Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
24#include <linux/timer.h>
25#include <linux/bug.h>
26#include <linux/delay.h>
27#include <linux/atomic.h>
28#include <linux/prefetch.h>
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31#include <asm/local.h>
32
33#include <linux/percpu.h>
34#include <linux/rculist.h>
35#include <linux/workqueue.h>
36#include <linux/dynamic_queue_limits.h>
37
38#include <net/net_namespace.h>
39#ifdef CONFIG_DCB
40#include <net/dcbnl.h>
41#endif
42#include <net/netprio_cgroup.h>
43#include <linux/netdev_features.h>
44#include <linux/neighbour.h>
45#include <linux/netdevice_xmit.h>
46#include <uapi/linux/netdevice.h>
47#include <uapi/linux/if_bonding.h>
48#include <uapi/linux/pkt_cls.h>
49#include <uapi/linux/netdev.h>
50#include <linux/hashtable.h>
51#include <linux/rbtree.h>
52#include <net/net_trackers.h>
53#include <net/net_debug.h>
54#include <net/dropreason-core.h>
55
56struct netpoll_info;
57struct device;
58struct ethtool_ops;
59struct kernel_hwtstamp_config;
60struct phy_device;
61struct dsa_port;
62struct ip_tunnel_parm_kern;
63struct macsec_context;
64struct macsec_ops;
65struct netdev_name_node;
66struct sd_flow_limit;
67struct sfp_bus;
68/* 802.11 specific */
69struct wireless_dev;
70/* 802.15.4 specific */
71struct wpan_dev;
72struct mpls_dev;
73/* UDP Tunnel offloads */
74struct udp_tunnel_info;
75struct udp_tunnel_nic_info;
76struct udp_tunnel_nic;
77struct bpf_prog;
78struct xdp_buff;
79struct xdp_frame;
80struct xdp_metadata_ops;
81struct xdp_md;
82struct ethtool_netdev_state;
83struct phy_link_topology;
84
85typedef u32 xdp_features_t;
86
87void synchronize_net(void);
88void netdev_set_default_ethtool_ops(struct net_device *dev,
89 const struct ethtool_ops *ops);
90void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
91
92/* Backlog congestion levels */
93#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
94#define NET_RX_DROP 1 /* packet dropped */
95
96#define MAX_NEST_DEV 8
97
98/*
99 * Transmit return codes: transmit return codes originate from three different
100 * namespaces:
101 *
102 * - qdisc return codes
103 * - driver transmit return codes
104 * - errno values
105 *
106 * Drivers are allowed to return any one of those in their hard_start_xmit()
107 * function. Real network devices commonly used with qdiscs should only return
108 * the driver transmit return codes though - when qdiscs are used, the actual
109 * transmission happens asynchronously, so the value is not propagated to
110 * higher layers. Virtual network devices transmit synchronously; in this case
111 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
112 * others are propagated to higher layers.
113 */
114
115/* qdisc ->enqueue() return codes. */
116#define NET_XMIT_SUCCESS 0x00
117#define NET_XMIT_DROP 0x01 /* skb dropped */
118#define NET_XMIT_CN 0x02 /* congestion notification */
119#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
120
121/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
122 * indicates that the device will soon be dropping packets, or already drops
123 * some packets of the same priority; prompting us to send less aggressively. */
124#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
125#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
126
127/* Driver transmit return codes */
128#define NETDEV_TX_MASK 0xf0
129
130enum netdev_tx {
131 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
132 NETDEV_TX_OK = 0x00, /* driver took care of packet */
133 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
134};
135typedef enum netdev_tx netdev_tx_t;
136
137/*
138 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
139 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
140 */
141static inline bool dev_xmit_complete(int rc)
142{
143 /*
144 * Positive cases with an skb consumed by a driver:
145 * - successful transmission (rc == NETDEV_TX_OK)
146 * - error while transmitting (rc < 0)
147 * - error while queueing to a different device (rc & NET_XMIT_MASK)
148 */
149 if (likely(rc < NET_XMIT_MASK))
150 return true;
151
152 return false;
153}
154
155/*
156 * Compute the worst-case header length according to the protocols
157 * used.
158 */
159
160#if defined(CONFIG_HYPERV_NET)
161# define LL_MAX_HEADER 128
162#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
163# if defined(CONFIG_MAC80211_MESH)
164# define LL_MAX_HEADER 128
165# else
166# define LL_MAX_HEADER 96
167# endif
168#else
169# define LL_MAX_HEADER 32
170#endif
171
172#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
173 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
174#define MAX_HEADER LL_MAX_HEADER
175#else
176#define MAX_HEADER (LL_MAX_HEADER + 48)
177#endif
178
179/*
180 * Old network device statistics. Fields are native words
181 * (unsigned long) so they can be read and written atomically.
182 */
183
184#define NET_DEV_STAT(FIELD) \
185 union { \
186 unsigned long FIELD; \
187 atomic_long_t __##FIELD; \
188 }
189
190struct net_device_stats {
191 NET_DEV_STAT(rx_packets);
192 NET_DEV_STAT(tx_packets);
193 NET_DEV_STAT(rx_bytes);
194 NET_DEV_STAT(tx_bytes);
195 NET_DEV_STAT(rx_errors);
196 NET_DEV_STAT(tx_errors);
197 NET_DEV_STAT(rx_dropped);
198 NET_DEV_STAT(tx_dropped);
199 NET_DEV_STAT(multicast);
200 NET_DEV_STAT(collisions);
201 NET_DEV_STAT(rx_length_errors);
202 NET_DEV_STAT(rx_over_errors);
203 NET_DEV_STAT(rx_crc_errors);
204 NET_DEV_STAT(rx_frame_errors);
205 NET_DEV_STAT(rx_fifo_errors);
206 NET_DEV_STAT(rx_missed_errors);
207 NET_DEV_STAT(tx_aborted_errors);
208 NET_DEV_STAT(tx_carrier_errors);
209 NET_DEV_STAT(tx_fifo_errors);
210 NET_DEV_STAT(tx_heartbeat_errors);
211 NET_DEV_STAT(tx_window_errors);
212 NET_DEV_STAT(rx_compressed);
213 NET_DEV_STAT(tx_compressed);
214};
215#undef NET_DEV_STAT
216
217/* per-cpu stats, allocated on demand.
218 * Try to fit them in a single cache line, for dev_get_stats() sake.
219 */
220struct net_device_core_stats {
221 unsigned long rx_dropped;
222 unsigned long tx_dropped;
223 unsigned long rx_nohandler;
224 unsigned long rx_otherhost_dropped;
225} __aligned(4 * sizeof(unsigned long));
226
227#include <linux/cache.h>
228#include <linux/skbuff.h>
229
230struct neighbour;
231struct neigh_parms;
232struct sk_buff;
233
234struct netdev_hw_addr {
235 struct list_head list;
236 struct rb_node node;
237 unsigned char addr[MAX_ADDR_LEN];
238 unsigned char type;
239#define NETDEV_HW_ADDR_T_LAN 1
240#define NETDEV_HW_ADDR_T_SAN 2
241#define NETDEV_HW_ADDR_T_UNICAST 3
242#define NETDEV_HW_ADDR_T_MULTICAST 4
243 bool global_use;
244 int sync_cnt;
245 int refcount;
246 int synced;
247 struct rcu_head rcu_head;
248};
249
250struct netdev_hw_addr_list {
251 struct list_head list;
252 int count;
253
254 /* Auxiliary tree for faster lookup on addition and deletion */
255 struct rb_root tree;
256};
257
258#define netdev_hw_addr_list_count(l) ((l)->count)
259#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
260#define netdev_hw_addr_list_for_each(ha, l) \
261 list_for_each_entry(ha, &(l)->list, list)
262
263#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
264#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
265#define netdev_for_each_uc_addr(ha, dev) \
266 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
267#define netdev_for_each_synced_uc_addr(_ha, _dev) \
268 netdev_for_each_uc_addr((_ha), (_dev)) \
269 if ((_ha)->sync_cnt)
270
271#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
272#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
273#define netdev_for_each_mc_addr(ha, dev) \
274 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
275#define netdev_for_each_synced_mc_addr(_ha, _dev) \
276 netdev_for_each_mc_addr((_ha), (_dev)) \
277 if ((_ha)->sync_cnt)
278
279struct hh_cache {
280 unsigned int hh_len;
281 seqlock_t hh_lock;
282
283 /* cached hardware header; allow for machine alignment needs. */
284#define HH_DATA_MOD 16
285#define HH_DATA_OFF(__len) \
286 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
287#define HH_DATA_ALIGN(__len) \
288 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
289 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
290};
291
292/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
293 * Alternative is:
294 * dev->hard_header_len ? (dev->hard_header_len +
295 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
296 *
297 * We could use other alignment values, but we must maintain the
298 * relationship HH alignment <= LL alignment.
299 */
300#define LL_RESERVED_SPACE(dev) \
301 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
302 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
303#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
304 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
305 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
306
307struct header_ops {
308 int (*create) (struct sk_buff *skb, struct net_device *dev,
309 unsigned short type, const void *daddr,
310 const void *saddr, unsigned int len);
311 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
312 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
313 void (*cache_update)(struct hh_cache *hh,
314 const struct net_device *dev,
315 const unsigned char *haddr);
316 bool (*validate)(const char *ll_header, unsigned int len);
317 __be16 (*parse_protocol)(const struct sk_buff *skb);
318};
319
320/* These flag bits are private to the generic network queueing
321 * layer; they may not be explicitly referenced by any other
322 * code.
323 */
324
325enum netdev_state_t {
326 __LINK_STATE_START,
327 __LINK_STATE_PRESENT,
328 __LINK_STATE_NOCARRIER,
329 __LINK_STATE_LINKWATCH_PENDING,
330 __LINK_STATE_DORMANT,
331 __LINK_STATE_TESTING,
332};
333
334struct gro_list {
335 struct list_head list;
336 int count;
337};
338
339/*
340 * size of gro hash buckets, must less than bit number of
341 * napi_struct::gro_bitmask
342 */
343#define GRO_HASH_BUCKETS 8
344
345/*
346 * Structure for NAPI scheduling similar to tasklet but with weighting
347 */
348struct napi_struct {
349 /* The poll_list must only be managed by the entity which
350 * changes the state of the NAPI_STATE_SCHED bit. This means
351 * whoever atomically sets that bit can add this napi_struct
352 * to the per-CPU poll_list, and whoever clears that bit
353 * can remove from the list right before clearing the bit.
354 */
355 struct list_head poll_list;
356
357 unsigned long state;
358 int weight;
359 u32 defer_hard_irqs_count;
360 unsigned long gro_bitmask;
361 int (*poll)(struct napi_struct *, int);
362#ifdef CONFIG_NETPOLL
363 /* CPU actively polling if netpoll is configured */
364 int poll_owner;
365#endif
366 /* CPU on which NAPI has been scheduled for processing */
367 int list_owner;
368 struct net_device *dev;
369 struct gro_list gro_hash[GRO_HASH_BUCKETS];
370 struct sk_buff *skb;
371 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
372 int rx_count; /* length of rx_list */
373 unsigned int napi_id;
374 struct hrtimer timer;
375 struct task_struct *thread;
376 /* control-path-only fields follow */
377 struct list_head dev_list;
378 struct hlist_node napi_hash_node;
379 int irq;
380};
381
382enum {
383 NAPI_STATE_SCHED, /* Poll is scheduled */
384 NAPI_STATE_MISSED, /* reschedule a napi */
385 NAPI_STATE_DISABLE, /* Disable pending */
386 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
387 NAPI_STATE_LISTED, /* NAPI added to system lists */
388 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
389 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
390 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
391 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
392 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
393};
394
395enum {
396 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
397 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
398 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
399 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
400 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
401 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
402 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
403 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
404 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
405 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
406};
407
408enum gro_result {
409 GRO_MERGED,
410 GRO_MERGED_FREE,
411 GRO_HELD,
412 GRO_NORMAL,
413 GRO_CONSUMED,
414};
415typedef enum gro_result gro_result_t;
416
417/*
418 * enum rx_handler_result - Possible return values for rx_handlers.
419 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
420 * further.
421 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
422 * case skb->dev was changed by rx_handler.
423 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
424 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
425 *
426 * rx_handlers are functions called from inside __netif_receive_skb(), to do
427 * special processing of the skb, prior to delivery to protocol handlers.
428 *
429 * Currently, a net_device can only have a single rx_handler registered. Trying
430 * to register a second rx_handler will return -EBUSY.
431 *
432 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
433 * To unregister a rx_handler on a net_device, use
434 * netdev_rx_handler_unregister().
435 *
436 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
437 * do with the skb.
438 *
439 * If the rx_handler consumed the skb in some way, it should return
440 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
441 * the skb to be delivered in some other way.
442 *
443 * If the rx_handler changed skb->dev, to divert the skb to another
444 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
445 * new device will be called if it exists.
446 *
447 * If the rx_handler decides the skb should be ignored, it should return
448 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
449 * are registered on exact device (ptype->dev == skb->dev).
450 *
451 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
452 * delivered, it should return RX_HANDLER_PASS.
453 *
454 * A device without a registered rx_handler will behave as if rx_handler
455 * returned RX_HANDLER_PASS.
456 */
457
458enum rx_handler_result {
459 RX_HANDLER_CONSUMED,
460 RX_HANDLER_ANOTHER,
461 RX_HANDLER_EXACT,
462 RX_HANDLER_PASS,
463};
464typedef enum rx_handler_result rx_handler_result_t;
465typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
466
467void __napi_schedule(struct napi_struct *n);
468void __napi_schedule_irqoff(struct napi_struct *n);
469
470static inline bool napi_disable_pending(struct napi_struct *n)
471{
472 return test_bit(NAPI_STATE_DISABLE, &n->state);
473}
474
475static inline bool napi_prefer_busy_poll(struct napi_struct *n)
476{
477 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
478}
479
480/**
481 * napi_is_scheduled - test if NAPI is scheduled
482 * @n: NAPI context
483 *
484 * This check is "best-effort". With no locking implemented,
485 * a NAPI can be scheduled or terminate right after this check
486 * and produce not precise results.
487 *
488 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
489 * should not be used normally and napi_schedule should be
490 * used instead.
491 *
492 * Use only if the driver really needs to check if a NAPI
493 * is scheduled for example in the context of delayed timer
494 * that can be skipped if a NAPI is already scheduled.
495 *
496 * Return True if NAPI is scheduled, False otherwise.
497 */
498static inline bool napi_is_scheduled(struct napi_struct *n)
499{
500 return test_bit(NAPI_STATE_SCHED, &n->state);
501}
502
503bool napi_schedule_prep(struct napi_struct *n);
504
505/**
506 * napi_schedule - schedule NAPI poll
507 * @n: NAPI context
508 *
509 * Schedule NAPI poll routine to be called if it is not already
510 * running.
511 * Return true if we schedule a NAPI or false if not.
512 * Refer to napi_schedule_prep() for additional reason on why
513 * a NAPI might not be scheduled.
514 */
515static inline bool napi_schedule(struct napi_struct *n)
516{
517 if (napi_schedule_prep(n)) {
518 __napi_schedule(n);
519 return true;
520 }
521
522 return false;
523}
524
525/**
526 * napi_schedule_irqoff - schedule NAPI poll
527 * @n: NAPI context
528 *
529 * Variant of napi_schedule(), assuming hard irqs are masked.
530 */
531static inline void napi_schedule_irqoff(struct napi_struct *n)
532{
533 if (napi_schedule_prep(n))
534 __napi_schedule_irqoff(n);
535}
536
537/**
538 * napi_complete_done - NAPI processing complete
539 * @n: NAPI context
540 * @work_done: number of packets processed
541 *
542 * Mark NAPI processing as complete. Should only be called if poll budget
543 * has not been completely consumed.
544 * Prefer over napi_complete().
545 * Return false if device should avoid rearming interrupts.
546 */
547bool napi_complete_done(struct napi_struct *n, int work_done);
548
549static inline bool napi_complete(struct napi_struct *n)
550{
551 return napi_complete_done(n, 0);
552}
553
554int dev_set_threaded(struct net_device *dev, bool threaded);
555
556/**
557 * napi_disable - prevent NAPI from scheduling
558 * @n: NAPI context
559 *
560 * Stop NAPI from being scheduled on this context.
561 * Waits till any outstanding processing completes.
562 */
563void napi_disable(struct napi_struct *n);
564
565void napi_enable(struct napi_struct *n);
566
567/**
568 * napi_synchronize - wait until NAPI is not running
569 * @n: NAPI context
570 *
571 * Wait until NAPI is done being scheduled on this context.
572 * Waits till any outstanding processing completes but
573 * does not disable future activations.
574 */
575static inline void napi_synchronize(const struct napi_struct *n)
576{
577 if (IS_ENABLED(CONFIG_SMP))
578 while (test_bit(NAPI_STATE_SCHED, &n->state))
579 msleep(1);
580 else
581 barrier();
582}
583
584/**
585 * napi_if_scheduled_mark_missed - if napi is running, set the
586 * NAPIF_STATE_MISSED
587 * @n: NAPI context
588 *
589 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
590 * NAPI is scheduled.
591 **/
592static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
593{
594 unsigned long val, new;
595
596 val = READ_ONCE(n->state);
597 do {
598 if (val & NAPIF_STATE_DISABLE)
599 return true;
600
601 if (!(val & NAPIF_STATE_SCHED))
602 return false;
603
604 new = val | NAPIF_STATE_MISSED;
605 } while (!try_cmpxchg(&n->state, &val, new));
606
607 return true;
608}
609
610enum netdev_queue_state_t {
611 __QUEUE_STATE_DRV_XOFF,
612 __QUEUE_STATE_STACK_XOFF,
613 __QUEUE_STATE_FROZEN,
614};
615
616#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
617#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
618#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
619
620#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
621#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
622 QUEUE_STATE_FROZEN)
623#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
624 QUEUE_STATE_FROZEN)
625
626/*
627 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
628 * netif_tx_* functions below are used to manipulate this flag. The
629 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
630 * queue independently. The netif_xmit_*stopped functions below are called
631 * to check if the queue has been stopped by the driver or stack (either
632 * of the XOFF bits are set in the state). Drivers should not need to call
633 * netif_xmit*stopped functions, they should only be using netif_tx_*.
634 */
635
636struct netdev_queue {
637/*
638 * read-mostly part
639 */
640 struct net_device *dev;
641 netdevice_tracker dev_tracker;
642
643 struct Qdisc __rcu *qdisc;
644 struct Qdisc __rcu *qdisc_sleeping;
645#ifdef CONFIG_SYSFS
646 struct kobject kobj;
647#endif
648 unsigned long tx_maxrate;
649 /*
650 * Number of TX timeouts for this queue
651 * (/sys/class/net/DEV/Q/trans_timeout)
652 */
653 atomic_long_t trans_timeout;
654
655 /* Subordinate device that the queue has been assigned to */
656 struct net_device *sb_dev;
657#ifdef CONFIG_XDP_SOCKETS
658 struct xsk_buff_pool *pool;
659#endif
660
661/*
662 * write-mostly part
663 */
664#ifdef CONFIG_BQL
665 struct dql dql;
666#endif
667 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
668 int xmit_lock_owner;
669 /*
670 * Time (in jiffies) of last Tx
671 */
672 unsigned long trans_start;
673
674 unsigned long state;
675
676/*
677 * slow- / control-path part
678 */
679 /* NAPI instance for the queue
680 * Readers and writers must hold RTNL
681 */
682 struct napi_struct *napi;
683
684#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
685 int numa_node;
686#endif
687} ____cacheline_aligned_in_smp;
688
689extern int sysctl_fb_tunnels_only_for_init_net;
690extern int sysctl_devconf_inherit_init_net;
691
692/*
693 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
694 * == 1 : For initns only
695 * == 2 : For none.
696 */
697static inline bool net_has_fallback_tunnels(const struct net *net)
698{
699#if IS_ENABLED(CONFIG_SYSCTL)
700 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
701
702 return !fb_tunnels_only_for_init_net ||
703 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
704#else
705 return true;
706#endif
707}
708
709static inline int net_inherit_devconf(void)
710{
711#if IS_ENABLED(CONFIG_SYSCTL)
712 return READ_ONCE(sysctl_devconf_inherit_init_net);
713#else
714 return 0;
715#endif
716}
717
718static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
719{
720#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
721 return q->numa_node;
722#else
723 return NUMA_NO_NODE;
724#endif
725}
726
727static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
728{
729#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
730 q->numa_node = node;
731#endif
732}
733
734#ifdef CONFIG_RFS_ACCEL
735bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
736 u16 filter_id);
737#endif
738
739/* XPS map type and offset of the xps map within net_device->xps_maps[]. */
740enum xps_map_type {
741 XPS_CPUS = 0,
742 XPS_RXQS,
743 XPS_MAPS_MAX,
744};
745
746#ifdef CONFIG_XPS
747/*
748 * This structure holds an XPS map which can be of variable length. The
749 * map is an array of queues.
750 */
751struct xps_map {
752 unsigned int len;
753 unsigned int alloc_len;
754 struct rcu_head rcu;
755 u16 queues[];
756};
757#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
758#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
759 - sizeof(struct xps_map)) / sizeof(u16))
760
761/*
762 * This structure holds all XPS maps for device. Maps are indexed by CPU.
763 *
764 * We keep track of the number of cpus/rxqs used when the struct is allocated,
765 * in nr_ids. This will help not accessing out-of-bound memory.
766 *
767 * We keep track of the number of traffic classes used when the struct is
768 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
769 * not crossing its upper bound, as the original dev->num_tc can be updated in
770 * the meantime.
771 */
772struct xps_dev_maps {
773 struct rcu_head rcu;
774 unsigned int nr_ids;
775 s16 num_tc;
776 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
777};
778
779#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
780 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
781
782#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
783 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
784
785#endif /* CONFIG_XPS */
786
787#define TC_MAX_QUEUE 16
788#define TC_BITMASK 15
789/* HW offloaded queuing disciplines txq count and offset maps */
790struct netdev_tc_txq {
791 u16 count;
792 u16 offset;
793};
794
795#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
796/*
797 * This structure is to hold information about the device
798 * configured to run FCoE protocol stack.
799 */
800struct netdev_fcoe_hbainfo {
801 char manufacturer[64];
802 char serial_number[64];
803 char hardware_version[64];
804 char driver_version[64];
805 char optionrom_version[64];
806 char firmware_version[64];
807 char model[256];
808 char model_description[256];
809};
810#endif
811
812#define MAX_PHYS_ITEM_ID_LEN 32
813
814/* This structure holds a unique identifier to identify some
815 * physical item (port for example) used by a netdevice.
816 */
817struct netdev_phys_item_id {
818 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
819 unsigned char id_len;
820};
821
822static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
823 struct netdev_phys_item_id *b)
824{
825 return a->id_len == b->id_len &&
826 memcmp(a->id, b->id, a->id_len) == 0;
827}
828
829typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
830 struct sk_buff *skb,
831 struct net_device *sb_dev);
832
833enum net_device_path_type {
834 DEV_PATH_ETHERNET = 0,
835 DEV_PATH_VLAN,
836 DEV_PATH_BRIDGE,
837 DEV_PATH_PPPOE,
838 DEV_PATH_DSA,
839 DEV_PATH_MTK_WDMA,
840};
841
842struct net_device_path {
843 enum net_device_path_type type;
844 const struct net_device *dev;
845 union {
846 struct {
847 u16 id;
848 __be16 proto;
849 u8 h_dest[ETH_ALEN];
850 } encap;
851 struct {
852 enum {
853 DEV_PATH_BR_VLAN_KEEP,
854 DEV_PATH_BR_VLAN_TAG,
855 DEV_PATH_BR_VLAN_UNTAG,
856 DEV_PATH_BR_VLAN_UNTAG_HW,
857 } vlan_mode;
858 u16 vlan_id;
859 __be16 vlan_proto;
860 } bridge;
861 struct {
862 int port;
863 u16 proto;
864 } dsa;
865 struct {
866 u8 wdma_idx;
867 u8 queue;
868 u16 wcid;
869 u8 bss;
870 u8 amsdu;
871 } mtk_wdma;
872 };
873};
874
875#define NET_DEVICE_PATH_STACK_MAX 5
876#define NET_DEVICE_PATH_VLAN_MAX 2
877
878struct net_device_path_stack {
879 int num_paths;
880 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
881};
882
883struct net_device_path_ctx {
884 const struct net_device *dev;
885 u8 daddr[ETH_ALEN];
886
887 int num_vlans;
888 struct {
889 u16 id;
890 __be16 proto;
891 } vlan[NET_DEVICE_PATH_VLAN_MAX];
892};
893
894enum tc_setup_type {
895 TC_QUERY_CAPS,
896 TC_SETUP_QDISC_MQPRIO,
897 TC_SETUP_CLSU32,
898 TC_SETUP_CLSFLOWER,
899 TC_SETUP_CLSMATCHALL,
900 TC_SETUP_CLSBPF,
901 TC_SETUP_BLOCK,
902 TC_SETUP_QDISC_CBS,
903 TC_SETUP_QDISC_RED,
904 TC_SETUP_QDISC_PRIO,
905 TC_SETUP_QDISC_MQ,
906 TC_SETUP_QDISC_ETF,
907 TC_SETUP_ROOT_QDISC,
908 TC_SETUP_QDISC_GRED,
909 TC_SETUP_QDISC_TAPRIO,
910 TC_SETUP_FT,
911 TC_SETUP_QDISC_ETS,
912 TC_SETUP_QDISC_TBF,
913 TC_SETUP_QDISC_FIFO,
914 TC_SETUP_QDISC_HTB,
915 TC_SETUP_ACT,
916};
917
918/* These structures hold the attributes of bpf state that are being passed
919 * to the netdevice through the bpf op.
920 */
921enum bpf_netdev_command {
922 /* Set or clear a bpf program used in the earliest stages of packet
923 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
924 * is responsible for calling bpf_prog_put on any old progs that are
925 * stored. In case of error, the callee need not release the new prog
926 * reference, but on success it takes ownership and must bpf_prog_put
927 * when it is no longer used.
928 */
929 XDP_SETUP_PROG,
930 XDP_SETUP_PROG_HW,
931 /* BPF program for offload callbacks, invoked at program load time. */
932 BPF_OFFLOAD_MAP_ALLOC,
933 BPF_OFFLOAD_MAP_FREE,
934 XDP_SETUP_XSK_POOL,
935};
936
937struct bpf_prog_offload_ops;
938struct netlink_ext_ack;
939struct xdp_umem;
940struct xdp_dev_bulk_queue;
941struct bpf_xdp_link;
942
943enum bpf_xdp_mode {
944 XDP_MODE_SKB = 0,
945 XDP_MODE_DRV = 1,
946 XDP_MODE_HW = 2,
947 __MAX_XDP_MODE
948};
949
950struct bpf_xdp_entity {
951 struct bpf_prog *prog;
952 struct bpf_xdp_link *link;
953};
954
955struct netdev_bpf {
956 enum bpf_netdev_command command;
957 union {
958 /* XDP_SETUP_PROG */
959 struct {
960 u32 flags;
961 struct bpf_prog *prog;
962 struct netlink_ext_ack *extack;
963 };
964 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
965 struct {
966 struct bpf_offloaded_map *offmap;
967 };
968 /* XDP_SETUP_XSK_POOL */
969 struct {
970 struct xsk_buff_pool *pool;
971 u16 queue_id;
972 } xsk;
973 };
974};
975
976/* Flags for ndo_xsk_wakeup. */
977#define XDP_WAKEUP_RX (1 << 0)
978#define XDP_WAKEUP_TX (1 << 1)
979
980#ifdef CONFIG_XFRM_OFFLOAD
981struct xfrmdev_ops {
982 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
983 void (*xdo_dev_state_delete) (struct xfrm_state *x);
984 void (*xdo_dev_state_free) (struct xfrm_state *x);
985 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
986 struct xfrm_state *x);
987 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
988 void (*xdo_dev_state_update_stats) (struct xfrm_state *x);
989 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
990 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
991 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
992};
993#endif
994
995struct dev_ifalias {
996 struct rcu_head rcuhead;
997 char ifalias[];
998};
999
1000struct devlink;
1001struct tlsdev_ops;
1002
1003struct netdev_net_notifier {
1004 struct list_head list;
1005 struct notifier_block *nb;
1006};
1007
1008/*
1009 * This structure defines the management hooks for network devices.
1010 * The following hooks can be defined; unless noted otherwise, they are
1011 * optional and can be filled with a null pointer.
1012 *
1013 * int (*ndo_init)(struct net_device *dev);
1014 * This function is called once when a network device is registered.
1015 * The network device can use this for any late stage initialization
1016 * or semantic validation. It can fail with an error code which will
1017 * be propagated back to register_netdev.
1018 *
1019 * void (*ndo_uninit)(struct net_device *dev);
1020 * This function is called when device is unregistered or when registration
1021 * fails. It is not called if init fails.
1022 *
1023 * int (*ndo_open)(struct net_device *dev);
1024 * This function is called when a network device transitions to the up
1025 * state.
1026 *
1027 * int (*ndo_stop)(struct net_device *dev);
1028 * This function is called when a network device transitions to the down
1029 * state.
1030 *
1031 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1032 * struct net_device *dev);
1033 * Called when a packet needs to be transmitted.
1034 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1035 * the queue before that can happen; it's for obsolete devices and weird
1036 * corner cases, but the stack really does a non-trivial amount
1037 * of useless work if you return NETDEV_TX_BUSY.
1038 * Required; cannot be NULL.
1039 *
1040 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1041 * struct net_device *dev
1042 * netdev_features_t features);
1043 * Called by core transmit path to determine if device is capable of
1044 * performing offload operations on a given packet. This is to give
1045 * the device an opportunity to implement any restrictions that cannot
1046 * be otherwise expressed by feature flags. The check is called with
1047 * the set of features that the stack has calculated and it returns
1048 * those the driver believes to be appropriate.
1049 *
1050 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1051 * struct net_device *sb_dev);
1052 * Called to decide which queue to use when device supports multiple
1053 * transmit queues.
1054 *
1055 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1056 * This function is called to allow device receiver to make
1057 * changes to configuration when multicast or promiscuous is enabled.
1058 *
1059 * void (*ndo_set_rx_mode)(struct net_device *dev);
1060 * This function is called device changes address list filtering.
1061 * If driver handles unicast address filtering, it should set
1062 * IFF_UNICAST_FLT in its priv_flags.
1063 *
1064 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1065 * This function is called when the Media Access Control address
1066 * needs to be changed. If this interface is not defined, the
1067 * MAC address can not be changed.
1068 *
1069 * int (*ndo_validate_addr)(struct net_device *dev);
1070 * Test if Media Access Control address is valid for the device.
1071 *
1072 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1073 * Old-style ioctl entry point. This is used internally by the
1074 * appletalk and ieee802154 subsystems but is no longer called by
1075 * the device ioctl handler.
1076 *
1077 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1078 * Used by the bonding driver for its device specific ioctls:
1079 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1080 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1081 *
1082 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1083 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1084 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1085 *
1086 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1087 * Used to set network devices bus interface parameters. This interface
1088 * is retained for legacy reasons; new devices should use the bus
1089 * interface (PCI) for low level management.
1090 *
1091 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1092 * Called when a user wants to change the Maximum Transfer Unit
1093 * of a device.
1094 *
1095 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1096 * Callback used when the transmitter has not made any progress
1097 * for dev->watchdog ticks.
1098 *
1099 * void (*ndo_get_stats64)(struct net_device *dev,
1100 * struct rtnl_link_stats64 *storage);
1101 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1102 * Called when a user wants to get the network device usage
1103 * statistics. Drivers must do one of the following:
1104 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1105 * rtnl_link_stats64 structure passed by the caller.
1106 * 2. Define @ndo_get_stats to update a net_device_stats structure
1107 * (which should normally be dev->stats) and return a pointer to
1108 * it. The structure may be changed asynchronously only if each
1109 * field is written atomically.
1110 * 3. Update dev->stats asynchronously and atomically, and define
1111 * neither operation.
1112 *
1113 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1114 * Return true if this device supports offload stats of this attr_id.
1115 *
1116 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1117 * void *attr_data)
1118 * Get statistics for offload operations by attr_id. Write it into the
1119 * attr_data pointer.
1120 *
1121 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1122 * If device supports VLAN filtering this function is called when a
1123 * VLAN id is registered.
1124 *
1125 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1126 * If device supports VLAN filtering this function is called when a
1127 * VLAN id is unregistered.
1128 *
1129 * void (*ndo_poll_controller)(struct net_device *dev);
1130 *
1131 * SR-IOV management functions.
1132 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1133 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1134 * u8 qos, __be16 proto);
1135 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1136 * int max_tx_rate);
1137 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1138 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1139 * int (*ndo_get_vf_config)(struct net_device *dev,
1140 * int vf, struct ifla_vf_info *ivf);
1141 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1142 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1143 * struct nlattr *port[]);
1144 *
1145 * Enable or disable the VF ability to query its RSS Redirection Table and
1146 * Hash Key. This is needed since on some devices VF share this information
1147 * with PF and querying it may introduce a theoretical security risk.
1148 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1149 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1150 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1151 * void *type_data);
1152 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1153 * This is always called from the stack with the rtnl lock held and netif
1154 * tx queues stopped. This allows the netdevice to perform queue
1155 * management safely.
1156 *
1157 * Fiber Channel over Ethernet (FCoE) offload functions.
1158 * int (*ndo_fcoe_enable)(struct net_device *dev);
1159 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1160 * so the underlying device can perform whatever needed configuration or
1161 * initialization to support acceleration of FCoE traffic.
1162 *
1163 * int (*ndo_fcoe_disable)(struct net_device *dev);
1164 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1165 * so the underlying device can perform whatever needed clean-ups to
1166 * stop supporting acceleration of FCoE traffic.
1167 *
1168 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1169 * struct scatterlist *sgl, unsigned int sgc);
1170 * Called when the FCoE Initiator wants to initialize an I/O that
1171 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1172 * perform necessary setup and returns 1 to indicate the device is set up
1173 * successfully to perform DDP on this I/O, otherwise this returns 0.
1174 *
1175 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1176 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1177 * indicated by the FC exchange id 'xid', so the underlying device can
1178 * clean up and reuse resources for later DDP requests.
1179 *
1180 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1181 * struct scatterlist *sgl, unsigned int sgc);
1182 * Called when the FCoE Target wants to initialize an I/O that
1183 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1184 * perform necessary setup and returns 1 to indicate the device is set up
1185 * successfully to perform DDP on this I/O, otherwise this returns 0.
1186 *
1187 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1188 * struct netdev_fcoe_hbainfo *hbainfo);
1189 * Called when the FCoE Protocol stack wants information on the underlying
1190 * device. This information is utilized by the FCoE protocol stack to
1191 * register attributes with Fiber Channel management service as per the
1192 * FC-GS Fabric Device Management Information(FDMI) specification.
1193 *
1194 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1195 * Called when the underlying device wants to override default World Wide
1196 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1197 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1198 * protocol stack to use.
1199 *
1200 * RFS acceleration.
1201 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1202 * u16 rxq_index, u32 flow_id);
1203 * Set hardware filter for RFS. rxq_index is the target queue index;
1204 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1205 * Return the filter ID on success, or a negative error code.
1206 *
1207 * Slave management functions (for bridge, bonding, etc).
1208 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1209 * Called to make another netdev an underling.
1210 *
1211 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1212 * Called to release previously enslaved netdev.
1213 *
1214 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1215 * struct sk_buff *skb,
1216 * bool all_slaves);
1217 * Get the xmit slave of master device. If all_slaves is true, function
1218 * assume all the slaves can transmit.
1219 *
1220 * Feature/offload setting functions.
1221 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1222 * netdev_features_t features);
1223 * Adjusts the requested feature flags according to device-specific
1224 * constraints, and returns the resulting flags. Must not modify
1225 * the device state.
1226 *
1227 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1228 * Called to update device configuration to new features. Passed
1229 * feature set might be less than what was returned by ndo_fix_features()).
1230 * Must return >0 or -errno if it changed dev->features itself.
1231 *
1232 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1233 * struct net_device *dev,
1234 * const unsigned char *addr, u16 vid, u16 flags,
1235 * struct netlink_ext_ack *extack);
1236 * Adds an FDB entry to dev for addr.
1237 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1238 * struct net_device *dev,
1239 * const unsigned char *addr, u16 vid)
1240 * Deletes the FDB entry from dev corresponding to addr.
1241 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1242 * struct netlink_ext_ack *extack);
1243 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1244 * struct net_device *dev, struct net_device *filter_dev,
1245 * int *idx)
1246 * Used to add FDB entries to dump requests. Implementers should add
1247 * entries to skb and update idx with the number of entries.
1248 *
1249 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1250 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1251 * Adds an MDB entry to dev.
1252 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1253 * struct netlink_ext_ack *extack);
1254 * Deletes the MDB entry from dev.
1255 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1256 * struct netlink_ext_ack *extack);
1257 * Bulk deletes MDB entries from dev.
1258 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1259 * struct netlink_callback *cb);
1260 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1261 * callback is used by core rtnetlink code.
1262 *
1263 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1264 * u16 flags, struct netlink_ext_ack *extack)
1265 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1266 * struct net_device *dev, u32 filter_mask,
1267 * int nlflags)
1268 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1269 * u16 flags);
1270 *
1271 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1272 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1273 * which do not represent real hardware may define this to allow their
1274 * userspace components to manage their virtual carrier state. Devices
1275 * that determine carrier state from physical hardware properties (eg
1276 * network cables) or protocol-dependent mechanisms (eg
1277 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1278 *
1279 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1280 * struct netdev_phys_item_id *ppid);
1281 * Called to get ID of physical port of this device. If driver does
1282 * not implement this, it is assumed that the hw is not able to have
1283 * multiple net devices on single physical port.
1284 *
1285 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1286 * struct netdev_phys_item_id *ppid)
1287 * Called to get the parent ID of the physical port of this device.
1288 *
1289 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1290 * struct net_device *dev)
1291 * Called by upper layer devices to accelerate switching or other
1292 * station functionality into hardware. 'pdev is the lowerdev
1293 * to use for the offload and 'dev' is the net device that will
1294 * back the offload. Returns a pointer to the private structure
1295 * the upper layer will maintain.
1296 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1297 * Called by upper layer device to delete the station created
1298 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1299 * the station and priv is the structure returned by the add
1300 * operation.
1301 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1302 * int queue_index, u32 maxrate);
1303 * Called when a user wants to set a max-rate limitation of specific
1304 * TX queue.
1305 * int (*ndo_get_iflink)(const struct net_device *dev);
1306 * Called to get the iflink value of this device.
1307 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1308 * This function is used to get egress tunnel information for given skb.
1309 * This is useful for retrieving outer tunnel header parameters while
1310 * sampling packet.
1311 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1312 * This function is used to specify the headroom that the skb must
1313 * consider when allocation skb during packet reception. Setting
1314 * appropriate rx headroom value allows avoiding skb head copy on
1315 * forward. Setting a negative value resets the rx headroom to the
1316 * default value.
1317 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1318 * This function is used to set or query state related to XDP on the
1319 * netdevice and manage BPF offload. See definition of
1320 * enum bpf_netdev_command for details.
1321 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1322 * u32 flags);
1323 * This function is used to submit @n XDP packets for transmit on a
1324 * netdevice. Returns number of frames successfully transmitted, frames
1325 * that got dropped are freed/returned via xdp_return_frame().
1326 * Returns negative number, means general error invoking ndo, meaning
1327 * no frames were xmit'ed and core-caller will free all frames.
1328 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1329 * struct xdp_buff *xdp);
1330 * Get the xmit slave of master device based on the xdp_buff.
1331 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1332 * This function is used to wake up the softirq, ksoftirqd or kthread
1333 * responsible for sending and/or receiving packets on a specific
1334 * queue id bound to an AF_XDP socket. The flags field specifies if
1335 * only RX, only Tx, or both should be woken up using the flags
1336 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1337 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1338 * int cmd);
1339 * Add, change, delete or get information on an IPv4 tunnel.
1340 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1341 * If a device is paired with a peer device, return the peer instance.
1342 * The caller must be under RCU read context.
1343 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1344 * Get the forwarding path to reach the real device from the HW destination address
1345 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1346 * const struct skb_shared_hwtstamps *hwtstamps,
1347 * bool cycles);
1348 * Get hardware timestamp based on normal/adjustable time or free running
1349 * cycle counter. This function is required if physical clock supports a
1350 * free running cycle counter.
1351 *
1352 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1353 * struct kernel_hwtstamp_config *kernel_config);
1354 * Get the currently configured hardware timestamping parameters for the
1355 * NIC device.
1356 *
1357 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1358 * struct kernel_hwtstamp_config *kernel_config,
1359 * struct netlink_ext_ack *extack);
1360 * Change the hardware timestamping parameters for NIC device.
1361 */
1362struct net_device_ops {
1363 int (*ndo_init)(struct net_device *dev);
1364 void (*ndo_uninit)(struct net_device *dev);
1365 int (*ndo_open)(struct net_device *dev);
1366 int (*ndo_stop)(struct net_device *dev);
1367 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1368 struct net_device *dev);
1369 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1370 struct net_device *dev,
1371 netdev_features_t features);
1372 u16 (*ndo_select_queue)(struct net_device *dev,
1373 struct sk_buff *skb,
1374 struct net_device *sb_dev);
1375 void (*ndo_change_rx_flags)(struct net_device *dev,
1376 int flags);
1377 void (*ndo_set_rx_mode)(struct net_device *dev);
1378 int (*ndo_set_mac_address)(struct net_device *dev,
1379 void *addr);
1380 int (*ndo_validate_addr)(struct net_device *dev);
1381 int (*ndo_do_ioctl)(struct net_device *dev,
1382 struct ifreq *ifr, int cmd);
1383 int (*ndo_eth_ioctl)(struct net_device *dev,
1384 struct ifreq *ifr, int cmd);
1385 int (*ndo_siocbond)(struct net_device *dev,
1386 struct ifreq *ifr, int cmd);
1387 int (*ndo_siocwandev)(struct net_device *dev,
1388 struct if_settings *ifs);
1389 int (*ndo_siocdevprivate)(struct net_device *dev,
1390 struct ifreq *ifr,
1391 void __user *data, int cmd);
1392 int (*ndo_set_config)(struct net_device *dev,
1393 struct ifmap *map);
1394 int (*ndo_change_mtu)(struct net_device *dev,
1395 int new_mtu);
1396 int (*ndo_neigh_setup)(struct net_device *dev,
1397 struct neigh_parms *);
1398 void (*ndo_tx_timeout) (struct net_device *dev,
1399 unsigned int txqueue);
1400
1401 void (*ndo_get_stats64)(struct net_device *dev,
1402 struct rtnl_link_stats64 *storage);
1403 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1404 int (*ndo_get_offload_stats)(int attr_id,
1405 const struct net_device *dev,
1406 void *attr_data);
1407 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1408
1409 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1410 __be16 proto, u16 vid);
1411 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1412 __be16 proto, u16 vid);
1413#ifdef CONFIG_NET_POLL_CONTROLLER
1414 void (*ndo_poll_controller)(struct net_device *dev);
1415 int (*ndo_netpoll_setup)(struct net_device *dev,
1416 struct netpoll_info *info);
1417 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1418#endif
1419 int (*ndo_set_vf_mac)(struct net_device *dev,
1420 int queue, u8 *mac);
1421 int (*ndo_set_vf_vlan)(struct net_device *dev,
1422 int queue, u16 vlan,
1423 u8 qos, __be16 proto);
1424 int (*ndo_set_vf_rate)(struct net_device *dev,
1425 int vf, int min_tx_rate,
1426 int max_tx_rate);
1427 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1428 int vf, bool setting);
1429 int (*ndo_set_vf_trust)(struct net_device *dev,
1430 int vf, bool setting);
1431 int (*ndo_get_vf_config)(struct net_device *dev,
1432 int vf,
1433 struct ifla_vf_info *ivf);
1434 int (*ndo_set_vf_link_state)(struct net_device *dev,
1435 int vf, int link_state);
1436 int (*ndo_get_vf_stats)(struct net_device *dev,
1437 int vf,
1438 struct ifla_vf_stats
1439 *vf_stats);
1440 int (*ndo_set_vf_port)(struct net_device *dev,
1441 int vf,
1442 struct nlattr *port[]);
1443 int (*ndo_get_vf_port)(struct net_device *dev,
1444 int vf, struct sk_buff *skb);
1445 int (*ndo_get_vf_guid)(struct net_device *dev,
1446 int vf,
1447 struct ifla_vf_guid *node_guid,
1448 struct ifla_vf_guid *port_guid);
1449 int (*ndo_set_vf_guid)(struct net_device *dev,
1450 int vf, u64 guid,
1451 int guid_type);
1452 int (*ndo_set_vf_rss_query_en)(
1453 struct net_device *dev,
1454 int vf, bool setting);
1455 int (*ndo_setup_tc)(struct net_device *dev,
1456 enum tc_setup_type type,
1457 void *type_data);
1458#if IS_ENABLED(CONFIG_FCOE)
1459 int (*ndo_fcoe_enable)(struct net_device *dev);
1460 int (*ndo_fcoe_disable)(struct net_device *dev);
1461 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1462 u16 xid,
1463 struct scatterlist *sgl,
1464 unsigned int sgc);
1465 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1466 u16 xid);
1467 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1468 u16 xid,
1469 struct scatterlist *sgl,
1470 unsigned int sgc);
1471 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1472 struct netdev_fcoe_hbainfo *hbainfo);
1473#endif
1474
1475#if IS_ENABLED(CONFIG_LIBFCOE)
1476#define NETDEV_FCOE_WWNN 0
1477#define NETDEV_FCOE_WWPN 1
1478 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1479 u64 *wwn, int type);
1480#endif
1481
1482#ifdef CONFIG_RFS_ACCEL
1483 int (*ndo_rx_flow_steer)(struct net_device *dev,
1484 const struct sk_buff *skb,
1485 u16 rxq_index,
1486 u32 flow_id);
1487#endif
1488 int (*ndo_add_slave)(struct net_device *dev,
1489 struct net_device *slave_dev,
1490 struct netlink_ext_ack *extack);
1491 int (*ndo_del_slave)(struct net_device *dev,
1492 struct net_device *slave_dev);
1493 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1494 struct sk_buff *skb,
1495 bool all_slaves);
1496 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1497 struct sock *sk);
1498 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1499 netdev_features_t features);
1500 int (*ndo_set_features)(struct net_device *dev,
1501 netdev_features_t features);
1502 int (*ndo_neigh_construct)(struct net_device *dev,
1503 struct neighbour *n);
1504 void (*ndo_neigh_destroy)(struct net_device *dev,
1505 struct neighbour *n);
1506
1507 int (*ndo_fdb_add)(struct ndmsg *ndm,
1508 struct nlattr *tb[],
1509 struct net_device *dev,
1510 const unsigned char *addr,
1511 u16 vid,
1512 u16 flags,
1513 struct netlink_ext_ack *extack);
1514 int (*ndo_fdb_del)(struct ndmsg *ndm,
1515 struct nlattr *tb[],
1516 struct net_device *dev,
1517 const unsigned char *addr,
1518 u16 vid, struct netlink_ext_ack *extack);
1519 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1520 struct net_device *dev,
1521 struct netlink_ext_ack *extack);
1522 int (*ndo_fdb_dump)(struct sk_buff *skb,
1523 struct netlink_callback *cb,
1524 struct net_device *dev,
1525 struct net_device *filter_dev,
1526 int *idx);
1527 int (*ndo_fdb_get)(struct sk_buff *skb,
1528 struct nlattr *tb[],
1529 struct net_device *dev,
1530 const unsigned char *addr,
1531 u16 vid, u32 portid, u32 seq,
1532 struct netlink_ext_ack *extack);
1533 int (*ndo_mdb_add)(struct net_device *dev,
1534 struct nlattr *tb[],
1535 u16 nlmsg_flags,
1536 struct netlink_ext_ack *extack);
1537 int (*ndo_mdb_del)(struct net_device *dev,
1538 struct nlattr *tb[],
1539 struct netlink_ext_ack *extack);
1540 int (*ndo_mdb_del_bulk)(struct net_device *dev,
1541 struct nlattr *tb[],
1542 struct netlink_ext_ack *extack);
1543 int (*ndo_mdb_dump)(struct net_device *dev,
1544 struct sk_buff *skb,
1545 struct netlink_callback *cb);
1546 int (*ndo_mdb_get)(struct net_device *dev,
1547 struct nlattr *tb[], u32 portid,
1548 u32 seq,
1549 struct netlink_ext_ack *extack);
1550 int (*ndo_bridge_setlink)(struct net_device *dev,
1551 struct nlmsghdr *nlh,
1552 u16 flags,
1553 struct netlink_ext_ack *extack);
1554 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1555 u32 pid, u32 seq,
1556 struct net_device *dev,
1557 u32 filter_mask,
1558 int nlflags);
1559 int (*ndo_bridge_dellink)(struct net_device *dev,
1560 struct nlmsghdr *nlh,
1561 u16 flags);
1562 int (*ndo_change_carrier)(struct net_device *dev,
1563 bool new_carrier);
1564 int (*ndo_get_phys_port_id)(struct net_device *dev,
1565 struct netdev_phys_item_id *ppid);
1566 int (*ndo_get_port_parent_id)(struct net_device *dev,
1567 struct netdev_phys_item_id *ppid);
1568 int (*ndo_get_phys_port_name)(struct net_device *dev,
1569 char *name, size_t len);
1570 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1571 struct net_device *dev);
1572 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1573 void *priv);
1574
1575 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1576 int queue_index,
1577 u32 maxrate);
1578 int (*ndo_get_iflink)(const struct net_device *dev);
1579 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1580 struct sk_buff *skb);
1581 void (*ndo_set_rx_headroom)(struct net_device *dev,
1582 int needed_headroom);
1583 int (*ndo_bpf)(struct net_device *dev,
1584 struct netdev_bpf *bpf);
1585 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1586 struct xdp_frame **xdp,
1587 u32 flags);
1588 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1589 struct xdp_buff *xdp);
1590 int (*ndo_xsk_wakeup)(struct net_device *dev,
1591 u32 queue_id, u32 flags);
1592 int (*ndo_tunnel_ctl)(struct net_device *dev,
1593 struct ip_tunnel_parm_kern *p,
1594 int cmd);
1595 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1596 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1597 struct net_device_path *path);
1598 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1599 const struct skb_shared_hwtstamps *hwtstamps,
1600 bool cycles);
1601 int (*ndo_hwtstamp_get)(struct net_device *dev,
1602 struct kernel_hwtstamp_config *kernel_config);
1603 int (*ndo_hwtstamp_set)(struct net_device *dev,
1604 struct kernel_hwtstamp_config *kernel_config,
1605 struct netlink_ext_ack *extack);
1606};
1607
1608/**
1609 * enum netdev_priv_flags - &struct net_device priv_flags
1610 *
1611 * These are the &struct net_device, they are only set internally
1612 * by drivers and used in the kernel. These flags are invisible to
1613 * userspace; this means that the order of these flags can change
1614 * during any kernel release.
1615 *
1616 * You should add bitfield booleans after either net_device::priv_flags
1617 * (hotpath) or ::threaded (slowpath) instead of extending these flags.
1618 *
1619 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1620 * @IFF_EBRIDGE: Ethernet bridging device
1621 * @IFF_BONDING: bonding master or slave
1622 * @IFF_ISATAP: ISATAP interface (RFC4214)
1623 * @IFF_WAN_HDLC: WAN HDLC device
1624 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1625 * release skb->dst
1626 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1627 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1628 * @IFF_MACVLAN_PORT: device used as macvlan port
1629 * @IFF_BRIDGE_PORT: device used as bridge port
1630 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1631 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1632 * @IFF_UNICAST_FLT: Supports unicast filtering
1633 * @IFF_TEAM_PORT: device used as team port
1634 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1635 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1636 * change when it's running
1637 * @IFF_MACVLAN: Macvlan device
1638 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1639 * underlying stacked devices
1640 * @IFF_L3MDEV_MASTER: device is an L3 master device
1641 * @IFF_NO_QUEUE: device can run without qdisc attached
1642 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1643 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1644 * @IFF_TEAM: device is a team device
1645 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1646 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1647 * entity (i.e. the master device for bridged veth)
1648 * @IFF_MACSEC: device is a MACsec device
1649 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1650 * @IFF_FAILOVER: device is a failover master device
1651 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1652 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1653 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1654 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1655 * skb_headlen(skb) == 0 (data starts from frag0)
1656 */
1657enum netdev_priv_flags {
1658 IFF_802_1Q_VLAN = 1<<0,
1659 IFF_EBRIDGE = 1<<1,
1660 IFF_BONDING = 1<<2,
1661 IFF_ISATAP = 1<<3,
1662 IFF_WAN_HDLC = 1<<4,
1663 IFF_XMIT_DST_RELEASE = 1<<5,
1664 IFF_DONT_BRIDGE = 1<<6,
1665 IFF_DISABLE_NETPOLL = 1<<7,
1666 IFF_MACVLAN_PORT = 1<<8,
1667 IFF_BRIDGE_PORT = 1<<9,
1668 IFF_OVS_DATAPATH = 1<<10,
1669 IFF_TX_SKB_SHARING = 1<<11,
1670 IFF_UNICAST_FLT = 1<<12,
1671 IFF_TEAM_PORT = 1<<13,
1672 IFF_SUPP_NOFCS = 1<<14,
1673 IFF_LIVE_ADDR_CHANGE = 1<<15,
1674 IFF_MACVLAN = 1<<16,
1675 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1676 IFF_L3MDEV_MASTER = 1<<18,
1677 IFF_NO_QUEUE = 1<<19,
1678 IFF_OPENVSWITCH = 1<<20,
1679 IFF_L3MDEV_SLAVE = 1<<21,
1680 IFF_TEAM = 1<<22,
1681 IFF_RXFH_CONFIGURED = 1<<23,
1682 IFF_PHONY_HEADROOM = 1<<24,
1683 IFF_MACSEC = 1<<25,
1684 IFF_NO_RX_HANDLER = 1<<26,
1685 IFF_FAILOVER = 1<<27,
1686 IFF_FAILOVER_SLAVE = 1<<28,
1687 IFF_L3MDEV_RX_HANDLER = 1<<29,
1688 IFF_NO_ADDRCONF = BIT_ULL(30),
1689 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1690};
1691
1692/* Specifies the type of the struct net_device::ml_priv pointer */
1693enum netdev_ml_priv_type {
1694 ML_PRIV_NONE,
1695 ML_PRIV_CAN,
1696};
1697
1698enum netdev_stat_type {
1699 NETDEV_PCPU_STAT_NONE,
1700 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1701 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1702 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1703};
1704
1705enum netdev_reg_state {
1706 NETREG_UNINITIALIZED = 0,
1707 NETREG_REGISTERED, /* completed register_netdevice */
1708 NETREG_UNREGISTERING, /* called unregister_netdevice */
1709 NETREG_UNREGISTERED, /* completed unregister todo */
1710 NETREG_RELEASED, /* called free_netdev */
1711 NETREG_DUMMY, /* dummy device for NAPI poll */
1712};
1713
1714/**
1715 * struct net_device - The DEVICE structure.
1716 *
1717 * Actually, this whole structure is a big mistake. It mixes I/O
1718 * data with strictly "high-level" data, and it has to know about
1719 * almost every data structure used in the INET module.
1720 *
1721 * @priv_flags: flags invisible to userspace defined as bits, see
1722 * enum netdev_priv_flags for the definitions
1723 * @lltx: device supports lockless Tx. Deprecated for real HW
1724 * drivers. Mainly used by logical interfaces, such as
1725 * bonding and tunnels
1726 *
1727 * @name: This is the first field of the "visible" part of this structure
1728 * (i.e. as seen by users in the "Space.c" file). It is the name
1729 * of the interface.
1730 *
1731 * @name_node: Name hashlist node
1732 * @ifalias: SNMP alias
1733 * @mem_end: Shared memory end
1734 * @mem_start: Shared memory start
1735 * @base_addr: Device I/O address
1736 * @irq: Device IRQ number
1737 *
1738 * @state: Generic network queuing layer state, see netdev_state_t
1739 * @dev_list: The global list of network devices
1740 * @napi_list: List entry used for polling NAPI devices
1741 * @unreg_list: List entry when we are unregistering the
1742 * device; see the function unregister_netdev
1743 * @close_list: List entry used when we are closing the device
1744 * @ptype_all: Device-specific packet handlers for all protocols
1745 * @ptype_specific: Device-specific, protocol-specific packet handlers
1746 *
1747 * @adj_list: Directly linked devices, like slaves for bonding
1748 * @features: Currently active device features
1749 * @hw_features: User-changeable features
1750 *
1751 * @wanted_features: User-requested features
1752 * @vlan_features: Mask of features inheritable by VLAN devices
1753 *
1754 * @hw_enc_features: Mask of features inherited by encapsulating devices
1755 * This field indicates what encapsulation
1756 * offloads the hardware is capable of doing,
1757 * and drivers will need to set them appropriately.
1758 *
1759 * @mpls_features: Mask of features inheritable by MPLS
1760 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1761 *
1762 * @ifindex: interface index
1763 * @group: The group the device belongs to
1764 *
1765 * @stats: Statistics struct, which was left as a legacy, use
1766 * rtnl_link_stats64 instead
1767 *
1768 * @core_stats: core networking counters,
1769 * do not use this in drivers
1770 * @carrier_up_count: Number of times the carrier has been up
1771 * @carrier_down_count: Number of times the carrier has been down
1772 *
1773 * @wireless_handlers: List of functions to handle Wireless Extensions,
1774 * instead of ioctl,
1775 * see <net/iw_handler.h> for details.
1776 * @wireless_data: Instance data managed by the core of wireless extensions
1777 *
1778 * @netdev_ops: Includes several pointers to callbacks,
1779 * if one wants to override the ndo_*() functions
1780 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1781 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks.
1782 * @ethtool_ops: Management operations
1783 * @l3mdev_ops: Layer 3 master device operations
1784 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1785 * discovery handling. Necessary for e.g. 6LoWPAN.
1786 * @xfrmdev_ops: Transformation offload operations
1787 * @tlsdev_ops: Transport Layer Security offload operations
1788 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1789 * of Layer 2 headers.
1790 *
1791 * @flags: Interface flags (a la BSD)
1792 * @xdp_features: XDP capability supported by the device
1793 * @gflags: Global flags ( kept as legacy )
1794 * @priv_len: Size of the ->priv flexible array
1795 * @priv: Flexible array containing private data
1796 * @operstate: RFC2863 operstate
1797 * @link_mode: Mapping policy to operstate
1798 * @if_port: Selectable AUI, TP, ...
1799 * @dma: DMA channel
1800 * @mtu: Interface MTU value
1801 * @min_mtu: Interface Minimum MTU value
1802 * @max_mtu: Interface Maximum MTU value
1803 * @type: Interface hardware type
1804 * @hard_header_len: Maximum hardware header length.
1805 * @min_header_len: Minimum hardware header length
1806 *
1807 * @needed_headroom: Extra headroom the hardware may need, but not in all
1808 * cases can this be guaranteed
1809 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1810 * cases can this be guaranteed. Some cases also use
1811 * LL_MAX_HEADER instead to allocate the skb
1812 *
1813 * interface address info:
1814 *
1815 * @perm_addr: Permanent hw address
1816 * @addr_assign_type: Hw address assignment type
1817 * @addr_len: Hardware address length
1818 * @upper_level: Maximum depth level of upper devices.
1819 * @lower_level: Maximum depth level of lower devices.
1820 * @neigh_priv_len: Used in neigh_alloc()
1821 * @dev_id: Used to differentiate devices that share
1822 * the same link layer address
1823 * @dev_port: Used to differentiate devices that share
1824 * the same function
1825 * @addr_list_lock: XXX: need comments on this one
1826 * @name_assign_type: network interface name assignment type
1827 * @uc_promisc: Counter that indicates promiscuous mode
1828 * has been enabled due to the need to listen to
1829 * additional unicast addresses in a device that
1830 * does not implement ndo_set_rx_mode()
1831 * @uc: unicast mac addresses
1832 * @mc: multicast mac addresses
1833 * @dev_addrs: list of device hw addresses
1834 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1835 * @promiscuity: Number of times the NIC is told to work in
1836 * promiscuous mode; if it becomes 0 the NIC will
1837 * exit promiscuous mode
1838 * @allmulti: Counter, enables or disables allmulticast mode
1839 *
1840 * @vlan_info: VLAN info
1841 * @dsa_ptr: dsa specific data
1842 * @tipc_ptr: TIPC specific data
1843 * @atalk_ptr: AppleTalk link
1844 * @ip_ptr: IPv4 specific data
1845 * @ip6_ptr: IPv6 specific data
1846 * @ax25_ptr: AX.25 specific data
1847 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1848 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1849 * device struct
1850 * @mpls_ptr: mpls_dev struct pointer
1851 * @mctp_ptr: MCTP specific data
1852 *
1853 * @dev_addr: Hw address (before bcast,
1854 * because most packets are unicast)
1855 *
1856 * @_rx: Array of RX queues
1857 * @num_rx_queues: Number of RX queues
1858 * allocated at register_netdev() time
1859 * @real_num_rx_queues: Number of RX queues currently active in device
1860 * @xdp_prog: XDP sockets filter program pointer
1861 * @gro_flush_timeout: timeout for GRO layer in NAPI
1862 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1863 * allow to avoid NIC hard IRQ, on busy queues.
1864 *
1865 * @rx_handler: handler for received packets
1866 * @rx_handler_data: XXX: need comments on this one
1867 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1868 * @ingress_queue: XXX: need comments on this one
1869 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1870 * @broadcast: hw bcast address
1871 *
1872 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1873 * indexed by RX queue number. Assigned by driver.
1874 * This must only be set if the ndo_rx_flow_steer
1875 * operation is defined
1876 * @index_hlist: Device index hash chain
1877 *
1878 * @_tx: Array of TX queues
1879 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1880 * @real_num_tx_queues: Number of TX queues currently active in device
1881 * @qdisc: Root qdisc from userspace point of view
1882 * @tx_queue_len: Max frames per queue allowed
1883 * @tx_global_lock: XXX: need comments on this one
1884 * @xdp_bulkq: XDP device bulk queue
1885 * @xps_maps: all CPUs/RXQs maps for XPS device
1886 *
1887 * @xps_maps: XXX: need comments on this one
1888 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1889 * @nf_hooks_egress: netfilter hooks executed for egress packets
1890 * @qdisc_hash: qdisc hash table
1891 * @watchdog_timeo: Represents the timeout that is used by
1892 * the watchdog (see dev_watchdog())
1893 * @watchdog_timer: List of timers
1894 *
1895 * @proto_down_reason: reason a netdev interface is held down
1896 * @pcpu_refcnt: Number of references to this device
1897 * @dev_refcnt: Number of references to this device
1898 * @refcnt_tracker: Tracker directory for tracked references to this device
1899 * @todo_list: Delayed register/unregister
1900 * @link_watch_list: XXX: need comments on this one
1901 *
1902 * @reg_state: Register/unregister state machine
1903 * @dismantle: Device is going to be freed
1904 * @rtnl_link_state: This enum represents the phases of creating
1905 * a new link
1906 *
1907 * @needs_free_netdev: Should unregister perform free_netdev?
1908 * @priv_destructor: Called from unregister
1909 * @npinfo: XXX: need comments on this one
1910 * @nd_net: Network namespace this network device is inside
1911 *
1912 * @ml_priv: Mid-layer private
1913 * @ml_priv_type: Mid-layer private type
1914 *
1915 * @pcpu_stat_type: Type of device statistics which the core should
1916 * allocate/free: none, lstats, tstats, dstats. none
1917 * means the driver is handling statistics allocation/
1918 * freeing internally.
1919 * @lstats: Loopback statistics: packets, bytes
1920 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes
1921 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes
1922 *
1923 * @garp_port: GARP
1924 * @mrp_port: MRP
1925 *
1926 * @dm_private: Drop monitor private
1927 *
1928 * @dev: Class/net/name entry
1929 * @sysfs_groups: Space for optional device, statistics and wireless
1930 * sysfs groups
1931 *
1932 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1933 * @rtnl_link_ops: Rtnl_link_ops
1934 * @stat_ops: Optional ops for queue-aware statistics
1935 * @queue_mgmt_ops: Optional ops for queue management
1936 *
1937 * @gso_max_size: Maximum size of generic segmentation offload
1938 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1939 * @gso_max_segs: Maximum number of segments that can be passed to the
1940 * NIC for GSO
1941 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1942 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
1943 * for IPv4.
1944 *
1945 * @dcbnl_ops: Data Center Bridging netlink ops
1946 * @num_tc: Number of traffic classes in the net device
1947 * @tc_to_txq: XXX: need comments on this one
1948 * @prio_tc_map: XXX: need comments on this one
1949 *
1950 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1951 *
1952 * @priomap: XXX: need comments on this one
1953 * @link_topo: Physical link topology tracking attached PHYs
1954 * @phydev: Physical device may attach itself
1955 * for hardware timestamping
1956 * @sfp_bus: attached &struct sfp_bus structure.
1957 *
1958 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1959 *
1960 * @proto_down: protocol port state information can be sent to the
1961 * switch driver and used to set the phys state of the
1962 * switch port.
1963 *
1964 * @threaded: napi threaded mode is enabled
1965 *
1966 * @see_all_hwtstamp_requests: device wants to see calls to
1967 * ndo_hwtstamp_set() for all timestamp requests
1968 * regardless of source, even if those aren't
1969 * HWTSTAMP_SOURCE_NETDEV
1970 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN
1971 * @netns_local: interface can't change network namespaces
1972 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes
1973 *
1974 * @net_notifier_list: List of per-net netdev notifier block
1975 * that follow this device when it is moved
1976 * to another network namespace.
1977 *
1978 * @macsec_ops: MACsec offloading ops
1979 *
1980 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1981 * offload capabilities of the device
1982 * @udp_tunnel_nic: UDP tunnel offload state
1983 * @ethtool: ethtool related state
1984 * @xdp_state: stores info on attached XDP BPF programs
1985 *
1986 * @nested_level: Used as a parameter of spin_lock_nested() of
1987 * dev->addr_list_lock.
1988 * @unlink_list: As netif_addr_lock() can be called recursively,
1989 * keep a list of interfaces to be deleted.
1990 * @gro_max_size: Maximum size of aggregated packet in generic
1991 * receive offload (GRO)
1992 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
1993 * receive offload (GRO), for IPv4.
1994 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
1995 * zero copy driver
1996 *
1997 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
1998 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
1999 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2000 * @dev_registered_tracker: tracker for reference held while
2001 * registered
2002 * @offload_xstats_l3: L3 HW stats for this netdevice.
2003 *
2004 * @devlink_port: Pointer to related devlink port structure.
2005 * Assigned by a driver before netdev registration using
2006 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2007 * during the time netdevice is registered.
2008 *
2009 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2010 * where the clock is recovered.
2011 *
2012 * FIXME: cleanup struct net_device such that network protocol info
2013 * moves out.
2014 */
2015
2016struct net_device {
2017 /* Cacheline organization can be found documented in
2018 * Documentation/networking/net_cachelines/net_device.rst.
2019 * Please update the document when adding new fields.
2020 */
2021
2022 /* TX read-mostly hotpath */
2023 __cacheline_group_begin(net_device_read_tx);
2024 struct_group(priv_flags_fast,
2025 unsigned long priv_flags:32;
2026 unsigned long lltx:1;
2027 );
2028 const struct net_device_ops *netdev_ops;
2029 const struct header_ops *header_ops;
2030 struct netdev_queue *_tx;
2031 netdev_features_t gso_partial_features;
2032 unsigned int real_num_tx_queues;
2033 unsigned int gso_max_size;
2034 unsigned int gso_ipv4_max_size;
2035 u16 gso_max_segs;
2036 s16 num_tc;
2037 /* Note : dev->mtu is often read without holding a lock.
2038 * Writers usually hold RTNL.
2039 * It is recommended to use READ_ONCE() to annotate the reads,
2040 * and to use WRITE_ONCE() to annotate the writes.
2041 */
2042 unsigned int mtu;
2043 unsigned short needed_headroom;
2044 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2045#ifdef CONFIG_XPS
2046 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2047#endif
2048#ifdef CONFIG_NETFILTER_EGRESS
2049 struct nf_hook_entries __rcu *nf_hooks_egress;
2050#endif
2051#ifdef CONFIG_NET_XGRESS
2052 struct bpf_mprog_entry __rcu *tcx_egress;
2053#endif
2054 __cacheline_group_end(net_device_read_tx);
2055
2056 /* TXRX read-mostly hotpath */
2057 __cacheline_group_begin(net_device_read_txrx);
2058 union {
2059 struct pcpu_lstats __percpu *lstats;
2060 struct pcpu_sw_netstats __percpu *tstats;
2061 struct pcpu_dstats __percpu *dstats;
2062 };
2063 unsigned long state;
2064 unsigned int flags;
2065 unsigned short hard_header_len;
2066 netdev_features_t features;
2067 struct inet6_dev __rcu *ip6_ptr;
2068 __cacheline_group_end(net_device_read_txrx);
2069
2070 /* RX read-mostly hotpath */
2071 __cacheline_group_begin(net_device_read_rx);
2072 struct bpf_prog __rcu *xdp_prog;
2073 struct list_head ptype_specific;
2074 int ifindex;
2075 unsigned int real_num_rx_queues;
2076 struct netdev_rx_queue *_rx;
2077 unsigned long gro_flush_timeout;
2078 u32 napi_defer_hard_irqs;
2079 unsigned int gro_max_size;
2080 unsigned int gro_ipv4_max_size;
2081 rx_handler_func_t __rcu *rx_handler;
2082 void __rcu *rx_handler_data;
2083 possible_net_t nd_net;
2084#ifdef CONFIG_NETPOLL
2085 struct netpoll_info __rcu *npinfo;
2086#endif
2087#ifdef CONFIG_NET_XGRESS
2088 struct bpf_mprog_entry __rcu *tcx_ingress;
2089#endif
2090 __cacheline_group_end(net_device_read_rx);
2091
2092 char name[IFNAMSIZ];
2093 struct netdev_name_node *name_node;
2094 struct dev_ifalias __rcu *ifalias;
2095 /*
2096 * I/O specific fields
2097 * FIXME: Merge these and struct ifmap into one
2098 */
2099 unsigned long mem_end;
2100 unsigned long mem_start;
2101 unsigned long base_addr;
2102
2103 /*
2104 * Some hardware also needs these fields (state,dev_list,
2105 * napi_list,unreg_list,close_list) but they are not
2106 * part of the usual set specified in Space.c.
2107 */
2108
2109
2110 struct list_head dev_list;
2111 struct list_head napi_list;
2112 struct list_head unreg_list;
2113 struct list_head close_list;
2114 struct list_head ptype_all;
2115
2116 struct {
2117 struct list_head upper;
2118 struct list_head lower;
2119 } adj_list;
2120
2121 /* Read-mostly cache-line for fast-path access */
2122 xdp_features_t xdp_features;
2123 const struct xdp_metadata_ops *xdp_metadata_ops;
2124 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2125 unsigned short gflags;
2126
2127 unsigned short needed_tailroom;
2128
2129 netdev_features_t hw_features;
2130 netdev_features_t wanted_features;
2131 netdev_features_t vlan_features;
2132 netdev_features_t hw_enc_features;
2133 netdev_features_t mpls_features;
2134
2135 unsigned int min_mtu;
2136 unsigned int max_mtu;
2137 unsigned short type;
2138 unsigned char min_header_len;
2139 unsigned char name_assign_type;
2140
2141 int group;
2142
2143 struct net_device_stats stats; /* not used by modern drivers */
2144
2145 struct net_device_core_stats __percpu *core_stats;
2146
2147 /* Stats to monitor link on/off, flapping */
2148 atomic_t carrier_up_count;
2149 atomic_t carrier_down_count;
2150
2151#ifdef CONFIG_WIRELESS_EXT
2152 const struct iw_handler_def *wireless_handlers;
2153 struct iw_public_data *wireless_data;
2154#endif
2155 const struct ethtool_ops *ethtool_ops;
2156#ifdef CONFIG_NET_L3_MASTER_DEV
2157 const struct l3mdev_ops *l3mdev_ops;
2158#endif
2159#if IS_ENABLED(CONFIG_IPV6)
2160 const struct ndisc_ops *ndisc_ops;
2161#endif
2162
2163#ifdef CONFIG_XFRM_OFFLOAD
2164 const struct xfrmdev_ops *xfrmdev_ops;
2165#endif
2166
2167#if IS_ENABLED(CONFIG_TLS_DEVICE)
2168 const struct tlsdev_ops *tlsdev_ops;
2169#endif
2170
2171 unsigned int operstate;
2172 unsigned char link_mode;
2173
2174 unsigned char if_port;
2175 unsigned char dma;
2176
2177 /* Interface address info. */
2178 unsigned char perm_addr[MAX_ADDR_LEN];
2179 unsigned char addr_assign_type;
2180 unsigned char addr_len;
2181 unsigned char upper_level;
2182 unsigned char lower_level;
2183
2184 unsigned short neigh_priv_len;
2185 unsigned short dev_id;
2186 unsigned short dev_port;
2187 int irq;
2188 u32 priv_len;
2189
2190 spinlock_t addr_list_lock;
2191
2192 struct netdev_hw_addr_list uc;
2193 struct netdev_hw_addr_list mc;
2194 struct netdev_hw_addr_list dev_addrs;
2195
2196#ifdef CONFIG_SYSFS
2197 struct kset *queues_kset;
2198#endif
2199#ifdef CONFIG_LOCKDEP
2200 struct list_head unlink_list;
2201#endif
2202 unsigned int promiscuity;
2203 unsigned int allmulti;
2204 bool uc_promisc;
2205#ifdef CONFIG_LOCKDEP
2206 unsigned char nested_level;
2207#endif
2208
2209
2210 /* Protocol-specific pointers */
2211 struct in_device __rcu *ip_ptr;
2212#if IS_ENABLED(CONFIG_VLAN_8021Q)
2213 struct vlan_info __rcu *vlan_info;
2214#endif
2215#if IS_ENABLED(CONFIG_NET_DSA)
2216 struct dsa_port *dsa_ptr;
2217#endif
2218#if IS_ENABLED(CONFIG_TIPC)
2219 struct tipc_bearer __rcu *tipc_ptr;
2220#endif
2221#if IS_ENABLED(CONFIG_ATALK)
2222 void *atalk_ptr;
2223#endif
2224#if IS_ENABLED(CONFIG_AX25)
2225 void *ax25_ptr;
2226#endif
2227#if IS_ENABLED(CONFIG_CFG80211)
2228 struct wireless_dev *ieee80211_ptr;
2229#endif
2230#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2231 struct wpan_dev *ieee802154_ptr;
2232#endif
2233#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2234 struct mpls_dev __rcu *mpls_ptr;
2235#endif
2236#if IS_ENABLED(CONFIG_MCTP)
2237 struct mctp_dev __rcu *mctp_ptr;
2238#endif
2239
2240/*
2241 * Cache lines mostly used on receive path (including eth_type_trans())
2242 */
2243 /* Interface address info used in eth_type_trans() */
2244 const unsigned char *dev_addr;
2245
2246 unsigned int num_rx_queues;
2247#define GRO_LEGACY_MAX_SIZE 65536u
2248/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2249 * and shinfo->gso_segs is a 16bit field.
2250 */
2251#define GRO_MAX_SIZE (8 * 65535u)
2252 unsigned int xdp_zc_max_segs;
2253 struct netdev_queue __rcu *ingress_queue;
2254#ifdef CONFIG_NETFILTER_INGRESS
2255 struct nf_hook_entries __rcu *nf_hooks_ingress;
2256#endif
2257
2258 unsigned char broadcast[MAX_ADDR_LEN];
2259#ifdef CONFIG_RFS_ACCEL
2260 struct cpu_rmap *rx_cpu_rmap;
2261#endif
2262 struct hlist_node index_hlist;
2263
2264/*
2265 * Cache lines mostly used on transmit path
2266 */
2267 unsigned int num_tx_queues;
2268 struct Qdisc __rcu *qdisc;
2269 unsigned int tx_queue_len;
2270 spinlock_t tx_global_lock;
2271
2272 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2273
2274#ifdef CONFIG_NET_SCHED
2275 DECLARE_HASHTABLE (qdisc_hash, 4);
2276#endif
2277 /* These may be needed for future network-power-down code. */
2278 struct timer_list watchdog_timer;
2279 int watchdog_timeo;
2280
2281 u32 proto_down_reason;
2282
2283 struct list_head todo_list;
2284
2285#ifdef CONFIG_PCPU_DEV_REFCNT
2286 int __percpu *pcpu_refcnt;
2287#else
2288 refcount_t dev_refcnt;
2289#endif
2290 struct ref_tracker_dir refcnt_tracker;
2291
2292 struct list_head link_watch_list;
2293
2294 u8 reg_state;
2295
2296 bool dismantle;
2297
2298 enum {
2299 RTNL_LINK_INITIALIZED,
2300 RTNL_LINK_INITIALIZING,
2301 } rtnl_link_state:16;
2302
2303 bool needs_free_netdev;
2304 void (*priv_destructor)(struct net_device *dev);
2305
2306 /* mid-layer private */
2307 void *ml_priv;
2308 enum netdev_ml_priv_type ml_priv_type;
2309
2310 enum netdev_stat_type pcpu_stat_type:8;
2311
2312#if IS_ENABLED(CONFIG_GARP)
2313 struct garp_port __rcu *garp_port;
2314#endif
2315#if IS_ENABLED(CONFIG_MRP)
2316 struct mrp_port __rcu *mrp_port;
2317#endif
2318#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2319 struct dm_hw_stat_delta __rcu *dm_private;
2320#endif
2321 struct device dev;
2322 const struct attribute_group *sysfs_groups[4];
2323 const struct attribute_group *sysfs_rx_queue_group;
2324
2325 const struct rtnl_link_ops *rtnl_link_ops;
2326
2327 const struct netdev_stat_ops *stat_ops;
2328
2329 const struct netdev_queue_mgmt_ops *queue_mgmt_ops;
2330
2331 /* for setting kernel sock attribute on TCP connection setup */
2332#define GSO_MAX_SEGS 65535u
2333#define GSO_LEGACY_MAX_SIZE 65536u
2334/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2335 * and shinfo->gso_segs is a 16bit field.
2336 */
2337#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2338
2339#define TSO_LEGACY_MAX_SIZE 65536
2340#define TSO_MAX_SIZE UINT_MAX
2341 unsigned int tso_max_size;
2342#define TSO_MAX_SEGS U16_MAX
2343 u16 tso_max_segs;
2344
2345#ifdef CONFIG_DCB
2346 const struct dcbnl_rtnl_ops *dcbnl_ops;
2347#endif
2348 u8 prio_tc_map[TC_BITMASK + 1];
2349
2350#if IS_ENABLED(CONFIG_FCOE)
2351 unsigned int fcoe_ddp_xid;
2352#endif
2353#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2354 struct netprio_map __rcu *priomap;
2355#endif
2356 struct phy_link_topology *link_topo;
2357 struct phy_device *phydev;
2358 struct sfp_bus *sfp_bus;
2359 struct lock_class_key *qdisc_tx_busylock;
2360 bool proto_down;
2361 bool threaded;
2362
2363 /* priv_flags_slow, ungrouped to save space */
2364 unsigned long see_all_hwtstamp_requests:1;
2365 unsigned long change_proto_down:1;
2366 unsigned long netns_local:1;
2367 unsigned long fcoe_mtu:1;
2368
2369 struct list_head net_notifier_list;
2370
2371#if IS_ENABLED(CONFIG_MACSEC)
2372 /* MACsec management functions */
2373 const struct macsec_ops *macsec_ops;
2374#endif
2375 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2376 struct udp_tunnel_nic *udp_tunnel_nic;
2377
2378 struct ethtool_netdev_state *ethtool;
2379
2380 /* protected by rtnl_lock */
2381 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2382
2383 u8 dev_addr_shadow[MAX_ADDR_LEN];
2384 netdevice_tracker linkwatch_dev_tracker;
2385 netdevice_tracker watchdog_dev_tracker;
2386 netdevice_tracker dev_registered_tracker;
2387 struct rtnl_hw_stats64 *offload_xstats_l3;
2388
2389 struct devlink_port *devlink_port;
2390
2391#if IS_ENABLED(CONFIG_DPLL)
2392 struct dpll_pin __rcu *dpll_pin;
2393#endif
2394#if IS_ENABLED(CONFIG_PAGE_POOL)
2395 /** @page_pools: page pools created for this netdevice */
2396 struct hlist_head page_pools;
2397#endif
2398
2399 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */
2400 struct dim_irq_moder *irq_moder;
2401
2402 u8 priv[] ____cacheline_aligned
2403 __counted_by(priv_len);
2404} ____cacheline_aligned;
2405#define to_net_dev(d) container_of(d, struct net_device, dev)
2406
2407/*
2408 * Driver should use this to assign devlink port instance to a netdevice
2409 * before it registers the netdevice. Therefore devlink_port is static
2410 * during the netdev lifetime after it is registered.
2411 */
2412#define SET_NETDEV_DEVLINK_PORT(dev, port) \
2413({ \
2414 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2415 ((dev)->devlink_port = (port)); \
2416})
2417
2418static inline bool netif_elide_gro(const struct net_device *dev)
2419{
2420 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2421 return true;
2422 return false;
2423}
2424
2425#define NETDEV_ALIGN 32
2426
2427static inline
2428int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2429{
2430 return dev->prio_tc_map[prio & TC_BITMASK];
2431}
2432
2433static inline
2434int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2435{
2436 if (tc >= dev->num_tc)
2437 return -EINVAL;
2438
2439 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2440 return 0;
2441}
2442
2443int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2444void netdev_reset_tc(struct net_device *dev);
2445int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2446int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2447
2448static inline
2449int netdev_get_num_tc(struct net_device *dev)
2450{
2451 return dev->num_tc;
2452}
2453
2454static inline void net_prefetch(void *p)
2455{
2456 prefetch(p);
2457#if L1_CACHE_BYTES < 128
2458 prefetch((u8 *)p + L1_CACHE_BYTES);
2459#endif
2460}
2461
2462static inline void net_prefetchw(void *p)
2463{
2464 prefetchw(p);
2465#if L1_CACHE_BYTES < 128
2466 prefetchw((u8 *)p + L1_CACHE_BYTES);
2467#endif
2468}
2469
2470void netdev_unbind_sb_channel(struct net_device *dev,
2471 struct net_device *sb_dev);
2472int netdev_bind_sb_channel_queue(struct net_device *dev,
2473 struct net_device *sb_dev,
2474 u8 tc, u16 count, u16 offset);
2475int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2476static inline int netdev_get_sb_channel(struct net_device *dev)
2477{
2478 return max_t(int, -dev->num_tc, 0);
2479}
2480
2481static inline
2482struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2483 unsigned int index)
2484{
2485 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2486 return &dev->_tx[index];
2487}
2488
2489static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2490 const struct sk_buff *skb)
2491{
2492 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2493}
2494
2495static inline void netdev_for_each_tx_queue(struct net_device *dev,
2496 void (*f)(struct net_device *,
2497 struct netdev_queue *,
2498 void *),
2499 void *arg)
2500{
2501 unsigned int i;
2502
2503 for (i = 0; i < dev->num_tx_queues; i++)
2504 f(dev, &dev->_tx[i], arg);
2505}
2506
2507#define netdev_lockdep_set_classes(dev) \
2508{ \
2509 static struct lock_class_key qdisc_tx_busylock_key; \
2510 static struct lock_class_key qdisc_xmit_lock_key; \
2511 static struct lock_class_key dev_addr_list_lock_key; \
2512 unsigned int i; \
2513 \
2514 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2515 lockdep_set_class(&(dev)->addr_list_lock, \
2516 &dev_addr_list_lock_key); \
2517 for (i = 0; i < (dev)->num_tx_queues; i++) \
2518 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2519 &qdisc_xmit_lock_key); \
2520}
2521
2522u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2523 struct net_device *sb_dev);
2524struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2525 struct sk_buff *skb,
2526 struct net_device *sb_dev);
2527
2528/* returns the headroom that the master device needs to take in account
2529 * when forwarding to this dev
2530 */
2531static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2532{
2533 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2534}
2535
2536static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2537{
2538 if (dev->netdev_ops->ndo_set_rx_headroom)
2539 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2540}
2541
2542/* set the device rx headroom to the dev's default */
2543static inline void netdev_reset_rx_headroom(struct net_device *dev)
2544{
2545 netdev_set_rx_headroom(dev, -1);
2546}
2547
2548static inline void *netdev_get_ml_priv(struct net_device *dev,
2549 enum netdev_ml_priv_type type)
2550{
2551 if (dev->ml_priv_type != type)
2552 return NULL;
2553
2554 return dev->ml_priv;
2555}
2556
2557static inline void netdev_set_ml_priv(struct net_device *dev,
2558 void *ml_priv,
2559 enum netdev_ml_priv_type type)
2560{
2561 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2562 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2563 dev->ml_priv_type, type);
2564 WARN(!dev->ml_priv_type && dev->ml_priv,
2565 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2566
2567 dev->ml_priv = ml_priv;
2568 dev->ml_priv_type = type;
2569}
2570
2571/*
2572 * Net namespace inlines
2573 */
2574static inline
2575struct net *dev_net(const struct net_device *dev)
2576{
2577 return read_pnet(&dev->nd_net);
2578}
2579
2580static inline
2581void dev_net_set(struct net_device *dev, struct net *net)
2582{
2583 write_pnet(&dev->nd_net, net);
2584}
2585
2586/**
2587 * netdev_priv - access network device private data
2588 * @dev: network device
2589 *
2590 * Get network device private data
2591 */
2592static inline void *netdev_priv(const struct net_device *dev)
2593{
2594 return (void *)dev->priv;
2595}
2596
2597/* Set the sysfs physical device reference for the network logical device
2598 * if set prior to registration will cause a symlink during initialization.
2599 */
2600#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2601
2602/* Set the sysfs device type for the network logical device to allow
2603 * fine-grained identification of different network device types. For
2604 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2605 */
2606#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2607
2608void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2609 enum netdev_queue_type type,
2610 struct napi_struct *napi);
2611
2612static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2613{
2614 napi->irq = irq;
2615}
2616
2617/* Default NAPI poll() weight
2618 * Device drivers are strongly advised to not use bigger value
2619 */
2620#define NAPI_POLL_WEIGHT 64
2621
2622void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2623 int (*poll)(struct napi_struct *, int), int weight);
2624
2625/**
2626 * netif_napi_add() - initialize a NAPI context
2627 * @dev: network device
2628 * @napi: NAPI context
2629 * @poll: polling function
2630 *
2631 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2632 * *any* of the other NAPI-related functions.
2633 */
2634static inline void
2635netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2636 int (*poll)(struct napi_struct *, int))
2637{
2638 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2639}
2640
2641static inline void
2642netif_napi_add_tx_weight(struct net_device *dev,
2643 struct napi_struct *napi,
2644 int (*poll)(struct napi_struct *, int),
2645 int weight)
2646{
2647 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2648 netif_napi_add_weight(dev, napi, poll, weight);
2649}
2650
2651/**
2652 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2653 * @dev: network device
2654 * @napi: NAPI context
2655 * @poll: polling function
2656 *
2657 * This variant of netif_napi_add() should be used from drivers using NAPI
2658 * to exclusively poll a TX queue.
2659 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2660 */
2661static inline void netif_napi_add_tx(struct net_device *dev,
2662 struct napi_struct *napi,
2663 int (*poll)(struct napi_struct *, int))
2664{
2665 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2666}
2667
2668/**
2669 * __netif_napi_del - remove a NAPI context
2670 * @napi: NAPI context
2671 *
2672 * Warning: caller must observe RCU grace period before freeing memory
2673 * containing @napi. Drivers might want to call this helper to combine
2674 * all the needed RCU grace periods into a single one.
2675 */
2676void __netif_napi_del(struct napi_struct *napi);
2677
2678/**
2679 * netif_napi_del - remove a NAPI context
2680 * @napi: NAPI context
2681 *
2682 * netif_napi_del() removes a NAPI context from the network device NAPI list
2683 */
2684static inline void netif_napi_del(struct napi_struct *napi)
2685{
2686 __netif_napi_del(napi);
2687 synchronize_net();
2688}
2689
2690struct packet_type {
2691 __be16 type; /* This is really htons(ether_type). */
2692 bool ignore_outgoing;
2693 struct net_device *dev; /* NULL is wildcarded here */
2694 netdevice_tracker dev_tracker;
2695 int (*func) (struct sk_buff *,
2696 struct net_device *,
2697 struct packet_type *,
2698 struct net_device *);
2699 void (*list_func) (struct list_head *,
2700 struct packet_type *,
2701 struct net_device *);
2702 bool (*id_match)(struct packet_type *ptype,
2703 struct sock *sk);
2704 struct net *af_packet_net;
2705 void *af_packet_priv;
2706 struct list_head list;
2707};
2708
2709struct offload_callbacks {
2710 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2711 netdev_features_t features);
2712 struct sk_buff *(*gro_receive)(struct list_head *head,
2713 struct sk_buff *skb);
2714 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2715};
2716
2717struct packet_offload {
2718 __be16 type; /* This is really htons(ether_type). */
2719 u16 priority;
2720 struct offload_callbacks callbacks;
2721 struct list_head list;
2722};
2723
2724/* often modified stats are per-CPU, other are shared (netdev->stats) */
2725struct pcpu_sw_netstats {
2726 u64_stats_t rx_packets;
2727 u64_stats_t rx_bytes;
2728 u64_stats_t tx_packets;
2729 u64_stats_t tx_bytes;
2730 struct u64_stats_sync syncp;
2731} __aligned(4 * sizeof(u64));
2732
2733struct pcpu_dstats {
2734 u64_stats_t rx_packets;
2735 u64_stats_t rx_bytes;
2736 u64_stats_t rx_drops;
2737 u64_stats_t tx_packets;
2738 u64_stats_t tx_bytes;
2739 u64_stats_t tx_drops;
2740 struct u64_stats_sync syncp;
2741} __aligned(8 * sizeof(u64));
2742
2743struct pcpu_lstats {
2744 u64_stats_t packets;
2745 u64_stats_t bytes;
2746 struct u64_stats_sync syncp;
2747} __aligned(2 * sizeof(u64));
2748
2749void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2750
2751static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2752{
2753 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2754
2755 u64_stats_update_begin(&tstats->syncp);
2756 u64_stats_add(&tstats->rx_bytes, len);
2757 u64_stats_inc(&tstats->rx_packets);
2758 u64_stats_update_end(&tstats->syncp);
2759}
2760
2761static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2762 unsigned int packets,
2763 unsigned int len)
2764{
2765 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2766
2767 u64_stats_update_begin(&tstats->syncp);
2768 u64_stats_add(&tstats->tx_bytes, len);
2769 u64_stats_add(&tstats->tx_packets, packets);
2770 u64_stats_update_end(&tstats->syncp);
2771}
2772
2773static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2774{
2775 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2776
2777 u64_stats_update_begin(&lstats->syncp);
2778 u64_stats_add(&lstats->bytes, len);
2779 u64_stats_inc(&lstats->packets);
2780 u64_stats_update_end(&lstats->syncp);
2781}
2782
2783#define __netdev_alloc_pcpu_stats(type, gfp) \
2784({ \
2785 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2786 if (pcpu_stats) { \
2787 int __cpu; \
2788 for_each_possible_cpu(__cpu) { \
2789 typeof(type) *stat; \
2790 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2791 u64_stats_init(&stat->syncp); \
2792 } \
2793 } \
2794 pcpu_stats; \
2795})
2796
2797#define netdev_alloc_pcpu_stats(type) \
2798 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2799
2800#define devm_netdev_alloc_pcpu_stats(dev, type) \
2801({ \
2802 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2803 if (pcpu_stats) { \
2804 int __cpu; \
2805 for_each_possible_cpu(__cpu) { \
2806 typeof(type) *stat; \
2807 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2808 u64_stats_init(&stat->syncp); \
2809 } \
2810 } \
2811 pcpu_stats; \
2812})
2813
2814enum netdev_lag_tx_type {
2815 NETDEV_LAG_TX_TYPE_UNKNOWN,
2816 NETDEV_LAG_TX_TYPE_RANDOM,
2817 NETDEV_LAG_TX_TYPE_BROADCAST,
2818 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2819 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2820 NETDEV_LAG_TX_TYPE_HASH,
2821};
2822
2823enum netdev_lag_hash {
2824 NETDEV_LAG_HASH_NONE,
2825 NETDEV_LAG_HASH_L2,
2826 NETDEV_LAG_HASH_L34,
2827 NETDEV_LAG_HASH_L23,
2828 NETDEV_LAG_HASH_E23,
2829 NETDEV_LAG_HASH_E34,
2830 NETDEV_LAG_HASH_VLAN_SRCMAC,
2831 NETDEV_LAG_HASH_UNKNOWN,
2832};
2833
2834struct netdev_lag_upper_info {
2835 enum netdev_lag_tx_type tx_type;
2836 enum netdev_lag_hash hash_type;
2837};
2838
2839struct netdev_lag_lower_state_info {
2840 u8 link_up : 1,
2841 tx_enabled : 1;
2842};
2843
2844#include <linux/notifier.h>
2845
2846/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2847 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2848 * adding new types.
2849 */
2850enum netdev_cmd {
2851 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2852 NETDEV_DOWN,
2853 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2854 detected a hardware crash and restarted
2855 - we can use this eg to kick tcp sessions
2856 once done */
2857 NETDEV_CHANGE, /* Notify device state change */
2858 NETDEV_REGISTER,
2859 NETDEV_UNREGISTER,
2860 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2861 NETDEV_CHANGEADDR, /* notify after the address change */
2862 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2863 NETDEV_GOING_DOWN,
2864 NETDEV_CHANGENAME,
2865 NETDEV_FEAT_CHANGE,
2866 NETDEV_BONDING_FAILOVER,
2867 NETDEV_PRE_UP,
2868 NETDEV_PRE_TYPE_CHANGE,
2869 NETDEV_POST_TYPE_CHANGE,
2870 NETDEV_POST_INIT,
2871 NETDEV_PRE_UNINIT,
2872 NETDEV_RELEASE,
2873 NETDEV_NOTIFY_PEERS,
2874 NETDEV_JOIN,
2875 NETDEV_CHANGEUPPER,
2876 NETDEV_RESEND_IGMP,
2877 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2878 NETDEV_CHANGEINFODATA,
2879 NETDEV_BONDING_INFO,
2880 NETDEV_PRECHANGEUPPER,
2881 NETDEV_CHANGELOWERSTATE,
2882 NETDEV_UDP_TUNNEL_PUSH_INFO,
2883 NETDEV_UDP_TUNNEL_DROP_INFO,
2884 NETDEV_CHANGE_TX_QUEUE_LEN,
2885 NETDEV_CVLAN_FILTER_PUSH_INFO,
2886 NETDEV_CVLAN_FILTER_DROP_INFO,
2887 NETDEV_SVLAN_FILTER_PUSH_INFO,
2888 NETDEV_SVLAN_FILTER_DROP_INFO,
2889 NETDEV_OFFLOAD_XSTATS_ENABLE,
2890 NETDEV_OFFLOAD_XSTATS_DISABLE,
2891 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2892 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2893 NETDEV_XDP_FEAT_CHANGE,
2894};
2895const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2896
2897int register_netdevice_notifier(struct notifier_block *nb);
2898int unregister_netdevice_notifier(struct notifier_block *nb);
2899int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2900int unregister_netdevice_notifier_net(struct net *net,
2901 struct notifier_block *nb);
2902int register_netdevice_notifier_dev_net(struct net_device *dev,
2903 struct notifier_block *nb,
2904 struct netdev_net_notifier *nn);
2905int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2906 struct notifier_block *nb,
2907 struct netdev_net_notifier *nn);
2908
2909struct netdev_notifier_info {
2910 struct net_device *dev;
2911 struct netlink_ext_ack *extack;
2912};
2913
2914struct netdev_notifier_info_ext {
2915 struct netdev_notifier_info info; /* must be first */
2916 union {
2917 u32 mtu;
2918 } ext;
2919};
2920
2921struct netdev_notifier_change_info {
2922 struct netdev_notifier_info info; /* must be first */
2923 unsigned int flags_changed;
2924};
2925
2926struct netdev_notifier_changeupper_info {
2927 struct netdev_notifier_info info; /* must be first */
2928 struct net_device *upper_dev; /* new upper dev */
2929 bool master; /* is upper dev master */
2930 bool linking; /* is the notification for link or unlink */
2931 void *upper_info; /* upper dev info */
2932};
2933
2934struct netdev_notifier_changelowerstate_info {
2935 struct netdev_notifier_info info; /* must be first */
2936 void *lower_state_info; /* is lower dev state */
2937};
2938
2939struct netdev_notifier_pre_changeaddr_info {
2940 struct netdev_notifier_info info; /* must be first */
2941 const unsigned char *dev_addr;
2942};
2943
2944enum netdev_offload_xstats_type {
2945 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2946};
2947
2948struct netdev_notifier_offload_xstats_info {
2949 struct netdev_notifier_info info; /* must be first */
2950 enum netdev_offload_xstats_type type;
2951
2952 union {
2953 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2954 struct netdev_notifier_offload_xstats_rd *report_delta;
2955 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2956 struct netdev_notifier_offload_xstats_ru *report_used;
2957 };
2958};
2959
2960int netdev_offload_xstats_enable(struct net_device *dev,
2961 enum netdev_offload_xstats_type type,
2962 struct netlink_ext_ack *extack);
2963int netdev_offload_xstats_disable(struct net_device *dev,
2964 enum netdev_offload_xstats_type type);
2965bool netdev_offload_xstats_enabled(const struct net_device *dev,
2966 enum netdev_offload_xstats_type type);
2967int netdev_offload_xstats_get(struct net_device *dev,
2968 enum netdev_offload_xstats_type type,
2969 struct rtnl_hw_stats64 *stats, bool *used,
2970 struct netlink_ext_ack *extack);
2971void
2972netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2973 const struct rtnl_hw_stats64 *stats);
2974void
2975netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2976void netdev_offload_xstats_push_delta(struct net_device *dev,
2977 enum netdev_offload_xstats_type type,
2978 const struct rtnl_hw_stats64 *stats);
2979
2980static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2981 struct net_device *dev)
2982{
2983 info->dev = dev;
2984 info->extack = NULL;
2985}
2986
2987static inline struct net_device *
2988netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2989{
2990 return info->dev;
2991}
2992
2993static inline struct netlink_ext_ack *
2994netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2995{
2996 return info->extack;
2997}
2998
2999int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3000int call_netdevice_notifiers_info(unsigned long val,
3001 struct netdev_notifier_info *info);
3002
3003#define for_each_netdev(net, d) \
3004 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3005#define for_each_netdev_reverse(net, d) \
3006 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3007#define for_each_netdev_rcu(net, d) \
3008 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3009#define for_each_netdev_safe(net, d, n) \
3010 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3011#define for_each_netdev_continue(net, d) \
3012 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3013#define for_each_netdev_continue_reverse(net, d) \
3014 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3015 dev_list)
3016#define for_each_netdev_continue_rcu(net, d) \
3017 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3018#define for_each_netdev_in_bond_rcu(bond, slave) \
3019 for_each_netdev_rcu(&init_net, slave) \
3020 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3021#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3022
3023#define for_each_netdev_dump(net, d, ifindex) \
3024 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \
3025 ULONG_MAX, XA_PRESENT)); ifindex++)
3026
3027static inline struct net_device *next_net_device(struct net_device *dev)
3028{
3029 struct list_head *lh;
3030 struct net *net;
3031
3032 net = dev_net(dev);
3033 lh = dev->dev_list.next;
3034 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3035}
3036
3037static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3038{
3039 struct list_head *lh;
3040 struct net *net;
3041
3042 net = dev_net(dev);
3043 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3044 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3045}
3046
3047static inline struct net_device *first_net_device(struct net *net)
3048{
3049 return list_empty(&net->dev_base_head) ? NULL :
3050 net_device_entry(net->dev_base_head.next);
3051}
3052
3053static inline struct net_device *first_net_device_rcu(struct net *net)
3054{
3055 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3056
3057 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3058}
3059
3060int netdev_boot_setup_check(struct net_device *dev);
3061struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3062 const char *hwaddr);
3063struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3064void dev_add_pack(struct packet_type *pt);
3065void dev_remove_pack(struct packet_type *pt);
3066void __dev_remove_pack(struct packet_type *pt);
3067void dev_add_offload(struct packet_offload *po);
3068void dev_remove_offload(struct packet_offload *po);
3069
3070int dev_get_iflink(const struct net_device *dev);
3071int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3072int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3073 struct net_device_path_stack *stack);
3074struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3075 unsigned short mask);
3076struct net_device *dev_get_by_name(struct net *net, const char *name);
3077struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3078struct net_device *__dev_get_by_name(struct net *net, const char *name);
3079bool netdev_name_in_use(struct net *net, const char *name);
3080int dev_alloc_name(struct net_device *dev, const char *name);
3081int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3082void dev_close(struct net_device *dev);
3083void dev_close_many(struct list_head *head, bool unlink);
3084void dev_disable_lro(struct net_device *dev);
3085int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3086u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3087 struct net_device *sb_dev);
3088
3089int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3090int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3091
3092static inline int dev_queue_xmit(struct sk_buff *skb)
3093{
3094 return __dev_queue_xmit(skb, NULL);
3095}
3096
3097static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3098 struct net_device *sb_dev)
3099{
3100 return __dev_queue_xmit(skb, sb_dev);
3101}
3102
3103static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3104{
3105 int ret;
3106
3107 ret = __dev_direct_xmit(skb, queue_id);
3108 if (!dev_xmit_complete(ret))
3109 kfree_skb(skb);
3110 return ret;
3111}
3112
3113int register_netdevice(struct net_device *dev);
3114void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3115void unregister_netdevice_many(struct list_head *head);
3116static inline void unregister_netdevice(struct net_device *dev)
3117{
3118 unregister_netdevice_queue(dev, NULL);
3119}
3120
3121int netdev_refcnt_read(const struct net_device *dev);
3122void free_netdev(struct net_device *dev);
3123void init_dummy_netdev(struct net_device *dev);
3124
3125struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3126 struct sk_buff *skb,
3127 bool all_slaves);
3128struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3129 struct sock *sk);
3130struct net_device *dev_get_by_index(struct net *net, int ifindex);
3131struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3132struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3133 netdevice_tracker *tracker, gfp_t gfp);
3134struct net_device *netdev_get_by_name(struct net *net, const char *name,
3135 netdevice_tracker *tracker, gfp_t gfp);
3136struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3137struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3138void netdev_copy_name(struct net_device *dev, char *name);
3139
3140static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3141 unsigned short type,
3142 const void *daddr, const void *saddr,
3143 unsigned int len)
3144{
3145 if (!dev->header_ops || !dev->header_ops->create)
3146 return 0;
3147
3148 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3149}
3150
3151static inline int dev_parse_header(const struct sk_buff *skb,
3152 unsigned char *haddr)
3153{
3154 const struct net_device *dev = skb->dev;
3155
3156 if (!dev->header_ops || !dev->header_ops->parse)
3157 return 0;
3158 return dev->header_ops->parse(skb, haddr);
3159}
3160
3161static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3162{
3163 const struct net_device *dev = skb->dev;
3164
3165 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3166 return 0;
3167 return dev->header_ops->parse_protocol(skb);
3168}
3169
3170/* ll_header must have at least hard_header_len allocated */
3171static inline bool dev_validate_header(const struct net_device *dev,
3172 char *ll_header, int len)
3173{
3174 if (likely(len >= dev->hard_header_len))
3175 return true;
3176 if (len < dev->min_header_len)
3177 return false;
3178
3179 if (capable(CAP_SYS_RAWIO)) {
3180 memset(ll_header + len, 0, dev->hard_header_len - len);
3181 return true;
3182 }
3183
3184 if (dev->header_ops && dev->header_ops->validate)
3185 return dev->header_ops->validate(ll_header, len);
3186
3187 return false;
3188}
3189
3190static inline bool dev_has_header(const struct net_device *dev)
3191{
3192 return dev->header_ops && dev->header_ops->create;
3193}
3194
3195/*
3196 * Incoming packets are placed on per-CPU queues
3197 */
3198struct softnet_data {
3199 struct list_head poll_list;
3200 struct sk_buff_head process_queue;
3201 local_lock_t process_queue_bh_lock;
3202
3203 /* stats */
3204 unsigned int processed;
3205 unsigned int time_squeeze;
3206#ifdef CONFIG_RPS
3207 struct softnet_data *rps_ipi_list;
3208#endif
3209
3210 unsigned int received_rps;
3211 bool in_net_rx_action;
3212 bool in_napi_threaded_poll;
3213
3214#ifdef CONFIG_NET_FLOW_LIMIT
3215 struct sd_flow_limit __rcu *flow_limit;
3216#endif
3217 struct Qdisc *output_queue;
3218 struct Qdisc **output_queue_tailp;
3219 struct sk_buff *completion_queue;
3220#ifdef CONFIG_XFRM_OFFLOAD
3221 struct sk_buff_head xfrm_backlog;
3222#endif
3223 /* written and read only by owning cpu: */
3224 struct netdev_xmit xmit;
3225#ifdef CONFIG_RPS
3226 /* input_queue_head should be written by cpu owning this struct,
3227 * and only read by other cpus. Worth using a cache line.
3228 */
3229 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3230
3231 /* Elements below can be accessed between CPUs for RPS/RFS */
3232 call_single_data_t csd ____cacheline_aligned_in_smp;
3233 struct softnet_data *rps_ipi_next;
3234 unsigned int cpu;
3235 unsigned int input_queue_tail;
3236#endif
3237 struct sk_buff_head input_pkt_queue;
3238 struct napi_struct backlog;
3239
3240 atomic_t dropped ____cacheline_aligned_in_smp;
3241
3242 /* Another possibly contended cache line */
3243 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3244 int defer_count;
3245 int defer_ipi_scheduled;
3246 struct sk_buff *defer_list;
3247 call_single_data_t defer_csd;
3248};
3249
3250DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3251
3252#ifndef CONFIG_PREEMPT_RT
3253static inline int dev_recursion_level(void)
3254{
3255 return this_cpu_read(softnet_data.xmit.recursion);
3256}
3257#else
3258static inline int dev_recursion_level(void)
3259{
3260 return current->net_xmit.recursion;
3261}
3262
3263#endif
3264
3265void __netif_schedule(struct Qdisc *q);
3266void netif_schedule_queue(struct netdev_queue *txq);
3267
3268static inline void netif_tx_schedule_all(struct net_device *dev)
3269{
3270 unsigned int i;
3271
3272 for (i = 0; i < dev->num_tx_queues; i++)
3273 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3274}
3275
3276static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3277{
3278 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3279}
3280
3281/**
3282 * netif_start_queue - allow transmit
3283 * @dev: network device
3284 *
3285 * Allow upper layers to call the device hard_start_xmit routine.
3286 */
3287static inline void netif_start_queue(struct net_device *dev)
3288{
3289 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3290}
3291
3292static inline void netif_tx_start_all_queues(struct net_device *dev)
3293{
3294 unsigned int i;
3295
3296 for (i = 0; i < dev->num_tx_queues; i++) {
3297 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3298 netif_tx_start_queue(txq);
3299 }
3300}
3301
3302void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3303
3304/**
3305 * netif_wake_queue - restart transmit
3306 * @dev: network device
3307 *
3308 * Allow upper layers to call the device hard_start_xmit routine.
3309 * Used for flow control when transmit resources are available.
3310 */
3311static inline void netif_wake_queue(struct net_device *dev)
3312{
3313 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3314}
3315
3316static inline void netif_tx_wake_all_queues(struct net_device *dev)
3317{
3318 unsigned int i;
3319
3320 for (i = 0; i < dev->num_tx_queues; i++) {
3321 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3322 netif_tx_wake_queue(txq);
3323 }
3324}
3325
3326static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3327{
3328 /* Must be an atomic op see netif_txq_try_stop() */
3329 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3330}
3331
3332/**
3333 * netif_stop_queue - stop transmitted packets
3334 * @dev: network device
3335 *
3336 * Stop upper layers calling the device hard_start_xmit routine.
3337 * Used for flow control when transmit resources are unavailable.
3338 */
3339static inline void netif_stop_queue(struct net_device *dev)
3340{
3341 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3342}
3343
3344void netif_tx_stop_all_queues(struct net_device *dev);
3345
3346static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3347{
3348 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3349}
3350
3351/**
3352 * netif_queue_stopped - test if transmit queue is flowblocked
3353 * @dev: network device
3354 *
3355 * Test if transmit queue on device is currently unable to send.
3356 */
3357static inline bool netif_queue_stopped(const struct net_device *dev)
3358{
3359 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3360}
3361
3362static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3363{
3364 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3365}
3366
3367static inline bool
3368netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3369{
3370 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3371}
3372
3373static inline bool
3374netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3375{
3376 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3377}
3378
3379/**
3380 * netdev_queue_set_dql_min_limit - set dql minimum limit
3381 * @dev_queue: pointer to transmit queue
3382 * @min_limit: dql minimum limit
3383 *
3384 * Forces xmit_more() to return true until the minimum threshold
3385 * defined by @min_limit is reached (or until the tx queue is
3386 * empty). Warning: to be use with care, misuse will impact the
3387 * latency.
3388 */
3389static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3390 unsigned int min_limit)
3391{
3392#ifdef CONFIG_BQL
3393 dev_queue->dql.min_limit = min_limit;
3394#endif
3395}
3396
3397static inline int netdev_queue_dql_avail(const struct netdev_queue *txq)
3398{
3399#ifdef CONFIG_BQL
3400 /* Non-BQL migrated drivers will return 0, too. */
3401 return dql_avail(&txq->dql);
3402#else
3403 return 0;
3404#endif
3405}
3406
3407/**
3408 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3409 * @dev_queue: pointer to transmit queue
3410 *
3411 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3412 * to give appropriate hint to the CPU.
3413 */
3414static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3415{
3416#ifdef CONFIG_BQL
3417 prefetchw(&dev_queue->dql.num_queued);
3418#endif
3419}
3420
3421/**
3422 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3423 * @dev_queue: pointer to transmit queue
3424 *
3425 * BQL enabled drivers might use this helper in their TX completion path,
3426 * to give appropriate hint to the CPU.
3427 */
3428static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3429{
3430#ifdef CONFIG_BQL
3431 prefetchw(&dev_queue->dql.limit);
3432#endif
3433}
3434
3435/**
3436 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3437 * @dev_queue: network device queue
3438 * @bytes: number of bytes queued to the device queue
3439 *
3440 * Report the number of bytes queued for sending/completion to the network
3441 * device hardware queue. @bytes should be a good approximation and should
3442 * exactly match netdev_completed_queue() @bytes.
3443 * This is typically called once per packet, from ndo_start_xmit().
3444 */
3445static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3446 unsigned int bytes)
3447{
3448#ifdef CONFIG_BQL
3449 dql_queued(&dev_queue->dql, bytes);
3450
3451 if (likely(dql_avail(&dev_queue->dql) >= 0))
3452 return;
3453
3454 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3455
3456 /*
3457 * The XOFF flag must be set before checking the dql_avail below,
3458 * because in netdev_tx_completed_queue we update the dql_completed
3459 * before checking the XOFF flag.
3460 */
3461 smp_mb();
3462
3463 /* check again in case another CPU has just made room avail */
3464 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3465 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3466#endif
3467}
3468
3469/* Variant of netdev_tx_sent_queue() for drivers that are aware
3470 * that they should not test BQL status themselves.
3471 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3472 * skb of a batch.
3473 * Returns true if the doorbell must be used to kick the NIC.
3474 */
3475static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3476 unsigned int bytes,
3477 bool xmit_more)
3478{
3479 if (xmit_more) {
3480#ifdef CONFIG_BQL
3481 dql_queued(&dev_queue->dql, bytes);
3482#endif
3483 return netif_tx_queue_stopped(dev_queue);
3484 }
3485 netdev_tx_sent_queue(dev_queue, bytes);
3486 return true;
3487}
3488
3489/**
3490 * netdev_sent_queue - report the number of bytes queued to hardware
3491 * @dev: network device
3492 * @bytes: number of bytes queued to the hardware device queue
3493 *
3494 * Report the number of bytes queued for sending/completion to the network
3495 * device hardware queue#0. @bytes should be a good approximation and should
3496 * exactly match netdev_completed_queue() @bytes.
3497 * This is typically called once per packet, from ndo_start_xmit().
3498 */
3499static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3500{
3501 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3502}
3503
3504static inline bool __netdev_sent_queue(struct net_device *dev,
3505 unsigned int bytes,
3506 bool xmit_more)
3507{
3508 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3509 xmit_more);
3510}
3511
3512/**
3513 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3514 * @dev_queue: network device queue
3515 * @pkts: number of packets (currently ignored)
3516 * @bytes: number of bytes dequeued from the device queue
3517 *
3518 * Must be called at most once per TX completion round (and not per
3519 * individual packet), so that BQL can adjust its limits appropriately.
3520 */
3521static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3522 unsigned int pkts, unsigned int bytes)
3523{
3524#ifdef CONFIG_BQL
3525 if (unlikely(!bytes))
3526 return;
3527
3528 dql_completed(&dev_queue->dql, bytes);
3529
3530 /*
3531 * Without the memory barrier there is a small possibility that
3532 * netdev_tx_sent_queue will miss the update and cause the queue to
3533 * be stopped forever
3534 */
3535 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3536
3537 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3538 return;
3539
3540 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3541 netif_schedule_queue(dev_queue);
3542#endif
3543}
3544
3545/**
3546 * netdev_completed_queue - report bytes and packets completed by device
3547 * @dev: network device
3548 * @pkts: actual number of packets sent over the medium
3549 * @bytes: actual number of bytes sent over the medium
3550 *
3551 * Report the number of bytes and packets transmitted by the network device
3552 * hardware queue over the physical medium, @bytes must exactly match the
3553 * @bytes amount passed to netdev_sent_queue()
3554 */
3555static inline void netdev_completed_queue(struct net_device *dev,
3556 unsigned int pkts, unsigned int bytes)
3557{
3558 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3559}
3560
3561static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3562{
3563#ifdef CONFIG_BQL
3564 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3565 dql_reset(&q->dql);
3566#endif
3567}
3568
3569/**
3570 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue
3571 * @dev: network device
3572 * @qid: stack index of the queue to reset
3573 */
3574static inline void netdev_tx_reset_subqueue(const struct net_device *dev,
3575 u32 qid)
3576{
3577 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid));
3578}
3579
3580/**
3581 * netdev_reset_queue - reset the packets and bytes count of a network device
3582 * @dev_queue: network device
3583 *
3584 * Reset the bytes and packet count of a network device and clear the
3585 * software flow control OFF bit for this network device
3586 */
3587static inline void netdev_reset_queue(struct net_device *dev_queue)
3588{
3589 netdev_tx_reset_subqueue(dev_queue, 0);
3590}
3591
3592/**
3593 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3594 * @dev: network device
3595 * @queue_index: given tx queue index
3596 *
3597 * Returns 0 if given tx queue index >= number of device tx queues,
3598 * otherwise returns the originally passed tx queue index.
3599 */
3600static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3601{
3602 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3603 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3604 dev->name, queue_index,
3605 dev->real_num_tx_queues);
3606 return 0;
3607 }
3608
3609 return queue_index;
3610}
3611
3612/**
3613 * netif_running - test if up
3614 * @dev: network device
3615 *
3616 * Test if the device has been brought up.
3617 */
3618static inline bool netif_running(const struct net_device *dev)
3619{
3620 return test_bit(__LINK_STATE_START, &dev->state);
3621}
3622
3623/*
3624 * Routines to manage the subqueues on a device. We only need start,
3625 * stop, and a check if it's stopped. All other device management is
3626 * done at the overall netdevice level.
3627 * Also test the device if we're multiqueue.
3628 */
3629
3630/**
3631 * netif_start_subqueue - allow sending packets on subqueue
3632 * @dev: network device
3633 * @queue_index: sub queue index
3634 *
3635 * Start individual transmit queue of a device with multiple transmit queues.
3636 */
3637static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3638{
3639 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3640
3641 netif_tx_start_queue(txq);
3642}
3643
3644/**
3645 * netif_stop_subqueue - stop sending packets on subqueue
3646 * @dev: network device
3647 * @queue_index: sub queue index
3648 *
3649 * Stop individual transmit queue of a device with multiple transmit queues.
3650 */
3651static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3652{
3653 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3654 netif_tx_stop_queue(txq);
3655}
3656
3657/**
3658 * __netif_subqueue_stopped - test status of subqueue
3659 * @dev: network device
3660 * @queue_index: sub queue index
3661 *
3662 * Check individual transmit queue of a device with multiple transmit queues.
3663 */
3664static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3665 u16 queue_index)
3666{
3667 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3668
3669 return netif_tx_queue_stopped(txq);
3670}
3671
3672/**
3673 * netif_subqueue_stopped - test status of subqueue
3674 * @dev: network device
3675 * @skb: sub queue buffer pointer
3676 *
3677 * Check individual transmit queue of a device with multiple transmit queues.
3678 */
3679static inline bool netif_subqueue_stopped(const struct net_device *dev,
3680 struct sk_buff *skb)
3681{
3682 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3683}
3684
3685/**
3686 * netif_wake_subqueue - allow sending packets on subqueue
3687 * @dev: network device
3688 * @queue_index: sub queue index
3689 *
3690 * Resume individual transmit queue of a device with multiple transmit queues.
3691 */
3692static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3693{
3694 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3695
3696 netif_tx_wake_queue(txq);
3697}
3698
3699#ifdef CONFIG_XPS
3700int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3701 u16 index);
3702int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3703 u16 index, enum xps_map_type type);
3704
3705/**
3706 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3707 * @j: CPU/Rx queue index
3708 * @mask: bitmask of all cpus/rx queues
3709 * @nr_bits: number of bits in the bitmask
3710 *
3711 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3712 */
3713static inline bool netif_attr_test_mask(unsigned long j,
3714 const unsigned long *mask,
3715 unsigned int nr_bits)
3716{
3717 cpu_max_bits_warn(j, nr_bits);
3718 return test_bit(j, mask);
3719}
3720
3721/**
3722 * netif_attr_test_online - Test for online CPU/Rx queue
3723 * @j: CPU/Rx queue index
3724 * @online_mask: bitmask for CPUs/Rx queues that are online
3725 * @nr_bits: number of bits in the bitmask
3726 *
3727 * Returns true if a CPU/Rx queue is online.
3728 */
3729static inline bool netif_attr_test_online(unsigned long j,
3730 const unsigned long *online_mask,
3731 unsigned int nr_bits)
3732{
3733 cpu_max_bits_warn(j, nr_bits);
3734
3735 if (online_mask)
3736 return test_bit(j, online_mask);
3737
3738 return (j < nr_bits);
3739}
3740
3741/**
3742 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3743 * @n: CPU/Rx queue index
3744 * @srcp: the cpumask/Rx queue mask pointer
3745 * @nr_bits: number of bits in the bitmask
3746 *
3747 * Returns >= nr_bits if no further CPUs/Rx queues set.
3748 */
3749static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3750 unsigned int nr_bits)
3751{
3752 /* -1 is a legal arg here. */
3753 if (n != -1)
3754 cpu_max_bits_warn(n, nr_bits);
3755
3756 if (srcp)
3757 return find_next_bit(srcp, nr_bits, n + 1);
3758
3759 return n + 1;
3760}
3761
3762/**
3763 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3764 * @n: CPU/Rx queue index
3765 * @src1p: the first CPUs/Rx queues mask pointer
3766 * @src2p: the second CPUs/Rx queues mask pointer
3767 * @nr_bits: number of bits in the bitmask
3768 *
3769 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3770 */
3771static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3772 const unsigned long *src2p,
3773 unsigned int nr_bits)
3774{
3775 /* -1 is a legal arg here. */
3776 if (n != -1)
3777 cpu_max_bits_warn(n, nr_bits);
3778
3779 if (src1p && src2p)
3780 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3781 else if (src1p)
3782 return find_next_bit(src1p, nr_bits, n + 1);
3783 else if (src2p)
3784 return find_next_bit(src2p, nr_bits, n + 1);
3785
3786 return n + 1;
3787}
3788#else
3789static inline int netif_set_xps_queue(struct net_device *dev,
3790 const struct cpumask *mask,
3791 u16 index)
3792{
3793 return 0;
3794}
3795
3796static inline int __netif_set_xps_queue(struct net_device *dev,
3797 const unsigned long *mask,
3798 u16 index, enum xps_map_type type)
3799{
3800 return 0;
3801}
3802#endif
3803
3804/**
3805 * netif_is_multiqueue - test if device has multiple transmit queues
3806 * @dev: network device
3807 *
3808 * Check if device has multiple transmit queues
3809 */
3810static inline bool netif_is_multiqueue(const struct net_device *dev)
3811{
3812 return dev->num_tx_queues > 1;
3813}
3814
3815int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3816
3817#ifdef CONFIG_SYSFS
3818int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3819#else
3820static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3821 unsigned int rxqs)
3822{
3823 dev->real_num_rx_queues = rxqs;
3824 return 0;
3825}
3826#endif
3827int netif_set_real_num_queues(struct net_device *dev,
3828 unsigned int txq, unsigned int rxq);
3829
3830int netif_get_num_default_rss_queues(void);
3831
3832void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3833void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3834
3835/*
3836 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3837 * interrupt context or with hardware interrupts being disabled.
3838 * (in_hardirq() || irqs_disabled())
3839 *
3840 * We provide four helpers that can be used in following contexts :
3841 *
3842 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3843 * replacing kfree_skb(skb)
3844 *
3845 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3846 * Typically used in place of consume_skb(skb) in TX completion path
3847 *
3848 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3849 * replacing kfree_skb(skb)
3850 *
3851 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3852 * and consumed a packet. Used in place of consume_skb(skb)
3853 */
3854static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3855{
3856 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3857}
3858
3859static inline void dev_consume_skb_irq(struct sk_buff *skb)
3860{
3861 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3862}
3863
3864static inline void dev_kfree_skb_any(struct sk_buff *skb)
3865{
3866 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3867}
3868
3869static inline void dev_consume_skb_any(struct sk_buff *skb)
3870{
3871 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3872}
3873
3874u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3875 struct bpf_prog *xdp_prog);
3876void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3877int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb);
3878int netif_rx(struct sk_buff *skb);
3879int __netif_rx(struct sk_buff *skb);
3880
3881int netif_receive_skb(struct sk_buff *skb);
3882int netif_receive_skb_core(struct sk_buff *skb);
3883void netif_receive_skb_list_internal(struct list_head *head);
3884void netif_receive_skb_list(struct list_head *head);
3885gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3886void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3887struct sk_buff *napi_get_frags(struct napi_struct *napi);
3888void napi_get_frags_check(struct napi_struct *napi);
3889gro_result_t napi_gro_frags(struct napi_struct *napi);
3890
3891static inline void napi_free_frags(struct napi_struct *napi)
3892{
3893 kfree_skb(napi->skb);
3894 napi->skb = NULL;
3895}
3896
3897bool netdev_is_rx_handler_busy(struct net_device *dev);
3898int netdev_rx_handler_register(struct net_device *dev,
3899 rx_handler_func_t *rx_handler,
3900 void *rx_handler_data);
3901void netdev_rx_handler_unregister(struct net_device *dev);
3902
3903bool dev_valid_name(const char *name);
3904static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3905{
3906 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3907}
3908int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3909int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3910int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3911 void __user *data, bool *need_copyout);
3912int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3913int generic_hwtstamp_get_lower(struct net_device *dev,
3914 struct kernel_hwtstamp_config *kernel_cfg);
3915int generic_hwtstamp_set_lower(struct net_device *dev,
3916 struct kernel_hwtstamp_config *kernel_cfg,
3917 struct netlink_ext_ack *extack);
3918int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3919unsigned int dev_get_flags(const struct net_device *);
3920int __dev_change_flags(struct net_device *dev, unsigned int flags,
3921 struct netlink_ext_ack *extack);
3922int dev_change_flags(struct net_device *dev, unsigned int flags,
3923 struct netlink_ext_ack *extack);
3924int dev_set_alias(struct net_device *, const char *, size_t);
3925int dev_get_alias(const struct net_device *, char *, size_t);
3926int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3927 const char *pat, int new_ifindex);
3928static inline
3929int dev_change_net_namespace(struct net_device *dev, struct net *net,
3930 const char *pat)
3931{
3932 return __dev_change_net_namespace(dev, net, pat, 0);
3933}
3934int __dev_set_mtu(struct net_device *, int);
3935int dev_set_mtu(struct net_device *, int);
3936int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3937 struct netlink_ext_ack *extack);
3938int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3939 struct netlink_ext_ack *extack);
3940int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3941 struct netlink_ext_ack *extack);
3942int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3943int dev_get_port_parent_id(struct net_device *dev,
3944 struct netdev_phys_item_id *ppid, bool recurse);
3945bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3946
3947struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3948struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3949 struct netdev_queue *txq, int *ret);
3950
3951int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3952u8 dev_xdp_prog_count(struct net_device *dev);
3953int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf);
3954u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3955
3956u32 dev_get_min_mp_channel_count(const struct net_device *dev);
3957
3958int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3959int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3960int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3961bool is_skb_forwardable(const struct net_device *dev,
3962 const struct sk_buff *skb);
3963
3964static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3965 const struct sk_buff *skb,
3966 const bool check_mtu)
3967{
3968 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3969 unsigned int len;
3970
3971 if (!(dev->flags & IFF_UP))
3972 return false;
3973
3974 if (!check_mtu)
3975 return true;
3976
3977 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3978 if (skb->len <= len)
3979 return true;
3980
3981 /* if TSO is enabled, we don't care about the length as the packet
3982 * could be forwarded without being segmented before
3983 */
3984 if (skb_is_gso(skb))
3985 return true;
3986
3987 return false;
3988}
3989
3990void netdev_core_stats_inc(struct net_device *dev, u32 offset);
3991
3992#define DEV_CORE_STATS_INC(FIELD) \
3993static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3994{ \
3995 netdev_core_stats_inc(dev, \
3996 offsetof(struct net_device_core_stats, FIELD)); \
3997}
3998DEV_CORE_STATS_INC(rx_dropped)
3999DEV_CORE_STATS_INC(tx_dropped)
4000DEV_CORE_STATS_INC(rx_nohandler)
4001DEV_CORE_STATS_INC(rx_otherhost_dropped)
4002#undef DEV_CORE_STATS_INC
4003
4004static __always_inline int ____dev_forward_skb(struct net_device *dev,
4005 struct sk_buff *skb,
4006 const bool check_mtu)
4007{
4008 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4009 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4010 dev_core_stats_rx_dropped_inc(dev);
4011 kfree_skb(skb);
4012 return NET_RX_DROP;
4013 }
4014
4015 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4016 skb->priority = 0;
4017 return 0;
4018}
4019
4020bool dev_nit_active(struct net_device *dev);
4021void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4022
4023static inline void __dev_put(struct net_device *dev)
4024{
4025 if (dev) {
4026#ifdef CONFIG_PCPU_DEV_REFCNT
4027 this_cpu_dec(*dev->pcpu_refcnt);
4028#else
4029 refcount_dec(&dev->dev_refcnt);
4030#endif
4031 }
4032}
4033
4034static inline void __dev_hold(struct net_device *dev)
4035{
4036 if (dev) {
4037#ifdef CONFIG_PCPU_DEV_REFCNT
4038 this_cpu_inc(*dev->pcpu_refcnt);
4039#else
4040 refcount_inc(&dev->dev_refcnt);
4041#endif
4042 }
4043}
4044
4045static inline void __netdev_tracker_alloc(struct net_device *dev,
4046 netdevice_tracker *tracker,
4047 gfp_t gfp)
4048{
4049#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4050 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4051#endif
4052}
4053
4054/* netdev_tracker_alloc() can upgrade a prior untracked reference
4055 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4056 */
4057static inline void netdev_tracker_alloc(struct net_device *dev,
4058 netdevice_tracker *tracker, gfp_t gfp)
4059{
4060#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4061 refcount_dec(&dev->refcnt_tracker.no_tracker);
4062 __netdev_tracker_alloc(dev, tracker, gfp);
4063#endif
4064}
4065
4066static inline void netdev_tracker_free(struct net_device *dev,
4067 netdevice_tracker *tracker)
4068{
4069#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4070 ref_tracker_free(&dev->refcnt_tracker, tracker);
4071#endif
4072}
4073
4074static inline void netdev_hold(struct net_device *dev,
4075 netdevice_tracker *tracker, gfp_t gfp)
4076{
4077 if (dev) {
4078 __dev_hold(dev);
4079 __netdev_tracker_alloc(dev, tracker, gfp);
4080 }
4081}
4082
4083static inline void netdev_put(struct net_device *dev,
4084 netdevice_tracker *tracker)
4085{
4086 if (dev) {
4087 netdev_tracker_free(dev, tracker);
4088 __dev_put(dev);
4089 }
4090}
4091
4092/**
4093 * dev_hold - get reference to device
4094 * @dev: network device
4095 *
4096 * Hold reference to device to keep it from being freed.
4097 * Try using netdev_hold() instead.
4098 */
4099static inline void dev_hold(struct net_device *dev)
4100{
4101 netdev_hold(dev, NULL, GFP_ATOMIC);
4102}
4103
4104/**
4105 * dev_put - release reference to device
4106 * @dev: network device
4107 *
4108 * Release reference to device to allow it to be freed.
4109 * Try using netdev_put() instead.
4110 */
4111static inline void dev_put(struct net_device *dev)
4112{
4113 netdev_put(dev, NULL);
4114}
4115
4116DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T))
4117
4118static inline void netdev_ref_replace(struct net_device *odev,
4119 struct net_device *ndev,
4120 netdevice_tracker *tracker,
4121 gfp_t gfp)
4122{
4123 if (odev)
4124 netdev_tracker_free(odev, tracker);
4125
4126 __dev_hold(ndev);
4127 __dev_put(odev);
4128
4129 if (ndev)
4130 __netdev_tracker_alloc(ndev, tracker, gfp);
4131}
4132
4133/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4134 * and _off may be called from IRQ context, but it is caller
4135 * who is responsible for serialization of these calls.
4136 *
4137 * The name carrier is inappropriate, these functions should really be
4138 * called netif_lowerlayer_*() because they represent the state of any
4139 * kind of lower layer not just hardware media.
4140 */
4141void linkwatch_fire_event(struct net_device *dev);
4142
4143/**
4144 * linkwatch_sync_dev - sync linkwatch for the given device
4145 * @dev: network device to sync linkwatch for
4146 *
4147 * Sync linkwatch for the given device, removing it from the
4148 * pending work list (if queued).
4149 */
4150void linkwatch_sync_dev(struct net_device *dev);
4151
4152/**
4153 * netif_carrier_ok - test if carrier present
4154 * @dev: network device
4155 *
4156 * Check if carrier is present on device
4157 */
4158static inline bool netif_carrier_ok(const struct net_device *dev)
4159{
4160 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4161}
4162
4163unsigned long dev_trans_start(struct net_device *dev);
4164
4165void __netdev_watchdog_up(struct net_device *dev);
4166
4167void netif_carrier_on(struct net_device *dev);
4168void netif_carrier_off(struct net_device *dev);
4169void netif_carrier_event(struct net_device *dev);
4170
4171/**
4172 * netif_dormant_on - mark device as dormant.
4173 * @dev: network device
4174 *
4175 * Mark device as dormant (as per RFC2863).
4176 *
4177 * The dormant state indicates that the relevant interface is not
4178 * actually in a condition to pass packets (i.e., it is not 'up') but is
4179 * in a "pending" state, waiting for some external event. For "on-
4180 * demand" interfaces, this new state identifies the situation where the
4181 * interface is waiting for events to place it in the up state.
4182 */
4183static inline void netif_dormant_on(struct net_device *dev)
4184{
4185 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4186 linkwatch_fire_event(dev);
4187}
4188
4189/**
4190 * netif_dormant_off - set device as not dormant.
4191 * @dev: network device
4192 *
4193 * Device is not in dormant state.
4194 */
4195static inline void netif_dormant_off(struct net_device *dev)
4196{
4197 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4198 linkwatch_fire_event(dev);
4199}
4200
4201/**
4202 * netif_dormant - test if device is dormant
4203 * @dev: network device
4204 *
4205 * Check if device is dormant.
4206 */
4207static inline bool netif_dormant(const struct net_device *dev)
4208{
4209 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4210}
4211
4212
4213/**
4214 * netif_testing_on - mark device as under test.
4215 * @dev: network device
4216 *
4217 * Mark device as under test (as per RFC2863).
4218 *
4219 * The testing state indicates that some test(s) must be performed on
4220 * the interface. After completion, of the test, the interface state
4221 * will change to up, dormant, or down, as appropriate.
4222 */
4223static inline void netif_testing_on(struct net_device *dev)
4224{
4225 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4226 linkwatch_fire_event(dev);
4227}
4228
4229/**
4230 * netif_testing_off - set device as not under test.
4231 * @dev: network device
4232 *
4233 * Device is not in testing state.
4234 */
4235static inline void netif_testing_off(struct net_device *dev)
4236{
4237 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4238 linkwatch_fire_event(dev);
4239}
4240
4241/**
4242 * netif_testing - test if device is under test
4243 * @dev: network device
4244 *
4245 * Check if device is under test
4246 */
4247static inline bool netif_testing(const struct net_device *dev)
4248{
4249 return test_bit(__LINK_STATE_TESTING, &dev->state);
4250}
4251
4252
4253/**
4254 * netif_oper_up - test if device is operational
4255 * @dev: network device
4256 *
4257 * Check if carrier is operational
4258 */
4259static inline bool netif_oper_up(const struct net_device *dev)
4260{
4261 unsigned int operstate = READ_ONCE(dev->operstate);
4262
4263 return operstate == IF_OPER_UP ||
4264 operstate == IF_OPER_UNKNOWN /* backward compat */;
4265}
4266
4267/**
4268 * netif_device_present - is device available or removed
4269 * @dev: network device
4270 *
4271 * Check if device has not been removed from system.
4272 */
4273static inline bool netif_device_present(const struct net_device *dev)
4274{
4275 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4276}
4277
4278void netif_device_detach(struct net_device *dev);
4279
4280void netif_device_attach(struct net_device *dev);
4281
4282/*
4283 * Network interface message level settings
4284 */
4285
4286enum {
4287 NETIF_MSG_DRV_BIT,
4288 NETIF_MSG_PROBE_BIT,
4289 NETIF_MSG_LINK_BIT,
4290 NETIF_MSG_TIMER_BIT,
4291 NETIF_MSG_IFDOWN_BIT,
4292 NETIF_MSG_IFUP_BIT,
4293 NETIF_MSG_RX_ERR_BIT,
4294 NETIF_MSG_TX_ERR_BIT,
4295 NETIF_MSG_TX_QUEUED_BIT,
4296 NETIF_MSG_INTR_BIT,
4297 NETIF_MSG_TX_DONE_BIT,
4298 NETIF_MSG_RX_STATUS_BIT,
4299 NETIF_MSG_PKTDATA_BIT,
4300 NETIF_MSG_HW_BIT,
4301 NETIF_MSG_WOL_BIT,
4302
4303 /* When you add a new bit above, update netif_msg_class_names array
4304 * in net/ethtool/common.c
4305 */
4306 NETIF_MSG_CLASS_COUNT,
4307};
4308/* Both ethtool_ops interface and internal driver implementation use u32 */
4309static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4310
4311#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4312#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4313
4314#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4315#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4316#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4317#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4318#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4319#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4320#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4321#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4322#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4323#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4324#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4325#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4326#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4327#define NETIF_MSG_HW __NETIF_MSG(HW)
4328#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4329
4330#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4331#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4332#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4333#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4334#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4335#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4336#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4337#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4338#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4339#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4340#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4341#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4342#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4343#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4344#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4345
4346static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4347{
4348 /* use default */
4349 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4350 return default_msg_enable_bits;
4351 if (debug_value == 0) /* no output */
4352 return 0;
4353 /* set low N bits */
4354 return (1U << debug_value) - 1;
4355}
4356
4357static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4358{
4359 spin_lock(&txq->_xmit_lock);
4360 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4361 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4362}
4363
4364static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4365{
4366 __acquire(&txq->_xmit_lock);
4367 return true;
4368}
4369
4370static inline void __netif_tx_release(struct netdev_queue *txq)
4371{
4372 __release(&txq->_xmit_lock);
4373}
4374
4375static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4376{
4377 spin_lock_bh(&txq->_xmit_lock);
4378 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4379 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4380}
4381
4382static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4383{
4384 bool ok = spin_trylock(&txq->_xmit_lock);
4385
4386 if (likely(ok)) {
4387 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4388 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4389 }
4390 return ok;
4391}
4392
4393static inline void __netif_tx_unlock(struct netdev_queue *txq)
4394{
4395 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4396 WRITE_ONCE(txq->xmit_lock_owner, -1);
4397 spin_unlock(&txq->_xmit_lock);
4398}
4399
4400static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4401{
4402 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4403 WRITE_ONCE(txq->xmit_lock_owner, -1);
4404 spin_unlock_bh(&txq->_xmit_lock);
4405}
4406
4407/*
4408 * txq->trans_start can be read locklessly from dev_watchdog()
4409 */
4410static inline void txq_trans_update(struct netdev_queue *txq)
4411{
4412 if (txq->xmit_lock_owner != -1)
4413 WRITE_ONCE(txq->trans_start, jiffies);
4414}
4415
4416static inline void txq_trans_cond_update(struct netdev_queue *txq)
4417{
4418 unsigned long now = jiffies;
4419
4420 if (READ_ONCE(txq->trans_start) != now)
4421 WRITE_ONCE(txq->trans_start, now);
4422}
4423
4424/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4425static inline void netif_trans_update(struct net_device *dev)
4426{
4427 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4428
4429 txq_trans_cond_update(txq);
4430}
4431
4432/**
4433 * netif_tx_lock - grab network device transmit lock
4434 * @dev: network device
4435 *
4436 * Get network device transmit lock
4437 */
4438void netif_tx_lock(struct net_device *dev);
4439
4440static inline void netif_tx_lock_bh(struct net_device *dev)
4441{
4442 local_bh_disable();
4443 netif_tx_lock(dev);
4444}
4445
4446void netif_tx_unlock(struct net_device *dev);
4447
4448static inline void netif_tx_unlock_bh(struct net_device *dev)
4449{
4450 netif_tx_unlock(dev);
4451 local_bh_enable();
4452}
4453
4454#define HARD_TX_LOCK(dev, txq, cpu) { \
4455 if (!(dev)->lltx) { \
4456 __netif_tx_lock(txq, cpu); \
4457 } else { \
4458 __netif_tx_acquire(txq); \
4459 } \
4460}
4461
4462#define HARD_TX_TRYLOCK(dev, txq) \
4463 (!(dev)->lltx ? \
4464 __netif_tx_trylock(txq) : \
4465 __netif_tx_acquire(txq))
4466
4467#define HARD_TX_UNLOCK(dev, txq) { \
4468 if (!(dev)->lltx) { \
4469 __netif_tx_unlock(txq); \
4470 } else { \
4471 __netif_tx_release(txq); \
4472 } \
4473}
4474
4475static inline void netif_tx_disable(struct net_device *dev)
4476{
4477 unsigned int i;
4478 int cpu;
4479
4480 local_bh_disable();
4481 cpu = smp_processor_id();
4482 spin_lock(&dev->tx_global_lock);
4483 for (i = 0; i < dev->num_tx_queues; i++) {
4484 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4485
4486 __netif_tx_lock(txq, cpu);
4487 netif_tx_stop_queue(txq);
4488 __netif_tx_unlock(txq);
4489 }
4490 spin_unlock(&dev->tx_global_lock);
4491 local_bh_enable();
4492}
4493
4494static inline void netif_addr_lock(struct net_device *dev)
4495{
4496 unsigned char nest_level = 0;
4497
4498#ifdef CONFIG_LOCKDEP
4499 nest_level = dev->nested_level;
4500#endif
4501 spin_lock_nested(&dev->addr_list_lock, nest_level);
4502}
4503
4504static inline void netif_addr_lock_bh(struct net_device *dev)
4505{
4506 unsigned char nest_level = 0;
4507
4508#ifdef CONFIG_LOCKDEP
4509 nest_level = dev->nested_level;
4510#endif
4511 local_bh_disable();
4512 spin_lock_nested(&dev->addr_list_lock, nest_level);
4513}
4514
4515static inline void netif_addr_unlock(struct net_device *dev)
4516{
4517 spin_unlock(&dev->addr_list_lock);
4518}
4519
4520static inline void netif_addr_unlock_bh(struct net_device *dev)
4521{
4522 spin_unlock_bh(&dev->addr_list_lock);
4523}
4524
4525/*
4526 * dev_addrs walker. Should be used only for read access. Call with
4527 * rcu_read_lock held.
4528 */
4529#define for_each_dev_addr(dev, ha) \
4530 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4531
4532/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4533
4534void ether_setup(struct net_device *dev);
4535
4536/* Allocate dummy net_device */
4537struct net_device *alloc_netdev_dummy(int sizeof_priv);
4538
4539/* Support for loadable net-drivers */
4540struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4541 unsigned char name_assign_type,
4542 void (*setup)(struct net_device *),
4543 unsigned int txqs, unsigned int rxqs);
4544#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4545 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4546
4547#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4548 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4549 count)
4550
4551int register_netdev(struct net_device *dev);
4552void unregister_netdev(struct net_device *dev);
4553
4554int devm_register_netdev(struct device *dev, struct net_device *ndev);
4555
4556/* General hardware address lists handling functions */
4557int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4558 struct netdev_hw_addr_list *from_list, int addr_len);
4559void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4560 struct netdev_hw_addr_list *from_list, int addr_len);
4561int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4562 struct net_device *dev,
4563 int (*sync)(struct net_device *, const unsigned char *),
4564 int (*unsync)(struct net_device *,
4565 const unsigned char *));
4566int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4567 struct net_device *dev,
4568 int (*sync)(struct net_device *,
4569 const unsigned char *, int),
4570 int (*unsync)(struct net_device *,
4571 const unsigned char *, int));
4572void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4573 struct net_device *dev,
4574 int (*unsync)(struct net_device *,
4575 const unsigned char *, int));
4576void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4577 struct net_device *dev,
4578 int (*unsync)(struct net_device *,
4579 const unsigned char *));
4580void __hw_addr_init(struct netdev_hw_addr_list *list);
4581
4582/* Functions used for device addresses handling */
4583void dev_addr_mod(struct net_device *dev, unsigned int offset,
4584 const void *addr, size_t len);
4585
4586static inline void
4587__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4588{
4589 dev_addr_mod(dev, 0, addr, len);
4590}
4591
4592static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4593{
4594 __dev_addr_set(dev, addr, dev->addr_len);
4595}
4596
4597int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4598 unsigned char addr_type);
4599int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4600 unsigned char addr_type);
4601
4602/* Functions used for unicast addresses handling */
4603int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4604int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4605int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4606int dev_uc_sync(struct net_device *to, struct net_device *from);
4607int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4608void dev_uc_unsync(struct net_device *to, struct net_device *from);
4609void dev_uc_flush(struct net_device *dev);
4610void dev_uc_init(struct net_device *dev);
4611
4612/**
4613 * __dev_uc_sync - Synchronize device's unicast list
4614 * @dev: device to sync
4615 * @sync: function to call if address should be added
4616 * @unsync: function to call if address should be removed
4617 *
4618 * Add newly added addresses to the interface, and release
4619 * addresses that have been deleted.
4620 */
4621static inline int __dev_uc_sync(struct net_device *dev,
4622 int (*sync)(struct net_device *,
4623 const unsigned char *),
4624 int (*unsync)(struct net_device *,
4625 const unsigned char *))
4626{
4627 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4628}
4629
4630/**
4631 * __dev_uc_unsync - Remove synchronized addresses from device
4632 * @dev: device to sync
4633 * @unsync: function to call if address should be removed
4634 *
4635 * Remove all addresses that were added to the device by dev_uc_sync().
4636 */
4637static inline void __dev_uc_unsync(struct net_device *dev,
4638 int (*unsync)(struct net_device *,
4639 const unsigned char *))
4640{
4641 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4642}
4643
4644/* Functions used for multicast addresses handling */
4645int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4646int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4647int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4648int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4649int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4650int dev_mc_sync(struct net_device *to, struct net_device *from);
4651int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4652void dev_mc_unsync(struct net_device *to, struct net_device *from);
4653void dev_mc_flush(struct net_device *dev);
4654void dev_mc_init(struct net_device *dev);
4655
4656/**
4657 * __dev_mc_sync - Synchronize device's multicast list
4658 * @dev: device to sync
4659 * @sync: function to call if address should be added
4660 * @unsync: function to call if address should be removed
4661 *
4662 * Add newly added addresses to the interface, and release
4663 * addresses that have been deleted.
4664 */
4665static inline int __dev_mc_sync(struct net_device *dev,
4666 int (*sync)(struct net_device *,
4667 const unsigned char *),
4668 int (*unsync)(struct net_device *,
4669 const unsigned char *))
4670{
4671 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4672}
4673
4674/**
4675 * __dev_mc_unsync - Remove synchronized addresses from device
4676 * @dev: device to sync
4677 * @unsync: function to call if address should be removed
4678 *
4679 * Remove all addresses that were added to the device by dev_mc_sync().
4680 */
4681static inline void __dev_mc_unsync(struct net_device *dev,
4682 int (*unsync)(struct net_device *,
4683 const unsigned char *))
4684{
4685 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4686}
4687
4688/* Functions used for secondary unicast and multicast support */
4689void dev_set_rx_mode(struct net_device *dev);
4690int dev_set_promiscuity(struct net_device *dev, int inc);
4691int dev_set_allmulti(struct net_device *dev, int inc);
4692void netdev_state_change(struct net_device *dev);
4693void __netdev_notify_peers(struct net_device *dev);
4694void netdev_notify_peers(struct net_device *dev);
4695void netdev_features_change(struct net_device *dev);
4696/* Load a device via the kmod */
4697void dev_load(struct net *net, const char *name);
4698struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4699 struct rtnl_link_stats64 *storage);
4700void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4701 const struct net_device_stats *netdev_stats);
4702void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4703 const struct pcpu_sw_netstats __percpu *netstats);
4704void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4705
4706enum {
4707 NESTED_SYNC_IMM_BIT,
4708 NESTED_SYNC_TODO_BIT,
4709};
4710
4711#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4712#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4713
4714#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4715#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4716
4717struct netdev_nested_priv {
4718 unsigned char flags;
4719 void *data;
4720};
4721
4722bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4723struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4724 struct list_head **iter);
4725
4726/* iterate through upper list, must be called under RCU read lock */
4727#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4728 for (iter = &(dev)->adj_list.upper, \
4729 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4730 updev; \
4731 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4732
4733int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4734 int (*fn)(struct net_device *upper_dev,
4735 struct netdev_nested_priv *priv),
4736 struct netdev_nested_priv *priv);
4737
4738bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4739 struct net_device *upper_dev);
4740
4741bool netdev_has_any_upper_dev(struct net_device *dev);
4742
4743void *netdev_lower_get_next_private(struct net_device *dev,
4744 struct list_head **iter);
4745void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4746 struct list_head **iter);
4747
4748#define netdev_for_each_lower_private(dev, priv, iter) \
4749 for (iter = (dev)->adj_list.lower.next, \
4750 priv = netdev_lower_get_next_private(dev, &(iter)); \
4751 priv; \
4752 priv = netdev_lower_get_next_private(dev, &(iter)))
4753
4754#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4755 for (iter = &(dev)->adj_list.lower, \
4756 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4757 priv; \
4758 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4759
4760void *netdev_lower_get_next(struct net_device *dev,
4761 struct list_head **iter);
4762
4763#define netdev_for_each_lower_dev(dev, ldev, iter) \
4764 for (iter = (dev)->adj_list.lower.next, \
4765 ldev = netdev_lower_get_next(dev, &(iter)); \
4766 ldev; \
4767 ldev = netdev_lower_get_next(dev, &(iter)))
4768
4769struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4770 struct list_head **iter);
4771int netdev_walk_all_lower_dev(struct net_device *dev,
4772 int (*fn)(struct net_device *lower_dev,
4773 struct netdev_nested_priv *priv),
4774 struct netdev_nested_priv *priv);
4775int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4776 int (*fn)(struct net_device *lower_dev,
4777 struct netdev_nested_priv *priv),
4778 struct netdev_nested_priv *priv);
4779
4780void *netdev_adjacent_get_private(struct list_head *adj_list);
4781void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4782struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4783struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4784int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4785 struct netlink_ext_ack *extack);
4786int netdev_master_upper_dev_link(struct net_device *dev,
4787 struct net_device *upper_dev,
4788 void *upper_priv, void *upper_info,
4789 struct netlink_ext_ack *extack);
4790void netdev_upper_dev_unlink(struct net_device *dev,
4791 struct net_device *upper_dev);
4792int netdev_adjacent_change_prepare(struct net_device *old_dev,
4793 struct net_device *new_dev,
4794 struct net_device *dev,
4795 struct netlink_ext_ack *extack);
4796void netdev_adjacent_change_commit(struct net_device *old_dev,
4797 struct net_device *new_dev,
4798 struct net_device *dev);
4799void netdev_adjacent_change_abort(struct net_device *old_dev,
4800 struct net_device *new_dev,
4801 struct net_device *dev);
4802void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4803void *netdev_lower_dev_get_private(struct net_device *dev,
4804 struct net_device *lower_dev);
4805void netdev_lower_state_changed(struct net_device *lower_dev,
4806 void *lower_state_info);
4807
4808/* RSS keys are 40 or 52 bytes long */
4809#define NETDEV_RSS_KEY_LEN 52
4810extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4811void netdev_rss_key_fill(void *buffer, size_t len);
4812
4813int skb_checksum_help(struct sk_buff *skb);
4814int skb_crc32c_csum_help(struct sk_buff *skb);
4815int skb_csum_hwoffload_help(struct sk_buff *skb,
4816 const netdev_features_t features);
4817
4818struct netdev_bonding_info {
4819 ifslave slave;
4820 ifbond master;
4821};
4822
4823struct netdev_notifier_bonding_info {
4824 struct netdev_notifier_info info; /* must be first */
4825 struct netdev_bonding_info bonding_info;
4826};
4827
4828void netdev_bonding_info_change(struct net_device *dev,
4829 struct netdev_bonding_info *bonding_info);
4830
4831#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4832void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4833#else
4834static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4835 const void *data)
4836{
4837}
4838#endif
4839
4840__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4841
4842static inline bool can_checksum_protocol(netdev_features_t features,
4843 __be16 protocol)
4844{
4845 if (protocol == htons(ETH_P_FCOE))
4846 return !!(features & NETIF_F_FCOE_CRC);
4847
4848 /* Assume this is an IP checksum (not SCTP CRC) */
4849
4850 if (features & NETIF_F_HW_CSUM) {
4851 /* Can checksum everything */
4852 return true;
4853 }
4854
4855 switch (protocol) {
4856 case htons(ETH_P_IP):
4857 return !!(features & NETIF_F_IP_CSUM);
4858 case htons(ETH_P_IPV6):
4859 return !!(features & NETIF_F_IPV6_CSUM);
4860 default:
4861 return false;
4862 }
4863}
4864
4865#ifdef CONFIG_BUG
4866void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4867#else
4868static inline void netdev_rx_csum_fault(struct net_device *dev,
4869 struct sk_buff *skb)
4870{
4871}
4872#endif
4873/* rx skb timestamps */
4874void net_enable_timestamp(void);
4875void net_disable_timestamp(void);
4876
4877static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4878 const struct skb_shared_hwtstamps *hwtstamps,
4879 bool cycles)
4880{
4881 const struct net_device_ops *ops = dev->netdev_ops;
4882
4883 if (ops->ndo_get_tstamp)
4884 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4885
4886 return hwtstamps->hwtstamp;
4887}
4888
4889#ifndef CONFIG_PREEMPT_RT
4890static inline void netdev_xmit_set_more(bool more)
4891{
4892 __this_cpu_write(softnet_data.xmit.more, more);
4893}
4894
4895static inline bool netdev_xmit_more(void)
4896{
4897 return __this_cpu_read(softnet_data.xmit.more);
4898}
4899#else
4900static inline void netdev_xmit_set_more(bool more)
4901{
4902 current->net_xmit.more = more;
4903}
4904
4905static inline bool netdev_xmit_more(void)
4906{
4907 return current->net_xmit.more;
4908}
4909#endif
4910
4911static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4912 struct sk_buff *skb, struct net_device *dev,
4913 bool more)
4914{
4915 netdev_xmit_set_more(more);
4916 return ops->ndo_start_xmit(skb, dev);
4917}
4918
4919static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4920 struct netdev_queue *txq, bool more)
4921{
4922 const struct net_device_ops *ops = dev->netdev_ops;
4923 netdev_tx_t rc;
4924
4925 rc = __netdev_start_xmit(ops, skb, dev, more);
4926 if (rc == NETDEV_TX_OK)
4927 txq_trans_update(txq);
4928
4929 return rc;
4930}
4931
4932int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4933 const void *ns);
4934void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4935 const void *ns);
4936
4937extern const struct kobj_ns_type_operations net_ns_type_operations;
4938
4939const char *netdev_drivername(const struct net_device *dev);
4940
4941static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4942 netdev_features_t f2)
4943{
4944 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4945 if (f1 & NETIF_F_HW_CSUM)
4946 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4947 else
4948 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4949 }
4950
4951 return f1 & f2;
4952}
4953
4954static inline netdev_features_t netdev_get_wanted_features(
4955 struct net_device *dev)
4956{
4957 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4958}
4959netdev_features_t netdev_increment_features(netdev_features_t all,
4960 netdev_features_t one, netdev_features_t mask);
4961
4962/* Allow TSO being used on stacked device :
4963 * Performing the GSO segmentation before last device
4964 * is a performance improvement.
4965 */
4966static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4967 netdev_features_t mask)
4968{
4969 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4970}
4971
4972int __netdev_update_features(struct net_device *dev);
4973void netdev_update_features(struct net_device *dev);
4974void netdev_change_features(struct net_device *dev);
4975
4976void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4977 struct net_device *dev);
4978
4979netdev_features_t passthru_features_check(struct sk_buff *skb,
4980 struct net_device *dev,
4981 netdev_features_t features);
4982netdev_features_t netif_skb_features(struct sk_buff *skb);
4983void skb_warn_bad_offload(const struct sk_buff *skb);
4984
4985static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4986{
4987 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4988
4989 /* check flags correspondence */
4990 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4991 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4992 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4993 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4994 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4995 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4996 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4997 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4998 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4999 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5000 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5001 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5002 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5003 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5004 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5005 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5006 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5007 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5008 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5009
5010 return (features & feature) == feature;
5011}
5012
5013static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5014{
5015 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5016 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5017}
5018
5019static inline bool netif_needs_gso(struct sk_buff *skb,
5020 netdev_features_t features)
5021{
5022 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5023 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5024 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5025}
5026
5027void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5028void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5029void netif_inherit_tso_max(struct net_device *to,
5030 const struct net_device *from);
5031
5032static inline unsigned int
5033netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb)
5034{
5035 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */
5036 return skb->protocol == htons(ETH_P_IPV6) ?
5037 READ_ONCE(dev->gro_max_size) :
5038 READ_ONCE(dev->gro_ipv4_max_size);
5039}
5040
5041static inline unsigned int
5042netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb)
5043{
5044 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */
5045 return skb->protocol == htons(ETH_P_IPV6) ?
5046 READ_ONCE(dev->gso_max_size) :
5047 READ_ONCE(dev->gso_ipv4_max_size);
5048}
5049
5050static inline bool netif_is_macsec(const struct net_device *dev)
5051{
5052 return dev->priv_flags & IFF_MACSEC;
5053}
5054
5055static inline bool netif_is_macvlan(const struct net_device *dev)
5056{
5057 return dev->priv_flags & IFF_MACVLAN;
5058}
5059
5060static inline bool netif_is_macvlan_port(const struct net_device *dev)
5061{
5062 return dev->priv_flags & IFF_MACVLAN_PORT;
5063}
5064
5065static inline bool netif_is_bond_master(const struct net_device *dev)
5066{
5067 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5068}
5069
5070static inline bool netif_is_bond_slave(const struct net_device *dev)
5071{
5072 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5073}
5074
5075static inline bool netif_supports_nofcs(struct net_device *dev)
5076{
5077 return dev->priv_flags & IFF_SUPP_NOFCS;
5078}
5079
5080static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5081{
5082 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5083}
5084
5085static inline bool netif_is_l3_master(const struct net_device *dev)
5086{
5087 return dev->priv_flags & IFF_L3MDEV_MASTER;
5088}
5089
5090static inline bool netif_is_l3_slave(const struct net_device *dev)
5091{
5092 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5093}
5094
5095static inline int dev_sdif(const struct net_device *dev)
5096{
5097#ifdef CONFIG_NET_L3_MASTER_DEV
5098 if (netif_is_l3_slave(dev))
5099 return dev->ifindex;
5100#endif
5101 return 0;
5102}
5103
5104static inline bool netif_is_bridge_master(const struct net_device *dev)
5105{
5106 return dev->priv_flags & IFF_EBRIDGE;
5107}
5108
5109static inline bool netif_is_bridge_port(const struct net_device *dev)
5110{
5111 return dev->priv_flags & IFF_BRIDGE_PORT;
5112}
5113
5114static inline bool netif_is_ovs_master(const struct net_device *dev)
5115{
5116 return dev->priv_flags & IFF_OPENVSWITCH;
5117}
5118
5119static inline bool netif_is_ovs_port(const struct net_device *dev)
5120{
5121 return dev->priv_flags & IFF_OVS_DATAPATH;
5122}
5123
5124static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5125{
5126 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5127}
5128
5129static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5130{
5131 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5132}
5133
5134static inline bool netif_is_team_master(const struct net_device *dev)
5135{
5136 return dev->priv_flags & IFF_TEAM;
5137}
5138
5139static inline bool netif_is_team_port(const struct net_device *dev)
5140{
5141 return dev->priv_flags & IFF_TEAM_PORT;
5142}
5143
5144static inline bool netif_is_lag_master(const struct net_device *dev)
5145{
5146 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5147}
5148
5149static inline bool netif_is_lag_port(const struct net_device *dev)
5150{
5151 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5152}
5153
5154static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5155{
5156 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5157}
5158
5159static inline bool netif_is_failover(const struct net_device *dev)
5160{
5161 return dev->priv_flags & IFF_FAILOVER;
5162}
5163
5164static inline bool netif_is_failover_slave(const struct net_device *dev)
5165{
5166 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5167}
5168
5169/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5170static inline void netif_keep_dst(struct net_device *dev)
5171{
5172 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5173}
5174
5175/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5176static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5177{
5178 /* TODO: reserve and use an additional IFF bit, if we get more users */
5179 return netif_is_macsec(dev);
5180}
5181
5182extern struct pernet_operations __net_initdata loopback_net_ops;
5183
5184/* Logging, debugging and troubleshooting/diagnostic helpers. */
5185
5186/* netdev_printk helpers, similar to dev_printk */
5187
5188static inline const char *netdev_name(const struct net_device *dev)
5189{
5190 if (!dev->name[0] || strchr(dev->name, '%'))
5191 return "(unnamed net_device)";
5192 return dev->name;
5193}
5194
5195static inline const char *netdev_reg_state(const struct net_device *dev)
5196{
5197 u8 reg_state = READ_ONCE(dev->reg_state);
5198
5199 switch (reg_state) {
5200 case NETREG_UNINITIALIZED: return " (uninitialized)";
5201 case NETREG_REGISTERED: return "";
5202 case NETREG_UNREGISTERING: return " (unregistering)";
5203 case NETREG_UNREGISTERED: return " (unregistered)";
5204 case NETREG_RELEASED: return " (released)";
5205 case NETREG_DUMMY: return " (dummy)";
5206 }
5207
5208 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state);
5209 return " (unknown)";
5210}
5211
5212#define MODULE_ALIAS_NETDEV(device) \
5213 MODULE_ALIAS("netdev-" device)
5214
5215/*
5216 * netdev_WARN() acts like dev_printk(), but with the key difference
5217 * of using a WARN/WARN_ON to get the message out, including the
5218 * file/line information and a backtrace.
5219 */
5220#define netdev_WARN(dev, format, args...) \
5221 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5222 netdev_reg_state(dev), ##args)
5223
5224#define netdev_WARN_ONCE(dev, format, args...) \
5225 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5226 netdev_reg_state(dev), ##args)
5227
5228/*
5229 * The list of packet types we will receive (as opposed to discard)
5230 * and the routines to invoke.
5231 *
5232 * Why 16. Because with 16 the only overlap we get on a hash of the
5233 * low nibble of the protocol value is RARP/SNAP/X.25.
5234 *
5235 * 0800 IP
5236 * 0001 802.3
5237 * 0002 AX.25
5238 * 0004 802.2
5239 * 8035 RARP
5240 * 0005 SNAP
5241 * 0805 X.25
5242 * 0806 ARP
5243 * 8137 IPX
5244 * 0009 Localtalk
5245 * 86DD IPv6
5246 */
5247#define PTYPE_HASH_SIZE (16)
5248#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5249
5250extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5251
5252extern struct net_device *blackhole_netdev;
5253
5254/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5255#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5256#define DEV_STATS_ADD(DEV, FIELD, VAL) \
5257 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5258#define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5259
5260#endif /* _LINUX_NETDEVICE_H */