Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
24#include <linux/timer.h>
25#include <linux/bug.h>
26#include <linux/delay.h>
27#include <linux/atomic.h>
28#include <linux/prefetch.h>
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31#include <asm/local.h>
32
33#include <linux/percpu.h>
34#include <linux/rculist.h>
35#include <linux/workqueue.h>
36#include <linux/dynamic_queue_limits.h>
37
38#include <net/net_namespace.h>
39#ifdef CONFIG_DCB
40#include <net/dcbnl.h>
41#endif
42#include <net/netprio_cgroup.h>
43
44#include <linux/netdev_features.h>
45#include <linux/neighbour.h>
46#include <uapi/linux/netdevice.h>
47#include <uapi/linux/if_bonding.h>
48#include <uapi/linux/pkt_cls.h>
49#include <uapi/linux/netdev.h>
50#include <linux/hashtable.h>
51#include <linux/rbtree.h>
52#include <net/net_trackers.h>
53#include <net/net_debug.h>
54#include <net/dropreason-core.h>
55
56struct netpoll_info;
57struct device;
58struct ethtool_ops;
59struct kernel_hwtstamp_config;
60struct phy_device;
61struct dsa_port;
62struct ip_tunnel_parm;
63struct macsec_context;
64struct macsec_ops;
65struct netdev_name_node;
66struct sd_flow_limit;
67struct sfp_bus;
68/* 802.11 specific */
69struct wireless_dev;
70/* 802.15.4 specific */
71struct wpan_dev;
72struct mpls_dev;
73/* UDP Tunnel offloads */
74struct udp_tunnel_info;
75struct udp_tunnel_nic_info;
76struct udp_tunnel_nic;
77struct bpf_prog;
78struct xdp_buff;
79struct xdp_frame;
80struct xdp_metadata_ops;
81struct xdp_md;
82/* DPLL specific */
83struct dpll_pin;
84
85typedef u32 xdp_features_t;
86
87void synchronize_net(void);
88void netdev_set_default_ethtool_ops(struct net_device *dev,
89 const struct ethtool_ops *ops);
90void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
91
92/* Backlog congestion levels */
93#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
94#define NET_RX_DROP 1 /* packet dropped */
95
96#define MAX_NEST_DEV 8
97
98/*
99 * Transmit return codes: transmit return codes originate from three different
100 * namespaces:
101 *
102 * - qdisc return codes
103 * - driver transmit return codes
104 * - errno values
105 *
106 * Drivers are allowed to return any one of those in their hard_start_xmit()
107 * function. Real network devices commonly used with qdiscs should only return
108 * the driver transmit return codes though - when qdiscs are used, the actual
109 * transmission happens asynchronously, so the value is not propagated to
110 * higher layers. Virtual network devices transmit synchronously; in this case
111 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
112 * others are propagated to higher layers.
113 */
114
115/* qdisc ->enqueue() return codes. */
116#define NET_XMIT_SUCCESS 0x00
117#define NET_XMIT_DROP 0x01 /* skb dropped */
118#define NET_XMIT_CN 0x02 /* congestion notification */
119#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
120
121/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
122 * indicates that the device will soon be dropping packets, or already drops
123 * some packets of the same priority; prompting us to send less aggressively. */
124#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
125#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
126
127/* Driver transmit return codes */
128#define NETDEV_TX_MASK 0xf0
129
130enum netdev_tx {
131 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
132 NETDEV_TX_OK = 0x00, /* driver took care of packet */
133 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
134};
135typedef enum netdev_tx netdev_tx_t;
136
137/*
138 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
139 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
140 */
141static inline bool dev_xmit_complete(int rc)
142{
143 /*
144 * Positive cases with an skb consumed by a driver:
145 * - successful transmission (rc == NETDEV_TX_OK)
146 * - error while transmitting (rc < 0)
147 * - error while queueing to a different device (rc & NET_XMIT_MASK)
148 */
149 if (likely(rc < NET_XMIT_MASK))
150 return true;
151
152 return false;
153}
154
155/*
156 * Compute the worst-case header length according to the protocols
157 * used.
158 */
159
160#if defined(CONFIG_HYPERV_NET)
161# define LL_MAX_HEADER 128
162#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
163# if defined(CONFIG_MAC80211_MESH)
164# define LL_MAX_HEADER 128
165# else
166# define LL_MAX_HEADER 96
167# endif
168#else
169# define LL_MAX_HEADER 32
170#endif
171
172#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
173 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
174#define MAX_HEADER LL_MAX_HEADER
175#else
176#define MAX_HEADER (LL_MAX_HEADER + 48)
177#endif
178
179/*
180 * Old network device statistics. Fields are native words
181 * (unsigned long) so they can be read and written atomically.
182 */
183
184#define NET_DEV_STAT(FIELD) \
185 union { \
186 unsigned long FIELD; \
187 atomic_long_t __##FIELD; \
188 }
189
190struct net_device_stats {
191 NET_DEV_STAT(rx_packets);
192 NET_DEV_STAT(tx_packets);
193 NET_DEV_STAT(rx_bytes);
194 NET_DEV_STAT(tx_bytes);
195 NET_DEV_STAT(rx_errors);
196 NET_DEV_STAT(tx_errors);
197 NET_DEV_STAT(rx_dropped);
198 NET_DEV_STAT(tx_dropped);
199 NET_DEV_STAT(multicast);
200 NET_DEV_STAT(collisions);
201 NET_DEV_STAT(rx_length_errors);
202 NET_DEV_STAT(rx_over_errors);
203 NET_DEV_STAT(rx_crc_errors);
204 NET_DEV_STAT(rx_frame_errors);
205 NET_DEV_STAT(rx_fifo_errors);
206 NET_DEV_STAT(rx_missed_errors);
207 NET_DEV_STAT(tx_aborted_errors);
208 NET_DEV_STAT(tx_carrier_errors);
209 NET_DEV_STAT(tx_fifo_errors);
210 NET_DEV_STAT(tx_heartbeat_errors);
211 NET_DEV_STAT(tx_window_errors);
212 NET_DEV_STAT(rx_compressed);
213 NET_DEV_STAT(tx_compressed);
214};
215#undef NET_DEV_STAT
216
217/* per-cpu stats, allocated on demand.
218 * Try to fit them in a single cache line, for dev_get_stats() sake.
219 */
220struct net_device_core_stats {
221 unsigned long rx_dropped;
222 unsigned long tx_dropped;
223 unsigned long rx_nohandler;
224 unsigned long rx_otherhost_dropped;
225} __aligned(4 * sizeof(unsigned long));
226
227#include <linux/cache.h>
228#include <linux/skbuff.h>
229
230#ifdef CONFIG_RPS
231#include <linux/static_key.h>
232extern struct static_key_false rps_needed;
233extern struct static_key_false rfs_needed;
234#endif
235
236struct neighbour;
237struct neigh_parms;
238struct sk_buff;
239
240struct netdev_hw_addr {
241 struct list_head list;
242 struct rb_node node;
243 unsigned char addr[MAX_ADDR_LEN];
244 unsigned char type;
245#define NETDEV_HW_ADDR_T_LAN 1
246#define NETDEV_HW_ADDR_T_SAN 2
247#define NETDEV_HW_ADDR_T_UNICAST 3
248#define NETDEV_HW_ADDR_T_MULTICAST 4
249 bool global_use;
250 int sync_cnt;
251 int refcount;
252 int synced;
253 struct rcu_head rcu_head;
254};
255
256struct netdev_hw_addr_list {
257 struct list_head list;
258 int count;
259
260 /* Auxiliary tree for faster lookup on addition and deletion */
261 struct rb_root tree;
262};
263
264#define netdev_hw_addr_list_count(l) ((l)->count)
265#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
266#define netdev_hw_addr_list_for_each(ha, l) \
267 list_for_each_entry(ha, &(l)->list, list)
268
269#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
270#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
271#define netdev_for_each_uc_addr(ha, dev) \
272 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
273#define netdev_for_each_synced_uc_addr(_ha, _dev) \
274 netdev_for_each_uc_addr((_ha), (_dev)) \
275 if ((_ha)->sync_cnt)
276
277#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
278#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
279#define netdev_for_each_mc_addr(ha, dev) \
280 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
281#define netdev_for_each_synced_mc_addr(_ha, _dev) \
282 netdev_for_each_mc_addr((_ha), (_dev)) \
283 if ((_ha)->sync_cnt)
284
285struct hh_cache {
286 unsigned int hh_len;
287 seqlock_t hh_lock;
288
289 /* cached hardware header; allow for machine alignment needs. */
290#define HH_DATA_MOD 16
291#define HH_DATA_OFF(__len) \
292 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
293#define HH_DATA_ALIGN(__len) \
294 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
295 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
296};
297
298/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
299 * Alternative is:
300 * dev->hard_header_len ? (dev->hard_header_len +
301 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
302 *
303 * We could use other alignment values, but we must maintain the
304 * relationship HH alignment <= LL alignment.
305 */
306#define LL_RESERVED_SPACE(dev) \
307 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
308 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
309#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
310 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
311 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
312
313struct header_ops {
314 int (*create) (struct sk_buff *skb, struct net_device *dev,
315 unsigned short type, const void *daddr,
316 const void *saddr, unsigned int len);
317 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
318 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
319 void (*cache_update)(struct hh_cache *hh,
320 const struct net_device *dev,
321 const unsigned char *haddr);
322 bool (*validate)(const char *ll_header, unsigned int len);
323 __be16 (*parse_protocol)(const struct sk_buff *skb);
324};
325
326/* These flag bits are private to the generic network queueing
327 * layer; they may not be explicitly referenced by any other
328 * code.
329 */
330
331enum netdev_state_t {
332 __LINK_STATE_START,
333 __LINK_STATE_PRESENT,
334 __LINK_STATE_NOCARRIER,
335 __LINK_STATE_LINKWATCH_PENDING,
336 __LINK_STATE_DORMANT,
337 __LINK_STATE_TESTING,
338};
339
340struct gro_list {
341 struct list_head list;
342 int count;
343};
344
345/*
346 * size of gro hash buckets, must less than bit number of
347 * napi_struct::gro_bitmask
348 */
349#define GRO_HASH_BUCKETS 8
350
351/*
352 * Structure for NAPI scheduling similar to tasklet but with weighting
353 */
354struct napi_struct {
355 /* The poll_list must only be managed by the entity which
356 * changes the state of the NAPI_STATE_SCHED bit. This means
357 * whoever atomically sets that bit can add this napi_struct
358 * to the per-CPU poll_list, and whoever clears that bit
359 * can remove from the list right before clearing the bit.
360 */
361 struct list_head poll_list;
362
363 unsigned long state;
364 int weight;
365 int defer_hard_irqs_count;
366 unsigned long gro_bitmask;
367 int (*poll)(struct napi_struct *, int);
368#ifdef CONFIG_NETPOLL
369 /* CPU actively polling if netpoll is configured */
370 int poll_owner;
371#endif
372 /* CPU on which NAPI has been scheduled for processing */
373 int list_owner;
374 struct net_device *dev;
375 struct gro_list gro_hash[GRO_HASH_BUCKETS];
376 struct sk_buff *skb;
377 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
378 int rx_count; /* length of rx_list */
379 unsigned int napi_id;
380 struct hrtimer timer;
381 struct task_struct *thread;
382 /* control-path-only fields follow */
383 struct list_head dev_list;
384 struct hlist_node napi_hash_node;
385};
386
387enum {
388 NAPI_STATE_SCHED, /* Poll is scheduled */
389 NAPI_STATE_MISSED, /* reschedule a napi */
390 NAPI_STATE_DISABLE, /* Disable pending */
391 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
392 NAPI_STATE_LISTED, /* NAPI added to system lists */
393 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
394 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
395 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
396 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
397 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
398};
399
400enum {
401 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
402 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
403 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
404 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
405 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
406 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
407 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
408 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
409 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
410 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
411};
412
413enum gro_result {
414 GRO_MERGED,
415 GRO_MERGED_FREE,
416 GRO_HELD,
417 GRO_NORMAL,
418 GRO_CONSUMED,
419};
420typedef enum gro_result gro_result_t;
421
422/*
423 * enum rx_handler_result - Possible return values for rx_handlers.
424 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
425 * further.
426 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
427 * case skb->dev was changed by rx_handler.
428 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
429 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
430 *
431 * rx_handlers are functions called from inside __netif_receive_skb(), to do
432 * special processing of the skb, prior to delivery to protocol handlers.
433 *
434 * Currently, a net_device can only have a single rx_handler registered. Trying
435 * to register a second rx_handler will return -EBUSY.
436 *
437 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
438 * To unregister a rx_handler on a net_device, use
439 * netdev_rx_handler_unregister().
440 *
441 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
442 * do with the skb.
443 *
444 * If the rx_handler consumed the skb in some way, it should return
445 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
446 * the skb to be delivered in some other way.
447 *
448 * If the rx_handler changed skb->dev, to divert the skb to another
449 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
450 * new device will be called if it exists.
451 *
452 * If the rx_handler decides the skb should be ignored, it should return
453 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
454 * are registered on exact device (ptype->dev == skb->dev).
455 *
456 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
457 * delivered, it should return RX_HANDLER_PASS.
458 *
459 * A device without a registered rx_handler will behave as if rx_handler
460 * returned RX_HANDLER_PASS.
461 */
462
463enum rx_handler_result {
464 RX_HANDLER_CONSUMED,
465 RX_HANDLER_ANOTHER,
466 RX_HANDLER_EXACT,
467 RX_HANDLER_PASS,
468};
469typedef enum rx_handler_result rx_handler_result_t;
470typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
471
472void __napi_schedule(struct napi_struct *n);
473void __napi_schedule_irqoff(struct napi_struct *n);
474
475static inline bool napi_disable_pending(struct napi_struct *n)
476{
477 return test_bit(NAPI_STATE_DISABLE, &n->state);
478}
479
480static inline bool napi_prefer_busy_poll(struct napi_struct *n)
481{
482 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
483}
484
485/**
486 * napi_is_scheduled - test if NAPI is scheduled
487 * @n: NAPI context
488 *
489 * This check is "best-effort". With no locking implemented,
490 * a NAPI can be scheduled or terminate right after this check
491 * and produce not precise results.
492 *
493 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
494 * should not be used normally and napi_schedule should be
495 * used instead.
496 *
497 * Use only if the driver really needs to check if a NAPI
498 * is scheduled for example in the context of delayed timer
499 * that can be skipped if a NAPI is already scheduled.
500 *
501 * Return True if NAPI is scheduled, False otherwise.
502 */
503static inline bool napi_is_scheduled(struct napi_struct *n)
504{
505 return test_bit(NAPI_STATE_SCHED, &n->state);
506}
507
508bool napi_schedule_prep(struct napi_struct *n);
509
510/**
511 * napi_schedule - schedule NAPI poll
512 * @n: NAPI context
513 *
514 * Schedule NAPI poll routine to be called if it is not already
515 * running.
516 * Return true if we schedule a NAPI or false if not.
517 * Refer to napi_schedule_prep() for additional reason on why
518 * a NAPI might not be scheduled.
519 */
520static inline bool napi_schedule(struct napi_struct *n)
521{
522 if (napi_schedule_prep(n)) {
523 __napi_schedule(n);
524 return true;
525 }
526
527 return false;
528}
529
530/**
531 * napi_schedule_irqoff - schedule NAPI poll
532 * @n: NAPI context
533 *
534 * Variant of napi_schedule(), assuming hard irqs are masked.
535 */
536static inline void napi_schedule_irqoff(struct napi_struct *n)
537{
538 if (napi_schedule_prep(n))
539 __napi_schedule_irqoff(n);
540}
541
542/**
543 * napi_complete_done - NAPI processing complete
544 * @n: NAPI context
545 * @work_done: number of packets processed
546 *
547 * Mark NAPI processing as complete. Should only be called if poll budget
548 * has not been completely consumed.
549 * Prefer over napi_complete().
550 * Return false if device should avoid rearming interrupts.
551 */
552bool napi_complete_done(struct napi_struct *n, int work_done);
553
554static inline bool napi_complete(struct napi_struct *n)
555{
556 return napi_complete_done(n, 0);
557}
558
559int dev_set_threaded(struct net_device *dev, bool threaded);
560
561/**
562 * napi_disable - prevent NAPI from scheduling
563 * @n: NAPI context
564 *
565 * Stop NAPI from being scheduled on this context.
566 * Waits till any outstanding processing completes.
567 */
568void napi_disable(struct napi_struct *n);
569
570void napi_enable(struct napi_struct *n);
571
572/**
573 * napi_synchronize - wait until NAPI is not running
574 * @n: NAPI context
575 *
576 * Wait until NAPI is done being scheduled on this context.
577 * Waits till any outstanding processing completes but
578 * does not disable future activations.
579 */
580static inline void napi_synchronize(const struct napi_struct *n)
581{
582 if (IS_ENABLED(CONFIG_SMP))
583 while (test_bit(NAPI_STATE_SCHED, &n->state))
584 msleep(1);
585 else
586 barrier();
587}
588
589/**
590 * napi_if_scheduled_mark_missed - if napi is running, set the
591 * NAPIF_STATE_MISSED
592 * @n: NAPI context
593 *
594 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
595 * NAPI is scheduled.
596 **/
597static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
598{
599 unsigned long val, new;
600
601 val = READ_ONCE(n->state);
602 do {
603 if (val & NAPIF_STATE_DISABLE)
604 return true;
605
606 if (!(val & NAPIF_STATE_SCHED))
607 return false;
608
609 new = val | NAPIF_STATE_MISSED;
610 } while (!try_cmpxchg(&n->state, &val, new));
611
612 return true;
613}
614
615enum netdev_queue_state_t {
616 __QUEUE_STATE_DRV_XOFF,
617 __QUEUE_STATE_STACK_XOFF,
618 __QUEUE_STATE_FROZEN,
619};
620
621#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
622#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
623#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
624
625#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
626#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
627 QUEUE_STATE_FROZEN)
628#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
629 QUEUE_STATE_FROZEN)
630
631/*
632 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
633 * netif_tx_* functions below are used to manipulate this flag. The
634 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
635 * queue independently. The netif_xmit_*stopped functions below are called
636 * to check if the queue has been stopped by the driver or stack (either
637 * of the XOFF bits are set in the state). Drivers should not need to call
638 * netif_xmit*stopped functions, they should only be using netif_tx_*.
639 */
640
641struct netdev_queue {
642/*
643 * read-mostly part
644 */
645 struct net_device *dev;
646 netdevice_tracker dev_tracker;
647
648 struct Qdisc __rcu *qdisc;
649 struct Qdisc __rcu *qdisc_sleeping;
650#ifdef CONFIG_SYSFS
651 struct kobject kobj;
652#endif
653#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
654 int numa_node;
655#endif
656 unsigned long tx_maxrate;
657 /*
658 * Number of TX timeouts for this queue
659 * (/sys/class/net/DEV/Q/trans_timeout)
660 */
661 atomic_long_t trans_timeout;
662
663 /* Subordinate device that the queue has been assigned to */
664 struct net_device *sb_dev;
665#ifdef CONFIG_XDP_SOCKETS
666 struct xsk_buff_pool *pool;
667#endif
668/*
669 * write-mostly part
670 */
671 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
672 int xmit_lock_owner;
673 /*
674 * Time (in jiffies) of last Tx
675 */
676 unsigned long trans_start;
677
678 unsigned long state;
679
680#ifdef CONFIG_BQL
681 struct dql dql;
682#endif
683} ____cacheline_aligned_in_smp;
684
685extern int sysctl_fb_tunnels_only_for_init_net;
686extern int sysctl_devconf_inherit_init_net;
687
688/*
689 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
690 * == 1 : For initns only
691 * == 2 : For none.
692 */
693static inline bool net_has_fallback_tunnels(const struct net *net)
694{
695#if IS_ENABLED(CONFIG_SYSCTL)
696 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
697
698 return !fb_tunnels_only_for_init_net ||
699 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
700#else
701 return true;
702#endif
703}
704
705static inline int net_inherit_devconf(void)
706{
707#if IS_ENABLED(CONFIG_SYSCTL)
708 return READ_ONCE(sysctl_devconf_inherit_init_net);
709#else
710 return 0;
711#endif
712}
713
714static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
715{
716#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
717 return q->numa_node;
718#else
719 return NUMA_NO_NODE;
720#endif
721}
722
723static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
724{
725#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
726 q->numa_node = node;
727#endif
728}
729
730#ifdef CONFIG_RPS
731/*
732 * This structure holds an RPS map which can be of variable length. The
733 * map is an array of CPUs.
734 */
735struct rps_map {
736 unsigned int len;
737 struct rcu_head rcu;
738 u16 cpus[];
739};
740#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
741
742/*
743 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
744 * tail pointer for that CPU's input queue at the time of last enqueue, and
745 * a hardware filter index.
746 */
747struct rps_dev_flow {
748 u16 cpu;
749 u16 filter;
750 unsigned int last_qtail;
751};
752#define RPS_NO_FILTER 0xffff
753
754/*
755 * The rps_dev_flow_table structure contains a table of flow mappings.
756 */
757struct rps_dev_flow_table {
758 unsigned int mask;
759 struct rcu_head rcu;
760 struct rps_dev_flow flows[];
761};
762#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
763 ((_num) * sizeof(struct rps_dev_flow)))
764
765/*
766 * The rps_sock_flow_table contains mappings of flows to the last CPU
767 * on which they were processed by the application (set in recvmsg).
768 * Each entry is a 32bit value. Upper part is the high-order bits
769 * of flow hash, lower part is CPU number.
770 * rps_cpu_mask is used to partition the space, depending on number of
771 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
772 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
773 * meaning we use 32-6=26 bits for the hash.
774 */
775struct rps_sock_flow_table {
776 u32 mask;
777
778 u32 ents[] ____cacheline_aligned_in_smp;
779};
780#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
781
782#define RPS_NO_CPU 0xffff
783
784extern u32 rps_cpu_mask;
785extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
786
787static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
788 u32 hash)
789{
790 if (table && hash) {
791 unsigned int index = hash & table->mask;
792 u32 val = hash & ~rps_cpu_mask;
793
794 /* We only give a hint, preemption can change CPU under us */
795 val |= raw_smp_processor_id();
796
797 /* The following WRITE_ONCE() is paired with the READ_ONCE()
798 * here, and another one in get_rps_cpu().
799 */
800 if (READ_ONCE(table->ents[index]) != val)
801 WRITE_ONCE(table->ents[index], val);
802 }
803}
804
805#ifdef CONFIG_RFS_ACCEL
806bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
807 u16 filter_id);
808#endif
809#endif /* CONFIG_RPS */
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 u8 amsdu;
943 } mtk_wdma;
944 };
945};
946
947#define NET_DEVICE_PATH_STACK_MAX 5
948#define NET_DEVICE_PATH_VLAN_MAX 2
949
950struct net_device_path_stack {
951 int num_paths;
952 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
953};
954
955struct net_device_path_ctx {
956 const struct net_device *dev;
957 u8 daddr[ETH_ALEN];
958
959 int num_vlans;
960 struct {
961 u16 id;
962 __be16 proto;
963 } vlan[NET_DEVICE_PATH_VLAN_MAX];
964};
965
966enum tc_setup_type {
967 TC_QUERY_CAPS,
968 TC_SETUP_QDISC_MQPRIO,
969 TC_SETUP_CLSU32,
970 TC_SETUP_CLSFLOWER,
971 TC_SETUP_CLSMATCHALL,
972 TC_SETUP_CLSBPF,
973 TC_SETUP_BLOCK,
974 TC_SETUP_QDISC_CBS,
975 TC_SETUP_QDISC_RED,
976 TC_SETUP_QDISC_PRIO,
977 TC_SETUP_QDISC_MQ,
978 TC_SETUP_QDISC_ETF,
979 TC_SETUP_ROOT_QDISC,
980 TC_SETUP_QDISC_GRED,
981 TC_SETUP_QDISC_TAPRIO,
982 TC_SETUP_FT,
983 TC_SETUP_QDISC_ETS,
984 TC_SETUP_QDISC_TBF,
985 TC_SETUP_QDISC_FIFO,
986 TC_SETUP_QDISC_HTB,
987 TC_SETUP_ACT,
988};
989
990/* These structures hold the attributes of bpf state that are being passed
991 * to the netdevice through the bpf op.
992 */
993enum bpf_netdev_command {
994 /* Set or clear a bpf program used in the earliest stages of packet
995 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
996 * is responsible for calling bpf_prog_put on any old progs that are
997 * stored. In case of error, the callee need not release the new prog
998 * reference, but on success it takes ownership and must bpf_prog_put
999 * when it is no longer used.
1000 */
1001 XDP_SETUP_PROG,
1002 XDP_SETUP_PROG_HW,
1003 /* BPF program for offload callbacks, invoked at program load time. */
1004 BPF_OFFLOAD_MAP_ALLOC,
1005 BPF_OFFLOAD_MAP_FREE,
1006 XDP_SETUP_XSK_POOL,
1007};
1008
1009struct bpf_prog_offload_ops;
1010struct netlink_ext_ack;
1011struct xdp_umem;
1012struct xdp_dev_bulk_queue;
1013struct bpf_xdp_link;
1014
1015enum bpf_xdp_mode {
1016 XDP_MODE_SKB = 0,
1017 XDP_MODE_DRV = 1,
1018 XDP_MODE_HW = 2,
1019 __MAX_XDP_MODE
1020};
1021
1022struct bpf_xdp_entity {
1023 struct bpf_prog *prog;
1024 struct bpf_xdp_link *link;
1025};
1026
1027struct netdev_bpf {
1028 enum bpf_netdev_command command;
1029 union {
1030 /* XDP_SETUP_PROG */
1031 struct {
1032 u32 flags;
1033 struct bpf_prog *prog;
1034 struct netlink_ext_ack *extack;
1035 };
1036 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1037 struct {
1038 struct bpf_offloaded_map *offmap;
1039 };
1040 /* XDP_SETUP_XSK_POOL */
1041 struct {
1042 struct xsk_buff_pool *pool;
1043 u16 queue_id;
1044 } xsk;
1045 };
1046};
1047
1048/* Flags for ndo_xsk_wakeup. */
1049#define XDP_WAKEUP_RX (1 << 0)
1050#define XDP_WAKEUP_TX (1 << 1)
1051
1052#ifdef CONFIG_XFRM_OFFLOAD
1053struct xfrmdev_ops {
1054 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1055 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1056 void (*xdo_dev_state_free) (struct xfrm_state *x);
1057 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1058 struct xfrm_state *x);
1059 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1060 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1061 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1062 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1063 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1064};
1065#endif
1066
1067struct dev_ifalias {
1068 struct rcu_head rcuhead;
1069 char ifalias[];
1070};
1071
1072struct devlink;
1073struct tlsdev_ops;
1074
1075struct netdev_net_notifier {
1076 struct list_head list;
1077 struct notifier_block *nb;
1078};
1079
1080/*
1081 * This structure defines the management hooks for network devices.
1082 * The following hooks can be defined; unless noted otherwise, they are
1083 * optional and can be filled with a null pointer.
1084 *
1085 * int (*ndo_init)(struct net_device *dev);
1086 * This function is called once when a network device is registered.
1087 * The network device can use this for any late stage initialization
1088 * or semantic validation. It can fail with an error code which will
1089 * be propagated back to register_netdev.
1090 *
1091 * void (*ndo_uninit)(struct net_device *dev);
1092 * This function is called when device is unregistered or when registration
1093 * fails. It is not called if init fails.
1094 *
1095 * int (*ndo_open)(struct net_device *dev);
1096 * This function is called when a network device transitions to the up
1097 * state.
1098 *
1099 * int (*ndo_stop)(struct net_device *dev);
1100 * This function is called when a network device transitions to the down
1101 * state.
1102 *
1103 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1104 * struct net_device *dev);
1105 * Called when a packet needs to be transmitted.
1106 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1107 * the queue before that can happen; it's for obsolete devices and weird
1108 * corner cases, but the stack really does a non-trivial amount
1109 * of useless work if you return NETDEV_TX_BUSY.
1110 * Required; cannot be NULL.
1111 *
1112 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1113 * struct net_device *dev
1114 * netdev_features_t features);
1115 * Called by core transmit path to determine if device is capable of
1116 * performing offload operations on a given packet. This is to give
1117 * the device an opportunity to implement any restrictions that cannot
1118 * be otherwise expressed by feature flags. The check is called with
1119 * the set of features that the stack has calculated and it returns
1120 * those the driver believes to be appropriate.
1121 *
1122 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1123 * struct net_device *sb_dev);
1124 * Called to decide which queue to use when device supports multiple
1125 * transmit queues.
1126 *
1127 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1128 * This function is called to allow device receiver to make
1129 * changes to configuration when multicast or promiscuous is enabled.
1130 *
1131 * void (*ndo_set_rx_mode)(struct net_device *dev);
1132 * This function is called device changes address list filtering.
1133 * If driver handles unicast address filtering, it should set
1134 * IFF_UNICAST_FLT in its priv_flags.
1135 *
1136 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1137 * This function is called when the Media Access Control address
1138 * needs to be changed. If this interface is not defined, the
1139 * MAC address can not be changed.
1140 *
1141 * int (*ndo_validate_addr)(struct net_device *dev);
1142 * Test if Media Access Control address is valid for the device.
1143 *
1144 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1145 * Old-style ioctl entry point. This is used internally by the
1146 * appletalk and ieee802154 subsystems but is no longer called by
1147 * the device ioctl handler.
1148 *
1149 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1150 * Used by the bonding driver for its device specific ioctls:
1151 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1152 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1153 *
1154 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1155 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1156 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1157 *
1158 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1159 * Used to set network devices bus interface parameters. This interface
1160 * is retained for legacy reasons; new devices should use the bus
1161 * interface (PCI) for low level management.
1162 *
1163 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1164 * Called when a user wants to change the Maximum Transfer Unit
1165 * of a device.
1166 *
1167 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1168 * Callback used when the transmitter has not made any progress
1169 * for dev->watchdog ticks.
1170 *
1171 * void (*ndo_get_stats64)(struct net_device *dev,
1172 * struct rtnl_link_stats64 *storage);
1173 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1174 * Called when a user wants to get the network device usage
1175 * statistics. Drivers must do one of the following:
1176 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1177 * rtnl_link_stats64 structure passed by the caller.
1178 * 2. Define @ndo_get_stats to update a net_device_stats structure
1179 * (which should normally be dev->stats) and return a pointer to
1180 * it. The structure may be changed asynchronously only if each
1181 * field is written atomically.
1182 * 3. Update dev->stats asynchronously and atomically, and define
1183 * neither operation.
1184 *
1185 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1186 * Return true if this device supports offload stats of this attr_id.
1187 *
1188 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1189 * void *attr_data)
1190 * Get statistics for offload operations by attr_id. Write it into the
1191 * attr_data pointer.
1192 *
1193 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1194 * If device supports VLAN filtering this function is called when a
1195 * VLAN id is registered.
1196 *
1197 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1198 * If device supports VLAN filtering this function is called when a
1199 * VLAN id is unregistered.
1200 *
1201 * void (*ndo_poll_controller)(struct net_device *dev);
1202 *
1203 * SR-IOV management functions.
1204 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1205 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1206 * u8 qos, __be16 proto);
1207 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1208 * int max_tx_rate);
1209 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1210 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1211 * int (*ndo_get_vf_config)(struct net_device *dev,
1212 * int vf, struct ifla_vf_info *ivf);
1213 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1214 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1215 * struct nlattr *port[]);
1216 *
1217 * Enable or disable the VF ability to query its RSS Redirection Table and
1218 * Hash Key. This is needed since on some devices VF share this information
1219 * with PF and querying it may introduce a theoretical security risk.
1220 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1221 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1222 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1223 * void *type_data);
1224 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1225 * This is always called from the stack with the rtnl lock held and netif
1226 * tx queues stopped. This allows the netdevice to perform queue
1227 * management safely.
1228 *
1229 * Fiber Channel over Ethernet (FCoE) offload functions.
1230 * int (*ndo_fcoe_enable)(struct net_device *dev);
1231 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1232 * so the underlying device can perform whatever needed configuration or
1233 * initialization to support acceleration of FCoE traffic.
1234 *
1235 * int (*ndo_fcoe_disable)(struct net_device *dev);
1236 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1237 * so the underlying device can perform whatever needed clean-ups to
1238 * stop supporting acceleration of FCoE traffic.
1239 *
1240 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1241 * struct scatterlist *sgl, unsigned int sgc);
1242 * Called when the FCoE Initiator wants to initialize an I/O that
1243 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1244 * perform necessary setup and returns 1 to indicate the device is set up
1245 * successfully to perform DDP on this I/O, otherwise this returns 0.
1246 *
1247 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1248 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1249 * indicated by the FC exchange id 'xid', so the underlying device can
1250 * clean up and reuse resources for later DDP requests.
1251 *
1252 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1253 * struct scatterlist *sgl, unsigned int sgc);
1254 * Called when the FCoE Target wants to initialize an I/O that
1255 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1256 * perform necessary setup and returns 1 to indicate the device is set up
1257 * successfully to perform DDP on this I/O, otherwise this returns 0.
1258 *
1259 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1260 * struct netdev_fcoe_hbainfo *hbainfo);
1261 * Called when the FCoE Protocol stack wants information on the underlying
1262 * device. This information is utilized by the FCoE protocol stack to
1263 * register attributes with Fiber Channel management service as per the
1264 * FC-GS Fabric Device Management Information(FDMI) specification.
1265 *
1266 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1267 * Called when the underlying device wants to override default World Wide
1268 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1269 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1270 * protocol stack to use.
1271 *
1272 * RFS acceleration.
1273 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1274 * u16 rxq_index, u32 flow_id);
1275 * Set hardware filter for RFS. rxq_index is the target queue index;
1276 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1277 * Return the filter ID on success, or a negative error code.
1278 *
1279 * Slave management functions (for bridge, bonding, etc).
1280 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1281 * Called to make another netdev an underling.
1282 *
1283 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1284 * Called to release previously enslaved netdev.
1285 *
1286 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1287 * struct sk_buff *skb,
1288 * bool all_slaves);
1289 * Get the xmit slave of master device. If all_slaves is true, function
1290 * assume all the slaves can transmit.
1291 *
1292 * Feature/offload setting functions.
1293 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1294 * netdev_features_t features);
1295 * Adjusts the requested feature flags according to device-specific
1296 * constraints, and returns the resulting flags. Must not modify
1297 * the device state.
1298 *
1299 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1300 * Called to update device configuration to new features. Passed
1301 * feature set might be less than what was returned by ndo_fix_features()).
1302 * Must return >0 or -errno if it changed dev->features itself.
1303 *
1304 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1305 * struct net_device *dev,
1306 * const unsigned char *addr, u16 vid, u16 flags,
1307 * struct netlink_ext_ack *extack);
1308 * Adds an FDB entry to dev for addr.
1309 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1310 * struct net_device *dev,
1311 * const unsigned char *addr, u16 vid)
1312 * Deletes the FDB entry from dev coresponding to addr.
1313 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1314 * struct netlink_ext_ack *extack);
1315 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1316 * struct net_device *dev, struct net_device *filter_dev,
1317 * int *idx)
1318 * Used to add FDB entries to dump requests. Implementers should add
1319 * entries to skb and update idx with the number of entries.
1320 *
1321 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1322 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1323 * Adds an MDB entry to dev.
1324 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1325 * struct netlink_ext_ack *extack);
1326 * Deletes the MDB entry from dev.
1327 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1328 * struct netlink_callback *cb);
1329 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1330 * callback is used by core rtnetlink code.
1331 *
1332 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1333 * u16 flags, struct netlink_ext_ack *extack)
1334 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1335 * struct net_device *dev, u32 filter_mask,
1336 * int nlflags)
1337 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1338 * u16 flags);
1339 *
1340 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1341 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1342 * which do not represent real hardware may define this to allow their
1343 * userspace components to manage their virtual carrier state. Devices
1344 * that determine carrier state from physical hardware properties (eg
1345 * network cables) or protocol-dependent mechanisms (eg
1346 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1347 *
1348 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1349 * struct netdev_phys_item_id *ppid);
1350 * Called to get ID of physical port of this device. If driver does
1351 * not implement this, it is assumed that the hw is not able to have
1352 * multiple net devices on single physical port.
1353 *
1354 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1355 * struct netdev_phys_item_id *ppid)
1356 * Called to get the parent ID of the physical port of this device.
1357 *
1358 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1359 * struct net_device *dev)
1360 * Called by upper layer devices to accelerate switching or other
1361 * station functionality into hardware. 'pdev is the lowerdev
1362 * to use for the offload and 'dev' is the net device that will
1363 * back the offload. Returns a pointer to the private structure
1364 * the upper layer will maintain.
1365 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1366 * Called by upper layer device to delete the station created
1367 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1368 * the station and priv is the structure returned by the add
1369 * operation.
1370 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1371 * int queue_index, u32 maxrate);
1372 * Called when a user wants to set a max-rate limitation of specific
1373 * TX queue.
1374 * int (*ndo_get_iflink)(const struct net_device *dev);
1375 * Called to get the iflink value of this device.
1376 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1377 * This function is used to get egress tunnel information for given skb.
1378 * This is useful for retrieving outer tunnel header parameters while
1379 * sampling packet.
1380 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1381 * This function is used to specify the headroom that the skb must
1382 * consider when allocation skb during packet reception. Setting
1383 * appropriate rx headroom value allows avoiding skb head copy on
1384 * forward. Setting a negative value resets the rx headroom to the
1385 * default value.
1386 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1387 * This function is used to set or query state related to XDP on the
1388 * netdevice and manage BPF offload. See definition of
1389 * enum bpf_netdev_command for details.
1390 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1391 * u32 flags);
1392 * This function is used to submit @n XDP packets for transmit on a
1393 * netdevice. Returns number of frames successfully transmitted, frames
1394 * that got dropped are freed/returned via xdp_return_frame().
1395 * Returns negative number, means general error invoking ndo, meaning
1396 * no frames were xmit'ed and core-caller will free all frames.
1397 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1398 * struct xdp_buff *xdp);
1399 * Get the xmit slave of master device based on the xdp_buff.
1400 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1401 * This function is used to wake up the softirq, ksoftirqd or kthread
1402 * responsible for sending and/or receiving packets on a specific
1403 * queue id bound to an AF_XDP socket. The flags field specifies if
1404 * only RX, only Tx, or both should be woken up using the flags
1405 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1406 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1407 * int cmd);
1408 * Add, change, delete or get information on an IPv4 tunnel.
1409 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1410 * If a device is paired with a peer device, return the peer instance.
1411 * The caller must be under RCU read context.
1412 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1413 * Get the forwarding path to reach the real device from the HW destination address
1414 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1415 * const struct skb_shared_hwtstamps *hwtstamps,
1416 * bool cycles);
1417 * Get hardware timestamp based on normal/adjustable time or free running
1418 * cycle counter. This function is required if physical clock supports a
1419 * free running cycle counter.
1420 *
1421 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1422 * struct kernel_hwtstamp_config *kernel_config);
1423 * Get the currently configured hardware timestamping parameters for the
1424 * NIC device.
1425 *
1426 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1427 * struct kernel_hwtstamp_config *kernel_config,
1428 * struct netlink_ext_ack *extack);
1429 * Change the hardware timestamping parameters for NIC device.
1430 */
1431struct net_device_ops {
1432 int (*ndo_init)(struct net_device *dev);
1433 void (*ndo_uninit)(struct net_device *dev);
1434 int (*ndo_open)(struct net_device *dev);
1435 int (*ndo_stop)(struct net_device *dev);
1436 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1437 struct net_device *dev);
1438 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1439 struct net_device *dev,
1440 netdev_features_t features);
1441 u16 (*ndo_select_queue)(struct net_device *dev,
1442 struct sk_buff *skb,
1443 struct net_device *sb_dev);
1444 void (*ndo_change_rx_flags)(struct net_device *dev,
1445 int flags);
1446 void (*ndo_set_rx_mode)(struct net_device *dev);
1447 int (*ndo_set_mac_address)(struct net_device *dev,
1448 void *addr);
1449 int (*ndo_validate_addr)(struct net_device *dev);
1450 int (*ndo_do_ioctl)(struct net_device *dev,
1451 struct ifreq *ifr, int cmd);
1452 int (*ndo_eth_ioctl)(struct net_device *dev,
1453 struct ifreq *ifr, int cmd);
1454 int (*ndo_siocbond)(struct net_device *dev,
1455 struct ifreq *ifr, int cmd);
1456 int (*ndo_siocwandev)(struct net_device *dev,
1457 struct if_settings *ifs);
1458 int (*ndo_siocdevprivate)(struct net_device *dev,
1459 struct ifreq *ifr,
1460 void __user *data, int cmd);
1461 int (*ndo_set_config)(struct net_device *dev,
1462 struct ifmap *map);
1463 int (*ndo_change_mtu)(struct net_device *dev,
1464 int new_mtu);
1465 int (*ndo_neigh_setup)(struct net_device *dev,
1466 struct neigh_parms *);
1467 void (*ndo_tx_timeout) (struct net_device *dev,
1468 unsigned int txqueue);
1469
1470 void (*ndo_get_stats64)(struct net_device *dev,
1471 struct rtnl_link_stats64 *storage);
1472 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1473 int (*ndo_get_offload_stats)(int attr_id,
1474 const struct net_device *dev,
1475 void *attr_data);
1476 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1477
1478 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1479 __be16 proto, u16 vid);
1480 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1481 __be16 proto, u16 vid);
1482#ifdef CONFIG_NET_POLL_CONTROLLER
1483 void (*ndo_poll_controller)(struct net_device *dev);
1484 int (*ndo_netpoll_setup)(struct net_device *dev,
1485 struct netpoll_info *info);
1486 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1487#endif
1488 int (*ndo_set_vf_mac)(struct net_device *dev,
1489 int queue, u8 *mac);
1490 int (*ndo_set_vf_vlan)(struct net_device *dev,
1491 int queue, u16 vlan,
1492 u8 qos, __be16 proto);
1493 int (*ndo_set_vf_rate)(struct net_device *dev,
1494 int vf, int min_tx_rate,
1495 int max_tx_rate);
1496 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1497 int vf, bool setting);
1498 int (*ndo_set_vf_trust)(struct net_device *dev,
1499 int vf, bool setting);
1500 int (*ndo_get_vf_config)(struct net_device *dev,
1501 int vf,
1502 struct ifla_vf_info *ivf);
1503 int (*ndo_set_vf_link_state)(struct net_device *dev,
1504 int vf, int link_state);
1505 int (*ndo_get_vf_stats)(struct net_device *dev,
1506 int vf,
1507 struct ifla_vf_stats
1508 *vf_stats);
1509 int (*ndo_set_vf_port)(struct net_device *dev,
1510 int vf,
1511 struct nlattr *port[]);
1512 int (*ndo_get_vf_port)(struct net_device *dev,
1513 int vf, struct sk_buff *skb);
1514 int (*ndo_get_vf_guid)(struct net_device *dev,
1515 int vf,
1516 struct ifla_vf_guid *node_guid,
1517 struct ifla_vf_guid *port_guid);
1518 int (*ndo_set_vf_guid)(struct net_device *dev,
1519 int vf, u64 guid,
1520 int guid_type);
1521 int (*ndo_set_vf_rss_query_en)(
1522 struct net_device *dev,
1523 int vf, bool setting);
1524 int (*ndo_setup_tc)(struct net_device *dev,
1525 enum tc_setup_type type,
1526 void *type_data);
1527#if IS_ENABLED(CONFIG_FCOE)
1528 int (*ndo_fcoe_enable)(struct net_device *dev);
1529 int (*ndo_fcoe_disable)(struct net_device *dev);
1530 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1531 u16 xid,
1532 struct scatterlist *sgl,
1533 unsigned int sgc);
1534 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1535 u16 xid);
1536 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1537 u16 xid,
1538 struct scatterlist *sgl,
1539 unsigned int sgc);
1540 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1541 struct netdev_fcoe_hbainfo *hbainfo);
1542#endif
1543
1544#if IS_ENABLED(CONFIG_LIBFCOE)
1545#define NETDEV_FCOE_WWNN 0
1546#define NETDEV_FCOE_WWPN 1
1547 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1548 u64 *wwn, int type);
1549#endif
1550
1551#ifdef CONFIG_RFS_ACCEL
1552 int (*ndo_rx_flow_steer)(struct net_device *dev,
1553 const struct sk_buff *skb,
1554 u16 rxq_index,
1555 u32 flow_id);
1556#endif
1557 int (*ndo_add_slave)(struct net_device *dev,
1558 struct net_device *slave_dev,
1559 struct netlink_ext_ack *extack);
1560 int (*ndo_del_slave)(struct net_device *dev,
1561 struct net_device *slave_dev);
1562 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1563 struct sk_buff *skb,
1564 bool all_slaves);
1565 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1566 struct sock *sk);
1567 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1568 netdev_features_t features);
1569 int (*ndo_set_features)(struct net_device *dev,
1570 netdev_features_t features);
1571 int (*ndo_neigh_construct)(struct net_device *dev,
1572 struct neighbour *n);
1573 void (*ndo_neigh_destroy)(struct net_device *dev,
1574 struct neighbour *n);
1575
1576 int (*ndo_fdb_add)(struct ndmsg *ndm,
1577 struct nlattr *tb[],
1578 struct net_device *dev,
1579 const unsigned char *addr,
1580 u16 vid,
1581 u16 flags,
1582 struct netlink_ext_ack *extack);
1583 int (*ndo_fdb_del)(struct ndmsg *ndm,
1584 struct nlattr *tb[],
1585 struct net_device *dev,
1586 const unsigned char *addr,
1587 u16 vid, struct netlink_ext_ack *extack);
1588 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1589 struct net_device *dev,
1590 struct netlink_ext_ack *extack);
1591 int (*ndo_fdb_dump)(struct sk_buff *skb,
1592 struct netlink_callback *cb,
1593 struct net_device *dev,
1594 struct net_device *filter_dev,
1595 int *idx);
1596 int (*ndo_fdb_get)(struct sk_buff *skb,
1597 struct nlattr *tb[],
1598 struct net_device *dev,
1599 const unsigned char *addr,
1600 u16 vid, u32 portid, u32 seq,
1601 struct netlink_ext_ack *extack);
1602 int (*ndo_mdb_add)(struct net_device *dev,
1603 struct nlattr *tb[],
1604 u16 nlmsg_flags,
1605 struct netlink_ext_ack *extack);
1606 int (*ndo_mdb_del)(struct net_device *dev,
1607 struct nlattr *tb[],
1608 struct netlink_ext_ack *extack);
1609 int (*ndo_mdb_dump)(struct net_device *dev,
1610 struct sk_buff *skb,
1611 struct netlink_callback *cb);
1612 int (*ndo_mdb_get)(struct net_device *dev,
1613 struct nlattr *tb[], u32 portid,
1614 u32 seq,
1615 struct netlink_ext_ack *extack);
1616 int (*ndo_bridge_setlink)(struct net_device *dev,
1617 struct nlmsghdr *nlh,
1618 u16 flags,
1619 struct netlink_ext_ack *extack);
1620 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1621 u32 pid, u32 seq,
1622 struct net_device *dev,
1623 u32 filter_mask,
1624 int nlflags);
1625 int (*ndo_bridge_dellink)(struct net_device *dev,
1626 struct nlmsghdr *nlh,
1627 u16 flags);
1628 int (*ndo_change_carrier)(struct net_device *dev,
1629 bool new_carrier);
1630 int (*ndo_get_phys_port_id)(struct net_device *dev,
1631 struct netdev_phys_item_id *ppid);
1632 int (*ndo_get_port_parent_id)(struct net_device *dev,
1633 struct netdev_phys_item_id *ppid);
1634 int (*ndo_get_phys_port_name)(struct net_device *dev,
1635 char *name, size_t len);
1636 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1637 struct net_device *dev);
1638 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1639 void *priv);
1640
1641 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1642 int queue_index,
1643 u32 maxrate);
1644 int (*ndo_get_iflink)(const struct net_device *dev);
1645 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1646 struct sk_buff *skb);
1647 void (*ndo_set_rx_headroom)(struct net_device *dev,
1648 int needed_headroom);
1649 int (*ndo_bpf)(struct net_device *dev,
1650 struct netdev_bpf *bpf);
1651 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1652 struct xdp_frame **xdp,
1653 u32 flags);
1654 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1655 struct xdp_buff *xdp);
1656 int (*ndo_xsk_wakeup)(struct net_device *dev,
1657 u32 queue_id, u32 flags);
1658 int (*ndo_tunnel_ctl)(struct net_device *dev,
1659 struct ip_tunnel_parm *p, int cmd);
1660 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1661 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1662 struct net_device_path *path);
1663 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1664 const struct skb_shared_hwtstamps *hwtstamps,
1665 bool cycles);
1666 int (*ndo_hwtstamp_get)(struct net_device *dev,
1667 struct kernel_hwtstamp_config *kernel_config);
1668 int (*ndo_hwtstamp_set)(struct net_device *dev,
1669 struct kernel_hwtstamp_config *kernel_config,
1670 struct netlink_ext_ack *extack);
1671};
1672
1673/**
1674 * enum netdev_priv_flags - &struct net_device priv_flags
1675 *
1676 * These are the &struct net_device, they are only set internally
1677 * by drivers and used in the kernel. These flags are invisible to
1678 * userspace; this means that the order of these flags can change
1679 * during any kernel release.
1680 *
1681 * You should have a pretty good reason to be extending these flags.
1682 *
1683 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1684 * @IFF_EBRIDGE: Ethernet bridging device
1685 * @IFF_BONDING: bonding master or slave
1686 * @IFF_ISATAP: ISATAP interface (RFC4214)
1687 * @IFF_WAN_HDLC: WAN HDLC device
1688 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1689 * release skb->dst
1690 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1691 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1692 * @IFF_MACVLAN_PORT: device used as macvlan port
1693 * @IFF_BRIDGE_PORT: device used as bridge port
1694 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1695 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1696 * @IFF_UNICAST_FLT: Supports unicast filtering
1697 * @IFF_TEAM_PORT: device used as team port
1698 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1699 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1700 * change when it's running
1701 * @IFF_MACVLAN: Macvlan device
1702 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1703 * underlying stacked devices
1704 * @IFF_L3MDEV_MASTER: device is an L3 master device
1705 * @IFF_NO_QUEUE: device can run without qdisc attached
1706 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1707 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1708 * @IFF_TEAM: device is a team device
1709 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1710 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1711 * entity (i.e. the master device for bridged veth)
1712 * @IFF_MACSEC: device is a MACsec device
1713 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1714 * @IFF_FAILOVER: device is a failover master device
1715 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1716 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1717 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1718 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1719 * skb_headlen(skb) == 0 (data starts from frag0)
1720 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1721 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1722 * ndo_hwtstamp_set() for all timestamp requests regardless of source,
1723 * even if those aren't HWTSTAMP_SOURCE_NETDEV.
1724 */
1725enum netdev_priv_flags {
1726 IFF_802_1Q_VLAN = 1<<0,
1727 IFF_EBRIDGE = 1<<1,
1728 IFF_BONDING = 1<<2,
1729 IFF_ISATAP = 1<<3,
1730 IFF_WAN_HDLC = 1<<4,
1731 IFF_XMIT_DST_RELEASE = 1<<5,
1732 IFF_DONT_BRIDGE = 1<<6,
1733 IFF_DISABLE_NETPOLL = 1<<7,
1734 IFF_MACVLAN_PORT = 1<<8,
1735 IFF_BRIDGE_PORT = 1<<9,
1736 IFF_OVS_DATAPATH = 1<<10,
1737 IFF_TX_SKB_SHARING = 1<<11,
1738 IFF_UNICAST_FLT = 1<<12,
1739 IFF_TEAM_PORT = 1<<13,
1740 IFF_SUPP_NOFCS = 1<<14,
1741 IFF_LIVE_ADDR_CHANGE = 1<<15,
1742 IFF_MACVLAN = 1<<16,
1743 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1744 IFF_L3MDEV_MASTER = 1<<18,
1745 IFF_NO_QUEUE = 1<<19,
1746 IFF_OPENVSWITCH = 1<<20,
1747 IFF_L3MDEV_SLAVE = 1<<21,
1748 IFF_TEAM = 1<<22,
1749 IFF_RXFH_CONFIGURED = 1<<23,
1750 IFF_PHONY_HEADROOM = 1<<24,
1751 IFF_MACSEC = 1<<25,
1752 IFF_NO_RX_HANDLER = 1<<26,
1753 IFF_FAILOVER = 1<<27,
1754 IFF_FAILOVER_SLAVE = 1<<28,
1755 IFF_L3MDEV_RX_HANDLER = 1<<29,
1756 IFF_NO_ADDRCONF = BIT_ULL(30),
1757 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1758 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1759 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33),
1760};
1761
1762#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1763#define IFF_EBRIDGE IFF_EBRIDGE
1764#define IFF_BONDING IFF_BONDING
1765#define IFF_ISATAP IFF_ISATAP
1766#define IFF_WAN_HDLC IFF_WAN_HDLC
1767#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1768#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1769#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1770#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1771#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1772#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1773#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1774#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1775#define IFF_TEAM_PORT IFF_TEAM_PORT
1776#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1777#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1778#define IFF_MACVLAN IFF_MACVLAN
1779#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1780#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1781#define IFF_NO_QUEUE IFF_NO_QUEUE
1782#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1783#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1784#define IFF_TEAM IFF_TEAM
1785#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1786#define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1787#define IFF_MACSEC IFF_MACSEC
1788#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1789#define IFF_FAILOVER IFF_FAILOVER
1790#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1791#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1792#define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1793
1794/* Specifies the type of the struct net_device::ml_priv pointer */
1795enum netdev_ml_priv_type {
1796 ML_PRIV_NONE,
1797 ML_PRIV_CAN,
1798};
1799
1800enum netdev_stat_type {
1801 NETDEV_PCPU_STAT_NONE,
1802 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1803 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1804 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1805};
1806
1807/**
1808 * struct net_device - The DEVICE structure.
1809 *
1810 * Actually, this whole structure is a big mistake. It mixes I/O
1811 * data with strictly "high-level" data, and it has to know about
1812 * almost every data structure used in the INET module.
1813 *
1814 * @name: This is the first field of the "visible" part of this structure
1815 * (i.e. as seen by users in the "Space.c" file). It is the name
1816 * of the interface.
1817 *
1818 * @name_node: Name hashlist node
1819 * @ifalias: SNMP alias
1820 * @mem_end: Shared memory end
1821 * @mem_start: Shared memory start
1822 * @base_addr: Device I/O address
1823 * @irq: Device IRQ number
1824 *
1825 * @state: Generic network queuing layer state, see netdev_state_t
1826 * @dev_list: The global list of network devices
1827 * @napi_list: List entry used for polling NAPI devices
1828 * @unreg_list: List entry when we are unregistering the
1829 * device; see the function unregister_netdev
1830 * @close_list: List entry used when we are closing the device
1831 * @ptype_all: Device-specific packet handlers for all protocols
1832 * @ptype_specific: Device-specific, protocol-specific packet handlers
1833 *
1834 * @adj_list: Directly linked devices, like slaves for bonding
1835 * @features: Currently active device features
1836 * @hw_features: User-changeable features
1837 *
1838 * @wanted_features: User-requested features
1839 * @vlan_features: Mask of features inheritable by VLAN devices
1840 *
1841 * @hw_enc_features: Mask of features inherited by encapsulating devices
1842 * This field indicates what encapsulation
1843 * offloads the hardware is capable of doing,
1844 * and drivers will need to set them appropriately.
1845 *
1846 * @mpls_features: Mask of features inheritable by MPLS
1847 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1848 *
1849 * @ifindex: interface index
1850 * @group: The group the device belongs to
1851 *
1852 * @stats: Statistics struct, which was left as a legacy, use
1853 * rtnl_link_stats64 instead
1854 *
1855 * @core_stats: core networking counters,
1856 * do not use this in drivers
1857 * @carrier_up_count: Number of times the carrier has been up
1858 * @carrier_down_count: Number of times the carrier has been down
1859 *
1860 * @wireless_handlers: List of functions to handle Wireless Extensions,
1861 * instead of ioctl,
1862 * see <net/iw_handler.h> for details.
1863 * @wireless_data: Instance data managed by the core of wireless extensions
1864 *
1865 * @netdev_ops: Includes several pointers to callbacks,
1866 * if one wants to override the ndo_*() functions
1867 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1868 * @ethtool_ops: Management operations
1869 * @l3mdev_ops: Layer 3 master device operations
1870 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1871 * discovery handling. Necessary for e.g. 6LoWPAN.
1872 * @xfrmdev_ops: Transformation offload operations
1873 * @tlsdev_ops: Transport Layer Security offload operations
1874 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1875 * of Layer 2 headers.
1876 *
1877 * @flags: Interface flags (a la BSD)
1878 * @xdp_features: XDP capability supported by the device
1879 * @priv_flags: Like 'flags' but invisible to userspace,
1880 * see if.h for the definitions
1881 * @gflags: Global flags ( kept as legacy )
1882 * @padded: How much padding added by alloc_netdev()
1883 * @operstate: RFC2863 operstate
1884 * @link_mode: Mapping policy to operstate
1885 * @if_port: Selectable AUI, TP, ...
1886 * @dma: DMA channel
1887 * @mtu: Interface MTU value
1888 * @min_mtu: Interface Minimum MTU value
1889 * @max_mtu: Interface Maximum MTU value
1890 * @type: Interface hardware type
1891 * @hard_header_len: Maximum hardware header length.
1892 * @min_header_len: Minimum hardware header length
1893 *
1894 * @needed_headroom: Extra headroom the hardware may need, but not in all
1895 * cases can this be guaranteed
1896 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1897 * cases can this be guaranteed. Some cases also use
1898 * LL_MAX_HEADER instead to allocate the skb
1899 *
1900 * interface address info:
1901 *
1902 * @perm_addr: Permanent hw address
1903 * @addr_assign_type: Hw address assignment type
1904 * @addr_len: Hardware address length
1905 * @upper_level: Maximum depth level of upper devices.
1906 * @lower_level: Maximum depth level of lower devices.
1907 * @neigh_priv_len: Used in neigh_alloc()
1908 * @dev_id: Used to differentiate devices that share
1909 * the same link layer address
1910 * @dev_port: Used to differentiate devices that share
1911 * the same function
1912 * @addr_list_lock: XXX: need comments on this one
1913 * @name_assign_type: network interface name assignment type
1914 * @uc_promisc: Counter that indicates promiscuous mode
1915 * has been enabled due to the need to listen to
1916 * additional unicast addresses in a device that
1917 * does not implement ndo_set_rx_mode()
1918 * @uc: unicast mac addresses
1919 * @mc: multicast mac addresses
1920 * @dev_addrs: list of device hw addresses
1921 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1922 * @promiscuity: Number of times the NIC is told to work in
1923 * promiscuous mode; if it becomes 0 the NIC will
1924 * exit promiscuous mode
1925 * @allmulti: Counter, enables or disables allmulticast mode
1926 *
1927 * @vlan_info: VLAN info
1928 * @dsa_ptr: dsa specific data
1929 * @tipc_ptr: TIPC specific data
1930 * @atalk_ptr: AppleTalk link
1931 * @ip_ptr: IPv4 specific data
1932 * @ip6_ptr: IPv6 specific data
1933 * @ax25_ptr: AX.25 specific data
1934 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1935 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1936 * device struct
1937 * @mpls_ptr: mpls_dev struct pointer
1938 * @mctp_ptr: MCTP specific data
1939 *
1940 * @dev_addr: Hw address (before bcast,
1941 * because most packets are unicast)
1942 *
1943 * @_rx: Array of RX queues
1944 * @num_rx_queues: Number of RX queues
1945 * allocated at register_netdev() time
1946 * @real_num_rx_queues: Number of RX queues currently active in device
1947 * @xdp_prog: XDP sockets filter program pointer
1948 * @gro_flush_timeout: timeout for GRO layer in NAPI
1949 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1950 * allow to avoid NIC hard IRQ, on busy queues.
1951 *
1952 * @rx_handler: handler for received packets
1953 * @rx_handler_data: XXX: need comments on this one
1954 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1955 * @ingress_queue: XXX: need comments on this one
1956 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1957 * @broadcast: hw bcast address
1958 *
1959 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1960 * indexed by RX queue number. Assigned by driver.
1961 * This must only be set if the ndo_rx_flow_steer
1962 * operation is defined
1963 * @index_hlist: Device index hash chain
1964 *
1965 * @_tx: Array of TX queues
1966 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1967 * @real_num_tx_queues: Number of TX queues currently active in device
1968 * @qdisc: Root qdisc from userspace point of view
1969 * @tx_queue_len: Max frames per queue allowed
1970 * @tx_global_lock: XXX: need comments on this one
1971 * @xdp_bulkq: XDP device bulk queue
1972 * @xps_maps: all CPUs/RXQs maps for XPS device
1973 *
1974 * @xps_maps: XXX: need comments on this one
1975 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1976 * @nf_hooks_egress: netfilter hooks executed for egress packets
1977 * @qdisc_hash: qdisc hash table
1978 * @watchdog_timeo: Represents the timeout that is used by
1979 * the watchdog (see dev_watchdog())
1980 * @watchdog_timer: List of timers
1981 *
1982 * @proto_down_reason: reason a netdev interface is held down
1983 * @pcpu_refcnt: Number of references to this device
1984 * @dev_refcnt: Number of references to this device
1985 * @refcnt_tracker: Tracker directory for tracked references to this device
1986 * @todo_list: Delayed register/unregister
1987 * @link_watch_list: XXX: need comments on this one
1988 *
1989 * @reg_state: Register/unregister state machine
1990 * @dismantle: Device is going to be freed
1991 * @rtnl_link_state: This enum represents the phases of creating
1992 * a new link
1993 *
1994 * @needs_free_netdev: Should unregister perform free_netdev?
1995 * @priv_destructor: Called from unregister
1996 * @npinfo: XXX: need comments on this one
1997 * @nd_net: Network namespace this network device is inside
1998 *
1999 * @ml_priv: Mid-layer private
2000 * @ml_priv_type: Mid-layer private type
2001 *
2002 * @pcpu_stat_type: Type of device statistics which the core should
2003 * allocate/free: none, lstats, tstats, dstats. none
2004 * means the driver is handling statistics allocation/
2005 * freeing internally.
2006 * @lstats: Loopback statistics: packets, bytes
2007 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes
2008 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes
2009 *
2010 * @garp_port: GARP
2011 * @mrp_port: MRP
2012 *
2013 * @dm_private: Drop monitor private
2014 *
2015 * @dev: Class/net/name entry
2016 * @sysfs_groups: Space for optional device, statistics and wireless
2017 * sysfs groups
2018 *
2019 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
2020 * @rtnl_link_ops: Rtnl_link_ops
2021 *
2022 * @gso_max_size: Maximum size of generic segmentation offload
2023 * @tso_max_size: Device (as in HW) limit on the max TSO request size
2024 * @gso_max_segs: Maximum number of segments that can be passed to the
2025 * NIC for GSO
2026 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
2027 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
2028 * for IPv4.
2029 *
2030 * @dcbnl_ops: Data Center Bridging netlink ops
2031 * @num_tc: Number of traffic classes in the net device
2032 * @tc_to_txq: XXX: need comments on this one
2033 * @prio_tc_map: XXX: need comments on this one
2034 *
2035 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
2036 *
2037 * @priomap: XXX: need comments on this one
2038 * @phydev: Physical device may attach itself
2039 * for hardware timestamping
2040 * @sfp_bus: attached &struct sfp_bus structure.
2041 *
2042 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2043 *
2044 * @proto_down: protocol port state information can be sent to the
2045 * switch driver and used to set the phys state of the
2046 * switch port.
2047 *
2048 * @wol_enabled: Wake-on-LAN is enabled
2049 *
2050 * @threaded: napi threaded mode is enabled
2051 *
2052 * @net_notifier_list: List of per-net netdev notifier block
2053 * that follow this device when it is moved
2054 * to another network namespace.
2055 *
2056 * @macsec_ops: MACsec offloading ops
2057 *
2058 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2059 * offload capabilities of the device
2060 * @udp_tunnel_nic: UDP tunnel offload state
2061 * @xdp_state: stores info on attached XDP BPF programs
2062 *
2063 * @nested_level: Used as a parameter of spin_lock_nested() of
2064 * dev->addr_list_lock.
2065 * @unlink_list: As netif_addr_lock() can be called recursively,
2066 * keep a list of interfaces to be deleted.
2067 * @gro_max_size: Maximum size of aggregated packet in generic
2068 * receive offload (GRO)
2069 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2070 * receive offload (GRO), for IPv4.
2071 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2072 * zero copy driver
2073 *
2074 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2075 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2076 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2077 * @dev_registered_tracker: tracker for reference held while
2078 * registered
2079 * @offload_xstats_l3: L3 HW stats for this netdevice.
2080 *
2081 * @devlink_port: Pointer to related devlink port structure.
2082 * Assigned by a driver before netdev registration using
2083 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2084 * during the time netdevice is registered.
2085 *
2086 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2087 * where the clock is recovered.
2088 *
2089 * FIXME: cleanup struct net_device such that network protocol info
2090 * moves out.
2091 */
2092
2093struct net_device {
2094 char name[IFNAMSIZ];
2095 struct netdev_name_node *name_node;
2096 struct dev_ifalias __rcu *ifalias;
2097 /*
2098 * I/O specific fields
2099 * FIXME: Merge these and struct ifmap into one
2100 */
2101 unsigned long mem_end;
2102 unsigned long mem_start;
2103 unsigned long base_addr;
2104
2105 /*
2106 * Some hardware also needs these fields (state,dev_list,
2107 * napi_list,unreg_list,close_list) but they are not
2108 * part of the usual set specified in Space.c.
2109 */
2110
2111 unsigned long state;
2112
2113 struct list_head dev_list;
2114 struct list_head napi_list;
2115 struct list_head unreg_list;
2116 struct list_head close_list;
2117 struct list_head ptype_all;
2118 struct list_head ptype_specific;
2119
2120 struct {
2121 struct list_head upper;
2122 struct list_head lower;
2123 } adj_list;
2124
2125 /* Read-mostly cache-line for fast-path access */
2126 unsigned int flags;
2127 xdp_features_t xdp_features;
2128 unsigned long long priv_flags;
2129 const struct net_device_ops *netdev_ops;
2130 const struct xdp_metadata_ops *xdp_metadata_ops;
2131 int ifindex;
2132 unsigned short gflags;
2133 unsigned short hard_header_len;
2134
2135 /* Note : dev->mtu is often read without holding a lock.
2136 * Writers usually hold RTNL.
2137 * It is recommended to use READ_ONCE() to annotate the reads,
2138 * and to use WRITE_ONCE() to annotate the writes.
2139 */
2140 unsigned int mtu;
2141 unsigned short needed_headroom;
2142 unsigned short needed_tailroom;
2143
2144 netdev_features_t features;
2145 netdev_features_t hw_features;
2146 netdev_features_t wanted_features;
2147 netdev_features_t vlan_features;
2148 netdev_features_t hw_enc_features;
2149 netdev_features_t mpls_features;
2150 netdev_features_t gso_partial_features;
2151
2152 unsigned int min_mtu;
2153 unsigned int max_mtu;
2154 unsigned short type;
2155 unsigned char min_header_len;
2156 unsigned char name_assign_type;
2157
2158 int group;
2159
2160 struct net_device_stats stats; /* not used by modern drivers */
2161
2162 struct net_device_core_stats __percpu *core_stats;
2163
2164 /* Stats to monitor link on/off, flapping */
2165 atomic_t carrier_up_count;
2166 atomic_t carrier_down_count;
2167
2168#ifdef CONFIG_WIRELESS_EXT
2169 const struct iw_handler_def *wireless_handlers;
2170 struct iw_public_data *wireless_data;
2171#endif
2172 const struct ethtool_ops *ethtool_ops;
2173#ifdef CONFIG_NET_L3_MASTER_DEV
2174 const struct l3mdev_ops *l3mdev_ops;
2175#endif
2176#if IS_ENABLED(CONFIG_IPV6)
2177 const struct ndisc_ops *ndisc_ops;
2178#endif
2179
2180#ifdef CONFIG_XFRM_OFFLOAD
2181 const struct xfrmdev_ops *xfrmdev_ops;
2182#endif
2183
2184#if IS_ENABLED(CONFIG_TLS_DEVICE)
2185 const struct tlsdev_ops *tlsdev_ops;
2186#endif
2187
2188 const struct header_ops *header_ops;
2189
2190 unsigned char operstate;
2191 unsigned char link_mode;
2192
2193 unsigned char if_port;
2194 unsigned char dma;
2195
2196 /* Interface address info. */
2197 unsigned char perm_addr[MAX_ADDR_LEN];
2198 unsigned char addr_assign_type;
2199 unsigned char addr_len;
2200 unsigned char upper_level;
2201 unsigned char lower_level;
2202
2203 unsigned short neigh_priv_len;
2204 unsigned short dev_id;
2205 unsigned short dev_port;
2206 unsigned short padded;
2207
2208 spinlock_t addr_list_lock;
2209 int irq;
2210
2211 struct netdev_hw_addr_list uc;
2212 struct netdev_hw_addr_list mc;
2213 struct netdev_hw_addr_list dev_addrs;
2214
2215#ifdef CONFIG_SYSFS
2216 struct kset *queues_kset;
2217#endif
2218#ifdef CONFIG_LOCKDEP
2219 struct list_head unlink_list;
2220#endif
2221 unsigned int promiscuity;
2222 unsigned int allmulti;
2223 bool uc_promisc;
2224#ifdef CONFIG_LOCKDEP
2225 unsigned char nested_level;
2226#endif
2227
2228
2229 /* Protocol-specific pointers */
2230
2231 struct in_device __rcu *ip_ptr;
2232 struct inet6_dev __rcu *ip6_ptr;
2233#if IS_ENABLED(CONFIG_VLAN_8021Q)
2234 struct vlan_info __rcu *vlan_info;
2235#endif
2236#if IS_ENABLED(CONFIG_NET_DSA)
2237 struct dsa_port *dsa_ptr;
2238#endif
2239#if IS_ENABLED(CONFIG_TIPC)
2240 struct tipc_bearer __rcu *tipc_ptr;
2241#endif
2242#if IS_ENABLED(CONFIG_ATALK)
2243 void *atalk_ptr;
2244#endif
2245#if IS_ENABLED(CONFIG_AX25)
2246 void *ax25_ptr;
2247#endif
2248#if IS_ENABLED(CONFIG_CFG80211)
2249 struct wireless_dev *ieee80211_ptr;
2250#endif
2251#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2252 struct wpan_dev *ieee802154_ptr;
2253#endif
2254#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2255 struct mpls_dev __rcu *mpls_ptr;
2256#endif
2257#if IS_ENABLED(CONFIG_MCTP)
2258 struct mctp_dev __rcu *mctp_ptr;
2259#endif
2260
2261/*
2262 * Cache lines mostly used on receive path (including eth_type_trans())
2263 */
2264 /* Interface address info used in eth_type_trans() */
2265 const unsigned char *dev_addr;
2266
2267 struct netdev_rx_queue *_rx;
2268 unsigned int num_rx_queues;
2269 unsigned int real_num_rx_queues;
2270
2271 struct bpf_prog __rcu *xdp_prog;
2272 unsigned long gro_flush_timeout;
2273 int napi_defer_hard_irqs;
2274#define GRO_LEGACY_MAX_SIZE 65536u
2275/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2276 * and shinfo->gso_segs is a 16bit field.
2277 */
2278#define GRO_MAX_SIZE (8 * 65535u)
2279 unsigned int gro_max_size;
2280 unsigned int gro_ipv4_max_size;
2281 unsigned int xdp_zc_max_segs;
2282 rx_handler_func_t __rcu *rx_handler;
2283 void __rcu *rx_handler_data;
2284#ifdef CONFIG_NET_XGRESS
2285 struct bpf_mprog_entry __rcu *tcx_ingress;
2286#endif
2287 struct netdev_queue __rcu *ingress_queue;
2288#ifdef CONFIG_NETFILTER_INGRESS
2289 struct nf_hook_entries __rcu *nf_hooks_ingress;
2290#endif
2291
2292 unsigned char broadcast[MAX_ADDR_LEN];
2293#ifdef CONFIG_RFS_ACCEL
2294 struct cpu_rmap *rx_cpu_rmap;
2295#endif
2296 struct hlist_node index_hlist;
2297
2298/*
2299 * Cache lines mostly used on transmit path
2300 */
2301 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2302 unsigned int num_tx_queues;
2303 unsigned int real_num_tx_queues;
2304 struct Qdisc __rcu *qdisc;
2305 unsigned int tx_queue_len;
2306 spinlock_t tx_global_lock;
2307
2308 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2309
2310#ifdef CONFIG_XPS
2311 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2312#endif
2313#ifdef CONFIG_NET_XGRESS
2314 struct bpf_mprog_entry __rcu *tcx_egress;
2315#endif
2316#ifdef CONFIG_NETFILTER_EGRESS
2317 struct nf_hook_entries __rcu *nf_hooks_egress;
2318#endif
2319
2320#ifdef CONFIG_NET_SCHED
2321 DECLARE_HASHTABLE (qdisc_hash, 4);
2322#endif
2323 /* These may be needed for future network-power-down code. */
2324 struct timer_list watchdog_timer;
2325 int watchdog_timeo;
2326
2327 u32 proto_down_reason;
2328
2329 struct list_head todo_list;
2330
2331#ifdef CONFIG_PCPU_DEV_REFCNT
2332 int __percpu *pcpu_refcnt;
2333#else
2334 refcount_t dev_refcnt;
2335#endif
2336 struct ref_tracker_dir refcnt_tracker;
2337
2338 struct list_head link_watch_list;
2339
2340 enum { NETREG_UNINITIALIZED=0,
2341 NETREG_REGISTERED, /* completed register_netdevice */
2342 NETREG_UNREGISTERING, /* called unregister_netdevice */
2343 NETREG_UNREGISTERED, /* completed unregister todo */
2344 NETREG_RELEASED, /* called free_netdev */
2345 NETREG_DUMMY, /* dummy device for NAPI poll */
2346 } reg_state:8;
2347
2348 bool dismantle;
2349
2350 enum {
2351 RTNL_LINK_INITIALIZED,
2352 RTNL_LINK_INITIALIZING,
2353 } rtnl_link_state:16;
2354
2355 bool needs_free_netdev;
2356 void (*priv_destructor)(struct net_device *dev);
2357
2358#ifdef CONFIG_NETPOLL
2359 struct netpoll_info __rcu *npinfo;
2360#endif
2361
2362 possible_net_t nd_net;
2363
2364 /* mid-layer private */
2365 void *ml_priv;
2366 enum netdev_ml_priv_type ml_priv_type;
2367
2368 enum netdev_stat_type pcpu_stat_type:8;
2369 union {
2370 struct pcpu_lstats __percpu *lstats;
2371 struct pcpu_sw_netstats __percpu *tstats;
2372 struct pcpu_dstats __percpu *dstats;
2373 };
2374
2375#if IS_ENABLED(CONFIG_GARP)
2376 struct garp_port __rcu *garp_port;
2377#endif
2378#if IS_ENABLED(CONFIG_MRP)
2379 struct mrp_port __rcu *mrp_port;
2380#endif
2381#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2382 struct dm_hw_stat_delta __rcu *dm_private;
2383#endif
2384 struct device dev;
2385 const struct attribute_group *sysfs_groups[4];
2386 const struct attribute_group *sysfs_rx_queue_group;
2387
2388 const struct rtnl_link_ops *rtnl_link_ops;
2389
2390 /* for setting kernel sock attribute on TCP connection setup */
2391#define GSO_MAX_SEGS 65535u
2392#define GSO_LEGACY_MAX_SIZE 65536u
2393/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2394 * and shinfo->gso_segs is a 16bit field.
2395 */
2396#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2397
2398 unsigned int gso_max_size;
2399#define TSO_LEGACY_MAX_SIZE 65536
2400#define TSO_MAX_SIZE UINT_MAX
2401 unsigned int tso_max_size;
2402 u16 gso_max_segs;
2403#define TSO_MAX_SEGS U16_MAX
2404 u16 tso_max_segs;
2405 unsigned int gso_ipv4_max_size;
2406
2407#ifdef CONFIG_DCB
2408 const struct dcbnl_rtnl_ops *dcbnl_ops;
2409#endif
2410 s16 num_tc;
2411 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2412 u8 prio_tc_map[TC_BITMASK + 1];
2413
2414#if IS_ENABLED(CONFIG_FCOE)
2415 unsigned int fcoe_ddp_xid;
2416#endif
2417#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2418 struct netprio_map __rcu *priomap;
2419#endif
2420 struct phy_device *phydev;
2421 struct sfp_bus *sfp_bus;
2422 struct lock_class_key *qdisc_tx_busylock;
2423 bool proto_down;
2424 unsigned wol_enabled:1;
2425 unsigned threaded:1;
2426
2427 struct list_head net_notifier_list;
2428
2429#if IS_ENABLED(CONFIG_MACSEC)
2430 /* MACsec management functions */
2431 const struct macsec_ops *macsec_ops;
2432#endif
2433 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2434 struct udp_tunnel_nic *udp_tunnel_nic;
2435
2436 /* protected by rtnl_lock */
2437 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2438
2439 u8 dev_addr_shadow[MAX_ADDR_LEN];
2440 netdevice_tracker linkwatch_dev_tracker;
2441 netdevice_tracker watchdog_dev_tracker;
2442 netdevice_tracker dev_registered_tracker;
2443 struct rtnl_hw_stats64 *offload_xstats_l3;
2444
2445 struct devlink_port *devlink_port;
2446
2447#if IS_ENABLED(CONFIG_DPLL)
2448 struct dpll_pin *dpll_pin;
2449#endif
2450};
2451#define to_net_dev(d) container_of(d, struct net_device, dev)
2452
2453/*
2454 * Driver should use this to assign devlink port instance to a netdevice
2455 * before it registers the netdevice. Therefore devlink_port is static
2456 * during the netdev lifetime after it is registered.
2457 */
2458#define SET_NETDEV_DEVLINK_PORT(dev, port) \
2459({ \
2460 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2461 ((dev)->devlink_port = (port)); \
2462})
2463
2464static inline bool netif_elide_gro(const struct net_device *dev)
2465{
2466 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2467 return true;
2468 return false;
2469}
2470
2471#define NETDEV_ALIGN 32
2472
2473static inline
2474int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2475{
2476 return dev->prio_tc_map[prio & TC_BITMASK];
2477}
2478
2479static inline
2480int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2481{
2482 if (tc >= dev->num_tc)
2483 return -EINVAL;
2484
2485 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2486 return 0;
2487}
2488
2489int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2490void netdev_reset_tc(struct net_device *dev);
2491int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2492int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2493
2494static inline
2495int netdev_get_num_tc(struct net_device *dev)
2496{
2497 return dev->num_tc;
2498}
2499
2500static inline void net_prefetch(void *p)
2501{
2502 prefetch(p);
2503#if L1_CACHE_BYTES < 128
2504 prefetch((u8 *)p + L1_CACHE_BYTES);
2505#endif
2506}
2507
2508static inline void net_prefetchw(void *p)
2509{
2510 prefetchw(p);
2511#if L1_CACHE_BYTES < 128
2512 prefetchw((u8 *)p + L1_CACHE_BYTES);
2513#endif
2514}
2515
2516void netdev_unbind_sb_channel(struct net_device *dev,
2517 struct net_device *sb_dev);
2518int netdev_bind_sb_channel_queue(struct net_device *dev,
2519 struct net_device *sb_dev,
2520 u8 tc, u16 count, u16 offset);
2521int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2522static inline int netdev_get_sb_channel(struct net_device *dev)
2523{
2524 return max_t(int, -dev->num_tc, 0);
2525}
2526
2527static inline
2528struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2529 unsigned int index)
2530{
2531 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2532 return &dev->_tx[index];
2533}
2534
2535static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2536 const struct sk_buff *skb)
2537{
2538 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2539}
2540
2541static inline void netdev_for_each_tx_queue(struct net_device *dev,
2542 void (*f)(struct net_device *,
2543 struct netdev_queue *,
2544 void *),
2545 void *arg)
2546{
2547 unsigned int i;
2548
2549 for (i = 0; i < dev->num_tx_queues; i++)
2550 f(dev, &dev->_tx[i], arg);
2551}
2552
2553#define netdev_lockdep_set_classes(dev) \
2554{ \
2555 static struct lock_class_key qdisc_tx_busylock_key; \
2556 static struct lock_class_key qdisc_xmit_lock_key; \
2557 static struct lock_class_key dev_addr_list_lock_key; \
2558 unsigned int i; \
2559 \
2560 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2561 lockdep_set_class(&(dev)->addr_list_lock, \
2562 &dev_addr_list_lock_key); \
2563 for (i = 0; i < (dev)->num_tx_queues; i++) \
2564 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2565 &qdisc_xmit_lock_key); \
2566}
2567
2568u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2569 struct net_device *sb_dev);
2570struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2571 struct sk_buff *skb,
2572 struct net_device *sb_dev);
2573
2574/* returns the headroom that the master device needs to take in account
2575 * when forwarding to this dev
2576 */
2577static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2578{
2579 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2580}
2581
2582static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2583{
2584 if (dev->netdev_ops->ndo_set_rx_headroom)
2585 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2586}
2587
2588/* set the device rx headroom to the dev's default */
2589static inline void netdev_reset_rx_headroom(struct net_device *dev)
2590{
2591 netdev_set_rx_headroom(dev, -1);
2592}
2593
2594static inline void *netdev_get_ml_priv(struct net_device *dev,
2595 enum netdev_ml_priv_type type)
2596{
2597 if (dev->ml_priv_type != type)
2598 return NULL;
2599
2600 return dev->ml_priv;
2601}
2602
2603static inline void netdev_set_ml_priv(struct net_device *dev,
2604 void *ml_priv,
2605 enum netdev_ml_priv_type type)
2606{
2607 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2608 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2609 dev->ml_priv_type, type);
2610 WARN(!dev->ml_priv_type && dev->ml_priv,
2611 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2612
2613 dev->ml_priv = ml_priv;
2614 dev->ml_priv_type = type;
2615}
2616
2617/*
2618 * Net namespace inlines
2619 */
2620static inline
2621struct net *dev_net(const struct net_device *dev)
2622{
2623 return read_pnet(&dev->nd_net);
2624}
2625
2626static inline
2627void dev_net_set(struct net_device *dev, struct net *net)
2628{
2629 write_pnet(&dev->nd_net, net);
2630}
2631
2632/**
2633 * netdev_priv - access network device private data
2634 * @dev: network device
2635 *
2636 * Get network device private data
2637 */
2638static inline void *netdev_priv(const struct net_device *dev)
2639{
2640 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2641}
2642
2643/* Set the sysfs physical device reference for the network logical device
2644 * if set prior to registration will cause a symlink during initialization.
2645 */
2646#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2647
2648/* Set the sysfs device type for the network logical device to allow
2649 * fine-grained identification of different network device types. For
2650 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2651 */
2652#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2653
2654/* Default NAPI poll() weight
2655 * Device drivers are strongly advised to not use bigger value
2656 */
2657#define NAPI_POLL_WEIGHT 64
2658
2659void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2660 int (*poll)(struct napi_struct *, int), int weight);
2661
2662/**
2663 * netif_napi_add() - initialize a NAPI context
2664 * @dev: network device
2665 * @napi: NAPI context
2666 * @poll: polling function
2667 *
2668 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2669 * *any* of the other NAPI-related functions.
2670 */
2671static inline void
2672netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2673 int (*poll)(struct napi_struct *, int))
2674{
2675 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2676}
2677
2678static inline void
2679netif_napi_add_tx_weight(struct net_device *dev,
2680 struct napi_struct *napi,
2681 int (*poll)(struct napi_struct *, int),
2682 int weight)
2683{
2684 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2685 netif_napi_add_weight(dev, napi, poll, weight);
2686}
2687
2688/**
2689 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2690 * @dev: network device
2691 * @napi: NAPI context
2692 * @poll: polling function
2693 *
2694 * This variant of netif_napi_add() should be used from drivers using NAPI
2695 * to exclusively poll a TX queue.
2696 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2697 */
2698static inline void netif_napi_add_tx(struct net_device *dev,
2699 struct napi_struct *napi,
2700 int (*poll)(struct napi_struct *, int))
2701{
2702 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2703}
2704
2705/**
2706 * __netif_napi_del - remove a NAPI context
2707 * @napi: NAPI context
2708 *
2709 * Warning: caller must observe RCU grace period before freeing memory
2710 * containing @napi. Drivers might want to call this helper to combine
2711 * all the needed RCU grace periods into a single one.
2712 */
2713void __netif_napi_del(struct napi_struct *napi);
2714
2715/**
2716 * netif_napi_del - remove a NAPI context
2717 * @napi: NAPI context
2718 *
2719 * netif_napi_del() removes a NAPI context from the network device NAPI list
2720 */
2721static inline void netif_napi_del(struct napi_struct *napi)
2722{
2723 __netif_napi_del(napi);
2724 synchronize_net();
2725}
2726
2727struct packet_type {
2728 __be16 type; /* This is really htons(ether_type). */
2729 bool ignore_outgoing;
2730 struct net_device *dev; /* NULL is wildcarded here */
2731 netdevice_tracker dev_tracker;
2732 int (*func) (struct sk_buff *,
2733 struct net_device *,
2734 struct packet_type *,
2735 struct net_device *);
2736 void (*list_func) (struct list_head *,
2737 struct packet_type *,
2738 struct net_device *);
2739 bool (*id_match)(struct packet_type *ptype,
2740 struct sock *sk);
2741 struct net *af_packet_net;
2742 void *af_packet_priv;
2743 struct list_head list;
2744};
2745
2746struct offload_callbacks {
2747 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2748 netdev_features_t features);
2749 struct sk_buff *(*gro_receive)(struct list_head *head,
2750 struct sk_buff *skb);
2751 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2752};
2753
2754struct packet_offload {
2755 __be16 type; /* This is really htons(ether_type). */
2756 u16 priority;
2757 struct offload_callbacks callbacks;
2758 struct list_head list;
2759};
2760
2761/* often modified stats are per-CPU, other are shared (netdev->stats) */
2762struct pcpu_sw_netstats {
2763 u64_stats_t rx_packets;
2764 u64_stats_t rx_bytes;
2765 u64_stats_t tx_packets;
2766 u64_stats_t tx_bytes;
2767 struct u64_stats_sync syncp;
2768} __aligned(4 * sizeof(u64));
2769
2770struct pcpu_dstats {
2771 u64 rx_packets;
2772 u64 rx_bytes;
2773 u64 rx_drops;
2774 u64 tx_packets;
2775 u64 tx_bytes;
2776 u64 tx_drops;
2777 struct u64_stats_sync syncp;
2778} __aligned(8 * sizeof(u64));
2779
2780struct pcpu_lstats {
2781 u64_stats_t packets;
2782 u64_stats_t bytes;
2783 struct u64_stats_sync syncp;
2784} __aligned(2 * sizeof(u64));
2785
2786void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2787
2788static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2789{
2790 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2791
2792 u64_stats_update_begin(&tstats->syncp);
2793 u64_stats_add(&tstats->rx_bytes, len);
2794 u64_stats_inc(&tstats->rx_packets);
2795 u64_stats_update_end(&tstats->syncp);
2796}
2797
2798static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2799 unsigned int packets,
2800 unsigned int len)
2801{
2802 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2803
2804 u64_stats_update_begin(&tstats->syncp);
2805 u64_stats_add(&tstats->tx_bytes, len);
2806 u64_stats_add(&tstats->tx_packets, packets);
2807 u64_stats_update_end(&tstats->syncp);
2808}
2809
2810static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2811{
2812 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2813
2814 u64_stats_update_begin(&lstats->syncp);
2815 u64_stats_add(&lstats->bytes, len);
2816 u64_stats_inc(&lstats->packets);
2817 u64_stats_update_end(&lstats->syncp);
2818}
2819
2820#define __netdev_alloc_pcpu_stats(type, gfp) \
2821({ \
2822 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2823 if (pcpu_stats) { \
2824 int __cpu; \
2825 for_each_possible_cpu(__cpu) { \
2826 typeof(type) *stat; \
2827 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2828 u64_stats_init(&stat->syncp); \
2829 } \
2830 } \
2831 pcpu_stats; \
2832})
2833
2834#define netdev_alloc_pcpu_stats(type) \
2835 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2836
2837#define devm_netdev_alloc_pcpu_stats(dev, type) \
2838({ \
2839 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2840 if (pcpu_stats) { \
2841 int __cpu; \
2842 for_each_possible_cpu(__cpu) { \
2843 typeof(type) *stat; \
2844 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2845 u64_stats_init(&stat->syncp); \
2846 } \
2847 } \
2848 pcpu_stats; \
2849})
2850
2851enum netdev_lag_tx_type {
2852 NETDEV_LAG_TX_TYPE_UNKNOWN,
2853 NETDEV_LAG_TX_TYPE_RANDOM,
2854 NETDEV_LAG_TX_TYPE_BROADCAST,
2855 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2856 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2857 NETDEV_LAG_TX_TYPE_HASH,
2858};
2859
2860enum netdev_lag_hash {
2861 NETDEV_LAG_HASH_NONE,
2862 NETDEV_LAG_HASH_L2,
2863 NETDEV_LAG_HASH_L34,
2864 NETDEV_LAG_HASH_L23,
2865 NETDEV_LAG_HASH_E23,
2866 NETDEV_LAG_HASH_E34,
2867 NETDEV_LAG_HASH_VLAN_SRCMAC,
2868 NETDEV_LAG_HASH_UNKNOWN,
2869};
2870
2871struct netdev_lag_upper_info {
2872 enum netdev_lag_tx_type tx_type;
2873 enum netdev_lag_hash hash_type;
2874};
2875
2876struct netdev_lag_lower_state_info {
2877 u8 link_up : 1,
2878 tx_enabled : 1;
2879};
2880
2881#include <linux/notifier.h>
2882
2883/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2884 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2885 * adding new types.
2886 */
2887enum netdev_cmd {
2888 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2889 NETDEV_DOWN,
2890 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2891 detected a hardware crash and restarted
2892 - we can use this eg to kick tcp sessions
2893 once done */
2894 NETDEV_CHANGE, /* Notify device state change */
2895 NETDEV_REGISTER,
2896 NETDEV_UNREGISTER,
2897 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2898 NETDEV_CHANGEADDR, /* notify after the address change */
2899 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2900 NETDEV_GOING_DOWN,
2901 NETDEV_CHANGENAME,
2902 NETDEV_FEAT_CHANGE,
2903 NETDEV_BONDING_FAILOVER,
2904 NETDEV_PRE_UP,
2905 NETDEV_PRE_TYPE_CHANGE,
2906 NETDEV_POST_TYPE_CHANGE,
2907 NETDEV_POST_INIT,
2908 NETDEV_PRE_UNINIT,
2909 NETDEV_RELEASE,
2910 NETDEV_NOTIFY_PEERS,
2911 NETDEV_JOIN,
2912 NETDEV_CHANGEUPPER,
2913 NETDEV_RESEND_IGMP,
2914 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2915 NETDEV_CHANGEINFODATA,
2916 NETDEV_BONDING_INFO,
2917 NETDEV_PRECHANGEUPPER,
2918 NETDEV_CHANGELOWERSTATE,
2919 NETDEV_UDP_TUNNEL_PUSH_INFO,
2920 NETDEV_UDP_TUNNEL_DROP_INFO,
2921 NETDEV_CHANGE_TX_QUEUE_LEN,
2922 NETDEV_CVLAN_FILTER_PUSH_INFO,
2923 NETDEV_CVLAN_FILTER_DROP_INFO,
2924 NETDEV_SVLAN_FILTER_PUSH_INFO,
2925 NETDEV_SVLAN_FILTER_DROP_INFO,
2926 NETDEV_OFFLOAD_XSTATS_ENABLE,
2927 NETDEV_OFFLOAD_XSTATS_DISABLE,
2928 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2929 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2930 NETDEV_XDP_FEAT_CHANGE,
2931};
2932const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2933
2934int register_netdevice_notifier(struct notifier_block *nb);
2935int unregister_netdevice_notifier(struct notifier_block *nb);
2936int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2937int unregister_netdevice_notifier_net(struct net *net,
2938 struct notifier_block *nb);
2939int register_netdevice_notifier_dev_net(struct net_device *dev,
2940 struct notifier_block *nb,
2941 struct netdev_net_notifier *nn);
2942int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2943 struct notifier_block *nb,
2944 struct netdev_net_notifier *nn);
2945
2946struct netdev_notifier_info {
2947 struct net_device *dev;
2948 struct netlink_ext_ack *extack;
2949};
2950
2951struct netdev_notifier_info_ext {
2952 struct netdev_notifier_info info; /* must be first */
2953 union {
2954 u32 mtu;
2955 } ext;
2956};
2957
2958struct netdev_notifier_change_info {
2959 struct netdev_notifier_info info; /* must be first */
2960 unsigned int flags_changed;
2961};
2962
2963struct netdev_notifier_changeupper_info {
2964 struct netdev_notifier_info info; /* must be first */
2965 struct net_device *upper_dev; /* new upper dev */
2966 bool master; /* is upper dev master */
2967 bool linking; /* is the notification for link or unlink */
2968 void *upper_info; /* upper dev info */
2969};
2970
2971struct netdev_notifier_changelowerstate_info {
2972 struct netdev_notifier_info info; /* must be first */
2973 void *lower_state_info; /* is lower dev state */
2974};
2975
2976struct netdev_notifier_pre_changeaddr_info {
2977 struct netdev_notifier_info info; /* must be first */
2978 const unsigned char *dev_addr;
2979};
2980
2981enum netdev_offload_xstats_type {
2982 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2983};
2984
2985struct netdev_notifier_offload_xstats_info {
2986 struct netdev_notifier_info info; /* must be first */
2987 enum netdev_offload_xstats_type type;
2988
2989 union {
2990 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2991 struct netdev_notifier_offload_xstats_rd *report_delta;
2992 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2993 struct netdev_notifier_offload_xstats_ru *report_used;
2994 };
2995};
2996
2997int netdev_offload_xstats_enable(struct net_device *dev,
2998 enum netdev_offload_xstats_type type,
2999 struct netlink_ext_ack *extack);
3000int netdev_offload_xstats_disable(struct net_device *dev,
3001 enum netdev_offload_xstats_type type);
3002bool netdev_offload_xstats_enabled(const struct net_device *dev,
3003 enum netdev_offload_xstats_type type);
3004int netdev_offload_xstats_get(struct net_device *dev,
3005 enum netdev_offload_xstats_type type,
3006 struct rtnl_hw_stats64 *stats, bool *used,
3007 struct netlink_ext_ack *extack);
3008void
3009netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3010 const struct rtnl_hw_stats64 *stats);
3011void
3012netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3013void netdev_offload_xstats_push_delta(struct net_device *dev,
3014 enum netdev_offload_xstats_type type,
3015 const struct rtnl_hw_stats64 *stats);
3016
3017static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3018 struct net_device *dev)
3019{
3020 info->dev = dev;
3021 info->extack = NULL;
3022}
3023
3024static inline struct net_device *
3025netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3026{
3027 return info->dev;
3028}
3029
3030static inline struct netlink_ext_ack *
3031netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3032{
3033 return info->extack;
3034}
3035
3036int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3037int call_netdevice_notifiers_info(unsigned long val,
3038 struct netdev_notifier_info *info);
3039
3040extern rwlock_t dev_base_lock; /* Device list lock */
3041
3042#define for_each_netdev(net, d) \
3043 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3044#define for_each_netdev_reverse(net, d) \
3045 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3046#define for_each_netdev_rcu(net, d) \
3047 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3048#define for_each_netdev_safe(net, d, n) \
3049 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3050#define for_each_netdev_continue(net, d) \
3051 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3052#define for_each_netdev_continue_reverse(net, d) \
3053 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3054 dev_list)
3055#define for_each_netdev_continue_rcu(net, d) \
3056 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3057#define for_each_netdev_in_bond_rcu(bond, slave) \
3058 for_each_netdev_rcu(&init_net, slave) \
3059 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3060#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3061
3062#define for_each_netdev_dump(net, d, ifindex) \
3063 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3064
3065static inline struct net_device *next_net_device(struct net_device *dev)
3066{
3067 struct list_head *lh;
3068 struct net *net;
3069
3070 net = dev_net(dev);
3071 lh = dev->dev_list.next;
3072 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3073}
3074
3075static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3076{
3077 struct list_head *lh;
3078 struct net *net;
3079
3080 net = dev_net(dev);
3081 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3082 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3083}
3084
3085static inline struct net_device *first_net_device(struct net *net)
3086{
3087 return list_empty(&net->dev_base_head) ? NULL :
3088 net_device_entry(net->dev_base_head.next);
3089}
3090
3091static inline struct net_device *first_net_device_rcu(struct net *net)
3092{
3093 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3094
3095 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3096}
3097
3098int netdev_boot_setup_check(struct net_device *dev);
3099struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3100 const char *hwaddr);
3101struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3102void dev_add_pack(struct packet_type *pt);
3103void dev_remove_pack(struct packet_type *pt);
3104void __dev_remove_pack(struct packet_type *pt);
3105void dev_add_offload(struct packet_offload *po);
3106void dev_remove_offload(struct packet_offload *po);
3107
3108int dev_get_iflink(const struct net_device *dev);
3109int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3110int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3111 struct net_device_path_stack *stack);
3112struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3113 unsigned short mask);
3114struct net_device *dev_get_by_name(struct net *net, const char *name);
3115struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3116struct net_device *__dev_get_by_name(struct net *net, const char *name);
3117bool netdev_name_in_use(struct net *net, const char *name);
3118int dev_alloc_name(struct net_device *dev, const char *name);
3119int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3120void dev_close(struct net_device *dev);
3121void dev_close_many(struct list_head *head, bool unlink);
3122void dev_disable_lro(struct net_device *dev);
3123int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3124u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3125 struct net_device *sb_dev);
3126u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3127 struct net_device *sb_dev);
3128
3129int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3130int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3131
3132static inline int dev_queue_xmit(struct sk_buff *skb)
3133{
3134 return __dev_queue_xmit(skb, NULL);
3135}
3136
3137static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3138 struct net_device *sb_dev)
3139{
3140 return __dev_queue_xmit(skb, sb_dev);
3141}
3142
3143static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3144{
3145 int ret;
3146
3147 ret = __dev_direct_xmit(skb, queue_id);
3148 if (!dev_xmit_complete(ret))
3149 kfree_skb(skb);
3150 return ret;
3151}
3152
3153int register_netdevice(struct net_device *dev);
3154void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3155void unregister_netdevice_many(struct list_head *head);
3156static inline void unregister_netdevice(struct net_device *dev)
3157{
3158 unregister_netdevice_queue(dev, NULL);
3159}
3160
3161int netdev_refcnt_read(const struct net_device *dev);
3162void free_netdev(struct net_device *dev);
3163void netdev_freemem(struct net_device *dev);
3164int init_dummy_netdev(struct net_device *dev);
3165
3166struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3167 struct sk_buff *skb,
3168 bool all_slaves);
3169struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3170 struct sock *sk);
3171struct net_device *dev_get_by_index(struct net *net, int ifindex);
3172struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3173struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3174 netdevice_tracker *tracker, gfp_t gfp);
3175struct net_device *netdev_get_by_name(struct net *net, const char *name,
3176 netdevice_tracker *tracker, gfp_t gfp);
3177struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3178struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3179
3180static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3181 unsigned short type,
3182 const void *daddr, const void *saddr,
3183 unsigned int len)
3184{
3185 if (!dev->header_ops || !dev->header_ops->create)
3186 return 0;
3187
3188 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3189}
3190
3191static inline int dev_parse_header(const struct sk_buff *skb,
3192 unsigned char *haddr)
3193{
3194 const struct net_device *dev = skb->dev;
3195
3196 if (!dev->header_ops || !dev->header_ops->parse)
3197 return 0;
3198 return dev->header_ops->parse(skb, haddr);
3199}
3200
3201static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3202{
3203 const struct net_device *dev = skb->dev;
3204
3205 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3206 return 0;
3207 return dev->header_ops->parse_protocol(skb);
3208}
3209
3210/* ll_header must have at least hard_header_len allocated */
3211static inline bool dev_validate_header(const struct net_device *dev,
3212 char *ll_header, int len)
3213{
3214 if (likely(len >= dev->hard_header_len))
3215 return true;
3216 if (len < dev->min_header_len)
3217 return false;
3218
3219 if (capable(CAP_SYS_RAWIO)) {
3220 memset(ll_header + len, 0, dev->hard_header_len - len);
3221 return true;
3222 }
3223
3224 if (dev->header_ops && dev->header_ops->validate)
3225 return dev->header_ops->validate(ll_header, len);
3226
3227 return false;
3228}
3229
3230static inline bool dev_has_header(const struct net_device *dev)
3231{
3232 return dev->header_ops && dev->header_ops->create;
3233}
3234
3235/*
3236 * Incoming packets are placed on per-CPU queues
3237 */
3238struct softnet_data {
3239 struct list_head poll_list;
3240 struct sk_buff_head process_queue;
3241
3242 /* stats */
3243 unsigned int processed;
3244 unsigned int time_squeeze;
3245#ifdef CONFIG_RPS
3246 struct softnet_data *rps_ipi_list;
3247#endif
3248
3249 bool in_net_rx_action;
3250 bool in_napi_threaded_poll;
3251
3252#ifdef CONFIG_NET_FLOW_LIMIT
3253 struct sd_flow_limit __rcu *flow_limit;
3254#endif
3255 struct Qdisc *output_queue;
3256 struct Qdisc **output_queue_tailp;
3257 struct sk_buff *completion_queue;
3258#ifdef CONFIG_XFRM_OFFLOAD
3259 struct sk_buff_head xfrm_backlog;
3260#endif
3261 /* written and read only by owning cpu: */
3262 struct {
3263 u16 recursion;
3264 u8 more;
3265#ifdef CONFIG_NET_EGRESS
3266 u8 skip_txqueue;
3267#endif
3268 } xmit;
3269#ifdef CONFIG_RPS
3270 /* input_queue_head should be written by cpu owning this struct,
3271 * and only read by other cpus. Worth using a cache line.
3272 */
3273 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3274
3275 /* Elements below can be accessed between CPUs for RPS/RFS */
3276 call_single_data_t csd ____cacheline_aligned_in_smp;
3277 struct softnet_data *rps_ipi_next;
3278 unsigned int cpu;
3279 unsigned int input_queue_tail;
3280#endif
3281 unsigned int received_rps;
3282 unsigned int dropped;
3283 struct sk_buff_head input_pkt_queue;
3284 struct napi_struct backlog;
3285
3286 /* Another possibly contended cache line */
3287 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3288 int defer_count;
3289 int defer_ipi_scheduled;
3290 struct sk_buff *defer_list;
3291 call_single_data_t defer_csd;
3292};
3293
3294static inline void input_queue_head_incr(struct softnet_data *sd)
3295{
3296#ifdef CONFIG_RPS
3297 sd->input_queue_head++;
3298#endif
3299}
3300
3301static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3302 unsigned int *qtail)
3303{
3304#ifdef CONFIG_RPS
3305 *qtail = ++sd->input_queue_tail;
3306#endif
3307}
3308
3309DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3310
3311static inline int dev_recursion_level(void)
3312{
3313 return this_cpu_read(softnet_data.xmit.recursion);
3314}
3315
3316#define XMIT_RECURSION_LIMIT 8
3317static inline bool dev_xmit_recursion(void)
3318{
3319 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3320 XMIT_RECURSION_LIMIT);
3321}
3322
3323static inline void dev_xmit_recursion_inc(void)
3324{
3325 __this_cpu_inc(softnet_data.xmit.recursion);
3326}
3327
3328static inline void dev_xmit_recursion_dec(void)
3329{
3330 __this_cpu_dec(softnet_data.xmit.recursion);
3331}
3332
3333void __netif_schedule(struct Qdisc *q);
3334void netif_schedule_queue(struct netdev_queue *txq);
3335
3336static inline void netif_tx_schedule_all(struct net_device *dev)
3337{
3338 unsigned int i;
3339
3340 for (i = 0; i < dev->num_tx_queues; i++)
3341 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3342}
3343
3344static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3345{
3346 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3347}
3348
3349/**
3350 * netif_start_queue - allow transmit
3351 * @dev: network device
3352 *
3353 * Allow upper layers to call the device hard_start_xmit routine.
3354 */
3355static inline void netif_start_queue(struct net_device *dev)
3356{
3357 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3358}
3359
3360static inline void netif_tx_start_all_queues(struct net_device *dev)
3361{
3362 unsigned int i;
3363
3364 for (i = 0; i < dev->num_tx_queues; i++) {
3365 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3366 netif_tx_start_queue(txq);
3367 }
3368}
3369
3370void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3371
3372/**
3373 * netif_wake_queue - restart transmit
3374 * @dev: network device
3375 *
3376 * Allow upper layers to call the device hard_start_xmit routine.
3377 * Used for flow control when transmit resources are available.
3378 */
3379static inline void netif_wake_queue(struct net_device *dev)
3380{
3381 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3382}
3383
3384static inline void netif_tx_wake_all_queues(struct net_device *dev)
3385{
3386 unsigned int i;
3387
3388 for (i = 0; i < dev->num_tx_queues; i++) {
3389 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3390 netif_tx_wake_queue(txq);
3391 }
3392}
3393
3394static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3395{
3396 /* Must be an atomic op see netif_txq_try_stop() */
3397 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3398}
3399
3400/**
3401 * netif_stop_queue - stop transmitted packets
3402 * @dev: network device
3403 *
3404 * Stop upper layers calling the device hard_start_xmit routine.
3405 * Used for flow control when transmit resources are unavailable.
3406 */
3407static inline void netif_stop_queue(struct net_device *dev)
3408{
3409 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3410}
3411
3412void netif_tx_stop_all_queues(struct net_device *dev);
3413
3414static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3415{
3416 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3417}
3418
3419/**
3420 * netif_queue_stopped - test if transmit queue is flowblocked
3421 * @dev: network device
3422 *
3423 * Test if transmit queue on device is currently unable to send.
3424 */
3425static inline bool netif_queue_stopped(const struct net_device *dev)
3426{
3427 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3428}
3429
3430static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3431{
3432 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3433}
3434
3435static inline bool
3436netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3437{
3438 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3439}
3440
3441static inline bool
3442netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3443{
3444 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3445}
3446
3447/**
3448 * netdev_queue_set_dql_min_limit - set dql minimum limit
3449 * @dev_queue: pointer to transmit queue
3450 * @min_limit: dql minimum limit
3451 *
3452 * Forces xmit_more() to return true until the minimum threshold
3453 * defined by @min_limit is reached (or until the tx queue is
3454 * empty). Warning: to be use with care, misuse will impact the
3455 * latency.
3456 */
3457static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3458 unsigned int min_limit)
3459{
3460#ifdef CONFIG_BQL
3461 dev_queue->dql.min_limit = min_limit;
3462#endif
3463}
3464
3465/**
3466 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3467 * @dev_queue: pointer to transmit queue
3468 *
3469 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3470 * to give appropriate hint to the CPU.
3471 */
3472static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3473{
3474#ifdef CONFIG_BQL
3475 prefetchw(&dev_queue->dql.num_queued);
3476#endif
3477}
3478
3479/**
3480 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3481 * @dev_queue: pointer to transmit queue
3482 *
3483 * BQL enabled drivers might use this helper in their TX completion path,
3484 * to give appropriate hint to the CPU.
3485 */
3486static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3487{
3488#ifdef CONFIG_BQL
3489 prefetchw(&dev_queue->dql.limit);
3490#endif
3491}
3492
3493/**
3494 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3495 * @dev_queue: network device queue
3496 * @bytes: number of bytes queued to the device queue
3497 *
3498 * Report the number of bytes queued for sending/completion to the network
3499 * device hardware queue. @bytes should be a good approximation and should
3500 * exactly match netdev_completed_queue() @bytes.
3501 * This is typically called once per packet, from ndo_start_xmit().
3502 */
3503static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3504 unsigned int bytes)
3505{
3506#ifdef CONFIG_BQL
3507 dql_queued(&dev_queue->dql, bytes);
3508
3509 if (likely(dql_avail(&dev_queue->dql) >= 0))
3510 return;
3511
3512 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3513
3514 /*
3515 * The XOFF flag must be set before checking the dql_avail below,
3516 * because in netdev_tx_completed_queue we update the dql_completed
3517 * before checking the XOFF flag.
3518 */
3519 smp_mb();
3520
3521 /* check again in case another CPU has just made room avail */
3522 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3523 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3524#endif
3525}
3526
3527/* Variant of netdev_tx_sent_queue() for drivers that are aware
3528 * that they should not test BQL status themselves.
3529 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3530 * skb of a batch.
3531 * Returns true if the doorbell must be used to kick the NIC.
3532 */
3533static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3534 unsigned int bytes,
3535 bool xmit_more)
3536{
3537 if (xmit_more) {
3538#ifdef CONFIG_BQL
3539 dql_queued(&dev_queue->dql, bytes);
3540#endif
3541 return netif_tx_queue_stopped(dev_queue);
3542 }
3543 netdev_tx_sent_queue(dev_queue, bytes);
3544 return true;
3545}
3546
3547/**
3548 * netdev_sent_queue - report the number of bytes queued to hardware
3549 * @dev: network device
3550 * @bytes: number of bytes queued to the hardware device queue
3551 *
3552 * Report the number of bytes queued for sending/completion to the network
3553 * device hardware queue#0. @bytes should be a good approximation and should
3554 * exactly match netdev_completed_queue() @bytes.
3555 * This is typically called once per packet, from ndo_start_xmit().
3556 */
3557static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3558{
3559 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3560}
3561
3562static inline bool __netdev_sent_queue(struct net_device *dev,
3563 unsigned int bytes,
3564 bool xmit_more)
3565{
3566 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3567 xmit_more);
3568}
3569
3570/**
3571 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3572 * @dev_queue: network device queue
3573 * @pkts: number of packets (currently ignored)
3574 * @bytes: number of bytes dequeued from the device queue
3575 *
3576 * Must be called at most once per TX completion round (and not per
3577 * individual packet), so that BQL can adjust its limits appropriately.
3578 */
3579static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3580 unsigned int pkts, unsigned int bytes)
3581{
3582#ifdef CONFIG_BQL
3583 if (unlikely(!bytes))
3584 return;
3585
3586 dql_completed(&dev_queue->dql, bytes);
3587
3588 /*
3589 * Without the memory barrier there is a small possiblity that
3590 * netdev_tx_sent_queue will miss the update and cause the queue to
3591 * be stopped forever
3592 */
3593 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3594
3595 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3596 return;
3597
3598 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3599 netif_schedule_queue(dev_queue);
3600#endif
3601}
3602
3603/**
3604 * netdev_completed_queue - report bytes and packets completed by device
3605 * @dev: network device
3606 * @pkts: actual number of packets sent over the medium
3607 * @bytes: actual number of bytes sent over the medium
3608 *
3609 * Report the number of bytes and packets transmitted by the network device
3610 * hardware queue over the physical medium, @bytes must exactly match the
3611 * @bytes amount passed to netdev_sent_queue()
3612 */
3613static inline void netdev_completed_queue(struct net_device *dev,
3614 unsigned int pkts, unsigned int bytes)
3615{
3616 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3617}
3618
3619static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3620{
3621#ifdef CONFIG_BQL
3622 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3623 dql_reset(&q->dql);
3624#endif
3625}
3626
3627/**
3628 * netdev_reset_queue - reset the packets and bytes count of a network device
3629 * @dev_queue: network device
3630 *
3631 * Reset the bytes and packet count of a network device and clear the
3632 * software flow control OFF bit for this network device
3633 */
3634static inline void netdev_reset_queue(struct net_device *dev_queue)
3635{
3636 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3637}
3638
3639/**
3640 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3641 * @dev: network device
3642 * @queue_index: given tx queue index
3643 *
3644 * Returns 0 if given tx queue index >= number of device tx queues,
3645 * otherwise returns the originally passed tx queue index.
3646 */
3647static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3648{
3649 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3650 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3651 dev->name, queue_index,
3652 dev->real_num_tx_queues);
3653 return 0;
3654 }
3655
3656 return queue_index;
3657}
3658
3659/**
3660 * netif_running - test if up
3661 * @dev: network device
3662 *
3663 * Test if the device has been brought up.
3664 */
3665static inline bool netif_running(const struct net_device *dev)
3666{
3667 return test_bit(__LINK_STATE_START, &dev->state);
3668}
3669
3670/*
3671 * Routines to manage the subqueues on a device. We only need start,
3672 * stop, and a check if it's stopped. All other device management is
3673 * done at the overall netdevice level.
3674 * Also test the device if we're multiqueue.
3675 */
3676
3677/**
3678 * netif_start_subqueue - allow sending packets on subqueue
3679 * @dev: network device
3680 * @queue_index: sub queue index
3681 *
3682 * Start individual transmit queue of a device with multiple transmit queues.
3683 */
3684static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3685{
3686 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3687
3688 netif_tx_start_queue(txq);
3689}
3690
3691/**
3692 * netif_stop_subqueue - stop sending packets on subqueue
3693 * @dev: network device
3694 * @queue_index: sub queue index
3695 *
3696 * Stop individual transmit queue of a device with multiple transmit queues.
3697 */
3698static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3699{
3700 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3701 netif_tx_stop_queue(txq);
3702}
3703
3704/**
3705 * __netif_subqueue_stopped - test status of subqueue
3706 * @dev: network device
3707 * @queue_index: sub queue index
3708 *
3709 * Check individual transmit queue of a device with multiple transmit queues.
3710 */
3711static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3712 u16 queue_index)
3713{
3714 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3715
3716 return netif_tx_queue_stopped(txq);
3717}
3718
3719/**
3720 * netif_subqueue_stopped - test status of subqueue
3721 * @dev: network device
3722 * @skb: sub queue buffer pointer
3723 *
3724 * Check individual transmit queue of a device with multiple transmit queues.
3725 */
3726static inline bool netif_subqueue_stopped(const struct net_device *dev,
3727 struct sk_buff *skb)
3728{
3729 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3730}
3731
3732/**
3733 * netif_wake_subqueue - allow sending packets on subqueue
3734 * @dev: network device
3735 * @queue_index: sub queue index
3736 *
3737 * Resume individual transmit queue of a device with multiple transmit queues.
3738 */
3739static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3740{
3741 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3742
3743 netif_tx_wake_queue(txq);
3744}
3745
3746#ifdef CONFIG_XPS
3747int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3748 u16 index);
3749int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3750 u16 index, enum xps_map_type type);
3751
3752/**
3753 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3754 * @j: CPU/Rx queue index
3755 * @mask: bitmask of all cpus/rx queues
3756 * @nr_bits: number of bits in the bitmask
3757 *
3758 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3759 */
3760static inline bool netif_attr_test_mask(unsigned long j,
3761 const unsigned long *mask,
3762 unsigned int nr_bits)
3763{
3764 cpu_max_bits_warn(j, nr_bits);
3765 return test_bit(j, mask);
3766}
3767
3768/**
3769 * netif_attr_test_online - Test for online CPU/Rx queue
3770 * @j: CPU/Rx queue index
3771 * @online_mask: bitmask for CPUs/Rx queues that are online
3772 * @nr_bits: number of bits in the bitmask
3773 *
3774 * Returns true if a CPU/Rx queue is online.
3775 */
3776static inline bool netif_attr_test_online(unsigned long j,
3777 const unsigned long *online_mask,
3778 unsigned int nr_bits)
3779{
3780 cpu_max_bits_warn(j, nr_bits);
3781
3782 if (online_mask)
3783 return test_bit(j, online_mask);
3784
3785 return (j < nr_bits);
3786}
3787
3788/**
3789 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3790 * @n: CPU/Rx queue index
3791 * @srcp: the cpumask/Rx queue mask pointer
3792 * @nr_bits: number of bits in the bitmask
3793 *
3794 * Returns >= nr_bits if no further CPUs/Rx queues set.
3795 */
3796static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3797 unsigned int nr_bits)
3798{
3799 /* -1 is a legal arg here. */
3800 if (n != -1)
3801 cpu_max_bits_warn(n, nr_bits);
3802
3803 if (srcp)
3804 return find_next_bit(srcp, nr_bits, n + 1);
3805
3806 return n + 1;
3807}
3808
3809/**
3810 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3811 * @n: CPU/Rx queue index
3812 * @src1p: the first CPUs/Rx queues mask pointer
3813 * @src2p: the second CPUs/Rx queues mask pointer
3814 * @nr_bits: number of bits in the bitmask
3815 *
3816 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3817 */
3818static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3819 const unsigned long *src2p,
3820 unsigned int nr_bits)
3821{
3822 /* -1 is a legal arg here. */
3823 if (n != -1)
3824 cpu_max_bits_warn(n, nr_bits);
3825
3826 if (src1p && src2p)
3827 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3828 else if (src1p)
3829 return find_next_bit(src1p, nr_bits, n + 1);
3830 else if (src2p)
3831 return find_next_bit(src2p, nr_bits, n + 1);
3832
3833 return n + 1;
3834}
3835#else
3836static inline int netif_set_xps_queue(struct net_device *dev,
3837 const struct cpumask *mask,
3838 u16 index)
3839{
3840 return 0;
3841}
3842
3843static inline int __netif_set_xps_queue(struct net_device *dev,
3844 const unsigned long *mask,
3845 u16 index, enum xps_map_type type)
3846{
3847 return 0;
3848}
3849#endif
3850
3851/**
3852 * netif_is_multiqueue - test if device has multiple transmit queues
3853 * @dev: network device
3854 *
3855 * Check if device has multiple transmit queues
3856 */
3857static inline bool netif_is_multiqueue(const struct net_device *dev)
3858{
3859 return dev->num_tx_queues > 1;
3860}
3861
3862int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3863
3864#ifdef CONFIG_SYSFS
3865int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3866#else
3867static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3868 unsigned int rxqs)
3869{
3870 dev->real_num_rx_queues = rxqs;
3871 return 0;
3872}
3873#endif
3874int netif_set_real_num_queues(struct net_device *dev,
3875 unsigned int txq, unsigned int rxq);
3876
3877int netif_get_num_default_rss_queues(void);
3878
3879void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3880void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3881
3882/*
3883 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3884 * interrupt context or with hardware interrupts being disabled.
3885 * (in_hardirq() || irqs_disabled())
3886 *
3887 * We provide four helpers that can be used in following contexts :
3888 *
3889 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3890 * replacing kfree_skb(skb)
3891 *
3892 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3893 * Typically used in place of consume_skb(skb) in TX completion path
3894 *
3895 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3896 * replacing kfree_skb(skb)
3897 *
3898 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3899 * and consumed a packet. Used in place of consume_skb(skb)
3900 */
3901static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3902{
3903 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3904}
3905
3906static inline void dev_consume_skb_irq(struct sk_buff *skb)
3907{
3908 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3909}
3910
3911static inline void dev_kfree_skb_any(struct sk_buff *skb)
3912{
3913 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3914}
3915
3916static inline void dev_consume_skb_any(struct sk_buff *skb)
3917{
3918 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3919}
3920
3921u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3922 struct bpf_prog *xdp_prog);
3923void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3924int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3925int netif_rx(struct sk_buff *skb);
3926int __netif_rx(struct sk_buff *skb);
3927
3928int netif_receive_skb(struct sk_buff *skb);
3929int netif_receive_skb_core(struct sk_buff *skb);
3930void netif_receive_skb_list_internal(struct list_head *head);
3931void netif_receive_skb_list(struct list_head *head);
3932gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3933void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3934struct sk_buff *napi_get_frags(struct napi_struct *napi);
3935void napi_get_frags_check(struct napi_struct *napi);
3936gro_result_t napi_gro_frags(struct napi_struct *napi);
3937struct packet_offload *gro_find_receive_by_type(__be16 type);
3938struct packet_offload *gro_find_complete_by_type(__be16 type);
3939
3940static inline void napi_free_frags(struct napi_struct *napi)
3941{
3942 kfree_skb(napi->skb);
3943 napi->skb = NULL;
3944}
3945
3946bool netdev_is_rx_handler_busy(struct net_device *dev);
3947int netdev_rx_handler_register(struct net_device *dev,
3948 rx_handler_func_t *rx_handler,
3949 void *rx_handler_data);
3950void netdev_rx_handler_unregister(struct net_device *dev);
3951
3952bool dev_valid_name(const char *name);
3953static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3954{
3955 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3956}
3957int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3958int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3959int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3960 void __user *data, bool *need_copyout);
3961int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3962int generic_hwtstamp_get_lower(struct net_device *dev,
3963 struct kernel_hwtstamp_config *kernel_cfg);
3964int generic_hwtstamp_set_lower(struct net_device *dev,
3965 struct kernel_hwtstamp_config *kernel_cfg,
3966 struct netlink_ext_ack *extack);
3967int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3968unsigned int dev_get_flags(const struct net_device *);
3969int __dev_change_flags(struct net_device *dev, unsigned int flags,
3970 struct netlink_ext_ack *extack);
3971int dev_change_flags(struct net_device *dev, unsigned int flags,
3972 struct netlink_ext_ack *extack);
3973int dev_set_alias(struct net_device *, const char *, size_t);
3974int dev_get_alias(const struct net_device *, char *, size_t);
3975int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3976 const char *pat, int new_ifindex);
3977static inline
3978int dev_change_net_namespace(struct net_device *dev, struct net *net,
3979 const char *pat)
3980{
3981 return __dev_change_net_namespace(dev, net, pat, 0);
3982}
3983int __dev_set_mtu(struct net_device *, int);
3984int dev_set_mtu(struct net_device *, int);
3985int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3986 struct netlink_ext_ack *extack);
3987int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3988 struct netlink_ext_ack *extack);
3989int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3990 struct netlink_ext_ack *extack);
3991int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3992int dev_get_port_parent_id(struct net_device *dev,
3993 struct netdev_phys_item_id *ppid, bool recurse);
3994bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3995void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin);
3996void netdev_dpll_pin_clear(struct net_device *dev);
3997
3998static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev)
3999{
4000#if IS_ENABLED(CONFIG_DPLL)
4001 return dev->dpll_pin;
4002#else
4003 return NULL;
4004#endif
4005}
4006
4007struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4008struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4009 struct netdev_queue *txq, int *ret);
4010
4011int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4012u8 dev_xdp_prog_count(struct net_device *dev);
4013u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4014
4015int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4016int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4017int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4018bool is_skb_forwardable(const struct net_device *dev,
4019 const struct sk_buff *skb);
4020
4021static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4022 const struct sk_buff *skb,
4023 const bool check_mtu)
4024{
4025 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4026 unsigned int len;
4027
4028 if (!(dev->flags & IFF_UP))
4029 return false;
4030
4031 if (!check_mtu)
4032 return true;
4033
4034 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4035 if (skb->len <= len)
4036 return true;
4037
4038 /* if TSO is enabled, we don't care about the length as the packet
4039 * could be forwarded without being segmented before
4040 */
4041 if (skb_is_gso(skb))
4042 return true;
4043
4044 return false;
4045}
4046
4047void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4048
4049#define DEV_CORE_STATS_INC(FIELD) \
4050static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4051{ \
4052 netdev_core_stats_inc(dev, \
4053 offsetof(struct net_device_core_stats, FIELD)); \
4054}
4055DEV_CORE_STATS_INC(rx_dropped)
4056DEV_CORE_STATS_INC(tx_dropped)
4057DEV_CORE_STATS_INC(rx_nohandler)
4058DEV_CORE_STATS_INC(rx_otherhost_dropped)
4059#undef DEV_CORE_STATS_INC
4060
4061static __always_inline int ____dev_forward_skb(struct net_device *dev,
4062 struct sk_buff *skb,
4063 const bool check_mtu)
4064{
4065 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4066 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4067 dev_core_stats_rx_dropped_inc(dev);
4068 kfree_skb(skb);
4069 return NET_RX_DROP;
4070 }
4071
4072 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4073 skb->priority = 0;
4074 return 0;
4075}
4076
4077bool dev_nit_active(struct net_device *dev);
4078void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4079
4080static inline void __dev_put(struct net_device *dev)
4081{
4082 if (dev) {
4083#ifdef CONFIG_PCPU_DEV_REFCNT
4084 this_cpu_dec(*dev->pcpu_refcnt);
4085#else
4086 refcount_dec(&dev->dev_refcnt);
4087#endif
4088 }
4089}
4090
4091static inline void __dev_hold(struct net_device *dev)
4092{
4093 if (dev) {
4094#ifdef CONFIG_PCPU_DEV_REFCNT
4095 this_cpu_inc(*dev->pcpu_refcnt);
4096#else
4097 refcount_inc(&dev->dev_refcnt);
4098#endif
4099 }
4100}
4101
4102static inline void __netdev_tracker_alloc(struct net_device *dev,
4103 netdevice_tracker *tracker,
4104 gfp_t gfp)
4105{
4106#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4107 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4108#endif
4109}
4110
4111/* netdev_tracker_alloc() can upgrade a prior untracked reference
4112 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4113 */
4114static inline void netdev_tracker_alloc(struct net_device *dev,
4115 netdevice_tracker *tracker, gfp_t gfp)
4116{
4117#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4118 refcount_dec(&dev->refcnt_tracker.no_tracker);
4119 __netdev_tracker_alloc(dev, tracker, gfp);
4120#endif
4121}
4122
4123static inline void netdev_tracker_free(struct net_device *dev,
4124 netdevice_tracker *tracker)
4125{
4126#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4127 ref_tracker_free(&dev->refcnt_tracker, tracker);
4128#endif
4129}
4130
4131static inline void netdev_hold(struct net_device *dev,
4132 netdevice_tracker *tracker, gfp_t gfp)
4133{
4134 if (dev) {
4135 __dev_hold(dev);
4136 __netdev_tracker_alloc(dev, tracker, gfp);
4137 }
4138}
4139
4140static inline void netdev_put(struct net_device *dev,
4141 netdevice_tracker *tracker)
4142{
4143 if (dev) {
4144 netdev_tracker_free(dev, tracker);
4145 __dev_put(dev);
4146 }
4147}
4148
4149/**
4150 * dev_hold - get reference to device
4151 * @dev: network device
4152 *
4153 * Hold reference to device to keep it from being freed.
4154 * Try using netdev_hold() instead.
4155 */
4156static inline void dev_hold(struct net_device *dev)
4157{
4158 netdev_hold(dev, NULL, GFP_ATOMIC);
4159}
4160
4161/**
4162 * dev_put - release reference to device
4163 * @dev: network device
4164 *
4165 * Release reference to device to allow it to be freed.
4166 * Try using netdev_put() instead.
4167 */
4168static inline void dev_put(struct net_device *dev)
4169{
4170 netdev_put(dev, NULL);
4171}
4172
4173static inline void netdev_ref_replace(struct net_device *odev,
4174 struct net_device *ndev,
4175 netdevice_tracker *tracker,
4176 gfp_t gfp)
4177{
4178 if (odev)
4179 netdev_tracker_free(odev, tracker);
4180
4181 __dev_hold(ndev);
4182 __dev_put(odev);
4183
4184 if (ndev)
4185 __netdev_tracker_alloc(ndev, tracker, gfp);
4186}
4187
4188/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4189 * and _off may be called from IRQ context, but it is caller
4190 * who is responsible for serialization of these calls.
4191 *
4192 * The name carrier is inappropriate, these functions should really be
4193 * called netif_lowerlayer_*() because they represent the state of any
4194 * kind of lower layer not just hardware media.
4195 */
4196void linkwatch_fire_event(struct net_device *dev);
4197
4198/**
4199 * netif_carrier_ok - test if carrier present
4200 * @dev: network device
4201 *
4202 * Check if carrier is present on device
4203 */
4204static inline bool netif_carrier_ok(const struct net_device *dev)
4205{
4206 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4207}
4208
4209unsigned long dev_trans_start(struct net_device *dev);
4210
4211void __netdev_watchdog_up(struct net_device *dev);
4212
4213void netif_carrier_on(struct net_device *dev);
4214void netif_carrier_off(struct net_device *dev);
4215void netif_carrier_event(struct net_device *dev);
4216
4217/**
4218 * netif_dormant_on - mark device as dormant.
4219 * @dev: network device
4220 *
4221 * Mark device as dormant (as per RFC2863).
4222 *
4223 * The dormant state indicates that the relevant interface is not
4224 * actually in a condition to pass packets (i.e., it is not 'up') but is
4225 * in a "pending" state, waiting for some external event. For "on-
4226 * demand" interfaces, this new state identifies the situation where the
4227 * interface is waiting for events to place it in the up state.
4228 */
4229static inline void netif_dormant_on(struct net_device *dev)
4230{
4231 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4232 linkwatch_fire_event(dev);
4233}
4234
4235/**
4236 * netif_dormant_off - set device as not dormant.
4237 * @dev: network device
4238 *
4239 * Device is not in dormant state.
4240 */
4241static inline void netif_dormant_off(struct net_device *dev)
4242{
4243 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4244 linkwatch_fire_event(dev);
4245}
4246
4247/**
4248 * netif_dormant - test if device is dormant
4249 * @dev: network device
4250 *
4251 * Check if device is dormant.
4252 */
4253static inline bool netif_dormant(const struct net_device *dev)
4254{
4255 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4256}
4257
4258
4259/**
4260 * netif_testing_on - mark device as under test.
4261 * @dev: network device
4262 *
4263 * Mark device as under test (as per RFC2863).
4264 *
4265 * The testing state indicates that some test(s) must be performed on
4266 * the interface. After completion, of the test, the interface state
4267 * will change to up, dormant, or down, as appropriate.
4268 */
4269static inline void netif_testing_on(struct net_device *dev)
4270{
4271 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4272 linkwatch_fire_event(dev);
4273}
4274
4275/**
4276 * netif_testing_off - set device as not under test.
4277 * @dev: network device
4278 *
4279 * Device is not in testing state.
4280 */
4281static inline void netif_testing_off(struct net_device *dev)
4282{
4283 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4284 linkwatch_fire_event(dev);
4285}
4286
4287/**
4288 * netif_testing - test if device is under test
4289 * @dev: network device
4290 *
4291 * Check if device is under test
4292 */
4293static inline bool netif_testing(const struct net_device *dev)
4294{
4295 return test_bit(__LINK_STATE_TESTING, &dev->state);
4296}
4297
4298
4299/**
4300 * netif_oper_up - test if device is operational
4301 * @dev: network device
4302 *
4303 * Check if carrier is operational
4304 */
4305static inline bool netif_oper_up(const struct net_device *dev)
4306{
4307 return (dev->operstate == IF_OPER_UP ||
4308 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4309}
4310
4311/**
4312 * netif_device_present - is device available or removed
4313 * @dev: network device
4314 *
4315 * Check if device has not been removed from system.
4316 */
4317static inline bool netif_device_present(const struct net_device *dev)
4318{
4319 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4320}
4321
4322void netif_device_detach(struct net_device *dev);
4323
4324void netif_device_attach(struct net_device *dev);
4325
4326/*
4327 * Network interface message level settings
4328 */
4329
4330enum {
4331 NETIF_MSG_DRV_BIT,
4332 NETIF_MSG_PROBE_BIT,
4333 NETIF_MSG_LINK_BIT,
4334 NETIF_MSG_TIMER_BIT,
4335 NETIF_MSG_IFDOWN_BIT,
4336 NETIF_MSG_IFUP_BIT,
4337 NETIF_MSG_RX_ERR_BIT,
4338 NETIF_MSG_TX_ERR_BIT,
4339 NETIF_MSG_TX_QUEUED_BIT,
4340 NETIF_MSG_INTR_BIT,
4341 NETIF_MSG_TX_DONE_BIT,
4342 NETIF_MSG_RX_STATUS_BIT,
4343 NETIF_MSG_PKTDATA_BIT,
4344 NETIF_MSG_HW_BIT,
4345 NETIF_MSG_WOL_BIT,
4346
4347 /* When you add a new bit above, update netif_msg_class_names array
4348 * in net/ethtool/common.c
4349 */
4350 NETIF_MSG_CLASS_COUNT,
4351};
4352/* Both ethtool_ops interface and internal driver implementation use u32 */
4353static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4354
4355#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4356#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4357
4358#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4359#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4360#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4361#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4362#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4363#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4364#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4365#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4366#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4367#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4368#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4369#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4370#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4371#define NETIF_MSG_HW __NETIF_MSG(HW)
4372#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4373
4374#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4375#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4376#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4377#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4378#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4379#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4380#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4381#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4382#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4383#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4384#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4385#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4386#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4387#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4388#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4389
4390static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4391{
4392 /* use default */
4393 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4394 return default_msg_enable_bits;
4395 if (debug_value == 0) /* no output */
4396 return 0;
4397 /* set low N bits */
4398 return (1U << debug_value) - 1;
4399}
4400
4401static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4402{
4403 spin_lock(&txq->_xmit_lock);
4404 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4405 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4406}
4407
4408static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4409{
4410 __acquire(&txq->_xmit_lock);
4411 return true;
4412}
4413
4414static inline void __netif_tx_release(struct netdev_queue *txq)
4415{
4416 __release(&txq->_xmit_lock);
4417}
4418
4419static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4420{
4421 spin_lock_bh(&txq->_xmit_lock);
4422 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4423 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4424}
4425
4426static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4427{
4428 bool ok = spin_trylock(&txq->_xmit_lock);
4429
4430 if (likely(ok)) {
4431 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4432 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4433 }
4434 return ok;
4435}
4436
4437static inline void __netif_tx_unlock(struct netdev_queue *txq)
4438{
4439 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4440 WRITE_ONCE(txq->xmit_lock_owner, -1);
4441 spin_unlock(&txq->_xmit_lock);
4442}
4443
4444static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4445{
4446 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4447 WRITE_ONCE(txq->xmit_lock_owner, -1);
4448 spin_unlock_bh(&txq->_xmit_lock);
4449}
4450
4451/*
4452 * txq->trans_start can be read locklessly from dev_watchdog()
4453 */
4454static inline void txq_trans_update(struct netdev_queue *txq)
4455{
4456 if (txq->xmit_lock_owner != -1)
4457 WRITE_ONCE(txq->trans_start, jiffies);
4458}
4459
4460static inline void txq_trans_cond_update(struct netdev_queue *txq)
4461{
4462 unsigned long now = jiffies;
4463
4464 if (READ_ONCE(txq->trans_start) != now)
4465 WRITE_ONCE(txq->trans_start, now);
4466}
4467
4468/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4469static inline void netif_trans_update(struct net_device *dev)
4470{
4471 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4472
4473 txq_trans_cond_update(txq);
4474}
4475
4476/**
4477 * netif_tx_lock - grab network device transmit lock
4478 * @dev: network device
4479 *
4480 * Get network device transmit lock
4481 */
4482void netif_tx_lock(struct net_device *dev);
4483
4484static inline void netif_tx_lock_bh(struct net_device *dev)
4485{
4486 local_bh_disable();
4487 netif_tx_lock(dev);
4488}
4489
4490void netif_tx_unlock(struct net_device *dev);
4491
4492static inline void netif_tx_unlock_bh(struct net_device *dev)
4493{
4494 netif_tx_unlock(dev);
4495 local_bh_enable();
4496}
4497
4498#define HARD_TX_LOCK(dev, txq, cpu) { \
4499 if ((dev->features & NETIF_F_LLTX) == 0) { \
4500 __netif_tx_lock(txq, cpu); \
4501 } else { \
4502 __netif_tx_acquire(txq); \
4503 } \
4504}
4505
4506#define HARD_TX_TRYLOCK(dev, txq) \
4507 (((dev->features & NETIF_F_LLTX) == 0) ? \
4508 __netif_tx_trylock(txq) : \
4509 __netif_tx_acquire(txq))
4510
4511#define HARD_TX_UNLOCK(dev, txq) { \
4512 if ((dev->features & NETIF_F_LLTX) == 0) { \
4513 __netif_tx_unlock(txq); \
4514 } else { \
4515 __netif_tx_release(txq); \
4516 } \
4517}
4518
4519static inline void netif_tx_disable(struct net_device *dev)
4520{
4521 unsigned int i;
4522 int cpu;
4523
4524 local_bh_disable();
4525 cpu = smp_processor_id();
4526 spin_lock(&dev->tx_global_lock);
4527 for (i = 0; i < dev->num_tx_queues; i++) {
4528 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4529
4530 __netif_tx_lock(txq, cpu);
4531 netif_tx_stop_queue(txq);
4532 __netif_tx_unlock(txq);
4533 }
4534 spin_unlock(&dev->tx_global_lock);
4535 local_bh_enable();
4536}
4537
4538static inline void netif_addr_lock(struct net_device *dev)
4539{
4540 unsigned char nest_level = 0;
4541
4542#ifdef CONFIG_LOCKDEP
4543 nest_level = dev->nested_level;
4544#endif
4545 spin_lock_nested(&dev->addr_list_lock, nest_level);
4546}
4547
4548static inline void netif_addr_lock_bh(struct net_device *dev)
4549{
4550 unsigned char nest_level = 0;
4551
4552#ifdef CONFIG_LOCKDEP
4553 nest_level = dev->nested_level;
4554#endif
4555 local_bh_disable();
4556 spin_lock_nested(&dev->addr_list_lock, nest_level);
4557}
4558
4559static inline void netif_addr_unlock(struct net_device *dev)
4560{
4561 spin_unlock(&dev->addr_list_lock);
4562}
4563
4564static inline void netif_addr_unlock_bh(struct net_device *dev)
4565{
4566 spin_unlock_bh(&dev->addr_list_lock);
4567}
4568
4569/*
4570 * dev_addrs walker. Should be used only for read access. Call with
4571 * rcu_read_lock held.
4572 */
4573#define for_each_dev_addr(dev, ha) \
4574 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4575
4576/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4577
4578void ether_setup(struct net_device *dev);
4579
4580/* Support for loadable net-drivers */
4581struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4582 unsigned char name_assign_type,
4583 void (*setup)(struct net_device *),
4584 unsigned int txqs, unsigned int rxqs);
4585#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4586 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4587
4588#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4589 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4590 count)
4591
4592int register_netdev(struct net_device *dev);
4593void unregister_netdev(struct net_device *dev);
4594
4595int devm_register_netdev(struct device *dev, struct net_device *ndev);
4596
4597/* General hardware address lists handling functions */
4598int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4599 struct netdev_hw_addr_list *from_list, int addr_len);
4600void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4601 struct netdev_hw_addr_list *from_list, int addr_len);
4602int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4603 struct net_device *dev,
4604 int (*sync)(struct net_device *, const unsigned char *),
4605 int (*unsync)(struct net_device *,
4606 const unsigned char *));
4607int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4608 struct net_device *dev,
4609 int (*sync)(struct net_device *,
4610 const unsigned char *, int),
4611 int (*unsync)(struct net_device *,
4612 const unsigned char *, int));
4613void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4614 struct net_device *dev,
4615 int (*unsync)(struct net_device *,
4616 const unsigned char *, int));
4617void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4618 struct net_device *dev,
4619 int (*unsync)(struct net_device *,
4620 const unsigned char *));
4621void __hw_addr_init(struct netdev_hw_addr_list *list);
4622
4623/* Functions used for device addresses handling */
4624void dev_addr_mod(struct net_device *dev, unsigned int offset,
4625 const void *addr, size_t len);
4626
4627static inline void
4628__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4629{
4630 dev_addr_mod(dev, 0, addr, len);
4631}
4632
4633static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4634{
4635 __dev_addr_set(dev, addr, dev->addr_len);
4636}
4637
4638int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4639 unsigned char addr_type);
4640int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4641 unsigned char addr_type);
4642
4643/* Functions used for unicast addresses handling */
4644int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4645int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4646int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4647int dev_uc_sync(struct net_device *to, struct net_device *from);
4648int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4649void dev_uc_unsync(struct net_device *to, struct net_device *from);
4650void dev_uc_flush(struct net_device *dev);
4651void dev_uc_init(struct net_device *dev);
4652
4653/**
4654 * __dev_uc_sync - Synchonize device's unicast list
4655 * @dev: device to sync
4656 * @sync: function to call if address should be added
4657 * @unsync: function to call if address should be removed
4658 *
4659 * Add newly added addresses to the interface, and release
4660 * addresses that have been deleted.
4661 */
4662static inline int __dev_uc_sync(struct net_device *dev,
4663 int (*sync)(struct net_device *,
4664 const unsigned char *),
4665 int (*unsync)(struct net_device *,
4666 const unsigned char *))
4667{
4668 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4669}
4670
4671/**
4672 * __dev_uc_unsync - Remove synchronized addresses from device
4673 * @dev: device to sync
4674 * @unsync: function to call if address should be removed
4675 *
4676 * Remove all addresses that were added to the device by dev_uc_sync().
4677 */
4678static inline void __dev_uc_unsync(struct net_device *dev,
4679 int (*unsync)(struct net_device *,
4680 const unsigned char *))
4681{
4682 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4683}
4684
4685/* Functions used for multicast addresses handling */
4686int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4687int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4688int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4689int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4690int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4691int dev_mc_sync(struct net_device *to, struct net_device *from);
4692int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4693void dev_mc_unsync(struct net_device *to, struct net_device *from);
4694void dev_mc_flush(struct net_device *dev);
4695void dev_mc_init(struct net_device *dev);
4696
4697/**
4698 * __dev_mc_sync - Synchonize device's multicast list
4699 * @dev: device to sync
4700 * @sync: function to call if address should be added
4701 * @unsync: function to call if address should be removed
4702 *
4703 * Add newly added addresses to the interface, and release
4704 * addresses that have been deleted.
4705 */
4706static inline int __dev_mc_sync(struct net_device *dev,
4707 int (*sync)(struct net_device *,
4708 const unsigned char *),
4709 int (*unsync)(struct net_device *,
4710 const unsigned char *))
4711{
4712 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4713}
4714
4715/**
4716 * __dev_mc_unsync - Remove synchronized addresses from device
4717 * @dev: device to sync
4718 * @unsync: function to call if address should be removed
4719 *
4720 * Remove all addresses that were added to the device by dev_mc_sync().
4721 */
4722static inline void __dev_mc_unsync(struct net_device *dev,
4723 int (*unsync)(struct net_device *,
4724 const unsigned char *))
4725{
4726 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4727}
4728
4729/* Functions used for secondary unicast and multicast support */
4730void dev_set_rx_mode(struct net_device *dev);
4731int dev_set_promiscuity(struct net_device *dev, int inc);
4732int dev_set_allmulti(struct net_device *dev, int inc);
4733void netdev_state_change(struct net_device *dev);
4734void __netdev_notify_peers(struct net_device *dev);
4735void netdev_notify_peers(struct net_device *dev);
4736void netdev_features_change(struct net_device *dev);
4737/* Load a device via the kmod */
4738void dev_load(struct net *net, const char *name);
4739struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4740 struct rtnl_link_stats64 *storage);
4741void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4742 const struct net_device_stats *netdev_stats);
4743void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4744 const struct pcpu_sw_netstats __percpu *netstats);
4745void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4746
4747extern int netdev_max_backlog;
4748extern int dev_rx_weight;
4749extern int dev_tx_weight;
4750extern int gro_normal_batch;
4751
4752enum {
4753 NESTED_SYNC_IMM_BIT,
4754 NESTED_SYNC_TODO_BIT,
4755};
4756
4757#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4758#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4759
4760#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4761#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4762
4763struct netdev_nested_priv {
4764 unsigned char flags;
4765 void *data;
4766};
4767
4768bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4769struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4770 struct list_head **iter);
4771
4772/* iterate through upper list, must be called under RCU read lock */
4773#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4774 for (iter = &(dev)->adj_list.upper, \
4775 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4776 updev; \
4777 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4778
4779int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4780 int (*fn)(struct net_device *upper_dev,
4781 struct netdev_nested_priv *priv),
4782 struct netdev_nested_priv *priv);
4783
4784bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4785 struct net_device *upper_dev);
4786
4787bool netdev_has_any_upper_dev(struct net_device *dev);
4788
4789void *netdev_lower_get_next_private(struct net_device *dev,
4790 struct list_head **iter);
4791void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4792 struct list_head **iter);
4793
4794#define netdev_for_each_lower_private(dev, priv, iter) \
4795 for (iter = (dev)->adj_list.lower.next, \
4796 priv = netdev_lower_get_next_private(dev, &(iter)); \
4797 priv; \
4798 priv = netdev_lower_get_next_private(dev, &(iter)))
4799
4800#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4801 for (iter = &(dev)->adj_list.lower, \
4802 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4803 priv; \
4804 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4805
4806void *netdev_lower_get_next(struct net_device *dev,
4807 struct list_head **iter);
4808
4809#define netdev_for_each_lower_dev(dev, ldev, iter) \
4810 for (iter = (dev)->adj_list.lower.next, \
4811 ldev = netdev_lower_get_next(dev, &(iter)); \
4812 ldev; \
4813 ldev = netdev_lower_get_next(dev, &(iter)))
4814
4815struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4816 struct list_head **iter);
4817int netdev_walk_all_lower_dev(struct net_device *dev,
4818 int (*fn)(struct net_device *lower_dev,
4819 struct netdev_nested_priv *priv),
4820 struct netdev_nested_priv *priv);
4821int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4822 int (*fn)(struct net_device *lower_dev,
4823 struct netdev_nested_priv *priv),
4824 struct netdev_nested_priv *priv);
4825
4826void *netdev_adjacent_get_private(struct list_head *adj_list);
4827void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4828struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4829struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4830int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4831 struct netlink_ext_ack *extack);
4832int netdev_master_upper_dev_link(struct net_device *dev,
4833 struct net_device *upper_dev,
4834 void *upper_priv, void *upper_info,
4835 struct netlink_ext_ack *extack);
4836void netdev_upper_dev_unlink(struct net_device *dev,
4837 struct net_device *upper_dev);
4838int netdev_adjacent_change_prepare(struct net_device *old_dev,
4839 struct net_device *new_dev,
4840 struct net_device *dev,
4841 struct netlink_ext_ack *extack);
4842void netdev_adjacent_change_commit(struct net_device *old_dev,
4843 struct net_device *new_dev,
4844 struct net_device *dev);
4845void netdev_adjacent_change_abort(struct net_device *old_dev,
4846 struct net_device *new_dev,
4847 struct net_device *dev);
4848void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4849void *netdev_lower_dev_get_private(struct net_device *dev,
4850 struct net_device *lower_dev);
4851void netdev_lower_state_changed(struct net_device *lower_dev,
4852 void *lower_state_info);
4853
4854/* RSS keys are 40 or 52 bytes long */
4855#define NETDEV_RSS_KEY_LEN 52
4856extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4857void netdev_rss_key_fill(void *buffer, size_t len);
4858
4859int skb_checksum_help(struct sk_buff *skb);
4860int skb_crc32c_csum_help(struct sk_buff *skb);
4861int skb_csum_hwoffload_help(struct sk_buff *skb,
4862 const netdev_features_t features);
4863
4864struct netdev_bonding_info {
4865 ifslave slave;
4866 ifbond master;
4867};
4868
4869struct netdev_notifier_bonding_info {
4870 struct netdev_notifier_info info; /* must be first */
4871 struct netdev_bonding_info bonding_info;
4872};
4873
4874void netdev_bonding_info_change(struct net_device *dev,
4875 struct netdev_bonding_info *bonding_info);
4876
4877#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4878void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4879#else
4880static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4881 const void *data)
4882{
4883}
4884#endif
4885
4886__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4887
4888static inline bool can_checksum_protocol(netdev_features_t features,
4889 __be16 protocol)
4890{
4891 if (protocol == htons(ETH_P_FCOE))
4892 return !!(features & NETIF_F_FCOE_CRC);
4893
4894 /* Assume this is an IP checksum (not SCTP CRC) */
4895
4896 if (features & NETIF_F_HW_CSUM) {
4897 /* Can checksum everything */
4898 return true;
4899 }
4900
4901 switch (protocol) {
4902 case htons(ETH_P_IP):
4903 return !!(features & NETIF_F_IP_CSUM);
4904 case htons(ETH_P_IPV6):
4905 return !!(features & NETIF_F_IPV6_CSUM);
4906 default:
4907 return false;
4908 }
4909}
4910
4911#ifdef CONFIG_BUG
4912void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4913#else
4914static inline void netdev_rx_csum_fault(struct net_device *dev,
4915 struct sk_buff *skb)
4916{
4917}
4918#endif
4919/* rx skb timestamps */
4920void net_enable_timestamp(void);
4921void net_disable_timestamp(void);
4922
4923static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4924 const struct skb_shared_hwtstamps *hwtstamps,
4925 bool cycles)
4926{
4927 const struct net_device_ops *ops = dev->netdev_ops;
4928
4929 if (ops->ndo_get_tstamp)
4930 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4931
4932 return hwtstamps->hwtstamp;
4933}
4934
4935static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4936 struct sk_buff *skb, struct net_device *dev,
4937 bool more)
4938{
4939 __this_cpu_write(softnet_data.xmit.more, more);
4940 return ops->ndo_start_xmit(skb, dev);
4941}
4942
4943static inline bool netdev_xmit_more(void)
4944{
4945 return __this_cpu_read(softnet_data.xmit.more);
4946}
4947
4948static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4949 struct netdev_queue *txq, bool more)
4950{
4951 const struct net_device_ops *ops = dev->netdev_ops;
4952 netdev_tx_t rc;
4953
4954 rc = __netdev_start_xmit(ops, skb, dev, more);
4955 if (rc == NETDEV_TX_OK)
4956 txq_trans_update(txq);
4957
4958 return rc;
4959}
4960
4961int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4962 const void *ns);
4963void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4964 const void *ns);
4965
4966extern const struct kobj_ns_type_operations net_ns_type_operations;
4967
4968const char *netdev_drivername(const struct net_device *dev);
4969
4970static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4971 netdev_features_t f2)
4972{
4973 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4974 if (f1 & NETIF_F_HW_CSUM)
4975 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4976 else
4977 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4978 }
4979
4980 return f1 & f2;
4981}
4982
4983static inline netdev_features_t netdev_get_wanted_features(
4984 struct net_device *dev)
4985{
4986 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4987}
4988netdev_features_t netdev_increment_features(netdev_features_t all,
4989 netdev_features_t one, netdev_features_t mask);
4990
4991/* Allow TSO being used on stacked device :
4992 * Performing the GSO segmentation before last device
4993 * is a performance improvement.
4994 */
4995static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4996 netdev_features_t mask)
4997{
4998 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4999}
5000
5001int __netdev_update_features(struct net_device *dev);
5002void netdev_update_features(struct net_device *dev);
5003void netdev_change_features(struct net_device *dev);
5004
5005void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5006 struct net_device *dev);
5007
5008netdev_features_t passthru_features_check(struct sk_buff *skb,
5009 struct net_device *dev,
5010 netdev_features_t features);
5011netdev_features_t netif_skb_features(struct sk_buff *skb);
5012void skb_warn_bad_offload(const struct sk_buff *skb);
5013
5014static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5015{
5016 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5017
5018 /* check flags correspondence */
5019 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5020 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5021 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5022 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5023 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5024 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5025 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5026 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5027 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5028 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5029 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5030 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5031 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5032 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5033 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5034 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5035 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5036 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5037 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5038
5039 return (features & feature) == feature;
5040}
5041
5042static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5043{
5044 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5045 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5046}
5047
5048static inline bool netif_needs_gso(struct sk_buff *skb,
5049 netdev_features_t features)
5050{
5051 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5052 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5053 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5054}
5055
5056void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5057void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5058void netif_inherit_tso_max(struct net_device *to,
5059 const struct net_device *from);
5060
5061static inline bool netif_is_macsec(const struct net_device *dev)
5062{
5063 return dev->priv_flags & IFF_MACSEC;
5064}
5065
5066static inline bool netif_is_macvlan(const struct net_device *dev)
5067{
5068 return dev->priv_flags & IFF_MACVLAN;
5069}
5070
5071static inline bool netif_is_macvlan_port(const struct net_device *dev)
5072{
5073 return dev->priv_flags & IFF_MACVLAN_PORT;
5074}
5075
5076static inline bool netif_is_bond_master(const struct net_device *dev)
5077{
5078 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5079}
5080
5081static inline bool netif_is_bond_slave(const struct net_device *dev)
5082{
5083 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5084}
5085
5086static inline bool netif_supports_nofcs(struct net_device *dev)
5087{
5088 return dev->priv_flags & IFF_SUPP_NOFCS;
5089}
5090
5091static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5092{
5093 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5094}
5095
5096static inline bool netif_is_l3_master(const struct net_device *dev)
5097{
5098 return dev->priv_flags & IFF_L3MDEV_MASTER;
5099}
5100
5101static inline bool netif_is_l3_slave(const struct net_device *dev)
5102{
5103 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5104}
5105
5106static inline int dev_sdif(const struct net_device *dev)
5107{
5108#ifdef CONFIG_NET_L3_MASTER_DEV
5109 if (netif_is_l3_slave(dev))
5110 return dev->ifindex;
5111#endif
5112 return 0;
5113}
5114
5115static inline bool netif_is_bridge_master(const struct net_device *dev)
5116{
5117 return dev->priv_flags & IFF_EBRIDGE;
5118}
5119
5120static inline bool netif_is_bridge_port(const struct net_device *dev)
5121{
5122 return dev->priv_flags & IFF_BRIDGE_PORT;
5123}
5124
5125static inline bool netif_is_ovs_master(const struct net_device *dev)
5126{
5127 return dev->priv_flags & IFF_OPENVSWITCH;
5128}
5129
5130static inline bool netif_is_ovs_port(const struct net_device *dev)
5131{
5132 return dev->priv_flags & IFF_OVS_DATAPATH;
5133}
5134
5135static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5136{
5137 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5138}
5139
5140static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5141{
5142 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5143}
5144
5145static inline bool netif_is_team_master(const struct net_device *dev)
5146{
5147 return dev->priv_flags & IFF_TEAM;
5148}
5149
5150static inline bool netif_is_team_port(const struct net_device *dev)
5151{
5152 return dev->priv_flags & IFF_TEAM_PORT;
5153}
5154
5155static inline bool netif_is_lag_master(const struct net_device *dev)
5156{
5157 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5158}
5159
5160static inline bool netif_is_lag_port(const struct net_device *dev)
5161{
5162 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5163}
5164
5165static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5166{
5167 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5168}
5169
5170static inline bool netif_is_failover(const struct net_device *dev)
5171{
5172 return dev->priv_flags & IFF_FAILOVER;
5173}
5174
5175static inline bool netif_is_failover_slave(const struct net_device *dev)
5176{
5177 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5178}
5179
5180/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5181static inline void netif_keep_dst(struct net_device *dev)
5182{
5183 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5184}
5185
5186/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5187static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5188{
5189 /* TODO: reserve and use an additional IFF bit, if we get more users */
5190 return netif_is_macsec(dev);
5191}
5192
5193extern struct pernet_operations __net_initdata loopback_net_ops;
5194
5195/* Logging, debugging and troubleshooting/diagnostic helpers. */
5196
5197/* netdev_printk helpers, similar to dev_printk */
5198
5199static inline const char *netdev_name(const struct net_device *dev)
5200{
5201 if (!dev->name[0] || strchr(dev->name, '%'))
5202 return "(unnamed net_device)";
5203 return dev->name;
5204}
5205
5206static inline const char *netdev_reg_state(const struct net_device *dev)
5207{
5208 switch (dev->reg_state) {
5209 case NETREG_UNINITIALIZED: return " (uninitialized)";
5210 case NETREG_REGISTERED: return "";
5211 case NETREG_UNREGISTERING: return " (unregistering)";
5212 case NETREG_UNREGISTERED: return " (unregistered)";
5213 case NETREG_RELEASED: return " (released)";
5214 case NETREG_DUMMY: return " (dummy)";
5215 }
5216
5217 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5218 return " (unknown)";
5219}
5220
5221#define MODULE_ALIAS_NETDEV(device) \
5222 MODULE_ALIAS("netdev-" device)
5223
5224/*
5225 * netdev_WARN() acts like dev_printk(), but with the key difference
5226 * of using a WARN/WARN_ON to get the message out, including the
5227 * file/line information and a backtrace.
5228 */
5229#define netdev_WARN(dev, format, args...) \
5230 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5231 netdev_reg_state(dev), ##args)
5232
5233#define netdev_WARN_ONCE(dev, format, args...) \
5234 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5235 netdev_reg_state(dev), ##args)
5236
5237/*
5238 * The list of packet types we will receive (as opposed to discard)
5239 * and the routines to invoke.
5240 *
5241 * Why 16. Because with 16 the only overlap we get on a hash of the
5242 * low nibble of the protocol value is RARP/SNAP/X.25.
5243 *
5244 * 0800 IP
5245 * 0001 802.3
5246 * 0002 AX.25
5247 * 0004 802.2
5248 * 8035 RARP
5249 * 0005 SNAP
5250 * 0805 X.25
5251 * 0806 ARP
5252 * 8137 IPX
5253 * 0009 Localtalk
5254 * 86DD IPv6
5255 */
5256#define PTYPE_HASH_SIZE (16)
5257#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5258
5259extern struct list_head ptype_all __read_mostly;
5260extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5261
5262extern struct net_device *blackhole_netdev;
5263
5264/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5265#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5266#define DEV_STATS_ADD(DEV, FIELD, VAL) \
5267 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5268#define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5269
5270#endif /* _LINUX_NETDEVICE_H */