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