669 void (*list_func) (struct list_head *,
2670 struct packet_type *,
2671 struct net_device *);
2672 bool (*id_match)(struct packet_type *ptype,
2673 struct sock *sk);
2674 struct net *af_packet_net;
2675 void *af_packet_priv;
2676 struct list_head list;
2677};
2678
2679struct offload_callbacks {
2680 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2681 netdev_features_t features);
2682 struct sk_buff *(*gro_receive)(struct list_head *head,
2683 struct sk_buff *skb);
2684 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2685};
2686
2687struct packet_offload {
2688 __be16 type; /* This is really htons(ether_type). */
2689 u16 priority;
2690 struct offload_callbacks callbacks;
2691 struct list_head list;
2692};
2693
2694/* often modified stats are per-CPU, other are shared (netdev->stats) */
2695struct pcpu_sw_netstats {
2696 u64_stats_t rx_packets;
2697 u64_stats_t rx_bytes;
2698 u64_stats_t tx_packets;
2699 u64_stats_t tx_bytes;
2700 struct u64_stats_sync syncp;
2701} __aligned(4 * sizeof(u64));
2702
2703struct pcpu_lstats {
2704 u64_stats_t packets;
2705 u64_stats_t bytes;
2706 struct u64_stats_sync syncp;
2707} __aligned(2 * sizeof(u64));
2708
2709void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2710
2711static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2712{
2713 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2714
2715 u64_stats_update_begin(&tstats->syncp);
2716 u64_stats_add(&tstats->rx_bytes, len);
2717 u64_stats_inc(&tstats->rx_packets);
2718 u64_stats_update_end(&tstats->syncp);
2719}
2720
2721static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2722 unsigned int packets,
2723 unsigned int len)
2724{
2725 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2726
2727 u64_stats_update_begin(&tstats->syncp);
2728 u64_stats_add(&tstats->tx_bytes, len);
2729 u64_stats_add(&tstats->tx_packets, packets);
2730 u64_stats_update_end(&tstats->syncp);
2731}
2732
2733static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2734{
2735 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2736
2737 u64_stats_update_begin(&lstats->syncp);
2738 u64_stats_add(&lstats->bytes, len);
2739 u64_stats_inc(&lstats->packets);
2740 u64_stats_update_end(&lstats->syncp);
2741}
2742
2743#define __netdev_alloc_pcpu_stats(type, gfp) \
2744({ \
2745 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2746 if (pcpu_stats) { \
2747 int __cpu; \
2748 for_each_possible_cpu(__cpu) { \
2749 typeof(type) *stat; \
2750 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2751 u64_stats_init(&stat->syncp); \
2752 } \
2753 } \
2754 pcpu_stats; \
2755})
2756
2757#define netdev_alloc_pcpu_stats(type) \
2758 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2759
2760#define devm_netdev_alloc_pcpu_stats(dev, type) \
2761({ \
2762 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2763 if (pcpu_stats) { \
2764 int __cpu; \
2765 for_each_possible_cpu(__cpu) { \
2766 typeof(type) *stat; \
2767 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2768 u64_stats_init(&stat->syncp); \
2769 } \
2770 } \
2771 pcpu_stats; \
2772})
2773
2774enum netdev_lag_tx_type {
2775 NETDEV_LAG_TX_TYPE_UNKNOWN,
2776 NETDEV_LAG_TX_TYPE_RANDOM,
2777 NETDEV_LAG_TX_TYPE_BROADCAST,
2778 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2779 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2780 NETDEV_LAG_TX_TYPE_HASH,
2781};
2782
2783enum netdev_lag_hash {
2784 NETDEV_LAG_HASH_NONE,
2785 NETDEV_LAG_HASH_L2,
2786 NETDEV_LAG_HASH_L34,
2787 NETDEV_LAG_HASH_L23,
2788 NETDEV_LAG_HASH_E23,
2789 NETDEV_LAG_HASH_E34,
2790 NETDEV_LAG_HASH_VLAN_SRCMAC,
2791 NETDEV_LAG_HASH_UNKNOWN,
2792};
2793
2794struct netdev_lag_upper_info {
2795 enum netdev_lag_tx_type tx_type;
2796 enum netdev_lag_hash hash_type;
2797};
2798
2799struct netdev_lag_lower_state_info {
2800 u8 link_up : 1,
2801 tx_enabled : 1;
2802};
2803
2804#include <linux/notifier.h>
2805
2806/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2807 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2808 * adding new types.
2809 */
2810enum netdev_cmd {
2811 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2812 NETDEV_DOWN,
2813 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2814 detected a hardware crash and restarted
2815 - we can use this eg to kick tcp sessions
2816 once done */
2817 NETDEV_CHANGE, /* Notify device state change */
2818 NETDEV_REGISTER,
2819 NETDEV_UNREGISTER,
2820 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2821 NETDEV_CHANGEADDR, /* notify after the address change */
2822 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2823 NETDEV_GOING_DOWN,
2824 NETDEV_CHANGENAME,
2825 NETDEV_FEAT_CHANGE,
2826 NETDEV_BONDING_FAILOVER,
2827 NETDEV_PRE_UP,
2828 NETDEV_PRE_TYPE_CHANGE,
2829 NETDEV_POST_TYPE_CHANGE,
2830 NETDEV_POST_INIT,
2831 NETDEV_PRE_UNINIT,
2832 NETDEV_RELEASE,
2833 NETDEV_NOTIFY_PEERS,
2834 NETDEV_JOIN,
2835 NETDEV_CHANGEUPPER,
2836 NETDEV_RESEND_IGMP,
2837 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2838 NETDEV_CHANGEINFODATA,
2839 NETDEV_BONDING_INFO,
2840 NETDEV_PRECHANGEUPPER,
2841 NETDEV_CHANGELOWERSTATE,
2842 NETDEV_UDP_TUNNEL_PUSH_INFO,
2843 NETDEV_UDP_TUNNEL_DROP_INFO,
2844 NETDEV_CHANGE_TX_QUEUE_LEN,
2845 NETDEV_CVLAN_FILTER_PUSH_INFO,
2846 NETDEV_CVLAN_FILTER_DROP_INFO,
2847 NETDEV_SVLAN_FILTER_PUSH_INFO,
2848 NETDEV_SVLAN_FILTER_DROP_INFO,
2849 NETDEV_OFFLOAD_XSTATS_ENABLE,
2850 NETDEV_OFFLOAD_XSTATS_DISABLE,
2851 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2852 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2853 NETDEV_XDP_FEAT_CHANGE,
2854};
2855const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2856
2857int register_netdevice_notifier(struct notifier_block *nb);
2858int unregister_netdevice_notifier(struct notifier_block *nb);
2859int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2860int unregister_netdevice_notifier_net(struct net *net,
2861 struct notifier_block *nb);
2862int register_netdevice_notifier_dev_net(struct net_device *dev,
2863 struct notifier_block *nb,
2864 struct netdev_net_notifier *nn);
2865int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2866 struct notifier_block *nb,
2867 struct netdev_net_notifier *nn);
2868
2869struct netdev_notifier_info {
2870 struct net_device *dev;
2871 struct netlink_ext_ack *extack;
2872};
2873
2874struct netdev_notifier_info_ext {
2875 struct netdev_notifier_info info; /* must be first */
2876 union {
2877 u32 mtu;
2878 } ext;
2879};
2880
2881struct netdev_notifier_change_info {
2882 struct netdev_notifier_info info; /* must be first */
2883 unsigned int flags_changed;
2884};
2885
2886struct netdev_notifier_changeupper_info {
2887 struct netdev_notifier_info info; /* must be first */
2888 struct net_device *upper_dev; /* new upper dev */
2889 bool master; /* is upper dev master */
2890 bool linking; /* is the notification for link or unlink */
2891 void *upper_info; /* upper dev info */
2892};
2893
2894struct netdev_notifier_changelowerstate_info {
2895 struct netdev_notifier_info info; /* must be first */
2896 void *lower_state_info; /* is lower dev state */
2897};
2898
2899struct netdev_notifier_pre_changeaddr_info {
2900 struct netdev_notifier_info info; /* must be first */
2901 const unsigned char *dev_addr;
2902};
2903
2904enum netdev_offload_xstats_type {
2905 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2906};
2907
2908struct netdev_notifier_offload_xstats_info {
2909 struct netdev_notifier_info info; /* must be first */
2910 enum netdev_offload_xstats_type type;
2911
2912 union {
2913 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2914 struct netdev_notifier_offload_xstats_rd *report_delta;
2915 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2916 struct netdev_notifier_offload_xstats_ru *report_used;
2917 };
2918};
2919
2920int netdev_offload_xstats_enable(struct net_device *dev,
2921 enum netdev_offload_xstats_type type,
2922 struct netlink_ext_ack *extack);
2923int netdev_offload_xstats_disable(struct net_device *dev,
2924 enum netdev_offload_xstats_type type);
2925bool netdev_offload_xstats_enabled(const struct net_device *dev,
2926 enum netdev_offload_xstats_type type);
2927int netdev_offload_xstats_get(struct net_device *dev,
2928 enum netdev_offload_xstats_type type,
2929 struct rtnl_hw_stats64 *stats, bool *used,
2930 struct netlink_ext_ack *extack);
2931void
2932netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2933 const struct rtnl_hw_stats64 *stats);
2934void
2935netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2936void netdev_offload_xstats_push_delta(struct net_device *dev,
2937 enum netdev_offload_xstats_type type,
2938 const struct rtnl_hw_stats64 *stats);
2939
2940static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2941 struct net_device *dev)
2942{
2943 info->dev = dev;
2944 info->extack = NULL;
2945}
2946
2947static inline struct net_device *
2948netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2949{
2950 return info->dev;
2951}
2952
2953static inline struct netlink_ext_ack *
2954netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2955{
2956 return info->extack;
2957}
2958
2959int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2960
2961
2962extern rwlock_t dev_base_lock; /* Device list lock */
2963
2964#define for_each_netdev(net, d) \
2965 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2966#define for_each_netdev_reverse(net, d) \
2967 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2968#define for_each_netdev_rcu(net, d) \
2969 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2970#define for_each_netdev_safe(net, d, n) \
2971 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2972#define for_each_netdev_continue(net, d) \
2973 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2974#define for_each_netdev_continue_reverse(net, d) \
2975 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2976 dev_list)
2977#define for_each_netdev_continue_rcu(net, d) \
2978 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2979#define for_each_netdev_in_bond_rcu(bond, slave) \
2980 for_each_netdev_rcu(&init_net, slave) \
2981 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2982#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2983
2984static inline struct net_device *next_net_device(struct net_device *dev)
2985{
2986 struct list_head *lh;
2987 struct net *net;
2988
2989 net = dev_net(dev);
2990 lh = dev->dev_list.next;
2991 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2992}
2993
2994static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2995{
2996 struct list_head *lh;
2997 struct net *net;
2998
2999 net = dev_net(dev);
3000 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3001 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3002}
3003
3004static inline struct net_device *first_net_device(struct net *net)
3005{
3006 return list_empty(&net->dev_base_head) ? NULL :
3007 net_device_entry(net->dev_base_head.next);
3008}
3009
3010static inline struct net_device *first_net_device_rcu(struct net *net)
3011{
3012 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3013
3014 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3015}
3016
3017int netdev_boot_setup_check(struct net_device *dev);
3018struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3019 const char *hwaddr);
3020struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3021void dev_add_pack(struct packet_type *pt);
3022void dev_remove_pack(struct packet_type *pt);
3023void __dev_remove_pack(struct packet_type *pt);
3024void dev_add_offload(struct packet_offload *po);
3025void dev_remove_offload(struct packet_offload *po);
3026
3027int dev_get_iflink(const struct net_device *dev);
3028int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3029int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3030 struct net_device_path_stack *stack);
3031struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3032 unsigned short mask);
3033struct net_device *dev_get_by_name(struct net *net, const char *name);
3034struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3035struct net_device *__dev_get_by_name(struct net *net, const char *name);
3036bool netdev_name_in_use(struct net *net, const char *name);
3037int dev_alloc_name(struct net_device *dev, const char *name);
3038int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3039void dev_close(struct net_device *dev);
3040void dev_close_many(struct list_head *head, bool unlink);
3041void dev_disable_lro(struct net_device *dev);
3042int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3043u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3044 struct net_device *sb_dev);
3045u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3046 struct net_device *sb_dev);
3047
3048int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3049int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3050
3051static inline int dev_queue_xmit(struct sk_buff *skb)
3052{
3053 return __dev_queue_xmit(skb, NULL);
3054}
3055
3056static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3057 struct net_device *sb_dev)
3058{
3059 return __dev_queue_xmit(skb, sb_dev);
3060}
3061
3062static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3063{
3064 int ret;
3065
3066 ret = __dev_direct_xmit(skb, queue_id);
3067 if (!dev_xmit_complete(ret))
3068 kfree_skb(skb);
3069 return ret;
3070}
3071
3072int register_netdevice(struct net_device *dev);
3073void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3074void unregister_netdevice_many(struct list_head *head);
3075static inline void unregister_netdevice(struct net_device *dev)
3076{
3077 unregister_netdevice_queue(dev, NULL);
3078}
3079
3080int netdev_refcnt_read(const struct net_device *dev);
3081void free_netdev(struct net_device *dev);
3082void netdev_freemem(struct net_device *dev);
3083int init_dummy_netdev(struct net_device *dev);
3084
3085struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3086 struct sk_buff *skb,
3087 bool all_slaves);
3088struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3089 struct sock *sk);
3090struct net_device *dev_get_by_index(struct net *net, int ifindex);
3091struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3092struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3093struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3094int dev_restart(struct net_device *dev);
3095
3096
3097static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3098 unsigned short type,
3099 const void *daddr, const void *saddr,
3100 unsigned int len)
3101{
3102 if (!dev->header_ops || !dev->header_ops->create)
3103 return 0;
3104
3105 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3106}
3107
3108static inline int dev_parse_header(const struct sk_buff *skb,
3109 unsigned char *haddr)
3110{
3111 const struct net_device *dev = skb->dev;
3112
3113 if (!dev->header_ops || !dev->header_ops->parse)
3114 return 0;
3115 return dev->header_ops->parse(skb, haddr);
3116}
3117
3118static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3119{
3120 const struct net_device *dev = skb->dev;
3121
3122 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3123 return 0;
3124 return dev->header_ops->parse_protocol(skb);
3125}
3126
3127/* ll_header must have at least hard_header_len allocated */
3128static inline bool dev_validate_header(const struct net_device *dev,
3129 char *ll_header, int len)
3130{
3131 if (likely(len >= dev->hard_header_len))
3132 return true;
3133 if (len < dev->min_header_len)
3134 return false;
3135
3136 if (capable(CAP_SYS_RAWIO)) {
3137 memset(ll_header + len, 0, dev->hard_header_len - len);
3138 return true;
3139 }
3140
3141 if (dev->header_ops && dev->header_ops->validate)
3142 return dev->header_ops->validate(ll_header, len);
3143
3144 return false;
3145}
3146
3147static inline bool dev_has_header(const struct net_device *dev)
3148{
3149 return dev->header_ops && dev->header_ops->create;
3150}
3151
3152/*
3153 * Incoming packets are placed on per-CPU queues
3154 */
3155struct softnet_data {
3156 struct list_head poll_list;
3157 struct sk_buff_head process_queue;
3158
3159 /* stats */
3160 unsigned int processed;
3161 unsigned int time_squeeze;
3162#ifdef CONFIG_RPS
3163 struct softnet_data *rps_ipi_list;
3164#endif
3165#ifdef CONFIG_NET_FLOW_LIMIT
3166 struct sd_flow_limit __rcu *flow_limit;
3167#endif
3168 struct Qdisc *output_queue;
3169 struct Qdisc **output_queue_tailp;
3170 struct sk_buff *completion_queue;
3171#ifdef CONFIG_XFRM_OFFLOAD
3172 struct sk_buff_head xfrm_backlog;
3173#endif
3174 /* written and read only by owning cpu: */
3175 struct {
3176 u16 recursion;
3177 u8 more;
3178#ifdef CONFIG_NET_EGRESS
3179 u8 skip_txqueue;
3180#endif
3181 } xmit;
3182#ifdef CONFIG_RPS
3183 /* input_queue_head should be written by cpu owning this struct,
3184 * and only read by other cpus. Worth using a cache line.
3185 */
3186 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3187
3188 /* Elements below can be accessed between CPUs for RPS/RFS */
3189 call_single_data_t csd ____cacheline_aligned_in_smp;
3190 struct softnet_data *rps_ipi_next;
3191 unsigned int cpu;
3192 unsigned int input_queue_tail;
3193#endif
3194 unsigned int received_rps;
3195 unsigned int dropped;
3196 struct sk_buff_head input_pkt_queue;
3197 struct napi_struct backlog;
3198
3199 /* Another possibly contended cache line */
3200 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3201 int defer_count;
3202 int defer_ipi_scheduled;
3203 struct sk_buff *defer_list;
3204 call_single_data_t defer_csd;
3205};
3206
3207static inline void input_queue_head_incr(struct softnet_data *sd)
3208{
3209#ifdef CONFIG_RPS
3210 sd->input_queue_head++;
3211#endif
3212}
3213
3214static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3215 unsigned int *qtail)
3216{
3217#ifdef CONFIG_RPS
3218 *qtail = ++sd->input_queue_tail;
3219#endif
3220}
3221
3222DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3223
3224static inline int dev_recursion_level(void)
3225{
3226 return this_cpu_read(softnet_data.xmit.recursion);
3227}
3228
3229#define XMIT_RECURSION_LIMIT 8
3230static inline bool dev_xmit_recursion(void)
3231{
3232 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3233 XMIT_RECURSION_LIMIT);
3234}
3235
3236static inline void dev_xmit_recursion_inc(void)
3237{
3238 __this_cpu_inc(softnet_data.xmit.recursion);
3239}
3240
3241static inline void dev_xmit_recursion_dec(void)
3242{
3243 __this_cpu_dec(softnet_data.xmit.recursion);
3244}
3245
3246void __netif_schedule(struct Qdisc *q);
3247void netif_schedule_queue(struct netdev_queue *txq);
3248
3249static inline void netif_tx_schedule_all(struct net_device *dev)
3250{
3251 unsigned int i;
3252
3253 for (i = 0; i < dev->num_tx_queues; i++)
3254 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3255}
3256
3257static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3258{
3259 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3260}
3261
3262/**
3263 * netif_start_queue - allow transmit
3264 * @dev: network device
3265 *
3266 * Allow upper layers to call the device hard_start_xmit routine.
3267 */
3268static inline void netif_start_queue(struct net_device *dev)
3269{
3270 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3271}
3272
3273static inline void netif_tx_start_all_queues(struct net_device *dev)
3274{
3275 unsigned int i;
3276
3277 for (i = 0; i < dev->num_tx_queues; i++) {
3278 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3279 netif_tx_start_queue(txq);
3280 }
3281}
3282
3283void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3284
3285/**
3286 * netif_wake_queue - restart transmit
3287 * @dev: network device
3288 *
3289 * Allow upper layers to call the device hard_start_xmit routine.
3290 * Used for flow control when transmit resources are available.
3291 */
3292static inline void netif_wake_queue(struct net_device *dev)
3293{
3294 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3295}
3296
3297static inline void netif_tx_wake_all_queues(struct net_device *dev)
3298{
3299 unsigned int i;
3300
3301 for (i = 0; i < dev->num_tx_queues; i++) {
3302 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3303 netif_tx_wake_queue(txq);
3304 }
3305}
3306
3307static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3308{
3309 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3310}
3311
3312/**
3313 * netif_stop_queue - stop transmitted packets
3314 * @dev: network device
3315 *
3316 * Stop upper layers calling the device hard_start_xmit routine.
3317 * Used for flow control when transmit resources are unavailable.
3318 */
3319static inline void netif_stop_queue(struct net_device *dev)
3320{
3321 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3322}
3323
3324void netif_tx_stop_all_queues(struct net_device *dev);
3325
3326static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3327{
3328 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3329}
3330
3331/**
3332 * netif_queue_stopped - test if transmit queue is flowblocked
3333 * @dev: network device
3334 *
3335 * Test if transmit queue on device is currently unable to send.
3336 */
3337static inline bool netif_queue_stopped(const struct net_device *dev)
3338{
3339 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3340}
3341
3342static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3343{
3344 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3345}
3346
3347static inline bool
3348netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3349{
3350 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3351}
3352
3353static inline bool
3354netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3355{
3356 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3357}
3358
3359/**
3360 * netdev_queue_set_dql_min_limit - set dql minimum limit
3361 * @dev_queue: pointer to transmit queue
3362 * @min_limit: dql minimum limit
3363 *
3364 * Forces xmit_more() to return true until the minimum threshold
3365 * defined by @min_limit is reached (or until the tx queue is
3366 * empty). Warning: to be use with care, misuse will impact the
3367 * latency.
3368 */
3369static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3370 unsigned int min_limit)
3371{
3372#ifdef CONFIG_BQL
3373 dev_queue->dql.min_limit = min_limit;
3374#endif
3375}
3376
3377/**
3378 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3379 * @dev_queue: pointer to transmit queue
3380 *
3381 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3382 * to give appropriate hint to the CPU.
3383 */
3384static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3385{
3386#ifdef CONFIG_BQL
3387 prefetchw(&dev_queue->dql.num_queued);
3388#endif
3389}
3390
3391/**
3392 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3393 * @dev_queue: pointer to transmit queue
3394 *
3395 * BQL enabled drivers might use this helper in their TX completion path,
3396 * to give appropriate hint to the CPU.
3397 */
3398static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3399{
3400#ifdef CONFIG_BQL
3401 prefetchw(&dev_queue->dql.limit);
3402#endif
3403}
3404
3405/**
3406 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3407 * @dev_queue: network device queue
3408 * @bytes: number of bytes queued to the device queue
3409 *
3410 * Report the number of bytes queued for sending/completion to the network
3411 * device hardware queue. @bytes should be a good approximation and should
3412 * exactly match netdev_completed_queue() @bytes.
3413 * This is typically called once per packet, from ndo_start_xmit().
3414 */
3415static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3416 unsigned int bytes)
3417{
3418#ifdef CONFIG_BQL
3419 dql_queued(&dev_queue->dql, bytes);
3420
3421 if (likely(dql_avail(&dev_queue->dql) >= 0))
3422 return;
3423
3424 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3425
3426 /*
3427 * The XOFF flag must be set before checking the dql_avail below,
3428 * because in netdev_tx_completed_queue we update the dql_completed
3429 * before checking the XOFF flag.
3430 */
3431 smp_mb();
3432
3433 /* check again in case another CPU has just made room avail */
3434 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3435 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3436#endif
3437}
3438
3439/* Variant of netdev_tx_sent_queue() for drivers that are aware
3440 * that they should not test BQL status themselves.
3441 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3442 * skb of a batch.
3443 * Returns true if the doorbell must be used to kick the NIC.
3444 */
3445static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3446 unsigned int bytes,
3447 bool xmit_more)
3448{
3449 if (xmit_more) {
3450#ifdef CONFIG_BQL
3451 dql_queued(&dev_queue->dql, bytes);
3452#endif
3453 return netif_tx_queue_stopped(dev_queue);
3454 }
3455 netdev_tx_sent_queue(dev_queue, bytes);
3456 return true;
3457}
3458
3459/**
3460 * netdev_sent_queue - report the number of bytes queued to hardware
3461 * @dev: network device
3462 * @bytes: number of bytes queued to the hardware device queue
3463 *
3464 * Report the number of bytes queued for sending/completion to the network
3465 * device hardware queue#0. @bytes should be a good approximation and should
3466 * exactly match netdev_completed_queue() @bytes.
3467 * This is typically called once per packet, from ndo_start_xmit().
3468 */
3469static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3470{
3471 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3472}
3473
3474static inline bool __netdev_sent_queue(struct net_device *dev,
3475 unsigned int bytes,
3476 bool xmit_more)
3477{
3478 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3479 xmit_more);
3480}
3481
3482/**
3483 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3484 * @dev_queue: network device queue
3485 * @pkts: number of packets (currently ignored)
3486 * @bytes: number of bytes dequeued from the device queue
3487 *
3488 * Must be called at most once per TX completion round (and not per
3489 * individual packet), so that BQL can adjust its limits appropriately.
3490 */
3491static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3492 unsigned int pkts, unsigned int bytes)
3493{
3494#ifdef CONFIG_BQL
3495 if (unlikely(!bytes))
3496 return;
3497
3498 dql_completed(&dev_queue->dql, bytes);
3499
3500 /*
3501 * Without the memory barrier there is a small possiblity that
3502 * netdev_tx_sent_queue will miss the update and cause the queue to
3503 * be stopped forever
3504 */
3505 smp_mb();
3506
3507 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3508 return;
3509
3510 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3511 netif_schedule_queue(dev_queue);
3512#endif
3513}
3514
3515/**
3516 * netdev_completed_queue - report bytes and packets completed by device
3517 * @dev: network device
3518 * @pkts: actual number of packets sent over the medium
3519 * @bytes: actual number of bytes sent over the medium
3520 *
3521 * Report the number of bytes and packets transmitted by the network device
3522 * hardware queue over the physical medium, @bytes must exactly match the
3523 * @bytes amount passed to netdev_sent_queue()
3524 */
3525static inline void netdev_completed_queue(struct net_device *dev,
3526 unsigned int pkts, unsigned int bytes)
3527{
3528 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3529}
3530
3531static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3532{
3533#ifdef CONFIG_BQL
3534 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3535 dql_reset(&q->dql);
3536#endif
3537}
3538
3539/**
3540 * netdev_reset_queue - reset the packets and bytes count of a network device
3541 * @dev_queue: network device
3542 *
3543 * Reset the bytes and packet count of a network device and clear the
3544 * software flow control OFF bit for this network device
3545 */
3546static inline void netdev_reset_queue(struct net_device *dev_queue)
3547{
3548 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3549}
3550
3551/**
3552 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3553 * @dev: network device
3554 * @queue_index: given tx queue index
3555 *
3556 * Returns 0 if given tx queue index >= number of device tx queues,
3557 * otherwise returns the originally passed tx queue index.
3558 */
3559static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3560{
3561 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3562 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3563 dev->name, queue_index,
3564 dev->real_num_tx_queues);
3565 return 0;
3566 }
3567
3568 return queue_index;
3569}
3570
3571/**
3572 * netif_running - test if up
3573 * @dev: network device
3574 *
3575 * Test if the device has been brought up.
3576 */
3577static inline bool netif_running(const struct net_device *dev)
3578{
3579 return test_bit(__LINK_STATE_START, &dev->state);
3580}
3581
3582/*
3583 * Routines to manage the subqueues on a device. We only need start,
3584 * stop, and a check if it's stopped. All other device management is
3585 * done at the overall netdevice level.
3586 * Also test the device if we're multiqueue.
3587 */
3588
3589/**
3590 * netif_start_subqueue - allow sending packets on subqueue
3591 * @dev: network device
3592 * @queue_index: sub queue index
3593 *
3594 * Start individual transmit queue of a device with multiple transmit queues.
3595 */
3596static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3597{
3598 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3599
3600 netif_tx_start_queue(txq);
3601}
3602
3603/**
3604 * netif_stop_subqueue - stop sending packets on subqueue
3605 * @dev: network device
3606 * @queue_index: sub queue index
3607 *
3608 * Stop individual transmit queue of a device with multiple transmit queues.
3609 */
3610static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3611{
3612 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3613 netif_tx_stop_queue(txq);
3614}
3615
3616/**
3617 * __netif_subqueue_stopped - test status of subqueue
3618 * @dev: network device
3619 * @queue_index: sub queue index
3620 *
3621 * Check individual transmit queue of a device with multiple transmit queues.
3622 */
3623static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3624 u16 queue_index)
3625{
3626 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3627
3628 return netif_tx_queue_stopped(txq);
3629}
3630
3631/**
3632 * netif_subqueue_stopped - test status of subqueue
3633 * @dev: network device
3634 * @skb: sub queue buffer pointer
3635 *
3636 * Check individual transmit queue of a device with multiple transmit queues.
3637 */
3638static inline bool netif_subqueue_stopped(const struct net_device *dev,
3639 struct sk_buff *skb)
3640{
3641 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3642}
3643
3644/**
3645 * netif_wake_subqueue - allow sending packets on subqueue
3646 * @dev: network device
3647 * @queue_index: sub queue index
3648 *
3649 * Resume individual transmit queue of a device with multiple transmit queues.
3650 */
3651static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3652{
3653 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3654
3655 netif_tx_wake_queue(txq);
3656}
3657
3658#ifdef CONFIG_XPS
3659int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3660 u16 index);
3661int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3662 u16 index, enum xps_map_type type);
3663
3664/**
3665 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3666 * @j: CPU/Rx queue index
3667 * @mask: bitmask of all cpus/rx queues
3668 * @nr_bits: number of bits in the bitmask
3669 *
3670 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3671 */
3672static inline bool netif_attr_test_mask(unsigned long j,
3673 const unsigned long *mask,
3674 unsigned int nr_bits)
3675{
3676 cpu_max_bits_warn(j, nr_bits);
3677 return test_bit(j, mask);
3678}
3679
3680/**
3681 * netif_attr_test_online - Test for online CPU/Rx queue
3682 * @j: CPU/Rx queue index
3683 * @online_mask: bitmask for CPUs/Rx queues that are online
3684 * @nr_bits: number of bits in the bitmask
3685 *
3686 * Returns true if a CPU/Rx queue is online.
3687 */
3688static inline bool netif_attr_test_online(unsigned long j,
3689 const unsigned long *online_mask,
3690 unsigned int nr_bits)
3691{
3692 cpu_max_bits_warn(j, nr_bits);
3693
3694 if (online_mask)
3695 return test_bit(j, online_mask);
3696
3697 return (j < nr_bits);
3698}
3699
3700/**
3701 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3702 * @n: CPU/Rx queue index
3703 * @srcp: the cpumask/Rx queue mask pointer
3704 * @nr_bits: number of bits in the bitmask
3705 *
3706 * Returns >= nr_bits if no further CPUs/Rx queues set.
3707 */
3708static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3709 unsigned int nr_bits)
3710{
3711 /* -1 is a legal arg here. */
3712 if (n != -1)
3713 cpu_max_bits_warn(n, nr_bits);
3714
3715 if (srcp)
3716 return find_next_bit(srcp, nr_bits, n + 1);
3717
3718 return n + 1;
3719}
3720
3721/**
3722 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3723 * @n: CPU/Rx queue index
3724 * @src1p: the first CPUs/Rx queues mask pointer
3725 * @src2p: the second CPUs/Rx queues mask pointer
3726 * @nr_bits: number of bits in the bitmask
3727 *
3728 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3729 */
3730static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3731 const unsigned long *src2p,
3732 unsigned int nr_bits)
3733{
3734 /* -1 is a legal arg here. */
3735 if (n != -1)
3736 cpu_max_bits_warn(n, nr_bits);
3737
3738 if (src1p && src2p)
3739 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3740 else if (src1p)
3741 return find_next_bit(src1p, nr_bits, n + 1);
3742 else if (src2p)
3743 return find_next_bit(src2p, nr_bits, n + 1);
3744
3745 return n + 1;
3746}
3747#else
3748static inline int netif_set_xps_queue(struct net_device *dev,
3749 const struct cpumask *mask,
3750 u16 index)
3751{
3752 return 0;
3753}
3754
3755static inline int __netif_set_xps_queue(struct net_device *dev,
3756 const unsigned long *mask,
3757 u16 index, enum xps_map_type type)
3758{
3759 return 0;
3760}
3761#endif
3762
3763/**
3764 * netif_is_multiqueue - test if device has multiple transmit queues
3765 * @dev: network device
3766 *
3767 * Check if device has multiple transmit queues
3768 */
3769static inline bool netif_is_multiqueue(const struct net_device *dev)
3770{
3771 return dev->num_tx_queues > 1;
3772}
3773
3774int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3775
3776#ifdef CONFIG_SYSFS
3777int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3778#else
3779static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3780 unsigned int rxqs)
3781{
3782 dev->real_num_rx_queues = rxqs;
3783 return 0;
3784}
3785#endif
3786int netif_set_real_num_queues(struct net_device *dev,
3787 unsigned int txq, unsigned int rxq);
3788
3789static inline struct netdev_rx_queue *
3790__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3791{
3792 return dev->_rx + rxq;
3793}
3794
3795#ifdef CONFIG_SYSFS
3796static inline unsigned int get_netdev_rx_queue_index(
3797 struct netdev_rx_queue *queue)
3798{
3799 struct net_device *dev = queue->dev;
3800 int index = queue - dev->_rx;
3801
3802 BUG_ON(index >= dev->num_rx_queues);
3803 return index;
3804}
3805#endif
3806
3807int netif_get_num_default_rss_queues(void);
3808
3809enum skb_free_reason {
3810 SKB_REASON_CONSUMED,
3811 SKB_REASON_DROPPED,
3812};
3813
3814void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3815void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3816
3817/*
3818 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3819 * interrupt context or with hardware interrupts being disabled.
3820 * (in_hardirq() || irqs_disabled())
3821 *
3822 * We provide four helpers that can be used in following contexts :
3823 *
3824 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3825 * replacing kfree_skb(skb)
3826 *
3827 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3828 * Typically used in place of consume_skb(skb) in TX completion path
3829 *
3830 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3831 * replacing kfree_skb(skb)
3832 *
3833 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3834 * and consumed a packet. Used in place of consume_skb(skb)
3835 */
3836static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3837{
3838 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3839}
3840
3841static inline void dev_consume_skb_irq(struct sk_buff *skb)
3842{
3843 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3844}
3845
3846static inline void dev_kfree_skb_any(struct sk_buff *skb)
3847{
3848 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3849}
3850
3851static inline void dev_consume_skb_any(struct sk_buff *skb)
3852{
3853 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3854}
3855
3856u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3857 struct bpf_prog *xdp_prog);
3858void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3859int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3860int netif_rx(struct sk_buff *skb);
3861int __netif_rx(struct sk_buff *skb);
3862
3863int netif_receive_skb(struct sk_buff *skb);
3864int netif_receive_skb_core(struct sk_buff *skb);
3865void netif_receive_skb_list_internal(struct list_head *head);
3866void netif_receive_skb_list(struct list_head *head);
3867gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3868void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3869struct sk_buff *napi_get_frags(struct napi_struct *napi);
3870void napi_get_frags_check(struct napi_struct *napi);
3871gro_result_t napi_gro_frags(struct napi_struct *napi);
3872struct packet_offload *gro_find_receive_by_type(__be16 type);
3873struct packet_offload *gro_find_complete_by_type(__be16 type);
3874
3875static inline void napi_free_frags(struct napi_struct *napi)
3876{
3877 kfree_skb(napi->skb);
3878 napi->skb = NULL;
3879}
3880
3881bool netdev_is_rx_handler_busy(struct net_device *dev);
3882int netdev_rx_handler_register(struct net_device *dev,
3883 rx_handler_func_t *rx_handler,
3884 void *rx_handler_data);
3885void netdev_rx_handler_unregister(struct net_device *dev);
3886
3887bool dev_valid_name(const char *name);
3888static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3889{
3890 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3891}
3892int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3893int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3894int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3895 void __user *data, bool *need_copyout);
3896int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3897int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3898unsigned int dev_get_flags(const struct net_device *);
3899int __dev_change_flags(struct net_device *dev, unsigned int flags,
3900 struct netlink_ext_ack *extack);
3901int dev_change_flags(struct net_device *dev, unsigned int flags,
3902 struct netlink_ext_ack *extack);
3903int dev_set_alias(struct net_device *, const char *, size_t);
3904int dev_get_alias(const struct net_device *, char *, size_t);
3905int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3906 const char *pat, int new_ifindex);
3907static inline
3908int dev_change_net_namespace(struct net_device *dev, struct net *net,
3909 const char *pat)
3910{
3911 return __dev_change_net_namespace(dev, net, pat, 0);
3912}
3913int __dev_set_mtu(struct net_device *, int);
3914int dev_set_mtu(struct net_device *, int);
3915int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3916 struct netlink_ext_ack *extack);
3917int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3918 struct netlink_ext_ack *extack);
3919int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3920 struct netlink_ext_ack *extack);
3921int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3922int dev_get_port_parent_id(struct net_device *dev,
3923 struct netdev_phys_item_id *ppid, bool recurse);
3924bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3925struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3926struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3927 struct netdev_queue *txq, int *ret);
3928
3929int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3930u8 dev_xdp_prog_count(struct net_device *dev);
3931u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3932
3933int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3934int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3935int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3936bool is_skb_forwardable(const struct net_device *dev,
3937 const struct sk_buff *skb);
3938
3939static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3940 const struct sk_buff *skb,
3941 const bool check_mtu)
3942{
3943 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3944 unsigned int len;
3945
3946 if (!(dev->flags & IFF_UP))
3947 return false;
3948
3949 if (!check_mtu)
3950 return true;
3951
3952 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3953 if (skb->len <= len)
3954 return true;
3955
3956 /* if TSO is enabled, we don't care about the length as the packet
3957 * could be forwarded without being segmented before
3958 */
3959 if (skb_is_gso(skb))
3960 return true;
3961
3962 return false;
3963}
3964
3965struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
3966
3967static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
3968{
3969 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
3970 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
3971
3972 if (likely(p))
3973 return p;
3974
3975 return netdev_core_stats_alloc(dev);
3976}
3977
3978#define DEV_CORE_STATS_INC(FIELD) \
3979static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3980{ \
3981 struct net_device_core_stats __percpu *p; \
3982 \
3983 p = dev_core_stats(dev); \
3984 if (p) \
3985 this_cpu_inc(p->FIELD); \
3986}
3987DEV_CORE_STATS_INC(rx_dropped)
3988DEV_CORE_STATS_INC(tx_dropped)
3989DEV_CORE_STATS_INC(rx_nohandler)
3990DEV_CORE_STATS_INC(rx_otherhost_dropped)
3991
3992static __always_inline int ____dev_forward_skb(struct net_device *dev,
3993 struct sk_buff *skb,
3994 const bool check_mtu)
3995{
3996 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3997 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3998 dev_core_stats_rx_dropped_inc(dev);
3999 kfree_skb(skb);
4000 return NET_RX_DROP;
4001 }
4002
4003 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4004 skb->priority = 0;
4005 return 0;
4006}
4007
4008bool dev_nit_active(struct net_device *dev);
4009void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4010
4011static inline void __dev_put(struct net_device *dev)
4012{
4013 if (dev) {
4014#ifdef CONFIG_PCPU_DEV_REFCNT
4015 this_cpu_dec(*dev->pcpu_refcnt);
4016#else
4017 refcount_dec(&dev->dev_refcnt);
4018#endif
4019 }
4020}
4021
4022static inline void __dev_hold(struct net_device *dev)
4023{
4024 if (dev) {
4025#ifdef CONFIG_PCPU_DEV_REFCNT
4026 this_cpu_inc(*dev->pcpu_refcnt);
4027#else
4028 refcount_inc(&dev->dev_refcnt);
4029#endif
4030 }
4031}
4032
4033static inline void __netdev_tracker_alloc(struct net_device *dev,
4034 netdevice_tracker *tracker,
4035 gfp_t gfp)
4036{
4037#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4038 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4039#endif
4040}
4041
4042/* netdev_tracker_alloc() can upgrade a prior untracked reference
4043 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4044 */
4045static inline void netdev_tracker_alloc(struct net_device *dev,
4046 netdevice_tracker *tracker, gfp_t gfp)
4047{
4048#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4049 refcount_dec(&dev->refcnt_tracker.no_tracker);
4050 __netdev_tracker_alloc(dev, tracker, gfp);
4051#endif
4052}
4053
4054static inline void netdev_tracker_free(struct net_device *dev,
4055 netdevice_tracker *tracker)
4056{
4057#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4058 ref_tracker_free(&dev->refcnt_tracker, tracker);
4059#endif
4060}
4061
4062static inline void netdev_hold(struct net_device *dev,
4063 netdevice_tracker *tracker, gfp_t gfp)
4064{
4065 if (dev) {
4066 __dev_hold(dev);
4067 __netdev_tracker_alloc(dev, tracker, gfp);
4068 }
4069}
4070
4071static inline void netdev_put(struct net_device *dev,
4072 netdevice_tracker *tracker)
4073{
4074 if (dev) {
4075 netdev_tracker_free(dev, tracker);
4076 __dev_put(dev);
4077 }
4078}
4079
4080/**
4081 * dev_hold - get reference to device
4082 * @dev: network device
4083 *
4084 * Hold reference to device to keep it from being freed.
4085 * Try using netdev_hold() instead.
4086 */
4087static inline void dev_hold(struct net_device *dev)
4088{
4089 netdev_hold(dev, NULL, GFP_ATOMIC);
4090}
4091
4092/**
4093 * dev_put - release reference to device
4094 * @dev: network device
4095 *
4096 * Release reference to device to allow it to be freed.
4097 * Try using netdev_put() instead.
4098 */
4099static inline void dev_put(struct net_device *dev)
4100{
4101 netdev_put(dev, NULL);
4102}
4103
4104static inline void netdev_ref_replace(struct net_device *odev,
4105 struct net_device *ndev,
4106 netdevice_tracker *tracker,
4107 gfp_t gfp)
4108{
4109 if (odev)
4110 netdev_tracker_free(odev, tracker);
4111
4112 __dev_hold(ndev);
4113 __dev_put(odev);
4114
4115 if (ndev)
4116 __netdev_tracker_alloc(ndev, tracker, gfp);
4117}
4118
4119/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4120 * and _off may be called from IRQ context, but it is caller
4121 * who is responsible for serialization of these calls.
4122 *
4123 * The name carrier is inappropriate, these functions should really be
4124 * called netif_lowerlayer_*() because they represent the state of any
4125 * kind of lower layer not just hardware media.
4126 */
4127void linkwatch_fire_event(struct net_device *dev);
4128
4129/**
4130 * netif_carrier_ok - test if carrier present
4131 * @dev: network device
4132 *
4133 * Check if carrier is present on device
4134 */
4135static inline bool netif_carrier_ok(const struct net_device *dev)
4136{
4137 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4138}
4139
4140unsigned long dev_trans_start(struct net_device *dev);
4141
4142void __netdev_watchdog_up(struct net_device *dev);
4143
4144void netif_carrier_on(struct net_device *dev);
4145void netif_carrier_off(struct net_device *dev);
4146void netif_carrier_event(struct net_device *dev);
4147
4148/**
4149 * netif_dormant_on - mark device as dormant.
4150 * @dev: network device
4151 *
4152 * Mark device as dormant (as per RFC2863).
4153 *
4154 * The dormant state indicates that the relevant interface is not
4155 * actually in a condition to pass packets (i.e., it is not 'up') but is
4156 * in a "pending" state, waiting for some external event. For "on-
4157 * demand" interfaces, this new state identifies the situation where the
4158 * interface is waiting for events to place it in the up state.
4159 */
4160static inline void netif_dormant_on(struct net_device *dev)
4161{
4162 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4163 linkwatch_fire_event(dev);
4164}
4165
4166/**
4167 * netif_dormant_off - set device as not dormant.
4168 * @dev: network device
4169 *
4170 * Device is not in dormant state.
4171 */
4172static inline void netif_dormant_off(struct net_device *dev)
4173{
4174 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4175 linkwatch_fire_event(dev);
4176}
4177
4178/**
4179 * netif_dormant - test if device is dormant
4180 * @dev: network device
4181 *
4182 * Check if device is dormant.
4183 */
4184static inline bool netif_dormant(const struct net_device *dev)
4185{
4186 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4187}
4188
4189
4190/**
4191 * netif_testing_on - mark device as under test.
4192 * @dev: network device
4193 *
4194 * Mark device as under test (as per RFC2863).
4195 *
4196 * The testing state indicates that some test(s) must be performed on
4197 * the interface. After completion, of the test, the interface state
4198 * will change to up, dormant, or down, as appropriate.
4199 */
4200static inline void netif_testing_on(struct net_device *dev)
4201{
4202 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4203 linkwatch_fire_event(dev);
4204}
4205
4206/**
4207 * netif_testing_off - set device as not under test.
4208 * @dev: network device
4209 *
4210 * Device is not in testing state.
4211 */
4212static inline void netif_testing_off(struct net_device *dev)
4213{
4214 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4215 linkwatch_fire_event(dev);
4216}
4217
4218/**
4219 * netif_testing - test if device is under test
4220 * @dev: network device
4221 *
4222 * Check if device is under test
4223 */
4224static inline bool netif_testing(const struct net_device *dev)
4225{
4226 return test_bit(__LINK_STATE_TESTING, &dev->state);
4227}
4228
4229
4230/**
4231 * netif_oper_up - test if device is operational
4232 * @dev: network device
4233 *
4234 * Check if carrier is operational
4235 */
4236static inline bool netif_oper_up(const struct net_device *dev)
4237{
4238 return (dev->operstate == IF_OPER_UP ||
4239 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4240}
4241
4242/**
4243 * netif_device_present - is device available or removed
4244 * @dev: network device
4245 *
4246 * Check if device has not been removed from system.
4247 */
4248static inline bool netif_device_present(const struct net_device *dev)
4249{
4250 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4251}
4252
4253void netif_device_detach(struct net_device *dev);
4254
4255void netif_device_attach(struct net_device *dev);
4256
4257/*
4258 * Network interface message level settings
4259 */
4260
4261enum {
4262 NETIF_MSG_DRV_BIT,
4263 NETIF_MSG_PROBE_BIT,
4264 NETIF_MSG_LINK_BIT,
4265 NETIF_MSG_TIMER_BIT,
4266 NETIF_MSG_IFDOWN_BIT,
4267 NETIF_MSG_IFUP_BIT,
4268 NETIF_MSG_RX_ERR_BIT,
4269 NETIF_MSG_TX_ERR_BIT,
4270 NETIF_MSG_TX_QUEUED_BIT,
4271 NETIF_MSG_INTR_BIT,
4272 NETIF_MSG_TX_DONE_BIT,
4273 NETIF_MSG_RX_STATUS_BIT,
4274 NETIF_MSG_PKTDATA_BIT,
4275 NETIF_MSG_HW_BIT,
4276 NETIF_MSG_WOL_BIT,
4277
4278 /* When you add a new bit above, update netif_msg_class_names array
4279 * in net/ethtool/common.c
4280 */
4281 NETIF_MSG_CLASS_COUNT,
4282};
4283/* Both ethtool_ops interface and internal driver implementation use u32 */
4284static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4285
4286#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4287#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4288
4289#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4290#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4291#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4292#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4293#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4294#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4295#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4296#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4297#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4298#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4299#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4300#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4301#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4302#define NETIF_MSG_HW __NETIF_MSG(HW)
4303#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4304
4305#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4306#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4307#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4308#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4309#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4310#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4311#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4312#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4313#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4314#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4315#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4316#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4317#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4318#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4319#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4320
4321static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4322{
4323 /* use default */
4324 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4325 return default_msg_enable_bits;
4326 if (debug_value == 0) /* no output */
4327 return 0;
4328 /* set low N bits */
4329 return (1U << debug_value) - 1;
4330}
4331
4332static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4333{
4334 spin_lock(&txq->_xmit_lock);
4335 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4336 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4337}
4338
4339static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4340{
4341 __acquire(&txq->_xmit_lock);
4342 return true;
4343}
4344
4345static inline void __netif_tx_release(struct netdev_queue *txq)
4346{
4347 __release(&txq->_xmit_lock);
4348}
4349
4350static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4351{
4352 spin_lock_bh(&txq->_xmit_lock);
4353 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4354 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4355}
4356
4357static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4358{
4359 bool ok = spin_trylock(&txq->_xmit_lock);
4360
4361 if (likely(ok)) {
4362 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4363 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4364 }
4365 return ok;
4366}
4367
4368static inline void __netif_tx_unlock(struct netdev_queue *txq)
4369{
4370 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4371 WRITE_ONCE(txq->xmit_lock_owner, -1);
4372 spin_unlock(&txq->_xmit_lock);
4373}
4374
4375static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4376{
4377 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4378 WRITE_ONCE(txq->xmit_lock_owner, -1);
4379 spin_unlock_bh(&txq->_xmit_lock);
4380}
4381
4382/*
4383 * txq->trans_start can be read locklessly from dev_watchdog()
4384 */
4385static inline void txq_trans_update(struct netdev_queue *txq)
4386{
4387 if (txq->xmit_lock_owner != -1)
4388 WRITE_ONCE(txq->trans_start, jiffies);
4389}
4390
4391static inline void txq_trans_cond_update(struct netdev_queue *txq)
4392{
4393 unsigned long now = jiffies;
4394
4395 if (READ_ONCE(txq->trans_start) != now)
4396 WRITE_ONCE(txq->trans_start, now);
4397}
4398
4399/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4400static inline void netif_trans_update(struct net_device *dev)
4401{
4402 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4403
4404 txq_trans_cond_update(txq);
4405}
4406
4407/**
4408 * netif_tx_lock - grab network device transmit lock
4409 * @dev: network device
4410 *
4411 * Get network device transmit lock
4412 */
4413void netif_tx_lock(struct net_device *dev);
4414
4415static inline void netif_tx_lock_bh(struct net_device *dev)
4416{
4417 local_bh_disable();
4418 netif_tx_lock(dev);
4419}
4420
4421void netif_tx_unlock(struct net_device *dev);
4422
4423static inline void netif_tx_unlock_bh(struct net_device *dev)
4424{
4425 netif_tx_unlock(dev);
4426 local_bh_enable();
4427}
4428
4429#define HARD_TX_LOCK(dev, txq, cpu) { \
4430 if ((dev->features & NETIF_F_LLTX) == 0) { \
4431 __netif_tx_lock(txq, cpu); \
4432 } else { \
4433 __netif_tx_acquire(txq); \
4434 } \
4435}
4436
4437#define HARD_TX_TRYLOCK(dev, txq) \
4438 (((dev->features & NETIF_F_LLTX) == 0) ? \
4439 __netif_tx_trylock(txq) : \
4440 __netif_tx_acquire(txq))
4441
4442#define HARD_TX_UNLOCK(dev, txq) { \
4443 if ((dev->features & NETIF_F_LLTX) == 0) { \
4444 __netif_tx_unlock(txq); \
4445 } else { \
4446 __netif_tx_release(txq); \
4447 } \
4448}
4449
4450static inline void netif_tx_disable(struct net_device *dev)
4451{
4452 unsigned int i;
4453 int cpu;
4454
4455 local_bh_disable();
4456 cpu = smp_processor_id();
4457 spin_lock(&dev->tx_global_lock);
4458 for (i = 0; i < dev->num_tx_queues; i++) {
4459 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4460
4461 __netif_tx_lock(txq, cpu);
4462 netif_tx_stop_queue(txq);
4463 __netif_tx_unlock(txq);
4464 }
4465 spin_unlock(&dev->tx_global_lock);
4466 local_bh_enable();
4467}
4468
4469static inline void netif_addr_lock(struct net_device *dev)
4470{
4471 unsigned char nest_level = 0;
4472
4473#ifdef CONFIG_LOCKDEP
4474 nest_level = dev->nested_level;
4475#endif
4476 spin_lock_nested(&dev->addr_list_lock, nest_level);
4477}
4478
4479static inline void netif_addr_lock_bh(struct net_device *dev)
4480{
4481 unsigned char nest_level = 0;
4482
4483#ifdef CONFIG_LOCKDEP
4484 nest_level = dev->nested_level;
4485#endif
4486 local_bh_disable();
4487 spin_lock_nested(&dev->addr_list_lock, nest_level);
4488}
4489
4490static inline void netif_addr_unlock(struct net_device *dev)
4491{
4492 spin_unlock(&dev->addr_list_lock);
4493}
4494
4495static inline void netif_addr_unlock_bh(struct net_device *dev)
4496{
4497 spin_unlock_bh(&dev->addr_list_lock);
4498}
4499
4500/*
4501 * dev_addrs walker. Should be used only for read access. Call with
4502 * rcu_read_lock held.
4503 */
4504#define for_each_dev_addr(dev, ha) \
4505 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4506
4507/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4508
4509void ether_setup(struct net_device *dev);
4510
4511/* Support for loadable net-drivers */
4512struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4513 unsigned char name_assign_type,
4514 void (*setup)(struct net_device *),
4515 unsigned int txqs, unsigned int rxqs);
4516#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4517 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4518
4519#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4520 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4521 count)
4522
4523int register_netdev(struct net_device *dev);
4524void unregister_netdev(struct net_device *dev);
4525
4526int devm_register_netdev(struct device *dev, struct net_device *ndev);
4527
4528/* General hardware address lists handling functions */
4529int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4530 struct netdev_hw_addr_list *from_list, int addr_len);
4531void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4532 struct netdev_hw_addr_list *from_list, int addr_len);
4533int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4534 struct net_device *dev,
4535 int (*sync)(struct net_device *, const unsigned char *),
4536 int (*unsync)(struct net_device *,
4537 const unsigned char *));
4538int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4539 struct net_device *dev,
4540 int (*sync)(struct net_device *,
4541 const unsigned char *, int),
4542 int (*unsync)(struct net_device *,
4543 const unsigned char *, int));
4544void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4545 struct net_device *dev,
4546 int (*unsync)(struct net_device *,
4547 const unsigned char *, int));
4548void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4549 struct net_device *dev,
4550 int (*unsync)(struct net_device *,
4551 const unsigned char *));
4552void __hw_addr_init(struct netdev_hw_addr_list *list);
4553
4554/* Functions used for device addresses handling */
4555void dev_addr_mod(struct net_device *dev, unsigned int offset,
4556 const void *addr, size_t len);
4557
4558static inline void
4559__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4560{
4561 dev_addr_mod(dev, 0, addr, len);
4562}
4563
4564static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4565{
4566 __dev_addr_set(dev, addr, dev->addr_len);
4567}
4568
4569int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4570 unsigned char addr_type);
4571int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4572 unsigned char addr_type);
4573
4574/* Functions used for unicast addresses handling */
4575int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4576int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4577int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4578int dev_uc_sync(struct net_device *to, struct net_device *from);
4579int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4580void dev_uc_unsync(struct net_device *to, struct net_device *from);
4581void dev_uc_flush(struct net_device *dev);
4582void dev_uc_init(struct net_device *dev);
4583
4584/**
4585 * __dev_uc_sync - Synchonize device's unicast list
4586 * @dev: device to sync
4587 * @sync: function to call if address should be added
4588 * @unsync: function to call if address should be removed
4589 *
4590 * Add newly added addresses to the interface, and release
4591 * addresses that have been deleted.
4592 */
4593static inline int __dev_uc_sync(struct net_device *dev,
4594 int (*sync)(struct net_device *,
4595 const unsigned char *),
4596 int (*unsync)(struct net_device *,
4597 const unsigned char *))
4598{
4599 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4600}
4601
4602/**
4603 * __dev_uc_unsync - Remove synchronized addresses from device
4604 * @dev: device to sync
4605 * @unsync: function to call if address should be removed
4606 *
4607 * Remove all addresses that were added to the device by dev_uc_sync().
4608 */
4609static inline void __dev_uc_unsync(struct net_device *dev,
4610 int (*unsync)(struct net_device *,
4611 const unsigned char *))
4612{
4613 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4614}
4615
4616/* Functions used for multicast addresses handling */
4617int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4618int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4619int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4620int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4621int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4622int dev_mc_sync(struct net_device *to, struct net_device *from);
4623int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4624void dev_mc_unsync(struct net_device *to, struct net_device *from);
4625void dev_mc_flush(struct net_device *dev);
4626void dev_mc_init(struct net_device *dev);
4627
4628/**
4629 * __dev_mc_sync - Synchonize device's multicast list
4630 * @dev: device to sync
4631 * @sync: function to call if address should be added
4632 * @unsync: function to call if address should be removed
4633 *
4634 * Add newly added addresses to the interface, and release
4635 * addresses that have been deleted.
4636 */
4637static inline int __dev_mc_sync(struct net_device *dev,
4638 int (*sync)(struct net_device *,
4639 const unsigned char *),
4640 int (*unsync)(struct net_device *,
4641 const unsigned char *))
4642{
4643 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4644}
4645
4646/**
4647 * __dev_mc_unsync - Remove synchronized addresses from device
4648 * @dev: device to sync
4649 * @unsync: function to call if address should be removed
4650 *
4651 * Remove all addresses that were added to the device by dev_mc_sync().
4652 */
4653static inline void __dev_mc_unsync(struct net_device *dev,
4654 int (*unsync)(struct net_device *,
4655 const unsigned char *))
4656{
4657 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4658}
4659
4660/* Functions used for secondary unicast and multicast support */
4661void dev_set_rx_mode(struct net_device *dev);
4662int dev_set_promiscuity(struct net_device *dev, int inc);
4663int dev_set_allmulti(struct net_device *dev, int inc);
4664void netdev_state_change(struct net_device *dev);
4665void __netdev_notify_peers(struct net_device *dev);
4666void netdev_notify_peers(struct net_device *dev);
4667void netdev_features_change(struct net_device *dev);
4668/* Load a device via the kmod */
4669void dev_load(struct net *net, const char *name);
4670struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4671 struct rtnl_link_stats64 *storage);
4672void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4673 const struct net_device_stats *netdev_stats);
4674void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4675 const struct pcpu_sw_netstats __percpu *netstats);
4676void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4677
4678extern int netdev_max_backlog;
4679extern int dev_rx_weight;
4680extern int dev_tx_weight;
4681extern int gro_normal_batch;
4682
4683enum {
4684 NESTED_SYNC_IMM_BIT,
4685 NESTED_SYNC_TODO_BIT,
4686};
4687
4688#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4689#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4690
4691#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4692#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4693
4694struct netdev_nested_priv {
4695 unsigned char flags;
4696 void *data;
4697};
4698
4699bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4700struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4701 struct list_head **iter);
4702
4703/* iterate through upper list, must be called under RCU read lock */
4704#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4705 for (iter = &(dev)->adj_list.upper, \
4706 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4707 updev; \
4708 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4709
4710int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4711 int (*fn)(struct net_device *upper_dev,
4712 struct netdev_nested_priv *priv),
4713 struct netdev_nested_priv *priv);
4714
4715bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4716 struct net_device *upper_dev);
4717
4718bool netdev_has_any_upper_dev(struct net_device *dev);
4719
4720void *netdev_lower_get_next_private(struct net_device *dev,
4721 struct list_head **iter);
4722void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4723 struct list_head **iter);
4724
4725#define netdev_for_each_lower_private(dev, priv, iter) \
4726 for (iter = (dev)->adj_list.lower.next, \
4727 priv = netdev_lower_get_next_private(dev, &(iter)); \
4728 priv; \
4729 priv = netdev_lower_get_next_private(dev, &(iter)))
4730
4731#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4732 for (iter = &(dev)->adj_list.lower, \
4733 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4734 priv; \
4735 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4736
4737void *netdev_lower_get_next(struct net_device *dev,
4738 struct list_head **iter);
4739
4740#define netdev_for_each_lower_dev(dev, ldev, iter) \
4741 for (iter = (dev)->adj_list.lower.next, \
4742 ldev = netdev_lower_get_next(dev, &(iter)); \
4743 ldev; \
4744 ldev = netdev_lower_get_next(dev, &(iter)))
4745
4746struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4747 struct list_head **iter);
4748int netdev_walk_all_lower_dev(struct net_device *dev,
4749 int (*fn)(struct net_device *lower_dev,
4750 struct netdev_nested_priv *priv),
4751 struct netdev_nested_priv *priv);
4752int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4753 int (*fn)(struct net_device *lower_dev,
4754 struct netdev_nested_priv *priv),
4755 struct netdev_nested_priv *priv);
4756
4757void *netdev_adjacent_get_private(struct list_head *adj_list);
4758void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4759struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4760struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4761int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4762 struct netlink_ext_ack *extack);
4763int netdev_master_upper_dev_link(struct net_device *dev,
4764 struct net_device *upper_dev,
4765 void *upper_priv, void *upper_info,
4766 struct netlink_ext_ack *extack);
4767void netdev_upper_dev_unlink(struct net_device *dev,
4768 struct net_device *upper_dev);
4769int netdev_adjacent_change_prepare(struct net_device *old_dev,
4770 struct net_device *new_dev,
4771 struct net_device *dev,
4772 struct netlink_ext_ack *extack);
4773void netdev_adjacent_change_commit(struct net_device *old_dev,
4774 struct net_device *new_dev,
4775 struct net_device *dev);
4776void netdev_adjacent_change_abort(struct net_device *old_dev,
4777 struct net_device *new_dev,
4778 struct net_device *dev);
4779void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4780void *netdev_lower_dev_get_private(struct net_device *dev,
4781 struct net_device *lower_dev);
4782void netdev_lower_state_changed(struct net_device *lower_dev,
4783 void *lower_state_info);
4784
4785/* RSS keys are 40 or 52 bytes long */
4786#define NETDEV_RSS_KEY_LEN 52
4787extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4788void netdev_rss_key_fill(void *buffer, size_t len);
4789
4790int skb_checksum_help(struct sk_buff *skb);
4791int skb_crc32c_csum_help(struct sk_buff *skb);
4792int skb_csum_hwoffload_help(struct sk_buff *skb,
4793 const netdev_features_t features);
4794
4795struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4796 netdev_features_t features, bool tx_path);
4797struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb,
4798 netdev_features_t features, __be16 type);
4799struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4800 netdev_features_t features);
4801
4802struct netdev_bonding_info {
4803 ifslave slave;
4804 ifbond master;
4805};
4806
4807struct netdev_notifier_bonding_info {
4808 struct netdev_notifier_info info; /* must be first */
4809 struct netdev_bonding_info bonding_info;
4810};
4811
4812void netdev_bonding_info_change(struct net_device *dev,
4813 struct netdev_bonding_info *bonding_info);
4814
4815#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4816void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4817#else
4818static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4819 const void *data)
4820{
4821}
4822#endif
4823
4824static inline
4825struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4826{
4827 return __skb_gso_segment(skb, features, true);
4828}
4829__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4830
4831static inline bool can_checksum_protocol(netdev_features_t features,
4832 __be16 protocol)
4833{
4834 if (protocol == htons(ETH_P_FCOE))
4835 return !!(features & NETIF_F_FCOE_CRC);
4836
4837 /* Assume this is an IP checksum (not SCTP CRC) */
4838
4839 if (features & NETIF_F_HW_CSUM) {
4840 /* Can checksum everything */
4841 return true;
4842 }
4843
4844 switch (protocol) {
4845 case htons(ETH_P_IP):
4846 return !!(features & NETIF_F_IP_CSUM);
4847 case htons(ETH_P_IPV6):
4848 return !!(features & NETIF_F_IPV6_CSUM);
4849 default:
4850 return false;
4851 }
4852}
4853
4854#ifdef CONFIG_BUG
4855void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4856#else
4857static inline void netdev_rx_csum_fault(struct net_device *dev,
4858 struct sk_buff *skb)
4859{
4860}
4861#endif
4862/* rx skb timestamps */
4863void net_enable_timestamp(void);
4864void net_disable_timestamp(void);
4865
4866static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4867 const struct skb_shared_hwtstamps *hwtstamps,
4868 bool cycles)
4869{
4870 const struct net_device_ops *ops = dev->netdev_ops;
4871
4872 if (ops->ndo_get_tstamp)
4873 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4874
4875 return hwtstamps->hwtstamp;
4876}
4877
4878static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4879 struct sk_buff *skb, struct net_device *dev,
4880 bool more)
4881{
4882 __this_cpu_write(softnet_data.xmit.more, more);
4883 return ops->ndo_start_xmit(skb, dev);
4884}
4885
4886static inline bool netdev_xmit_more(void)
4887{
4888 return __this_cpu_read(softnet_data.xmit.more);
4889}
4890
4891static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4892 struct netdev_queue *txq, bool more)
4893{
4894 const struct net_device_ops *ops = dev->netdev_ops;
4895 netdev_tx_t rc;
4896
4897 rc = __netdev_start_xmit(ops, skb, dev, more);
4898 if (rc == NETDEV_TX_OK)
4899 txq_trans_update(txq);
4900
4901 return rc;
4902}
4903
4904int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4905 const void *ns);
4906void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4907 const void *ns);
4908
4909extern const struct kobj_ns_type_operations net_ns_type_operations;
4910
4911const char *netdev_drivername(const struct net_device *dev);
4912
4913static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4914 netdev_features_t f2)
4915{
4916 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4917 if (f1 & NETIF_F_HW_CSUM)
4918 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4919 else
4920 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4921 }
4922
4923 return f1 & f2;
4924}
4925
4926static inline netdev_features_t netdev_get_wanted_features(
4927 struct net_device *dev)
4928{
4929 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4930}
4931netdev_features_t netdev_increment_features(netdev_features_t all,
4932 netdev_features_t one, netdev_features_t mask);
4933
4934/* Allow TSO being used on stacked device :
4935 * Performing the GSO segmentation before last device
4936 * is a performance improvement.
4937 */
4938static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4939 netdev_features_t mask)
4940{
4941 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4942}
4943
4944int __netdev_update_features(struct net_device *dev);
4945void netdev_update_features(struct net_device *dev);
4946void netdev_change_features(struct net_device *dev);
4947
4948void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4949 struct net_device *dev);
4950
4951netdev_features_t passthru_features_check(struct sk_buff *skb,
4952 struct net_device *dev,
4953 netdev_features_t features);
4954netdev_features_t netif_skb_features(struct sk_buff *skb);
4955
4956static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4957{
4958 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4959
4960 /* check flags correspondence */
4961 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4962 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4963 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4964 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4965 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4966 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4967 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4968 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4969 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4970 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4971 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4972 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4973 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4974 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4975 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4976 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4977 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4978 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4979 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4980
4981 return (features & feature) == feature;
4982}
4983
4984static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4985{
4986 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4987 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4988}
4989
4990static inline bool netif_needs_gso(struct sk_buff *skb,
4991 netdev_features_t features)
4992{
4993 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4994 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4995 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4996}
4997
4998void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
4999void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5000void netif_inherit_tso_max(struct net_device *to,
5001 const struct net_device *from);
5002
5003static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5004 int pulled_hlen, u16 mac_offset,
5005 int mac_len)
5006{
5007 skb->protocol = protocol;
5008 skb->encapsulation = 1;
5009 skb_push(skb, pulled_hlen);
5010 skb_reset_transport_header(skb);
5011 skb->mac_header = mac_offset;
5012 skb->network_header = skb->mac_header + mac_len;
5013 skb->mac_len = mac_len;
5014}
5015
5016static inline bool netif_is_macsec(const struct net_device *dev)
5017{
5018 return dev->priv_flags & IFF_MACSEC;
5019}
5020
5021static inline bool netif_is_macvlan(const struct net_device *dev)
5022{
5023 return dev->priv_flags & IFF_MACVLAN;
5024}
5025
5026static inline bool netif_is_macvlan_port(const struct net_device *dev)
5027{
5028 return dev->priv_flags & IFF_MACVLAN_PORT;
5029}
5030
5031static inline bool netif_is_bond_master(const struct net_device *dev)
5032{
5033 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5034}
5035
5036static inline bool netif_is_bond_slave(const struct net_device *dev)
5037{
5038 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5039}
5040
5041static inline bool netif_supports_nofcs(struct net_device *dev)
5042{
5043 return dev->priv_flags & IFF_SUPP_NOFCS;
5044}
5045
5046static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5047{
5048 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5049}
5050
5051static inline bool netif_is_l3_master(const struct net_device *dev)
5052{
5053 return dev->priv_flags & IFF_L3MDEV_MASTER;
5054}
5055
5056static inline bool netif_is_l3_slave(const struct net_device *dev)
5057{
5058 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5059}
5060
5061static inline bool netif_is_bridge_master(const struct net_device *dev)
5062{
5063 return dev->priv_flags & IFF_EBRIDGE;
5064}
5065
5066static inline bool netif_is_bridge_port(const struct net_device *dev)
5067{
5068 return dev->priv_flags & IFF_BRIDGE_PORT;
5069}
5070
5071static inline bool netif_is_ovs_master(const struct net_device *dev)
5072{
5073 return dev->priv_flags & IFF_OPENVSWITCH;
5074}
5075
5076static inline bool netif_is_ovs_port(const struct net_device *dev)
5077{
5078 return dev->priv_flags & IFF_OVS_DATAPATH;
5079}
5080
5081static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5082{
5083 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5084}
5085
5086static inline bool netif_is_team_master(const struct net_device *dev)
5087{
5088 return dev->priv_flags & IFF_TEAM;
5089}
5090
5091static inline bool netif_is_team_port(const struct net_device *dev)
5092{
5093 return dev->priv_flags & IFF_TEAM_PORT;
5094}
5095
5096static inline bool netif_is_lag_master(const struct net_device *dev)
5097{
5098 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5099}
5100
5101static inline bool netif_is_lag_port(const struct net_device *dev)
5102{
5103 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5104}
5105
5106static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5107{
5108 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5109}
5110
5111static inline bool netif_is_failover(const struct net_device *dev)
5112{
5113 return dev->priv_flags & IFF_FAILOVER;
5114}
5115
5116static inline bool netif_is_failover_slave(const struct net_device *dev)
5117{
5118 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5119}
5120
5121/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5122static inline void netif_keep_dst(struct net_device *dev)
5123{
5124 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5125}
5126
5127/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5128static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5129{
5130 /* TODO: reserve and use an additional IFF bit, if we get more users */
5131 return netif_is_macsec(dev);
5132}
5133
5134extern struct pernet_operations __net_initdata loopback_net_ops;
5135
5136/* Logging, debugging and troubleshooting/diagnostic helpers. */
5137
5138/* netdev_printk helpers, similar to dev_printk */
5139
5140static inline const char *netdev_name(const struct net_device *dev)
5141{
5142 if (!dev->name[0] || strchr(dev->name, '%'))
5143 return "(unnamed net_device)";
5144 return dev->name;
5145}
5146
5147static inline const char *netdev_reg_state(const struct net_device *dev)
5148{
5149 switch (dev->reg_state) {
5150 case NETREG_UNINITIALIZED: return " (uninitialized)";
5151 case NETREG_REGISTERED: return "";
5152 case NETREG_UNREGISTERING: return " (unregistering)";
5153 case NETREG_UNREGISTERED: return " (unregistered)";
5154 case NETREG_RELEASED: return " (released)";
5155 case NETREG_DUMMY: return " (dummy)";
5156 }
5157
5158 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5159 return " (unknown)";
5160}
5161
5162#define MODULE_ALIAS_NETDEV(device) \
5163 MODULE_ALIAS("netdev-" device)
5164
5165/*
5166 * netdev_WARN() acts like dev_printk(), but with the key difference
5167 * of using a WARN/WARN_ON to get the message out, including the
5168 * file/line information and a backtrace.
5169 */
5170#define netdev_WARN(dev, format, args...) \
5171 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5172 netdev_reg_state(dev), ##args)
5173
5174#define netdev_WARN_ONCE(dev, format, args...) \
5175 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5176 netdev_reg_state(dev), ##args)
5177
5178/*
5179 * The list of packet types we will receive (as opposed to discard)
5180 * and the routines to invoke.
5181 *
5182 * Why 16. Because with 16 the only overlap we get on a hash of the
5183 * low nibble of the protocol value is RARP/SNAP/X.25.
5184 *
5185 * 0800 IP
5186 * 0001 802.3
5187 * 0002 AX.25
5188 * 0004 802.2
5189 * 8035 RARP
5190 * 0005 SNAP
5191 * 0805 X.25
5192 * 0806 ARP
5193 * 8137 IPX
5194 * 0009 Localtalk
5195 * 86DD IPv6
5196 */
5197#define PTYPE_HASH_SIZE (16)
5198#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5199
5200extern struct list_head ptype_all __read_mostly;
5201extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5202
5203extern struct net_device *blackhole_netdev;
5204
5205/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5206#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5207#define DEV_STATS_ADD(DEV, FIELD, VAL) \
5208 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5209
5210#endif /* _LINUX_NETDEVICE_H */