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