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 int irq;
386};
387
388enum {
389 NAPI_STATE_SCHED, /* Poll is scheduled */
390 NAPI_STATE_MISSED, /* reschedule a napi */
391 NAPI_STATE_DISABLE, /* Disable pending */
392 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
393 NAPI_STATE_LISTED, /* NAPI added to system lists */
394 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
395 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
396 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
397 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
398 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
399};
400
401enum {
402 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
403 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
404 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
405 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
406 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
407 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
408 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
409 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
410 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
411 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
412};
413
414enum gro_result {
415 GRO_MERGED,
416 GRO_MERGED_FREE,
417 GRO_HELD,
418 GRO_NORMAL,
419 GRO_CONSUMED,
420};
421typedef enum gro_result gro_result_t;
422
423/*
424 * enum rx_handler_result - Possible return values for rx_handlers.
425 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
426 * further.
427 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
428 * case skb->dev was changed by rx_handler.
429 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
430 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
431 *
432 * rx_handlers are functions called from inside __netif_receive_skb(), to do
433 * special processing of the skb, prior to delivery to protocol handlers.
434 *
435 * Currently, a net_device can only have a single rx_handler registered. Trying
436 * to register a second rx_handler will return -EBUSY.
437 *
438 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
439 * To unregister a rx_handler on a net_device, use
440 * netdev_rx_handler_unregister().
441 *
442 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
443 * do with the skb.
444 *
445 * If the rx_handler consumed the skb in some way, it should return
446 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
447 * the skb to be delivered in some other way.
448 *
449 * If the rx_handler changed skb->dev, to divert the skb to another
450 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
451 * new device will be called if it exists.
452 *
453 * If the rx_handler decides the skb should be ignored, it should return
454 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
455 * are registered on exact device (ptype->dev == skb->dev).
456 *
457 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
458 * delivered, it should return RX_HANDLER_PASS.
459 *
460 * A device without a registered rx_handler will behave as if rx_handler
461 * returned RX_HANDLER_PASS.
462 */
463
464enum rx_handler_result {
465 RX_HANDLER_CONSUMED,
466 RX_HANDLER_ANOTHER,
467 RX_HANDLER_EXACT,
468 RX_HANDLER_PASS,
469};
470typedef enum rx_handler_result rx_handler_result_t;
471typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
472
473void __napi_schedule(struct napi_struct *n);
474void __napi_schedule_irqoff(struct napi_struct *n);
475
476static inline bool napi_disable_pending(struct napi_struct *n)
477{
478 return test_bit(NAPI_STATE_DISABLE, &n->state);
479}
480
481static inline bool napi_prefer_busy_poll(struct napi_struct *n)
482{
483 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
484}
485
486/**
487 * napi_is_scheduled - test if NAPI is scheduled
488 * @n: NAPI context
489 *
490 * This check is "best-effort". With no locking implemented,
491 * a NAPI can be scheduled or terminate right after this check
492 * and produce not precise results.
493 *
494 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
495 * should not be used normally and napi_schedule should be
496 * used instead.
497 *
498 * Use only if the driver really needs to check if a NAPI
499 * is scheduled for example in the context of delayed timer
500 * that can be skipped if a NAPI is already scheduled.
501 *
502 * Return True if NAPI is scheduled, False otherwise.
503 */
504static inline bool napi_is_scheduled(struct napi_struct *n)
505{
506 return test_bit(NAPI_STATE_SCHED, &n->state);
507}
508
509bool napi_schedule_prep(struct napi_struct *n);
510
511/**
512 * napi_schedule - schedule NAPI poll
513 * @n: NAPI context
514 *
515 * Schedule NAPI poll routine to be called if it is not already
516 * running.
517 * Return true if we schedule a NAPI or false if not.
518 * Refer to napi_schedule_prep() for additional reason on why
519 * a NAPI might not be scheduled.
520 */
521static inline bool napi_schedule(struct napi_struct *n)
522{
523 if (napi_schedule_prep(n)) {
524 __napi_schedule(n);
525 return true;
526 }
527
528 return false;
529}
530
531/**
532 * napi_schedule_irqoff - schedule NAPI poll
533 * @n: NAPI context
534 *
535 * Variant of napi_schedule(), assuming hard irqs are masked.
536 */
537static inline void napi_schedule_irqoff(struct napi_struct *n)
538{
539 if (napi_schedule_prep(n))
540 __napi_schedule_irqoff(n);
541}
542
543/**
544 * napi_complete_done - NAPI processing complete
545 * @n: NAPI context
546 * @work_done: number of packets processed
547 *
548 * Mark NAPI processing as complete. Should only be called if poll budget
549 * has not been completely consumed.
550 * Prefer over napi_complete().
551 * Return false if device should avoid rearming interrupts.
552 */
553bool napi_complete_done(struct napi_struct *n, int work_done);
554
555static inline bool napi_complete(struct napi_struct *n)
556{
557 return napi_complete_done(n, 0);
558}
559
560int dev_set_threaded(struct net_device *dev, bool threaded);
561
562/**
563 * napi_disable - prevent NAPI from scheduling
564 * @n: NAPI context
565 *
566 * Stop NAPI from being scheduled on this context.
567 * Waits till any outstanding processing completes.
568 */
569void napi_disable(struct napi_struct *n);
570
571void napi_enable(struct napi_struct *n);
572
573/**
574 * napi_synchronize - wait until NAPI is not running
575 * @n: NAPI context
576 *
577 * Wait until NAPI is done being scheduled on this context.
578 * Waits till any outstanding processing completes but
579 * does not disable future activations.
580 */
581static inline void napi_synchronize(const struct napi_struct *n)
582{
583 if (IS_ENABLED(CONFIG_SMP))
584 while (test_bit(NAPI_STATE_SCHED, &n->state))
585 msleep(1);
586 else
587 barrier();
588}
589
590/**
591 * napi_if_scheduled_mark_missed - if napi is running, set the
592 * NAPIF_STATE_MISSED
593 * @n: NAPI context
594 *
595 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
596 * NAPI is scheduled.
597 **/
598static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
599{
600 unsigned long val, new;
601
602 val = READ_ONCE(n->state);
603 do {
604 if (val & NAPIF_STATE_DISABLE)
605 return true;
606
607 if (!(val & NAPIF_STATE_SCHED))
608 return false;
609
610 new = val | NAPIF_STATE_MISSED;
611 } while (!try_cmpxchg(&n->state, &val, new));
612
613 return true;
614}
615
616enum netdev_queue_state_t {
617 __QUEUE_STATE_DRV_XOFF,
618 __QUEUE_STATE_STACK_XOFF,
619 __QUEUE_STATE_FROZEN,
620};
621
622#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
623#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
624#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
625
626#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
627#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
628 QUEUE_STATE_FROZEN)
629#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
630 QUEUE_STATE_FROZEN)
631
632/*
633 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
634 * netif_tx_* functions below are used to manipulate this flag. The
635 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
636 * queue independently. The netif_xmit_*stopped functions below are called
637 * to check if the queue has been stopped by the driver or stack (either
638 * of the XOFF bits are set in the state). Drivers should not need to call
639 * netif_xmit*stopped functions, they should only be using netif_tx_*.
640 */
641
642struct netdev_queue {
643/*
644 * read-mostly part
645 */
646 struct net_device *dev;
647 netdevice_tracker dev_tracker;
648
649 struct Qdisc __rcu *qdisc;
650 struct Qdisc __rcu *qdisc_sleeping;
651#ifdef CONFIG_SYSFS
652 struct kobject kobj;
653#endif
654#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
655 int numa_node;
656#endif
657 unsigned long tx_maxrate;
658 /*
659 * Number of TX timeouts for this queue
660 * (/sys/class/net/DEV/Q/trans_timeout)
661 */
662 atomic_long_t trans_timeout;
663
664 /* Subordinate device that the queue has been assigned to */
665 struct net_device *sb_dev;
666#ifdef CONFIG_XDP_SOCKETS
667 struct xsk_buff_pool *pool;
668#endif
669 /* NAPI instance for the queue
670 * Readers and writers must hold RTNL
671 */
672 struct napi_struct *napi;
673/*
674 * write-mostly part
675 */
676 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
677 int xmit_lock_owner;
678 /*
679 * Time (in jiffies) of last Tx
680 */
681 unsigned long trans_start;
682
683 unsigned long state;
684
685#ifdef CONFIG_BQL
686 struct dql dql;
687#endif
688} ____cacheline_aligned_in_smp;
689
690extern int sysctl_fb_tunnels_only_for_init_net;
691extern int sysctl_devconf_inherit_init_net;
692
693/*
694 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
695 * == 1 : For initns only
696 * == 2 : For none.
697 */
698static inline bool net_has_fallback_tunnels(const struct net *net)
699{
700#if IS_ENABLED(CONFIG_SYSCTL)
701 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
702
703 return !fb_tunnels_only_for_init_net ||
704 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
705#else
706 return true;
707#endif
708}
709
710static inline int net_inherit_devconf(void)
711{
712#if IS_ENABLED(CONFIG_SYSCTL)
713 return READ_ONCE(sysctl_devconf_inherit_init_net);
714#else
715 return 0;
716#endif
717}
718
719static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
720{
721#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
722 return q->numa_node;
723#else
724 return NUMA_NO_NODE;
725#endif
726}
727
728static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
729{
730#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
731 q->numa_node = node;
732#endif
733}
734
735#ifdef CONFIG_RPS
736/*
737 * This structure holds an RPS map which can be of variable length. The
738 * map is an array of CPUs.
739 */
740struct rps_map {
741 unsigned int len;
742 struct rcu_head rcu;
743 u16 cpus[];
744};
745#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
746
747/*
748 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
749 * tail pointer for that CPU's input queue at the time of last enqueue, and
750 * a hardware filter index.
751 */
752struct rps_dev_flow {
753 u16 cpu;
754 u16 filter;
755 unsigned int last_qtail;
756};
757#define RPS_NO_FILTER 0xffff
758
759/*
760 * The rps_dev_flow_table structure contains a table of flow mappings.
761 */
762struct rps_dev_flow_table {
763 unsigned int mask;
764 struct rcu_head rcu;
765 struct rps_dev_flow flows[];
766};
767#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
768 ((_num) * sizeof(struct rps_dev_flow)))
769
770/*
771 * The rps_sock_flow_table contains mappings of flows to the last CPU
772 * on which they were processed by the application (set in recvmsg).
773 * Each entry is a 32bit value. Upper part is the high-order bits
774 * of flow hash, lower part is CPU number.
775 * rps_cpu_mask is used to partition the space, depending on number of
776 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
777 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
778 * meaning we use 32-6=26 bits for the hash.
779 */
780struct rps_sock_flow_table {
781 u32 mask;
782
783 u32 ents[] ____cacheline_aligned_in_smp;
784};
785#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
786
787#define RPS_NO_CPU 0xffff
788
789extern u32 rps_cpu_mask;
790extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
791
792static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
793 u32 hash)
794{
795 if (table && hash) {
796 unsigned int index = hash & table->mask;
797 u32 val = hash & ~rps_cpu_mask;
798
799 /* We only give a hint, preemption can change CPU under us */
800 val |= raw_smp_processor_id();
801
802 /* The following WRITE_ONCE() is paired with the READ_ONCE()
803 * here, and another one in get_rps_cpu().
804 */
805 if (READ_ONCE(table->ents[index]) != val)
806 WRITE_ONCE(table->ents[index], val);
807 }
808}
809
810#ifdef CONFIG_RFS_ACCEL
811bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
812 u16 filter_id);
813#endif
814#endif /* CONFIG_RPS */
815
816/* XPS map type and offset of the xps map within net_device->xps_maps[]. */
817enum xps_map_type {
818 XPS_CPUS = 0,
819 XPS_RXQS,
820 XPS_MAPS_MAX,
821};
822
823#ifdef CONFIG_XPS
824/*
825 * This structure holds an XPS map which can be of variable length. The
826 * map is an array of queues.
827 */
828struct xps_map {
829 unsigned int len;
830 unsigned int alloc_len;
831 struct rcu_head rcu;
832 u16 queues[];
833};
834#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
835#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
836 - sizeof(struct xps_map)) / sizeof(u16))
837
838/*
839 * This structure holds all XPS maps for device. Maps are indexed by CPU.
840 *
841 * We keep track of the number of cpus/rxqs used when the struct is allocated,
842 * in nr_ids. This will help not accessing out-of-bound memory.
843 *
844 * We keep track of the number of traffic classes used when the struct is
845 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
846 * not crossing its upper bound, as the original dev->num_tc can be updated in
847 * the meantime.
848 */
849struct xps_dev_maps {
850 struct rcu_head rcu;
851 unsigned int nr_ids;
852 s16 num_tc;
853 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
854};
855
856#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
857 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
858
859#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
860 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
861
862#endif /* CONFIG_XPS */
863
864#define TC_MAX_QUEUE 16
865#define TC_BITMASK 15
866/* HW offloaded queuing disciplines txq count and offset maps */
867struct netdev_tc_txq {
868 u16 count;
869 u16 offset;
870};
871
872#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
873/*
874 * This structure is to hold information about the device
875 * configured to run FCoE protocol stack.
876 */
877struct netdev_fcoe_hbainfo {
878 char manufacturer[64];
879 char serial_number[64];
880 char hardware_version[64];
881 char driver_version[64];
882 char optionrom_version[64];
883 char firmware_version[64];
884 char model[256];
885 char model_description[256];
886};
887#endif
888
889#define MAX_PHYS_ITEM_ID_LEN 32
890
891/* This structure holds a unique identifier to identify some
892 * physical item (port for example) used by a netdevice.
893 */
894struct netdev_phys_item_id {
895 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
896 unsigned char id_len;
897};
898
899static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
900 struct netdev_phys_item_id *b)
901{
902 return a->id_len == b->id_len &&
903 memcmp(a->id, b->id, a->id_len) == 0;
904}
905
906typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
907 struct sk_buff *skb,
908 struct net_device *sb_dev);
909
910enum net_device_path_type {
911 DEV_PATH_ETHERNET = 0,
912 DEV_PATH_VLAN,
913 DEV_PATH_BRIDGE,
914 DEV_PATH_PPPOE,
915 DEV_PATH_DSA,
916 DEV_PATH_MTK_WDMA,
917};
918
919struct net_device_path {
920 enum net_device_path_type type;
921 const struct net_device *dev;
922 union {
923 struct {
924 u16 id;
925 __be16 proto;
926 u8 h_dest[ETH_ALEN];
927 } encap;
928 struct {
929 enum {
930 DEV_PATH_BR_VLAN_KEEP,
931 DEV_PATH_BR_VLAN_TAG,
932 DEV_PATH_BR_VLAN_UNTAG,
933 DEV_PATH_BR_VLAN_UNTAG_HW,
934 } vlan_mode;
935 u16 vlan_id;
936 __be16 vlan_proto;
937 } bridge;
938 struct {
939 int port;
940 u16 proto;
941 } dsa;
942 struct {
943 u8 wdma_idx;
944 u8 queue;
945 u16 wcid;
946 u8 bss;
947 u8 amsdu;
948 } mtk_wdma;
949 };
950};
951
952#define NET_DEVICE_PATH_STACK_MAX 5
953#define NET_DEVICE_PATH_VLAN_MAX 2
954
955struct net_device_path_stack {
956 int num_paths;
957 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
958};
959
960struct net_device_path_ctx {
961 const struct net_device *dev;
962 u8 daddr[ETH_ALEN];
963
964 int num_vlans;
965 struct {
966 u16 id;
967 __be16 proto;
968 } vlan[NET_DEVICE_PATH_VLAN_MAX];
969};
970
971enum tc_setup_type {
972 TC_QUERY_CAPS,
973 TC_SETUP_QDISC_MQPRIO,
974 TC_SETUP_CLSU32,
975 TC_SETUP_CLSFLOWER,
976 TC_SETUP_CLSMATCHALL,
977 TC_SETUP_CLSBPF,
978 TC_SETUP_BLOCK,
979 TC_SETUP_QDISC_CBS,
980 TC_SETUP_QDISC_RED,
981 TC_SETUP_QDISC_PRIO,
982 TC_SETUP_QDISC_MQ,
983 TC_SETUP_QDISC_ETF,
984 TC_SETUP_ROOT_QDISC,
985 TC_SETUP_QDISC_GRED,
986 TC_SETUP_QDISC_TAPRIO,
987 TC_SETUP_FT,
988 TC_SETUP_QDISC_ETS,
989 TC_SETUP_QDISC_TBF,
990 TC_SETUP_QDISC_FIFO,
991 TC_SETUP_QDISC_HTB,
992 TC_SETUP_ACT,
993};
994
995/* These structures hold the attributes of bpf state that are being passed
996 * to the netdevice through the bpf op.
997 */
998enum bpf_netdev_command {
999 /* Set or clear a bpf program used in the earliest stages of packet
1000 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
1001 * is responsible for calling bpf_prog_put on any old progs that are
1002 * stored. In case of error, the callee need not release the new prog
1003 * reference, but on success it takes ownership and must bpf_prog_put
1004 * when it is no longer used.
1005 */
1006 XDP_SETUP_PROG,
1007 XDP_SETUP_PROG_HW,
1008 /* BPF program for offload callbacks, invoked at program load time. */
1009 BPF_OFFLOAD_MAP_ALLOC,
1010 BPF_OFFLOAD_MAP_FREE,
1011 XDP_SETUP_XSK_POOL,
1012};
1013
1014struct bpf_prog_offload_ops;
1015struct netlink_ext_ack;
1016struct xdp_umem;
1017struct xdp_dev_bulk_queue;
1018struct bpf_xdp_link;
1019
1020enum bpf_xdp_mode {
1021 XDP_MODE_SKB = 0,
1022 XDP_MODE_DRV = 1,
1023 XDP_MODE_HW = 2,
1024 __MAX_XDP_MODE
1025};
1026
1027struct bpf_xdp_entity {
1028 struct bpf_prog *prog;
1029 struct bpf_xdp_link *link;
1030};
1031
1032struct netdev_bpf {
1033 enum bpf_netdev_command command;
1034 union {
1035 /* XDP_SETUP_PROG */
1036 struct {
1037 u32 flags;
1038 struct bpf_prog *prog;
1039 struct netlink_ext_ack *extack;
1040 };
1041 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1042 struct {
1043 struct bpf_offloaded_map *offmap;
1044 };
1045 /* XDP_SETUP_XSK_POOL */
1046 struct {
1047 struct xsk_buff_pool *pool;
1048 u16 queue_id;
1049 } xsk;
1050 };
1051};
1052
1053/* Flags for ndo_xsk_wakeup. */
1054#define XDP_WAKEUP_RX (1 << 0)
1055#define XDP_WAKEUP_TX (1 << 1)
1056
1057#ifdef CONFIG_XFRM_OFFLOAD
1058struct xfrmdev_ops {
1059 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1060 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1061 void (*xdo_dev_state_free) (struct xfrm_state *x);
1062 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1063 struct xfrm_state *x);
1064 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1065 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1066 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1067 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1068 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1069};
1070#endif
1071
1072struct dev_ifalias {
1073 struct rcu_head rcuhead;
1074 char ifalias[];
1075};
1076
1077struct devlink;
1078struct tlsdev_ops;
1079
1080struct netdev_net_notifier {
1081 struct list_head list;
1082 struct notifier_block *nb;
1083};
1084
1085/*
1086 * This structure defines the management hooks for network devices.
1087 * The following hooks can be defined; unless noted otherwise, they are
1088 * optional and can be filled with a null pointer.
1089 *
1090 * int (*ndo_init)(struct net_device *dev);
1091 * This function is called once when a network device is registered.
1092 * The network device can use this for any late stage initialization
1093 * or semantic validation. It can fail with an error code which will
1094 * be propagated back to register_netdev.
1095 *
1096 * void (*ndo_uninit)(struct net_device *dev);
1097 * This function is called when device is unregistered or when registration
1098 * fails. It is not called if init fails.
1099 *
1100 * int (*ndo_open)(struct net_device *dev);
1101 * This function is called when a network device transitions to the up
1102 * state.
1103 *
1104 * int (*ndo_stop)(struct net_device *dev);
1105 * This function is called when a network device transitions to the down
1106 * state.
1107 *
1108 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1109 * struct net_device *dev);
1110 * Called when a packet needs to be transmitted.
1111 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1112 * the queue before that can happen; it's for obsolete devices and weird
1113 * corner cases, but the stack really does a non-trivial amount
1114 * of useless work if you return NETDEV_TX_BUSY.
1115 * Required; cannot be NULL.
1116 *
1117 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1118 * struct net_device *dev
1119 * netdev_features_t features);
1120 * Called by core transmit path to determine if device is capable of
1121 * performing offload operations on a given packet. This is to give
1122 * the device an opportunity to implement any restrictions that cannot
1123 * be otherwise expressed by feature flags. The check is called with
1124 * the set of features that the stack has calculated and it returns
1125 * those the driver believes to be appropriate.
1126 *
1127 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1128 * struct net_device *sb_dev);
1129 * Called to decide which queue to use when device supports multiple
1130 * transmit queues.
1131 *
1132 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1133 * This function is called to allow device receiver to make
1134 * changes to configuration when multicast or promiscuous is enabled.
1135 *
1136 * void (*ndo_set_rx_mode)(struct net_device *dev);
1137 * This function is called device changes address list filtering.
1138 * If driver handles unicast address filtering, it should set
1139 * IFF_UNICAST_FLT in its priv_flags.
1140 *
1141 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1142 * This function is called when the Media Access Control address
1143 * needs to be changed. If this interface is not defined, the
1144 * MAC address can not be changed.
1145 *
1146 * int (*ndo_validate_addr)(struct net_device *dev);
1147 * Test if Media Access Control address is valid for the device.
1148 *
1149 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1150 * Old-style ioctl entry point. This is used internally by the
1151 * appletalk and ieee802154 subsystems but is no longer called by
1152 * the device ioctl handler.
1153 *
1154 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1155 * Used by the bonding driver for its device specific ioctls:
1156 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1157 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1158 *
1159 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1160 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1161 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1162 *
1163 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1164 * Used to set network devices bus interface parameters. This interface
1165 * is retained for legacy reasons; new devices should use the bus
1166 * interface (PCI) for low level management.
1167 *
1168 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1169 * Called when a user wants to change the Maximum Transfer Unit
1170 * of a device.
1171 *
1172 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1173 * Callback used when the transmitter has not made any progress
1174 * for dev->watchdog ticks.
1175 *
1176 * void (*ndo_get_stats64)(struct net_device *dev,
1177 * struct rtnl_link_stats64 *storage);
1178 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1179 * Called when a user wants to get the network device usage
1180 * statistics. Drivers must do one of the following:
1181 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1182 * rtnl_link_stats64 structure passed by the caller.
1183 * 2. Define @ndo_get_stats to update a net_device_stats structure
1184 * (which should normally be dev->stats) and return a pointer to
1185 * it. The structure may be changed asynchronously only if each
1186 * field is written atomically.
1187 * 3. Update dev->stats asynchronously and atomically, and define
1188 * neither operation.
1189 *
1190 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1191 * Return true if this device supports offload stats of this attr_id.
1192 *
1193 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1194 * void *attr_data)
1195 * Get statistics for offload operations by attr_id. Write it into the
1196 * attr_data pointer.
1197 *
1198 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1199 * If device supports VLAN filtering this function is called when a
1200 * VLAN id is registered.
1201 *
1202 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1203 * If device supports VLAN filtering this function is called when a
1204 * VLAN id is unregistered.
1205 *
1206 * void (*ndo_poll_controller)(struct net_device *dev);
1207 *
1208 * SR-IOV management functions.
1209 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1210 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1211 * u8 qos, __be16 proto);
1212 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1213 * int max_tx_rate);
1214 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1215 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1216 * int (*ndo_get_vf_config)(struct net_device *dev,
1217 * int vf, struct ifla_vf_info *ivf);
1218 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1219 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1220 * struct nlattr *port[]);
1221 *
1222 * Enable or disable the VF ability to query its RSS Redirection Table and
1223 * Hash Key. This is needed since on some devices VF share this information
1224 * with PF and querying it may introduce a theoretical security risk.
1225 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1226 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1227 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1228 * void *type_data);
1229 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1230 * This is always called from the stack with the rtnl lock held and netif
1231 * tx queues stopped. This allows the netdevice to perform queue
1232 * management safely.
1233 *
1234 * Fiber Channel over Ethernet (FCoE) offload functions.
1235 * int (*ndo_fcoe_enable)(struct net_device *dev);
1236 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1237 * so the underlying device can perform whatever needed configuration or
1238 * initialization to support acceleration of FCoE traffic.
1239 *
1240 * int (*ndo_fcoe_disable)(struct net_device *dev);
1241 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1242 * so the underlying device can perform whatever needed clean-ups to
1243 * stop supporting acceleration of FCoE traffic.
1244 *
1245 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1246 * struct scatterlist *sgl, unsigned int sgc);
1247 * Called when the FCoE Initiator wants to initialize an I/O that
1248 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1249 * perform necessary setup and returns 1 to indicate the device is set up
1250 * successfully to perform DDP on this I/O, otherwise this returns 0.
1251 *
1252 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1253 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1254 * indicated by the FC exchange id 'xid', so the underlying device can
1255 * clean up and reuse resources for later DDP requests.
1256 *
1257 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1258 * struct scatterlist *sgl, unsigned int sgc);
1259 * Called when the FCoE Target wants to initialize an I/O that
1260 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1261 * perform necessary setup and returns 1 to indicate the device is set up
1262 * successfully to perform DDP on this I/O, otherwise this returns 0.
1263 *
1264 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1265 * struct netdev_fcoe_hbainfo *hbainfo);
1266 * Called when the FCoE Protocol stack wants information on the underlying
1267 * device. This information is utilized by the FCoE protocol stack to
1268 * register attributes with Fiber Channel management service as per the
1269 * FC-GS Fabric Device Management Information(FDMI) specification.
1270 *
1271 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1272 * Called when the underlying device wants to override default World Wide
1273 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1274 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1275 * protocol stack to use.
1276 *
1277 * RFS acceleration.
1278 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1279 * u16 rxq_index, u32 flow_id);
1280 * Set hardware filter for RFS. rxq_index is the target queue index;
1281 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1282 * Return the filter ID on success, or a negative error code.
1283 *
1284 * Slave management functions (for bridge, bonding, etc).
1285 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1286 * Called to make another netdev an underling.
1287 *
1288 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1289 * Called to release previously enslaved netdev.
1290 *
1291 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1292 * struct sk_buff *skb,
1293 * bool all_slaves);
1294 * Get the xmit slave of master device. If all_slaves is true, function
1295 * assume all the slaves can transmit.
1296 *
1297 * Feature/offload setting functions.
1298 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1299 * netdev_features_t features);
1300 * Adjusts the requested feature flags according to device-specific
1301 * constraints, and returns the resulting flags. Must not modify
1302 * the device state.
1303 *
1304 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1305 * Called to update device configuration to new features. Passed
1306 * feature set might be less than what was returned by ndo_fix_features()).
1307 * Must return >0 or -errno if it changed dev->features itself.
1308 *
1309 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1310 * struct net_device *dev,
1311 * const unsigned char *addr, u16 vid, u16 flags,
1312 * struct netlink_ext_ack *extack);
1313 * Adds an FDB entry to dev for addr.
1314 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1315 * struct net_device *dev,
1316 * const unsigned char *addr, u16 vid)
1317 * Deletes the FDB entry from dev coresponding to addr.
1318 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1319 * struct netlink_ext_ack *extack);
1320 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1321 * struct net_device *dev, struct net_device *filter_dev,
1322 * int *idx)
1323 * Used to add FDB entries to dump requests. Implementers should add
1324 * entries to skb and update idx with the number of entries.
1325 *
1326 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1327 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1328 * Adds an MDB entry to dev.
1329 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1330 * struct netlink_ext_ack *extack);
1331 * Deletes the MDB entry from dev.
1332 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1333 * struct netlink_ext_ack *extack);
1334 * Bulk deletes MDB entries from dev.
1335 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1336 * struct netlink_callback *cb);
1337 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1338 * callback is used by core rtnetlink code.
1339 *
1340 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1341 * u16 flags, struct netlink_ext_ack *extack)
1342 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1343 * struct net_device *dev, u32 filter_mask,
1344 * int nlflags)
1345 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1346 * u16 flags);
1347 *
1348 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1349 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1350 * which do not represent real hardware may define this to allow their
1351 * userspace components to manage their virtual carrier state. Devices
1352 * that determine carrier state from physical hardware properties (eg
1353 * network cables) or protocol-dependent mechanisms (eg
1354 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1355 *
1356 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1357 * struct netdev_phys_item_id *ppid);
1358 * Called to get ID of physical port of this device. If driver does
1359 * not implement this, it is assumed that the hw is not able to have
1360 * multiple net devices on single physical port.
1361 *
1362 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1363 * struct netdev_phys_item_id *ppid)
1364 * Called to get the parent ID of the physical port of this device.
1365 *
1366 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1367 * struct net_device *dev)
1368 * Called by upper layer devices to accelerate switching or other
1369 * station functionality into hardware. 'pdev is the lowerdev
1370 * to use for the offload and 'dev' is the net device that will
1371 * back the offload. Returns a pointer to the private structure
1372 * the upper layer will maintain.
1373 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1374 * Called by upper layer device to delete the station created
1375 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1376 * the station and priv is the structure returned by the add
1377 * operation.
1378 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1379 * int queue_index, u32 maxrate);
1380 * Called when a user wants to set a max-rate limitation of specific
1381 * TX queue.
1382 * int (*ndo_get_iflink)(const struct net_device *dev);
1383 * Called to get the iflink value of this device.
1384 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1385 * This function is used to get egress tunnel information for given skb.
1386 * This is useful for retrieving outer tunnel header parameters while
1387 * sampling packet.
1388 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1389 * This function is used to specify the headroom that the skb must
1390 * consider when allocation skb during packet reception. Setting
1391 * appropriate rx headroom value allows avoiding skb head copy on
1392 * forward. Setting a negative value resets the rx headroom to the
1393 * default value.
1394 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1395 * This function is used to set or query state related to XDP on the
1396 * netdevice and manage BPF offload. See definition of
1397 * enum bpf_netdev_command for details.
1398 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1399 * u32 flags);
1400 * This function is used to submit @n XDP packets for transmit on a
1401 * netdevice. Returns number of frames successfully transmitted, frames
1402 * that got dropped are freed/returned via xdp_return_frame().
1403 * Returns negative number, means general error invoking ndo, meaning
1404 * no frames were xmit'ed and core-caller will free all frames.
1405 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1406 * struct xdp_buff *xdp);
1407 * Get the xmit slave of master device based on the xdp_buff.
1408 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1409 * This function is used to wake up the softirq, ksoftirqd or kthread
1410 * responsible for sending and/or receiving packets on a specific
1411 * queue id bound to an AF_XDP socket. The flags field specifies if
1412 * only RX, only Tx, or both should be woken up using the flags
1413 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1414 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1415 * int cmd);
1416 * Add, change, delete or get information on an IPv4 tunnel.
1417 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1418 * If a device is paired with a peer device, return the peer instance.
1419 * The caller must be under RCU read context.
1420 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1421 * Get the forwarding path to reach the real device from the HW destination address
1422 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1423 * const struct skb_shared_hwtstamps *hwtstamps,
1424 * bool cycles);
1425 * Get hardware timestamp based on normal/adjustable time or free running
1426 * cycle counter. This function is required if physical clock supports a
1427 * free running cycle counter.
1428 *
1429 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1430 * struct kernel_hwtstamp_config *kernel_config);
1431 * Get the currently configured hardware timestamping parameters for the
1432 * NIC device.
1433 *
1434 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1435 * struct kernel_hwtstamp_config *kernel_config,
1436 * struct netlink_ext_ack *extack);
1437 * Change the hardware timestamping parameters for NIC device.
1438 */
1439struct net_device_ops {
1440 int (*ndo_init)(struct net_device *dev);
1441 void (*ndo_uninit)(struct net_device *dev);
1442 int (*ndo_open)(struct net_device *dev);
1443 int (*ndo_stop)(struct net_device *dev);
1444 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1445 struct net_device *dev);
1446 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1447 struct net_device *dev,
1448 netdev_features_t features);
1449 u16 (*ndo_select_queue)(struct net_device *dev,
1450 struct sk_buff *skb,
1451 struct net_device *sb_dev);
1452 void (*ndo_change_rx_flags)(struct net_device *dev,
1453 int flags);
1454 void (*ndo_set_rx_mode)(struct net_device *dev);
1455 int (*ndo_set_mac_address)(struct net_device *dev,
1456 void *addr);
1457 int (*ndo_validate_addr)(struct net_device *dev);
1458 int (*ndo_do_ioctl)(struct net_device *dev,
1459 struct ifreq *ifr, int cmd);
1460 int (*ndo_eth_ioctl)(struct net_device *dev,
1461 struct ifreq *ifr, int cmd);
1462 int (*ndo_siocbond)(struct net_device *dev,
1463 struct ifreq *ifr, int cmd);
1464 int (*ndo_siocwandev)(struct net_device *dev,
1465 struct if_settings *ifs);
1466 int (*ndo_siocdevprivate)(struct net_device *dev,
1467 struct ifreq *ifr,
1468 void __user *data, int cmd);
1469 int (*ndo_set_config)(struct net_device *dev,
1470 struct ifmap *map);
1471 int (*ndo_change_mtu)(struct net_device *dev,
1472 int new_mtu);
1473 int (*ndo_neigh_setup)(struct net_device *dev,
1474 struct neigh_parms *);
1475 void (*ndo_tx_timeout) (struct net_device *dev,
1476 unsigned int txqueue);
1477
1478 void (*ndo_get_stats64)(struct net_device *dev,
1479 struct rtnl_link_stats64 *storage);
1480 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1481 int (*ndo_get_offload_stats)(int attr_id,
1482 const struct net_device *dev,
1483 void *attr_data);
1484 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1485
1486 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1487 __be16 proto, u16 vid);
1488 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1489 __be16 proto, u16 vid);
1490#ifdef CONFIG_NET_POLL_CONTROLLER
1491 void (*ndo_poll_controller)(struct net_device *dev);
1492 int (*ndo_netpoll_setup)(struct net_device *dev,
1493 struct netpoll_info *info);
1494 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1495#endif
1496 int (*ndo_set_vf_mac)(struct net_device *dev,
1497 int queue, u8 *mac);
1498 int (*ndo_set_vf_vlan)(struct net_device *dev,
1499 int queue, u16 vlan,
1500 u8 qos, __be16 proto);
1501 int (*ndo_set_vf_rate)(struct net_device *dev,
1502 int vf, int min_tx_rate,
1503 int max_tx_rate);
1504 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1505 int vf, bool setting);
1506 int (*ndo_set_vf_trust)(struct net_device *dev,
1507 int vf, bool setting);
1508 int (*ndo_get_vf_config)(struct net_device *dev,
1509 int vf,
1510 struct ifla_vf_info *ivf);
1511 int (*ndo_set_vf_link_state)(struct net_device *dev,
1512 int vf, int link_state);
1513 int (*ndo_get_vf_stats)(struct net_device *dev,
1514 int vf,
1515 struct ifla_vf_stats
1516 *vf_stats);
1517 int (*ndo_set_vf_port)(struct net_device *dev,
1518 int vf,
1519 struct nlattr *port[]);
1520 int (*ndo_get_vf_port)(struct net_device *dev,
1521 int vf, struct sk_buff *skb);
1522 int (*ndo_get_vf_guid)(struct net_device *dev,
1523 int vf,
1524 struct ifla_vf_guid *node_guid,
1525 struct ifla_vf_guid *port_guid);
1526 int (*ndo_set_vf_guid)(struct net_device *dev,
1527 int vf, u64 guid,
1528 int guid_type);
1529 int (*ndo_set_vf_rss_query_en)(
1530 struct net_device *dev,
1531 int vf, bool setting);
1532 int (*ndo_setup_tc)(struct net_device *dev,
1533 enum tc_setup_type type,
1534 void *type_data);
1535#if IS_ENABLED(CONFIG_FCOE)
1536 int (*ndo_fcoe_enable)(struct net_device *dev);
1537 int (*ndo_fcoe_disable)(struct net_device *dev);
1538 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1539 u16 xid,
1540 struct scatterlist *sgl,
1541 unsigned int sgc);
1542 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1543 u16 xid);
1544 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1545 u16 xid,
1546 struct scatterlist *sgl,
1547 unsigned int sgc);
1548 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1549 struct netdev_fcoe_hbainfo *hbainfo);
1550#endif
1551
1552#if IS_ENABLED(CONFIG_LIBFCOE)
1553#define NETDEV_FCOE_WWNN 0
1554#define NETDEV_FCOE_WWPN 1
1555 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1556 u64 *wwn, int type);
1557#endif
1558
1559#ifdef CONFIG_RFS_ACCEL
1560 int (*ndo_rx_flow_steer)(struct net_device *dev,
1561 const struct sk_buff *skb,
1562 u16 rxq_index,
1563 u32 flow_id);
1564#endif
1565 int (*ndo_add_slave)(struct net_device *dev,
1566 struct net_device *slave_dev,
1567 struct netlink_ext_ack *extack);
1568 int (*ndo_del_slave)(struct net_device *dev,
1569 struct net_device *slave_dev);
1570 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1571 struct sk_buff *skb,
1572 bool all_slaves);
1573 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1574 struct sock *sk);
1575 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1576 netdev_features_t features);
1577 int (*ndo_set_features)(struct net_device *dev,
1578 netdev_features_t features);
1579 int (*ndo_neigh_construct)(struct net_device *dev,
1580 struct neighbour *n);
1581 void (*ndo_neigh_destroy)(struct net_device *dev,
1582 struct neighbour *n);
1583
1584 int (*ndo_fdb_add)(struct ndmsg *ndm,
1585 struct nlattr *tb[],
1586 struct net_device *dev,
1587 const unsigned char *addr,
1588 u16 vid,
1589 u16 flags,
1590 struct netlink_ext_ack *extack);
1591 int (*ndo_fdb_del)(struct ndmsg *ndm,
1592 struct nlattr *tb[],
1593 struct net_device *dev,
1594 const unsigned char *addr,
1595 u16 vid, struct netlink_ext_ack *extack);
1596 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1597 struct net_device *dev,
1598 struct netlink_ext_ack *extack);
1599 int (*ndo_fdb_dump)(struct sk_buff *skb,
1600 struct netlink_callback *cb,
1601 struct net_device *dev,
1602 struct net_device *filter_dev,
1603 int *idx);
1604 int (*ndo_fdb_get)(struct sk_buff *skb,
1605 struct nlattr *tb[],
1606 struct net_device *dev,
1607 const unsigned char *addr,
1608 u16 vid, u32 portid, u32 seq,
1609 struct netlink_ext_ack *extack);
1610 int (*ndo_mdb_add)(struct net_device *dev,
1611 struct nlattr *tb[],
1612 u16 nlmsg_flags,
1613 struct netlink_ext_ack *extack);
1614 int (*ndo_mdb_del)(struct net_device *dev,
1615 struct nlattr *tb[],
1616 struct netlink_ext_ack *extack);
1617 int (*ndo_mdb_del_bulk)(struct net_device *dev,
1618 struct nlattr *tb[],
1619 struct netlink_ext_ack *extack);
1620 int (*ndo_mdb_dump)(struct net_device *dev,
1621 struct sk_buff *skb,
1622 struct netlink_callback *cb);
1623 int (*ndo_mdb_get)(struct net_device *dev,
1624 struct nlattr *tb[], u32 portid,
1625 u32 seq,
1626 struct netlink_ext_ack *extack);
1627 int (*ndo_bridge_setlink)(struct net_device *dev,
1628 struct nlmsghdr *nlh,
1629 u16 flags,
1630 struct netlink_ext_ack *extack);
1631 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1632 u32 pid, u32 seq,
1633 struct net_device *dev,
1634 u32 filter_mask,
1635 int nlflags);
1636 int (*ndo_bridge_dellink)(struct net_device *dev,
1637 struct nlmsghdr *nlh,
1638 u16 flags);
1639 int (*ndo_change_carrier)(struct net_device *dev,
1640 bool new_carrier);
1641 int (*ndo_get_phys_port_id)(struct net_device *dev,
1642 struct netdev_phys_item_id *ppid);
1643 int (*ndo_get_port_parent_id)(struct net_device *dev,
1644 struct netdev_phys_item_id *ppid);
1645 int (*ndo_get_phys_port_name)(struct net_device *dev,
1646 char *name, size_t len);
1647 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1648 struct net_device *dev);
1649 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1650 void *priv);
1651
1652 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1653 int queue_index,
1654 u32 maxrate);
1655 int (*ndo_get_iflink)(const struct net_device *dev);
1656 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1657 struct sk_buff *skb);
1658 void (*ndo_set_rx_headroom)(struct net_device *dev,
1659 int needed_headroom);
1660 int (*ndo_bpf)(struct net_device *dev,
1661 struct netdev_bpf *bpf);
1662 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1663 struct xdp_frame **xdp,
1664 u32 flags);
1665 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1666 struct xdp_buff *xdp);
1667 int (*ndo_xsk_wakeup)(struct net_device *dev,
1668 u32 queue_id, u32 flags);
1669 int (*ndo_tunnel_ctl)(struct net_device *dev,
1670 struct ip_tunnel_parm *p, int cmd);
1671 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1672 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1673 struct net_device_path *path);
1674 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1675 const struct skb_shared_hwtstamps *hwtstamps,
1676 bool cycles);
1677 int (*ndo_hwtstamp_get)(struct net_device *dev,
1678 struct kernel_hwtstamp_config *kernel_config);
1679 int (*ndo_hwtstamp_set)(struct net_device *dev,
1680 struct kernel_hwtstamp_config *kernel_config,
1681 struct netlink_ext_ack *extack);
1682};
1683
1684/**
1685 * enum netdev_priv_flags - &struct net_device priv_flags
1686 *
1687 * These are the &struct net_device, they are only set internally
1688 * by drivers and used in the kernel. These flags are invisible to
1689 * userspace; this means that the order of these flags can change
1690 * during any kernel release.
1691 *
1692 * You should have a pretty good reason to be extending these flags.
1693 *
1694 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1695 * @IFF_EBRIDGE: Ethernet bridging device
1696 * @IFF_BONDING: bonding master or slave
1697 * @IFF_ISATAP: ISATAP interface (RFC4214)
1698 * @IFF_WAN_HDLC: WAN HDLC device
1699 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1700 * release skb->dst
1701 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1702 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1703 * @IFF_MACVLAN_PORT: device used as macvlan port
1704 * @IFF_BRIDGE_PORT: device used as bridge port
1705 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1706 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1707 * @IFF_UNICAST_FLT: Supports unicast filtering
1708 * @IFF_TEAM_PORT: device used as team port
1709 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1710 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1711 * change when it's running
1712 * @IFF_MACVLAN: Macvlan device
1713 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1714 * underlying stacked devices
1715 * @IFF_L3MDEV_MASTER: device is an L3 master device
1716 * @IFF_NO_QUEUE: device can run without qdisc attached
1717 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1718 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1719 * @IFF_TEAM: device is a team device
1720 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1721 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1722 * entity (i.e. the master device for bridged veth)
1723 * @IFF_MACSEC: device is a MACsec device
1724 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1725 * @IFF_FAILOVER: device is a failover master device
1726 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1727 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1728 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1729 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1730 * skb_headlen(skb) == 0 (data starts from frag0)
1731 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1732 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1733 * ndo_hwtstamp_set() for all timestamp requests regardless of source,
1734 * even if those aren't HWTSTAMP_SOURCE_NETDEV.
1735 */
1736enum netdev_priv_flags {
1737 IFF_802_1Q_VLAN = 1<<0,
1738 IFF_EBRIDGE = 1<<1,
1739 IFF_BONDING = 1<<2,
1740 IFF_ISATAP = 1<<3,
1741 IFF_WAN_HDLC = 1<<4,
1742 IFF_XMIT_DST_RELEASE = 1<<5,
1743 IFF_DONT_BRIDGE = 1<<6,
1744 IFF_DISABLE_NETPOLL = 1<<7,
1745 IFF_MACVLAN_PORT = 1<<8,
1746 IFF_BRIDGE_PORT = 1<<9,
1747 IFF_OVS_DATAPATH = 1<<10,
1748 IFF_TX_SKB_SHARING = 1<<11,
1749 IFF_UNICAST_FLT = 1<<12,
1750 IFF_TEAM_PORT = 1<<13,
1751 IFF_SUPP_NOFCS = 1<<14,
1752 IFF_LIVE_ADDR_CHANGE = 1<<15,
1753 IFF_MACVLAN = 1<<16,
1754 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1755 IFF_L3MDEV_MASTER = 1<<18,
1756 IFF_NO_QUEUE = 1<<19,
1757 IFF_OPENVSWITCH = 1<<20,
1758 IFF_L3MDEV_SLAVE = 1<<21,
1759 IFF_TEAM = 1<<22,
1760 IFF_RXFH_CONFIGURED = 1<<23,
1761 IFF_PHONY_HEADROOM = 1<<24,
1762 IFF_MACSEC = 1<<25,
1763 IFF_NO_RX_HANDLER = 1<<26,
1764 IFF_FAILOVER = 1<<27,
1765 IFF_FAILOVER_SLAVE = 1<<28,
1766 IFF_L3MDEV_RX_HANDLER = 1<<29,
1767 IFF_NO_ADDRCONF = BIT_ULL(30),
1768 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1769 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1770 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33),
1771};
1772
1773#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1774#define IFF_EBRIDGE IFF_EBRIDGE
1775#define IFF_BONDING IFF_BONDING
1776#define IFF_ISATAP IFF_ISATAP
1777#define IFF_WAN_HDLC IFF_WAN_HDLC
1778#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1779#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1780#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1781#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1782#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1783#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1784#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1785#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1786#define IFF_TEAM_PORT IFF_TEAM_PORT
1787#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1788#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1789#define IFF_MACVLAN IFF_MACVLAN
1790#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1791#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1792#define IFF_NO_QUEUE IFF_NO_QUEUE
1793#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1794#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1795#define IFF_TEAM IFF_TEAM
1796#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1797#define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1798#define IFF_MACSEC IFF_MACSEC
1799#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1800#define IFF_FAILOVER IFF_FAILOVER
1801#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1802#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1803#define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1804
1805/* Specifies the type of the struct net_device::ml_priv pointer */
1806enum netdev_ml_priv_type {
1807 ML_PRIV_NONE,
1808 ML_PRIV_CAN,
1809};
1810
1811enum netdev_stat_type {
1812 NETDEV_PCPU_STAT_NONE,
1813 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1814 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1815 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1816};
1817
1818/**
1819 * struct net_device - The DEVICE structure.
1820 *
1821 * Actually, this whole structure is a big mistake. It mixes I/O
1822 * data with strictly "high-level" data, and it has to know about
1823 * almost every data structure used in the INET module.
1824 *
1825 * @name: This is the first field of the "visible" part of this structure
1826 * (i.e. as seen by users in the "Space.c" file). It is the name
1827 * of the interface.
1828 *
1829 * @name_node: Name hashlist node
1830 * @ifalias: SNMP alias
1831 * @mem_end: Shared memory end
1832 * @mem_start: Shared memory start
1833 * @base_addr: Device I/O address
1834 * @irq: Device IRQ number
1835 *
1836 * @state: Generic network queuing layer state, see netdev_state_t
1837 * @dev_list: The global list of network devices
1838 * @napi_list: List entry used for polling NAPI devices
1839 * @unreg_list: List entry when we are unregistering the
1840 * device; see the function unregister_netdev
1841 * @close_list: List entry used when we are closing the device
1842 * @ptype_all: Device-specific packet handlers for all protocols
1843 * @ptype_specific: Device-specific, protocol-specific packet handlers
1844 *
1845 * @adj_list: Directly linked devices, like slaves for bonding
1846 * @features: Currently active device features
1847 * @hw_features: User-changeable features
1848 *
1849 * @wanted_features: User-requested features
1850 * @vlan_features: Mask of features inheritable by VLAN devices
1851 *
1852 * @hw_enc_features: Mask of features inherited by encapsulating devices
1853 * This field indicates what encapsulation
1854 * offloads the hardware is capable of doing,
1855 * and drivers will need to set them appropriately.
1856 *
1857 * @mpls_features: Mask of features inheritable by MPLS
1858 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1859 *
1860 * @ifindex: interface index
1861 * @group: The group the device belongs to
1862 *
1863 * @stats: Statistics struct, which was left as a legacy, use
1864 * rtnl_link_stats64 instead
1865 *
1866 * @core_stats: core networking counters,
1867 * do not use this in drivers
1868 * @carrier_up_count: Number of times the carrier has been up
1869 * @carrier_down_count: Number of times the carrier has been down
1870 *
1871 * @wireless_handlers: List of functions to handle Wireless Extensions,
1872 * instead of ioctl,
1873 * see <net/iw_handler.h> for details.
1874 * @wireless_data: Instance data managed by the core of wireless extensions
1875 *
1876 * @netdev_ops: Includes several pointers to callbacks,
1877 * if one wants to override the ndo_*() functions
1878 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1879 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks.
1880 * @ethtool_ops: Management operations
1881 * @l3mdev_ops: Layer 3 master device operations
1882 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1883 * discovery handling. Necessary for e.g. 6LoWPAN.
1884 * @xfrmdev_ops: Transformation offload operations
1885 * @tlsdev_ops: Transport Layer Security offload operations
1886 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1887 * of Layer 2 headers.
1888 *
1889 * @flags: Interface flags (a la BSD)
1890 * @xdp_features: XDP capability supported by the device
1891 * @priv_flags: Like 'flags' but invisible to userspace,
1892 * see if.h for the definitions
1893 * @gflags: Global flags ( kept as legacy )
1894 * @padded: How much padding added by alloc_netdev()
1895 * @operstate: RFC2863 operstate
1896 * @link_mode: Mapping policy to operstate
1897 * @if_port: Selectable AUI, TP, ...
1898 * @dma: DMA channel
1899 * @mtu: Interface MTU value
1900 * @min_mtu: Interface Minimum MTU value
1901 * @max_mtu: Interface Maximum MTU value
1902 * @type: Interface hardware type
1903 * @hard_header_len: Maximum hardware header length.
1904 * @min_header_len: Minimum hardware header length
1905 *
1906 * @needed_headroom: Extra headroom the hardware may need, but not in all
1907 * cases can this be guaranteed
1908 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1909 * cases can this be guaranteed. Some cases also use
1910 * LL_MAX_HEADER instead to allocate the skb
1911 *
1912 * interface address info:
1913 *
1914 * @perm_addr: Permanent hw address
1915 * @addr_assign_type: Hw address assignment type
1916 * @addr_len: Hardware address length
1917 * @upper_level: Maximum depth level of upper devices.
1918 * @lower_level: Maximum depth level of lower devices.
1919 * @neigh_priv_len: Used in neigh_alloc()
1920 * @dev_id: Used to differentiate devices that share
1921 * the same link layer address
1922 * @dev_port: Used to differentiate devices that share
1923 * the same function
1924 * @addr_list_lock: XXX: need comments on this one
1925 * @name_assign_type: network interface name assignment type
1926 * @uc_promisc: Counter that indicates promiscuous mode
1927 * has been enabled due to the need to listen to
1928 * additional unicast addresses in a device that
1929 * does not implement ndo_set_rx_mode()
1930 * @uc: unicast mac addresses
1931 * @mc: multicast mac addresses
1932 * @dev_addrs: list of device hw addresses
1933 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1934 * @promiscuity: Number of times the NIC is told to work in
1935 * promiscuous mode; if it becomes 0 the NIC will
1936 * exit promiscuous mode
1937 * @allmulti: Counter, enables or disables allmulticast mode
1938 *
1939 * @vlan_info: VLAN info
1940 * @dsa_ptr: dsa specific data
1941 * @tipc_ptr: TIPC specific data
1942 * @atalk_ptr: AppleTalk link
1943 * @ip_ptr: IPv4 specific data
1944 * @ip6_ptr: IPv6 specific data
1945 * @ax25_ptr: AX.25 specific data
1946 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1947 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1948 * device struct
1949 * @mpls_ptr: mpls_dev struct pointer
1950 * @mctp_ptr: MCTP specific data
1951 *
1952 * @dev_addr: Hw address (before bcast,
1953 * because most packets are unicast)
1954 *
1955 * @_rx: Array of RX queues
1956 * @num_rx_queues: Number of RX queues
1957 * allocated at register_netdev() time
1958 * @real_num_rx_queues: Number of RX queues currently active in device
1959 * @xdp_prog: XDP sockets filter program pointer
1960 * @gro_flush_timeout: timeout for GRO layer in NAPI
1961 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1962 * allow to avoid NIC hard IRQ, on busy queues.
1963 *
1964 * @rx_handler: handler for received packets
1965 * @rx_handler_data: XXX: need comments on this one
1966 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1967 * @ingress_queue: XXX: need comments on this one
1968 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1969 * @broadcast: hw bcast address
1970 *
1971 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1972 * indexed by RX queue number. Assigned by driver.
1973 * This must only be set if the ndo_rx_flow_steer
1974 * operation is defined
1975 * @index_hlist: Device index hash chain
1976 *
1977 * @_tx: Array of TX queues
1978 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1979 * @real_num_tx_queues: Number of TX queues currently active in device
1980 * @qdisc: Root qdisc from userspace point of view
1981 * @tx_queue_len: Max frames per queue allowed
1982 * @tx_global_lock: XXX: need comments on this one
1983 * @xdp_bulkq: XDP device bulk queue
1984 * @xps_maps: all CPUs/RXQs maps for XPS device
1985 *
1986 * @xps_maps: XXX: need comments on this one
1987 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1988 * @nf_hooks_egress: netfilter hooks executed for egress packets
1989 * @qdisc_hash: qdisc hash table
1990 * @watchdog_timeo: Represents the timeout that is used by
1991 * the watchdog (see dev_watchdog())
1992 * @watchdog_timer: List of timers
1993 *
1994 * @proto_down_reason: reason a netdev interface is held down
1995 * @pcpu_refcnt: Number of references to this device
1996 * @dev_refcnt: Number of references to this device
1997 * @refcnt_tracker: Tracker directory for tracked references to this device
1998 * @todo_list: Delayed register/unregister
1999 * @link_watch_list: XXX: need comments on this one
2000 *
2001 * @reg_state: Register/unregister state machine
2002 * @dismantle: Device is going to be freed
2003 * @rtnl_link_state: This enum represents the phases of creating
2004 * a new link
2005 *
2006 * @needs_free_netdev: Should unregister perform free_netdev?
2007 * @priv_destructor: Called from unregister
2008 * @npinfo: XXX: need comments on this one
2009 * @nd_net: Network namespace this network device is inside
2010 *
2011 * @ml_priv: Mid-layer private
2012 * @ml_priv_type: Mid-layer private type
2013 *
2014 * @pcpu_stat_type: Type of device statistics which the core should
2015 * allocate/free: none, lstats, tstats, dstats. none
2016 * means the driver is handling statistics allocation/
2017 * freeing internally.
2018 * @lstats: Loopback statistics: packets, bytes
2019 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes
2020 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes
2021 *
2022 * @garp_port: GARP
2023 * @mrp_port: MRP
2024 *
2025 * @dm_private: Drop monitor private
2026 *
2027 * @dev: Class/net/name entry
2028 * @sysfs_groups: Space for optional device, statistics and wireless
2029 * sysfs groups
2030 *
2031 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
2032 * @rtnl_link_ops: Rtnl_link_ops
2033 *
2034 * @gso_max_size: Maximum size of generic segmentation offload
2035 * @tso_max_size: Device (as in HW) limit on the max TSO request size
2036 * @gso_max_segs: Maximum number of segments that can be passed to the
2037 * NIC for GSO
2038 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
2039 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
2040 * for IPv4.
2041 *
2042 * @dcbnl_ops: Data Center Bridging netlink ops
2043 * @num_tc: Number of traffic classes in the net device
2044 * @tc_to_txq: XXX: need comments on this one
2045 * @prio_tc_map: XXX: need comments on this one
2046 *
2047 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
2048 *
2049 * @priomap: XXX: need comments on this one
2050 * @phydev: Physical device may attach itself
2051 * for hardware timestamping
2052 * @sfp_bus: attached &struct sfp_bus structure.
2053 *
2054 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2055 *
2056 * @proto_down: protocol port state information can be sent to the
2057 * switch driver and used to set the phys state of the
2058 * switch port.
2059 *
2060 * @wol_enabled: Wake-on-LAN is enabled
2061 *
2062 * @threaded: napi threaded mode is enabled
2063 *
2064 * @net_notifier_list: List of per-net netdev notifier block
2065 * that follow this device when it is moved
2066 * to another network namespace.
2067 *
2068 * @macsec_ops: MACsec offloading ops
2069 *
2070 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2071 * offload capabilities of the device
2072 * @udp_tunnel_nic: UDP tunnel offload state
2073 * @xdp_state: stores info on attached XDP BPF programs
2074 *
2075 * @nested_level: Used as a parameter of spin_lock_nested() of
2076 * dev->addr_list_lock.
2077 * @unlink_list: As netif_addr_lock() can be called recursively,
2078 * keep a list of interfaces to be deleted.
2079 * @gro_max_size: Maximum size of aggregated packet in generic
2080 * receive offload (GRO)
2081 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2082 * receive offload (GRO), for IPv4.
2083 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2084 * zero copy driver
2085 *
2086 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2087 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2088 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2089 * @dev_registered_tracker: tracker for reference held while
2090 * registered
2091 * @offload_xstats_l3: L3 HW stats for this netdevice.
2092 *
2093 * @devlink_port: Pointer to related devlink port structure.
2094 * Assigned by a driver before netdev registration using
2095 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2096 * during the time netdevice is registered.
2097 *
2098 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2099 * where the clock is recovered.
2100 *
2101 * FIXME: cleanup struct net_device such that network protocol info
2102 * moves out.
2103 */
2104
2105struct net_device {
2106 /* Cacheline organization can be found documented in
2107 * Documentation/networking/net_cachelines/net_device.rst.
2108 * Please update the document when adding new fields.
2109 */
2110
2111 /* TX read-mostly hotpath */
2112 __cacheline_group_begin(net_device_read_tx);
2113 unsigned long long priv_flags;
2114 const struct net_device_ops *netdev_ops;
2115 const struct header_ops *header_ops;
2116 struct netdev_queue *_tx;
2117 netdev_features_t gso_partial_features;
2118 unsigned int real_num_tx_queues;
2119 unsigned int gso_max_size;
2120 unsigned int gso_ipv4_max_size;
2121 u16 gso_max_segs;
2122 s16 num_tc;
2123 /* Note : dev->mtu is often read without holding a lock.
2124 * Writers usually hold RTNL.
2125 * It is recommended to use READ_ONCE() to annotate the reads,
2126 * and to use WRITE_ONCE() to annotate the writes.
2127 */
2128 unsigned int mtu;
2129 unsigned short needed_headroom;
2130 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2131#ifdef CONFIG_XPS
2132 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2133#endif
2134#ifdef CONFIG_NETFILTER_EGRESS
2135 struct nf_hook_entries __rcu *nf_hooks_egress;
2136#endif
2137#ifdef CONFIG_NET_XGRESS
2138 struct bpf_mprog_entry __rcu *tcx_egress;
2139#endif
2140 __cacheline_group_end(net_device_read_tx);
2141
2142 /* TXRX read-mostly hotpath */
2143 __cacheline_group_begin(net_device_read_txrx);
2144 unsigned int flags;
2145 unsigned short hard_header_len;
2146 netdev_features_t features;
2147 struct inet6_dev __rcu *ip6_ptr;
2148 __cacheline_group_end(net_device_read_txrx);
2149
2150 /* RX read-mostly hotpath */
2151 __cacheline_group_begin(net_device_read_rx);
2152 struct bpf_prog __rcu *xdp_prog;
2153 struct list_head ptype_specific;
2154 int ifindex;
2155 unsigned int real_num_rx_queues;
2156 struct netdev_rx_queue *_rx;
2157 unsigned long gro_flush_timeout;
2158 int napi_defer_hard_irqs;
2159 unsigned int gro_max_size;
2160 unsigned int gro_ipv4_max_size;
2161 rx_handler_func_t __rcu *rx_handler;
2162 void __rcu *rx_handler_data;
2163 possible_net_t nd_net;
2164#ifdef CONFIG_NETPOLL
2165 struct netpoll_info __rcu *npinfo;
2166#endif
2167#ifdef CONFIG_NET_XGRESS
2168 struct bpf_mprog_entry __rcu *tcx_ingress;
2169#endif
2170 __cacheline_group_end(net_device_read_rx);
2171
2172 char name[IFNAMSIZ];
2173 struct netdev_name_node *name_node;
2174 struct dev_ifalias __rcu *ifalias;
2175 /*
2176 * I/O specific fields
2177 * FIXME: Merge these and struct ifmap into one
2178 */
2179 unsigned long mem_end;
2180 unsigned long mem_start;
2181 unsigned long base_addr;
2182
2183 /*
2184 * Some hardware also needs these fields (state,dev_list,
2185 * napi_list,unreg_list,close_list) but they are not
2186 * part of the usual set specified in Space.c.
2187 */
2188
2189 unsigned long state;
2190
2191 struct list_head dev_list;
2192 struct list_head napi_list;
2193 struct list_head unreg_list;
2194 struct list_head close_list;
2195 struct list_head ptype_all;
2196
2197 struct {
2198 struct list_head upper;
2199 struct list_head lower;
2200 } adj_list;
2201
2202 /* Read-mostly cache-line for fast-path access */
2203 xdp_features_t xdp_features;
2204 const struct xdp_metadata_ops *xdp_metadata_ops;
2205 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2206 unsigned short gflags;
2207
2208 unsigned short needed_tailroom;
2209
2210 netdev_features_t hw_features;
2211 netdev_features_t wanted_features;
2212 netdev_features_t vlan_features;
2213 netdev_features_t hw_enc_features;
2214 netdev_features_t mpls_features;
2215
2216 unsigned int min_mtu;
2217 unsigned int max_mtu;
2218 unsigned short type;
2219 unsigned char min_header_len;
2220 unsigned char name_assign_type;
2221
2222 int group;
2223
2224 struct net_device_stats stats; /* not used by modern drivers */
2225
2226 struct net_device_core_stats __percpu *core_stats;
2227
2228 /* Stats to monitor link on/off, flapping */
2229 atomic_t carrier_up_count;
2230 atomic_t carrier_down_count;
2231
2232#ifdef CONFIG_WIRELESS_EXT
2233 const struct iw_handler_def *wireless_handlers;
2234 struct iw_public_data *wireless_data;
2235#endif
2236 const struct ethtool_ops *ethtool_ops;
2237#ifdef CONFIG_NET_L3_MASTER_DEV
2238 const struct l3mdev_ops *l3mdev_ops;
2239#endif
2240#if IS_ENABLED(CONFIG_IPV6)
2241 const struct ndisc_ops *ndisc_ops;
2242#endif
2243
2244#ifdef CONFIG_XFRM_OFFLOAD
2245 const struct xfrmdev_ops *xfrmdev_ops;
2246#endif
2247
2248#if IS_ENABLED(CONFIG_TLS_DEVICE)
2249 const struct tlsdev_ops *tlsdev_ops;
2250#endif
2251
2252 unsigned char operstate;
2253 unsigned char link_mode;
2254
2255 unsigned char if_port;
2256 unsigned char dma;
2257
2258 /* Interface address info. */
2259 unsigned char perm_addr[MAX_ADDR_LEN];
2260 unsigned char addr_assign_type;
2261 unsigned char addr_len;
2262 unsigned char upper_level;
2263 unsigned char lower_level;
2264
2265 unsigned short neigh_priv_len;
2266 unsigned short dev_id;
2267 unsigned short dev_port;
2268 unsigned short padded;
2269
2270 spinlock_t addr_list_lock;
2271 int irq;
2272
2273 struct netdev_hw_addr_list uc;
2274 struct netdev_hw_addr_list mc;
2275 struct netdev_hw_addr_list dev_addrs;
2276
2277#ifdef CONFIG_SYSFS
2278 struct kset *queues_kset;
2279#endif
2280#ifdef CONFIG_LOCKDEP
2281 struct list_head unlink_list;
2282#endif
2283 unsigned int promiscuity;
2284 unsigned int allmulti;
2285 bool uc_promisc;
2286#ifdef CONFIG_LOCKDEP
2287 unsigned char nested_level;
2288#endif
2289
2290
2291 /* Protocol-specific pointers */
2292 struct in_device __rcu *ip_ptr;
2293#if IS_ENABLED(CONFIG_VLAN_8021Q)
2294 struct vlan_info __rcu *vlan_info;
2295#endif
2296#if IS_ENABLED(CONFIG_NET_DSA)
2297 struct dsa_port *dsa_ptr;
2298#endif
2299#if IS_ENABLED(CONFIG_TIPC)
2300 struct tipc_bearer __rcu *tipc_ptr;
2301#endif
2302#if IS_ENABLED(CONFIG_ATALK)
2303 void *atalk_ptr;
2304#endif
2305#if IS_ENABLED(CONFIG_AX25)
2306 void *ax25_ptr;
2307#endif
2308#if IS_ENABLED(CONFIG_CFG80211)
2309 struct wireless_dev *ieee80211_ptr;
2310#endif
2311#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2312 struct wpan_dev *ieee802154_ptr;
2313#endif
2314#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2315 struct mpls_dev __rcu *mpls_ptr;
2316#endif
2317#if IS_ENABLED(CONFIG_MCTP)
2318 struct mctp_dev __rcu *mctp_ptr;
2319#endif
2320
2321/*
2322 * Cache lines mostly used on receive path (including eth_type_trans())
2323 */
2324 /* Interface address info used in eth_type_trans() */
2325 const unsigned char *dev_addr;
2326
2327 unsigned int num_rx_queues;
2328#define GRO_LEGACY_MAX_SIZE 65536u
2329/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2330 * and shinfo->gso_segs is a 16bit field.
2331 */
2332#define GRO_MAX_SIZE (8 * 65535u)
2333 unsigned int xdp_zc_max_segs;
2334 struct netdev_queue __rcu *ingress_queue;
2335#ifdef CONFIG_NETFILTER_INGRESS
2336 struct nf_hook_entries __rcu *nf_hooks_ingress;
2337#endif
2338
2339 unsigned char broadcast[MAX_ADDR_LEN];
2340#ifdef CONFIG_RFS_ACCEL
2341 struct cpu_rmap *rx_cpu_rmap;
2342#endif
2343 struct hlist_node index_hlist;
2344
2345/*
2346 * Cache lines mostly used on transmit path
2347 */
2348 unsigned int num_tx_queues;
2349 struct Qdisc __rcu *qdisc;
2350 unsigned int tx_queue_len;
2351 spinlock_t tx_global_lock;
2352
2353 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2354
2355#ifdef CONFIG_NET_SCHED
2356 DECLARE_HASHTABLE (qdisc_hash, 4);
2357#endif
2358 /* These may be needed for future network-power-down code. */
2359 struct timer_list watchdog_timer;
2360 int watchdog_timeo;
2361
2362 u32 proto_down_reason;
2363
2364 struct list_head todo_list;
2365
2366#ifdef CONFIG_PCPU_DEV_REFCNT
2367 int __percpu *pcpu_refcnt;
2368#else
2369 refcount_t dev_refcnt;
2370#endif
2371 struct ref_tracker_dir refcnt_tracker;
2372
2373 struct list_head link_watch_list;
2374
2375 enum { NETREG_UNINITIALIZED=0,
2376 NETREG_REGISTERED, /* completed register_netdevice */
2377 NETREG_UNREGISTERING, /* called unregister_netdevice */
2378 NETREG_UNREGISTERED, /* completed unregister todo */
2379 NETREG_RELEASED, /* called free_netdev */
2380 NETREG_DUMMY, /* dummy device for NAPI poll */
2381 } reg_state:8;
2382
2383 bool dismantle;
2384
2385 enum {
2386 RTNL_LINK_INITIALIZED,
2387 RTNL_LINK_INITIALIZING,
2388 } rtnl_link_state:16;
2389
2390 bool needs_free_netdev;
2391 void (*priv_destructor)(struct net_device *dev);
2392
2393 /* mid-layer private */
2394 void *ml_priv;
2395 enum netdev_ml_priv_type ml_priv_type;
2396
2397 enum netdev_stat_type pcpu_stat_type:8;
2398 union {
2399 struct pcpu_lstats __percpu *lstats;
2400 struct pcpu_sw_netstats __percpu *tstats;
2401 struct pcpu_dstats __percpu *dstats;
2402 };
2403
2404#if IS_ENABLED(CONFIG_GARP)
2405 struct garp_port __rcu *garp_port;
2406#endif
2407#if IS_ENABLED(CONFIG_MRP)
2408 struct mrp_port __rcu *mrp_port;
2409#endif
2410#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2411 struct dm_hw_stat_delta __rcu *dm_private;
2412#endif
2413 struct device dev;
2414 const struct attribute_group *sysfs_groups[4];
2415 const struct attribute_group *sysfs_rx_queue_group;
2416
2417 const struct rtnl_link_ops *rtnl_link_ops;
2418
2419 /* for setting kernel sock attribute on TCP connection setup */
2420#define GSO_MAX_SEGS 65535u
2421#define GSO_LEGACY_MAX_SIZE 65536u
2422/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2423 * and shinfo->gso_segs is a 16bit field.
2424 */
2425#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2426
2427#define TSO_LEGACY_MAX_SIZE 65536
2428#define TSO_MAX_SIZE UINT_MAX
2429 unsigned int tso_max_size;
2430#define TSO_MAX_SEGS U16_MAX
2431 u16 tso_max_segs;
2432
2433#ifdef CONFIG_DCB
2434 const struct dcbnl_rtnl_ops *dcbnl_ops;
2435#endif
2436 u8 prio_tc_map[TC_BITMASK + 1];
2437
2438#if IS_ENABLED(CONFIG_FCOE)
2439 unsigned int fcoe_ddp_xid;
2440#endif
2441#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2442 struct netprio_map __rcu *priomap;
2443#endif
2444 struct phy_device *phydev;
2445 struct sfp_bus *sfp_bus;
2446 struct lock_class_key *qdisc_tx_busylock;
2447 bool proto_down;
2448 unsigned wol_enabled:1;
2449 unsigned threaded:1;
2450
2451 struct list_head net_notifier_list;
2452
2453#if IS_ENABLED(CONFIG_MACSEC)
2454 /* MACsec management functions */
2455 const struct macsec_ops *macsec_ops;
2456#endif
2457 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2458 struct udp_tunnel_nic *udp_tunnel_nic;
2459
2460 /* protected by rtnl_lock */
2461 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2462
2463 u8 dev_addr_shadow[MAX_ADDR_LEN];
2464 netdevice_tracker linkwatch_dev_tracker;
2465 netdevice_tracker watchdog_dev_tracker;
2466 netdevice_tracker dev_registered_tracker;
2467 struct rtnl_hw_stats64 *offload_xstats_l3;
2468
2469 struct devlink_port *devlink_port;
2470
2471#if IS_ENABLED(CONFIG_DPLL)
2472 struct dpll_pin *dpll_pin;
2473#endif
2474#if IS_ENABLED(CONFIG_PAGE_POOL)
2475 /** @page_pools: page pools created for this netdevice */
2476 struct hlist_head page_pools;
2477#endif
2478};
2479#define to_net_dev(d) container_of(d, struct net_device, dev)
2480
2481/*
2482 * Driver should use this to assign devlink port instance to a netdevice
2483 * before it registers the netdevice. Therefore devlink_port is static
2484 * during the netdev lifetime after it is registered.
2485 */
2486#define SET_NETDEV_DEVLINK_PORT(dev, port) \
2487({ \
2488 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2489 ((dev)->devlink_port = (port)); \
2490})
2491
2492static inline bool netif_elide_gro(const struct net_device *dev)
2493{
2494 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2495 return true;
2496 return false;
2497}
2498
2499#define NETDEV_ALIGN 32
2500
2501static inline
2502int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2503{
2504 return dev->prio_tc_map[prio & TC_BITMASK];
2505}
2506
2507static inline
2508int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2509{
2510 if (tc >= dev->num_tc)
2511 return -EINVAL;
2512
2513 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2514 return 0;
2515}
2516
2517int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2518void netdev_reset_tc(struct net_device *dev);
2519int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2520int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2521
2522static inline
2523int netdev_get_num_tc(struct net_device *dev)
2524{
2525 return dev->num_tc;
2526}
2527
2528static inline void net_prefetch(void *p)
2529{
2530 prefetch(p);
2531#if L1_CACHE_BYTES < 128
2532 prefetch((u8 *)p + L1_CACHE_BYTES);
2533#endif
2534}
2535
2536static inline void net_prefetchw(void *p)
2537{
2538 prefetchw(p);
2539#if L1_CACHE_BYTES < 128
2540 prefetchw((u8 *)p + L1_CACHE_BYTES);
2541#endif
2542}
2543
2544void netdev_unbind_sb_channel(struct net_device *dev,
2545 struct net_device *sb_dev);
2546int netdev_bind_sb_channel_queue(struct net_device *dev,
2547 struct net_device *sb_dev,
2548 u8 tc, u16 count, u16 offset);
2549int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2550static inline int netdev_get_sb_channel(struct net_device *dev)
2551{
2552 return max_t(int, -dev->num_tc, 0);
2553}
2554
2555static inline
2556struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2557 unsigned int index)
2558{
2559 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2560 return &dev->_tx[index];
2561}
2562
2563static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2564 const struct sk_buff *skb)
2565{
2566 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2567}
2568
2569static inline void netdev_for_each_tx_queue(struct net_device *dev,
2570 void (*f)(struct net_device *,
2571 struct netdev_queue *,
2572 void *),
2573 void *arg)
2574{
2575 unsigned int i;
2576
2577 for (i = 0; i < dev->num_tx_queues; i++)
2578 f(dev, &dev->_tx[i], arg);
2579}
2580
2581#define netdev_lockdep_set_classes(dev) \
2582{ \
2583 static struct lock_class_key qdisc_tx_busylock_key; \
2584 static struct lock_class_key qdisc_xmit_lock_key; \
2585 static struct lock_class_key dev_addr_list_lock_key; \
2586 unsigned int i; \
2587 \
2588 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2589 lockdep_set_class(&(dev)->addr_list_lock, \
2590 &dev_addr_list_lock_key); \
2591 for (i = 0; i < (dev)->num_tx_queues; i++) \
2592 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2593 &qdisc_xmit_lock_key); \
2594}
2595
2596u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2597 struct net_device *sb_dev);
2598struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2599 struct sk_buff *skb,
2600 struct net_device *sb_dev);
2601
2602/* returns the headroom that the master device needs to take in account
2603 * when forwarding to this dev
2604 */
2605static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2606{
2607 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2608}
2609
2610static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2611{
2612 if (dev->netdev_ops->ndo_set_rx_headroom)
2613 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2614}
2615
2616/* set the device rx headroom to the dev's default */
2617static inline void netdev_reset_rx_headroom(struct net_device *dev)
2618{
2619 netdev_set_rx_headroom(dev, -1);
2620}
2621
2622static inline void *netdev_get_ml_priv(struct net_device *dev,
2623 enum netdev_ml_priv_type type)
2624{
2625 if (dev->ml_priv_type != type)
2626 return NULL;
2627
2628 return dev->ml_priv;
2629}
2630
2631static inline void netdev_set_ml_priv(struct net_device *dev,
2632 void *ml_priv,
2633 enum netdev_ml_priv_type type)
2634{
2635 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2636 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2637 dev->ml_priv_type, type);
2638 WARN(!dev->ml_priv_type && dev->ml_priv,
2639 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2640
2641 dev->ml_priv = ml_priv;
2642 dev->ml_priv_type = type;
2643}
2644
2645/*
2646 * Net namespace inlines
2647 */
2648static inline
2649struct net *dev_net(const struct net_device *dev)
2650{
2651 return read_pnet(&dev->nd_net);
2652}
2653
2654static inline
2655void dev_net_set(struct net_device *dev, struct net *net)
2656{
2657 write_pnet(&dev->nd_net, net);
2658}
2659
2660/**
2661 * netdev_priv - access network device private data
2662 * @dev: network device
2663 *
2664 * Get network device private data
2665 */
2666static inline void *netdev_priv(const struct net_device *dev)
2667{
2668 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2669}
2670
2671/* Set the sysfs physical device reference for the network logical device
2672 * if set prior to registration will cause a symlink during initialization.
2673 */
2674#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2675
2676/* Set the sysfs device type for the network logical device to allow
2677 * fine-grained identification of different network device types. For
2678 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2679 */
2680#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2681
2682void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2683 enum netdev_queue_type type,
2684 struct napi_struct *napi);
2685
2686static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2687{
2688 napi->irq = irq;
2689}
2690
2691/* Default NAPI poll() weight
2692 * Device drivers are strongly advised to not use bigger value
2693 */
2694#define NAPI_POLL_WEIGHT 64
2695
2696void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2697 int (*poll)(struct napi_struct *, int), int weight);
2698
2699/**
2700 * netif_napi_add() - initialize a NAPI context
2701 * @dev: network device
2702 * @napi: NAPI context
2703 * @poll: polling function
2704 *
2705 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2706 * *any* of the other NAPI-related functions.
2707 */
2708static inline void
2709netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2710 int (*poll)(struct napi_struct *, int))
2711{
2712 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2713}
2714
2715static inline void
2716netif_napi_add_tx_weight(struct net_device *dev,
2717 struct napi_struct *napi,
2718 int (*poll)(struct napi_struct *, int),
2719 int weight)
2720{
2721 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2722 netif_napi_add_weight(dev, napi, poll, weight);
2723}
2724
2725/**
2726 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2727 * @dev: network device
2728 * @napi: NAPI context
2729 * @poll: polling function
2730 *
2731 * This variant of netif_napi_add() should be used from drivers using NAPI
2732 * to exclusively poll a TX queue.
2733 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2734 */
2735static inline void netif_napi_add_tx(struct net_device *dev,
2736 struct napi_struct *napi,
2737 int (*poll)(struct napi_struct *, int))
2738{
2739 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2740}
2741
2742/**
2743 * __netif_napi_del - remove a NAPI context
2744 * @napi: NAPI context
2745 *
2746 * Warning: caller must observe RCU grace period before freeing memory
2747 * containing @napi. Drivers might want to call this helper to combine
2748 * all the needed RCU grace periods into a single one.
2749 */
2750void __netif_napi_del(struct napi_struct *napi);
2751
2752/**
2753 * netif_napi_del - remove a NAPI context
2754 * @napi: NAPI context
2755 *
2756 * netif_napi_del() removes a NAPI context from the network device NAPI list
2757 */
2758static inline void netif_napi_del(struct napi_struct *napi)
2759{
2760 __netif_napi_del(napi);
2761 synchronize_net();
2762}
2763
2764struct packet_type {
2765 __be16 type; /* This is really htons(ether_type). */
2766 bool ignore_outgoing;
2767 struct net_device *dev; /* NULL is wildcarded here */
2768 netdevice_tracker dev_tracker;
2769 int (*func) (struct sk_buff *,
2770 struct net_device *,
2771 struct packet_type *,
2772 struct net_device *);
2773 void (*list_func) (struct list_head *,
2774 struct packet_type *,
2775 struct net_device *);
2776 bool (*id_match)(struct packet_type *ptype,
2777 struct sock *sk);
2778 struct net *af_packet_net;
2779 void *af_packet_priv;
2780 struct list_head list;
2781};
2782
2783struct offload_callbacks {
2784 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2785 netdev_features_t features);
2786 struct sk_buff *(*gro_receive)(struct list_head *head,
2787 struct sk_buff *skb);
2788 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2789};
2790
2791struct packet_offload {
2792 __be16 type; /* This is really htons(ether_type). */
2793 u16 priority;
2794 struct offload_callbacks callbacks;
2795 struct list_head list;
2796};
2797
2798/* often modified stats are per-CPU, other are shared (netdev->stats) */
2799struct pcpu_sw_netstats {
2800 u64_stats_t rx_packets;
2801 u64_stats_t rx_bytes;
2802 u64_stats_t tx_packets;
2803 u64_stats_t tx_bytes;
2804 struct u64_stats_sync syncp;
2805} __aligned(4 * sizeof(u64));
2806
2807struct pcpu_dstats {
2808 u64 rx_packets;
2809 u64 rx_bytes;
2810 u64 rx_drops;
2811 u64 tx_packets;
2812 u64 tx_bytes;
2813 u64 tx_drops;
2814 struct u64_stats_sync syncp;
2815} __aligned(8 * sizeof(u64));
2816
2817struct pcpu_lstats {
2818 u64_stats_t packets;
2819 u64_stats_t bytes;
2820 struct u64_stats_sync syncp;
2821} __aligned(2 * sizeof(u64));
2822
2823void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2824
2825static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2826{
2827 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2828
2829 u64_stats_update_begin(&tstats->syncp);
2830 u64_stats_add(&tstats->rx_bytes, len);
2831 u64_stats_inc(&tstats->rx_packets);
2832 u64_stats_update_end(&tstats->syncp);
2833}
2834
2835static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2836 unsigned int packets,
2837 unsigned int len)
2838{
2839 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2840
2841 u64_stats_update_begin(&tstats->syncp);
2842 u64_stats_add(&tstats->tx_bytes, len);
2843 u64_stats_add(&tstats->tx_packets, packets);
2844 u64_stats_update_end(&tstats->syncp);
2845}
2846
2847static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2848{
2849 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2850
2851 u64_stats_update_begin(&lstats->syncp);
2852 u64_stats_add(&lstats->bytes, len);
2853 u64_stats_inc(&lstats->packets);
2854 u64_stats_update_end(&lstats->syncp);
2855}
2856
2857#define __netdev_alloc_pcpu_stats(type, gfp) \
2858({ \
2859 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2860 if (pcpu_stats) { \
2861 int __cpu; \
2862 for_each_possible_cpu(__cpu) { \
2863 typeof(type) *stat; \
2864 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2865 u64_stats_init(&stat->syncp); \
2866 } \
2867 } \
2868 pcpu_stats; \
2869})
2870
2871#define netdev_alloc_pcpu_stats(type) \
2872 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2873
2874#define devm_netdev_alloc_pcpu_stats(dev, type) \
2875({ \
2876 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2877 if (pcpu_stats) { \
2878 int __cpu; \
2879 for_each_possible_cpu(__cpu) { \
2880 typeof(type) *stat; \
2881 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2882 u64_stats_init(&stat->syncp); \
2883 } \
2884 } \
2885 pcpu_stats; \
2886})
2887
2888enum netdev_lag_tx_type {
2889 NETDEV_LAG_TX_TYPE_UNKNOWN,
2890 NETDEV_LAG_TX_TYPE_RANDOM,
2891 NETDEV_LAG_TX_TYPE_BROADCAST,
2892 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2893 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2894 NETDEV_LAG_TX_TYPE_HASH,
2895};
2896
2897enum netdev_lag_hash {
2898 NETDEV_LAG_HASH_NONE,
2899 NETDEV_LAG_HASH_L2,
2900 NETDEV_LAG_HASH_L34,
2901 NETDEV_LAG_HASH_L23,
2902 NETDEV_LAG_HASH_E23,
2903 NETDEV_LAG_HASH_E34,
2904 NETDEV_LAG_HASH_VLAN_SRCMAC,
2905 NETDEV_LAG_HASH_UNKNOWN,
2906};
2907
2908struct netdev_lag_upper_info {
2909 enum netdev_lag_tx_type tx_type;
2910 enum netdev_lag_hash hash_type;
2911};
2912
2913struct netdev_lag_lower_state_info {
2914 u8 link_up : 1,
2915 tx_enabled : 1;
2916};
2917
2918#include <linux/notifier.h>
2919
2920/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2921 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2922 * adding new types.
2923 */
2924enum netdev_cmd {
2925 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2926 NETDEV_DOWN,
2927 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2928 detected a hardware crash and restarted
2929 - we can use this eg to kick tcp sessions
2930 once done */
2931 NETDEV_CHANGE, /* Notify device state change */
2932 NETDEV_REGISTER,
2933 NETDEV_UNREGISTER,
2934 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2935 NETDEV_CHANGEADDR, /* notify after the address change */
2936 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2937 NETDEV_GOING_DOWN,
2938 NETDEV_CHANGENAME,
2939 NETDEV_FEAT_CHANGE,
2940 NETDEV_BONDING_FAILOVER,
2941 NETDEV_PRE_UP,
2942 NETDEV_PRE_TYPE_CHANGE,
2943 NETDEV_POST_TYPE_CHANGE,
2944 NETDEV_POST_INIT,
2945 NETDEV_PRE_UNINIT,
2946 NETDEV_RELEASE,
2947 NETDEV_NOTIFY_PEERS,
2948 NETDEV_JOIN,
2949 NETDEV_CHANGEUPPER,
2950 NETDEV_RESEND_IGMP,
2951 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2952 NETDEV_CHANGEINFODATA,
2953 NETDEV_BONDING_INFO,
2954 NETDEV_PRECHANGEUPPER,
2955 NETDEV_CHANGELOWERSTATE,
2956 NETDEV_UDP_TUNNEL_PUSH_INFO,
2957 NETDEV_UDP_TUNNEL_DROP_INFO,
2958 NETDEV_CHANGE_TX_QUEUE_LEN,
2959 NETDEV_CVLAN_FILTER_PUSH_INFO,
2960 NETDEV_CVLAN_FILTER_DROP_INFO,
2961 NETDEV_SVLAN_FILTER_PUSH_INFO,
2962 NETDEV_SVLAN_FILTER_DROP_INFO,
2963 NETDEV_OFFLOAD_XSTATS_ENABLE,
2964 NETDEV_OFFLOAD_XSTATS_DISABLE,
2965 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2966 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2967 NETDEV_XDP_FEAT_CHANGE,
2968};
2969const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2970
2971int register_netdevice_notifier(struct notifier_block *nb);
2972int unregister_netdevice_notifier(struct notifier_block *nb);
2973int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2974int unregister_netdevice_notifier_net(struct net *net,
2975 struct notifier_block *nb);
2976int register_netdevice_notifier_dev_net(struct net_device *dev,
2977 struct notifier_block *nb,
2978 struct netdev_net_notifier *nn);
2979int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2980 struct notifier_block *nb,
2981 struct netdev_net_notifier *nn);
2982
2983struct netdev_notifier_info {
2984 struct net_device *dev;
2985 struct netlink_ext_ack *extack;
2986};
2987
2988struct netdev_notifier_info_ext {
2989 struct netdev_notifier_info info; /* must be first */
2990 union {
2991 u32 mtu;
2992 } ext;
2993};
2994
2995struct netdev_notifier_change_info {
2996 struct netdev_notifier_info info; /* must be first */
2997 unsigned int flags_changed;
2998};
2999
3000struct netdev_notifier_changeupper_info {
3001 struct netdev_notifier_info info; /* must be first */
3002 struct net_device *upper_dev; /* new upper dev */
3003 bool master; /* is upper dev master */
3004 bool linking; /* is the notification for link or unlink */
3005 void *upper_info; /* upper dev info */
3006};
3007
3008struct netdev_notifier_changelowerstate_info {
3009 struct netdev_notifier_info info; /* must be first */
3010 void *lower_state_info; /* is lower dev state */
3011};
3012
3013struct netdev_notifier_pre_changeaddr_info {
3014 struct netdev_notifier_info info; /* must be first */
3015 const unsigned char *dev_addr;
3016};
3017
3018enum netdev_offload_xstats_type {
3019 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
3020};
3021
3022struct netdev_notifier_offload_xstats_info {
3023 struct netdev_notifier_info info; /* must be first */
3024 enum netdev_offload_xstats_type type;
3025
3026 union {
3027 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
3028 struct netdev_notifier_offload_xstats_rd *report_delta;
3029 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
3030 struct netdev_notifier_offload_xstats_ru *report_used;
3031 };
3032};
3033
3034int netdev_offload_xstats_enable(struct net_device *dev,
3035 enum netdev_offload_xstats_type type,
3036 struct netlink_ext_ack *extack);
3037int netdev_offload_xstats_disable(struct net_device *dev,
3038 enum netdev_offload_xstats_type type);
3039bool netdev_offload_xstats_enabled(const struct net_device *dev,
3040 enum netdev_offload_xstats_type type);
3041int netdev_offload_xstats_get(struct net_device *dev,
3042 enum netdev_offload_xstats_type type,
3043 struct rtnl_hw_stats64 *stats, bool *used,
3044 struct netlink_ext_ack *extack);
3045void
3046netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3047 const struct rtnl_hw_stats64 *stats);
3048void
3049netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3050void netdev_offload_xstats_push_delta(struct net_device *dev,
3051 enum netdev_offload_xstats_type type,
3052 const struct rtnl_hw_stats64 *stats);
3053
3054static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3055 struct net_device *dev)
3056{
3057 info->dev = dev;
3058 info->extack = NULL;
3059}
3060
3061static inline struct net_device *
3062netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3063{
3064 return info->dev;
3065}
3066
3067static inline struct netlink_ext_ack *
3068netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3069{
3070 return info->extack;
3071}
3072
3073int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3074int call_netdevice_notifiers_info(unsigned long val,
3075 struct netdev_notifier_info *info);
3076
3077extern rwlock_t dev_base_lock; /* Device list lock */
3078
3079#define for_each_netdev(net, d) \
3080 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3081#define for_each_netdev_reverse(net, d) \
3082 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3083#define for_each_netdev_rcu(net, d) \
3084 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3085#define for_each_netdev_safe(net, d, n) \
3086 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3087#define for_each_netdev_continue(net, d) \
3088 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3089#define for_each_netdev_continue_reverse(net, d) \
3090 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3091 dev_list)
3092#define for_each_netdev_continue_rcu(net, d) \
3093 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3094#define for_each_netdev_in_bond_rcu(bond, slave) \
3095 for_each_netdev_rcu(&init_net, slave) \
3096 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3097#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3098
3099#define for_each_netdev_dump(net, d, ifindex) \
3100 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3101
3102static inline struct net_device *next_net_device(struct net_device *dev)
3103{
3104 struct list_head *lh;
3105 struct net *net;
3106
3107 net = dev_net(dev);
3108 lh = dev->dev_list.next;
3109 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3110}
3111
3112static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3113{
3114 struct list_head *lh;
3115 struct net *net;
3116
3117 net = dev_net(dev);
3118 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3119 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3120}
3121
3122static inline struct net_device *first_net_device(struct net *net)
3123{
3124 return list_empty(&net->dev_base_head) ? NULL :
3125 net_device_entry(net->dev_base_head.next);
3126}
3127
3128static inline struct net_device *first_net_device_rcu(struct net *net)
3129{
3130 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3131
3132 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3133}
3134
3135int netdev_boot_setup_check(struct net_device *dev);
3136struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3137 const char *hwaddr);
3138struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3139void dev_add_pack(struct packet_type *pt);
3140void dev_remove_pack(struct packet_type *pt);
3141void __dev_remove_pack(struct packet_type *pt);
3142void dev_add_offload(struct packet_offload *po);
3143void dev_remove_offload(struct packet_offload *po);
3144
3145int dev_get_iflink(const struct net_device *dev);
3146int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3147int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3148 struct net_device_path_stack *stack);
3149struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3150 unsigned short mask);
3151struct net_device *dev_get_by_name(struct net *net, const char *name);
3152struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3153struct net_device *__dev_get_by_name(struct net *net, const char *name);
3154bool netdev_name_in_use(struct net *net, const char *name);
3155int dev_alloc_name(struct net_device *dev, const char *name);
3156int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3157void dev_close(struct net_device *dev);
3158void dev_close_many(struct list_head *head, bool unlink);
3159void dev_disable_lro(struct net_device *dev);
3160int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3161u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3162 struct net_device *sb_dev);
3163u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3164 struct net_device *sb_dev);
3165
3166int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3167int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3168
3169static inline int dev_queue_xmit(struct sk_buff *skb)
3170{
3171 return __dev_queue_xmit(skb, NULL);
3172}
3173
3174static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3175 struct net_device *sb_dev)
3176{
3177 return __dev_queue_xmit(skb, sb_dev);
3178}
3179
3180static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3181{
3182 int ret;
3183
3184 ret = __dev_direct_xmit(skb, queue_id);
3185 if (!dev_xmit_complete(ret))
3186 kfree_skb(skb);
3187 return ret;
3188}
3189
3190int register_netdevice(struct net_device *dev);
3191void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3192void unregister_netdevice_many(struct list_head *head);
3193static inline void unregister_netdevice(struct net_device *dev)
3194{
3195 unregister_netdevice_queue(dev, NULL);
3196}
3197
3198int netdev_refcnt_read(const struct net_device *dev);
3199void free_netdev(struct net_device *dev);
3200void netdev_freemem(struct net_device *dev);
3201int init_dummy_netdev(struct net_device *dev);
3202
3203struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3204 struct sk_buff *skb,
3205 bool all_slaves);
3206struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3207 struct sock *sk);
3208struct net_device *dev_get_by_index(struct net *net, int ifindex);
3209struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3210struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3211 netdevice_tracker *tracker, gfp_t gfp);
3212struct net_device *netdev_get_by_name(struct net *net, const char *name,
3213 netdevice_tracker *tracker, gfp_t gfp);
3214struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3215struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3216
3217static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3218 unsigned short type,
3219 const void *daddr, const void *saddr,
3220 unsigned int len)
3221{
3222 if (!dev->header_ops || !dev->header_ops->create)
3223 return 0;
3224
3225 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3226}
3227
3228static inline int dev_parse_header(const struct sk_buff *skb,
3229 unsigned char *haddr)
3230{
3231 const struct net_device *dev = skb->dev;
3232
3233 if (!dev->header_ops || !dev->header_ops->parse)
3234 return 0;
3235 return dev->header_ops->parse(skb, haddr);
3236}
3237
3238static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3239{
3240 const struct net_device *dev = skb->dev;
3241
3242 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3243 return 0;
3244 return dev->header_ops->parse_protocol(skb);
3245}
3246
3247/* ll_header must have at least hard_header_len allocated */
3248static inline bool dev_validate_header(const struct net_device *dev,
3249 char *ll_header, int len)
3250{
3251 if (likely(len >= dev->hard_header_len))
3252 return true;
3253 if (len < dev->min_header_len)
3254 return false;
3255
3256 if (capable(CAP_SYS_RAWIO)) {
3257 memset(ll_header + len, 0, dev->hard_header_len - len);
3258 return true;
3259 }
3260
3261 if (dev->header_ops && dev->header_ops->validate)
3262 return dev->header_ops->validate(ll_header, len);
3263
3264 return false;
3265}
3266
3267static inline bool dev_has_header(const struct net_device *dev)
3268{
3269 return dev->header_ops && dev->header_ops->create;
3270}
3271
3272/*
3273 * Incoming packets are placed on per-CPU queues
3274 */
3275struct softnet_data {
3276 struct list_head poll_list;
3277 struct sk_buff_head process_queue;
3278
3279 /* stats */
3280 unsigned int processed;
3281 unsigned int time_squeeze;
3282#ifdef CONFIG_RPS
3283 struct softnet_data *rps_ipi_list;
3284#endif
3285
3286 bool in_net_rx_action;
3287 bool in_napi_threaded_poll;
3288
3289#ifdef CONFIG_NET_FLOW_LIMIT
3290 struct sd_flow_limit __rcu *flow_limit;
3291#endif
3292 struct Qdisc *output_queue;
3293 struct Qdisc **output_queue_tailp;
3294 struct sk_buff *completion_queue;
3295#ifdef CONFIG_XFRM_OFFLOAD
3296 struct sk_buff_head xfrm_backlog;
3297#endif
3298 /* written and read only by owning cpu: */
3299 struct {
3300 u16 recursion;
3301 u8 more;
3302#ifdef CONFIG_NET_EGRESS
3303 u8 skip_txqueue;
3304#endif
3305 } xmit;
3306#ifdef CONFIG_RPS
3307 /* input_queue_head should be written by cpu owning this struct,
3308 * and only read by other cpus. Worth using a cache line.
3309 */
3310 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3311
3312 /* Elements below can be accessed between CPUs for RPS/RFS */
3313 call_single_data_t csd ____cacheline_aligned_in_smp;
3314 struct softnet_data *rps_ipi_next;
3315 unsigned int cpu;
3316 unsigned int input_queue_tail;
3317#endif
3318 unsigned int received_rps;
3319 unsigned int dropped;
3320 struct sk_buff_head input_pkt_queue;
3321 struct napi_struct backlog;
3322
3323 /* Another possibly contended cache line */
3324 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3325 int defer_count;
3326 int defer_ipi_scheduled;
3327 struct sk_buff *defer_list;
3328 call_single_data_t defer_csd;
3329};
3330
3331static inline void input_queue_head_incr(struct softnet_data *sd)
3332{
3333#ifdef CONFIG_RPS
3334 sd->input_queue_head++;
3335#endif
3336}
3337
3338static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3339 unsigned int *qtail)
3340{
3341#ifdef CONFIG_RPS
3342 *qtail = ++sd->input_queue_tail;
3343#endif
3344}
3345
3346DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3347
3348static inline int dev_recursion_level(void)
3349{
3350 return this_cpu_read(softnet_data.xmit.recursion);
3351}
3352
3353#define XMIT_RECURSION_LIMIT 8
3354static inline bool dev_xmit_recursion(void)
3355{
3356 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3357 XMIT_RECURSION_LIMIT);
3358}
3359
3360static inline void dev_xmit_recursion_inc(void)
3361{
3362 __this_cpu_inc(softnet_data.xmit.recursion);
3363}
3364
3365static inline void dev_xmit_recursion_dec(void)
3366{
3367 __this_cpu_dec(softnet_data.xmit.recursion);
3368}
3369
3370void __netif_schedule(struct Qdisc *q);
3371void netif_schedule_queue(struct netdev_queue *txq);
3372
3373static inline void netif_tx_schedule_all(struct net_device *dev)
3374{
3375 unsigned int i;
3376
3377 for (i = 0; i < dev->num_tx_queues; i++)
3378 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3379}
3380
3381static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3382{
3383 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3384}
3385
3386/**
3387 * netif_start_queue - allow transmit
3388 * @dev: network device
3389 *
3390 * Allow upper layers to call the device hard_start_xmit routine.
3391 */
3392static inline void netif_start_queue(struct net_device *dev)
3393{
3394 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3395}
3396
3397static inline void netif_tx_start_all_queues(struct net_device *dev)
3398{
3399 unsigned int i;
3400
3401 for (i = 0; i < dev->num_tx_queues; i++) {
3402 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3403 netif_tx_start_queue(txq);
3404 }
3405}
3406
3407void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3408
3409/**
3410 * netif_wake_queue - restart transmit
3411 * @dev: network device
3412 *
3413 * Allow upper layers to call the device hard_start_xmit routine.
3414 * Used for flow control when transmit resources are available.
3415 */
3416static inline void netif_wake_queue(struct net_device *dev)
3417{
3418 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3419}
3420
3421static inline void netif_tx_wake_all_queues(struct net_device *dev)
3422{
3423 unsigned int i;
3424
3425 for (i = 0; i < dev->num_tx_queues; i++) {
3426 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3427 netif_tx_wake_queue(txq);
3428 }
3429}
3430
3431static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3432{
3433 /* Must be an atomic op see netif_txq_try_stop() */
3434 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3435}
3436
3437/**
3438 * netif_stop_queue - stop transmitted packets
3439 * @dev: network device
3440 *
3441 * Stop upper layers calling the device hard_start_xmit routine.
3442 * Used for flow control when transmit resources are unavailable.
3443 */
3444static inline void netif_stop_queue(struct net_device *dev)
3445{
3446 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3447}
3448
3449void netif_tx_stop_all_queues(struct net_device *dev);
3450
3451static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3452{
3453 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3454}
3455
3456/**
3457 * netif_queue_stopped - test if transmit queue is flowblocked
3458 * @dev: network device
3459 *
3460 * Test if transmit queue on device is currently unable to send.
3461 */
3462static inline bool netif_queue_stopped(const struct net_device *dev)
3463{
3464 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3465}
3466
3467static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3468{
3469 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3470}
3471
3472static inline bool
3473netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3474{
3475 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3476}
3477
3478static inline bool
3479netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3480{
3481 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3482}
3483
3484/**
3485 * netdev_queue_set_dql_min_limit - set dql minimum limit
3486 * @dev_queue: pointer to transmit queue
3487 * @min_limit: dql minimum limit
3488 *
3489 * Forces xmit_more() to return true until the minimum threshold
3490 * defined by @min_limit is reached (or until the tx queue is
3491 * empty). Warning: to be use with care, misuse will impact the
3492 * latency.
3493 */
3494static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3495 unsigned int min_limit)
3496{
3497#ifdef CONFIG_BQL
3498 dev_queue->dql.min_limit = min_limit;
3499#endif
3500}
3501
3502/**
3503 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3504 * @dev_queue: pointer to transmit queue
3505 *
3506 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3507 * to give appropriate hint to the CPU.
3508 */
3509static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3510{
3511#ifdef CONFIG_BQL
3512 prefetchw(&dev_queue->dql.num_queued);
3513#endif
3514}
3515
3516/**
3517 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3518 * @dev_queue: pointer to transmit queue
3519 *
3520 * BQL enabled drivers might use this helper in their TX completion path,
3521 * to give appropriate hint to the CPU.
3522 */
3523static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3524{
3525#ifdef CONFIG_BQL
3526 prefetchw(&dev_queue->dql.limit);
3527#endif
3528}
3529
3530/**
3531 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3532 * @dev_queue: network device queue
3533 * @bytes: number of bytes queued to the device queue
3534 *
3535 * Report the number of bytes queued for sending/completion to the network
3536 * device hardware queue. @bytes should be a good approximation and should
3537 * exactly match netdev_completed_queue() @bytes.
3538 * This is typically called once per packet, from ndo_start_xmit().
3539 */
3540static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3541 unsigned int bytes)
3542{
3543#ifdef CONFIG_BQL
3544 dql_queued(&dev_queue->dql, bytes);
3545
3546 if (likely(dql_avail(&dev_queue->dql) >= 0))
3547 return;
3548
3549 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3550
3551 /*
3552 * The XOFF flag must be set before checking the dql_avail below,
3553 * because in netdev_tx_completed_queue we update the dql_completed
3554 * before checking the XOFF flag.
3555 */
3556 smp_mb();
3557
3558 /* check again in case another CPU has just made room avail */
3559 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3560 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3561#endif
3562}
3563
3564/* Variant of netdev_tx_sent_queue() for drivers that are aware
3565 * that they should not test BQL status themselves.
3566 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3567 * skb of a batch.
3568 * Returns true if the doorbell must be used to kick the NIC.
3569 */
3570static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3571 unsigned int bytes,
3572 bool xmit_more)
3573{
3574 if (xmit_more) {
3575#ifdef CONFIG_BQL
3576 dql_queued(&dev_queue->dql, bytes);
3577#endif
3578 return netif_tx_queue_stopped(dev_queue);
3579 }
3580 netdev_tx_sent_queue(dev_queue, bytes);
3581 return true;
3582}
3583
3584/**
3585 * netdev_sent_queue - report the number of bytes queued to hardware
3586 * @dev: network device
3587 * @bytes: number of bytes queued to the hardware device queue
3588 *
3589 * Report the number of bytes queued for sending/completion to the network
3590 * device hardware queue#0. @bytes should be a good approximation and should
3591 * exactly match netdev_completed_queue() @bytes.
3592 * This is typically called once per packet, from ndo_start_xmit().
3593 */
3594static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3595{
3596 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3597}
3598
3599static inline bool __netdev_sent_queue(struct net_device *dev,
3600 unsigned int bytes,
3601 bool xmit_more)
3602{
3603 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3604 xmit_more);
3605}
3606
3607/**
3608 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3609 * @dev_queue: network device queue
3610 * @pkts: number of packets (currently ignored)
3611 * @bytes: number of bytes dequeued from the device queue
3612 *
3613 * Must be called at most once per TX completion round (and not per
3614 * individual packet), so that BQL can adjust its limits appropriately.
3615 */
3616static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3617 unsigned int pkts, unsigned int bytes)
3618{
3619#ifdef CONFIG_BQL
3620 if (unlikely(!bytes))
3621 return;
3622
3623 dql_completed(&dev_queue->dql, bytes);
3624
3625 /*
3626 * Without the memory barrier there is a small possiblity that
3627 * netdev_tx_sent_queue will miss the update and cause the queue to
3628 * be stopped forever
3629 */
3630 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3631
3632 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3633 return;
3634
3635 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3636 netif_schedule_queue(dev_queue);
3637#endif
3638}
3639
3640/**
3641 * netdev_completed_queue - report bytes and packets completed by device
3642 * @dev: network device
3643 * @pkts: actual number of packets sent over the medium
3644 * @bytes: actual number of bytes sent over the medium
3645 *
3646 * Report the number of bytes and packets transmitted by the network device
3647 * hardware queue over the physical medium, @bytes must exactly match the
3648 * @bytes amount passed to netdev_sent_queue()
3649 */
3650static inline void netdev_completed_queue(struct net_device *dev,
3651 unsigned int pkts, unsigned int bytes)
3652{
3653 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3654}
3655
3656static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3657{
3658#ifdef CONFIG_BQL
3659 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3660 dql_reset(&q->dql);
3661#endif
3662}
3663
3664/**
3665 * netdev_reset_queue - reset the packets and bytes count of a network device
3666 * @dev_queue: network device
3667 *
3668 * Reset the bytes and packet count of a network device and clear the
3669 * software flow control OFF bit for this network device
3670 */
3671static inline void netdev_reset_queue(struct net_device *dev_queue)
3672{
3673 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3674}
3675
3676/**
3677 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3678 * @dev: network device
3679 * @queue_index: given tx queue index
3680 *
3681 * Returns 0 if given tx queue index >= number of device tx queues,
3682 * otherwise returns the originally passed tx queue index.
3683 */
3684static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3685{
3686 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3687 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3688 dev->name, queue_index,
3689 dev->real_num_tx_queues);
3690 return 0;
3691 }
3692
3693 return queue_index;
3694}
3695
3696/**
3697 * netif_running - test if up
3698 * @dev: network device
3699 *
3700 * Test if the device has been brought up.
3701 */
3702static inline bool netif_running(const struct net_device *dev)
3703{
3704 return test_bit(__LINK_STATE_START, &dev->state);
3705}
3706
3707/*
3708 * Routines to manage the subqueues on a device. We only need start,
3709 * stop, and a check if it's stopped. All other device management is
3710 * done at the overall netdevice level.
3711 * Also test the device if we're multiqueue.
3712 */
3713
3714/**
3715 * netif_start_subqueue - allow sending packets on subqueue
3716 * @dev: network device
3717 * @queue_index: sub queue index
3718 *
3719 * Start individual transmit queue of a device with multiple transmit queues.
3720 */
3721static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3722{
3723 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3724
3725 netif_tx_start_queue(txq);
3726}
3727
3728/**
3729 * netif_stop_subqueue - stop sending packets on subqueue
3730 * @dev: network device
3731 * @queue_index: sub queue index
3732 *
3733 * Stop individual transmit queue of a device with multiple transmit queues.
3734 */
3735static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3736{
3737 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3738 netif_tx_stop_queue(txq);
3739}
3740
3741/**
3742 * __netif_subqueue_stopped - test status of subqueue
3743 * @dev: network device
3744 * @queue_index: sub queue index
3745 *
3746 * Check individual transmit queue of a device with multiple transmit queues.
3747 */
3748static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3749 u16 queue_index)
3750{
3751 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3752
3753 return netif_tx_queue_stopped(txq);
3754}
3755
3756/**
3757 * netif_subqueue_stopped - test status of subqueue
3758 * @dev: network device
3759 * @skb: sub queue buffer pointer
3760 *
3761 * Check individual transmit queue of a device with multiple transmit queues.
3762 */
3763static inline bool netif_subqueue_stopped(const struct net_device *dev,
3764 struct sk_buff *skb)
3765{
3766 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3767}
3768
3769/**
3770 * netif_wake_subqueue - allow sending packets on subqueue
3771 * @dev: network device
3772 * @queue_index: sub queue index
3773 *
3774 * Resume individual transmit queue of a device with multiple transmit queues.
3775 */
3776static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3777{
3778 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3779
3780 netif_tx_wake_queue(txq);
3781}
3782
3783#ifdef CONFIG_XPS
3784int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3785 u16 index);
3786int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3787 u16 index, enum xps_map_type type);
3788
3789/**
3790 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3791 * @j: CPU/Rx queue index
3792 * @mask: bitmask of all cpus/rx queues
3793 * @nr_bits: number of bits in the bitmask
3794 *
3795 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3796 */
3797static inline bool netif_attr_test_mask(unsigned long j,
3798 const unsigned long *mask,
3799 unsigned int nr_bits)
3800{
3801 cpu_max_bits_warn(j, nr_bits);
3802 return test_bit(j, mask);
3803}
3804
3805/**
3806 * netif_attr_test_online - Test for online CPU/Rx queue
3807 * @j: CPU/Rx queue index
3808 * @online_mask: bitmask for CPUs/Rx queues that are online
3809 * @nr_bits: number of bits in the bitmask
3810 *
3811 * Returns true if a CPU/Rx queue is online.
3812 */
3813static inline bool netif_attr_test_online(unsigned long j,
3814 const unsigned long *online_mask,
3815 unsigned int nr_bits)
3816{
3817 cpu_max_bits_warn(j, nr_bits);
3818
3819 if (online_mask)
3820 return test_bit(j, online_mask);
3821
3822 return (j < nr_bits);
3823}
3824
3825/**
3826 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3827 * @n: CPU/Rx queue index
3828 * @srcp: the cpumask/Rx queue mask pointer
3829 * @nr_bits: number of bits in the bitmask
3830 *
3831 * Returns >= nr_bits if no further CPUs/Rx queues set.
3832 */
3833static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3834 unsigned int nr_bits)
3835{
3836 /* -1 is a legal arg here. */
3837 if (n != -1)
3838 cpu_max_bits_warn(n, nr_bits);
3839
3840 if (srcp)
3841 return find_next_bit(srcp, nr_bits, n + 1);
3842
3843 return n + 1;
3844}
3845
3846/**
3847 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3848 * @n: CPU/Rx queue index
3849 * @src1p: the first CPUs/Rx queues mask pointer
3850 * @src2p: the second CPUs/Rx queues mask pointer
3851 * @nr_bits: number of bits in the bitmask
3852 *
3853 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3854 */
3855static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3856 const unsigned long *src2p,
3857 unsigned int nr_bits)
3858{
3859 /* -1 is a legal arg here. */
3860 if (n != -1)
3861 cpu_max_bits_warn(n, nr_bits);
3862
3863 if (src1p && src2p)
3864 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3865 else if (src1p)
3866 return find_next_bit(src1p, nr_bits, n + 1);
3867 else if (src2p)
3868 return find_next_bit(src2p, nr_bits, n + 1);
3869
3870 return n + 1;
3871}
3872#else
3873static inline int netif_set_xps_queue(struct net_device *dev,
3874 const struct cpumask *mask,
3875 u16 index)
3876{
3877 return 0;
3878}
3879
3880static inline int __netif_set_xps_queue(struct net_device *dev,
3881 const unsigned long *mask,
3882 u16 index, enum xps_map_type type)
3883{
3884 return 0;
3885}
3886#endif
3887
3888/**
3889 * netif_is_multiqueue - test if device has multiple transmit queues
3890 * @dev: network device
3891 *
3892 * Check if device has multiple transmit queues
3893 */
3894static inline bool netif_is_multiqueue(const struct net_device *dev)
3895{
3896 return dev->num_tx_queues > 1;
3897}
3898
3899int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3900
3901#ifdef CONFIG_SYSFS
3902int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3903#else
3904static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3905 unsigned int rxqs)
3906{
3907 dev->real_num_rx_queues = rxqs;
3908 return 0;
3909}
3910#endif
3911int netif_set_real_num_queues(struct net_device *dev,
3912 unsigned int txq, unsigned int rxq);
3913
3914int netif_get_num_default_rss_queues(void);
3915
3916void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3917void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3918
3919/*
3920 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3921 * interrupt context or with hardware interrupts being disabled.
3922 * (in_hardirq() || irqs_disabled())
3923 *
3924 * We provide four helpers that can be used in following contexts :
3925 *
3926 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3927 * replacing kfree_skb(skb)
3928 *
3929 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3930 * Typically used in place of consume_skb(skb) in TX completion path
3931 *
3932 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3933 * replacing kfree_skb(skb)
3934 *
3935 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3936 * and consumed a packet. Used in place of consume_skb(skb)
3937 */
3938static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3939{
3940 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3941}
3942
3943static inline void dev_consume_skb_irq(struct sk_buff *skb)
3944{
3945 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3946}
3947
3948static inline void dev_kfree_skb_any(struct sk_buff *skb)
3949{
3950 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3951}
3952
3953static inline void dev_consume_skb_any(struct sk_buff *skb)
3954{
3955 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3956}
3957
3958u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3959 struct bpf_prog *xdp_prog);
3960void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3961int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3962int netif_rx(struct sk_buff *skb);
3963int __netif_rx(struct sk_buff *skb);
3964
3965int netif_receive_skb(struct sk_buff *skb);
3966int netif_receive_skb_core(struct sk_buff *skb);
3967void netif_receive_skb_list_internal(struct list_head *head);
3968void netif_receive_skb_list(struct list_head *head);
3969gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3970void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3971struct sk_buff *napi_get_frags(struct napi_struct *napi);
3972void napi_get_frags_check(struct napi_struct *napi);
3973gro_result_t napi_gro_frags(struct napi_struct *napi);
3974struct packet_offload *gro_find_receive_by_type(__be16 type);
3975struct packet_offload *gro_find_complete_by_type(__be16 type);
3976
3977static inline void napi_free_frags(struct napi_struct *napi)
3978{
3979 kfree_skb(napi->skb);
3980 napi->skb = NULL;
3981}
3982
3983bool netdev_is_rx_handler_busy(struct net_device *dev);
3984int netdev_rx_handler_register(struct net_device *dev,
3985 rx_handler_func_t *rx_handler,
3986 void *rx_handler_data);
3987void netdev_rx_handler_unregister(struct net_device *dev);
3988
3989bool dev_valid_name(const char *name);
3990static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3991{
3992 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3993}
3994int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3995int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3996int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3997 void __user *data, bool *need_copyout);
3998int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3999int generic_hwtstamp_get_lower(struct net_device *dev,
4000 struct kernel_hwtstamp_config *kernel_cfg);
4001int generic_hwtstamp_set_lower(struct net_device *dev,
4002 struct kernel_hwtstamp_config *kernel_cfg,
4003 struct netlink_ext_ack *extack);
4004int dev_set_hwtstamp_phylib(struct net_device *dev,
4005 struct kernel_hwtstamp_config *cfg,
4006 struct netlink_ext_ack *extack);
4007int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4008unsigned int dev_get_flags(const struct net_device *);
4009int __dev_change_flags(struct net_device *dev, unsigned int flags,
4010 struct netlink_ext_ack *extack);
4011int dev_change_flags(struct net_device *dev, unsigned int flags,
4012 struct netlink_ext_ack *extack);
4013int dev_set_alias(struct net_device *, const char *, size_t);
4014int dev_get_alias(const struct net_device *, char *, size_t);
4015int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4016 const char *pat, int new_ifindex);
4017static inline
4018int dev_change_net_namespace(struct net_device *dev, struct net *net,
4019 const char *pat)
4020{
4021 return __dev_change_net_namespace(dev, net, pat, 0);
4022}
4023int __dev_set_mtu(struct net_device *, int);
4024int dev_set_mtu(struct net_device *, int);
4025int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4026 struct netlink_ext_ack *extack);
4027int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4028 struct netlink_ext_ack *extack);
4029int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4030 struct netlink_ext_ack *extack);
4031int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4032int dev_get_port_parent_id(struct net_device *dev,
4033 struct netdev_phys_item_id *ppid, bool recurse);
4034bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4035void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin);
4036void netdev_dpll_pin_clear(struct net_device *dev);
4037
4038static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev)
4039{
4040#if IS_ENABLED(CONFIG_DPLL)
4041 return dev->dpll_pin;
4042#else
4043 return NULL;
4044#endif
4045}
4046
4047struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4048struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4049 struct netdev_queue *txq, int *ret);
4050
4051int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4052u8 dev_xdp_prog_count(struct net_device *dev);
4053u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4054
4055int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4056int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4057int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4058bool is_skb_forwardable(const struct net_device *dev,
4059 const struct sk_buff *skb);
4060
4061static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4062 const struct sk_buff *skb,
4063 const bool check_mtu)
4064{
4065 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4066 unsigned int len;
4067
4068 if (!(dev->flags & IFF_UP))
4069 return false;
4070
4071 if (!check_mtu)
4072 return true;
4073
4074 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4075 if (skb->len <= len)
4076 return true;
4077
4078 /* if TSO is enabled, we don't care about the length as the packet
4079 * could be forwarded without being segmented before
4080 */
4081 if (skb_is_gso(skb))
4082 return true;
4083
4084 return false;
4085}
4086
4087void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4088
4089#define DEV_CORE_STATS_INC(FIELD) \
4090static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4091{ \
4092 netdev_core_stats_inc(dev, \
4093 offsetof(struct net_device_core_stats, FIELD)); \
4094}
4095DEV_CORE_STATS_INC(rx_dropped)
4096DEV_CORE_STATS_INC(tx_dropped)
4097DEV_CORE_STATS_INC(rx_nohandler)
4098DEV_CORE_STATS_INC(rx_otherhost_dropped)
4099#undef DEV_CORE_STATS_INC
4100
4101static __always_inline int ____dev_forward_skb(struct net_device *dev,
4102 struct sk_buff *skb,
4103 const bool check_mtu)
4104{
4105 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4106 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4107 dev_core_stats_rx_dropped_inc(dev);
4108 kfree_skb(skb);
4109 return NET_RX_DROP;
4110 }
4111
4112 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4113 skb->priority = 0;
4114 return 0;
4115}
4116
4117bool dev_nit_active(struct net_device *dev);
4118void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4119
4120static inline void __dev_put(struct net_device *dev)
4121{
4122 if (dev) {
4123#ifdef CONFIG_PCPU_DEV_REFCNT
4124 this_cpu_dec(*dev->pcpu_refcnt);
4125#else
4126 refcount_dec(&dev->dev_refcnt);
4127#endif
4128 }
4129}
4130
4131static inline void __dev_hold(struct net_device *dev)
4132{
4133 if (dev) {
4134#ifdef CONFIG_PCPU_DEV_REFCNT
4135 this_cpu_inc(*dev->pcpu_refcnt);
4136#else
4137 refcount_inc(&dev->dev_refcnt);
4138#endif
4139 }
4140}
4141
4142static inline void __netdev_tracker_alloc(struct net_device *dev,
4143 netdevice_tracker *tracker,
4144 gfp_t gfp)
4145{
4146#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4147 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4148#endif
4149}
4150
4151/* netdev_tracker_alloc() can upgrade a prior untracked reference
4152 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4153 */
4154static inline void netdev_tracker_alloc(struct net_device *dev,
4155 netdevice_tracker *tracker, gfp_t gfp)
4156{
4157#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4158 refcount_dec(&dev->refcnt_tracker.no_tracker);
4159 __netdev_tracker_alloc(dev, tracker, gfp);
4160#endif
4161}
4162
4163static inline void netdev_tracker_free(struct net_device *dev,
4164 netdevice_tracker *tracker)
4165{
4166#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4167 ref_tracker_free(&dev->refcnt_tracker, tracker);
4168#endif
4169}
4170
4171static inline void netdev_hold(struct net_device *dev,
4172 netdevice_tracker *tracker, gfp_t gfp)
4173{
4174 if (dev) {
4175 __dev_hold(dev);
4176 __netdev_tracker_alloc(dev, tracker, gfp);
4177 }
4178}
4179
4180static inline void netdev_put(struct net_device *dev,
4181 netdevice_tracker *tracker)
4182{
4183 if (dev) {
4184 netdev_tracker_free(dev, tracker);
4185 __dev_put(dev);
4186 }
4187}
4188
4189/**
4190 * dev_hold - get reference to device
4191 * @dev: network device
4192 *
4193 * Hold reference to device to keep it from being freed.
4194 * Try using netdev_hold() instead.
4195 */
4196static inline void dev_hold(struct net_device *dev)
4197{
4198 netdev_hold(dev, NULL, GFP_ATOMIC);
4199}
4200
4201/**
4202 * dev_put - release reference to device
4203 * @dev: network device
4204 *
4205 * Release reference to device to allow it to be freed.
4206 * Try using netdev_put() instead.
4207 */
4208static inline void dev_put(struct net_device *dev)
4209{
4210 netdev_put(dev, NULL);
4211}
4212
4213static inline void netdev_ref_replace(struct net_device *odev,
4214 struct net_device *ndev,
4215 netdevice_tracker *tracker,
4216 gfp_t gfp)
4217{
4218 if (odev)
4219 netdev_tracker_free(odev, tracker);
4220
4221 __dev_hold(ndev);
4222 __dev_put(odev);
4223
4224 if (ndev)
4225 __netdev_tracker_alloc(ndev, tracker, gfp);
4226}
4227
4228/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4229 * and _off may be called from IRQ context, but it is caller
4230 * who is responsible for serialization of these calls.
4231 *
4232 * The name carrier is inappropriate, these functions should really be
4233 * called netif_lowerlayer_*() because they represent the state of any
4234 * kind of lower layer not just hardware media.
4235 */
4236void linkwatch_fire_event(struct net_device *dev);
4237
4238/**
4239 * linkwatch_sync_dev - sync linkwatch for the given device
4240 * @dev: network device to sync linkwatch for
4241 *
4242 * Sync linkwatch for the given device, removing it from the
4243 * pending work list (if queued).
4244 */
4245void linkwatch_sync_dev(struct net_device *dev);
4246
4247/**
4248 * netif_carrier_ok - test if carrier present
4249 * @dev: network device
4250 *
4251 * Check if carrier is present on device
4252 */
4253static inline bool netif_carrier_ok(const struct net_device *dev)
4254{
4255 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4256}
4257
4258unsigned long dev_trans_start(struct net_device *dev);
4259
4260void __netdev_watchdog_up(struct net_device *dev);
4261
4262void netif_carrier_on(struct net_device *dev);
4263void netif_carrier_off(struct net_device *dev);
4264void netif_carrier_event(struct net_device *dev);
4265
4266/**
4267 * netif_dormant_on - mark device as dormant.
4268 * @dev: network device
4269 *
4270 * Mark device as dormant (as per RFC2863).
4271 *
4272 * The dormant state indicates that the relevant interface is not
4273 * actually in a condition to pass packets (i.e., it is not 'up') but is
4274 * in a "pending" state, waiting for some external event. For "on-
4275 * demand" interfaces, this new state identifies the situation where the
4276 * interface is waiting for events to place it in the up state.
4277 */
4278static inline void netif_dormant_on(struct net_device *dev)
4279{
4280 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4281 linkwatch_fire_event(dev);
4282}
4283
4284/**
4285 * netif_dormant_off - set device as not dormant.
4286 * @dev: network device
4287 *
4288 * Device is not in dormant state.
4289 */
4290static inline void netif_dormant_off(struct net_device *dev)
4291{
4292 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4293 linkwatch_fire_event(dev);
4294}
4295
4296/**
4297 * netif_dormant - test if device is dormant
4298 * @dev: network device
4299 *
4300 * Check if device is dormant.
4301 */
4302static inline bool netif_dormant(const struct net_device *dev)
4303{
4304 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4305}
4306
4307
4308/**
4309 * netif_testing_on - mark device as under test.
4310 * @dev: network device
4311 *
4312 * Mark device as under test (as per RFC2863).
4313 *
4314 * The testing state indicates that some test(s) must be performed on
4315 * the interface. After completion, of the test, the interface state
4316 * will change to up, dormant, or down, as appropriate.
4317 */
4318static inline void netif_testing_on(struct net_device *dev)
4319{
4320 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4321 linkwatch_fire_event(dev);
4322}
4323
4324/**
4325 * netif_testing_off - set device as not under test.
4326 * @dev: network device
4327 *
4328 * Device is not in testing state.
4329 */
4330static inline void netif_testing_off(struct net_device *dev)
4331{
4332 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4333 linkwatch_fire_event(dev);
4334}
4335
4336/**
4337 * netif_testing - test if device is under test
4338 * @dev: network device
4339 *
4340 * Check if device is under test
4341 */
4342static inline bool netif_testing(const struct net_device *dev)
4343{
4344 return test_bit(__LINK_STATE_TESTING, &dev->state);
4345}
4346
4347
4348/**
4349 * netif_oper_up - test if device is operational
4350 * @dev: network device
4351 *
4352 * Check if carrier is operational
4353 */
4354static inline bool netif_oper_up(const struct net_device *dev)
4355{
4356 return (dev->operstate == IF_OPER_UP ||
4357 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4358}
4359
4360/**
4361 * netif_device_present - is device available or removed
4362 * @dev: network device
4363 *
4364 * Check if device has not been removed from system.
4365 */
4366static inline bool netif_device_present(const struct net_device *dev)
4367{
4368 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4369}
4370
4371void netif_device_detach(struct net_device *dev);
4372
4373void netif_device_attach(struct net_device *dev);
4374
4375/*
4376 * Network interface message level settings
4377 */
4378
4379enum {
4380 NETIF_MSG_DRV_BIT,
4381 NETIF_MSG_PROBE_BIT,
4382 NETIF_MSG_LINK_BIT,
4383 NETIF_MSG_TIMER_BIT,
4384 NETIF_MSG_IFDOWN_BIT,
4385 NETIF_MSG_IFUP_BIT,
4386 NETIF_MSG_RX_ERR_BIT,
4387 NETIF_MSG_TX_ERR_BIT,
4388 NETIF_MSG_TX_QUEUED_BIT,
4389 NETIF_MSG_INTR_BIT,
4390 NETIF_MSG_TX_DONE_BIT,
4391 NETIF_MSG_RX_STATUS_BIT,
4392 NETIF_MSG_PKTDATA_BIT,
4393 NETIF_MSG_HW_BIT,
4394 NETIF_MSG_WOL_BIT,
4395
4396 /* When you add a new bit above, update netif_msg_class_names array
4397 * in net/ethtool/common.c
4398 */
4399 NETIF_MSG_CLASS_COUNT,
4400};
4401/* Both ethtool_ops interface and internal driver implementation use u32 */
4402static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4403
4404#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4405#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4406
4407#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4408#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4409#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4410#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4411#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4412#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4413#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4414#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4415#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4416#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4417#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4418#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4419#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4420#define NETIF_MSG_HW __NETIF_MSG(HW)
4421#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4422
4423#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4424#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4425#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4426#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4427#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4428#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4429#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4430#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4431#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4432#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4433#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4434#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4435#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4436#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4437#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4438
4439static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4440{
4441 /* use default */
4442 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4443 return default_msg_enable_bits;
4444 if (debug_value == 0) /* no output */
4445 return 0;
4446 /* set low N bits */
4447 return (1U << debug_value) - 1;
4448}
4449
4450static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4451{
4452 spin_lock(&txq->_xmit_lock);
4453 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4454 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4455}
4456
4457static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4458{
4459 __acquire(&txq->_xmit_lock);
4460 return true;
4461}
4462
4463static inline void __netif_tx_release(struct netdev_queue *txq)
4464{
4465 __release(&txq->_xmit_lock);
4466}
4467
4468static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4469{
4470 spin_lock_bh(&txq->_xmit_lock);
4471 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4472 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4473}
4474
4475static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4476{
4477 bool ok = spin_trylock(&txq->_xmit_lock);
4478
4479 if (likely(ok)) {
4480 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4481 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4482 }
4483 return ok;
4484}
4485
4486static inline void __netif_tx_unlock(struct netdev_queue *txq)
4487{
4488 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4489 WRITE_ONCE(txq->xmit_lock_owner, -1);
4490 spin_unlock(&txq->_xmit_lock);
4491}
4492
4493static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4494{
4495 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4496 WRITE_ONCE(txq->xmit_lock_owner, -1);
4497 spin_unlock_bh(&txq->_xmit_lock);
4498}
4499
4500/*
4501 * txq->trans_start can be read locklessly from dev_watchdog()
4502 */
4503static inline void txq_trans_update(struct netdev_queue *txq)
4504{
4505 if (txq->xmit_lock_owner != -1)
4506 WRITE_ONCE(txq->trans_start, jiffies);
4507}
4508
4509static inline void txq_trans_cond_update(struct netdev_queue *txq)
4510{
4511 unsigned long now = jiffies;
4512
4513 if (READ_ONCE(txq->trans_start) != now)
4514 WRITE_ONCE(txq->trans_start, now);
4515}
4516
4517/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4518static inline void netif_trans_update(struct net_device *dev)
4519{
4520 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4521
4522 txq_trans_cond_update(txq);
4523}
4524
4525/**
4526 * netif_tx_lock - grab network device transmit lock
4527 * @dev: network device
4528 *
4529 * Get network device transmit lock
4530 */
4531void netif_tx_lock(struct net_device *dev);
4532
4533static inline void netif_tx_lock_bh(struct net_device *dev)
4534{
4535 local_bh_disable();
4536 netif_tx_lock(dev);
4537}
4538
4539void netif_tx_unlock(struct net_device *dev);
4540
4541static inline void netif_tx_unlock_bh(struct net_device *dev)
4542{
4543 netif_tx_unlock(dev);
4544 local_bh_enable();
4545}
4546
4547#define HARD_TX_LOCK(dev, txq, cpu) { \
4548 if ((dev->features & NETIF_F_LLTX) == 0) { \
4549 __netif_tx_lock(txq, cpu); \
4550 } else { \
4551 __netif_tx_acquire(txq); \
4552 } \
4553}
4554
4555#define HARD_TX_TRYLOCK(dev, txq) \
4556 (((dev->features & NETIF_F_LLTX) == 0) ? \
4557 __netif_tx_trylock(txq) : \
4558 __netif_tx_acquire(txq))
4559
4560#define HARD_TX_UNLOCK(dev, txq) { \
4561 if ((dev->features & NETIF_F_LLTX) == 0) { \
4562 __netif_tx_unlock(txq); \
4563 } else { \
4564 __netif_tx_release(txq); \
4565 } \
4566}
4567
4568static inline void netif_tx_disable(struct net_device *dev)
4569{
4570 unsigned int i;
4571 int cpu;
4572
4573 local_bh_disable();
4574 cpu = smp_processor_id();
4575 spin_lock(&dev->tx_global_lock);
4576 for (i = 0; i < dev->num_tx_queues; i++) {
4577 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4578
4579 __netif_tx_lock(txq, cpu);
4580 netif_tx_stop_queue(txq);
4581 __netif_tx_unlock(txq);
4582 }
4583 spin_unlock(&dev->tx_global_lock);
4584 local_bh_enable();
4585}
4586
4587static inline void netif_addr_lock(struct net_device *dev)
4588{
4589 unsigned char nest_level = 0;
4590
4591#ifdef CONFIG_LOCKDEP
4592 nest_level = dev->nested_level;
4593#endif
4594 spin_lock_nested(&dev->addr_list_lock, nest_level);
4595}
4596
4597static inline void netif_addr_lock_bh(struct net_device *dev)
4598{
4599 unsigned char nest_level = 0;
4600
4601#ifdef CONFIG_LOCKDEP
4602 nest_level = dev->nested_level;
4603#endif
4604 local_bh_disable();
4605 spin_lock_nested(&dev->addr_list_lock, nest_level);
4606}
4607
4608static inline void netif_addr_unlock(struct net_device *dev)
4609{
4610 spin_unlock(&dev->addr_list_lock);
4611}
4612
4613static inline void netif_addr_unlock_bh(struct net_device *dev)
4614{
4615 spin_unlock_bh(&dev->addr_list_lock);
4616}
4617
4618/*
4619 * dev_addrs walker. Should be used only for read access. Call with
4620 * rcu_read_lock held.
4621 */
4622#define for_each_dev_addr(dev, ha) \
4623 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4624
4625/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4626
4627void ether_setup(struct net_device *dev);
4628
4629/* Support for loadable net-drivers */
4630struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4631 unsigned char name_assign_type,
4632 void (*setup)(struct net_device *),
4633 unsigned int txqs, unsigned int rxqs);
4634#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4635 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4636
4637#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4638 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4639 count)
4640
4641int register_netdev(struct net_device *dev);
4642void unregister_netdev(struct net_device *dev);
4643
4644int devm_register_netdev(struct device *dev, struct net_device *ndev);
4645
4646/* General hardware address lists handling functions */
4647int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4648 struct netdev_hw_addr_list *from_list, int addr_len);
4649void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4650 struct netdev_hw_addr_list *from_list, int addr_len);
4651int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4652 struct net_device *dev,
4653 int (*sync)(struct net_device *, const unsigned char *),
4654 int (*unsync)(struct net_device *,
4655 const unsigned char *));
4656int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4657 struct net_device *dev,
4658 int (*sync)(struct net_device *,
4659 const unsigned char *, int),
4660 int (*unsync)(struct net_device *,
4661 const unsigned char *, int));
4662void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4663 struct net_device *dev,
4664 int (*unsync)(struct net_device *,
4665 const unsigned char *, int));
4666void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4667 struct net_device *dev,
4668 int (*unsync)(struct net_device *,
4669 const unsigned char *));
4670void __hw_addr_init(struct netdev_hw_addr_list *list);
4671
4672/* Functions used for device addresses handling */
4673void dev_addr_mod(struct net_device *dev, unsigned int offset,
4674 const void *addr, size_t len);
4675
4676static inline void
4677__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4678{
4679 dev_addr_mod(dev, 0, addr, len);
4680}
4681
4682static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4683{
4684 __dev_addr_set(dev, addr, dev->addr_len);
4685}
4686
4687int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4688 unsigned char addr_type);
4689int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4690 unsigned char addr_type);
4691
4692/* Functions used for unicast addresses handling */
4693int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4694int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4695int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4696int dev_uc_sync(struct net_device *to, struct net_device *from);
4697int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4698void dev_uc_unsync(struct net_device *to, struct net_device *from);
4699void dev_uc_flush(struct net_device *dev);
4700void dev_uc_init(struct net_device *dev);
4701
4702/**
4703 * __dev_uc_sync - Synchonize device's unicast list
4704 * @dev: device to sync
4705 * @sync: function to call if address should be added
4706 * @unsync: function to call if address should be removed
4707 *
4708 * Add newly added addresses to the interface, and release
4709 * addresses that have been deleted.
4710 */
4711static inline int __dev_uc_sync(struct net_device *dev,
4712 int (*sync)(struct net_device *,
4713 const unsigned char *),
4714 int (*unsync)(struct net_device *,
4715 const unsigned char *))
4716{
4717 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4718}
4719
4720/**
4721 * __dev_uc_unsync - Remove synchronized addresses from device
4722 * @dev: device to sync
4723 * @unsync: function to call if address should be removed
4724 *
4725 * Remove all addresses that were added to the device by dev_uc_sync().
4726 */
4727static inline void __dev_uc_unsync(struct net_device *dev,
4728 int (*unsync)(struct net_device *,
4729 const unsigned char *))
4730{
4731 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4732}
4733
4734/* Functions used for multicast addresses handling */
4735int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4736int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4737int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4738int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4739int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4740int dev_mc_sync(struct net_device *to, struct net_device *from);
4741int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4742void dev_mc_unsync(struct net_device *to, struct net_device *from);
4743void dev_mc_flush(struct net_device *dev);
4744void dev_mc_init(struct net_device *dev);
4745
4746/**
4747 * __dev_mc_sync - Synchonize device's multicast list
4748 * @dev: device to sync
4749 * @sync: function to call if address should be added
4750 * @unsync: function to call if address should be removed
4751 *
4752 * Add newly added addresses to the interface, and release
4753 * addresses that have been deleted.
4754 */
4755static inline int __dev_mc_sync(struct net_device *dev,
4756 int (*sync)(struct net_device *,
4757 const unsigned char *),
4758 int (*unsync)(struct net_device *,
4759 const unsigned char *))
4760{
4761 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4762}
4763
4764/**
4765 * __dev_mc_unsync - Remove synchronized addresses from device
4766 * @dev: device to sync
4767 * @unsync: function to call if address should be removed
4768 *
4769 * Remove all addresses that were added to the device by dev_mc_sync().
4770 */
4771static inline void __dev_mc_unsync(struct net_device *dev,
4772 int (*unsync)(struct net_device *,
4773 const unsigned char *))
4774{
4775 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4776}
4777
4778/* Functions used for secondary unicast and multicast support */
4779void dev_set_rx_mode(struct net_device *dev);
4780int dev_set_promiscuity(struct net_device *dev, int inc);
4781int dev_set_allmulti(struct net_device *dev, int inc);
4782void netdev_state_change(struct net_device *dev);
4783void __netdev_notify_peers(struct net_device *dev);
4784void netdev_notify_peers(struct net_device *dev);
4785void netdev_features_change(struct net_device *dev);
4786/* Load a device via the kmod */
4787void dev_load(struct net *net, const char *name);
4788struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4789 struct rtnl_link_stats64 *storage);
4790void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4791 const struct net_device_stats *netdev_stats);
4792void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4793 const struct pcpu_sw_netstats __percpu *netstats);
4794void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4795
4796extern int netdev_max_backlog;
4797extern int dev_rx_weight;
4798extern int dev_tx_weight;
4799extern int gro_normal_batch;
4800
4801enum {
4802 NESTED_SYNC_IMM_BIT,
4803 NESTED_SYNC_TODO_BIT,
4804};
4805
4806#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4807#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4808
4809#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4810#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4811
4812struct netdev_nested_priv {
4813 unsigned char flags;
4814 void *data;
4815};
4816
4817bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4818struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4819 struct list_head **iter);
4820
4821/* iterate through upper list, must be called under RCU read lock */
4822#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4823 for (iter = &(dev)->adj_list.upper, \
4824 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4825 updev; \
4826 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4827
4828int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4829 int (*fn)(struct net_device *upper_dev,
4830 struct netdev_nested_priv *priv),
4831 struct netdev_nested_priv *priv);
4832
4833bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4834 struct net_device *upper_dev);
4835
4836bool netdev_has_any_upper_dev(struct net_device *dev);
4837
4838void *netdev_lower_get_next_private(struct net_device *dev,
4839 struct list_head **iter);
4840void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4841 struct list_head **iter);
4842
4843#define netdev_for_each_lower_private(dev, priv, iter) \
4844 for (iter = (dev)->adj_list.lower.next, \
4845 priv = netdev_lower_get_next_private(dev, &(iter)); \
4846 priv; \
4847 priv = netdev_lower_get_next_private(dev, &(iter)))
4848
4849#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4850 for (iter = &(dev)->adj_list.lower, \
4851 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4852 priv; \
4853 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4854
4855void *netdev_lower_get_next(struct net_device *dev,
4856 struct list_head **iter);
4857
4858#define netdev_for_each_lower_dev(dev, ldev, iter) \
4859 for (iter = (dev)->adj_list.lower.next, \
4860 ldev = netdev_lower_get_next(dev, &(iter)); \
4861 ldev; \
4862 ldev = netdev_lower_get_next(dev, &(iter)))
4863
4864struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4865 struct list_head **iter);
4866int netdev_walk_all_lower_dev(struct net_device *dev,
4867 int (*fn)(struct net_device *lower_dev,
4868 struct netdev_nested_priv *priv),
4869 struct netdev_nested_priv *priv);
4870int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4871 int (*fn)(struct net_device *lower_dev,
4872 struct netdev_nested_priv *priv),
4873 struct netdev_nested_priv *priv);
4874
4875void *netdev_adjacent_get_private(struct list_head *adj_list);
4876void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4877struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4878struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4879int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4880 struct netlink_ext_ack *extack);
4881int netdev_master_upper_dev_link(struct net_device *dev,
4882 struct net_device *upper_dev,
4883 void *upper_priv, void *upper_info,
4884 struct netlink_ext_ack *extack);
4885void netdev_upper_dev_unlink(struct net_device *dev,
4886 struct net_device *upper_dev);
4887int netdev_adjacent_change_prepare(struct net_device *old_dev,
4888 struct net_device *new_dev,
4889 struct net_device *dev,
4890 struct netlink_ext_ack *extack);
4891void netdev_adjacent_change_commit(struct net_device *old_dev,
4892 struct net_device *new_dev,
4893 struct net_device *dev);
4894void netdev_adjacent_change_abort(struct net_device *old_dev,
4895 struct net_device *new_dev,
4896 struct net_device *dev);
4897void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4898void *netdev_lower_dev_get_private(struct net_device *dev,
4899 struct net_device *lower_dev);
4900void netdev_lower_state_changed(struct net_device *lower_dev,
4901 void *lower_state_info);
4902
4903/* RSS keys are 40 or 52 bytes long */
4904#define NETDEV_RSS_KEY_LEN 52
4905extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4906void netdev_rss_key_fill(void *buffer, size_t len);
4907
4908int skb_checksum_help(struct sk_buff *skb);
4909int skb_crc32c_csum_help(struct sk_buff *skb);
4910int skb_csum_hwoffload_help(struct sk_buff *skb,
4911 const netdev_features_t features);
4912
4913struct netdev_bonding_info {
4914 ifslave slave;
4915 ifbond master;
4916};
4917
4918struct netdev_notifier_bonding_info {
4919 struct netdev_notifier_info info; /* must be first */
4920 struct netdev_bonding_info bonding_info;
4921};
4922
4923void netdev_bonding_info_change(struct net_device *dev,
4924 struct netdev_bonding_info *bonding_info);
4925
4926#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4927void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4928#else
4929static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4930 const void *data)
4931{
4932}
4933#endif
4934
4935__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4936
4937static inline bool can_checksum_protocol(netdev_features_t features,
4938 __be16 protocol)
4939{
4940 if (protocol == htons(ETH_P_FCOE))
4941 return !!(features & NETIF_F_FCOE_CRC);
4942
4943 /* Assume this is an IP checksum (not SCTP CRC) */
4944
4945 if (features & NETIF_F_HW_CSUM) {
4946 /* Can checksum everything */
4947 return true;
4948 }
4949
4950 switch (protocol) {
4951 case htons(ETH_P_IP):
4952 return !!(features & NETIF_F_IP_CSUM);
4953 case htons(ETH_P_IPV6):
4954 return !!(features & NETIF_F_IPV6_CSUM);
4955 default:
4956 return false;
4957 }
4958}
4959
4960#ifdef CONFIG_BUG
4961void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4962#else
4963static inline void netdev_rx_csum_fault(struct net_device *dev,
4964 struct sk_buff *skb)
4965{
4966}
4967#endif
4968/* rx skb timestamps */
4969void net_enable_timestamp(void);
4970void net_disable_timestamp(void);
4971
4972static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4973 const struct skb_shared_hwtstamps *hwtstamps,
4974 bool cycles)
4975{
4976 const struct net_device_ops *ops = dev->netdev_ops;
4977
4978 if (ops->ndo_get_tstamp)
4979 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4980
4981 return hwtstamps->hwtstamp;
4982}
4983
4984static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4985 struct sk_buff *skb, struct net_device *dev,
4986 bool more)
4987{
4988 __this_cpu_write(softnet_data.xmit.more, more);
4989 return ops->ndo_start_xmit(skb, dev);
4990}
4991
4992static inline bool netdev_xmit_more(void)
4993{
4994 return __this_cpu_read(softnet_data.xmit.more);
4995}
4996
4997static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4998 struct netdev_queue *txq, bool more)
4999{
5000 const struct net_device_ops *ops = dev->netdev_ops;
5001 netdev_tx_t rc;
5002
5003 rc = __netdev_start_xmit(ops, skb, dev, more);
5004 if (rc == NETDEV_TX_OK)
5005 txq_trans_update(txq);
5006
5007 return rc;
5008}
5009
5010int netdev_class_create_file_ns(const struct class_attribute *class_attr,
5011 const void *ns);
5012void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
5013 const void *ns);
5014
5015extern const struct kobj_ns_type_operations net_ns_type_operations;
5016
5017const char *netdev_drivername(const struct net_device *dev);
5018
5019static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
5020 netdev_features_t f2)
5021{
5022 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
5023 if (f1 & NETIF_F_HW_CSUM)
5024 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5025 else
5026 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5027 }
5028
5029 return f1 & f2;
5030}
5031
5032static inline netdev_features_t netdev_get_wanted_features(
5033 struct net_device *dev)
5034{
5035 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5036}
5037netdev_features_t netdev_increment_features(netdev_features_t all,
5038 netdev_features_t one, netdev_features_t mask);
5039
5040/* Allow TSO being used on stacked device :
5041 * Performing the GSO segmentation before last device
5042 * is a performance improvement.
5043 */
5044static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5045 netdev_features_t mask)
5046{
5047 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5048}
5049
5050int __netdev_update_features(struct net_device *dev);
5051void netdev_update_features(struct net_device *dev);
5052void netdev_change_features(struct net_device *dev);
5053
5054void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5055 struct net_device *dev);
5056
5057netdev_features_t passthru_features_check(struct sk_buff *skb,
5058 struct net_device *dev,
5059 netdev_features_t features);
5060netdev_features_t netif_skb_features(struct sk_buff *skb);
5061void skb_warn_bad_offload(const struct sk_buff *skb);
5062
5063static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5064{
5065 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5066
5067 /* check flags correspondence */
5068 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5069 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5070 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5071 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5072 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5073 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5074 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5075 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5076 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5077 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5078 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5079 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5080 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5081 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5082 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5083 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5084 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5085 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5086 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5087
5088 return (features & feature) == feature;
5089}
5090
5091static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5092{
5093 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5094 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5095}
5096
5097static inline bool netif_needs_gso(struct sk_buff *skb,
5098 netdev_features_t features)
5099{
5100 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5101 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5102 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5103}
5104
5105void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5106void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5107void netif_inherit_tso_max(struct net_device *to,
5108 const struct net_device *from);
5109
5110static inline bool netif_is_macsec(const struct net_device *dev)
5111{
5112 return dev->priv_flags & IFF_MACSEC;
5113}
5114
5115static inline bool netif_is_macvlan(const struct net_device *dev)
5116{
5117 return dev->priv_flags & IFF_MACVLAN;
5118}
5119
5120static inline bool netif_is_macvlan_port(const struct net_device *dev)
5121{
5122 return dev->priv_flags & IFF_MACVLAN_PORT;
5123}
5124
5125static inline bool netif_is_bond_master(const struct net_device *dev)
5126{
5127 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5128}
5129
5130static inline bool netif_is_bond_slave(const struct net_device *dev)
5131{
5132 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5133}
5134
5135static inline bool netif_supports_nofcs(struct net_device *dev)
5136{
5137 return dev->priv_flags & IFF_SUPP_NOFCS;
5138}
5139
5140static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5141{
5142 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5143}
5144
5145static inline bool netif_is_l3_master(const struct net_device *dev)
5146{
5147 return dev->priv_flags & IFF_L3MDEV_MASTER;
5148}
5149
5150static inline bool netif_is_l3_slave(const struct net_device *dev)
5151{
5152 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5153}
5154
5155static inline int dev_sdif(const struct net_device *dev)
5156{
5157#ifdef CONFIG_NET_L3_MASTER_DEV
5158 if (netif_is_l3_slave(dev))
5159 return dev->ifindex;
5160#endif
5161 return 0;
5162}
5163
5164static inline bool netif_is_bridge_master(const struct net_device *dev)
5165{
5166 return dev->priv_flags & IFF_EBRIDGE;
5167}
5168
5169static inline bool netif_is_bridge_port(const struct net_device *dev)
5170{
5171 return dev->priv_flags & IFF_BRIDGE_PORT;
5172}
5173
5174static inline bool netif_is_ovs_master(const struct net_device *dev)
5175{
5176 return dev->priv_flags & IFF_OPENVSWITCH;
5177}
5178
5179static inline bool netif_is_ovs_port(const struct net_device *dev)
5180{
5181 return dev->priv_flags & IFF_OVS_DATAPATH;
5182}
5183
5184static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5185{
5186 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5187}
5188
5189static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5190{
5191 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5192}
5193
5194static inline bool netif_is_team_master(const struct net_device *dev)
5195{
5196 return dev->priv_flags & IFF_TEAM;
5197}
5198
5199static inline bool netif_is_team_port(const struct net_device *dev)
5200{
5201 return dev->priv_flags & IFF_TEAM_PORT;
5202}
5203
5204static inline bool netif_is_lag_master(const struct net_device *dev)
5205{
5206 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5207}
5208
5209static inline bool netif_is_lag_port(const struct net_device *dev)
5210{
5211 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5212}
5213
5214static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5215{
5216 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5217}
5218
5219static inline bool netif_is_failover(const struct net_device *dev)
5220{
5221 return dev->priv_flags & IFF_FAILOVER;
5222}
5223
5224static inline bool netif_is_failover_slave(const struct net_device *dev)
5225{
5226 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5227}
5228
5229/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5230static inline void netif_keep_dst(struct net_device *dev)
5231{
5232 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5233}
5234
5235/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5236static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5237{
5238 /* TODO: reserve and use an additional IFF bit, if we get more users */
5239 return netif_is_macsec(dev);
5240}
5241
5242extern struct pernet_operations __net_initdata loopback_net_ops;
5243
5244/* Logging, debugging and troubleshooting/diagnostic helpers. */
5245
5246/* netdev_printk helpers, similar to dev_printk */
5247
5248static inline const char *netdev_name(const struct net_device *dev)
5249{
5250 if (!dev->name[0] || strchr(dev->name, '%'))
5251 return "(unnamed net_device)";
5252 return dev->name;
5253}
5254
5255static inline const char *netdev_reg_state(const struct net_device *dev)
5256{
5257 switch (dev->reg_state) {
5258 case NETREG_UNINITIALIZED: return " (uninitialized)";
5259 case NETREG_REGISTERED: return "";
5260 case NETREG_UNREGISTERING: return " (unregistering)";
5261 case NETREG_UNREGISTERED: return " (unregistered)";
5262 case NETREG_RELEASED: return " (released)";
5263 case NETREG_DUMMY: return " (dummy)";
5264 }
5265
5266 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5267 return " (unknown)";
5268}
5269
5270#define MODULE_ALIAS_NETDEV(device) \
5271 MODULE_ALIAS("netdev-" device)
5272
5273/*
5274 * netdev_WARN() acts like dev_printk(), but with the key difference
5275 * of using a WARN/WARN_ON to get the message out, including the
5276 * file/line information and a backtrace.
5277 */
5278#define netdev_WARN(dev, format, args...) \
5279 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5280 netdev_reg_state(dev), ##args)
5281
5282#define netdev_WARN_ONCE(dev, format, args...) \
5283 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5284 netdev_reg_state(dev), ##args)
5285
5286/*
5287 * The list of packet types we will receive (as opposed to discard)
5288 * and the routines to invoke.
5289 *
5290 * Why 16. Because with 16 the only overlap we get on a hash of the
5291 * low nibble of the protocol value is RARP/SNAP/X.25.
5292 *
5293 * 0800 IP
5294 * 0001 802.3
5295 * 0002 AX.25
5296 * 0004 802.2
5297 * 8035 RARP
5298 * 0005 SNAP
5299 * 0805 X.25
5300 * 0806 ARP
5301 * 8137 IPX
5302 * 0009 Localtalk
5303 * 86DD IPv6
5304 */
5305#define PTYPE_HASH_SIZE (16)
5306#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5307
5308extern struct list_head ptype_all __read_mostly;
5309extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5310
5311extern struct net_device *blackhole_netdev;
5312
5313/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5314#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5315#define DEV_STATS_ADD(DEV, FIELD, VAL) \
5316 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5317#define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5318
5319#endif /* _LINUX_NETDEVICE_H */