Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
24#include <linux/timer.h>
25#include <linux/bug.h>
26#include <linux/delay.h>
27#include <linux/atomic.h>
28#include <linux/prefetch.h>
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31#include <asm/local.h>
32
33#include <linux/percpu.h>
34#include <linux/rculist.h>
35#include <linux/workqueue.h>
36#include <linux/dynamic_queue_limits.h>
37
38#include <net/net_namespace.h>
39#ifdef CONFIG_DCB
40#include <net/dcbnl.h>
41#endif
42#include <net/netprio_cgroup.h>
43#include <net/xdp.h>
44
45#include <linux/netdev_features.h>
46#include <linux/neighbour.h>
47#include <uapi/linux/netdevice.h>
48#include <uapi/linux/if_bonding.h>
49#include <uapi/linux/pkt_cls.h>
50#include <uapi/linux/netdev.h>
51#include <linux/hashtable.h>
52#include <linux/rbtree.h>
53#include <net/net_trackers.h>
54#include <net/net_debug.h>
55#include <net/dropreason-core.h>
56
57struct netpoll_info;
58struct device;
59struct ethtool_ops;
60struct phy_device;
61struct dsa_port;
62struct ip_tunnel_parm;
63struct macsec_context;
64struct macsec_ops;
65struct netdev_name_node;
66struct sd_flow_limit;
67struct sfp_bus;
68/* 802.11 specific */
69struct wireless_dev;
70/* 802.15.4 specific */
71struct wpan_dev;
72struct mpls_dev;
73/* UDP Tunnel offloads */
74struct udp_tunnel_info;
75struct udp_tunnel_nic_info;
76struct udp_tunnel_nic;
77struct bpf_prog;
78struct xdp_buff;
79struct xdp_md;
80
81void synchronize_net(void);
82void netdev_set_default_ethtool_ops(struct net_device *dev,
83 const struct ethtool_ops *ops);
84void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
85
86/* Backlog congestion levels */
87#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
88#define NET_RX_DROP 1 /* packet dropped */
89
90#define MAX_NEST_DEV 8
91
92/*
93 * Transmit return codes: transmit return codes originate from three different
94 * namespaces:
95 *
96 * - qdisc return codes
97 * - driver transmit return codes
98 * - errno values
99 *
100 * Drivers are allowed to return any one of those in their hard_start_xmit()
101 * function. Real network devices commonly used with qdiscs should only return
102 * the driver transmit return codes though - when qdiscs are used, the actual
103 * transmission happens asynchronously, so the value is not propagated to
104 * higher layers. Virtual network devices transmit synchronously; in this case
105 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
106 * others are propagated to higher layers.
107 */
108
109/* qdisc ->enqueue() return codes. */
110#define NET_XMIT_SUCCESS 0x00
111#define NET_XMIT_DROP 0x01 /* skb dropped */
112#define NET_XMIT_CN 0x02 /* congestion notification */
113#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
114
115/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
116 * indicates that the device will soon be dropping packets, or already drops
117 * some packets of the same priority; prompting us to send less aggressively. */
118#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
119#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
120
121/* Driver transmit return codes */
122#define NETDEV_TX_MASK 0xf0
123
124enum netdev_tx {
125 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
126 NETDEV_TX_OK = 0x00, /* driver took care of packet */
127 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
128};
129typedef enum netdev_tx netdev_tx_t;
130
131/*
132 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
133 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
134 */
135static inline bool dev_xmit_complete(int rc)
136{
137 /*
138 * Positive cases with an skb consumed by a driver:
139 * - successful transmission (rc == NETDEV_TX_OK)
140 * - error while transmitting (rc < 0)
141 * - error while queueing to a different device (rc & NET_XMIT_MASK)
142 */
143 if (likely(rc < NET_XMIT_MASK))
144 return true;
145
146 return false;
147}
148
149/*
150 * Compute the worst-case header length according to the protocols
151 * used.
152 */
153
154#if defined(CONFIG_HYPERV_NET)
155# define LL_MAX_HEADER 128
156#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
157# if defined(CONFIG_MAC80211_MESH)
158# define LL_MAX_HEADER 128
159# else
160# define LL_MAX_HEADER 96
161# endif
162#else
163# define LL_MAX_HEADER 32
164#endif
165
166#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
167 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
168#define MAX_HEADER LL_MAX_HEADER
169#else
170#define MAX_HEADER (LL_MAX_HEADER + 48)
171#endif
172
173/*
174 * Old network device statistics. Fields are native words
175 * (unsigned long) so they can be read and written atomically.
176 */
177
178#define NET_DEV_STAT(FIELD) \
179 union { \
180 unsigned long FIELD; \
181 atomic_long_t __##FIELD; \
182 }
183
184struct net_device_stats {
185 NET_DEV_STAT(rx_packets);
186 NET_DEV_STAT(tx_packets);
187 NET_DEV_STAT(rx_bytes);
188 NET_DEV_STAT(tx_bytes);
189 NET_DEV_STAT(rx_errors);
190 NET_DEV_STAT(tx_errors);
191 NET_DEV_STAT(rx_dropped);
192 NET_DEV_STAT(tx_dropped);
193 NET_DEV_STAT(multicast);
194 NET_DEV_STAT(collisions);
195 NET_DEV_STAT(rx_length_errors);
196 NET_DEV_STAT(rx_over_errors);
197 NET_DEV_STAT(rx_crc_errors);
198 NET_DEV_STAT(rx_frame_errors);
199 NET_DEV_STAT(rx_fifo_errors);
200 NET_DEV_STAT(rx_missed_errors);
201 NET_DEV_STAT(tx_aborted_errors);
202 NET_DEV_STAT(tx_carrier_errors);
203 NET_DEV_STAT(tx_fifo_errors);
204 NET_DEV_STAT(tx_heartbeat_errors);
205 NET_DEV_STAT(tx_window_errors);
206 NET_DEV_STAT(rx_compressed);
207 NET_DEV_STAT(tx_compressed);
208};
209#undef NET_DEV_STAT
210
211/* per-cpu stats, allocated on demand.
212 * Try to fit them in a single cache line, for dev_get_stats() sake.
213 */
214struct net_device_core_stats {
215 unsigned long rx_dropped;
216 unsigned long tx_dropped;
217 unsigned long rx_nohandler;
218 unsigned long rx_otherhost_dropped;
219} __aligned(4 * sizeof(unsigned long));
220
221#include <linux/cache.h>
222#include <linux/skbuff.h>
223
224#ifdef CONFIG_RPS
225#include <linux/static_key.h>
226extern struct static_key_false rps_needed;
227extern struct static_key_false rfs_needed;
228#endif
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};
380
381enum {
382 NAPI_STATE_SCHED, /* Poll is scheduled */
383 NAPI_STATE_MISSED, /* reschedule a napi */
384 NAPI_STATE_DISABLE, /* Disable pending */
385 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
386 NAPI_STATE_LISTED, /* NAPI added to system lists */
387 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
388 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
389 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
390 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
391 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
392};
393
394enum {
395 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
396 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
397 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
398 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
399 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
400 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
401 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
402 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
403 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
404 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
405};
406
407enum gro_result {
408 GRO_MERGED,
409 GRO_MERGED_FREE,
410 GRO_HELD,
411 GRO_NORMAL,
412 GRO_CONSUMED,
413};
414typedef enum gro_result gro_result_t;
415
416/*
417 * enum rx_handler_result - Possible return values for rx_handlers.
418 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
419 * further.
420 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
421 * case skb->dev was changed by rx_handler.
422 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
423 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
424 *
425 * rx_handlers are functions called from inside __netif_receive_skb(), to do
426 * special processing of the skb, prior to delivery to protocol handlers.
427 *
428 * Currently, a net_device can only have a single rx_handler registered. Trying
429 * to register a second rx_handler will return -EBUSY.
430 *
431 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
432 * To unregister a rx_handler on a net_device, use
433 * netdev_rx_handler_unregister().
434 *
435 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
436 * do with the skb.
437 *
438 * If the rx_handler consumed the skb in some way, it should return
439 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
440 * the skb to be delivered in some other way.
441 *
442 * If the rx_handler changed skb->dev, to divert the skb to another
443 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
444 * new device will be called if it exists.
445 *
446 * If the rx_handler decides the skb should be ignored, it should return
447 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
448 * are registered on exact device (ptype->dev == skb->dev).
449 *
450 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
451 * delivered, it should return RX_HANDLER_PASS.
452 *
453 * A device without a registered rx_handler will behave as if rx_handler
454 * returned RX_HANDLER_PASS.
455 */
456
457enum rx_handler_result {
458 RX_HANDLER_CONSUMED,
459 RX_HANDLER_ANOTHER,
460 RX_HANDLER_EXACT,
461 RX_HANDLER_PASS,
462};
463typedef enum rx_handler_result rx_handler_result_t;
464typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
465
466void __napi_schedule(struct napi_struct *n);
467void __napi_schedule_irqoff(struct napi_struct *n);
468
469static inline bool napi_disable_pending(struct napi_struct *n)
470{
471 return test_bit(NAPI_STATE_DISABLE, &n->state);
472}
473
474static inline bool napi_prefer_busy_poll(struct napi_struct *n)
475{
476 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
477}
478
479bool napi_schedule_prep(struct napi_struct *n);
480
481/**
482 * napi_schedule - schedule NAPI poll
483 * @n: NAPI context
484 *
485 * Schedule NAPI poll routine to be called if it is not already
486 * running.
487 */
488static inline void napi_schedule(struct napi_struct *n)
489{
490 if (napi_schedule_prep(n))
491 __napi_schedule(n);
492}
493
494/**
495 * napi_schedule_irqoff - schedule NAPI poll
496 * @n: NAPI context
497 *
498 * Variant of napi_schedule(), assuming hard irqs are masked.
499 */
500static inline void napi_schedule_irqoff(struct napi_struct *n)
501{
502 if (napi_schedule_prep(n))
503 __napi_schedule_irqoff(n);
504}
505
506/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
507static inline bool napi_reschedule(struct napi_struct *napi)
508{
509 if (napi_schedule_prep(napi)) {
510 __napi_schedule(napi);
511 return true;
512 }
513 return false;
514}
515
516/**
517 * napi_complete_done - NAPI processing complete
518 * @n: NAPI context
519 * @work_done: number of packets processed
520 *
521 * Mark NAPI processing as complete. Should only be called if poll budget
522 * has not been completely consumed.
523 * Prefer over napi_complete().
524 * Return false if device should avoid rearming interrupts.
525 */
526bool napi_complete_done(struct napi_struct *n, int work_done);
527
528static inline bool napi_complete(struct napi_struct *n)
529{
530 return napi_complete_done(n, 0);
531}
532
533int dev_set_threaded(struct net_device *dev, bool threaded);
534
535/**
536 * napi_disable - prevent NAPI from scheduling
537 * @n: NAPI context
538 *
539 * Stop NAPI from being scheduled on this context.
540 * Waits till any outstanding processing completes.
541 */
542void napi_disable(struct napi_struct *n);
543
544void napi_enable(struct napi_struct *n);
545
546/**
547 * napi_synchronize - wait until NAPI is not running
548 * @n: NAPI context
549 *
550 * Wait until NAPI is done being scheduled on this context.
551 * Waits till any outstanding processing completes but
552 * does not disable future activations.
553 */
554static inline void napi_synchronize(const struct napi_struct *n)
555{
556 if (IS_ENABLED(CONFIG_SMP))
557 while (test_bit(NAPI_STATE_SCHED, &n->state))
558 msleep(1);
559 else
560 barrier();
561}
562
563/**
564 * napi_if_scheduled_mark_missed - if napi is running, set the
565 * NAPIF_STATE_MISSED
566 * @n: NAPI context
567 *
568 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
569 * NAPI is scheduled.
570 **/
571static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
572{
573 unsigned long val, new;
574
575 val = READ_ONCE(n->state);
576 do {
577 if (val & NAPIF_STATE_DISABLE)
578 return true;
579
580 if (!(val & NAPIF_STATE_SCHED))
581 return false;
582
583 new = val | NAPIF_STATE_MISSED;
584 } while (!try_cmpxchg(&n->state, &val, new));
585
586 return true;
587}
588
589enum netdev_queue_state_t {
590 __QUEUE_STATE_DRV_XOFF,
591 __QUEUE_STATE_STACK_XOFF,
592 __QUEUE_STATE_FROZEN,
593};
594
595#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
596#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
597#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
598
599#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
600#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
601 QUEUE_STATE_FROZEN)
602#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
603 QUEUE_STATE_FROZEN)
604
605/*
606 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
607 * netif_tx_* functions below are used to manipulate this flag. The
608 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
609 * queue independently. The netif_xmit_*stopped functions below are called
610 * to check if the queue has been stopped by the driver or stack (either
611 * of the XOFF bits are set in the state). Drivers should not need to call
612 * netif_xmit*stopped functions, they should only be using netif_tx_*.
613 */
614
615struct netdev_queue {
616/*
617 * read-mostly part
618 */
619 struct net_device *dev;
620 netdevice_tracker dev_tracker;
621
622 struct Qdisc __rcu *qdisc;
623 struct Qdisc __rcu *qdisc_sleeping;
624#ifdef CONFIG_SYSFS
625 struct kobject kobj;
626#endif
627#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
628 int numa_node;
629#endif
630 unsigned long tx_maxrate;
631 /*
632 * Number of TX timeouts for this queue
633 * (/sys/class/net/DEV/Q/trans_timeout)
634 */
635 atomic_long_t trans_timeout;
636
637 /* Subordinate device that the queue has been assigned to */
638 struct net_device *sb_dev;
639#ifdef CONFIG_XDP_SOCKETS
640 struct xsk_buff_pool *pool;
641#endif
642/*
643 * write-mostly part
644 */
645 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
646 int xmit_lock_owner;
647 /*
648 * Time (in jiffies) of last Tx
649 */
650 unsigned long trans_start;
651
652 unsigned long state;
653
654#ifdef CONFIG_BQL
655 struct dql dql;
656#endif
657} ____cacheline_aligned_in_smp;
658
659extern int sysctl_fb_tunnels_only_for_init_net;
660extern int sysctl_devconf_inherit_init_net;
661
662/*
663 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
664 * == 1 : For initns only
665 * == 2 : For none.
666 */
667static inline bool net_has_fallback_tunnels(const struct net *net)
668{
669#if IS_ENABLED(CONFIG_SYSCTL)
670 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
671
672 return !fb_tunnels_only_for_init_net ||
673 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
674#else
675 return true;
676#endif
677}
678
679static inline int net_inherit_devconf(void)
680{
681#if IS_ENABLED(CONFIG_SYSCTL)
682 return READ_ONCE(sysctl_devconf_inherit_init_net);
683#else
684 return 0;
685#endif
686}
687
688static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
689{
690#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
691 return q->numa_node;
692#else
693 return NUMA_NO_NODE;
694#endif
695}
696
697static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
698{
699#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
700 q->numa_node = node;
701#endif
702}
703
704#ifdef CONFIG_RPS
705/*
706 * This structure holds an RPS map which can be of variable length. The
707 * map is an array of CPUs.
708 */
709struct rps_map {
710 unsigned int len;
711 struct rcu_head rcu;
712 u16 cpus[];
713};
714#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
715
716/*
717 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
718 * tail pointer for that CPU's input queue at the time of last enqueue, and
719 * a hardware filter index.
720 */
721struct rps_dev_flow {
722 u16 cpu;
723 u16 filter;
724 unsigned int last_qtail;
725};
726#define RPS_NO_FILTER 0xffff
727
728/*
729 * The rps_dev_flow_table structure contains a table of flow mappings.
730 */
731struct rps_dev_flow_table {
732 unsigned int mask;
733 struct rcu_head rcu;
734 struct rps_dev_flow flows[];
735};
736#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
737 ((_num) * sizeof(struct rps_dev_flow)))
738
739/*
740 * The rps_sock_flow_table contains mappings of flows to the last CPU
741 * on which they were processed by the application (set in recvmsg).
742 * Each entry is a 32bit value. Upper part is the high-order bits
743 * of flow hash, lower part is CPU number.
744 * rps_cpu_mask is used to partition the space, depending on number of
745 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
746 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
747 * meaning we use 32-6=26 bits for the hash.
748 */
749struct rps_sock_flow_table {
750 u32 mask;
751
752 u32 ents[] ____cacheline_aligned_in_smp;
753};
754#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
755
756#define RPS_NO_CPU 0xffff
757
758extern u32 rps_cpu_mask;
759extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
760
761static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
762 u32 hash)
763{
764 if (table && hash) {
765 unsigned int index = hash & table->mask;
766 u32 val = hash & ~rps_cpu_mask;
767
768 /* We only give a hint, preemption can change CPU under us */
769 val |= raw_smp_processor_id();
770
771 /* The following WRITE_ONCE() is paired with the READ_ONCE()
772 * here, and another one in get_rps_cpu().
773 */
774 if (READ_ONCE(table->ents[index]) != val)
775 WRITE_ONCE(table->ents[index], val);
776 }
777}
778
779#ifdef CONFIG_RFS_ACCEL
780bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
781 u16 filter_id);
782#endif
783#endif /* CONFIG_RPS */
784
785/* This structure contains an instance of an RX queue. */
786struct netdev_rx_queue {
787 struct xdp_rxq_info xdp_rxq;
788#ifdef CONFIG_RPS
789 struct rps_map __rcu *rps_map;
790 struct rps_dev_flow_table __rcu *rps_flow_table;
791#endif
792 struct kobject kobj;
793 struct net_device *dev;
794 netdevice_tracker dev_tracker;
795
796#ifdef CONFIG_XDP_SOCKETS
797 struct xsk_buff_pool *pool;
798#endif
799} ____cacheline_aligned_in_smp;
800
801/*
802 * RX queue sysfs structures and functions.
803 */
804struct rx_queue_attribute {
805 struct attribute attr;
806 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
807 ssize_t (*store)(struct netdev_rx_queue *queue,
808 const char *buf, size_t len);
809};
810
811/* XPS map type and offset of the xps map within net_device->xps_maps[]. */
812enum xps_map_type {
813 XPS_CPUS = 0,
814 XPS_RXQS,
815 XPS_MAPS_MAX,
816};
817
818#ifdef CONFIG_XPS
819/*
820 * This structure holds an XPS map which can be of variable length. The
821 * map is an array of queues.
822 */
823struct xps_map {
824 unsigned int len;
825 unsigned int alloc_len;
826 struct rcu_head rcu;
827 u16 queues[];
828};
829#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
830#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
831 - sizeof(struct xps_map)) / sizeof(u16))
832
833/*
834 * This structure holds all XPS maps for device. Maps are indexed by CPU.
835 *
836 * We keep track of the number of cpus/rxqs used when the struct is allocated,
837 * in nr_ids. This will help not accessing out-of-bound memory.
838 *
839 * We keep track of the number of traffic classes used when the struct is
840 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
841 * not crossing its upper bound, as the original dev->num_tc can be updated in
842 * the meantime.
843 */
844struct xps_dev_maps {
845 struct rcu_head rcu;
846 unsigned int nr_ids;
847 s16 num_tc;
848 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
849};
850
851#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
852 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
853
854#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
855 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
856
857#endif /* CONFIG_XPS */
858
859#define TC_MAX_QUEUE 16
860#define TC_BITMASK 15
861/* HW offloaded queuing disciplines txq count and offset maps */
862struct netdev_tc_txq {
863 u16 count;
864 u16 offset;
865};
866
867#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
868/*
869 * This structure is to hold information about the device
870 * configured to run FCoE protocol stack.
871 */
872struct netdev_fcoe_hbainfo {
873 char manufacturer[64];
874 char serial_number[64];
875 char hardware_version[64];
876 char driver_version[64];
877 char optionrom_version[64];
878 char firmware_version[64];
879 char model[256];
880 char model_description[256];
881};
882#endif
883
884#define MAX_PHYS_ITEM_ID_LEN 32
885
886/* This structure holds a unique identifier to identify some
887 * physical item (port for example) used by a netdevice.
888 */
889struct netdev_phys_item_id {
890 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
891 unsigned char id_len;
892};
893
894static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
895 struct netdev_phys_item_id *b)
896{
897 return a->id_len == b->id_len &&
898 memcmp(a->id, b->id, a->id_len) == 0;
899}
900
901typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
902 struct sk_buff *skb,
903 struct net_device *sb_dev);
904
905enum net_device_path_type {
906 DEV_PATH_ETHERNET = 0,
907 DEV_PATH_VLAN,
908 DEV_PATH_BRIDGE,
909 DEV_PATH_PPPOE,
910 DEV_PATH_DSA,
911 DEV_PATH_MTK_WDMA,
912};
913
914struct net_device_path {
915 enum net_device_path_type type;
916 const struct net_device *dev;
917 union {
918 struct {
919 u16 id;
920 __be16 proto;
921 u8 h_dest[ETH_ALEN];
922 } encap;
923 struct {
924 enum {
925 DEV_PATH_BR_VLAN_KEEP,
926 DEV_PATH_BR_VLAN_TAG,
927 DEV_PATH_BR_VLAN_UNTAG,
928 DEV_PATH_BR_VLAN_UNTAG_HW,
929 } vlan_mode;
930 u16 vlan_id;
931 __be16 vlan_proto;
932 } bridge;
933 struct {
934 int port;
935 u16 proto;
936 } dsa;
937 struct {
938 u8 wdma_idx;
939 u8 queue;
940 u16 wcid;
941 u8 bss;
942 } mtk_wdma;
943 };
944};
945
946#define NET_DEVICE_PATH_STACK_MAX 5
947#define NET_DEVICE_PATH_VLAN_MAX 2
948
949struct net_device_path_stack {
950 int num_paths;
951 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
952};
953
954struct net_device_path_ctx {
955 const struct net_device *dev;
956 u8 daddr[ETH_ALEN];
957
958 int num_vlans;
959 struct {
960 u16 id;
961 __be16 proto;
962 } vlan[NET_DEVICE_PATH_VLAN_MAX];
963};
964
965enum tc_setup_type {
966 TC_QUERY_CAPS,
967 TC_SETUP_QDISC_MQPRIO,
968 TC_SETUP_CLSU32,
969 TC_SETUP_CLSFLOWER,
970 TC_SETUP_CLSMATCHALL,
971 TC_SETUP_CLSBPF,
972 TC_SETUP_BLOCK,
973 TC_SETUP_QDISC_CBS,
974 TC_SETUP_QDISC_RED,
975 TC_SETUP_QDISC_PRIO,
976 TC_SETUP_QDISC_MQ,
977 TC_SETUP_QDISC_ETF,
978 TC_SETUP_ROOT_QDISC,
979 TC_SETUP_QDISC_GRED,
980 TC_SETUP_QDISC_TAPRIO,
981 TC_SETUP_FT,
982 TC_SETUP_QDISC_ETS,
983 TC_SETUP_QDISC_TBF,
984 TC_SETUP_QDISC_FIFO,
985 TC_SETUP_QDISC_HTB,
986 TC_SETUP_ACT,
987};
988
989/* These structures hold the attributes of bpf state that are being passed
990 * to the netdevice through the bpf op.
991 */
992enum bpf_netdev_command {
993 /* Set or clear a bpf program used in the earliest stages of packet
994 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
995 * is responsible for calling bpf_prog_put on any old progs that are
996 * stored. In case of error, the callee need not release the new prog
997 * reference, but on success it takes ownership and must bpf_prog_put
998 * when it is no longer used.
999 */
1000 XDP_SETUP_PROG,
1001 XDP_SETUP_PROG_HW,
1002 /* BPF program for offload callbacks, invoked at program load time. */
1003 BPF_OFFLOAD_MAP_ALLOC,
1004 BPF_OFFLOAD_MAP_FREE,
1005 XDP_SETUP_XSK_POOL,
1006};
1007
1008struct bpf_prog_offload_ops;
1009struct netlink_ext_ack;
1010struct xdp_umem;
1011struct xdp_dev_bulk_queue;
1012struct bpf_xdp_link;
1013
1014enum bpf_xdp_mode {
1015 XDP_MODE_SKB = 0,
1016 XDP_MODE_DRV = 1,
1017 XDP_MODE_HW = 2,
1018 __MAX_XDP_MODE
1019};
1020
1021struct bpf_xdp_entity {
1022 struct bpf_prog *prog;
1023 struct bpf_xdp_link *link;
1024};
1025
1026struct netdev_bpf {
1027 enum bpf_netdev_command command;
1028 union {
1029 /* XDP_SETUP_PROG */
1030 struct {
1031 u32 flags;
1032 struct bpf_prog *prog;
1033 struct netlink_ext_ack *extack;
1034 };
1035 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1036 struct {
1037 struct bpf_offloaded_map *offmap;
1038 };
1039 /* XDP_SETUP_XSK_POOL */
1040 struct {
1041 struct xsk_buff_pool *pool;
1042 u16 queue_id;
1043 } xsk;
1044 };
1045};
1046
1047/* Flags for ndo_xsk_wakeup. */
1048#define XDP_WAKEUP_RX (1 << 0)
1049#define XDP_WAKEUP_TX (1 << 1)
1050
1051#ifdef CONFIG_XFRM_OFFLOAD
1052struct xfrmdev_ops {
1053 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1054 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1055 void (*xdo_dev_state_free) (struct xfrm_state *x);
1056 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1057 struct xfrm_state *x);
1058 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1059 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1060 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1061 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1062 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1063};
1064#endif
1065
1066struct dev_ifalias {
1067 struct rcu_head rcuhead;
1068 char ifalias[];
1069};
1070
1071struct devlink;
1072struct tlsdev_ops;
1073
1074struct netdev_net_notifier {
1075 struct list_head list;
1076 struct notifier_block *nb;
1077};
1078
1079/*
1080 * This structure defines the management hooks for network devices.
1081 * The following hooks can be defined; unless noted otherwise, they are
1082 * optional and can be filled with a null pointer.
1083 *
1084 * int (*ndo_init)(struct net_device *dev);
1085 * This function is called once when a network device is registered.
1086 * The network device can use this for any late stage initialization
1087 * or semantic validation. It can fail with an error code which will
1088 * be propagated back to register_netdev.
1089 *
1090 * void (*ndo_uninit)(struct net_device *dev);
1091 * This function is called when device is unregistered or when registration
1092 * fails. It is not called if init fails.
1093 *
1094 * int (*ndo_open)(struct net_device *dev);
1095 * This function is called when a network device transitions to the up
1096 * state.
1097 *
1098 * int (*ndo_stop)(struct net_device *dev);
1099 * This function is called when a network device transitions to the down
1100 * state.
1101 *
1102 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1103 * struct net_device *dev);
1104 * Called when a packet needs to be transmitted.
1105 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1106 * the queue before that can happen; it's for obsolete devices and weird
1107 * corner cases, but the stack really does a non-trivial amount
1108 * of useless work if you return NETDEV_TX_BUSY.
1109 * Required; cannot be NULL.
1110 *
1111 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1112 * struct net_device *dev
1113 * netdev_features_t features);
1114 * Called by core transmit path to determine if device is capable of
1115 * performing offload operations on a given packet. This is to give
1116 * the device an opportunity to implement any restrictions that cannot
1117 * be otherwise expressed by feature flags. The check is called with
1118 * the set of features that the stack has calculated and it returns
1119 * those the driver believes to be appropriate.
1120 *
1121 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1122 * struct net_device *sb_dev);
1123 * Called to decide which queue to use when device supports multiple
1124 * transmit queues.
1125 *
1126 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1127 * This function is called to allow device receiver to make
1128 * changes to configuration when multicast or promiscuous is enabled.
1129 *
1130 * void (*ndo_set_rx_mode)(struct net_device *dev);
1131 * This function is called device changes address list filtering.
1132 * If driver handles unicast address filtering, it should set
1133 * IFF_UNICAST_FLT in its priv_flags.
1134 *
1135 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1136 * This function is called when the Media Access Control address
1137 * needs to be changed. If this interface is not defined, the
1138 * MAC address can not be changed.
1139 *
1140 * int (*ndo_validate_addr)(struct net_device *dev);
1141 * Test if Media Access Control address is valid for the device.
1142 *
1143 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1144 * Old-style ioctl entry point. This is used internally by the
1145 * appletalk and ieee802154 subsystems but is no longer called by
1146 * the device ioctl handler.
1147 *
1148 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1149 * Used by the bonding driver for its device specific ioctls:
1150 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1151 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1152 *
1153 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1154 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1155 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1156 *
1157 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1158 * Used to set network devices bus interface parameters. This interface
1159 * is retained for legacy reasons; new devices should use the bus
1160 * interface (PCI) for low level management.
1161 *
1162 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1163 * Called when a user wants to change the Maximum Transfer Unit
1164 * of a device.
1165 *
1166 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1167 * Callback used when the transmitter has not made any progress
1168 * for dev->watchdog ticks.
1169 *
1170 * void (*ndo_get_stats64)(struct net_device *dev,
1171 * struct rtnl_link_stats64 *storage);
1172 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1173 * Called when a user wants to get the network device usage
1174 * statistics. Drivers must do one of the following:
1175 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1176 * rtnl_link_stats64 structure passed by the caller.
1177 * 2. Define @ndo_get_stats to update a net_device_stats structure
1178 * (which should normally be dev->stats) and return a pointer to
1179 * it. The structure may be changed asynchronously only if each
1180 * field is written atomically.
1181 * 3. Update dev->stats asynchronously and atomically, and define
1182 * neither operation.
1183 *
1184 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1185 * Return true if this device supports offload stats of this attr_id.
1186 *
1187 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1188 * void *attr_data)
1189 * Get statistics for offload operations by attr_id. Write it into the
1190 * attr_data pointer.
1191 *
1192 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1193 * If device supports VLAN filtering this function is called when a
1194 * VLAN id is registered.
1195 *
1196 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1197 * If device supports VLAN filtering this function is called when a
1198 * VLAN id is unregistered.
1199 *
1200 * void (*ndo_poll_controller)(struct net_device *dev);
1201 *
1202 * SR-IOV management functions.
1203 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1204 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1205 * u8 qos, __be16 proto);
1206 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1207 * int max_tx_rate);
1208 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1209 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1210 * int (*ndo_get_vf_config)(struct net_device *dev,
1211 * int vf, struct ifla_vf_info *ivf);
1212 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1213 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1214 * struct nlattr *port[]);
1215 *
1216 * Enable or disable the VF ability to query its RSS Redirection Table and
1217 * Hash Key. This is needed since on some devices VF share this information
1218 * with PF and querying it may introduce a theoretical security risk.
1219 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1220 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1221 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1222 * void *type_data);
1223 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1224 * This is always called from the stack with the rtnl lock held and netif
1225 * tx queues stopped. This allows the netdevice to perform queue
1226 * management safely.
1227 *
1228 * Fiber Channel over Ethernet (FCoE) offload functions.
1229 * int (*ndo_fcoe_enable)(struct net_device *dev);
1230 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1231 * so the underlying device can perform whatever needed configuration or
1232 * initialization to support acceleration of FCoE traffic.
1233 *
1234 * int (*ndo_fcoe_disable)(struct net_device *dev);
1235 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1236 * so the underlying device can perform whatever needed clean-ups to
1237 * stop supporting acceleration of FCoE traffic.
1238 *
1239 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1240 * struct scatterlist *sgl, unsigned int sgc);
1241 * Called when the FCoE Initiator wants to initialize an I/O that
1242 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1243 * perform necessary setup and returns 1 to indicate the device is set up
1244 * successfully to perform DDP on this I/O, otherwise this returns 0.
1245 *
1246 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1247 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1248 * indicated by the FC exchange id 'xid', so the underlying device can
1249 * clean up and reuse resources for later DDP requests.
1250 *
1251 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1252 * struct scatterlist *sgl, unsigned int sgc);
1253 * Called when the FCoE Target wants to initialize an I/O that
1254 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1255 * perform necessary setup and returns 1 to indicate the device is set up
1256 * successfully to perform DDP on this I/O, otherwise this returns 0.
1257 *
1258 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1259 * struct netdev_fcoe_hbainfo *hbainfo);
1260 * Called when the FCoE Protocol stack wants information on the underlying
1261 * device. This information is utilized by the FCoE protocol stack to
1262 * register attributes with Fiber Channel management service as per the
1263 * FC-GS Fabric Device Management Information(FDMI) specification.
1264 *
1265 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1266 * Called when the underlying device wants to override default World Wide
1267 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1268 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1269 * protocol stack to use.
1270 *
1271 * RFS acceleration.
1272 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1273 * u16 rxq_index, u32 flow_id);
1274 * Set hardware filter for RFS. rxq_index is the target queue index;
1275 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1276 * Return the filter ID on success, or a negative error code.
1277 *
1278 * Slave management functions (for bridge, bonding, etc).
1279 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1280 * Called to make another netdev an underling.
1281 *
1282 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1283 * Called to release previously enslaved netdev.
1284 *
1285 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1286 * struct sk_buff *skb,
1287 * bool all_slaves);
1288 * Get the xmit slave of master device. If all_slaves is true, function
1289 * assume all the slaves can transmit.
1290 *
1291 * Feature/offload setting functions.
1292 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1293 * netdev_features_t features);
1294 * Adjusts the requested feature flags according to device-specific
1295 * constraints, and returns the resulting flags. Must not modify
1296 * the device state.
1297 *
1298 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1299 * Called to update device configuration to new features. Passed
1300 * feature set might be less than what was returned by ndo_fix_features()).
1301 * Must return >0 or -errno if it changed dev->features itself.
1302 *
1303 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1304 * struct net_device *dev,
1305 * const unsigned char *addr, u16 vid, u16 flags,
1306 * struct netlink_ext_ack *extack);
1307 * Adds an FDB entry to dev for addr.
1308 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1309 * struct net_device *dev,
1310 * const unsigned char *addr, u16 vid)
1311 * Deletes the FDB entry from dev coresponding to addr.
1312 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1313 * struct net_device *dev,
1314 * u16 vid,
1315 * struct netlink_ext_ack *extack);
1316 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1317 * struct net_device *dev, struct net_device *filter_dev,
1318 * int *idx)
1319 * Used to add FDB entries to dump requests. Implementers should add
1320 * entries to skb and update idx with the number of entries.
1321 *
1322 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1323 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1324 * Adds an MDB entry to dev.
1325 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1326 * struct netlink_ext_ack *extack);
1327 * Deletes the MDB entry from dev.
1328 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1329 * struct netlink_callback *cb);
1330 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1331 * callback is used by core rtnetlink code.
1332 *
1333 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1334 * u16 flags, struct netlink_ext_ack *extack)
1335 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1336 * struct net_device *dev, u32 filter_mask,
1337 * int nlflags)
1338 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1339 * u16 flags);
1340 *
1341 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1342 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1343 * which do not represent real hardware may define this to allow their
1344 * userspace components to manage their virtual carrier state. Devices
1345 * that determine carrier state from physical hardware properties (eg
1346 * network cables) or protocol-dependent mechanisms (eg
1347 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1348 *
1349 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1350 * struct netdev_phys_item_id *ppid);
1351 * Called to get ID of physical port of this device. If driver does
1352 * not implement this, it is assumed that the hw is not able to have
1353 * multiple net devices on single physical port.
1354 *
1355 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1356 * struct netdev_phys_item_id *ppid)
1357 * Called to get the parent ID of the physical port of this device.
1358 *
1359 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1360 * struct net_device *dev)
1361 * Called by upper layer devices to accelerate switching or other
1362 * station functionality into hardware. 'pdev is the lowerdev
1363 * to use for the offload and 'dev' is the net device that will
1364 * back the offload. Returns a pointer to the private structure
1365 * the upper layer will maintain.
1366 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1367 * Called by upper layer device to delete the station created
1368 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1369 * the station and priv is the structure returned by the add
1370 * operation.
1371 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1372 * int queue_index, u32 maxrate);
1373 * Called when a user wants to set a max-rate limitation of specific
1374 * TX queue.
1375 * int (*ndo_get_iflink)(const struct net_device *dev);
1376 * Called to get the iflink value of this device.
1377 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1378 * This function is used to get egress tunnel information for given skb.
1379 * This is useful for retrieving outer tunnel header parameters while
1380 * sampling packet.
1381 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1382 * This function is used to specify the headroom that the skb must
1383 * consider when allocation skb during packet reception. Setting
1384 * appropriate rx headroom value allows avoiding skb head copy on
1385 * forward. Setting a negative value resets the rx headroom to the
1386 * default value.
1387 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1388 * This function is used to set or query state related to XDP on the
1389 * netdevice and manage BPF offload. See definition of
1390 * enum bpf_netdev_command for details.
1391 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1392 * u32 flags);
1393 * This function is used to submit @n XDP packets for transmit on a
1394 * netdevice. Returns number of frames successfully transmitted, frames
1395 * that got dropped are freed/returned via xdp_return_frame().
1396 * Returns negative number, means general error invoking ndo, meaning
1397 * no frames were xmit'ed and core-caller will free all frames.
1398 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1399 * struct xdp_buff *xdp);
1400 * Get the xmit slave of master device based on the xdp_buff.
1401 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1402 * This function is used to wake up the softirq, ksoftirqd or kthread
1403 * responsible for sending and/or receiving packets on a specific
1404 * queue id bound to an AF_XDP socket. The flags field specifies if
1405 * only RX, only Tx, or both should be woken up using the flags
1406 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1407 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1408 * int cmd);
1409 * Add, change, delete or get information on an IPv4 tunnel.
1410 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1411 * If a device is paired with a peer device, return the peer instance.
1412 * The caller must be under RCU read context.
1413 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1414 * Get the forwarding path to reach the real device from the HW destination address
1415 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1416 * const struct skb_shared_hwtstamps *hwtstamps,
1417 * bool cycles);
1418 * Get hardware timestamp based on normal/adjustable time or free running
1419 * cycle counter. This function is required if physical clock supports a
1420 * free running cycle counter.
1421 */
1422struct net_device_ops {
1423 int (*ndo_init)(struct net_device *dev);
1424 void (*ndo_uninit)(struct net_device *dev);
1425 int (*ndo_open)(struct net_device *dev);
1426 int (*ndo_stop)(struct net_device *dev);
1427 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1428 struct net_device *dev);
1429 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1430 struct net_device *dev,
1431 netdev_features_t features);
1432 u16 (*ndo_select_queue)(struct net_device *dev,
1433 struct sk_buff *skb,
1434 struct net_device *sb_dev);
1435 void (*ndo_change_rx_flags)(struct net_device *dev,
1436 int flags);
1437 void (*ndo_set_rx_mode)(struct net_device *dev);
1438 int (*ndo_set_mac_address)(struct net_device *dev,
1439 void *addr);
1440 int (*ndo_validate_addr)(struct net_device *dev);
1441 int (*ndo_do_ioctl)(struct net_device *dev,
1442 struct ifreq *ifr, int cmd);
1443 int (*ndo_eth_ioctl)(struct net_device *dev,
1444 struct ifreq *ifr, int cmd);
1445 int (*ndo_siocbond)(struct net_device *dev,
1446 struct ifreq *ifr, int cmd);
1447 int (*ndo_siocwandev)(struct net_device *dev,
1448 struct if_settings *ifs);
1449 int (*ndo_siocdevprivate)(struct net_device *dev,
1450 struct ifreq *ifr,
1451 void __user *data, int cmd);
1452 int (*ndo_set_config)(struct net_device *dev,
1453 struct ifmap *map);
1454 int (*ndo_change_mtu)(struct net_device *dev,
1455 int new_mtu);
1456 int (*ndo_neigh_setup)(struct net_device *dev,
1457 struct neigh_parms *);
1458 void (*ndo_tx_timeout) (struct net_device *dev,
1459 unsigned int txqueue);
1460
1461 void (*ndo_get_stats64)(struct net_device *dev,
1462 struct rtnl_link_stats64 *storage);
1463 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1464 int (*ndo_get_offload_stats)(int attr_id,
1465 const struct net_device *dev,
1466 void *attr_data);
1467 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1468
1469 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1470 __be16 proto, u16 vid);
1471 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1472 __be16 proto, u16 vid);
1473#ifdef CONFIG_NET_POLL_CONTROLLER
1474 void (*ndo_poll_controller)(struct net_device *dev);
1475 int (*ndo_netpoll_setup)(struct net_device *dev,
1476 struct netpoll_info *info);
1477 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1478#endif
1479 int (*ndo_set_vf_mac)(struct net_device *dev,
1480 int queue, u8 *mac);
1481 int (*ndo_set_vf_vlan)(struct net_device *dev,
1482 int queue, u16 vlan,
1483 u8 qos, __be16 proto);
1484 int (*ndo_set_vf_rate)(struct net_device *dev,
1485 int vf, int min_tx_rate,
1486 int max_tx_rate);
1487 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1488 int vf, bool setting);
1489 int (*ndo_set_vf_trust)(struct net_device *dev,
1490 int vf, bool setting);
1491 int (*ndo_get_vf_config)(struct net_device *dev,
1492 int vf,
1493 struct ifla_vf_info *ivf);
1494 int (*ndo_set_vf_link_state)(struct net_device *dev,
1495 int vf, int link_state);
1496 int (*ndo_get_vf_stats)(struct net_device *dev,
1497 int vf,
1498 struct ifla_vf_stats
1499 *vf_stats);
1500 int (*ndo_set_vf_port)(struct net_device *dev,
1501 int vf,
1502 struct nlattr *port[]);
1503 int (*ndo_get_vf_port)(struct net_device *dev,
1504 int vf, struct sk_buff *skb);
1505 int (*ndo_get_vf_guid)(struct net_device *dev,
1506 int vf,
1507 struct ifla_vf_guid *node_guid,
1508 struct ifla_vf_guid *port_guid);
1509 int (*ndo_set_vf_guid)(struct net_device *dev,
1510 int vf, u64 guid,
1511 int guid_type);
1512 int (*ndo_set_vf_rss_query_en)(
1513 struct net_device *dev,
1514 int vf, bool setting);
1515 int (*ndo_setup_tc)(struct net_device *dev,
1516 enum tc_setup_type type,
1517 void *type_data);
1518#if IS_ENABLED(CONFIG_FCOE)
1519 int (*ndo_fcoe_enable)(struct net_device *dev);
1520 int (*ndo_fcoe_disable)(struct net_device *dev);
1521 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1522 u16 xid,
1523 struct scatterlist *sgl,
1524 unsigned int sgc);
1525 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1526 u16 xid);
1527 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1528 u16 xid,
1529 struct scatterlist *sgl,
1530 unsigned int sgc);
1531 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1532 struct netdev_fcoe_hbainfo *hbainfo);
1533#endif
1534
1535#if IS_ENABLED(CONFIG_LIBFCOE)
1536#define NETDEV_FCOE_WWNN 0
1537#define NETDEV_FCOE_WWPN 1
1538 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1539 u64 *wwn, int type);
1540#endif
1541
1542#ifdef CONFIG_RFS_ACCEL
1543 int (*ndo_rx_flow_steer)(struct net_device *dev,
1544 const struct sk_buff *skb,
1545 u16 rxq_index,
1546 u32 flow_id);
1547#endif
1548 int (*ndo_add_slave)(struct net_device *dev,
1549 struct net_device *slave_dev,
1550 struct netlink_ext_ack *extack);
1551 int (*ndo_del_slave)(struct net_device *dev,
1552 struct net_device *slave_dev);
1553 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1554 struct sk_buff *skb,
1555 bool all_slaves);
1556 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1557 struct sock *sk);
1558 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1559 netdev_features_t features);
1560 int (*ndo_set_features)(struct net_device *dev,
1561 netdev_features_t features);
1562 int (*ndo_neigh_construct)(struct net_device *dev,
1563 struct neighbour *n);
1564 void (*ndo_neigh_destroy)(struct net_device *dev,
1565 struct neighbour *n);
1566
1567 int (*ndo_fdb_add)(struct ndmsg *ndm,
1568 struct nlattr *tb[],
1569 struct net_device *dev,
1570 const unsigned char *addr,
1571 u16 vid,
1572 u16 flags,
1573 struct netlink_ext_ack *extack);
1574 int (*ndo_fdb_del)(struct ndmsg *ndm,
1575 struct nlattr *tb[],
1576 struct net_device *dev,
1577 const unsigned char *addr,
1578 u16 vid, struct netlink_ext_ack *extack);
1579 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1580 struct nlattr *tb[],
1581 struct net_device *dev,
1582 u16 vid,
1583 struct netlink_ext_ack *extack);
1584 int (*ndo_fdb_dump)(struct sk_buff *skb,
1585 struct netlink_callback *cb,
1586 struct net_device *dev,
1587 struct net_device *filter_dev,
1588 int *idx);
1589 int (*ndo_fdb_get)(struct sk_buff *skb,
1590 struct nlattr *tb[],
1591 struct net_device *dev,
1592 const unsigned char *addr,
1593 u16 vid, u32 portid, u32 seq,
1594 struct netlink_ext_ack *extack);
1595 int (*ndo_mdb_add)(struct net_device *dev,
1596 struct nlattr *tb[],
1597 u16 nlmsg_flags,
1598 struct netlink_ext_ack *extack);
1599 int (*ndo_mdb_del)(struct net_device *dev,
1600 struct nlattr *tb[],
1601 struct netlink_ext_ack *extack);
1602 int (*ndo_mdb_dump)(struct net_device *dev,
1603 struct sk_buff *skb,
1604 struct netlink_callback *cb);
1605 int (*ndo_bridge_setlink)(struct net_device *dev,
1606 struct nlmsghdr *nlh,
1607 u16 flags,
1608 struct netlink_ext_ack *extack);
1609 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1610 u32 pid, u32 seq,
1611 struct net_device *dev,
1612 u32 filter_mask,
1613 int nlflags);
1614 int (*ndo_bridge_dellink)(struct net_device *dev,
1615 struct nlmsghdr *nlh,
1616 u16 flags);
1617 int (*ndo_change_carrier)(struct net_device *dev,
1618 bool new_carrier);
1619 int (*ndo_get_phys_port_id)(struct net_device *dev,
1620 struct netdev_phys_item_id *ppid);
1621 int (*ndo_get_port_parent_id)(struct net_device *dev,
1622 struct netdev_phys_item_id *ppid);
1623 int (*ndo_get_phys_port_name)(struct net_device *dev,
1624 char *name, size_t len);
1625 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1626 struct net_device *dev);
1627 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1628 void *priv);
1629
1630 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1631 int queue_index,
1632 u32 maxrate);
1633 int (*ndo_get_iflink)(const struct net_device *dev);
1634 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1635 struct sk_buff *skb);
1636 void (*ndo_set_rx_headroom)(struct net_device *dev,
1637 int needed_headroom);
1638 int (*ndo_bpf)(struct net_device *dev,
1639 struct netdev_bpf *bpf);
1640 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1641 struct xdp_frame **xdp,
1642 u32 flags);
1643 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1644 struct xdp_buff *xdp);
1645 int (*ndo_xsk_wakeup)(struct net_device *dev,
1646 u32 queue_id, u32 flags);
1647 int (*ndo_tunnel_ctl)(struct net_device *dev,
1648 struct ip_tunnel_parm *p, int cmd);
1649 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1650 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1651 struct net_device_path *path);
1652 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1653 const struct skb_shared_hwtstamps *hwtstamps,
1654 bool cycles);
1655};
1656
1657struct xdp_metadata_ops {
1658 int (*xmo_rx_timestamp)(const struct xdp_md *ctx, u64 *timestamp);
1659 int (*xmo_rx_hash)(const struct xdp_md *ctx, u32 *hash,
1660 enum xdp_rss_hash_type *rss_type);
1661};
1662
1663/**
1664 * enum netdev_priv_flags - &struct net_device priv_flags
1665 *
1666 * These are the &struct net_device, they are only set internally
1667 * by drivers and used in the kernel. These flags are invisible to
1668 * userspace; this means that the order of these flags can change
1669 * during any kernel release.
1670 *
1671 * You should have a pretty good reason to be extending these flags.
1672 *
1673 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1674 * @IFF_EBRIDGE: Ethernet bridging device
1675 * @IFF_BONDING: bonding master or slave
1676 * @IFF_ISATAP: ISATAP interface (RFC4214)
1677 * @IFF_WAN_HDLC: WAN HDLC device
1678 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1679 * release skb->dst
1680 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1681 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1682 * @IFF_MACVLAN_PORT: device used as macvlan port
1683 * @IFF_BRIDGE_PORT: device used as bridge port
1684 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1685 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1686 * @IFF_UNICAST_FLT: Supports unicast filtering
1687 * @IFF_TEAM_PORT: device used as team port
1688 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1689 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1690 * change when it's running
1691 * @IFF_MACVLAN: Macvlan device
1692 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1693 * underlying stacked devices
1694 * @IFF_L3MDEV_MASTER: device is an L3 master device
1695 * @IFF_NO_QUEUE: device can run without qdisc attached
1696 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1697 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1698 * @IFF_TEAM: device is a team device
1699 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1700 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1701 * entity (i.e. the master device for bridged veth)
1702 * @IFF_MACSEC: device is a MACsec device
1703 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1704 * @IFF_FAILOVER: device is a failover master device
1705 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1706 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1707 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1708 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1709 * skb_headlen(skb) == 0 (data starts from frag0)
1710 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1711 */
1712enum netdev_priv_flags {
1713 IFF_802_1Q_VLAN = 1<<0,
1714 IFF_EBRIDGE = 1<<1,
1715 IFF_BONDING = 1<<2,
1716 IFF_ISATAP = 1<<3,
1717 IFF_WAN_HDLC = 1<<4,
1718 IFF_XMIT_DST_RELEASE = 1<<5,
1719 IFF_DONT_BRIDGE = 1<<6,
1720 IFF_DISABLE_NETPOLL = 1<<7,
1721 IFF_MACVLAN_PORT = 1<<8,
1722 IFF_BRIDGE_PORT = 1<<9,
1723 IFF_OVS_DATAPATH = 1<<10,
1724 IFF_TX_SKB_SHARING = 1<<11,
1725 IFF_UNICAST_FLT = 1<<12,
1726 IFF_TEAM_PORT = 1<<13,
1727 IFF_SUPP_NOFCS = 1<<14,
1728 IFF_LIVE_ADDR_CHANGE = 1<<15,
1729 IFF_MACVLAN = 1<<16,
1730 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1731 IFF_L3MDEV_MASTER = 1<<18,
1732 IFF_NO_QUEUE = 1<<19,
1733 IFF_OPENVSWITCH = 1<<20,
1734 IFF_L3MDEV_SLAVE = 1<<21,
1735 IFF_TEAM = 1<<22,
1736 IFF_RXFH_CONFIGURED = 1<<23,
1737 IFF_PHONY_HEADROOM = 1<<24,
1738 IFF_MACSEC = 1<<25,
1739 IFF_NO_RX_HANDLER = 1<<26,
1740 IFF_FAILOVER = 1<<27,
1741 IFF_FAILOVER_SLAVE = 1<<28,
1742 IFF_L3MDEV_RX_HANDLER = 1<<29,
1743 IFF_NO_ADDRCONF = BIT_ULL(30),
1744 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1745 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1746};
1747
1748#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1749#define IFF_EBRIDGE IFF_EBRIDGE
1750#define IFF_BONDING IFF_BONDING
1751#define IFF_ISATAP IFF_ISATAP
1752#define IFF_WAN_HDLC IFF_WAN_HDLC
1753#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1754#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1755#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1756#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1757#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1758#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1759#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1760#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1761#define IFF_TEAM_PORT IFF_TEAM_PORT
1762#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1763#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1764#define IFF_MACVLAN IFF_MACVLAN
1765#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1766#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1767#define IFF_NO_QUEUE IFF_NO_QUEUE
1768#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1769#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1770#define IFF_TEAM IFF_TEAM
1771#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1772#define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1773#define IFF_MACSEC IFF_MACSEC
1774#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1775#define IFF_FAILOVER IFF_FAILOVER
1776#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1777#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1778#define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1779
1780/* Specifies the type of the struct net_device::ml_priv pointer */
1781enum netdev_ml_priv_type {
1782 ML_PRIV_NONE,
1783 ML_PRIV_CAN,
1784};
1785
1786/**
1787 * struct net_device - The DEVICE structure.
1788 *
1789 * Actually, this whole structure is a big mistake. It mixes I/O
1790 * data with strictly "high-level" data, and it has to know about
1791 * almost every data structure used in the INET module.
1792 *
1793 * @name: This is the first field of the "visible" part of this structure
1794 * (i.e. as seen by users in the "Space.c" file). It is the name
1795 * of the interface.
1796 *
1797 * @name_node: Name hashlist node
1798 * @ifalias: SNMP alias
1799 * @mem_end: Shared memory end
1800 * @mem_start: Shared memory start
1801 * @base_addr: Device I/O address
1802 * @irq: Device IRQ number
1803 *
1804 * @state: Generic network queuing layer state, see netdev_state_t
1805 * @dev_list: The global list of network devices
1806 * @napi_list: List entry used for polling NAPI devices
1807 * @unreg_list: List entry when we are unregistering the
1808 * device; see the function unregister_netdev
1809 * @close_list: List entry used when we are closing the device
1810 * @ptype_all: Device-specific packet handlers for all protocols
1811 * @ptype_specific: Device-specific, protocol-specific packet handlers
1812 *
1813 * @adj_list: Directly linked devices, like slaves for bonding
1814 * @features: Currently active device features
1815 * @hw_features: User-changeable features
1816 *
1817 * @wanted_features: User-requested features
1818 * @vlan_features: Mask of features inheritable by VLAN devices
1819 *
1820 * @hw_enc_features: Mask of features inherited by encapsulating devices
1821 * This field indicates what encapsulation
1822 * offloads the hardware is capable of doing,
1823 * and drivers will need to set them appropriately.
1824 *
1825 * @mpls_features: Mask of features inheritable by MPLS
1826 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1827 *
1828 * @ifindex: interface index
1829 * @group: The group the device belongs to
1830 *
1831 * @stats: Statistics struct, which was left as a legacy, use
1832 * rtnl_link_stats64 instead
1833 *
1834 * @core_stats: core networking counters,
1835 * do not use this in drivers
1836 * @carrier_up_count: Number of times the carrier has been up
1837 * @carrier_down_count: Number of times the carrier has been down
1838 *
1839 * @wireless_handlers: List of functions to handle Wireless Extensions,
1840 * instead of ioctl,
1841 * see <net/iw_handler.h> for details.
1842 * @wireless_data: Instance data managed by the core of wireless extensions
1843 *
1844 * @netdev_ops: Includes several pointers to callbacks,
1845 * if one wants to override the ndo_*() functions
1846 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1847 * @ethtool_ops: Management operations
1848 * @l3mdev_ops: Layer 3 master device operations
1849 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1850 * discovery handling. Necessary for e.g. 6LoWPAN.
1851 * @xfrmdev_ops: Transformation offload operations
1852 * @tlsdev_ops: Transport Layer Security offload operations
1853 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1854 * of Layer 2 headers.
1855 *
1856 * @flags: Interface flags (a la BSD)
1857 * @xdp_features: XDP capability supported by the device
1858 * @priv_flags: Like 'flags' but invisible to userspace,
1859 * see if.h for the definitions
1860 * @gflags: Global flags ( kept as legacy )
1861 * @padded: How much padding added by alloc_netdev()
1862 * @operstate: RFC2863 operstate
1863 * @link_mode: Mapping policy to operstate
1864 * @if_port: Selectable AUI, TP, ...
1865 * @dma: DMA channel
1866 * @mtu: Interface MTU value
1867 * @min_mtu: Interface Minimum MTU value
1868 * @max_mtu: Interface Maximum MTU value
1869 * @type: Interface hardware type
1870 * @hard_header_len: Maximum hardware header length.
1871 * @min_header_len: Minimum hardware header length
1872 *
1873 * @needed_headroom: Extra headroom the hardware may need, but not in all
1874 * cases can this be guaranteed
1875 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1876 * cases can this be guaranteed. Some cases also use
1877 * LL_MAX_HEADER instead to allocate the skb
1878 *
1879 * interface address info:
1880 *
1881 * @perm_addr: Permanent hw address
1882 * @addr_assign_type: Hw address assignment type
1883 * @addr_len: Hardware address length
1884 * @upper_level: Maximum depth level of upper devices.
1885 * @lower_level: Maximum depth level of lower devices.
1886 * @neigh_priv_len: Used in neigh_alloc()
1887 * @dev_id: Used to differentiate devices that share
1888 * the same link layer address
1889 * @dev_port: Used to differentiate devices that share
1890 * the same function
1891 * @addr_list_lock: XXX: need comments on this one
1892 * @name_assign_type: network interface name assignment type
1893 * @uc_promisc: Counter that indicates promiscuous mode
1894 * has been enabled due to the need to listen to
1895 * additional unicast addresses in a device that
1896 * does not implement ndo_set_rx_mode()
1897 * @uc: unicast mac addresses
1898 * @mc: multicast mac addresses
1899 * @dev_addrs: list of device hw addresses
1900 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1901 * @promiscuity: Number of times the NIC is told to work in
1902 * promiscuous mode; if it becomes 0 the NIC will
1903 * exit promiscuous mode
1904 * @allmulti: Counter, enables or disables allmulticast mode
1905 *
1906 * @vlan_info: VLAN info
1907 * @dsa_ptr: dsa specific data
1908 * @tipc_ptr: TIPC specific data
1909 * @atalk_ptr: AppleTalk link
1910 * @ip_ptr: IPv4 specific data
1911 * @ip6_ptr: IPv6 specific data
1912 * @ax25_ptr: AX.25 specific data
1913 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1914 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1915 * device struct
1916 * @mpls_ptr: mpls_dev struct pointer
1917 * @mctp_ptr: MCTP specific data
1918 *
1919 * @dev_addr: Hw address (before bcast,
1920 * because most packets are unicast)
1921 *
1922 * @_rx: Array of RX queues
1923 * @num_rx_queues: Number of RX queues
1924 * allocated at register_netdev() time
1925 * @real_num_rx_queues: Number of RX queues currently active in device
1926 * @xdp_prog: XDP sockets filter program pointer
1927 * @gro_flush_timeout: timeout for GRO layer in NAPI
1928 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1929 * allow to avoid NIC hard IRQ, on busy queues.
1930 *
1931 * @rx_handler: handler for received packets
1932 * @rx_handler_data: XXX: need comments on this one
1933 * @miniq_ingress: ingress/clsact qdisc specific data for
1934 * ingress processing
1935 * @ingress_queue: XXX: need comments on this one
1936 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1937 * @broadcast: hw bcast address
1938 *
1939 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1940 * indexed by RX queue number. Assigned by driver.
1941 * This must only be set if the ndo_rx_flow_steer
1942 * operation is defined
1943 * @index_hlist: Device index hash chain
1944 *
1945 * @_tx: Array of TX queues
1946 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1947 * @real_num_tx_queues: Number of TX queues currently active in device
1948 * @qdisc: Root qdisc from userspace point of view
1949 * @tx_queue_len: Max frames per queue allowed
1950 * @tx_global_lock: XXX: need comments on this one
1951 * @xdp_bulkq: XDP device bulk queue
1952 * @xps_maps: all CPUs/RXQs maps for XPS device
1953 *
1954 * @xps_maps: XXX: need comments on this one
1955 * @miniq_egress: clsact qdisc specific data for
1956 * egress processing
1957 * @nf_hooks_egress: netfilter hooks executed for egress packets
1958 * @qdisc_hash: qdisc hash table
1959 * @watchdog_timeo: Represents the timeout that is used by
1960 * the watchdog (see dev_watchdog())
1961 * @watchdog_timer: List of timers
1962 *
1963 * @proto_down_reason: reason a netdev interface is held down
1964 * @pcpu_refcnt: Number of references to this device
1965 * @dev_refcnt: Number of references to this device
1966 * @refcnt_tracker: Tracker directory for tracked references to this device
1967 * @todo_list: Delayed register/unregister
1968 * @link_watch_list: XXX: need comments on this one
1969 *
1970 * @reg_state: Register/unregister state machine
1971 * @dismantle: Device is going to be freed
1972 * @rtnl_link_state: This enum represents the phases of creating
1973 * a new link
1974 *
1975 * @needs_free_netdev: Should unregister perform free_netdev?
1976 * @priv_destructor: Called from unregister
1977 * @npinfo: XXX: need comments on this one
1978 * @nd_net: Network namespace this network device is inside
1979 *
1980 * @ml_priv: Mid-layer private
1981 * @ml_priv_type: Mid-layer private type
1982 * @lstats: Loopback statistics
1983 * @tstats: Tunnel statistics
1984 * @dstats: Dummy statistics
1985 * @vstats: Virtual ethernet statistics
1986 *
1987 * @garp_port: GARP
1988 * @mrp_port: MRP
1989 *
1990 * @dm_private: Drop monitor private
1991 *
1992 * @dev: Class/net/name entry
1993 * @sysfs_groups: Space for optional device, statistics and wireless
1994 * sysfs groups
1995 *
1996 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1997 * @rtnl_link_ops: Rtnl_link_ops
1998 *
1999 * @gso_max_size: Maximum size of generic segmentation offload
2000 * @tso_max_size: Device (as in HW) limit on the max TSO request size
2001 * @gso_max_segs: Maximum number of segments that can be passed to the
2002 * NIC for GSO
2003 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
2004 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
2005 * for IPv4.
2006 *
2007 * @dcbnl_ops: Data Center Bridging netlink ops
2008 * @num_tc: Number of traffic classes in the net device
2009 * @tc_to_txq: XXX: need comments on this one
2010 * @prio_tc_map: XXX: need comments on this one
2011 *
2012 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
2013 *
2014 * @priomap: XXX: need comments on this one
2015 * @phydev: Physical device may attach itself
2016 * for hardware timestamping
2017 * @sfp_bus: attached &struct sfp_bus structure.
2018 *
2019 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2020 *
2021 * @proto_down: protocol port state information can be sent to the
2022 * switch driver and used to set the phys state of the
2023 * switch port.
2024 *
2025 * @wol_enabled: Wake-on-LAN is enabled
2026 *
2027 * @threaded: napi threaded mode is enabled
2028 *
2029 * @net_notifier_list: List of per-net netdev notifier block
2030 * that follow this device when it is moved
2031 * to another network namespace.
2032 *
2033 * @macsec_ops: MACsec offloading ops
2034 *
2035 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2036 * offload capabilities of the device
2037 * @udp_tunnel_nic: UDP tunnel offload state
2038 * @xdp_state: stores info on attached XDP BPF programs
2039 *
2040 * @nested_level: Used as a parameter of spin_lock_nested() of
2041 * dev->addr_list_lock.
2042 * @unlink_list: As netif_addr_lock() can be called recursively,
2043 * keep a list of interfaces to be deleted.
2044 * @gro_max_size: Maximum size of aggregated packet in generic
2045 * receive offload (GRO)
2046 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2047 * receive offload (GRO), for IPv4.
2048 *
2049 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2050 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2051 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2052 * @dev_registered_tracker: tracker for reference held while
2053 * registered
2054 * @offload_xstats_l3: L3 HW stats for this netdevice.
2055 *
2056 * @devlink_port: Pointer to related devlink port structure.
2057 * Assigned by a driver before netdev registration using
2058 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2059 * during the time netdevice is registered.
2060 *
2061 * FIXME: cleanup struct net_device such that network protocol info
2062 * moves out.
2063 */
2064
2065struct net_device {
2066 char name[IFNAMSIZ];
2067 struct netdev_name_node *name_node;
2068 struct dev_ifalias __rcu *ifalias;
2069 /*
2070 * I/O specific fields
2071 * FIXME: Merge these and struct ifmap into one
2072 */
2073 unsigned long mem_end;
2074 unsigned long mem_start;
2075 unsigned long base_addr;
2076
2077 /*
2078 * Some hardware also needs these fields (state,dev_list,
2079 * napi_list,unreg_list,close_list) but they are not
2080 * part of the usual set specified in Space.c.
2081 */
2082
2083 unsigned long state;
2084
2085 struct list_head dev_list;
2086 struct list_head napi_list;
2087 struct list_head unreg_list;
2088 struct list_head close_list;
2089 struct list_head ptype_all;
2090 struct list_head ptype_specific;
2091
2092 struct {
2093 struct list_head upper;
2094 struct list_head lower;
2095 } adj_list;
2096
2097 /* Read-mostly cache-line for fast-path access */
2098 unsigned int flags;
2099 xdp_features_t xdp_features;
2100 unsigned long long priv_flags;
2101 const struct net_device_ops *netdev_ops;
2102 const struct xdp_metadata_ops *xdp_metadata_ops;
2103 int ifindex;
2104 unsigned short gflags;
2105 unsigned short hard_header_len;
2106
2107 /* Note : dev->mtu is often read without holding a lock.
2108 * Writers usually hold RTNL.
2109 * It is recommended to use READ_ONCE() to annotate the reads,
2110 * and to use WRITE_ONCE() to annotate the writes.
2111 */
2112 unsigned int mtu;
2113 unsigned short needed_headroom;
2114 unsigned short needed_tailroom;
2115
2116 netdev_features_t features;
2117 netdev_features_t hw_features;
2118 netdev_features_t wanted_features;
2119 netdev_features_t vlan_features;
2120 netdev_features_t hw_enc_features;
2121 netdev_features_t mpls_features;
2122 netdev_features_t gso_partial_features;
2123
2124 unsigned int min_mtu;
2125 unsigned int max_mtu;
2126 unsigned short type;
2127 unsigned char min_header_len;
2128 unsigned char name_assign_type;
2129
2130 int group;
2131
2132 struct net_device_stats stats; /* not used by modern drivers */
2133
2134 struct net_device_core_stats __percpu *core_stats;
2135
2136 /* Stats to monitor link on/off, flapping */
2137 atomic_t carrier_up_count;
2138 atomic_t carrier_down_count;
2139
2140#ifdef CONFIG_WIRELESS_EXT
2141 const struct iw_handler_def *wireless_handlers;
2142 struct iw_public_data *wireless_data;
2143#endif
2144 const struct ethtool_ops *ethtool_ops;
2145#ifdef CONFIG_NET_L3_MASTER_DEV
2146 const struct l3mdev_ops *l3mdev_ops;
2147#endif
2148#if IS_ENABLED(CONFIG_IPV6)
2149 const struct ndisc_ops *ndisc_ops;
2150#endif
2151
2152#ifdef CONFIG_XFRM_OFFLOAD
2153 const struct xfrmdev_ops *xfrmdev_ops;
2154#endif
2155
2156#if IS_ENABLED(CONFIG_TLS_DEVICE)
2157 const struct tlsdev_ops *tlsdev_ops;
2158#endif
2159
2160 const struct header_ops *header_ops;
2161
2162 unsigned char operstate;
2163 unsigned char link_mode;
2164
2165 unsigned char if_port;
2166 unsigned char dma;
2167
2168 /* Interface address info. */
2169 unsigned char perm_addr[MAX_ADDR_LEN];
2170 unsigned char addr_assign_type;
2171 unsigned char addr_len;
2172 unsigned char upper_level;
2173 unsigned char lower_level;
2174
2175 unsigned short neigh_priv_len;
2176 unsigned short dev_id;
2177 unsigned short dev_port;
2178 unsigned short padded;
2179
2180 spinlock_t addr_list_lock;
2181 int irq;
2182
2183 struct netdev_hw_addr_list uc;
2184 struct netdev_hw_addr_list mc;
2185 struct netdev_hw_addr_list dev_addrs;
2186
2187#ifdef CONFIG_SYSFS
2188 struct kset *queues_kset;
2189#endif
2190#ifdef CONFIG_LOCKDEP
2191 struct list_head unlink_list;
2192#endif
2193 unsigned int promiscuity;
2194 unsigned int allmulti;
2195 bool uc_promisc;
2196#ifdef CONFIG_LOCKDEP
2197 unsigned char nested_level;
2198#endif
2199
2200
2201 /* Protocol-specific pointers */
2202
2203 struct in_device __rcu *ip_ptr;
2204 struct inet6_dev __rcu *ip6_ptr;
2205#if IS_ENABLED(CONFIG_VLAN_8021Q)
2206 struct vlan_info __rcu *vlan_info;
2207#endif
2208#if IS_ENABLED(CONFIG_NET_DSA)
2209 struct dsa_port *dsa_ptr;
2210#endif
2211#if IS_ENABLED(CONFIG_TIPC)
2212 struct tipc_bearer __rcu *tipc_ptr;
2213#endif
2214#if IS_ENABLED(CONFIG_ATALK)
2215 void *atalk_ptr;
2216#endif
2217#if IS_ENABLED(CONFIG_AX25)
2218 void *ax25_ptr;
2219#endif
2220#if IS_ENABLED(CONFIG_CFG80211)
2221 struct wireless_dev *ieee80211_ptr;
2222#endif
2223#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2224 struct wpan_dev *ieee802154_ptr;
2225#endif
2226#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2227 struct mpls_dev __rcu *mpls_ptr;
2228#endif
2229#if IS_ENABLED(CONFIG_MCTP)
2230 struct mctp_dev __rcu *mctp_ptr;
2231#endif
2232
2233/*
2234 * Cache lines mostly used on receive path (including eth_type_trans())
2235 */
2236 /* Interface address info used in eth_type_trans() */
2237 const unsigned char *dev_addr;
2238
2239 struct netdev_rx_queue *_rx;
2240 unsigned int num_rx_queues;
2241 unsigned int real_num_rx_queues;
2242
2243 struct bpf_prog __rcu *xdp_prog;
2244 unsigned long gro_flush_timeout;
2245 int napi_defer_hard_irqs;
2246#define GRO_LEGACY_MAX_SIZE 65536u
2247/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2248 * and shinfo->gso_segs is a 16bit field.
2249 */
2250#define GRO_MAX_SIZE (8 * 65535u)
2251 unsigned int gro_max_size;
2252 unsigned int gro_ipv4_max_size;
2253 rx_handler_func_t __rcu *rx_handler;
2254 void __rcu *rx_handler_data;
2255
2256#ifdef CONFIG_NET_CLS_ACT
2257 struct mini_Qdisc __rcu *miniq_ingress;
2258#endif
2259 struct netdev_queue __rcu *ingress_queue;
2260#ifdef CONFIG_NETFILTER_INGRESS
2261 struct nf_hook_entries __rcu *nf_hooks_ingress;
2262#endif
2263
2264 unsigned char broadcast[MAX_ADDR_LEN];
2265#ifdef CONFIG_RFS_ACCEL
2266 struct cpu_rmap *rx_cpu_rmap;
2267#endif
2268 struct hlist_node index_hlist;
2269
2270/*
2271 * Cache lines mostly used on transmit path
2272 */
2273 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2274 unsigned int num_tx_queues;
2275 unsigned int real_num_tx_queues;
2276 struct Qdisc __rcu *qdisc;
2277 unsigned int tx_queue_len;
2278 spinlock_t tx_global_lock;
2279
2280 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2281
2282#ifdef CONFIG_XPS
2283 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2284#endif
2285#ifdef CONFIG_NET_CLS_ACT
2286 struct mini_Qdisc __rcu *miniq_egress;
2287#endif
2288#ifdef CONFIG_NETFILTER_EGRESS
2289 struct nf_hook_entries __rcu *nf_hooks_egress;
2290#endif
2291
2292#ifdef CONFIG_NET_SCHED
2293 DECLARE_HASHTABLE (qdisc_hash, 4);
2294#endif
2295 /* These may be needed for future network-power-down code. */
2296 struct timer_list watchdog_timer;
2297 int watchdog_timeo;
2298
2299 u32 proto_down_reason;
2300
2301 struct list_head todo_list;
2302
2303#ifdef CONFIG_PCPU_DEV_REFCNT
2304 int __percpu *pcpu_refcnt;
2305#else
2306 refcount_t dev_refcnt;
2307#endif
2308 struct ref_tracker_dir refcnt_tracker;
2309
2310 struct list_head link_watch_list;
2311
2312 enum { NETREG_UNINITIALIZED=0,
2313 NETREG_REGISTERED, /* completed register_netdevice */
2314 NETREG_UNREGISTERING, /* called unregister_netdevice */
2315 NETREG_UNREGISTERED, /* completed unregister todo */
2316 NETREG_RELEASED, /* called free_netdev */
2317 NETREG_DUMMY, /* dummy device for NAPI poll */
2318 } reg_state:8;
2319
2320 bool dismantle;
2321
2322 enum {
2323 RTNL_LINK_INITIALIZED,
2324 RTNL_LINK_INITIALIZING,
2325 } rtnl_link_state:16;
2326
2327 bool needs_free_netdev;
2328 void (*priv_destructor)(struct net_device *dev);
2329
2330#ifdef CONFIG_NETPOLL
2331 struct netpoll_info __rcu *npinfo;
2332#endif
2333
2334 possible_net_t nd_net;
2335
2336 /* mid-layer private */
2337 void *ml_priv;
2338 enum netdev_ml_priv_type ml_priv_type;
2339
2340 union {
2341 struct pcpu_lstats __percpu *lstats;
2342 struct pcpu_sw_netstats __percpu *tstats;
2343 struct pcpu_dstats __percpu *dstats;
2344 };
2345
2346#if IS_ENABLED(CONFIG_GARP)
2347 struct garp_port __rcu *garp_port;
2348#endif
2349#if IS_ENABLED(CONFIG_MRP)
2350 struct mrp_port __rcu *mrp_port;
2351#endif
2352#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2353 struct dm_hw_stat_delta __rcu *dm_private;
2354#endif
2355 struct device dev;
2356 const struct attribute_group *sysfs_groups[4];
2357 const struct attribute_group *sysfs_rx_queue_group;
2358
2359 const struct rtnl_link_ops *rtnl_link_ops;
2360
2361 /* for setting kernel sock attribute on TCP connection setup */
2362#define GSO_MAX_SEGS 65535u
2363#define GSO_LEGACY_MAX_SIZE 65536u
2364/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2365 * and shinfo->gso_segs is a 16bit field.
2366 */
2367#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2368
2369 unsigned int gso_max_size;
2370#define TSO_LEGACY_MAX_SIZE 65536
2371#define TSO_MAX_SIZE UINT_MAX
2372 unsigned int tso_max_size;
2373 u16 gso_max_segs;
2374#define TSO_MAX_SEGS U16_MAX
2375 u16 tso_max_segs;
2376 unsigned int gso_ipv4_max_size;
2377
2378#ifdef CONFIG_DCB
2379 const struct dcbnl_rtnl_ops *dcbnl_ops;
2380#endif
2381 s16 num_tc;
2382 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2383 u8 prio_tc_map[TC_BITMASK + 1];
2384
2385#if IS_ENABLED(CONFIG_FCOE)
2386 unsigned int fcoe_ddp_xid;
2387#endif
2388#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2389 struct netprio_map __rcu *priomap;
2390#endif
2391 struct phy_device *phydev;
2392 struct sfp_bus *sfp_bus;
2393 struct lock_class_key *qdisc_tx_busylock;
2394 bool proto_down;
2395 unsigned wol_enabled:1;
2396 unsigned threaded:1;
2397
2398 struct list_head net_notifier_list;
2399
2400#if IS_ENABLED(CONFIG_MACSEC)
2401 /* MACsec management functions */
2402 const struct macsec_ops *macsec_ops;
2403#endif
2404 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2405 struct udp_tunnel_nic *udp_tunnel_nic;
2406
2407 /* protected by rtnl_lock */
2408 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2409
2410 u8 dev_addr_shadow[MAX_ADDR_LEN];
2411 netdevice_tracker linkwatch_dev_tracker;
2412 netdevice_tracker watchdog_dev_tracker;
2413 netdevice_tracker dev_registered_tracker;
2414 struct rtnl_hw_stats64 *offload_xstats_l3;
2415
2416 struct devlink_port *devlink_port;
2417};
2418#define to_net_dev(d) container_of(d, struct net_device, dev)
2419
2420/*
2421 * Driver should use this to assign devlink port instance to a netdevice
2422 * before it registers the netdevice. Therefore devlink_port is static
2423 * during the netdev lifetime after it is registered.
2424 */
2425#define SET_NETDEV_DEVLINK_PORT(dev, port) \
2426({ \
2427 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2428 ((dev)->devlink_port = (port)); \
2429})
2430
2431static inline bool netif_elide_gro(const struct net_device *dev)
2432{
2433 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2434 return true;
2435 return false;
2436}
2437
2438#define NETDEV_ALIGN 32
2439
2440static inline
2441int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2442{
2443 return dev->prio_tc_map[prio & TC_BITMASK];
2444}
2445
2446static inline
2447int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2448{
2449 if (tc >= dev->num_tc)
2450 return -EINVAL;
2451
2452 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2453 return 0;
2454}
2455
2456int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2457void netdev_reset_tc(struct net_device *dev);
2458int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2459int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2460
2461static inline
2462int netdev_get_num_tc(struct net_device *dev)
2463{
2464 return dev->num_tc;
2465}
2466
2467static inline void net_prefetch(void *p)
2468{
2469 prefetch(p);
2470#if L1_CACHE_BYTES < 128
2471 prefetch((u8 *)p + L1_CACHE_BYTES);
2472#endif
2473}
2474
2475static inline void net_prefetchw(void *p)
2476{
2477 prefetchw(p);
2478#if L1_CACHE_BYTES < 128
2479 prefetchw((u8 *)p + L1_CACHE_BYTES);
2480#endif
2481}
2482
2483void netdev_unbind_sb_channel(struct net_device *dev,
2484 struct net_device *sb_dev);
2485int netdev_bind_sb_channel_queue(struct net_device *dev,
2486 struct net_device *sb_dev,
2487 u8 tc, u16 count, u16 offset);
2488int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2489static inline int netdev_get_sb_channel(struct net_device *dev)
2490{
2491 return max_t(int, -dev->num_tc, 0);
2492}
2493
2494static inline
2495struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2496 unsigned int index)
2497{
2498 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2499 return &dev->_tx[index];
2500}
2501
2502static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2503 const struct sk_buff *skb)
2504{
2505 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2506}
2507
2508static inline void netdev_for_each_tx_queue(struct net_device *dev,
2509 void (*f)(struct net_device *,
2510 struct netdev_queue *,
2511 void *),
2512 void *arg)
2513{
2514 unsigned int i;
2515
2516 for (i = 0; i < dev->num_tx_queues; i++)
2517 f(dev, &dev->_tx[i], arg);
2518}
2519
2520#define netdev_lockdep_set_classes(dev) \
2521{ \
2522 static struct lock_class_key qdisc_tx_busylock_key; \
2523 static struct lock_class_key qdisc_xmit_lock_key; \
2524 static struct lock_class_key dev_addr_list_lock_key; \
2525 unsigned int i; \
2526 \
2527 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2528 lockdep_set_class(&(dev)->addr_list_lock, \
2529 &dev_addr_list_lock_key); \
2530 for (i = 0; i < (dev)->num_tx_queues; i++) \
2531 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2532 &qdisc_xmit_lock_key); \
2533}
2534
2535u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2536 struct net_device *sb_dev);
2537struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2538 struct sk_buff *skb,
2539 struct net_device *sb_dev);
2540
2541/* returns the headroom that the master device needs to take in account
2542 * when forwarding to this dev
2543 */
2544static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2545{
2546 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2547}
2548
2549static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2550{
2551 if (dev->netdev_ops->ndo_set_rx_headroom)
2552 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2553}
2554
2555/* set the device rx headroom to the dev's default */
2556static inline void netdev_reset_rx_headroom(struct net_device *dev)
2557{
2558 netdev_set_rx_headroom(dev, -1);
2559}
2560
2561static inline void *netdev_get_ml_priv(struct net_device *dev,
2562 enum netdev_ml_priv_type type)
2563{
2564 if (dev->ml_priv_type != type)
2565 return NULL;
2566
2567 return dev->ml_priv;
2568}
2569
2570static inline void netdev_set_ml_priv(struct net_device *dev,
2571 void *ml_priv,
2572 enum netdev_ml_priv_type type)
2573{
2574 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2575 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2576 dev->ml_priv_type, type);
2577 WARN(!dev->ml_priv_type && dev->ml_priv,
2578 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2579
2580 dev->ml_priv = ml_priv;
2581 dev->ml_priv_type = type;
2582}
2583
2584/*
2585 * Net namespace inlines
2586 */
2587static inline
2588struct net *dev_net(const struct net_device *dev)
2589{
2590 return read_pnet(&dev->nd_net);
2591}
2592
2593static inline
2594void dev_net_set(struct net_device *dev, struct net *net)
2595{
2596 write_pnet(&dev->nd_net, net);
2597}
2598
2599/**
2600 * netdev_priv - access network device private data
2601 * @dev: network device
2602 *
2603 * Get network device private data
2604 */
2605static inline void *netdev_priv(const struct net_device *dev)
2606{
2607 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2608}
2609
2610/* Set the sysfs physical device reference for the network logical device
2611 * if set prior to registration will cause a symlink during initialization.
2612 */
2613#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2614
2615/* Set the sysfs device type for the network logical device to allow
2616 * fine-grained identification of different network device types. For
2617 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2618 */
2619#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2620
2621/* Default NAPI poll() weight
2622 * Device drivers are strongly advised to not use bigger value
2623 */
2624#define NAPI_POLL_WEIGHT 64
2625
2626void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2627 int (*poll)(struct napi_struct *, int), int weight);
2628
2629/**
2630 * netif_napi_add() - initialize a NAPI context
2631 * @dev: network device
2632 * @napi: NAPI context
2633 * @poll: polling function
2634 *
2635 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2636 * *any* of the other NAPI-related functions.
2637 */
2638static inline void
2639netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2640 int (*poll)(struct napi_struct *, int))
2641{
2642 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2643}
2644
2645static inline void
2646netif_napi_add_tx_weight(struct net_device *dev,
2647 struct napi_struct *napi,
2648 int (*poll)(struct napi_struct *, int),
2649 int weight)
2650{
2651 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2652 netif_napi_add_weight(dev, napi, poll, weight);
2653}
2654
2655/**
2656 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2657 * @dev: network device
2658 * @napi: NAPI context
2659 * @poll: polling function
2660 *
2661 * This variant of netif_napi_add() should be used from drivers using NAPI
2662 * to exclusively poll a TX queue.
2663 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2664 */
2665static inline void netif_napi_add_tx(struct net_device *dev,
2666 struct napi_struct *napi,
2667 int (*poll)(struct napi_struct *, int))
2668{
2669 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2670}
2671
2672/**
2673 * __netif_napi_del - remove a NAPI context
2674 * @napi: NAPI context
2675 *
2676 * Warning: caller must observe RCU grace period before freeing memory
2677 * containing @napi. Drivers might want to call this helper to combine
2678 * all the needed RCU grace periods into a single one.
2679 */
2680void __netif_napi_del(struct napi_struct *napi);
2681
2682/**
2683 * netif_napi_del - remove a NAPI context
2684 * @napi: NAPI context
2685 *
2686 * netif_napi_del() removes a NAPI context from the network device NAPI list
2687 */
2688static inline void netif_napi_del(struct napi_struct *napi)
2689{
2690 __netif_napi_del(napi);
2691 synchronize_net();
2692}
2693
2694struct packet_type {
2695 __be16 type; /* This is really htons(ether_type). */
2696 bool ignore_outgoing;
2697 struct net_device *dev; /* NULL is wildcarded here */
2698 netdevice_tracker dev_tracker;
2699 int (*func) (struct sk_buff *,
2700 struct net_device *,
2701 struct packet_type *,
2702 struct net_device *);
2703 void (*list_func) (struct list_head *,
2704 struct packet_type *,
2705 struct net_device *);
2706 bool (*id_match)(struct packet_type *ptype,
2707 struct sock *sk);
2708 struct net *af_packet_net;
2709 void *af_packet_priv;
2710 struct list_head list;
2711};
2712
2713struct offload_callbacks {
2714 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2715 netdev_features_t features);
2716 struct sk_buff *(*gro_receive)(struct list_head *head,
2717 struct sk_buff *skb);
2718 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2719};
2720
2721struct packet_offload {
2722 __be16 type; /* This is really htons(ether_type). */
2723 u16 priority;
2724 struct offload_callbacks callbacks;
2725 struct list_head list;
2726};
2727
2728/* often modified stats are per-CPU, other are shared (netdev->stats) */
2729struct pcpu_sw_netstats {
2730 u64_stats_t rx_packets;
2731 u64_stats_t rx_bytes;
2732 u64_stats_t tx_packets;
2733 u64_stats_t tx_bytes;
2734 struct u64_stats_sync syncp;
2735} __aligned(4 * sizeof(u64));
2736
2737struct pcpu_lstats {
2738 u64_stats_t packets;
2739 u64_stats_t bytes;
2740 struct u64_stats_sync syncp;
2741} __aligned(2 * sizeof(u64));
2742
2743void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2744
2745static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2746{
2747 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2748
2749 u64_stats_update_begin(&tstats->syncp);
2750 u64_stats_add(&tstats->rx_bytes, len);
2751 u64_stats_inc(&tstats->rx_packets);
2752 u64_stats_update_end(&tstats->syncp);
2753}
2754
2755static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2756 unsigned int packets,
2757 unsigned int len)
2758{
2759 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2760
2761 u64_stats_update_begin(&tstats->syncp);
2762 u64_stats_add(&tstats->tx_bytes, len);
2763 u64_stats_add(&tstats->tx_packets, packets);
2764 u64_stats_update_end(&tstats->syncp);
2765}
2766
2767static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2768{
2769 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2770
2771 u64_stats_update_begin(&lstats->syncp);
2772 u64_stats_add(&lstats->bytes, len);
2773 u64_stats_inc(&lstats->packets);
2774 u64_stats_update_end(&lstats->syncp);
2775}
2776
2777#define __netdev_alloc_pcpu_stats(type, gfp) \
2778({ \
2779 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2780 if (pcpu_stats) { \
2781 int __cpu; \
2782 for_each_possible_cpu(__cpu) { \
2783 typeof(type) *stat; \
2784 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2785 u64_stats_init(&stat->syncp); \
2786 } \
2787 } \
2788 pcpu_stats; \
2789})
2790
2791#define netdev_alloc_pcpu_stats(type) \
2792 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2793
2794#define devm_netdev_alloc_pcpu_stats(dev, type) \
2795({ \
2796 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2797 if (pcpu_stats) { \
2798 int __cpu; \
2799 for_each_possible_cpu(__cpu) { \
2800 typeof(type) *stat; \
2801 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2802 u64_stats_init(&stat->syncp); \
2803 } \
2804 } \
2805 pcpu_stats; \
2806})
2807
2808enum netdev_lag_tx_type {
2809 NETDEV_LAG_TX_TYPE_UNKNOWN,
2810 NETDEV_LAG_TX_TYPE_RANDOM,
2811 NETDEV_LAG_TX_TYPE_BROADCAST,
2812 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2813 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2814 NETDEV_LAG_TX_TYPE_HASH,
2815};
2816
2817enum netdev_lag_hash {
2818 NETDEV_LAG_HASH_NONE,
2819 NETDEV_LAG_HASH_L2,
2820 NETDEV_LAG_HASH_L34,
2821 NETDEV_LAG_HASH_L23,
2822 NETDEV_LAG_HASH_E23,
2823 NETDEV_LAG_HASH_E34,
2824 NETDEV_LAG_HASH_VLAN_SRCMAC,
2825 NETDEV_LAG_HASH_UNKNOWN,
2826};
2827
2828struct netdev_lag_upper_info {
2829 enum netdev_lag_tx_type tx_type;
2830 enum netdev_lag_hash hash_type;
2831};
2832
2833struct netdev_lag_lower_state_info {
2834 u8 link_up : 1,
2835 tx_enabled : 1;
2836};
2837
2838#include <linux/notifier.h>
2839
2840/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2841 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2842 * adding new types.
2843 */
2844enum netdev_cmd {
2845 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2846 NETDEV_DOWN,
2847 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2848 detected a hardware crash and restarted
2849 - we can use this eg to kick tcp sessions
2850 once done */
2851 NETDEV_CHANGE, /* Notify device state change */
2852 NETDEV_REGISTER,
2853 NETDEV_UNREGISTER,
2854 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2855 NETDEV_CHANGEADDR, /* notify after the address change */
2856 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2857 NETDEV_GOING_DOWN,
2858 NETDEV_CHANGENAME,
2859 NETDEV_FEAT_CHANGE,
2860 NETDEV_BONDING_FAILOVER,
2861 NETDEV_PRE_UP,
2862 NETDEV_PRE_TYPE_CHANGE,
2863 NETDEV_POST_TYPE_CHANGE,
2864 NETDEV_POST_INIT,
2865 NETDEV_PRE_UNINIT,
2866 NETDEV_RELEASE,
2867 NETDEV_NOTIFY_PEERS,
2868 NETDEV_JOIN,
2869 NETDEV_CHANGEUPPER,
2870 NETDEV_RESEND_IGMP,
2871 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2872 NETDEV_CHANGEINFODATA,
2873 NETDEV_BONDING_INFO,
2874 NETDEV_PRECHANGEUPPER,
2875 NETDEV_CHANGELOWERSTATE,
2876 NETDEV_UDP_TUNNEL_PUSH_INFO,
2877 NETDEV_UDP_TUNNEL_DROP_INFO,
2878 NETDEV_CHANGE_TX_QUEUE_LEN,
2879 NETDEV_CVLAN_FILTER_PUSH_INFO,
2880 NETDEV_CVLAN_FILTER_DROP_INFO,
2881 NETDEV_SVLAN_FILTER_PUSH_INFO,
2882 NETDEV_SVLAN_FILTER_DROP_INFO,
2883 NETDEV_OFFLOAD_XSTATS_ENABLE,
2884 NETDEV_OFFLOAD_XSTATS_DISABLE,
2885 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2886 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2887 NETDEV_XDP_FEAT_CHANGE,
2888};
2889const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2890
2891int register_netdevice_notifier(struct notifier_block *nb);
2892int unregister_netdevice_notifier(struct notifier_block *nb);
2893int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2894int unregister_netdevice_notifier_net(struct net *net,
2895 struct notifier_block *nb);
2896int register_netdevice_notifier_dev_net(struct net_device *dev,
2897 struct notifier_block *nb,
2898 struct netdev_net_notifier *nn);
2899int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2900 struct notifier_block *nb,
2901 struct netdev_net_notifier *nn);
2902
2903struct netdev_notifier_info {
2904 struct net_device *dev;
2905 struct netlink_ext_ack *extack;
2906};
2907
2908struct netdev_notifier_info_ext {
2909 struct netdev_notifier_info info; /* must be first */
2910 union {
2911 u32 mtu;
2912 } ext;
2913};
2914
2915struct netdev_notifier_change_info {
2916 struct netdev_notifier_info info; /* must be first */
2917 unsigned int flags_changed;
2918};
2919
2920struct netdev_notifier_changeupper_info {
2921 struct netdev_notifier_info info; /* must be first */
2922 struct net_device *upper_dev; /* new upper dev */
2923 bool master; /* is upper dev master */
2924 bool linking; /* is the notification for link or unlink */
2925 void *upper_info; /* upper dev info */
2926};
2927
2928struct netdev_notifier_changelowerstate_info {
2929 struct netdev_notifier_info info; /* must be first */
2930 void *lower_state_info; /* is lower dev state */
2931};
2932
2933struct netdev_notifier_pre_changeaddr_info {
2934 struct netdev_notifier_info info; /* must be first */
2935 const unsigned char *dev_addr;
2936};
2937
2938enum netdev_offload_xstats_type {
2939 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2940};
2941
2942struct netdev_notifier_offload_xstats_info {
2943 struct netdev_notifier_info info; /* must be first */
2944 enum netdev_offload_xstats_type type;
2945
2946 union {
2947 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2948 struct netdev_notifier_offload_xstats_rd *report_delta;
2949 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2950 struct netdev_notifier_offload_xstats_ru *report_used;
2951 };
2952};
2953
2954int netdev_offload_xstats_enable(struct net_device *dev,
2955 enum netdev_offload_xstats_type type,
2956 struct netlink_ext_ack *extack);
2957int netdev_offload_xstats_disable(struct net_device *dev,
2958 enum netdev_offload_xstats_type type);
2959bool netdev_offload_xstats_enabled(const struct net_device *dev,
2960 enum netdev_offload_xstats_type type);
2961int netdev_offload_xstats_get(struct net_device *dev,
2962 enum netdev_offload_xstats_type type,
2963 struct rtnl_hw_stats64 *stats, bool *used,
2964 struct netlink_ext_ack *extack);
2965void
2966netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2967 const struct rtnl_hw_stats64 *stats);
2968void
2969netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2970void netdev_offload_xstats_push_delta(struct net_device *dev,
2971 enum netdev_offload_xstats_type type,
2972 const struct rtnl_hw_stats64 *stats);
2973
2974static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2975 struct net_device *dev)
2976{
2977 info->dev = dev;
2978 info->extack = NULL;
2979}
2980
2981static inline struct net_device *
2982netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2983{
2984 return info->dev;
2985}
2986
2987static inline struct netlink_ext_ack *
2988netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2989{
2990 return info->extack;
2991}
2992
2993int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2994int call_netdevice_notifiers_info(unsigned long val,
2995 struct netdev_notifier_info *info);
2996
2997extern rwlock_t dev_base_lock; /* Device list lock */
2998
2999#define for_each_netdev(net, d) \
3000 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3001#define for_each_netdev_reverse(net, d) \
3002 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3003#define for_each_netdev_rcu(net, d) \
3004 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3005#define for_each_netdev_safe(net, d, n) \
3006 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3007#define for_each_netdev_continue(net, d) \
3008 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3009#define for_each_netdev_continue_reverse(net, d) \
3010 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3011 dev_list)
3012#define for_each_netdev_continue_rcu(net, d) \
3013 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3014#define for_each_netdev_in_bond_rcu(bond, slave) \
3015 for_each_netdev_rcu(&init_net, slave) \
3016 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3017#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3018
3019static inline struct net_device *next_net_device(struct net_device *dev)
3020{
3021 struct list_head *lh;
3022 struct net *net;
3023
3024 net = dev_net(dev);
3025 lh = dev->dev_list.next;
3026 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3027}
3028
3029static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3030{
3031 struct list_head *lh;
3032 struct net *net;
3033
3034 net = dev_net(dev);
3035 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3036 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3037}
3038
3039static inline struct net_device *first_net_device(struct net *net)
3040{
3041 return list_empty(&net->dev_base_head) ? NULL :
3042 net_device_entry(net->dev_base_head.next);
3043}
3044
3045static inline struct net_device *first_net_device_rcu(struct net *net)
3046{
3047 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3048
3049 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3050}
3051
3052int netdev_boot_setup_check(struct net_device *dev);
3053struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3054 const char *hwaddr);
3055struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3056void dev_add_pack(struct packet_type *pt);
3057void dev_remove_pack(struct packet_type *pt);
3058void __dev_remove_pack(struct packet_type *pt);
3059void dev_add_offload(struct packet_offload *po);
3060void dev_remove_offload(struct packet_offload *po);
3061
3062int dev_get_iflink(const struct net_device *dev);
3063int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3064int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3065 struct net_device_path_stack *stack);
3066struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3067 unsigned short mask);
3068struct net_device *dev_get_by_name(struct net *net, const char *name);
3069struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3070struct net_device *__dev_get_by_name(struct net *net, const char *name);
3071bool netdev_name_in_use(struct net *net, const char *name);
3072int dev_alloc_name(struct net_device *dev, const char *name);
3073int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3074void dev_close(struct net_device *dev);
3075void dev_close_many(struct list_head *head, bool unlink);
3076void dev_disable_lro(struct net_device *dev);
3077int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3078u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3079 struct net_device *sb_dev);
3080u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3081 struct net_device *sb_dev);
3082
3083int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3084int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3085
3086static inline int dev_queue_xmit(struct sk_buff *skb)
3087{
3088 return __dev_queue_xmit(skb, NULL);
3089}
3090
3091static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3092 struct net_device *sb_dev)
3093{
3094 return __dev_queue_xmit(skb, sb_dev);
3095}
3096
3097static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3098{
3099 int ret;
3100
3101 ret = __dev_direct_xmit(skb, queue_id);
3102 if (!dev_xmit_complete(ret))
3103 kfree_skb(skb);
3104 return ret;
3105}
3106
3107int register_netdevice(struct net_device *dev);
3108void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3109void unregister_netdevice_many(struct list_head *head);
3110static inline void unregister_netdevice(struct net_device *dev)
3111{
3112 unregister_netdevice_queue(dev, NULL);
3113}
3114
3115int netdev_refcnt_read(const struct net_device *dev);
3116void free_netdev(struct net_device *dev);
3117void netdev_freemem(struct net_device *dev);
3118int init_dummy_netdev(struct net_device *dev);
3119
3120struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3121 struct sk_buff *skb,
3122 bool all_slaves);
3123struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3124 struct sock *sk);
3125struct net_device *dev_get_by_index(struct net *net, int ifindex);
3126struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3127struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3128 netdevice_tracker *tracker, gfp_t gfp);
3129struct net_device *netdev_get_by_name(struct net *net, const char *name,
3130 netdevice_tracker *tracker, gfp_t gfp);
3131struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3132struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3133int dev_restart(struct net_device *dev);
3134
3135
3136static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3137 unsigned short type,
3138 const void *daddr, const void *saddr,
3139 unsigned int len)
3140{
3141 if (!dev->header_ops || !dev->header_ops->create)
3142 return 0;
3143
3144 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3145}
3146
3147static inline int dev_parse_header(const struct sk_buff *skb,
3148 unsigned char *haddr)
3149{
3150 const struct net_device *dev = skb->dev;
3151
3152 if (!dev->header_ops || !dev->header_ops->parse)
3153 return 0;
3154 return dev->header_ops->parse(skb, haddr);
3155}
3156
3157static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3158{
3159 const struct net_device *dev = skb->dev;
3160
3161 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3162 return 0;
3163 return dev->header_ops->parse_protocol(skb);
3164}
3165
3166/* ll_header must have at least hard_header_len allocated */
3167static inline bool dev_validate_header(const struct net_device *dev,
3168 char *ll_header, int len)
3169{
3170 if (likely(len >= dev->hard_header_len))
3171 return true;
3172 if (len < dev->min_header_len)
3173 return false;
3174
3175 if (capable(CAP_SYS_RAWIO)) {
3176 memset(ll_header + len, 0, dev->hard_header_len - len);
3177 return true;
3178 }
3179
3180 if (dev->header_ops && dev->header_ops->validate)
3181 return dev->header_ops->validate(ll_header, len);
3182
3183 return false;
3184}
3185
3186static inline bool dev_has_header(const struct net_device *dev)
3187{
3188 return dev->header_ops && dev->header_ops->create;
3189}
3190
3191/*
3192 * Incoming packets are placed on per-CPU queues
3193 */
3194struct softnet_data {
3195 struct list_head poll_list;
3196 struct sk_buff_head process_queue;
3197
3198 /* stats */
3199 unsigned int processed;
3200 unsigned int time_squeeze;
3201#ifdef CONFIG_RPS
3202 struct softnet_data *rps_ipi_list;
3203#endif
3204
3205 bool in_net_rx_action;
3206 bool in_napi_threaded_poll;
3207
3208#ifdef CONFIG_NET_FLOW_LIMIT
3209 struct sd_flow_limit __rcu *flow_limit;
3210#endif
3211 struct Qdisc *output_queue;
3212 struct Qdisc **output_queue_tailp;
3213 struct sk_buff *completion_queue;
3214#ifdef CONFIG_XFRM_OFFLOAD
3215 struct sk_buff_head xfrm_backlog;
3216#endif
3217 /* written and read only by owning cpu: */
3218 struct {
3219 u16 recursion;
3220 u8 more;
3221#ifdef CONFIG_NET_EGRESS
3222 u8 skip_txqueue;
3223#endif
3224 } xmit;
3225#ifdef CONFIG_RPS
3226 /* input_queue_head should be written by cpu owning this struct,
3227 * and only read by other cpus. Worth using a cache line.
3228 */
3229 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3230
3231 /* Elements below can be accessed between CPUs for RPS/RFS */
3232 call_single_data_t csd ____cacheline_aligned_in_smp;
3233 struct softnet_data *rps_ipi_next;
3234 unsigned int cpu;
3235 unsigned int input_queue_tail;
3236#endif
3237 unsigned int received_rps;
3238 unsigned int dropped;
3239 struct sk_buff_head input_pkt_queue;
3240 struct napi_struct backlog;
3241
3242 /* Another possibly contended cache line */
3243 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3244 int defer_count;
3245 int defer_ipi_scheduled;
3246 struct sk_buff *defer_list;
3247 call_single_data_t defer_csd;
3248};
3249
3250static inline void input_queue_head_incr(struct softnet_data *sd)
3251{
3252#ifdef CONFIG_RPS
3253 sd->input_queue_head++;
3254#endif
3255}
3256
3257static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3258 unsigned int *qtail)
3259{
3260#ifdef CONFIG_RPS
3261 *qtail = ++sd->input_queue_tail;
3262#endif
3263}
3264
3265DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3266
3267static inline int dev_recursion_level(void)
3268{
3269 return this_cpu_read(softnet_data.xmit.recursion);
3270}
3271
3272#define XMIT_RECURSION_LIMIT 8
3273static inline bool dev_xmit_recursion(void)
3274{
3275 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3276 XMIT_RECURSION_LIMIT);
3277}
3278
3279static inline void dev_xmit_recursion_inc(void)
3280{
3281 __this_cpu_inc(softnet_data.xmit.recursion);
3282}
3283
3284static inline void dev_xmit_recursion_dec(void)
3285{
3286 __this_cpu_dec(softnet_data.xmit.recursion);
3287}
3288
3289void __netif_schedule(struct Qdisc *q);
3290void netif_schedule_queue(struct netdev_queue *txq);
3291
3292static inline void netif_tx_schedule_all(struct net_device *dev)
3293{
3294 unsigned int i;
3295
3296 for (i = 0; i < dev->num_tx_queues; i++)
3297 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3298}
3299
3300static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3301{
3302 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3303}
3304
3305/**
3306 * netif_start_queue - allow transmit
3307 * @dev: network device
3308 *
3309 * Allow upper layers to call the device hard_start_xmit routine.
3310 */
3311static inline void netif_start_queue(struct net_device *dev)
3312{
3313 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3314}
3315
3316static inline void netif_tx_start_all_queues(struct net_device *dev)
3317{
3318 unsigned int i;
3319
3320 for (i = 0; i < dev->num_tx_queues; i++) {
3321 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3322 netif_tx_start_queue(txq);
3323 }
3324}
3325
3326void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3327
3328/**
3329 * netif_wake_queue - restart transmit
3330 * @dev: network device
3331 *
3332 * Allow upper layers to call the device hard_start_xmit routine.
3333 * Used for flow control when transmit resources are available.
3334 */
3335static inline void netif_wake_queue(struct net_device *dev)
3336{
3337 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3338}
3339
3340static inline void netif_tx_wake_all_queues(struct net_device *dev)
3341{
3342 unsigned int i;
3343
3344 for (i = 0; i < dev->num_tx_queues; i++) {
3345 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3346 netif_tx_wake_queue(txq);
3347 }
3348}
3349
3350static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3351{
3352 /* Must be an atomic op see netif_txq_try_stop() */
3353 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3354}
3355
3356/**
3357 * netif_stop_queue - stop transmitted packets
3358 * @dev: network device
3359 *
3360 * Stop upper layers calling the device hard_start_xmit routine.
3361 * Used for flow control when transmit resources are unavailable.
3362 */
3363static inline void netif_stop_queue(struct net_device *dev)
3364{
3365 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3366}
3367
3368void netif_tx_stop_all_queues(struct net_device *dev);
3369
3370static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3371{
3372 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3373}
3374
3375/**
3376 * netif_queue_stopped - test if transmit queue is flowblocked
3377 * @dev: network device
3378 *
3379 * Test if transmit queue on device is currently unable to send.
3380 */
3381static inline bool netif_queue_stopped(const struct net_device *dev)
3382{
3383 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3384}
3385
3386static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3387{
3388 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3389}
3390
3391static inline bool
3392netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3393{
3394 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3395}
3396
3397static inline bool
3398netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3399{
3400 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3401}
3402
3403/**
3404 * netdev_queue_set_dql_min_limit - set dql minimum limit
3405 * @dev_queue: pointer to transmit queue
3406 * @min_limit: dql minimum limit
3407 *
3408 * Forces xmit_more() to return true until the minimum threshold
3409 * defined by @min_limit is reached (or until the tx queue is
3410 * empty). Warning: to be use with care, misuse will impact the
3411 * latency.
3412 */
3413static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3414 unsigned int min_limit)
3415{
3416#ifdef CONFIG_BQL
3417 dev_queue->dql.min_limit = min_limit;
3418#endif
3419}
3420
3421/**
3422 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3423 * @dev_queue: pointer to transmit queue
3424 *
3425 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3426 * to give appropriate hint to the CPU.
3427 */
3428static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3429{
3430#ifdef CONFIG_BQL
3431 prefetchw(&dev_queue->dql.num_queued);
3432#endif
3433}
3434
3435/**
3436 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3437 * @dev_queue: pointer to transmit queue
3438 *
3439 * BQL enabled drivers might use this helper in their TX completion path,
3440 * to give appropriate hint to the CPU.
3441 */
3442static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3443{
3444#ifdef CONFIG_BQL
3445 prefetchw(&dev_queue->dql.limit);
3446#endif
3447}
3448
3449/**
3450 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3451 * @dev_queue: network device queue
3452 * @bytes: number of bytes queued to the device queue
3453 *
3454 * Report the number of bytes queued for sending/completion to the network
3455 * device hardware queue. @bytes should be a good approximation and should
3456 * exactly match netdev_completed_queue() @bytes.
3457 * This is typically called once per packet, from ndo_start_xmit().
3458 */
3459static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3460 unsigned int bytes)
3461{
3462#ifdef CONFIG_BQL
3463 dql_queued(&dev_queue->dql, bytes);
3464
3465 if (likely(dql_avail(&dev_queue->dql) >= 0))
3466 return;
3467
3468 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3469
3470 /*
3471 * The XOFF flag must be set before checking the dql_avail below,
3472 * because in netdev_tx_completed_queue we update the dql_completed
3473 * before checking the XOFF flag.
3474 */
3475 smp_mb();
3476
3477 /* check again in case another CPU has just made room avail */
3478 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3479 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3480#endif
3481}
3482
3483/* Variant of netdev_tx_sent_queue() for drivers that are aware
3484 * that they should not test BQL status themselves.
3485 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3486 * skb of a batch.
3487 * Returns true if the doorbell must be used to kick the NIC.
3488 */
3489static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3490 unsigned int bytes,
3491 bool xmit_more)
3492{
3493 if (xmit_more) {
3494#ifdef CONFIG_BQL
3495 dql_queued(&dev_queue->dql, bytes);
3496#endif
3497 return netif_tx_queue_stopped(dev_queue);
3498 }
3499 netdev_tx_sent_queue(dev_queue, bytes);
3500 return true;
3501}
3502
3503/**
3504 * netdev_sent_queue - report the number of bytes queued to hardware
3505 * @dev: network device
3506 * @bytes: number of bytes queued to the hardware device queue
3507 *
3508 * Report the number of bytes queued for sending/completion to the network
3509 * device hardware queue#0. @bytes should be a good approximation and should
3510 * exactly match netdev_completed_queue() @bytes.
3511 * This is typically called once per packet, from ndo_start_xmit().
3512 */
3513static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3514{
3515 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3516}
3517
3518static inline bool __netdev_sent_queue(struct net_device *dev,
3519 unsigned int bytes,
3520 bool xmit_more)
3521{
3522 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3523 xmit_more);
3524}
3525
3526/**
3527 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3528 * @dev_queue: network device queue
3529 * @pkts: number of packets (currently ignored)
3530 * @bytes: number of bytes dequeued from the device queue
3531 *
3532 * Must be called at most once per TX completion round (and not per
3533 * individual packet), so that BQL can adjust its limits appropriately.
3534 */
3535static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3536 unsigned int pkts, unsigned int bytes)
3537{
3538#ifdef CONFIG_BQL
3539 if (unlikely(!bytes))
3540 return;
3541
3542 dql_completed(&dev_queue->dql, bytes);
3543
3544 /*
3545 * Without the memory barrier there is a small possiblity that
3546 * netdev_tx_sent_queue will miss the update and cause the queue to
3547 * be stopped forever
3548 */
3549 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3550
3551 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3552 return;
3553
3554 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3555 netif_schedule_queue(dev_queue);
3556#endif
3557}
3558
3559/**
3560 * netdev_completed_queue - report bytes and packets completed by device
3561 * @dev: network device
3562 * @pkts: actual number of packets sent over the medium
3563 * @bytes: actual number of bytes sent over the medium
3564 *
3565 * Report the number of bytes and packets transmitted by the network device
3566 * hardware queue over the physical medium, @bytes must exactly match the
3567 * @bytes amount passed to netdev_sent_queue()
3568 */
3569static inline void netdev_completed_queue(struct net_device *dev,
3570 unsigned int pkts, unsigned int bytes)
3571{
3572 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3573}
3574
3575static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3576{
3577#ifdef CONFIG_BQL
3578 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3579 dql_reset(&q->dql);
3580#endif
3581}
3582
3583/**
3584 * netdev_reset_queue - reset the packets and bytes count of a network device
3585 * @dev_queue: network device
3586 *
3587 * Reset the bytes and packet count of a network device and clear the
3588 * software flow control OFF bit for this network device
3589 */
3590static inline void netdev_reset_queue(struct net_device *dev_queue)
3591{
3592 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3593}
3594
3595/**
3596 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3597 * @dev: network device
3598 * @queue_index: given tx queue index
3599 *
3600 * Returns 0 if given tx queue index >= number of device tx queues,
3601 * otherwise returns the originally passed tx queue index.
3602 */
3603static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3604{
3605 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3606 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3607 dev->name, queue_index,
3608 dev->real_num_tx_queues);
3609 return 0;
3610 }
3611
3612 return queue_index;
3613}
3614
3615/**
3616 * netif_running - test if up
3617 * @dev: network device
3618 *
3619 * Test if the device has been brought up.
3620 */
3621static inline bool netif_running(const struct net_device *dev)
3622{
3623 return test_bit(__LINK_STATE_START, &dev->state);
3624}
3625
3626/*
3627 * Routines to manage the subqueues on a device. We only need start,
3628 * stop, and a check if it's stopped. All other device management is
3629 * done at the overall netdevice level.
3630 * Also test the device if we're multiqueue.
3631 */
3632
3633/**
3634 * netif_start_subqueue - allow sending packets on subqueue
3635 * @dev: network device
3636 * @queue_index: sub queue index
3637 *
3638 * Start individual transmit queue of a device with multiple transmit queues.
3639 */
3640static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3641{
3642 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3643
3644 netif_tx_start_queue(txq);
3645}
3646
3647/**
3648 * netif_stop_subqueue - stop sending packets on subqueue
3649 * @dev: network device
3650 * @queue_index: sub queue index
3651 *
3652 * Stop individual transmit queue of a device with multiple transmit queues.
3653 */
3654static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3655{
3656 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3657 netif_tx_stop_queue(txq);
3658}
3659
3660/**
3661 * __netif_subqueue_stopped - test status of subqueue
3662 * @dev: network device
3663 * @queue_index: sub queue index
3664 *
3665 * Check individual transmit queue of a device with multiple transmit queues.
3666 */
3667static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3668 u16 queue_index)
3669{
3670 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3671
3672 return netif_tx_queue_stopped(txq);
3673}
3674
3675/**
3676 * netif_subqueue_stopped - test status of subqueue
3677 * @dev: network device
3678 * @skb: sub queue buffer pointer
3679 *
3680 * Check individual transmit queue of a device with multiple transmit queues.
3681 */
3682static inline bool netif_subqueue_stopped(const struct net_device *dev,
3683 struct sk_buff *skb)
3684{
3685 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3686}
3687
3688/**
3689 * netif_wake_subqueue - allow sending packets on subqueue
3690 * @dev: network device
3691 * @queue_index: sub queue index
3692 *
3693 * Resume individual transmit queue of a device with multiple transmit queues.
3694 */
3695static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3696{
3697 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3698
3699 netif_tx_wake_queue(txq);
3700}
3701
3702#ifdef CONFIG_XPS
3703int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3704 u16 index);
3705int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3706 u16 index, enum xps_map_type type);
3707
3708/**
3709 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3710 * @j: CPU/Rx queue index
3711 * @mask: bitmask of all cpus/rx queues
3712 * @nr_bits: number of bits in the bitmask
3713 *
3714 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3715 */
3716static inline bool netif_attr_test_mask(unsigned long j,
3717 const unsigned long *mask,
3718 unsigned int nr_bits)
3719{
3720 cpu_max_bits_warn(j, nr_bits);
3721 return test_bit(j, mask);
3722}
3723
3724/**
3725 * netif_attr_test_online - Test for online CPU/Rx queue
3726 * @j: CPU/Rx queue index
3727 * @online_mask: bitmask for CPUs/Rx queues that are online
3728 * @nr_bits: number of bits in the bitmask
3729 *
3730 * Returns true if a CPU/Rx queue is online.
3731 */
3732static inline bool netif_attr_test_online(unsigned long j,
3733 const unsigned long *online_mask,
3734 unsigned int nr_bits)
3735{
3736 cpu_max_bits_warn(j, nr_bits);
3737
3738 if (online_mask)
3739 return test_bit(j, online_mask);
3740
3741 return (j < nr_bits);
3742}
3743
3744/**
3745 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3746 * @n: CPU/Rx queue index
3747 * @srcp: the cpumask/Rx queue mask pointer
3748 * @nr_bits: number of bits in the bitmask
3749 *
3750 * Returns >= nr_bits if no further CPUs/Rx queues set.
3751 */
3752static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3753 unsigned int nr_bits)
3754{
3755 /* -1 is a legal arg here. */
3756 if (n != -1)
3757 cpu_max_bits_warn(n, nr_bits);
3758
3759 if (srcp)
3760 return find_next_bit(srcp, nr_bits, n + 1);
3761
3762 return n + 1;
3763}
3764
3765/**
3766 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3767 * @n: CPU/Rx queue index
3768 * @src1p: the first CPUs/Rx queues mask pointer
3769 * @src2p: the second CPUs/Rx queues mask pointer
3770 * @nr_bits: number of bits in the bitmask
3771 *
3772 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3773 */
3774static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3775 const unsigned long *src2p,
3776 unsigned int nr_bits)
3777{
3778 /* -1 is a legal arg here. */
3779 if (n != -1)
3780 cpu_max_bits_warn(n, nr_bits);
3781
3782 if (src1p && src2p)
3783 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3784 else if (src1p)
3785 return find_next_bit(src1p, nr_bits, n + 1);
3786 else if (src2p)
3787 return find_next_bit(src2p, nr_bits, n + 1);
3788
3789 return n + 1;
3790}
3791#else
3792static inline int netif_set_xps_queue(struct net_device *dev,
3793 const struct cpumask *mask,
3794 u16 index)
3795{
3796 return 0;
3797}
3798
3799static inline int __netif_set_xps_queue(struct net_device *dev,
3800 const unsigned long *mask,
3801 u16 index, enum xps_map_type type)
3802{
3803 return 0;
3804}
3805#endif
3806
3807/**
3808 * netif_is_multiqueue - test if device has multiple transmit queues
3809 * @dev: network device
3810 *
3811 * Check if device has multiple transmit queues
3812 */
3813static inline bool netif_is_multiqueue(const struct net_device *dev)
3814{
3815 return dev->num_tx_queues > 1;
3816}
3817
3818int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3819
3820#ifdef CONFIG_SYSFS
3821int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3822#else
3823static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3824 unsigned int rxqs)
3825{
3826 dev->real_num_rx_queues = rxqs;
3827 return 0;
3828}
3829#endif
3830int netif_set_real_num_queues(struct net_device *dev,
3831 unsigned int txq, unsigned int rxq);
3832
3833static inline struct netdev_rx_queue *
3834__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3835{
3836 return dev->_rx + rxq;
3837}
3838
3839#ifdef CONFIG_SYSFS
3840static inline unsigned int get_netdev_rx_queue_index(
3841 struct netdev_rx_queue *queue)
3842{
3843 struct net_device *dev = queue->dev;
3844 int index = queue - dev->_rx;
3845
3846 BUG_ON(index >= dev->num_rx_queues);
3847 return index;
3848}
3849#endif
3850
3851int netif_get_num_default_rss_queues(void);
3852
3853void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3854void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3855
3856/*
3857 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3858 * interrupt context or with hardware interrupts being disabled.
3859 * (in_hardirq() || irqs_disabled())
3860 *
3861 * We provide four helpers that can be used in following contexts :
3862 *
3863 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3864 * replacing kfree_skb(skb)
3865 *
3866 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3867 * Typically used in place of consume_skb(skb) in TX completion path
3868 *
3869 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3870 * replacing kfree_skb(skb)
3871 *
3872 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3873 * and consumed a packet. Used in place of consume_skb(skb)
3874 */
3875static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3876{
3877 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3878}
3879
3880static inline void dev_consume_skb_irq(struct sk_buff *skb)
3881{
3882 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3883}
3884
3885static inline void dev_kfree_skb_any(struct sk_buff *skb)
3886{
3887 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3888}
3889
3890static inline void dev_consume_skb_any(struct sk_buff *skb)
3891{
3892 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3893}
3894
3895u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3896 struct bpf_prog *xdp_prog);
3897void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3898int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3899int netif_rx(struct sk_buff *skb);
3900int __netif_rx(struct sk_buff *skb);
3901
3902int netif_receive_skb(struct sk_buff *skb);
3903int netif_receive_skb_core(struct sk_buff *skb);
3904void netif_receive_skb_list_internal(struct list_head *head);
3905void netif_receive_skb_list(struct list_head *head);
3906gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3907void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3908struct sk_buff *napi_get_frags(struct napi_struct *napi);
3909void napi_get_frags_check(struct napi_struct *napi);
3910gro_result_t napi_gro_frags(struct napi_struct *napi);
3911struct packet_offload *gro_find_receive_by_type(__be16 type);
3912struct packet_offload *gro_find_complete_by_type(__be16 type);
3913
3914static inline void napi_free_frags(struct napi_struct *napi)
3915{
3916 kfree_skb(napi->skb);
3917 napi->skb = NULL;
3918}
3919
3920bool netdev_is_rx_handler_busy(struct net_device *dev);
3921int netdev_rx_handler_register(struct net_device *dev,
3922 rx_handler_func_t *rx_handler,
3923 void *rx_handler_data);
3924void netdev_rx_handler_unregister(struct net_device *dev);
3925
3926bool dev_valid_name(const char *name);
3927static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3928{
3929 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3930}
3931int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3932int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3933int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3934 void __user *data, bool *need_copyout);
3935int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3936int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3937unsigned int dev_get_flags(const struct net_device *);
3938int __dev_change_flags(struct net_device *dev, unsigned int flags,
3939 struct netlink_ext_ack *extack);
3940int dev_change_flags(struct net_device *dev, unsigned int flags,
3941 struct netlink_ext_ack *extack);
3942int dev_set_alias(struct net_device *, const char *, size_t);
3943int dev_get_alias(const struct net_device *, char *, size_t);
3944int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3945 const char *pat, int new_ifindex);
3946static inline
3947int dev_change_net_namespace(struct net_device *dev, struct net *net,
3948 const char *pat)
3949{
3950 return __dev_change_net_namespace(dev, net, pat, 0);
3951}
3952int __dev_set_mtu(struct net_device *, int);
3953int dev_set_mtu(struct net_device *, int);
3954int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3955 struct netlink_ext_ack *extack);
3956int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3957 struct netlink_ext_ack *extack);
3958int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3959 struct netlink_ext_ack *extack);
3960int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3961int dev_get_port_parent_id(struct net_device *dev,
3962 struct netdev_phys_item_id *ppid, bool recurse);
3963bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3964struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3965struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3966 struct netdev_queue *txq, int *ret);
3967
3968int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3969u8 dev_xdp_prog_count(struct net_device *dev);
3970u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3971
3972int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3973int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3974int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3975bool is_skb_forwardable(const struct net_device *dev,
3976 const struct sk_buff *skb);
3977
3978static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3979 const struct sk_buff *skb,
3980 const bool check_mtu)
3981{
3982 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3983 unsigned int len;
3984
3985 if (!(dev->flags & IFF_UP))
3986 return false;
3987
3988 if (!check_mtu)
3989 return true;
3990
3991 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3992 if (skb->len <= len)
3993 return true;
3994
3995 /* if TSO is enabled, we don't care about the length as the packet
3996 * could be forwarded without being segmented before
3997 */
3998 if (skb_is_gso(skb))
3999 return true;
4000
4001 return false;
4002}
4003
4004struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
4005
4006static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
4007{
4008 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
4009 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
4010
4011 if (likely(p))
4012 return p;
4013
4014 return netdev_core_stats_alloc(dev);
4015}
4016
4017#define DEV_CORE_STATS_INC(FIELD) \
4018static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4019{ \
4020 struct net_device_core_stats __percpu *p; \
4021 \
4022 p = dev_core_stats(dev); \
4023 if (p) \
4024 this_cpu_inc(p->FIELD); \
4025}
4026DEV_CORE_STATS_INC(rx_dropped)
4027DEV_CORE_STATS_INC(tx_dropped)
4028DEV_CORE_STATS_INC(rx_nohandler)
4029DEV_CORE_STATS_INC(rx_otherhost_dropped)
4030
4031static __always_inline int ____dev_forward_skb(struct net_device *dev,
4032 struct sk_buff *skb,
4033 const bool check_mtu)
4034{
4035 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4036 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4037 dev_core_stats_rx_dropped_inc(dev);
4038 kfree_skb(skb);
4039 return NET_RX_DROP;
4040 }
4041
4042 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4043 skb->priority = 0;
4044 return 0;
4045}
4046
4047bool dev_nit_active(struct net_device *dev);
4048void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4049
4050static inline void __dev_put(struct net_device *dev)
4051{
4052 if (dev) {
4053#ifdef CONFIG_PCPU_DEV_REFCNT
4054 this_cpu_dec(*dev->pcpu_refcnt);
4055#else
4056 refcount_dec(&dev->dev_refcnt);
4057#endif
4058 }
4059}
4060
4061static inline void __dev_hold(struct net_device *dev)
4062{
4063 if (dev) {
4064#ifdef CONFIG_PCPU_DEV_REFCNT
4065 this_cpu_inc(*dev->pcpu_refcnt);
4066#else
4067 refcount_inc(&dev->dev_refcnt);
4068#endif
4069 }
4070}
4071
4072static inline void __netdev_tracker_alloc(struct net_device *dev,
4073 netdevice_tracker *tracker,
4074 gfp_t gfp)
4075{
4076#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4077 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4078#endif
4079}
4080
4081/* netdev_tracker_alloc() can upgrade a prior untracked reference
4082 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4083 */
4084static inline void netdev_tracker_alloc(struct net_device *dev,
4085 netdevice_tracker *tracker, gfp_t gfp)
4086{
4087#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4088 refcount_dec(&dev->refcnt_tracker.no_tracker);
4089 __netdev_tracker_alloc(dev, tracker, gfp);
4090#endif
4091}
4092
4093static inline void netdev_tracker_free(struct net_device *dev,
4094 netdevice_tracker *tracker)
4095{
4096#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4097 ref_tracker_free(&dev->refcnt_tracker, tracker);
4098#endif
4099}
4100
4101static inline void netdev_hold(struct net_device *dev,
4102 netdevice_tracker *tracker, gfp_t gfp)
4103{
4104 if (dev) {
4105 __dev_hold(dev);
4106 __netdev_tracker_alloc(dev, tracker, gfp);
4107 }
4108}
4109
4110static inline void netdev_put(struct net_device *dev,
4111 netdevice_tracker *tracker)
4112{
4113 if (dev) {
4114 netdev_tracker_free(dev, tracker);
4115 __dev_put(dev);
4116 }
4117}
4118
4119/**
4120 * dev_hold - get reference to device
4121 * @dev: network device
4122 *
4123 * Hold reference to device to keep it from being freed.
4124 * Try using netdev_hold() instead.
4125 */
4126static inline void dev_hold(struct net_device *dev)
4127{
4128 netdev_hold(dev, NULL, GFP_ATOMIC);
4129}
4130
4131/**
4132 * dev_put - release reference to device
4133 * @dev: network device
4134 *
4135 * Release reference to device to allow it to be freed.
4136 * Try using netdev_put() instead.
4137 */
4138static inline void dev_put(struct net_device *dev)
4139{
4140 netdev_put(dev, NULL);
4141}
4142
4143static inline void netdev_ref_replace(struct net_device *odev,
4144 struct net_device *ndev,
4145 netdevice_tracker *tracker,
4146 gfp_t gfp)
4147{
4148 if (odev)
4149 netdev_tracker_free(odev, tracker);
4150
4151 __dev_hold(ndev);
4152 __dev_put(odev);
4153
4154 if (ndev)
4155 __netdev_tracker_alloc(ndev, tracker, gfp);
4156}
4157
4158/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4159 * and _off may be called from IRQ context, but it is caller
4160 * who is responsible for serialization of these calls.
4161 *
4162 * The name carrier is inappropriate, these functions should really be
4163 * called netif_lowerlayer_*() because they represent the state of any
4164 * kind of lower layer not just hardware media.
4165 */
4166void linkwatch_fire_event(struct net_device *dev);
4167
4168/**
4169 * netif_carrier_ok - test if carrier present
4170 * @dev: network device
4171 *
4172 * Check if carrier is present on device
4173 */
4174static inline bool netif_carrier_ok(const struct net_device *dev)
4175{
4176 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4177}
4178
4179unsigned long dev_trans_start(struct net_device *dev);
4180
4181void __netdev_watchdog_up(struct net_device *dev);
4182
4183void netif_carrier_on(struct net_device *dev);
4184void netif_carrier_off(struct net_device *dev);
4185void netif_carrier_event(struct net_device *dev);
4186
4187/**
4188 * netif_dormant_on - mark device as dormant.
4189 * @dev: network device
4190 *
4191 * Mark device as dormant (as per RFC2863).
4192 *
4193 * The dormant state indicates that the relevant interface is not
4194 * actually in a condition to pass packets (i.e., it is not 'up') but is
4195 * in a "pending" state, waiting for some external event. For "on-
4196 * demand" interfaces, this new state identifies the situation where the
4197 * interface is waiting for events to place it in the up state.
4198 */
4199static inline void netif_dormant_on(struct net_device *dev)
4200{
4201 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4202 linkwatch_fire_event(dev);
4203}
4204
4205/**
4206 * netif_dormant_off - set device as not dormant.
4207 * @dev: network device
4208 *
4209 * Device is not in dormant state.
4210 */
4211static inline void netif_dormant_off(struct net_device *dev)
4212{
4213 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4214 linkwatch_fire_event(dev);
4215}
4216
4217/**
4218 * netif_dormant - test if device is dormant
4219 * @dev: network device
4220 *
4221 * Check if device is dormant.
4222 */
4223static inline bool netif_dormant(const struct net_device *dev)
4224{
4225 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4226}
4227
4228
4229/**
4230 * netif_testing_on - mark device as under test.
4231 * @dev: network device
4232 *
4233 * Mark device as under test (as per RFC2863).
4234 *
4235 * The testing state indicates that some test(s) must be performed on
4236 * the interface. After completion, of the test, the interface state
4237 * will change to up, dormant, or down, as appropriate.
4238 */
4239static inline void netif_testing_on(struct net_device *dev)
4240{
4241 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4242 linkwatch_fire_event(dev);
4243}
4244
4245/**
4246 * netif_testing_off - set device as not under test.
4247 * @dev: network device
4248 *
4249 * Device is not in testing state.
4250 */
4251static inline void netif_testing_off(struct net_device *dev)
4252{
4253 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4254 linkwatch_fire_event(dev);
4255}
4256
4257/**
4258 * netif_testing - test if device is under test
4259 * @dev: network device
4260 *
4261 * Check if device is under test
4262 */
4263static inline bool netif_testing(const struct net_device *dev)
4264{
4265 return test_bit(__LINK_STATE_TESTING, &dev->state);
4266}
4267
4268
4269/**
4270 * netif_oper_up - test if device is operational
4271 * @dev: network device
4272 *
4273 * Check if carrier is operational
4274 */
4275static inline bool netif_oper_up(const struct net_device *dev)
4276{
4277 return (dev->operstate == IF_OPER_UP ||
4278 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4279}
4280
4281/**
4282 * netif_device_present - is device available or removed
4283 * @dev: network device
4284 *
4285 * Check if device has not been removed from system.
4286 */
4287static inline bool netif_device_present(const struct net_device *dev)
4288{
4289 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4290}
4291
4292void netif_device_detach(struct net_device *dev);
4293
4294void netif_device_attach(struct net_device *dev);
4295
4296/*
4297 * Network interface message level settings
4298 */
4299
4300enum {
4301 NETIF_MSG_DRV_BIT,
4302 NETIF_MSG_PROBE_BIT,
4303 NETIF_MSG_LINK_BIT,
4304 NETIF_MSG_TIMER_BIT,
4305 NETIF_MSG_IFDOWN_BIT,
4306 NETIF_MSG_IFUP_BIT,
4307 NETIF_MSG_RX_ERR_BIT,
4308 NETIF_MSG_TX_ERR_BIT,
4309 NETIF_MSG_TX_QUEUED_BIT,
4310 NETIF_MSG_INTR_BIT,
4311 NETIF_MSG_TX_DONE_BIT,
4312 NETIF_MSG_RX_STATUS_BIT,
4313 NETIF_MSG_PKTDATA_BIT,
4314 NETIF_MSG_HW_BIT,
4315 NETIF_MSG_WOL_BIT,
4316
4317 /* When you add a new bit above, update netif_msg_class_names array
4318 * in net/ethtool/common.c
4319 */
4320 NETIF_MSG_CLASS_COUNT,
4321};
4322/* Both ethtool_ops interface and internal driver implementation use u32 */
4323static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4324
4325#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4326#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4327
4328#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4329#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4330#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4331#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4332#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4333#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4334#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4335#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4336#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4337#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4338#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4339#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4340#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4341#define NETIF_MSG_HW __NETIF_MSG(HW)
4342#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4343
4344#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4345#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4346#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4347#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4348#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4349#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4350#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4351#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4352#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4353#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4354#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4355#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4356#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4357#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4358#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4359
4360static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4361{
4362 /* use default */
4363 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4364 return default_msg_enable_bits;
4365 if (debug_value == 0) /* no output */
4366 return 0;
4367 /* set low N bits */
4368 return (1U << debug_value) - 1;
4369}
4370
4371static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4372{
4373 spin_lock(&txq->_xmit_lock);
4374 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4375 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4376}
4377
4378static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4379{
4380 __acquire(&txq->_xmit_lock);
4381 return true;
4382}
4383
4384static inline void __netif_tx_release(struct netdev_queue *txq)
4385{
4386 __release(&txq->_xmit_lock);
4387}
4388
4389static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4390{
4391 spin_lock_bh(&txq->_xmit_lock);
4392 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4393 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4394}
4395
4396static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4397{
4398 bool ok = spin_trylock(&txq->_xmit_lock);
4399
4400 if (likely(ok)) {
4401 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4402 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4403 }
4404 return ok;
4405}
4406
4407static inline void __netif_tx_unlock(struct netdev_queue *txq)
4408{
4409 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4410 WRITE_ONCE(txq->xmit_lock_owner, -1);
4411 spin_unlock(&txq->_xmit_lock);
4412}
4413
4414static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4415{
4416 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4417 WRITE_ONCE(txq->xmit_lock_owner, -1);
4418 spin_unlock_bh(&txq->_xmit_lock);
4419}
4420
4421/*
4422 * txq->trans_start can be read locklessly from dev_watchdog()
4423 */
4424static inline void txq_trans_update(struct netdev_queue *txq)
4425{
4426 if (txq->xmit_lock_owner != -1)
4427 WRITE_ONCE(txq->trans_start, jiffies);
4428}
4429
4430static inline void txq_trans_cond_update(struct netdev_queue *txq)
4431{
4432 unsigned long now = jiffies;
4433
4434 if (READ_ONCE(txq->trans_start) != now)
4435 WRITE_ONCE(txq->trans_start, now);
4436}
4437
4438/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4439static inline void netif_trans_update(struct net_device *dev)
4440{
4441 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4442
4443 txq_trans_cond_update(txq);
4444}
4445
4446/**
4447 * netif_tx_lock - grab network device transmit lock
4448 * @dev: network device
4449 *
4450 * Get network device transmit lock
4451 */
4452void netif_tx_lock(struct net_device *dev);
4453
4454static inline void netif_tx_lock_bh(struct net_device *dev)
4455{
4456 local_bh_disable();
4457 netif_tx_lock(dev);
4458}
4459
4460void netif_tx_unlock(struct net_device *dev);
4461
4462static inline void netif_tx_unlock_bh(struct net_device *dev)
4463{
4464 netif_tx_unlock(dev);
4465 local_bh_enable();
4466}
4467
4468#define HARD_TX_LOCK(dev, txq, cpu) { \
4469 if ((dev->features & NETIF_F_LLTX) == 0) { \
4470 __netif_tx_lock(txq, cpu); \
4471 } else { \
4472 __netif_tx_acquire(txq); \
4473 } \
4474}
4475
4476#define HARD_TX_TRYLOCK(dev, txq) \
4477 (((dev->features & NETIF_F_LLTX) == 0) ? \
4478 __netif_tx_trylock(txq) : \
4479 __netif_tx_acquire(txq))
4480
4481#define HARD_TX_UNLOCK(dev, txq) { \
4482 if ((dev->features & NETIF_F_LLTX) == 0) { \
4483 __netif_tx_unlock(txq); \
4484 } else { \
4485 __netif_tx_release(txq); \
4486 } \
4487}
4488
4489static inline void netif_tx_disable(struct net_device *dev)
4490{
4491 unsigned int i;
4492 int cpu;
4493
4494 local_bh_disable();
4495 cpu = smp_processor_id();
4496 spin_lock(&dev->tx_global_lock);
4497 for (i = 0; i < dev->num_tx_queues; i++) {
4498 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4499
4500 __netif_tx_lock(txq, cpu);
4501 netif_tx_stop_queue(txq);
4502 __netif_tx_unlock(txq);
4503 }
4504 spin_unlock(&dev->tx_global_lock);
4505 local_bh_enable();
4506}
4507
4508static inline void netif_addr_lock(struct net_device *dev)
4509{
4510 unsigned char nest_level = 0;
4511
4512#ifdef CONFIG_LOCKDEP
4513 nest_level = dev->nested_level;
4514#endif
4515 spin_lock_nested(&dev->addr_list_lock, nest_level);
4516}
4517
4518static inline void netif_addr_lock_bh(struct net_device *dev)
4519{
4520 unsigned char nest_level = 0;
4521
4522#ifdef CONFIG_LOCKDEP
4523 nest_level = dev->nested_level;
4524#endif
4525 local_bh_disable();
4526 spin_lock_nested(&dev->addr_list_lock, nest_level);
4527}
4528
4529static inline void netif_addr_unlock(struct net_device *dev)
4530{
4531 spin_unlock(&dev->addr_list_lock);
4532}
4533
4534static inline void netif_addr_unlock_bh(struct net_device *dev)
4535{
4536 spin_unlock_bh(&dev->addr_list_lock);
4537}
4538
4539/*
4540 * dev_addrs walker. Should be used only for read access. Call with
4541 * rcu_read_lock held.
4542 */
4543#define for_each_dev_addr(dev, ha) \
4544 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4545
4546/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4547
4548void ether_setup(struct net_device *dev);
4549
4550/* Support for loadable net-drivers */
4551struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4552 unsigned char name_assign_type,
4553 void (*setup)(struct net_device *),
4554 unsigned int txqs, unsigned int rxqs);
4555#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4556 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4557
4558#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4559 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4560 count)
4561
4562int register_netdev(struct net_device *dev);
4563void unregister_netdev(struct net_device *dev);
4564
4565int devm_register_netdev(struct device *dev, struct net_device *ndev);
4566
4567/* General hardware address lists handling functions */
4568int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4569 struct netdev_hw_addr_list *from_list, int addr_len);
4570void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4571 struct netdev_hw_addr_list *from_list, int addr_len);
4572int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4573 struct net_device *dev,
4574 int (*sync)(struct net_device *, const unsigned char *),
4575 int (*unsync)(struct net_device *,
4576 const unsigned char *));
4577int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4578 struct net_device *dev,
4579 int (*sync)(struct net_device *,
4580 const unsigned char *, int),
4581 int (*unsync)(struct net_device *,
4582 const unsigned char *, int));
4583void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4584 struct net_device *dev,
4585 int (*unsync)(struct net_device *,
4586 const unsigned char *, int));
4587void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4588 struct net_device *dev,
4589 int (*unsync)(struct net_device *,
4590 const unsigned char *));
4591void __hw_addr_init(struct netdev_hw_addr_list *list);
4592
4593/* Functions used for device addresses handling */
4594void dev_addr_mod(struct net_device *dev, unsigned int offset,
4595 const void *addr, size_t len);
4596
4597static inline void
4598__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4599{
4600 dev_addr_mod(dev, 0, addr, len);
4601}
4602
4603static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4604{
4605 __dev_addr_set(dev, addr, dev->addr_len);
4606}
4607
4608int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4609 unsigned char addr_type);
4610int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4611 unsigned char addr_type);
4612
4613/* Functions used for unicast addresses handling */
4614int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4615int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4616int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4617int dev_uc_sync(struct net_device *to, struct net_device *from);
4618int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4619void dev_uc_unsync(struct net_device *to, struct net_device *from);
4620void dev_uc_flush(struct net_device *dev);
4621void dev_uc_init(struct net_device *dev);
4622
4623/**
4624 * __dev_uc_sync - Synchonize device's unicast list
4625 * @dev: device to sync
4626 * @sync: function to call if address should be added
4627 * @unsync: function to call if address should be removed
4628 *
4629 * Add newly added addresses to the interface, and release
4630 * addresses that have been deleted.
4631 */
4632static inline int __dev_uc_sync(struct net_device *dev,
4633 int (*sync)(struct net_device *,
4634 const unsigned char *),
4635 int (*unsync)(struct net_device *,
4636 const unsigned char *))
4637{
4638 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4639}
4640
4641/**
4642 * __dev_uc_unsync - Remove synchronized addresses from device
4643 * @dev: device to sync
4644 * @unsync: function to call if address should be removed
4645 *
4646 * Remove all addresses that were added to the device by dev_uc_sync().
4647 */
4648static inline void __dev_uc_unsync(struct net_device *dev,
4649 int (*unsync)(struct net_device *,
4650 const unsigned char *))
4651{
4652 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4653}
4654
4655/* Functions used for multicast addresses handling */
4656int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4657int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4658int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4659int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4660int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4661int dev_mc_sync(struct net_device *to, struct net_device *from);
4662int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4663void dev_mc_unsync(struct net_device *to, struct net_device *from);
4664void dev_mc_flush(struct net_device *dev);
4665void dev_mc_init(struct net_device *dev);
4666
4667/**
4668 * __dev_mc_sync - Synchonize device's multicast list
4669 * @dev: device to sync
4670 * @sync: function to call if address should be added
4671 * @unsync: function to call if address should be removed
4672 *
4673 * Add newly added addresses to the interface, and release
4674 * addresses that have been deleted.
4675 */
4676static inline int __dev_mc_sync(struct net_device *dev,
4677 int (*sync)(struct net_device *,
4678 const unsigned char *),
4679 int (*unsync)(struct net_device *,
4680 const unsigned char *))
4681{
4682 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4683}
4684
4685/**
4686 * __dev_mc_unsync - Remove synchronized addresses from device
4687 * @dev: device to sync
4688 * @unsync: function to call if address should be removed
4689 *
4690 * Remove all addresses that were added to the device by dev_mc_sync().
4691 */
4692static inline void __dev_mc_unsync(struct net_device *dev,
4693 int (*unsync)(struct net_device *,
4694 const unsigned char *))
4695{
4696 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4697}
4698
4699/* Functions used for secondary unicast and multicast support */
4700void dev_set_rx_mode(struct net_device *dev);
4701int dev_set_promiscuity(struct net_device *dev, int inc);
4702int dev_set_allmulti(struct net_device *dev, int inc);
4703void netdev_state_change(struct net_device *dev);
4704void __netdev_notify_peers(struct net_device *dev);
4705void netdev_notify_peers(struct net_device *dev);
4706void netdev_features_change(struct net_device *dev);
4707/* Load a device via the kmod */
4708void dev_load(struct net *net, const char *name);
4709struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4710 struct rtnl_link_stats64 *storage);
4711void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4712 const struct net_device_stats *netdev_stats);
4713void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4714 const struct pcpu_sw_netstats __percpu *netstats);
4715void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4716
4717extern int netdev_max_backlog;
4718extern int dev_rx_weight;
4719extern int dev_tx_weight;
4720extern int gro_normal_batch;
4721
4722enum {
4723 NESTED_SYNC_IMM_BIT,
4724 NESTED_SYNC_TODO_BIT,
4725};
4726
4727#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4728#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4729
4730#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4731#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4732
4733struct netdev_nested_priv {
4734 unsigned char flags;
4735 void *data;
4736};
4737
4738bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4739struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4740 struct list_head **iter);
4741
4742/* iterate through upper list, must be called under RCU read lock */
4743#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4744 for (iter = &(dev)->adj_list.upper, \
4745 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4746 updev; \
4747 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4748
4749int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4750 int (*fn)(struct net_device *upper_dev,
4751 struct netdev_nested_priv *priv),
4752 struct netdev_nested_priv *priv);
4753
4754bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4755 struct net_device *upper_dev);
4756
4757bool netdev_has_any_upper_dev(struct net_device *dev);
4758
4759void *netdev_lower_get_next_private(struct net_device *dev,
4760 struct list_head **iter);
4761void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4762 struct list_head **iter);
4763
4764#define netdev_for_each_lower_private(dev, priv, iter) \
4765 for (iter = (dev)->adj_list.lower.next, \
4766 priv = netdev_lower_get_next_private(dev, &(iter)); \
4767 priv; \
4768 priv = netdev_lower_get_next_private(dev, &(iter)))
4769
4770#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4771 for (iter = &(dev)->adj_list.lower, \
4772 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4773 priv; \
4774 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4775
4776void *netdev_lower_get_next(struct net_device *dev,
4777 struct list_head **iter);
4778
4779#define netdev_for_each_lower_dev(dev, ldev, iter) \
4780 for (iter = (dev)->adj_list.lower.next, \
4781 ldev = netdev_lower_get_next(dev, &(iter)); \
4782 ldev; \
4783 ldev = netdev_lower_get_next(dev, &(iter)))
4784
4785struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4786 struct list_head **iter);
4787int netdev_walk_all_lower_dev(struct net_device *dev,
4788 int (*fn)(struct net_device *lower_dev,
4789 struct netdev_nested_priv *priv),
4790 struct netdev_nested_priv *priv);
4791int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4792 int (*fn)(struct net_device *lower_dev,
4793 struct netdev_nested_priv *priv),
4794 struct netdev_nested_priv *priv);
4795
4796void *netdev_adjacent_get_private(struct list_head *adj_list);
4797void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4798struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4799struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4800int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4801 struct netlink_ext_ack *extack);
4802int netdev_master_upper_dev_link(struct net_device *dev,
4803 struct net_device *upper_dev,
4804 void *upper_priv, void *upper_info,
4805 struct netlink_ext_ack *extack);
4806void netdev_upper_dev_unlink(struct net_device *dev,
4807 struct net_device *upper_dev);
4808int netdev_adjacent_change_prepare(struct net_device *old_dev,
4809 struct net_device *new_dev,
4810 struct net_device *dev,
4811 struct netlink_ext_ack *extack);
4812void netdev_adjacent_change_commit(struct net_device *old_dev,
4813 struct net_device *new_dev,
4814 struct net_device *dev);
4815void netdev_adjacent_change_abort(struct net_device *old_dev,
4816 struct net_device *new_dev,
4817 struct net_device *dev);
4818void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4819void *netdev_lower_dev_get_private(struct net_device *dev,
4820 struct net_device *lower_dev);
4821void netdev_lower_state_changed(struct net_device *lower_dev,
4822 void *lower_state_info);
4823
4824/* RSS keys are 40 or 52 bytes long */
4825#define NETDEV_RSS_KEY_LEN 52
4826extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4827void netdev_rss_key_fill(void *buffer, size_t len);
4828
4829int skb_checksum_help(struct sk_buff *skb);
4830int skb_crc32c_csum_help(struct sk_buff *skb);
4831int skb_csum_hwoffload_help(struct sk_buff *skb,
4832 const netdev_features_t features);
4833
4834struct netdev_bonding_info {
4835 ifslave slave;
4836 ifbond master;
4837};
4838
4839struct netdev_notifier_bonding_info {
4840 struct netdev_notifier_info info; /* must be first */
4841 struct netdev_bonding_info bonding_info;
4842};
4843
4844void netdev_bonding_info_change(struct net_device *dev,
4845 struct netdev_bonding_info *bonding_info);
4846
4847#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4848void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4849#else
4850static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4851 const void *data)
4852{
4853}
4854#endif
4855
4856__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4857
4858static inline bool can_checksum_protocol(netdev_features_t features,
4859 __be16 protocol)
4860{
4861 if (protocol == htons(ETH_P_FCOE))
4862 return !!(features & NETIF_F_FCOE_CRC);
4863
4864 /* Assume this is an IP checksum (not SCTP CRC) */
4865
4866 if (features & NETIF_F_HW_CSUM) {
4867 /* Can checksum everything */
4868 return true;
4869 }
4870
4871 switch (protocol) {
4872 case htons(ETH_P_IP):
4873 return !!(features & NETIF_F_IP_CSUM);
4874 case htons(ETH_P_IPV6):
4875 return !!(features & NETIF_F_IPV6_CSUM);
4876 default:
4877 return false;
4878 }
4879}
4880
4881#ifdef CONFIG_BUG
4882void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4883#else
4884static inline void netdev_rx_csum_fault(struct net_device *dev,
4885 struct sk_buff *skb)
4886{
4887}
4888#endif
4889/* rx skb timestamps */
4890void net_enable_timestamp(void);
4891void net_disable_timestamp(void);
4892
4893static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4894 const struct skb_shared_hwtstamps *hwtstamps,
4895 bool cycles)
4896{
4897 const struct net_device_ops *ops = dev->netdev_ops;
4898
4899 if (ops->ndo_get_tstamp)
4900 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4901
4902 return hwtstamps->hwtstamp;
4903}
4904
4905static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4906 struct sk_buff *skb, struct net_device *dev,
4907 bool more)
4908{
4909 __this_cpu_write(softnet_data.xmit.more, more);
4910 return ops->ndo_start_xmit(skb, dev);
4911}
4912
4913static inline bool netdev_xmit_more(void)
4914{
4915 return __this_cpu_read(softnet_data.xmit.more);
4916}
4917
4918static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4919 struct netdev_queue *txq, bool more)
4920{
4921 const struct net_device_ops *ops = dev->netdev_ops;
4922 netdev_tx_t rc;
4923
4924 rc = __netdev_start_xmit(ops, skb, dev, more);
4925 if (rc == NETDEV_TX_OK)
4926 txq_trans_update(txq);
4927
4928 return rc;
4929}
4930
4931int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4932 const void *ns);
4933void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4934 const void *ns);
4935
4936extern const struct kobj_ns_type_operations net_ns_type_operations;
4937
4938const char *netdev_drivername(const struct net_device *dev);
4939
4940static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4941 netdev_features_t f2)
4942{
4943 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4944 if (f1 & NETIF_F_HW_CSUM)
4945 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4946 else
4947 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4948 }
4949
4950 return f1 & f2;
4951}
4952
4953static inline netdev_features_t netdev_get_wanted_features(
4954 struct net_device *dev)
4955{
4956 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4957}
4958netdev_features_t netdev_increment_features(netdev_features_t all,
4959 netdev_features_t one, netdev_features_t mask);
4960
4961/* Allow TSO being used on stacked device :
4962 * Performing the GSO segmentation before last device
4963 * is a performance improvement.
4964 */
4965static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4966 netdev_features_t mask)
4967{
4968 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4969}
4970
4971int __netdev_update_features(struct net_device *dev);
4972void netdev_update_features(struct net_device *dev);
4973void netdev_change_features(struct net_device *dev);
4974
4975void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4976 struct net_device *dev);
4977
4978netdev_features_t passthru_features_check(struct sk_buff *skb,
4979 struct net_device *dev,
4980 netdev_features_t features);
4981netdev_features_t netif_skb_features(struct sk_buff *skb);
4982void skb_warn_bad_offload(const struct sk_buff *skb);
4983
4984static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4985{
4986 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4987
4988 /* check flags correspondence */
4989 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4990 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4991 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4992 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4993 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4994 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4995 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4996 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4997 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4998 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4999 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5000 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5001 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5002 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5003 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5004 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5005 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5006 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5007 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5008
5009 return (features & feature) == feature;
5010}
5011
5012static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5013{
5014 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5015 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5016}
5017
5018static inline bool netif_needs_gso(struct sk_buff *skb,
5019 netdev_features_t features)
5020{
5021 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5022 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5023 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5024}
5025
5026void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5027void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5028void netif_inherit_tso_max(struct net_device *to,
5029 const struct net_device *from);
5030
5031static inline bool netif_is_macsec(const struct net_device *dev)
5032{
5033 return dev->priv_flags & IFF_MACSEC;
5034}
5035
5036static inline bool netif_is_macvlan(const struct net_device *dev)
5037{
5038 return dev->priv_flags & IFF_MACVLAN;
5039}
5040
5041static inline bool netif_is_macvlan_port(const struct net_device *dev)
5042{
5043 return dev->priv_flags & IFF_MACVLAN_PORT;
5044}
5045
5046static inline bool netif_is_bond_master(const struct net_device *dev)
5047{
5048 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5049}
5050
5051static inline bool netif_is_bond_slave(const struct net_device *dev)
5052{
5053 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5054}
5055
5056static inline bool netif_supports_nofcs(struct net_device *dev)
5057{
5058 return dev->priv_flags & IFF_SUPP_NOFCS;
5059}
5060
5061static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5062{
5063 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5064}
5065
5066static inline bool netif_is_l3_master(const struct net_device *dev)
5067{
5068 return dev->priv_flags & IFF_L3MDEV_MASTER;
5069}
5070
5071static inline bool netif_is_l3_slave(const struct net_device *dev)
5072{
5073 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5074}
5075
5076static inline int dev_sdif(const struct net_device *dev)
5077{
5078#ifdef CONFIG_NET_L3_MASTER_DEV
5079 if (netif_is_l3_slave(dev))
5080 return dev->ifindex;
5081#endif
5082 return 0;
5083}
5084
5085static inline bool netif_is_bridge_master(const struct net_device *dev)
5086{
5087 return dev->priv_flags & IFF_EBRIDGE;
5088}
5089
5090static inline bool netif_is_bridge_port(const struct net_device *dev)
5091{
5092 return dev->priv_flags & IFF_BRIDGE_PORT;
5093}
5094
5095static inline bool netif_is_ovs_master(const struct net_device *dev)
5096{
5097 return dev->priv_flags & IFF_OPENVSWITCH;
5098}
5099
5100static inline bool netif_is_ovs_port(const struct net_device *dev)
5101{
5102 return dev->priv_flags & IFF_OVS_DATAPATH;
5103}
5104
5105static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5106{
5107 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5108}
5109
5110static inline bool netif_is_team_master(const struct net_device *dev)
5111{
5112 return dev->priv_flags & IFF_TEAM;
5113}
5114
5115static inline bool netif_is_team_port(const struct net_device *dev)
5116{
5117 return dev->priv_flags & IFF_TEAM_PORT;
5118}
5119
5120static inline bool netif_is_lag_master(const struct net_device *dev)
5121{
5122 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5123}
5124
5125static inline bool netif_is_lag_port(const struct net_device *dev)
5126{
5127 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5128}
5129
5130static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5131{
5132 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5133}
5134
5135static inline bool netif_is_failover(const struct net_device *dev)
5136{
5137 return dev->priv_flags & IFF_FAILOVER;
5138}
5139
5140static inline bool netif_is_failover_slave(const struct net_device *dev)
5141{
5142 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5143}
5144
5145/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5146static inline void netif_keep_dst(struct net_device *dev)
5147{
5148 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5149}
5150
5151/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5152static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5153{
5154 /* TODO: reserve and use an additional IFF bit, if we get more users */
5155 return netif_is_macsec(dev);
5156}
5157
5158extern struct pernet_operations __net_initdata loopback_net_ops;
5159
5160/* Logging, debugging and troubleshooting/diagnostic helpers. */
5161
5162/* netdev_printk helpers, similar to dev_printk */
5163
5164static inline const char *netdev_name(const struct net_device *dev)
5165{
5166 if (!dev->name[0] || strchr(dev->name, '%'))
5167 return "(unnamed net_device)";
5168 return dev->name;
5169}
5170
5171static inline const char *netdev_reg_state(const struct net_device *dev)
5172{
5173 switch (dev->reg_state) {
5174 case NETREG_UNINITIALIZED: return " (uninitialized)";
5175 case NETREG_REGISTERED: return "";
5176 case NETREG_UNREGISTERING: return " (unregistering)";
5177 case NETREG_UNREGISTERED: return " (unregistered)";
5178 case NETREG_RELEASED: return " (released)";
5179 case NETREG_DUMMY: return " (dummy)";
5180 }
5181
5182 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5183 return " (unknown)";
5184}
5185
5186#define MODULE_ALIAS_NETDEV(device) \
5187 MODULE_ALIAS("netdev-" device)
5188
5189/*
5190 * netdev_WARN() acts like dev_printk(), but with the key difference
5191 * of using a WARN/WARN_ON to get the message out, including the
5192 * file/line information and a backtrace.
5193 */
5194#define netdev_WARN(dev, format, args...) \
5195 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5196 netdev_reg_state(dev), ##args)
5197
5198#define netdev_WARN_ONCE(dev, format, args...) \
5199 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5200 netdev_reg_state(dev), ##args)
5201
5202/*
5203 * The list of packet types we will receive (as opposed to discard)
5204 * and the routines to invoke.
5205 *
5206 * Why 16. Because with 16 the only overlap we get on a hash of the
5207 * low nibble of the protocol value is RARP/SNAP/X.25.
5208 *
5209 * 0800 IP
5210 * 0001 802.3
5211 * 0002 AX.25
5212 * 0004 802.2
5213 * 8035 RARP
5214 * 0005 SNAP
5215 * 0805 X.25
5216 * 0806 ARP
5217 * 8137 IPX
5218 * 0009 Localtalk
5219 * 86DD IPv6
5220 */
5221#define PTYPE_HASH_SIZE (16)
5222#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5223
5224extern struct list_head ptype_all __read_mostly;
5225extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5226
5227extern struct net_device *blackhole_netdev;
5228
5229/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5230#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5231#define DEV_STATS_ADD(DEV, FIELD, VAL) \
5232 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5233
5234#endif /* _LINUX_NETDEVICE_H */