Loading...
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
32#include <linux/percpu.h>
33#include <linux/rculist.h>
34#include <linux/workqueue.h>
35#include <linux/dynamic_queue_limits.h>
36
37#include <linux/ethtool.h>
38#include <net/net_namespace.h>
39#ifdef CONFIG_DCB
40#include <net/dcbnl.h>
41#endif
42#include <net/netprio_cgroup.h>
43#include <net/xdp.h>
44
45#include <linux/netdev_features.h>
46#include <linux/neighbour.h>
47#include <uapi/linux/netdevice.h>
48#include <uapi/linux/if_bonding.h>
49#include <uapi/linux/pkt_cls.h>
50#include <linux/hashtable.h>
51
52struct netpoll_info;
53struct device;
54struct phy_device;
55struct dsa_port;
56struct ip_tunnel_parm;
57struct macsec_context;
58struct macsec_ops;
59
60struct sfp_bus;
61/* 802.11 specific */
62struct wireless_dev;
63/* 802.15.4 specific */
64struct wpan_dev;
65struct mpls_dev;
66/* UDP Tunnel offloads */
67struct udp_tunnel_info;
68struct udp_tunnel_nic_info;
69struct udp_tunnel_nic;
70struct bpf_prog;
71struct xdp_buff;
72
73void netdev_set_default_ethtool_ops(struct net_device *dev,
74 const struct ethtool_ops *ops);
75
76/* Backlog congestion levels */
77#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
78#define NET_RX_DROP 1 /* packet dropped */
79
80#define MAX_NEST_DEV 8
81
82/*
83 * Transmit return codes: transmit return codes originate from three different
84 * namespaces:
85 *
86 * - qdisc return codes
87 * - driver transmit return codes
88 * - errno values
89 *
90 * Drivers are allowed to return any one of those in their hard_start_xmit()
91 * function. Real network devices commonly used with qdiscs should only return
92 * the driver transmit return codes though - when qdiscs are used, the actual
93 * transmission happens asynchronously, so the value is not propagated to
94 * higher layers. Virtual network devices transmit synchronously; in this case
95 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
96 * others are propagated to higher layers.
97 */
98
99/* qdisc ->enqueue() return codes. */
100#define NET_XMIT_SUCCESS 0x00
101#define NET_XMIT_DROP 0x01 /* skb dropped */
102#define NET_XMIT_CN 0x02 /* congestion notification */
103#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
104
105/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
106 * indicates that the device will soon be dropping packets, or already drops
107 * some packets of the same priority; prompting us to send less aggressively. */
108#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
109#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
110
111/* Driver transmit return codes */
112#define NETDEV_TX_MASK 0xf0
113
114enum netdev_tx {
115 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
116 NETDEV_TX_OK = 0x00, /* driver took care of packet */
117 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
118};
119typedef enum netdev_tx netdev_tx_t;
120
121/*
122 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
123 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
124 */
125static inline bool dev_xmit_complete(int rc)
126{
127 /*
128 * Positive cases with an skb consumed by a driver:
129 * - successful transmission (rc == NETDEV_TX_OK)
130 * - error while transmitting (rc < 0)
131 * - error while queueing to a different device (rc & NET_XMIT_MASK)
132 */
133 if (likely(rc < NET_XMIT_MASK))
134 return true;
135
136 return false;
137}
138
139/*
140 * Compute the worst-case header length according to the protocols
141 * used.
142 */
143
144#if defined(CONFIG_HYPERV_NET)
145# define LL_MAX_HEADER 128
146#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
147# if defined(CONFIG_MAC80211_MESH)
148# define LL_MAX_HEADER 128
149# else
150# define LL_MAX_HEADER 96
151# endif
152#else
153# define LL_MAX_HEADER 32
154#endif
155
156#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
157 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
158#define MAX_HEADER LL_MAX_HEADER
159#else
160#define MAX_HEADER (LL_MAX_HEADER + 48)
161#endif
162
163/*
164 * Old network device statistics. Fields are native words
165 * (unsigned long) so they can be read and written atomically.
166 */
167
168struct net_device_stats {
169 unsigned long rx_packets;
170 unsigned long tx_packets;
171 unsigned long rx_bytes;
172 unsigned long tx_bytes;
173 unsigned long rx_errors;
174 unsigned long tx_errors;
175 unsigned long rx_dropped;
176 unsigned long tx_dropped;
177 unsigned long multicast;
178 unsigned long collisions;
179 unsigned long rx_length_errors;
180 unsigned long rx_over_errors;
181 unsigned long rx_crc_errors;
182 unsigned long rx_frame_errors;
183 unsigned long rx_fifo_errors;
184 unsigned long rx_missed_errors;
185 unsigned long tx_aborted_errors;
186 unsigned long tx_carrier_errors;
187 unsigned long tx_fifo_errors;
188 unsigned long tx_heartbeat_errors;
189 unsigned long tx_window_errors;
190 unsigned long rx_compressed;
191 unsigned long tx_compressed;
192};
193
194
195#include <linux/cache.h>
196#include <linux/skbuff.h>
197
198#ifdef CONFIG_RPS
199#include <linux/static_key.h>
200extern struct static_key_false rps_needed;
201extern struct static_key_false rfs_needed;
202#endif
203
204struct neighbour;
205struct neigh_parms;
206struct sk_buff;
207
208struct netdev_hw_addr {
209 struct list_head list;
210 unsigned char addr[MAX_ADDR_LEN];
211 unsigned char type;
212#define NETDEV_HW_ADDR_T_LAN 1
213#define NETDEV_HW_ADDR_T_SAN 2
214#define NETDEV_HW_ADDR_T_SLAVE 3
215#define NETDEV_HW_ADDR_T_UNICAST 4
216#define NETDEV_HW_ADDR_T_MULTICAST 5
217 bool global_use;
218 int sync_cnt;
219 int refcount;
220 int synced;
221 struct rcu_head rcu_head;
222};
223
224struct netdev_hw_addr_list {
225 struct list_head list;
226 int count;
227};
228
229#define netdev_hw_addr_list_count(l) ((l)->count)
230#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
231#define netdev_hw_addr_list_for_each(ha, l) \
232 list_for_each_entry(ha, &(l)->list, list)
233
234#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
235#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
236#define netdev_for_each_uc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
238
239#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
240#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
241#define netdev_for_each_mc_addr(ha, dev) \
242 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
243
244struct hh_cache {
245 unsigned int hh_len;
246 seqlock_t hh_lock;
247
248 /* cached hardware header; allow for machine alignment needs. */
249#define HH_DATA_MOD 16
250#define HH_DATA_OFF(__len) \
251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
252#define HH_DATA_ALIGN(__len) \
253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
255};
256
257/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
258 * Alternative is:
259 * dev->hard_header_len ? (dev->hard_header_len +
260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
261 *
262 * We could use other alignment values, but we must maintain the
263 * relationship HH alignment <= LL alignment.
264 */
265#define LL_RESERVED_SPACE(dev) \
266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
267#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
269
270struct header_ops {
271 int (*create) (struct sk_buff *skb, struct net_device *dev,
272 unsigned short type, const void *daddr,
273 const void *saddr, unsigned int len);
274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
276 void (*cache_update)(struct hh_cache *hh,
277 const struct net_device *dev,
278 const unsigned char *haddr);
279 bool (*validate)(const char *ll_header, unsigned int len);
280 __be16 (*parse_protocol)(const struct sk_buff *skb);
281};
282
283/* These flag bits are private to the generic network queueing
284 * layer; they may not be explicitly referenced by any other
285 * code.
286 */
287
288enum netdev_state_t {
289 __LINK_STATE_START,
290 __LINK_STATE_PRESENT,
291 __LINK_STATE_NOCARRIER,
292 __LINK_STATE_LINKWATCH_PENDING,
293 __LINK_STATE_DORMANT,
294 __LINK_STATE_TESTING,
295};
296
297
298/*
299 * This structure holds boot-time configured netdevice settings. They
300 * are then used in the device probing.
301 */
302struct netdev_boot_setup {
303 char name[IFNAMSIZ];
304 struct ifmap map;
305};
306#define NETDEV_BOOT_SETUP_MAX 8
307
308int __init netdev_boot_setup(char *str);
309
310struct gro_list {
311 struct list_head list;
312 int count;
313};
314
315/*
316 * size of gro hash buckets, must less than bit number of
317 * napi_struct::gro_bitmask
318 */
319#define GRO_HASH_BUCKETS 8
320
321/*
322 * Structure for NAPI scheduling similar to tasklet but with weighting
323 */
324struct napi_struct {
325 /* The poll_list must only be managed by the entity which
326 * changes the state of the NAPI_STATE_SCHED bit. This means
327 * whoever atomically sets that bit can add this napi_struct
328 * to the per-CPU poll_list, and whoever clears that bit
329 * can remove from the list right before clearing the bit.
330 */
331 struct list_head poll_list;
332
333 unsigned long state;
334 int weight;
335 int defer_hard_irqs_count;
336 unsigned long gro_bitmask;
337 int (*poll)(struct napi_struct *, int);
338#ifdef CONFIG_NETPOLL
339 int poll_owner;
340#endif
341 struct net_device *dev;
342 struct gro_list gro_hash[GRO_HASH_BUCKETS];
343 struct sk_buff *skb;
344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
345 int rx_count; /* length of rx_list */
346 struct hrtimer timer;
347 struct list_head dev_list;
348 struct hlist_node napi_hash_node;
349 unsigned int napi_id;
350};
351
352enum {
353 NAPI_STATE_SCHED, /* Poll is scheduled */
354 NAPI_STATE_MISSED, /* reschedule a napi */
355 NAPI_STATE_DISABLE, /* Disable pending */
356 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
357 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
358 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
359 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
360};
361
362enum {
363 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
364 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
365 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
366 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
367 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
368 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
369 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
370};
371
372enum gro_result {
373 GRO_MERGED,
374 GRO_MERGED_FREE,
375 GRO_HELD,
376 GRO_NORMAL,
377 GRO_DROP,
378 GRO_CONSUMED,
379};
380typedef enum gro_result gro_result_t;
381
382/*
383 * enum rx_handler_result - Possible return values for rx_handlers.
384 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
385 * further.
386 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
387 * case skb->dev was changed by rx_handler.
388 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
389 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
390 *
391 * rx_handlers are functions called from inside __netif_receive_skb(), to do
392 * special processing of the skb, prior to delivery to protocol handlers.
393 *
394 * Currently, a net_device can only have a single rx_handler registered. Trying
395 * to register a second rx_handler will return -EBUSY.
396 *
397 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
398 * To unregister a rx_handler on a net_device, use
399 * netdev_rx_handler_unregister().
400 *
401 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
402 * do with the skb.
403 *
404 * If the rx_handler consumed the skb in some way, it should return
405 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
406 * the skb to be delivered in some other way.
407 *
408 * If the rx_handler changed skb->dev, to divert the skb to another
409 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
410 * new device will be called if it exists.
411 *
412 * If the rx_handler decides the skb should be ignored, it should return
413 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
414 * are registered on exact device (ptype->dev == skb->dev).
415 *
416 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
417 * delivered, it should return RX_HANDLER_PASS.
418 *
419 * A device without a registered rx_handler will behave as if rx_handler
420 * returned RX_HANDLER_PASS.
421 */
422
423enum rx_handler_result {
424 RX_HANDLER_CONSUMED,
425 RX_HANDLER_ANOTHER,
426 RX_HANDLER_EXACT,
427 RX_HANDLER_PASS,
428};
429typedef enum rx_handler_result rx_handler_result_t;
430typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
431
432void __napi_schedule(struct napi_struct *n);
433void __napi_schedule_irqoff(struct napi_struct *n);
434
435static inline bool napi_disable_pending(struct napi_struct *n)
436{
437 return test_bit(NAPI_STATE_DISABLE, &n->state);
438}
439
440bool napi_schedule_prep(struct napi_struct *n);
441
442/**
443 * napi_schedule - schedule NAPI poll
444 * @n: NAPI context
445 *
446 * Schedule NAPI poll routine to be called if it is not already
447 * running.
448 */
449static inline void napi_schedule(struct napi_struct *n)
450{
451 if (napi_schedule_prep(n))
452 __napi_schedule(n);
453}
454
455/**
456 * napi_schedule_irqoff - schedule NAPI poll
457 * @n: NAPI context
458 *
459 * Variant of napi_schedule(), assuming hard irqs are masked.
460 */
461static inline void napi_schedule_irqoff(struct napi_struct *n)
462{
463 if (napi_schedule_prep(n))
464 __napi_schedule_irqoff(n);
465}
466
467/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
468static inline bool napi_reschedule(struct napi_struct *napi)
469{
470 if (napi_schedule_prep(napi)) {
471 __napi_schedule(napi);
472 return true;
473 }
474 return false;
475}
476
477bool napi_complete_done(struct napi_struct *n, int work_done);
478/**
479 * napi_complete - NAPI processing complete
480 * @n: NAPI context
481 *
482 * Mark NAPI processing as complete.
483 * Consider using napi_complete_done() instead.
484 * Return false if device should avoid rearming interrupts.
485 */
486static inline bool napi_complete(struct napi_struct *n)
487{
488 return napi_complete_done(n, 0);
489}
490
491/**
492 * napi_hash_del - remove a NAPI from global table
493 * @napi: NAPI context
494 *
495 * Warning: caller must observe RCU grace period
496 * before freeing memory containing @napi, if
497 * this function returns true.
498 * Note: core networking stack automatically calls it
499 * from netif_napi_del().
500 * Drivers might want to call this helper to combine all
501 * the needed RCU grace periods into a single one.
502 */
503bool napi_hash_del(struct napi_struct *napi);
504
505/**
506 * napi_disable - prevent NAPI from scheduling
507 * @n: NAPI context
508 *
509 * Stop NAPI from being scheduled on this context.
510 * Waits till any outstanding processing completes.
511 */
512void napi_disable(struct napi_struct *n);
513
514/**
515 * napi_enable - enable NAPI scheduling
516 * @n: NAPI context
517 *
518 * Resume NAPI from being scheduled on this context.
519 * Must be paired with napi_disable.
520 */
521static inline void napi_enable(struct napi_struct *n)
522{
523 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
524 smp_mb__before_atomic();
525 clear_bit(NAPI_STATE_SCHED, &n->state);
526 clear_bit(NAPI_STATE_NPSVC, &n->state);
527}
528
529/**
530 * napi_synchronize - wait until NAPI is not running
531 * @n: NAPI context
532 *
533 * Wait until NAPI is done being scheduled on this context.
534 * Waits till any outstanding processing completes but
535 * does not disable future activations.
536 */
537static inline void napi_synchronize(const struct napi_struct *n)
538{
539 if (IS_ENABLED(CONFIG_SMP))
540 while (test_bit(NAPI_STATE_SCHED, &n->state))
541 msleep(1);
542 else
543 barrier();
544}
545
546/**
547 * napi_if_scheduled_mark_missed - if napi is running, set the
548 * NAPIF_STATE_MISSED
549 * @n: NAPI context
550 *
551 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
552 * NAPI is scheduled.
553 **/
554static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
555{
556 unsigned long val, new;
557
558 do {
559 val = READ_ONCE(n->state);
560 if (val & NAPIF_STATE_DISABLE)
561 return true;
562
563 if (!(val & NAPIF_STATE_SCHED))
564 return false;
565
566 new = val | NAPIF_STATE_MISSED;
567 } while (cmpxchg(&n->state, val, new) != val);
568
569 return true;
570}
571
572enum netdev_queue_state_t {
573 __QUEUE_STATE_DRV_XOFF,
574 __QUEUE_STATE_STACK_XOFF,
575 __QUEUE_STATE_FROZEN,
576};
577
578#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
579#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
580#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
581
582#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
583#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
584 QUEUE_STATE_FROZEN)
585#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
586 QUEUE_STATE_FROZEN)
587
588/*
589 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
590 * netif_tx_* functions below are used to manipulate this flag. The
591 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
592 * queue independently. The netif_xmit_*stopped functions below are called
593 * to check if the queue has been stopped by the driver or stack (either
594 * of the XOFF bits are set in the state). Drivers should not need to call
595 * netif_xmit*stopped functions, they should only be using netif_tx_*.
596 */
597
598struct netdev_queue {
599/*
600 * read-mostly part
601 */
602 struct net_device *dev;
603 struct Qdisc __rcu *qdisc;
604 struct Qdisc *qdisc_sleeping;
605#ifdef CONFIG_SYSFS
606 struct kobject kobj;
607#endif
608#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
609 int numa_node;
610#endif
611 unsigned long tx_maxrate;
612 /*
613 * Number of TX timeouts for this queue
614 * (/sys/class/net/DEV/Q/trans_timeout)
615 */
616 unsigned long trans_timeout;
617
618 /* Subordinate device that the queue has been assigned to */
619 struct net_device *sb_dev;
620#ifdef CONFIG_XDP_SOCKETS
621 struct xdp_umem *umem;
622#endif
623/*
624 * write-mostly part
625 */
626 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
627 int xmit_lock_owner;
628 /*
629 * Time (in jiffies) of last Tx
630 */
631 unsigned long trans_start;
632
633 unsigned long state;
634
635#ifdef CONFIG_BQL
636 struct dql dql;
637#endif
638} ____cacheline_aligned_in_smp;
639
640extern int sysctl_fb_tunnels_only_for_init_net;
641extern int sysctl_devconf_inherit_init_net;
642
643static inline bool net_has_fallback_tunnels(const struct net *net)
644{
645 return net == &init_net ||
646 !IS_ENABLED(CONFIG_SYSCTL) ||
647 !sysctl_fb_tunnels_only_for_init_net;
648}
649
650static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
651{
652#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
653 return q->numa_node;
654#else
655 return NUMA_NO_NODE;
656#endif
657}
658
659static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
660{
661#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
662 q->numa_node = node;
663#endif
664}
665
666#ifdef CONFIG_RPS
667/*
668 * This structure holds an RPS map which can be of variable length. The
669 * map is an array of CPUs.
670 */
671struct rps_map {
672 unsigned int len;
673 struct rcu_head rcu;
674 u16 cpus[];
675};
676#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
677
678/*
679 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
680 * tail pointer for that CPU's input queue at the time of last enqueue, and
681 * a hardware filter index.
682 */
683struct rps_dev_flow {
684 u16 cpu;
685 u16 filter;
686 unsigned int last_qtail;
687};
688#define RPS_NO_FILTER 0xffff
689
690/*
691 * The rps_dev_flow_table structure contains a table of flow mappings.
692 */
693struct rps_dev_flow_table {
694 unsigned int mask;
695 struct rcu_head rcu;
696 struct rps_dev_flow flows[];
697};
698#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
699 ((_num) * sizeof(struct rps_dev_flow)))
700
701/*
702 * The rps_sock_flow_table contains mappings of flows to the last CPU
703 * on which they were processed by the application (set in recvmsg).
704 * Each entry is a 32bit value. Upper part is the high-order bits
705 * of flow hash, lower part is CPU number.
706 * rps_cpu_mask is used to partition the space, depending on number of
707 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
708 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
709 * meaning we use 32-6=26 bits for the hash.
710 */
711struct rps_sock_flow_table {
712 u32 mask;
713
714 u32 ents[] ____cacheline_aligned_in_smp;
715};
716#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
717
718#define RPS_NO_CPU 0xffff
719
720extern u32 rps_cpu_mask;
721extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
722
723static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
724 u32 hash)
725{
726 if (table && hash) {
727 unsigned int index = hash & table->mask;
728 u32 val = hash & ~rps_cpu_mask;
729
730 /* We only give a hint, preemption can change CPU under us */
731 val |= raw_smp_processor_id();
732
733 if (table->ents[index] != val)
734 table->ents[index] = val;
735 }
736}
737
738#ifdef CONFIG_RFS_ACCEL
739bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
740 u16 filter_id);
741#endif
742#endif /* CONFIG_RPS */
743
744/* This structure contains an instance of an RX queue. */
745struct netdev_rx_queue {
746#ifdef CONFIG_RPS
747 struct rps_map __rcu *rps_map;
748 struct rps_dev_flow_table __rcu *rps_flow_table;
749#endif
750 struct kobject kobj;
751 struct net_device *dev;
752 struct xdp_rxq_info xdp_rxq;
753#ifdef CONFIG_XDP_SOCKETS
754 struct xdp_umem *umem;
755#endif
756} ____cacheline_aligned_in_smp;
757
758/*
759 * RX queue sysfs structures and functions.
760 */
761struct rx_queue_attribute {
762 struct attribute attr;
763 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
764 ssize_t (*store)(struct netdev_rx_queue *queue,
765 const char *buf, size_t len);
766};
767
768#ifdef CONFIG_XPS
769/*
770 * This structure holds an XPS map which can be of variable length. The
771 * map is an array of queues.
772 */
773struct xps_map {
774 unsigned int len;
775 unsigned int alloc_len;
776 struct rcu_head rcu;
777 u16 queues[];
778};
779#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
780#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
781 - sizeof(struct xps_map)) / sizeof(u16))
782
783/*
784 * This structure holds all XPS maps for device. Maps are indexed by CPU.
785 */
786struct xps_dev_maps {
787 struct rcu_head rcu;
788 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
789};
790
791#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
792 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
793
794#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
795 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
796
797#endif /* CONFIG_XPS */
798
799#define TC_MAX_QUEUE 16
800#define TC_BITMASK 15
801/* HW offloaded queuing disciplines txq count and offset maps */
802struct netdev_tc_txq {
803 u16 count;
804 u16 offset;
805};
806
807#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
808/*
809 * This structure is to hold information about the device
810 * configured to run FCoE protocol stack.
811 */
812struct netdev_fcoe_hbainfo {
813 char manufacturer[64];
814 char serial_number[64];
815 char hardware_version[64];
816 char driver_version[64];
817 char optionrom_version[64];
818 char firmware_version[64];
819 char model[256];
820 char model_description[256];
821};
822#endif
823
824#define MAX_PHYS_ITEM_ID_LEN 32
825
826/* This structure holds a unique identifier to identify some
827 * physical item (port for example) used by a netdevice.
828 */
829struct netdev_phys_item_id {
830 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
831 unsigned char id_len;
832};
833
834static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
835 struct netdev_phys_item_id *b)
836{
837 return a->id_len == b->id_len &&
838 memcmp(a->id, b->id, a->id_len) == 0;
839}
840
841typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
842 struct sk_buff *skb,
843 struct net_device *sb_dev);
844
845enum tc_setup_type {
846 TC_SETUP_QDISC_MQPRIO,
847 TC_SETUP_CLSU32,
848 TC_SETUP_CLSFLOWER,
849 TC_SETUP_CLSMATCHALL,
850 TC_SETUP_CLSBPF,
851 TC_SETUP_BLOCK,
852 TC_SETUP_QDISC_CBS,
853 TC_SETUP_QDISC_RED,
854 TC_SETUP_QDISC_PRIO,
855 TC_SETUP_QDISC_MQ,
856 TC_SETUP_QDISC_ETF,
857 TC_SETUP_ROOT_QDISC,
858 TC_SETUP_QDISC_GRED,
859 TC_SETUP_QDISC_TAPRIO,
860 TC_SETUP_FT,
861 TC_SETUP_QDISC_ETS,
862 TC_SETUP_QDISC_TBF,
863 TC_SETUP_QDISC_FIFO,
864};
865
866/* These structures hold the attributes of bpf state that are being passed
867 * to the netdevice through the bpf op.
868 */
869enum bpf_netdev_command {
870 /* Set or clear a bpf program used in the earliest stages of packet
871 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
872 * is responsible for calling bpf_prog_put on any old progs that are
873 * stored. In case of error, the callee need not release the new prog
874 * reference, but on success it takes ownership and must bpf_prog_put
875 * when it is no longer used.
876 */
877 XDP_SETUP_PROG,
878 XDP_SETUP_PROG_HW,
879 /* BPF program for offload callbacks, invoked at program load time. */
880 BPF_OFFLOAD_MAP_ALLOC,
881 BPF_OFFLOAD_MAP_FREE,
882 XDP_SETUP_XSK_UMEM,
883};
884
885struct bpf_prog_offload_ops;
886struct netlink_ext_ack;
887struct xdp_umem;
888struct xdp_dev_bulk_queue;
889struct bpf_xdp_link;
890
891enum bpf_xdp_mode {
892 XDP_MODE_SKB = 0,
893 XDP_MODE_DRV = 1,
894 XDP_MODE_HW = 2,
895 __MAX_XDP_MODE
896};
897
898struct bpf_xdp_entity {
899 struct bpf_prog *prog;
900 struct bpf_xdp_link *link;
901};
902
903struct netdev_bpf {
904 enum bpf_netdev_command command;
905 union {
906 /* XDP_SETUP_PROG */
907 struct {
908 u32 flags;
909 struct bpf_prog *prog;
910 struct netlink_ext_ack *extack;
911 };
912 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
913 struct {
914 struct bpf_offloaded_map *offmap;
915 };
916 /* XDP_SETUP_XSK_UMEM */
917 struct {
918 struct xdp_umem *umem;
919 u16 queue_id;
920 } xsk;
921 };
922};
923
924/* Flags for ndo_xsk_wakeup. */
925#define XDP_WAKEUP_RX (1 << 0)
926#define XDP_WAKEUP_TX (1 << 1)
927
928#ifdef CONFIG_XFRM_OFFLOAD
929struct xfrmdev_ops {
930 int (*xdo_dev_state_add) (struct xfrm_state *x);
931 void (*xdo_dev_state_delete) (struct xfrm_state *x);
932 void (*xdo_dev_state_free) (struct xfrm_state *x);
933 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
934 struct xfrm_state *x);
935 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
936};
937#endif
938
939struct dev_ifalias {
940 struct rcu_head rcuhead;
941 char ifalias[];
942};
943
944struct devlink;
945struct tlsdev_ops;
946
947struct netdev_name_node {
948 struct hlist_node hlist;
949 struct list_head list;
950 struct net_device *dev;
951 const char *name;
952};
953
954int netdev_name_node_alt_create(struct net_device *dev, const char *name);
955int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
956
957struct netdev_net_notifier {
958 struct list_head list;
959 struct notifier_block *nb;
960};
961
962/*
963 * This structure defines the management hooks for network devices.
964 * The following hooks can be defined; unless noted otherwise, they are
965 * optional and can be filled with a null pointer.
966 *
967 * int (*ndo_init)(struct net_device *dev);
968 * This function is called once when a network device is registered.
969 * The network device can use this for any late stage initialization
970 * or semantic validation. It can fail with an error code which will
971 * be propagated back to register_netdev.
972 *
973 * void (*ndo_uninit)(struct net_device *dev);
974 * This function is called when device is unregistered or when registration
975 * fails. It is not called if init fails.
976 *
977 * int (*ndo_open)(struct net_device *dev);
978 * This function is called when a network device transitions to the up
979 * state.
980 *
981 * int (*ndo_stop)(struct net_device *dev);
982 * This function is called when a network device transitions to the down
983 * state.
984 *
985 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
986 * struct net_device *dev);
987 * Called when a packet needs to be transmitted.
988 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
989 * the queue before that can happen; it's for obsolete devices and weird
990 * corner cases, but the stack really does a non-trivial amount
991 * of useless work if you return NETDEV_TX_BUSY.
992 * Required; cannot be NULL.
993 *
994 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
995 * struct net_device *dev
996 * netdev_features_t features);
997 * Called by core transmit path to determine if device is capable of
998 * performing offload operations on a given packet. This is to give
999 * the device an opportunity to implement any restrictions that cannot
1000 * be otherwise expressed by feature flags. The check is called with
1001 * the set of features that the stack has calculated and it returns
1002 * those the driver believes to be appropriate.
1003 *
1004 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1005 * struct net_device *sb_dev);
1006 * Called to decide which queue to use when device supports multiple
1007 * transmit queues.
1008 *
1009 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1010 * This function is called to allow device receiver to make
1011 * changes to configuration when multicast or promiscuous is enabled.
1012 *
1013 * void (*ndo_set_rx_mode)(struct net_device *dev);
1014 * This function is called device changes address list filtering.
1015 * If driver handles unicast address filtering, it should set
1016 * IFF_UNICAST_FLT in its priv_flags.
1017 *
1018 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1019 * This function is called when the Media Access Control address
1020 * needs to be changed. If this interface is not defined, the
1021 * MAC address can not be changed.
1022 *
1023 * int (*ndo_validate_addr)(struct net_device *dev);
1024 * Test if Media Access Control address is valid for the device.
1025 *
1026 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1027 * Called when a user requests an ioctl which can't be handled by
1028 * the generic interface code. If not defined ioctls return
1029 * not supported error code.
1030 *
1031 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1032 * Used to set network devices bus interface parameters. This interface
1033 * is retained for legacy reasons; new devices should use the bus
1034 * interface (PCI) for low level management.
1035 *
1036 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1037 * Called when a user wants to change the Maximum Transfer Unit
1038 * of a device.
1039 *
1040 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1041 * Callback used when the transmitter has not made any progress
1042 * for dev->watchdog ticks.
1043 *
1044 * void (*ndo_get_stats64)(struct net_device *dev,
1045 * struct rtnl_link_stats64 *storage);
1046 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1047 * Called when a user wants to get the network device usage
1048 * statistics. Drivers must do one of the following:
1049 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1050 * rtnl_link_stats64 structure passed by the caller.
1051 * 2. Define @ndo_get_stats to update a net_device_stats structure
1052 * (which should normally be dev->stats) and return a pointer to
1053 * it. The structure may be changed asynchronously only if each
1054 * field is written atomically.
1055 * 3. Update dev->stats asynchronously and atomically, and define
1056 * neither operation.
1057 *
1058 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1059 * Return true if this device supports offload stats of this attr_id.
1060 *
1061 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1062 * void *attr_data)
1063 * Get statistics for offload operations by attr_id. Write it into the
1064 * attr_data pointer.
1065 *
1066 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1067 * If device supports VLAN filtering this function is called when a
1068 * VLAN id is registered.
1069 *
1070 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1071 * If device supports VLAN filtering this function is called when a
1072 * VLAN id is unregistered.
1073 *
1074 * void (*ndo_poll_controller)(struct net_device *dev);
1075 *
1076 * SR-IOV management functions.
1077 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1078 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1079 * u8 qos, __be16 proto);
1080 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1081 * int max_tx_rate);
1082 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1083 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1084 * int (*ndo_get_vf_config)(struct net_device *dev,
1085 * int vf, struct ifla_vf_info *ivf);
1086 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1087 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1088 * struct nlattr *port[]);
1089 *
1090 * Enable or disable the VF ability to query its RSS Redirection Table and
1091 * Hash Key. This is needed since on some devices VF share this information
1092 * with PF and querying it may introduce a theoretical security risk.
1093 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1094 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1095 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1096 * void *type_data);
1097 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1098 * This is always called from the stack with the rtnl lock held and netif
1099 * tx queues stopped. This allows the netdevice to perform queue
1100 * management safely.
1101 *
1102 * Fiber Channel over Ethernet (FCoE) offload functions.
1103 * int (*ndo_fcoe_enable)(struct net_device *dev);
1104 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1105 * so the underlying device can perform whatever needed configuration or
1106 * initialization to support acceleration of FCoE traffic.
1107 *
1108 * int (*ndo_fcoe_disable)(struct net_device *dev);
1109 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1110 * so the underlying device can perform whatever needed clean-ups to
1111 * stop supporting acceleration of FCoE traffic.
1112 *
1113 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1114 * struct scatterlist *sgl, unsigned int sgc);
1115 * Called when the FCoE Initiator wants to initialize an I/O that
1116 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1117 * perform necessary setup and returns 1 to indicate the device is set up
1118 * successfully to perform DDP on this I/O, otherwise this returns 0.
1119 *
1120 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1121 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1122 * indicated by the FC exchange id 'xid', so the underlying device can
1123 * clean up and reuse resources for later DDP requests.
1124 *
1125 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1126 * struct scatterlist *sgl, unsigned int sgc);
1127 * Called when the FCoE Target wants to initialize an I/O that
1128 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1129 * perform necessary setup and returns 1 to indicate the device is set up
1130 * successfully to perform DDP on this I/O, otherwise this returns 0.
1131 *
1132 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1133 * struct netdev_fcoe_hbainfo *hbainfo);
1134 * Called when the FCoE Protocol stack wants information on the underlying
1135 * device. This information is utilized by the FCoE protocol stack to
1136 * register attributes with Fiber Channel management service as per the
1137 * FC-GS Fabric Device Management Information(FDMI) specification.
1138 *
1139 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1140 * Called when the underlying device wants to override default World Wide
1141 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1142 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1143 * protocol stack to use.
1144 *
1145 * RFS acceleration.
1146 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1147 * u16 rxq_index, u32 flow_id);
1148 * Set hardware filter for RFS. rxq_index is the target queue index;
1149 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1150 * Return the filter ID on success, or a negative error code.
1151 *
1152 * Slave management functions (for bridge, bonding, etc).
1153 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1154 * Called to make another netdev an underling.
1155 *
1156 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1157 * Called to release previously enslaved netdev.
1158 *
1159 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1160 * struct sk_buff *skb,
1161 * bool all_slaves);
1162 * Get the xmit slave of master device. If all_slaves is true, function
1163 * assume all the slaves can transmit.
1164 *
1165 * Feature/offload setting functions.
1166 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1167 * netdev_features_t features);
1168 * Adjusts the requested feature flags according to device-specific
1169 * constraints, and returns the resulting flags. Must not modify
1170 * the device state.
1171 *
1172 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1173 * Called to update device configuration to new features. Passed
1174 * feature set might be less than what was returned by ndo_fix_features()).
1175 * Must return >0 or -errno if it changed dev->features itself.
1176 *
1177 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1178 * struct net_device *dev,
1179 * const unsigned char *addr, u16 vid, u16 flags,
1180 * struct netlink_ext_ack *extack);
1181 * Adds an FDB entry to dev for addr.
1182 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1183 * struct net_device *dev,
1184 * const unsigned char *addr, u16 vid)
1185 * Deletes the FDB entry from dev coresponding to addr.
1186 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1187 * struct net_device *dev, struct net_device *filter_dev,
1188 * int *idx)
1189 * Used to add FDB entries to dump requests. Implementers should add
1190 * entries to skb and update idx with the number of entries.
1191 *
1192 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1193 * u16 flags, struct netlink_ext_ack *extack)
1194 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1195 * struct net_device *dev, u32 filter_mask,
1196 * int nlflags)
1197 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1198 * u16 flags);
1199 *
1200 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1201 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1202 * which do not represent real hardware may define this to allow their
1203 * userspace components to manage their virtual carrier state. Devices
1204 * that determine carrier state from physical hardware properties (eg
1205 * network cables) or protocol-dependent mechanisms (eg
1206 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1207 *
1208 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1209 * struct netdev_phys_item_id *ppid);
1210 * Called to get ID of physical port of this device. If driver does
1211 * not implement this, it is assumed that the hw is not able to have
1212 * multiple net devices on single physical port.
1213 *
1214 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1215 * struct netdev_phys_item_id *ppid)
1216 * Called to get the parent ID of the physical port of this device.
1217 *
1218 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1219 * struct udp_tunnel_info *ti);
1220 * Called by UDP tunnel to notify a driver about the UDP port and socket
1221 * address family that a UDP tunnel is listnening to. It is called only
1222 * when a new port starts listening. The operation is protected by the
1223 * RTNL.
1224 *
1225 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1226 * struct udp_tunnel_info *ti);
1227 * Called by UDP tunnel to notify the driver about a UDP port and socket
1228 * address family that the UDP tunnel is not listening to anymore. The
1229 * operation is protected by the RTNL.
1230 *
1231 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1232 * struct net_device *dev)
1233 * Called by upper layer devices to accelerate switching or other
1234 * station functionality into hardware. 'pdev is the lowerdev
1235 * to use for the offload and 'dev' is the net device that will
1236 * back the offload. Returns a pointer to the private structure
1237 * the upper layer will maintain.
1238 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1239 * Called by upper layer device to delete the station created
1240 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1241 * the station and priv is the structure returned by the add
1242 * operation.
1243 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1244 * int queue_index, u32 maxrate);
1245 * Called when a user wants to set a max-rate limitation of specific
1246 * TX queue.
1247 * int (*ndo_get_iflink)(const struct net_device *dev);
1248 * Called to get the iflink value of this device.
1249 * void (*ndo_change_proto_down)(struct net_device *dev,
1250 * bool proto_down);
1251 * This function is used to pass protocol port error state information
1252 * to the switch driver. The switch driver can react to the proto_down
1253 * by doing a phys down on the associated switch port.
1254 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1255 * This function is used to get egress tunnel information for given skb.
1256 * This is useful for retrieving outer tunnel header parameters while
1257 * sampling packet.
1258 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1259 * This function is used to specify the headroom that the skb must
1260 * consider when allocation skb during packet reception. Setting
1261 * appropriate rx headroom value allows avoiding skb head copy on
1262 * forward. Setting a negative value resets the rx headroom to the
1263 * default value.
1264 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1265 * This function is used to set or query state related to XDP on the
1266 * netdevice and manage BPF offload. See definition of
1267 * enum bpf_netdev_command for details.
1268 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1269 * u32 flags);
1270 * This function is used to submit @n XDP packets for transmit on a
1271 * netdevice. Returns number of frames successfully transmitted, frames
1272 * that got dropped are freed/returned via xdp_return_frame().
1273 * Returns negative number, means general error invoking ndo, meaning
1274 * no frames were xmit'ed and core-caller will free all frames.
1275 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1276 * This function is used to wake up the softirq, ksoftirqd or kthread
1277 * responsible for sending and/or receiving packets on a specific
1278 * queue id bound to an AF_XDP socket. The flags field specifies if
1279 * only RX, only Tx, or both should be woken up using the flags
1280 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1281 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1282 * Get devlink port instance associated with a given netdev.
1283 * Called with a reference on the netdevice and devlink locks only,
1284 * rtnl_lock is not held.
1285 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1286 * int cmd);
1287 * Add, change, delete or get information on an IPv4 tunnel.
1288 */
1289struct net_device_ops {
1290 int (*ndo_init)(struct net_device *dev);
1291 void (*ndo_uninit)(struct net_device *dev);
1292 int (*ndo_open)(struct net_device *dev);
1293 int (*ndo_stop)(struct net_device *dev);
1294 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1295 struct net_device *dev);
1296 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1297 struct net_device *dev,
1298 netdev_features_t features);
1299 u16 (*ndo_select_queue)(struct net_device *dev,
1300 struct sk_buff *skb,
1301 struct net_device *sb_dev);
1302 void (*ndo_change_rx_flags)(struct net_device *dev,
1303 int flags);
1304 void (*ndo_set_rx_mode)(struct net_device *dev);
1305 int (*ndo_set_mac_address)(struct net_device *dev,
1306 void *addr);
1307 int (*ndo_validate_addr)(struct net_device *dev);
1308 int (*ndo_do_ioctl)(struct net_device *dev,
1309 struct ifreq *ifr, int cmd);
1310 int (*ndo_set_config)(struct net_device *dev,
1311 struct ifmap *map);
1312 int (*ndo_change_mtu)(struct net_device *dev,
1313 int new_mtu);
1314 int (*ndo_neigh_setup)(struct net_device *dev,
1315 struct neigh_parms *);
1316 void (*ndo_tx_timeout) (struct net_device *dev,
1317 unsigned int txqueue);
1318
1319 void (*ndo_get_stats64)(struct net_device *dev,
1320 struct rtnl_link_stats64 *storage);
1321 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1322 int (*ndo_get_offload_stats)(int attr_id,
1323 const struct net_device *dev,
1324 void *attr_data);
1325 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1326
1327 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1328 __be16 proto, u16 vid);
1329 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1330 __be16 proto, u16 vid);
1331#ifdef CONFIG_NET_POLL_CONTROLLER
1332 void (*ndo_poll_controller)(struct net_device *dev);
1333 int (*ndo_netpoll_setup)(struct net_device *dev,
1334 struct netpoll_info *info);
1335 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1336#endif
1337 int (*ndo_set_vf_mac)(struct net_device *dev,
1338 int queue, u8 *mac);
1339 int (*ndo_set_vf_vlan)(struct net_device *dev,
1340 int queue, u16 vlan,
1341 u8 qos, __be16 proto);
1342 int (*ndo_set_vf_rate)(struct net_device *dev,
1343 int vf, int min_tx_rate,
1344 int max_tx_rate);
1345 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1346 int vf, bool setting);
1347 int (*ndo_set_vf_trust)(struct net_device *dev,
1348 int vf, bool setting);
1349 int (*ndo_get_vf_config)(struct net_device *dev,
1350 int vf,
1351 struct ifla_vf_info *ivf);
1352 int (*ndo_set_vf_link_state)(struct net_device *dev,
1353 int vf, int link_state);
1354 int (*ndo_get_vf_stats)(struct net_device *dev,
1355 int vf,
1356 struct ifla_vf_stats
1357 *vf_stats);
1358 int (*ndo_set_vf_port)(struct net_device *dev,
1359 int vf,
1360 struct nlattr *port[]);
1361 int (*ndo_get_vf_port)(struct net_device *dev,
1362 int vf, struct sk_buff *skb);
1363 int (*ndo_get_vf_guid)(struct net_device *dev,
1364 int vf,
1365 struct ifla_vf_guid *node_guid,
1366 struct ifla_vf_guid *port_guid);
1367 int (*ndo_set_vf_guid)(struct net_device *dev,
1368 int vf, u64 guid,
1369 int guid_type);
1370 int (*ndo_set_vf_rss_query_en)(
1371 struct net_device *dev,
1372 int vf, bool setting);
1373 int (*ndo_setup_tc)(struct net_device *dev,
1374 enum tc_setup_type type,
1375 void *type_data);
1376#if IS_ENABLED(CONFIG_FCOE)
1377 int (*ndo_fcoe_enable)(struct net_device *dev);
1378 int (*ndo_fcoe_disable)(struct net_device *dev);
1379 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1380 u16 xid,
1381 struct scatterlist *sgl,
1382 unsigned int sgc);
1383 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1384 u16 xid);
1385 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1386 u16 xid,
1387 struct scatterlist *sgl,
1388 unsigned int sgc);
1389 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1390 struct netdev_fcoe_hbainfo *hbainfo);
1391#endif
1392
1393#if IS_ENABLED(CONFIG_LIBFCOE)
1394#define NETDEV_FCOE_WWNN 0
1395#define NETDEV_FCOE_WWPN 1
1396 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1397 u64 *wwn, int type);
1398#endif
1399
1400#ifdef CONFIG_RFS_ACCEL
1401 int (*ndo_rx_flow_steer)(struct net_device *dev,
1402 const struct sk_buff *skb,
1403 u16 rxq_index,
1404 u32 flow_id);
1405#endif
1406 int (*ndo_add_slave)(struct net_device *dev,
1407 struct net_device *slave_dev,
1408 struct netlink_ext_ack *extack);
1409 int (*ndo_del_slave)(struct net_device *dev,
1410 struct net_device *slave_dev);
1411 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1412 struct sk_buff *skb,
1413 bool all_slaves);
1414 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1415 netdev_features_t features);
1416 int (*ndo_set_features)(struct net_device *dev,
1417 netdev_features_t features);
1418 int (*ndo_neigh_construct)(struct net_device *dev,
1419 struct neighbour *n);
1420 void (*ndo_neigh_destroy)(struct net_device *dev,
1421 struct neighbour *n);
1422
1423 int (*ndo_fdb_add)(struct ndmsg *ndm,
1424 struct nlattr *tb[],
1425 struct net_device *dev,
1426 const unsigned char *addr,
1427 u16 vid,
1428 u16 flags,
1429 struct netlink_ext_ack *extack);
1430 int (*ndo_fdb_del)(struct ndmsg *ndm,
1431 struct nlattr *tb[],
1432 struct net_device *dev,
1433 const unsigned char *addr,
1434 u16 vid);
1435 int (*ndo_fdb_dump)(struct sk_buff *skb,
1436 struct netlink_callback *cb,
1437 struct net_device *dev,
1438 struct net_device *filter_dev,
1439 int *idx);
1440 int (*ndo_fdb_get)(struct sk_buff *skb,
1441 struct nlattr *tb[],
1442 struct net_device *dev,
1443 const unsigned char *addr,
1444 u16 vid, u32 portid, u32 seq,
1445 struct netlink_ext_ack *extack);
1446 int (*ndo_bridge_setlink)(struct net_device *dev,
1447 struct nlmsghdr *nlh,
1448 u16 flags,
1449 struct netlink_ext_ack *extack);
1450 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1451 u32 pid, u32 seq,
1452 struct net_device *dev,
1453 u32 filter_mask,
1454 int nlflags);
1455 int (*ndo_bridge_dellink)(struct net_device *dev,
1456 struct nlmsghdr *nlh,
1457 u16 flags);
1458 int (*ndo_change_carrier)(struct net_device *dev,
1459 bool new_carrier);
1460 int (*ndo_get_phys_port_id)(struct net_device *dev,
1461 struct netdev_phys_item_id *ppid);
1462 int (*ndo_get_port_parent_id)(struct net_device *dev,
1463 struct netdev_phys_item_id *ppid);
1464 int (*ndo_get_phys_port_name)(struct net_device *dev,
1465 char *name, size_t len);
1466 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1467 struct udp_tunnel_info *ti);
1468 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1469 struct udp_tunnel_info *ti);
1470 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1471 struct net_device *dev);
1472 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1473 void *priv);
1474
1475 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1476 int queue_index,
1477 u32 maxrate);
1478 int (*ndo_get_iflink)(const struct net_device *dev);
1479 int (*ndo_change_proto_down)(struct net_device *dev,
1480 bool proto_down);
1481 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1482 struct sk_buff *skb);
1483 void (*ndo_set_rx_headroom)(struct net_device *dev,
1484 int needed_headroom);
1485 int (*ndo_bpf)(struct net_device *dev,
1486 struct netdev_bpf *bpf);
1487 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1488 struct xdp_frame **xdp,
1489 u32 flags);
1490 int (*ndo_xsk_wakeup)(struct net_device *dev,
1491 u32 queue_id, u32 flags);
1492 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1493 int (*ndo_tunnel_ctl)(struct net_device *dev,
1494 struct ip_tunnel_parm *p, int cmd);
1495};
1496
1497/**
1498 * enum net_device_priv_flags - &struct net_device priv_flags
1499 *
1500 * These are the &struct net_device, they are only set internally
1501 * by drivers and used in the kernel. These flags are invisible to
1502 * userspace; this means that the order of these flags can change
1503 * during any kernel release.
1504 *
1505 * You should have a pretty good reason to be extending these flags.
1506 *
1507 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1508 * @IFF_EBRIDGE: Ethernet bridging device
1509 * @IFF_BONDING: bonding master or slave
1510 * @IFF_ISATAP: ISATAP interface (RFC4214)
1511 * @IFF_WAN_HDLC: WAN HDLC device
1512 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1513 * release skb->dst
1514 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1515 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1516 * @IFF_MACVLAN_PORT: device used as macvlan port
1517 * @IFF_BRIDGE_PORT: device used as bridge port
1518 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1519 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1520 * @IFF_UNICAST_FLT: Supports unicast filtering
1521 * @IFF_TEAM_PORT: device used as team port
1522 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1523 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1524 * change when it's running
1525 * @IFF_MACVLAN: Macvlan device
1526 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1527 * underlying stacked devices
1528 * @IFF_L3MDEV_MASTER: device is an L3 master device
1529 * @IFF_NO_QUEUE: device can run without qdisc attached
1530 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1531 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1532 * @IFF_TEAM: device is a team device
1533 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1534 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1535 * entity (i.e. the master device for bridged veth)
1536 * @IFF_MACSEC: device is a MACsec device
1537 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1538 * @IFF_FAILOVER: device is a failover master device
1539 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1540 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1541 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1542 */
1543enum netdev_priv_flags {
1544 IFF_802_1Q_VLAN = 1<<0,
1545 IFF_EBRIDGE = 1<<1,
1546 IFF_BONDING = 1<<2,
1547 IFF_ISATAP = 1<<3,
1548 IFF_WAN_HDLC = 1<<4,
1549 IFF_XMIT_DST_RELEASE = 1<<5,
1550 IFF_DONT_BRIDGE = 1<<6,
1551 IFF_DISABLE_NETPOLL = 1<<7,
1552 IFF_MACVLAN_PORT = 1<<8,
1553 IFF_BRIDGE_PORT = 1<<9,
1554 IFF_OVS_DATAPATH = 1<<10,
1555 IFF_TX_SKB_SHARING = 1<<11,
1556 IFF_UNICAST_FLT = 1<<12,
1557 IFF_TEAM_PORT = 1<<13,
1558 IFF_SUPP_NOFCS = 1<<14,
1559 IFF_LIVE_ADDR_CHANGE = 1<<15,
1560 IFF_MACVLAN = 1<<16,
1561 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1562 IFF_L3MDEV_MASTER = 1<<18,
1563 IFF_NO_QUEUE = 1<<19,
1564 IFF_OPENVSWITCH = 1<<20,
1565 IFF_L3MDEV_SLAVE = 1<<21,
1566 IFF_TEAM = 1<<22,
1567 IFF_RXFH_CONFIGURED = 1<<23,
1568 IFF_PHONY_HEADROOM = 1<<24,
1569 IFF_MACSEC = 1<<25,
1570 IFF_NO_RX_HANDLER = 1<<26,
1571 IFF_FAILOVER = 1<<27,
1572 IFF_FAILOVER_SLAVE = 1<<28,
1573 IFF_L3MDEV_RX_HANDLER = 1<<29,
1574 IFF_LIVE_RENAME_OK = 1<<30,
1575};
1576
1577#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1578#define IFF_EBRIDGE IFF_EBRIDGE
1579#define IFF_BONDING IFF_BONDING
1580#define IFF_ISATAP IFF_ISATAP
1581#define IFF_WAN_HDLC IFF_WAN_HDLC
1582#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1583#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1584#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1585#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1586#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1587#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1588#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1589#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1590#define IFF_TEAM_PORT IFF_TEAM_PORT
1591#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1592#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1593#define IFF_MACVLAN IFF_MACVLAN
1594#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1595#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1596#define IFF_NO_QUEUE IFF_NO_QUEUE
1597#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1598#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1599#define IFF_TEAM IFF_TEAM
1600#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1601#define IFF_MACSEC IFF_MACSEC
1602#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1603#define IFF_FAILOVER IFF_FAILOVER
1604#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1605#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1606#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1607
1608/**
1609 * struct net_device - The DEVICE structure.
1610 *
1611 * Actually, this whole structure is a big mistake. It mixes I/O
1612 * data with strictly "high-level" data, and it has to know about
1613 * almost every data structure used in the INET module.
1614 *
1615 * @name: This is the first field of the "visible" part of this structure
1616 * (i.e. as seen by users in the "Space.c" file). It is the name
1617 * of the interface.
1618 *
1619 * @name_node: Name hashlist node
1620 * @ifalias: SNMP alias
1621 * @mem_end: Shared memory end
1622 * @mem_start: Shared memory start
1623 * @base_addr: Device I/O address
1624 * @irq: Device IRQ number
1625 *
1626 * @state: Generic network queuing layer state, see netdev_state_t
1627 * @dev_list: The global list of network devices
1628 * @napi_list: List entry used for polling NAPI devices
1629 * @unreg_list: List entry when we are unregistering the
1630 * device; see the function unregister_netdev
1631 * @close_list: List entry used when we are closing the device
1632 * @ptype_all: Device-specific packet handlers for all protocols
1633 * @ptype_specific: Device-specific, protocol-specific packet handlers
1634 *
1635 * @adj_list: Directly linked devices, like slaves for bonding
1636 * @features: Currently active device features
1637 * @hw_features: User-changeable features
1638 *
1639 * @wanted_features: User-requested features
1640 * @vlan_features: Mask of features inheritable by VLAN devices
1641 *
1642 * @hw_enc_features: Mask of features inherited by encapsulating devices
1643 * This field indicates what encapsulation
1644 * offloads the hardware is capable of doing,
1645 * and drivers will need to set them appropriately.
1646 *
1647 * @mpls_features: Mask of features inheritable by MPLS
1648 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1649 *
1650 * @ifindex: interface index
1651 * @group: The group the device belongs to
1652 *
1653 * @stats: Statistics struct, which was left as a legacy, use
1654 * rtnl_link_stats64 instead
1655 *
1656 * @rx_dropped: Dropped packets by core network,
1657 * do not use this in drivers
1658 * @tx_dropped: Dropped packets by core network,
1659 * do not use this in drivers
1660 * @rx_nohandler: nohandler dropped packets by core network on
1661 * inactive devices, do not use this in drivers
1662 * @carrier_up_count: Number of times the carrier has been up
1663 * @carrier_down_count: Number of times the carrier has been down
1664 *
1665 * @wireless_handlers: List of functions to handle Wireless Extensions,
1666 * instead of ioctl,
1667 * see <net/iw_handler.h> for details.
1668 * @wireless_data: Instance data managed by the core of wireless extensions
1669 *
1670 * @netdev_ops: Includes several pointers to callbacks,
1671 * if one wants to override the ndo_*() functions
1672 * @ethtool_ops: Management operations
1673 * @l3mdev_ops: Layer 3 master device operations
1674 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1675 * discovery handling. Necessary for e.g. 6LoWPAN.
1676 * @xfrmdev_ops: Transformation offload operations
1677 * @tlsdev_ops: Transport Layer Security offload operations
1678 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1679 * of Layer 2 headers.
1680 *
1681 * @flags: Interface flags (a la BSD)
1682 * @priv_flags: Like 'flags' but invisible to userspace,
1683 * see if.h for the definitions
1684 * @gflags: Global flags ( kept as legacy )
1685 * @padded: How much padding added by alloc_netdev()
1686 * @operstate: RFC2863 operstate
1687 * @link_mode: Mapping policy to operstate
1688 * @if_port: Selectable AUI, TP, ...
1689 * @dma: DMA channel
1690 * @mtu: Interface MTU value
1691 * @min_mtu: Interface Minimum MTU value
1692 * @max_mtu: Interface Maximum MTU value
1693 * @type: Interface hardware type
1694 * @hard_header_len: Maximum hardware header length.
1695 * @min_header_len: Minimum hardware header length
1696 *
1697 * @needed_headroom: Extra headroom the hardware may need, but not in all
1698 * cases can this be guaranteed
1699 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1700 * cases can this be guaranteed. Some cases also use
1701 * LL_MAX_HEADER instead to allocate the skb
1702 *
1703 * interface address info:
1704 *
1705 * @perm_addr: Permanent hw address
1706 * @addr_assign_type: Hw address assignment type
1707 * @addr_len: Hardware address length
1708 * @upper_level: Maximum depth level of upper devices.
1709 * @lower_level: Maximum depth level of lower devices.
1710 * @neigh_priv_len: Used in neigh_alloc()
1711 * @dev_id: Used to differentiate devices that share
1712 * the same link layer address
1713 * @dev_port: Used to differentiate devices that share
1714 * the same function
1715 * @addr_list_lock: XXX: need comments on this one
1716 * @name_assign_type: network interface name assignment type
1717 * @uc_promisc: Counter that indicates promiscuous mode
1718 * has been enabled due to the need to listen to
1719 * additional unicast addresses in a device that
1720 * does not implement ndo_set_rx_mode()
1721 * @uc: unicast mac addresses
1722 * @mc: multicast mac addresses
1723 * @dev_addrs: list of device hw addresses
1724 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1725 * @promiscuity: Number of times the NIC is told to work in
1726 * promiscuous mode; if it becomes 0 the NIC will
1727 * exit promiscuous mode
1728 * @allmulti: Counter, enables or disables allmulticast mode
1729 *
1730 * @vlan_info: VLAN info
1731 * @dsa_ptr: dsa specific data
1732 * @tipc_ptr: TIPC specific data
1733 * @atalk_ptr: AppleTalk link
1734 * @ip_ptr: IPv4 specific data
1735 * @dn_ptr: DECnet specific data
1736 * @ip6_ptr: IPv6 specific data
1737 * @ax25_ptr: AX.25 specific data
1738 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1739 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1740 * device struct
1741 * @mpls_ptr: mpls_dev struct pointer
1742 *
1743 * @dev_addr: Hw address (before bcast,
1744 * because most packets are unicast)
1745 *
1746 * @_rx: Array of RX queues
1747 * @num_rx_queues: Number of RX queues
1748 * allocated at register_netdev() time
1749 * @real_num_rx_queues: Number of RX queues currently active in device
1750 * @xdp_prog: XDP sockets filter program pointer
1751 * @gro_flush_timeout: timeout for GRO layer in NAPI
1752 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1753 * allow to avoid NIC hard IRQ, on busy queues.
1754 *
1755 * @rx_handler: handler for received packets
1756 * @rx_handler_data: XXX: need comments on this one
1757 * @miniq_ingress: ingress/clsact qdisc specific data for
1758 * ingress processing
1759 * @ingress_queue: XXX: need comments on this one
1760 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1761 * @broadcast: hw bcast address
1762 *
1763 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1764 * indexed by RX queue number. Assigned by driver.
1765 * This must only be set if the ndo_rx_flow_steer
1766 * operation is defined
1767 * @index_hlist: Device index hash chain
1768 *
1769 * @_tx: Array of TX queues
1770 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1771 * @real_num_tx_queues: Number of TX queues currently active in device
1772 * @qdisc: Root qdisc from userspace point of view
1773 * @tx_queue_len: Max frames per queue allowed
1774 * @tx_global_lock: XXX: need comments on this one
1775 * @xdp_bulkq: XDP device bulk queue
1776 * @xps_cpus_map: all CPUs map for XPS device
1777 * @xps_rxqs_map: all RXQs map for XPS device
1778 *
1779 * @xps_maps: XXX: need comments on this one
1780 * @miniq_egress: clsact qdisc specific data for
1781 * egress processing
1782 * @qdisc_hash: qdisc hash table
1783 * @watchdog_timeo: Represents the timeout that is used by
1784 * the watchdog (see dev_watchdog())
1785 * @watchdog_timer: List of timers
1786 *
1787 * @proto_down_reason: reason a netdev interface is held down
1788 * @pcpu_refcnt: Number of references to this device
1789 * @todo_list: Delayed register/unregister
1790 * @link_watch_list: XXX: need comments on this one
1791 *
1792 * @reg_state: Register/unregister state machine
1793 * @dismantle: Device is going to be freed
1794 * @rtnl_link_state: This enum represents the phases of creating
1795 * a new link
1796 *
1797 * @needs_free_netdev: Should unregister perform free_netdev?
1798 * @priv_destructor: Called from unregister
1799 * @npinfo: XXX: need comments on this one
1800 * @nd_net: Network namespace this network device is inside
1801 *
1802 * @ml_priv: Mid-layer private
1803 * @lstats: Loopback statistics
1804 * @tstats: Tunnel statistics
1805 * @dstats: Dummy statistics
1806 * @vstats: Virtual ethernet statistics
1807 *
1808 * @garp_port: GARP
1809 * @mrp_port: MRP
1810 *
1811 * @dev: Class/net/name entry
1812 * @sysfs_groups: Space for optional device, statistics and wireless
1813 * sysfs groups
1814 *
1815 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1816 * @rtnl_link_ops: Rtnl_link_ops
1817 *
1818 * @gso_max_size: Maximum size of generic segmentation offload
1819 * @gso_max_segs: Maximum number of segments that can be passed to the
1820 * NIC for GSO
1821 *
1822 * @dcbnl_ops: Data Center Bridging netlink ops
1823 * @num_tc: Number of traffic classes in the net device
1824 * @tc_to_txq: XXX: need comments on this one
1825 * @prio_tc_map: XXX: need comments on this one
1826 *
1827 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1828 *
1829 * @priomap: XXX: need comments on this one
1830 * @phydev: Physical device may attach itself
1831 * for hardware timestamping
1832 * @sfp_bus: attached &struct sfp_bus structure.
1833 *
1834 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1835 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1836 *
1837 * @proto_down: protocol port state information can be sent to the
1838 * switch driver and used to set the phys state of the
1839 * switch port.
1840 *
1841 * @wol_enabled: Wake-on-LAN is enabled
1842 *
1843 * @net_notifier_list: List of per-net netdev notifier block
1844 * that follow this device when it is moved
1845 * to another network namespace.
1846 *
1847 * @macsec_ops: MACsec offloading ops
1848 *
1849 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1850 * offload capabilities of the device
1851 * @udp_tunnel_nic: UDP tunnel offload state
1852 * @xdp_state: stores info on attached XDP BPF programs
1853 *
1854 * @nested_level: Used as as a parameter of spin_lock_nested() of
1855 * dev->addr_list_lock.
1856 * @unlink_list: As netif_addr_lock() can be called recursively,
1857 * keep a list of interfaces to be deleted.
1858 *
1859 * FIXME: cleanup struct net_device such that network protocol info
1860 * moves out.
1861 */
1862
1863struct net_device {
1864 char name[IFNAMSIZ];
1865 struct netdev_name_node *name_node;
1866 struct dev_ifalias __rcu *ifalias;
1867 /*
1868 * I/O specific fields
1869 * FIXME: Merge these and struct ifmap into one
1870 */
1871 unsigned long mem_end;
1872 unsigned long mem_start;
1873 unsigned long base_addr;
1874 int irq;
1875
1876 /*
1877 * Some hardware also needs these fields (state,dev_list,
1878 * napi_list,unreg_list,close_list) but they are not
1879 * part of the usual set specified in Space.c.
1880 */
1881
1882 unsigned long state;
1883
1884 struct list_head dev_list;
1885 struct list_head napi_list;
1886 struct list_head unreg_list;
1887 struct list_head close_list;
1888 struct list_head ptype_all;
1889 struct list_head ptype_specific;
1890
1891 struct {
1892 struct list_head upper;
1893 struct list_head lower;
1894 } adj_list;
1895
1896 netdev_features_t features;
1897 netdev_features_t hw_features;
1898 netdev_features_t wanted_features;
1899 netdev_features_t vlan_features;
1900 netdev_features_t hw_enc_features;
1901 netdev_features_t mpls_features;
1902 netdev_features_t gso_partial_features;
1903
1904 int ifindex;
1905 int group;
1906
1907 struct net_device_stats stats;
1908
1909 atomic_long_t rx_dropped;
1910 atomic_long_t tx_dropped;
1911 atomic_long_t rx_nohandler;
1912
1913 /* Stats to monitor link on/off, flapping */
1914 atomic_t carrier_up_count;
1915 atomic_t carrier_down_count;
1916
1917#ifdef CONFIG_WIRELESS_EXT
1918 const struct iw_handler_def *wireless_handlers;
1919 struct iw_public_data *wireless_data;
1920#endif
1921 const struct net_device_ops *netdev_ops;
1922 const struct ethtool_ops *ethtool_ops;
1923#ifdef CONFIG_NET_L3_MASTER_DEV
1924 const struct l3mdev_ops *l3mdev_ops;
1925#endif
1926#if IS_ENABLED(CONFIG_IPV6)
1927 const struct ndisc_ops *ndisc_ops;
1928#endif
1929
1930#ifdef CONFIG_XFRM_OFFLOAD
1931 const struct xfrmdev_ops *xfrmdev_ops;
1932#endif
1933
1934#if IS_ENABLED(CONFIG_TLS_DEVICE)
1935 const struct tlsdev_ops *tlsdev_ops;
1936#endif
1937
1938 const struct header_ops *header_ops;
1939
1940 unsigned int flags;
1941 unsigned int priv_flags;
1942
1943 unsigned short gflags;
1944 unsigned short padded;
1945
1946 unsigned char operstate;
1947 unsigned char link_mode;
1948
1949 unsigned char if_port;
1950 unsigned char dma;
1951
1952 /* Note : dev->mtu is often read without holding a lock.
1953 * Writers usually hold RTNL.
1954 * It is recommended to use READ_ONCE() to annotate the reads,
1955 * and to use WRITE_ONCE() to annotate the writes.
1956 */
1957 unsigned int mtu;
1958 unsigned int min_mtu;
1959 unsigned int max_mtu;
1960 unsigned short type;
1961 unsigned short hard_header_len;
1962 unsigned char min_header_len;
1963 unsigned char name_assign_type;
1964
1965 unsigned short needed_headroom;
1966 unsigned short needed_tailroom;
1967
1968 /* Interface address info. */
1969 unsigned char perm_addr[MAX_ADDR_LEN];
1970 unsigned char addr_assign_type;
1971 unsigned char addr_len;
1972 unsigned char upper_level;
1973 unsigned char lower_level;
1974
1975 unsigned short neigh_priv_len;
1976 unsigned short dev_id;
1977 unsigned short dev_port;
1978 spinlock_t addr_list_lock;
1979
1980 struct netdev_hw_addr_list uc;
1981 struct netdev_hw_addr_list mc;
1982 struct netdev_hw_addr_list dev_addrs;
1983
1984#ifdef CONFIG_SYSFS
1985 struct kset *queues_kset;
1986#endif
1987#ifdef CONFIG_LOCKDEP
1988 struct list_head unlink_list;
1989#endif
1990 unsigned int promiscuity;
1991 unsigned int allmulti;
1992 bool uc_promisc;
1993#ifdef CONFIG_LOCKDEP
1994 unsigned char nested_level;
1995#endif
1996
1997
1998 /* Protocol-specific pointers */
1999
2000#if IS_ENABLED(CONFIG_VLAN_8021Q)
2001 struct vlan_info __rcu *vlan_info;
2002#endif
2003#if IS_ENABLED(CONFIG_NET_DSA)
2004 struct dsa_port *dsa_ptr;
2005#endif
2006#if IS_ENABLED(CONFIG_TIPC)
2007 struct tipc_bearer __rcu *tipc_ptr;
2008#endif
2009#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2010 void *atalk_ptr;
2011#endif
2012 struct in_device __rcu *ip_ptr;
2013#if IS_ENABLED(CONFIG_DECNET)
2014 struct dn_dev __rcu *dn_ptr;
2015#endif
2016 struct inet6_dev __rcu *ip6_ptr;
2017#if IS_ENABLED(CONFIG_AX25)
2018 void *ax25_ptr;
2019#endif
2020 struct wireless_dev *ieee80211_ptr;
2021 struct wpan_dev *ieee802154_ptr;
2022#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2023 struct mpls_dev __rcu *mpls_ptr;
2024#endif
2025
2026/*
2027 * Cache lines mostly used on receive path (including eth_type_trans())
2028 */
2029 /* Interface address info used in eth_type_trans() */
2030 unsigned char *dev_addr;
2031
2032 struct netdev_rx_queue *_rx;
2033 unsigned int num_rx_queues;
2034 unsigned int real_num_rx_queues;
2035
2036 struct bpf_prog __rcu *xdp_prog;
2037 unsigned long gro_flush_timeout;
2038 int napi_defer_hard_irqs;
2039 rx_handler_func_t __rcu *rx_handler;
2040 void __rcu *rx_handler_data;
2041
2042#ifdef CONFIG_NET_CLS_ACT
2043 struct mini_Qdisc __rcu *miniq_ingress;
2044#endif
2045 struct netdev_queue __rcu *ingress_queue;
2046#ifdef CONFIG_NETFILTER_INGRESS
2047 struct nf_hook_entries __rcu *nf_hooks_ingress;
2048#endif
2049
2050 unsigned char broadcast[MAX_ADDR_LEN];
2051#ifdef CONFIG_RFS_ACCEL
2052 struct cpu_rmap *rx_cpu_rmap;
2053#endif
2054 struct hlist_node index_hlist;
2055
2056/*
2057 * Cache lines mostly used on transmit path
2058 */
2059 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2060 unsigned int num_tx_queues;
2061 unsigned int real_num_tx_queues;
2062 struct Qdisc *qdisc;
2063 unsigned int tx_queue_len;
2064 spinlock_t tx_global_lock;
2065
2066 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2067
2068#ifdef CONFIG_XPS
2069 struct xps_dev_maps __rcu *xps_cpus_map;
2070 struct xps_dev_maps __rcu *xps_rxqs_map;
2071#endif
2072#ifdef CONFIG_NET_CLS_ACT
2073 struct mini_Qdisc __rcu *miniq_egress;
2074#endif
2075
2076#ifdef CONFIG_NET_SCHED
2077 DECLARE_HASHTABLE (qdisc_hash, 4);
2078#endif
2079 /* These may be needed for future network-power-down code. */
2080 struct timer_list watchdog_timer;
2081 int watchdog_timeo;
2082
2083 u32 proto_down_reason;
2084
2085 struct list_head todo_list;
2086 int __percpu *pcpu_refcnt;
2087
2088 struct list_head link_watch_list;
2089
2090 enum { NETREG_UNINITIALIZED=0,
2091 NETREG_REGISTERED, /* completed register_netdevice */
2092 NETREG_UNREGISTERING, /* called unregister_netdevice */
2093 NETREG_UNREGISTERED, /* completed unregister todo */
2094 NETREG_RELEASED, /* called free_netdev */
2095 NETREG_DUMMY, /* dummy device for NAPI poll */
2096 } reg_state:8;
2097
2098 bool dismantle;
2099
2100 enum {
2101 RTNL_LINK_INITIALIZED,
2102 RTNL_LINK_INITIALIZING,
2103 } rtnl_link_state:16;
2104
2105 bool needs_free_netdev;
2106 void (*priv_destructor)(struct net_device *dev);
2107
2108#ifdef CONFIG_NETPOLL
2109 struct netpoll_info __rcu *npinfo;
2110#endif
2111
2112 possible_net_t nd_net;
2113
2114 /* mid-layer private */
2115 union {
2116 void *ml_priv;
2117 struct pcpu_lstats __percpu *lstats;
2118 struct pcpu_sw_netstats __percpu *tstats;
2119 struct pcpu_dstats __percpu *dstats;
2120 };
2121
2122#if IS_ENABLED(CONFIG_GARP)
2123 struct garp_port __rcu *garp_port;
2124#endif
2125#if IS_ENABLED(CONFIG_MRP)
2126 struct mrp_port __rcu *mrp_port;
2127#endif
2128
2129 struct device dev;
2130 const struct attribute_group *sysfs_groups[4];
2131 const struct attribute_group *sysfs_rx_queue_group;
2132
2133 const struct rtnl_link_ops *rtnl_link_ops;
2134
2135 /* for setting kernel sock attribute on TCP connection setup */
2136#define GSO_MAX_SIZE 65536
2137 unsigned int gso_max_size;
2138#define GSO_MAX_SEGS 65535
2139 u16 gso_max_segs;
2140
2141#ifdef CONFIG_DCB
2142 const struct dcbnl_rtnl_ops *dcbnl_ops;
2143#endif
2144 s16 num_tc;
2145 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2146 u8 prio_tc_map[TC_BITMASK + 1];
2147
2148#if IS_ENABLED(CONFIG_FCOE)
2149 unsigned int fcoe_ddp_xid;
2150#endif
2151#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2152 struct netprio_map __rcu *priomap;
2153#endif
2154 struct phy_device *phydev;
2155 struct sfp_bus *sfp_bus;
2156 struct lock_class_key *qdisc_tx_busylock;
2157 struct lock_class_key *qdisc_running_key;
2158 bool proto_down;
2159 unsigned wol_enabled:1;
2160
2161 struct list_head net_notifier_list;
2162
2163#if IS_ENABLED(CONFIG_MACSEC)
2164 /* MACsec management functions */
2165 const struct macsec_ops *macsec_ops;
2166#endif
2167 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2168 struct udp_tunnel_nic *udp_tunnel_nic;
2169
2170 /* protected by rtnl_lock */
2171 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2172};
2173#define to_net_dev(d) container_of(d, struct net_device, dev)
2174
2175static inline bool netif_elide_gro(const struct net_device *dev)
2176{
2177 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2178 return true;
2179 return false;
2180}
2181
2182#define NETDEV_ALIGN 32
2183
2184static inline
2185int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2186{
2187 return dev->prio_tc_map[prio & TC_BITMASK];
2188}
2189
2190static inline
2191int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2192{
2193 if (tc >= dev->num_tc)
2194 return -EINVAL;
2195
2196 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2197 return 0;
2198}
2199
2200int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2201void netdev_reset_tc(struct net_device *dev);
2202int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2203int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2204
2205static inline
2206int netdev_get_num_tc(struct net_device *dev)
2207{
2208 return dev->num_tc;
2209}
2210
2211void netdev_unbind_sb_channel(struct net_device *dev,
2212 struct net_device *sb_dev);
2213int netdev_bind_sb_channel_queue(struct net_device *dev,
2214 struct net_device *sb_dev,
2215 u8 tc, u16 count, u16 offset);
2216int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2217static inline int netdev_get_sb_channel(struct net_device *dev)
2218{
2219 return max_t(int, -dev->num_tc, 0);
2220}
2221
2222static inline
2223struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2224 unsigned int index)
2225{
2226 return &dev->_tx[index];
2227}
2228
2229static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2230 const struct sk_buff *skb)
2231{
2232 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2233}
2234
2235static inline void netdev_for_each_tx_queue(struct net_device *dev,
2236 void (*f)(struct net_device *,
2237 struct netdev_queue *,
2238 void *),
2239 void *arg)
2240{
2241 unsigned int i;
2242
2243 for (i = 0; i < dev->num_tx_queues; i++)
2244 f(dev, &dev->_tx[i], arg);
2245}
2246
2247#define netdev_lockdep_set_classes(dev) \
2248{ \
2249 static struct lock_class_key qdisc_tx_busylock_key; \
2250 static struct lock_class_key qdisc_running_key; \
2251 static struct lock_class_key qdisc_xmit_lock_key; \
2252 static struct lock_class_key dev_addr_list_lock_key; \
2253 unsigned int i; \
2254 \
2255 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2256 (dev)->qdisc_running_key = &qdisc_running_key; \
2257 lockdep_set_class(&(dev)->addr_list_lock, \
2258 &dev_addr_list_lock_key); \
2259 for (i = 0; i < (dev)->num_tx_queues; i++) \
2260 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2261 &qdisc_xmit_lock_key); \
2262}
2263
2264u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2265 struct net_device *sb_dev);
2266struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2267 struct sk_buff *skb,
2268 struct net_device *sb_dev);
2269
2270/* returns the headroom that the master device needs to take in account
2271 * when forwarding to this dev
2272 */
2273static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2274{
2275 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2276}
2277
2278static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2279{
2280 if (dev->netdev_ops->ndo_set_rx_headroom)
2281 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2282}
2283
2284/* set the device rx headroom to the dev's default */
2285static inline void netdev_reset_rx_headroom(struct net_device *dev)
2286{
2287 netdev_set_rx_headroom(dev, -1);
2288}
2289
2290/*
2291 * Net namespace inlines
2292 */
2293static inline
2294struct net *dev_net(const struct net_device *dev)
2295{
2296 return read_pnet(&dev->nd_net);
2297}
2298
2299static inline
2300void dev_net_set(struct net_device *dev, struct net *net)
2301{
2302 write_pnet(&dev->nd_net, net);
2303}
2304
2305/**
2306 * netdev_priv - access network device private data
2307 * @dev: network device
2308 *
2309 * Get network device private data
2310 */
2311static inline void *netdev_priv(const struct net_device *dev)
2312{
2313 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2314}
2315
2316/* Set the sysfs physical device reference for the network logical device
2317 * if set prior to registration will cause a symlink during initialization.
2318 */
2319#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2320
2321/* Set the sysfs device type for the network logical device to allow
2322 * fine-grained identification of different network device types. For
2323 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2324 */
2325#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2326
2327/* Default NAPI poll() weight
2328 * Device drivers are strongly advised to not use bigger value
2329 */
2330#define NAPI_POLL_WEIGHT 64
2331
2332/**
2333 * netif_napi_add - initialize a NAPI context
2334 * @dev: network device
2335 * @napi: NAPI context
2336 * @poll: polling function
2337 * @weight: default weight
2338 *
2339 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2340 * *any* of the other NAPI-related functions.
2341 */
2342void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2343 int (*poll)(struct napi_struct *, int), int weight);
2344
2345/**
2346 * netif_tx_napi_add - initialize a NAPI context
2347 * @dev: network device
2348 * @napi: NAPI context
2349 * @poll: polling function
2350 * @weight: default weight
2351 *
2352 * This variant of netif_napi_add() should be used from drivers using NAPI
2353 * to exclusively poll a TX queue.
2354 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2355 */
2356static inline void netif_tx_napi_add(struct net_device *dev,
2357 struct napi_struct *napi,
2358 int (*poll)(struct napi_struct *, int),
2359 int weight)
2360{
2361 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2362 netif_napi_add(dev, napi, poll, weight);
2363}
2364
2365/**
2366 * netif_napi_del - remove a NAPI context
2367 * @napi: NAPI context
2368 *
2369 * netif_napi_del() removes a NAPI context from the network device NAPI list
2370 */
2371void netif_napi_del(struct napi_struct *napi);
2372
2373struct napi_gro_cb {
2374 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2375 void *frag0;
2376
2377 /* Length of frag0. */
2378 unsigned int frag0_len;
2379
2380 /* This indicates where we are processing relative to skb->data. */
2381 int data_offset;
2382
2383 /* This is non-zero if the packet cannot be merged with the new skb. */
2384 u16 flush;
2385
2386 /* Save the IP ID here and check when we get to the transport layer */
2387 u16 flush_id;
2388
2389 /* Number of segments aggregated. */
2390 u16 count;
2391
2392 /* Start offset for remote checksum offload */
2393 u16 gro_remcsum_start;
2394
2395 /* jiffies when first packet was created/queued */
2396 unsigned long age;
2397
2398 /* Used in ipv6_gro_receive() and foo-over-udp */
2399 u16 proto;
2400
2401 /* This is non-zero if the packet may be of the same flow. */
2402 u8 same_flow:1;
2403
2404 /* Used in tunnel GRO receive */
2405 u8 encap_mark:1;
2406
2407 /* GRO checksum is valid */
2408 u8 csum_valid:1;
2409
2410 /* Number of checksums via CHECKSUM_UNNECESSARY */
2411 u8 csum_cnt:3;
2412
2413 /* Free the skb? */
2414 u8 free:2;
2415#define NAPI_GRO_FREE 1
2416#define NAPI_GRO_FREE_STOLEN_HEAD 2
2417
2418 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2419 u8 is_ipv6:1;
2420
2421 /* Used in GRE, set in fou/gue_gro_receive */
2422 u8 is_fou:1;
2423
2424 /* Used to determine if flush_id can be ignored */
2425 u8 is_atomic:1;
2426
2427 /* Number of gro_receive callbacks this packet already went through */
2428 u8 recursion_counter:4;
2429
2430 /* GRO is done by frag_list pointer chaining. */
2431 u8 is_flist:1;
2432
2433 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2434 __wsum csum;
2435
2436 /* used in skb_gro_receive() slow path */
2437 struct sk_buff *last;
2438};
2439
2440#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2441
2442#define GRO_RECURSION_LIMIT 15
2443static inline int gro_recursion_inc_test(struct sk_buff *skb)
2444{
2445 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2446}
2447
2448typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2449static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2450 struct list_head *head,
2451 struct sk_buff *skb)
2452{
2453 if (unlikely(gro_recursion_inc_test(skb))) {
2454 NAPI_GRO_CB(skb)->flush |= 1;
2455 return NULL;
2456 }
2457
2458 return cb(head, skb);
2459}
2460
2461typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2462 struct sk_buff *);
2463static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2464 struct sock *sk,
2465 struct list_head *head,
2466 struct sk_buff *skb)
2467{
2468 if (unlikely(gro_recursion_inc_test(skb))) {
2469 NAPI_GRO_CB(skb)->flush |= 1;
2470 return NULL;
2471 }
2472
2473 return cb(sk, head, skb);
2474}
2475
2476struct packet_type {
2477 __be16 type; /* This is really htons(ether_type). */
2478 bool ignore_outgoing;
2479 struct net_device *dev; /* NULL is wildcarded here */
2480 int (*func) (struct sk_buff *,
2481 struct net_device *,
2482 struct packet_type *,
2483 struct net_device *);
2484 void (*list_func) (struct list_head *,
2485 struct packet_type *,
2486 struct net_device *);
2487 bool (*id_match)(struct packet_type *ptype,
2488 struct sock *sk);
2489 void *af_packet_priv;
2490 struct list_head list;
2491};
2492
2493struct offload_callbacks {
2494 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2495 netdev_features_t features);
2496 struct sk_buff *(*gro_receive)(struct list_head *head,
2497 struct sk_buff *skb);
2498 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2499};
2500
2501struct packet_offload {
2502 __be16 type; /* This is really htons(ether_type). */
2503 u16 priority;
2504 struct offload_callbacks callbacks;
2505 struct list_head list;
2506};
2507
2508/* often modified stats are per-CPU, other are shared (netdev->stats) */
2509struct pcpu_sw_netstats {
2510 u64 rx_packets;
2511 u64 rx_bytes;
2512 u64 tx_packets;
2513 u64 tx_bytes;
2514 struct u64_stats_sync syncp;
2515} __aligned(4 * sizeof(u64));
2516
2517struct pcpu_lstats {
2518 u64_stats_t packets;
2519 u64_stats_t bytes;
2520 struct u64_stats_sync syncp;
2521} __aligned(2 * sizeof(u64));
2522
2523void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2524
2525static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2526{
2527 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2528
2529 u64_stats_update_begin(&lstats->syncp);
2530 u64_stats_add(&lstats->bytes, len);
2531 u64_stats_inc(&lstats->packets);
2532 u64_stats_update_end(&lstats->syncp);
2533}
2534
2535#define __netdev_alloc_pcpu_stats(type, gfp) \
2536({ \
2537 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2538 if (pcpu_stats) { \
2539 int __cpu; \
2540 for_each_possible_cpu(__cpu) { \
2541 typeof(type) *stat; \
2542 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2543 u64_stats_init(&stat->syncp); \
2544 } \
2545 } \
2546 pcpu_stats; \
2547})
2548
2549#define netdev_alloc_pcpu_stats(type) \
2550 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2551
2552enum netdev_lag_tx_type {
2553 NETDEV_LAG_TX_TYPE_UNKNOWN,
2554 NETDEV_LAG_TX_TYPE_RANDOM,
2555 NETDEV_LAG_TX_TYPE_BROADCAST,
2556 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2557 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2558 NETDEV_LAG_TX_TYPE_HASH,
2559};
2560
2561enum netdev_lag_hash {
2562 NETDEV_LAG_HASH_NONE,
2563 NETDEV_LAG_HASH_L2,
2564 NETDEV_LAG_HASH_L34,
2565 NETDEV_LAG_HASH_L23,
2566 NETDEV_LAG_HASH_E23,
2567 NETDEV_LAG_HASH_E34,
2568 NETDEV_LAG_HASH_UNKNOWN,
2569};
2570
2571struct netdev_lag_upper_info {
2572 enum netdev_lag_tx_type tx_type;
2573 enum netdev_lag_hash hash_type;
2574};
2575
2576struct netdev_lag_lower_state_info {
2577 u8 link_up : 1,
2578 tx_enabled : 1;
2579};
2580
2581#include <linux/notifier.h>
2582
2583/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2584 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2585 * adding new types.
2586 */
2587enum netdev_cmd {
2588 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2589 NETDEV_DOWN,
2590 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2591 detected a hardware crash and restarted
2592 - we can use this eg to kick tcp sessions
2593 once done */
2594 NETDEV_CHANGE, /* Notify device state change */
2595 NETDEV_REGISTER,
2596 NETDEV_UNREGISTER,
2597 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2598 NETDEV_CHANGEADDR, /* notify after the address change */
2599 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2600 NETDEV_GOING_DOWN,
2601 NETDEV_CHANGENAME,
2602 NETDEV_FEAT_CHANGE,
2603 NETDEV_BONDING_FAILOVER,
2604 NETDEV_PRE_UP,
2605 NETDEV_PRE_TYPE_CHANGE,
2606 NETDEV_POST_TYPE_CHANGE,
2607 NETDEV_POST_INIT,
2608 NETDEV_RELEASE,
2609 NETDEV_NOTIFY_PEERS,
2610 NETDEV_JOIN,
2611 NETDEV_CHANGEUPPER,
2612 NETDEV_RESEND_IGMP,
2613 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2614 NETDEV_CHANGEINFODATA,
2615 NETDEV_BONDING_INFO,
2616 NETDEV_PRECHANGEUPPER,
2617 NETDEV_CHANGELOWERSTATE,
2618 NETDEV_UDP_TUNNEL_PUSH_INFO,
2619 NETDEV_UDP_TUNNEL_DROP_INFO,
2620 NETDEV_CHANGE_TX_QUEUE_LEN,
2621 NETDEV_CVLAN_FILTER_PUSH_INFO,
2622 NETDEV_CVLAN_FILTER_DROP_INFO,
2623 NETDEV_SVLAN_FILTER_PUSH_INFO,
2624 NETDEV_SVLAN_FILTER_DROP_INFO,
2625};
2626const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2627
2628int register_netdevice_notifier(struct notifier_block *nb);
2629int unregister_netdevice_notifier(struct notifier_block *nb);
2630int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2631int unregister_netdevice_notifier_net(struct net *net,
2632 struct notifier_block *nb);
2633int register_netdevice_notifier_dev_net(struct net_device *dev,
2634 struct notifier_block *nb,
2635 struct netdev_net_notifier *nn);
2636int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2637 struct notifier_block *nb,
2638 struct netdev_net_notifier *nn);
2639
2640struct netdev_notifier_info {
2641 struct net_device *dev;
2642 struct netlink_ext_ack *extack;
2643};
2644
2645struct netdev_notifier_info_ext {
2646 struct netdev_notifier_info info; /* must be first */
2647 union {
2648 u32 mtu;
2649 } ext;
2650};
2651
2652struct netdev_notifier_change_info {
2653 struct netdev_notifier_info info; /* must be first */
2654 unsigned int flags_changed;
2655};
2656
2657struct netdev_notifier_changeupper_info {
2658 struct netdev_notifier_info info; /* must be first */
2659 struct net_device *upper_dev; /* new upper dev */
2660 bool master; /* is upper dev master */
2661 bool linking; /* is the notification for link or unlink */
2662 void *upper_info; /* upper dev info */
2663};
2664
2665struct netdev_notifier_changelowerstate_info {
2666 struct netdev_notifier_info info; /* must be first */
2667 void *lower_state_info; /* is lower dev state */
2668};
2669
2670struct netdev_notifier_pre_changeaddr_info {
2671 struct netdev_notifier_info info; /* must be first */
2672 const unsigned char *dev_addr;
2673};
2674
2675static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2676 struct net_device *dev)
2677{
2678 info->dev = dev;
2679 info->extack = NULL;
2680}
2681
2682static inline struct net_device *
2683netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2684{
2685 return info->dev;
2686}
2687
2688static inline struct netlink_ext_ack *
2689netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2690{
2691 return info->extack;
2692}
2693
2694int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2695
2696
2697extern rwlock_t dev_base_lock; /* Device list lock */
2698
2699#define for_each_netdev(net, d) \
2700 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2701#define for_each_netdev_reverse(net, d) \
2702 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2703#define for_each_netdev_rcu(net, d) \
2704 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2705#define for_each_netdev_safe(net, d, n) \
2706 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2707#define for_each_netdev_continue(net, d) \
2708 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2709#define for_each_netdev_continue_reverse(net, d) \
2710 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2711 dev_list)
2712#define for_each_netdev_continue_rcu(net, d) \
2713 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2714#define for_each_netdev_in_bond_rcu(bond, slave) \
2715 for_each_netdev_rcu(&init_net, slave) \
2716 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2717#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2718
2719static inline struct net_device *next_net_device(struct net_device *dev)
2720{
2721 struct list_head *lh;
2722 struct net *net;
2723
2724 net = dev_net(dev);
2725 lh = dev->dev_list.next;
2726 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2727}
2728
2729static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2730{
2731 struct list_head *lh;
2732 struct net *net;
2733
2734 net = dev_net(dev);
2735 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2736 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2737}
2738
2739static inline struct net_device *first_net_device(struct net *net)
2740{
2741 return list_empty(&net->dev_base_head) ? NULL :
2742 net_device_entry(net->dev_base_head.next);
2743}
2744
2745static inline struct net_device *first_net_device_rcu(struct net *net)
2746{
2747 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2748
2749 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2750}
2751
2752int netdev_boot_setup_check(struct net_device *dev);
2753unsigned long netdev_boot_base(const char *prefix, int unit);
2754struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2755 const char *hwaddr);
2756struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2757struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2758void dev_add_pack(struct packet_type *pt);
2759void dev_remove_pack(struct packet_type *pt);
2760void __dev_remove_pack(struct packet_type *pt);
2761void dev_add_offload(struct packet_offload *po);
2762void dev_remove_offload(struct packet_offload *po);
2763
2764int dev_get_iflink(const struct net_device *dev);
2765int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2766struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2767 unsigned short mask);
2768struct net_device *dev_get_by_name(struct net *net, const char *name);
2769struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2770struct net_device *__dev_get_by_name(struct net *net, const char *name);
2771int dev_alloc_name(struct net_device *dev, const char *name);
2772int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2773void dev_close(struct net_device *dev);
2774void dev_close_many(struct list_head *head, bool unlink);
2775void dev_disable_lro(struct net_device *dev);
2776int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2777u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2778 struct net_device *sb_dev);
2779u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2780 struct net_device *sb_dev);
2781int dev_queue_xmit(struct sk_buff *skb);
2782int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2783int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2784int register_netdevice(struct net_device *dev);
2785void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2786void unregister_netdevice_many(struct list_head *head);
2787static inline void unregister_netdevice(struct net_device *dev)
2788{
2789 unregister_netdevice_queue(dev, NULL);
2790}
2791
2792int netdev_refcnt_read(const struct net_device *dev);
2793void free_netdev(struct net_device *dev);
2794void netdev_freemem(struct net_device *dev);
2795void synchronize_net(void);
2796int init_dummy_netdev(struct net_device *dev);
2797
2798struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2799 struct sk_buff *skb,
2800 bool all_slaves);
2801struct net_device *dev_get_by_index(struct net *net, int ifindex);
2802struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2803struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2804struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2805int netdev_get_name(struct net *net, char *name, int ifindex);
2806int dev_restart(struct net_device *dev);
2807int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2808int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2809
2810static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2811{
2812 return NAPI_GRO_CB(skb)->data_offset;
2813}
2814
2815static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2816{
2817 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2818}
2819
2820static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2821{
2822 NAPI_GRO_CB(skb)->data_offset += len;
2823}
2824
2825static inline void *skb_gro_header_fast(struct sk_buff *skb,
2826 unsigned int offset)
2827{
2828 return NAPI_GRO_CB(skb)->frag0 + offset;
2829}
2830
2831static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2832{
2833 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2834}
2835
2836static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2837{
2838 NAPI_GRO_CB(skb)->frag0 = NULL;
2839 NAPI_GRO_CB(skb)->frag0_len = 0;
2840}
2841
2842static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2843 unsigned int offset)
2844{
2845 if (!pskb_may_pull(skb, hlen))
2846 return NULL;
2847
2848 skb_gro_frag0_invalidate(skb);
2849 return skb->data + offset;
2850}
2851
2852static inline void *skb_gro_network_header(struct sk_buff *skb)
2853{
2854 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2855 skb_network_offset(skb);
2856}
2857
2858static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2859 const void *start, unsigned int len)
2860{
2861 if (NAPI_GRO_CB(skb)->csum_valid)
2862 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2863 csum_partial(start, len, 0));
2864}
2865
2866/* GRO checksum functions. These are logical equivalents of the normal
2867 * checksum functions (in skbuff.h) except that they operate on the GRO
2868 * offsets and fields in sk_buff.
2869 */
2870
2871__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2872
2873static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2874{
2875 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2876}
2877
2878static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2879 bool zero_okay,
2880 __sum16 check)
2881{
2882 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2883 skb_checksum_start_offset(skb) <
2884 skb_gro_offset(skb)) &&
2885 !skb_at_gro_remcsum_start(skb) &&
2886 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2887 (!zero_okay || check));
2888}
2889
2890static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2891 __wsum psum)
2892{
2893 if (NAPI_GRO_CB(skb)->csum_valid &&
2894 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2895 return 0;
2896
2897 NAPI_GRO_CB(skb)->csum = psum;
2898
2899 return __skb_gro_checksum_complete(skb);
2900}
2901
2902static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2903{
2904 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2905 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2906 NAPI_GRO_CB(skb)->csum_cnt--;
2907 } else {
2908 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2909 * verified a new top level checksum or an encapsulated one
2910 * during GRO. This saves work if we fallback to normal path.
2911 */
2912 __skb_incr_checksum_unnecessary(skb);
2913 }
2914}
2915
2916#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2917 compute_pseudo) \
2918({ \
2919 __sum16 __ret = 0; \
2920 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2921 __ret = __skb_gro_checksum_validate_complete(skb, \
2922 compute_pseudo(skb, proto)); \
2923 if (!__ret) \
2924 skb_gro_incr_csum_unnecessary(skb); \
2925 __ret; \
2926})
2927
2928#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2929 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2930
2931#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2932 compute_pseudo) \
2933 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2934
2935#define skb_gro_checksum_simple_validate(skb) \
2936 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2937
2938static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2939{
2940 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2941 !NAPI_GRO_CB(skb)->csum_valid);
2942}
2943
2944static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2945 __wsum pseudo)
2946{
2947 NAPI_GRO_CB(skb)->csum = ~pseudo;
2948 NAPI_GRO_CB(skb)->csum_valid = 1;
2949}
2950
2951#define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
2952do { \
2953 if (__skb_gro_checksum_convert_check(skb)) \
2954 __skb_gro_checksum_convert(skb, \
2955 compute_pseudo(skb, proto)); \
2956} while (0)
2957
2958struct gro_remcsum {
2959 int offset;
2960 __wsum delta;
2961};
2962
2963static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2964{
2965 grc->offset = 0;
2966 grc->delta = 0;
2967}
2968
2969static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2970 unsigned int off, size_t hdrlen,
2971 int start, int offset,
2972 struct gro_remcsum *grc,
2973 bool nopartial)
2974{
2975 __wsum delta;
2976 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2977
2978 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2979
2980 if (!nopartial) {
2981 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2982 return ptr;
2983 }
2984
2985 ptr = skb_gro_header_fast(skb, off);
2986 if (skb_gro_header_hard(skb, off + plen)) {
2987 ptr = skb_gro_header_slow(skb, off + plen, off);
2988 if (!ptr)
2989 return NULL;
2990 }
2991
2992 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2993 start, offset);
2994
2995 /* Adjust skb->csum since we changed the packet */
2996 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2997
2998 grc->offset = off + hdrlen + offset;
2999 grc->delta = delta;
3000
3001 return ptr;
3002}
3003
3004static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3005 struct gro_remcsum *grc)
3006{
3007 void *ptr;
3008 size_t plen = grc->offset + sizeof(u16);
3009
3010 if (!grc->delta)
3011 return;
3012
3013 ptr = skb_gro_header_fast(skb, grc->offset);
3014 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3015 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3016 if (!ptr)
3017 return;
3018 }
3019
3020 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3021}
3022
3023#ifdef CONFIG_XFRM_OFFLOAD
3024static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3025{
3026 if (PTR_ERR(pp) != -EINPROGRESS)
3027 NAPI_GRO_CB(skb)->flush |= flush;
3028}
3029static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3030 struct sk_buff *pp,
3031 int flush,
3032 struct gro_remcsum *grc)
3033{
3034 if (PTR_ERR(pp) != -EINPROGRESS) {
3035 NAPI_GRO_CB(skb)->flush |= flush;
3036 skb_gro_remcsum_cleanup(skb, grc);
3037 skb->remcsum_offload = 0;
3038 }
3039}
3040#else
3041static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3042{
3043 NAPI_GRO_CB(skb)->flush |= flush;
3044}
3045static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3046 struct sk_buff *pp,
3047 int flush,
3048 struct gro_remcsum *grc)
3049{
3050 NAPI_GRO_CB(skb)->flush |= flush;
3051 skb_gro_remcsum_cleanup(skb, grc);
3052 skb->remcsum_offload = 0;
3053}
3054#endif
3055
3056static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3057 unsigned short type,
3058 const void *daddr, const void *saddr,
3059 unsigned int len)
3060{
3061 if (!dev->header_ops || !dev->header_ops->create)
3062 return 0;
3063
3064 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3065}
3066
3067static inline int dev_parse_header(const struct sk_buff *skb,
3068 unsigned char *haddr)
3069{
3070 const struct net_device *dev = skb->dev;
3071
3072 if (!dev->header_ops || !dev->header_ops->parse)
3073 return 0;
3074 return dev->header_ops->parse(skb, haddr);
3075}
3076
3077static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3078{
3079 const struct net_device *dev = skb->dev;
3080
3081 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3082 return 0;
3083 return dev->header_ops->parse_protocol(skb);
3084}
3085
3086/* ll_header must have at least hard_header_len allocated */
3087static inline bool dev_validate_header(const struct net_device *dev,
3088 char *ll_header, int len)
3089{
3090 if (likely(len >= dev->hard_header_len))
3091 return true;
3092 if (len < dev->min_header_len)
3093 return false;
3094
3095 if (capable(CAP_SYS_RAWIO)) {
3096 memset(ll_header + len, 0, dev->hard_header_len - len);
3097 return true;
3098 }
3099
3100 if (dev->header_ops && dev->header_ops->validate)
3101 return dev->header_ops->validate(ll_header, len);
3102
3103 return false;
3104}
3105
3106typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3107 int len, int size);
3108int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3109static inline int unregister_gifconf(unsigned int family)
3110{
3111 return register_gifconf(family, NULL);
3112}
3113
3114#ifdef CONFIG_NET_FLOW_LIMIT
3115#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3116struct sd_flow_limit {
3117 u64 count;
3118 unsigned int num_buckets;
3119 unsigned int history_head;
3120 u16 history[FLOW_LIMIT_HISTORY];
3121 u8 buckets[];
3122};
3123
3124extern int netdev_flow_limit_table_len;
3125#endif /* CONFIG_NET_FLOW_LIMIT */
3126
3127/*
3128 * Incoming packets are placed on per-CPU queues
3129 */
3130struct softnet_data {
3131 struct list_head poll_list;
3132 struct sk_buff_head process_queue;
3133
3134 /* stats */
3135 unsigned int processed;
3136 unsigned int time_squeeze;
3137 unsigned int received_rps;
3138#ifdef CONFIG_RPS
3139 struct softnet_data *rps_ipi_list;
3140#endif
3141#ifdef CONFIG_NET_FLOW_LIMIT
3142 struct sd_flow_limit __rcu *flow_limit;
3143#endif
3144 struct Qdisc *output_queue;
3145 struct Qdisc **output_queue_tailp;
3146 struct sk_buff *completion_queue;
3147#ifdef CONFIG_XFRM_OFFLOAD
3148 struct sk_buff_head xfrm_backlog;
3149#endif
3150 /* written and read only by owning cpu: */
3151 struct {
3152 u16 recursion;
3153 u8 more;
3154 } xmit;
3155#ifdef CONFIG_RPS
3156 /* input_queue_head should be written by cpu owning this struct,
3157 * and only read by other cpus. Worth using a cache line.
3158 */
3159 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3160
3161 /* Elements below can be accessed between CPUs for RPS/RFS */
3162 call_single_data_t csd ____cacheline_aligned_in_smp;
3163 struct softnet_data *rps_ipi_next;
3164 unsigned int cpu;
3165 unsigned int input_queue_tail;
3166#endif
3167 unsigned int dropped;
3168 struct sk_buff_head input_pkt_queue;
3169 struct napi_struct backlog;
3170
3171};
3172
3173static inline void input_queue_head_incr(struct softnet_data *sd)
3174{
3175#ifdef CONFIG_RPS
3176 sd->input_queue_head++;
3177#endif
3178}
3179
3180static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3181 unsigned int *qtail)
3182{
3183#ifdef CONFIG_RPS
3184 *qtail = ++sd->input_queue_tail;
3185#endif
3186}
3187
3188DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3189
3190static inline int dev_recursion_level(void)
3191{
3192 return this_cpu_read(softnet_data.xmit.recursion);
3193}
3194
3195#define XMIT_RECURSION_LIMIT 8
3196static inline bool dev_xmit_recursion(void)
3197{
3198 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3199 XMIT_RECURSION_LIMIT);
3200}
3201
3202static inline void dev_xmit_recursion_inc(void)
3203{
3204 __this_cpu_inc(softnet_data.xmit.recursion);
3205}
3206
3207static inline void dev_xmit_recursion_dec(void)
3208{
3209 __this_cpu_dec(softnet_data.xmit.recursion);
3210}
3211
3212void __netif_schedule(struct Qdisc *q);
3213void netif_schedule_queue(struct netdev_queue *txq);
3214
3215static inline void netif_tx_schedule_all(struct net_device *dev)
3216{
3217 unsigned int i;
3218
3219 for (i = 0; i < dev->num_tx_queues; i++)
3220 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3221}
3222
3223static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3224{
3225 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3226}
3227
3228/**
3229 * netif_start_queue - allow transmit
3230 * @dev: network device
3231 *
3232 * Allow upper layers to call the device hard_start_xmit routine.
3233 */
3234static inline void netif_start_queue(struct net_device *dev)
3235{
3236 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3237}
3238
3239static inline void netif_tx_start_all_queues(struct net_device *dev)
3240{
3241 unsigned int i;
3242
3243 for (i = 0; i < dev->num_tx_queues; i++) {
3244 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3245 netif_tx_start_queue(txq);
3246 }
3247}
3248
3249void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3250
3251/**
3252 * netif_wake_queue - restart transmit
3253 * @dev: network device
3254 *
3255 * Allow upper layers to call the device hard_start_xmit routine.
3256 * Used for flow control when transmit resources are available.
3257 */
3258static inline void netif_wake_queue(struct net_device *dev)
3259{
3260 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3261}
3262
3263static inline void netif_tx_wake_all_queues(struct net_device *dev)
3264{
3265 unsigned int i;
3266
3267 for (i = 0; i < dev->num_tx_queues; i++) {
3268 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3269 netif_tx_wake_queue(txq);
3270 }
3271}
3272
3273static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3274{
3275 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3276}
3277
3278/**
3279 * netif_stop_queue - stop transmitted packets
3280 * @dev: network device
3281 *
3282 * Stop upper layers calling the device hard_start_xmit routine.
3283 * Used for flow control when transmit resources are unavailable.
3284 */
3285static inline void netif_stop_queue(struct net_device *dev)
3286{
3287 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3288}
3289
3290void netif_tx_stop_all_queues(struct net_device *dev);
3291
3292static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3293{
3294 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3295}
3296
3297/**
3298 * netif_queue_stopped - test if transmit queue is flowblocked
3299 * @dev: network device
3300 *
3301 * Test if transmit queue on device is currently unable to send.
3302 */
3303static inline bool netif_queue_stopped(const struct net_device *dev)
3304{
3305 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3306}
3307
3308static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3309{
3310 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3311}
3312
3313static inline bool
3314netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3315{
3316 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3317}
3318
3319static inline bool
3320netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3321{
3322 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3323}
3324
3325/**
3326 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3327 * @dev_queue: pointer to transmit queue
3328 *
3329 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3330 * to give appropriate hint to the CPU.
3331 */
3332static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3333{
3334#ifdef CONFIG_BQL
3335 prefetchw(&dev_queue->dql.num_queued);
3336#endif
3337}
3338
3339/**
3340 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3341 * @dev_queue: pointer to transmit queue
3342 *
3343 * BQL enabled drivers might use this helper in their TX completion path,
3344 * to give appropriate hint to the CPU.
3345 */
3346static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3347{
3348#ifdef CONFIG_BQL
3349 prefetchw(&dev_queue->dql.limit);
3350#endif
3351}
3352
3353static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3354 unsigned int bytes)
3355{
3356#ifdef CONFIG_BQL
3357 dql_queued(&dev_queue->dql, bytes);
3358
3359 if (likely(dql_avail(&dev_queue->dql) >= 0))
3360 return;
3361
3362 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3363
3364 /*
3365 * The XOFF flag must be set before checking the dql_avail below,
3366 * because in netdev_tx_completed_queue we update the dql_completed
3367 * before checking the XOFF flag.
3368 */
3369 smp_mb();
3370
3371 /* check again in case another CPU has just made room avail */
3372 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3373 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3374#endif
3375}
3376
3377/* Variant of netdev_tx_sent_queue() for drivers that are aware
3378 * that they should not test BQL status themselves.
3379 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3380 * skb of a batch.
3381 * Returns true if the doorbell must be used to kick the NIC.
3382 */
3383static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3384 unsigned int bytes,
3385 bool xmit_more)
3386{
3387 if (xmit_more) {
3388#ifdef CONFIG_BQL
3389 dql_queued(&dev_queue->dql, bytes);
3390#endif
3391 return netif_tx_queue_stopped(dev_queue);
3392 }
3393 netdev_tx_sent_queue(dev_queue, bytes);
3394 return true;
3395}
3396
3397/**
3398 * netdev_sent_queue - report the number of bytes queued to hardware
3399 * @dev: network device
3400 * @bytes: number of bytes queued to the hardware device queue
3401 *
3402 * Report the number of bytes queued for sending/completion to the network
3403 * device hardware queue. @bytes should be a good approximation and should
3404 * exactly match netdev_completed_queue() @bytes
3405 */
3406static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3407{
3408 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3409}
3410
3411static inline bool __netdev_sent_queue(struct net_device *dev,
3412 unsigned int bytes,
3413 bool xmit_more)
3414{
3415 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3416 xmit_more);
3417}
3418
3419static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3420 unsigned int pkts, unsigned int bytes)
3421{
3422#ifdef CONFIG_BQL
3423 if (unlikely(!bytes))
3424 return;
3425
3426 dql_completed(&dev_queue->dql, bytes);
3427
3428 /*
3429 * Without the memory barrier there is a small possiblity that
3430 * netdev_tx_sent_queue will miss the update and cause the queue to
3431 * be stopped forever
3432 */
3433 smp_mb();
3434
3435 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3436 return;
3437
3438 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3439 netif_schedule_queue(dev_queue);
3440#endif
3441}
3442
3443/**
3444 * netdev_completed_queue - report bytes and packets completed by device
3445 * @dev: network device
3446 * @pkts: actual number of packets sent over the medium
3447 * @bytes: actual number of bytes sent over the medium
3448 *
3449 * Report the number of bytes and packets transmitted by the network device
3450 * hardware queue over the physical medium, @bytes must exactly match the
3451 * @bytes amount passed to netdev_sent_queue()
3452 */
3453static inline void netdev_completed_queue(struct net_device *dev,
3454 unsigned int pkts, unsigned int bytes)
3455{
3456 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3457}
3458
3459static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3460{
3461#ifdef CONFIG_BQL
3462 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3463 dql_reset(&q->dql);
3464#endif
3465}
3466
3467/**
3468 * netdev_reset_queue - reset the packets and bytes count of a network device
3469 * @dev_queue: network device
3470 *
3471 * Reset the bytes and packet count of a network device and clear the
3472 * software flow control OFF bit for this network device
3473 */
3474static inline void netdev_reset_queue(struct net_device *dev_queue)
3475{
3476 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3477}
3478
3479/**
3480 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3481 * @dev: network device
3482 * @queue_index: given tx queue index
3483 *
3484 * Returns 0 if given tx queue index >= number of device tx queues,
3485 * otherwise returns the originally passed tx queue index.
3486 */
3487static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3488{
3489 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3490 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3491 dev->name, queue_index,
3492 dev->real_num_tx_queues);
3493 return 0;
3494 }
3495
3496 return queue_index;
3497}
3498
3499/**
3500 * netif_running - test if up
3501 * @dev: network device
3502 *
3503 * Test if the device has been brought up.
3504 */
3505static inline bool netif_running(const struct net_device *dev)
3506{
3507 return test_bit(__LINK_STATE_START, &dev->state);
3508}
3509
3510/*
3511 * Routines to manage the subqueues on a device. We only need start,
3512 * stop, and a check if it's stopped. All other device management is
3513 * done at the overall netdevice level.
3514 * Also test the device if we're multiqueue.
3515 */
3516
3517/**
3518 * netif_start_subqueue - allow sending packets on subqueue
3519 * @dev: network device
3520 * @queue_index: sub queue index
3521 *
3522 * Start individual transmit queue of a device with multiple transmit queues.
3523 */
3524static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3525{
3526 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3527
3528 netif_tx_start_queue(txq);
3529}
3530
3531/**
3532 * netif_stop_subqueue - stop sending packets on subqueue
3533 * @dev: network device
3534 * @queue_index: sub queue index
3535 *
3536 * Stop individual transmit queue of a device with multiple transmit queues.
3537 */
3538static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3539{
3540 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3541 netif_tx_stop_queue(txq);
3542}
3543
3544/**
3545 * netif_subqueue_stopped - test status of subqueue
3546 * @dev: network device
3547 * @queue_index: sub queue index
3548 *
3549 * Check individual transmit queue of a device with multiple transmit queues.
3550 */
3551static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3552 u16 queue_index)
3553{
3554 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3555
3556 return netif_tx_queue_stopped(txq);
3557}
3558
3559static inline bool netif_subqueue_stopped(const struct net_device *dev,
3560 struct sk_buff *skb)
3561{
3562 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3563}
3564
3565/**
3566 * netif_wake_subqueue - allow sending packets on subqueue
3567 * @dev: network device
3568 * @queue_index: sub queue index
3569 *
3570 * Resume individual transmit queue of a device with multiple transmit queues.
3571 */
3572static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3573{
3574 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3575
3576 netif_tx_wake_queue(txq);
3577}
3578
3579#ifdef CONFIG_XPS
3580int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3581 u16 index);
3582int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3583 u16 index, bool is_rxqs_map);
3584
3585/**
3586 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3587 * @j: CPU/Rx queue index
3588 * @mask: bitmask of all cpus/rx queues
3589 * @nr_bits: number of bits in the bitmask
3590 *
3591 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3592 */
3593static inline bool netif_attr_test_mask(unsigned long j,
3594 const unsigned long *mask,
3595 unsigned int nr_bits)
3596{
3597 cpu_max_bits_warn(j, nr_bits);
3598 return test_bit(j, mask);
3599}
3600
3601/**
3602 * netif_attr_test_online - Test for online CPU/Rx queue
3603 * @j: CPU/Rx queue index
3604 * @online_mask: bitmask for CPUs/Rx queues that are online
3605 * @nr_bits: number of bits in the bitmask
3606 *
3607 * Returns true if a CPU/Rx queue is online.
3608 */
3609static inline bool netif_attr_test_online(unsigned long j,
3610 const unsigned long *online_mask,
3611 unsigned int nr_bits)
3612{
3613 cpu_max_bits_warn(j, nr_bits);
3614
3615 if (online_mask)
3616 return test_bit(j, online_mask);
3617
3618 return (j < nr_bits);
3619}
3620
3621/**
3622 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3623 * @n: CPU/Rx queue index
3624 * @srcp: the cpumask/Rx queue mask pointer
3625 * @nr_bits: number of bits in the bitmask
3626 *
3627 * Returns >= nr_bits if no further CPUs/Rx queues set.
3628 */
3629static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3630 unsigned int nr_bits)
3631{
3632 /* -1 is a legal arg here. */
3633 if (n != -1)
3634 cpu_max_bits_warn(n, nr_bits);
3635
3636 if (srcp)
3637 return find_next_bit(srcp, nr_bits, n + 1);
3638
3639 return n + 1;
3640}
3641
3642/**
3643 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3644 * @n: CPU/Rx queue index
3645 * @src1p: the first CPUs/Rx queues mask pointer
3646 * @src2p: the second CPUs/Rx queues mask pointer
3647 * @nr_bits: number of bits in the bitmask
3648 *
3649 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3650 */
3651static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3652 const unsigned long *src2p,
3653 unsigned int nr_bits)
3654{
3655 /* -1 is a legal arg here. */
3656 if (n != -1)
3657 cpu_max_bits_warn(n, nr_bits);
3658
3659 if (src1p && src2p)
3660 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3661 else if (src1p)
3662 return find_next_bit(src1p, nr_bits, n + 1);
3663 else if (src2p)
3664 return find_next_bit(src2p, nr_bits, n + 1);
3665
3666 return n + 1;
3667}
3668#else
3669static inline int netif_set_xps_queue(struct net_device *dev,
3670 const struct cpumask *mask,
3671 u16 index)
3672{
3673 return 0;
3674}
3675
3676static inline int __netif_set_xps_queue(struct net_device *dev,
3677 const unsigned long *mask,
3678 u16 index, bool is_rxqs_map)
3679{
3680 return 0;
3681}
3682#endif
3683
3684/**
3685 * netif_is_multiqueue - test if device has multiple transmit queues
3686 * @dev: network device
3687 *
3688 * Check if device has multiple transmit queues
3689 */
3690static inline bool netif_is_multiqueue(const struct net_device *dev)
3691{
3692 return dev->num_tx_queues > 1;
3693}
3694
3695int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3696
3697#ifdef CONFIG_SYSFS
3698int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3699#else
3700static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3701 unsigned int rxqs)
3702{
3703 dev->real_num_rx_queues = rxqs;
3704 return 0;
3705}
3706#endif
3707
3708static inline struct netdev_rx_queue *
3709__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3710{
3711 return dev->_rx + rxq;
3712}
3713
3714#ifdef CONFIG_SYSFS
3715static inline unsigned int get_netdev_rx_queue_index(
3716 struct netdev_rx_queue *queue)
3717{
3718 struct net_device *dev = queue->dev;
3719 int index = queue - dev->_rx;
3720
3721 BUG_ON(index >= dev->num_rx_queues);
3722 return index;
3723}
3724#endif
3725
3726#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3727int netif_get_num_default_rss_queues(void);
3728
3729enum skb_free_reason {
3730 SKB_REASON_CONSUMED,
3731 SKB_REASON_DROPPED,
3732};
3733
3734void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3735void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3736
3737/*
3738 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3739 * interrupt context or with hardware interrupts being disabled.
3740 * (in_irq() || irqs_disabled())
3741 *
3742 * We provide four helpers that can be used in following contexts :
3743 *
3744 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3745 * replacing kfree_skb(skb)
3746 *
3747 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3748 * Typically used in place of consume_skb(skb) in TX completion path
3749 *
3750 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3751 * replacing kfree_skb(skb)
3752 *
3753 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3754 * and consumed a packet. Used in place of consume_skb(skb)
3755 */
3756static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3757{
3758 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3759}
3760
3761static inline void dev_consume_skb_irq(struct sk_buff *skb)
3762{
3763 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3764}
3765
3766static inline void dev_kfree_skb_any(struct sk_buff *skb)
3767{
3768 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3769}
3770
3771static inline void dev_consume_skb_any(struct sk_buff *skb)
3772{
3773 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3774}
3775
3776void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3777int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3778int netif_rx(struct sk_buff *skb);
3779int netif_rx_ni(struct sk_buff *skb);
3780int netif_receive_skb(struct sk_buff *skb);
3781int netif_receive_skb_core(struct sk_buff *skb);
3782void netif_receive_skb_list(struct list_head *head);
3783gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3784void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3785struct sk_buff *napi_get_frags(struct napi_struct *napi);
3786gro_result_t napi_gro_frags(struct napi_struct *napi);
3787struct packet_offload *gro_find_receive_by_type(__be16 type);
3788struct packet_offload *gro_find_complete_by_type(__be16 type);
3789
3790static inline void napi_free_frags(struct napi_struct *napi)
3791{
3792 kfree_skb(napi->skb);
3793 napi->skb = NULL;
3794}
3795
3796bool netdev_is_rx_handler_busy(struct net_device *dev);
3797int netdev_rx_handler_register(struct net_device *dev,
3798 rx_handler_func_t *rx_handler,
3799 void *rx_handler_data);
3800void netdev_rx_handler_unregister(struct net_device *dev);
3801
3802bool dev_valid_name(const char *name);
3803int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3804 bool *need_copyout);
3805int dev_ifconf(struct net *net, struct ifconf *, int);
3806int dev_ethtool(struct net *net, struct ifreq *);
3807unsigned int dev_get_flags(const struct net_device *);
3808int __dev_change_flags(struct net_device *dev, unsigned int flags,
3809 struct netlink_ext_ack *extack);
3810int dev_change_flags(struct net_device *dev, unsigned int flags,
3811 struct netlink_ext_ack *extack);
3812void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3813 unsigned int gchanges);
3814int dev_change_name(struct net_device *, const char *);
3815int dev_set_alias(struct net_device *, const char *, size_t);
3816int dev_get_alias(const struct net_device *, char *, size_t);
3817int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3818int __dev_set_mtu(struct net_device *, int);
3819int dev_validate_mtu(struct net_device *dev, int mtu,
3820 struct netlink_ext_ack *extack);
3821int dev_set_mtu_ext(struct net_device *dev, int mtu,
3822 struct netlink_ext_ack *extack);
3823int dev_set_mtu(struct net_device *, int);
3824int dev_change_tx_queue_len(struct net_device *, unsigned long);
3825void dev_set_group(struct net_device *, int);
3826int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3827 struct netlink_ext_ack *extack);
3828int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3829 struct netlink_ext_ack *extack);
3830int dev_change_carrier(struct net_device *, bool new_carrier);
3831int dev_get_phys_port_id(struct net_device *dev,
3832 struct netdev_phys_item_id *ppid);
3833int dev_get_phys_port_name(struct net_device *dev,
3834 char *name, size_t len);
3835int dev_get_port_parent_id(struct net_device *dev,
3836 struct netdev_phys_item_id *ppid, bool recurse);
3837bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3838int dev_change_proto_down(struct net_device *dev, bool proto_down);
3839int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3840void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3841 u32 value);
3842struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3843struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3844 struct netdev_queue *txq, int *ret);
3845
3846typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3847int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3848 int fd, int expected_fd, u32 flags);
3849int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3850u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3851
3852int xdp_umem_query(struct net_device *dev, u16 queue_id);
3853
3854int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3855int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3856bool is_skb_forwardable(const struct net_device *dev,
3857 const struct sk_buff *skb);
3858
3859static __always_inline int ____dev_forward_skb(struct net_device *dev,
3860 struct sk_buff *skb)
3861{
3862 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3863 unlikely(!is_skb_forwardable(dev, skb))) {
3864 atomic_long_inc(&dev->rx_dropped);
3865 kfree_skb(skb);
3866 return NET_RX_DROP;
3867 }
3868
3869 skb_scrub_packet(skb, true);
3870 skb->priority = 0;
3871 return 0;
3872}
3873
3874bool dev_nit_active(struct net_device *dev);
3875void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3876
3877extern int netdev_budget;
3878extern unsigned int netdev_budget_usecs;
3879
3880/* Called by rtnetlink.c:rtnl_unlock() */
3881void netdev_run_todo(void);
3882
3883/**
3884 * dev_put - release reference to device
3885 * @dev: network device
3886 *
3887 * Release reference to device to allow it to be freed.
3888 */
3889static inline void dev_put(struct net_device *dev)
3890{
3891 this_cpu_dec(*dev->pcpu_refcnt);
3892}
3893
3894/**
3895 * dev_hold - get reference to device
3896 * @dev: network device
3897 *
3898 * Hold reference to device to keep it from being freed.
3899 */
3900static inline void dev_hold(struct net_device *dev)
3901{
3902 this_cpu_inc(*dev->pcpu_refcnt);
3903}
3904
3905/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3906 * and _off may be called from IRQ context, but it is caller
3907 * who is responsible for serialization of these calls.
3908 *
3909 * The name carrier is inappropriate, these functions should really be
3910 * called netif_lowerlayer_*() because they represent the state of any
3911 * kind of lower layer not just hardware media.
3912 */
3913
3914void linkwatch_init_dev(struct net_device *dev);
3915void linkwatch_fire_event(struct net_device *dev);
3916void linkwatch_forget_dev(struct net_device *dev);
3917
3918/**
3919 * netif_carrier_ok - test if carrier present
3920 * @dev: network device
3921 *
3922 * Check if carrier is present on device
3923 */
3924static inline bool netif_carrier_ok(const struct net_device *dev)
3925{
3926 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3927}
3928
3929unsigned long dev_trans_start(struct net_device *dev);
3930
3931void __netdev_watchdog_up(struct net_device *dev);
3932
3933void netif_carrier_on(struct net_device *dev);
3934
3935void netif_carrier_off(struct net_device *dev);
3936
3937/**
3938 * netif_dormant_on - mark device as dormant.
3939 * @dev: network device
3940 *
3941 * Mark device as dormant (as per RFC2863).
3942 *
3943 * The dormant state indicates that the relevant interface is not
3944 * actually in a condition to pass packets (i.e., it is not 'up') but is
3945 * in a "pending" state, waiting for some external event. For "on-
3946 * demand" interfaces, this new state identifies the situation where the
3947 * interface is waiting for events to place it in the up state.
3948 */
3949static inline void netif_dormant_on(struct net_device *dev)
3950{
3951 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3952 linkwatch_fire_event(dev);
3953}
3954
3955/**
3956 * netif_dormant_off - set device as not dormant.
3957 * @dev: network device
3958 *
3959 * Device is not in dormant state.
3960 */
3961static inline void netif_dormant_off(struct net_device *dev)
3962{
3963 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3964 linkwatch_fire_event(dev);
3965}
3966
3967/**
3968 * netif_dormant - test if device is dormant
3969 * @dev: network device
3970 *
3971 * Check if device is dormant.
3972 */
3973static inline bool netif_dormant(const struct net_device *dev)
3974{
3975 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3976}
3977
3978
3979/**
3980 * netif_testing_on - mark device as under test.
3981 * @dev: network device
3982 *
3983 * Mark device as under test (as per RFC2863).
3984 *
3985 * The testing state indicates that some test(s) must be performed on
3986 * the interface. After completion, of the test, the interface state
3987 * will change to up, dormant, or down, as appropriate.
3988 */
3989static inline void netif_testing_on(struct net_device *dev)
3990{
3991 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
3992 linkwatch_fire_event(dev);
3993}
3994
3995/**
3996 * netif_testing_off - set device as not under test.
3997 * @dev: network device
3998 *
3999 * Device is not in testing state.
4000 */
4001static inline void netif_testing_off(struct net_device *dev)
4002{
4003 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4004 linkwatch_fire_event(dev);
4005}
4006
4007/**
4008 * netif_testing - test if device is under test
4009 * @dev: network device
4010 *
4011 * Check if device is under test
4012 */
4013static inline bool netif_testing(const struct net_device *dev)
4014{
4015 return test_bit(__LINK_STATE_TESTING, &dev->state);
4016}
4017
4018
4019/**
4020 * netif_oper_up - test if device is operational
4021 * @dev: network device
4022 *
4023 * Check if carrier is operational
4024 */
4025static inline bool netif_oper_up(const struct net_device *dev)
4026{
4027 return (dev->operstate == IF_OPER_UP ||
4028 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4029}
4030
4031/**
4032 * netif_device_present - is device available or removed
4033 * @dev: network device
4034 *
4035 * Check if device has not been removed from system.
4036 */
4037static inline bool netif_device_present(struct net_device *dev)
4038{
4039 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4040}
4041
4042void netif_device_detach(struct net_device *dev);
4043
4044void netif_device_attach(struct net_device *dev);
4045
4046/*
4047 * Network interface message level settings
4048 */
4049
4050enum {
4051 NETIF_MSG_DRV_BIT,
4052 NETIF_MSG_PROBE_BIT,
4053 NETIF_MSG_LINK_BIT,
4054 NETIF_MSG_TIMER_BIT,
4055 NETIF_MSG_IFDOWN_BIT,
4056 NETIF_MSG_IFUP_BIT,
4057 NETIF_MSG_RX_ERR_BIT,
4058 NETIF_MSG_TX_ERR_BIT,
4059 NETIF_MSG_TX_QUEUED_BIT,
4060 NETIF_MSG_INTR_BIT,
4061 NETIF_MSG_TX_DONE_BIT,
4062 NETIF_MSG_RX_STATUS_BIT,
4063 NETIF_MSG_PKTDATA_BIT,
4064 NETIF_MSG_HW_BIT,
4065 NETIF_MSG_WOL_BIT,
4066
4067 /* When you add a new bit above, update netif_msg_class_names array
4068 * in net/ethtool/common.c
4069 */
4070 NETIF_MSG_CLASS_COUNT,
4071};
4072/* Both ethtool_ops interface and internal driver implementation use u32 */
4073static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4074
4075#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4076#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4077
4078#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4079#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4080#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4081#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4082#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4083#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4084#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4085#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4086#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4087#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4088#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4089#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4090#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4091#define NETIF_MSG_HW __NETIF_MSG(HW)
4092#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4093
4094#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4095#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4096#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4097#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4098#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4099#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4100#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4101#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4102#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4103#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4104#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4105#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4106#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4107#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4108#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4109
4110static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4111{
4112 /* use default */
4113 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4114 return default_msg_enable_bits;
4115 if (debug_value == 0) /* no output */
4116 return 0;
4117 /* set low N bits */
4118 return (1U << debug_value) - 1;
4119}
4120
4121static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4122{
4123 spin_lock(&txq->_xmit_lock);
4124 txq->xmit_lock_owner = cpu;
4125}
4126
4127static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4128{
4129 __acquire(&txq->_xmit_lock);
4130 return true;
4131}
4132
4133static inline void __netif_tx_release(struct netdev_queue *txq)
4134{
4135 __release(&txq->_xmit_lock);
4136}
4137
4138static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4139{
4140 spin_lock_bh(&txq->_xmit_lock);
4141 txq->xmit_lock_owner = smp_processor_id();
4142}
4143
4144static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4145{
4146 bool ok = spin_trylock(&txq->_xmit_lock);
4147 if (likely(ok))
4148 txq->xmit_lock_owner = smp_processor_id();
4149 return ok;
4150}
4151
4152static inline void __netif_tx_unlock(struct netdev_queue *txq)
4153{
4154 txq->xmit_lock_owner = -1;
4155 spin_unlock(&txq->_xmit_lock);
4156}
4157
4158static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4159{
4160 txq->xmit_lock_owner = -1;
4161 spin_unlock_bh(&txq->_xmit_lock);
4162}
4163
4164static inline void txq_trans_update(struct netdev_queue *txq)
4165{
4166 if (txq->xmit_lock_owner != -1)
4167 txq->trans_start = jiffies;
4168}
4169
4170/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4171static inline void netif_trans_update(struct net_device *dev)
4172{
4173 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4174
4175 if (txq->trans_start != jiffies)
4176 txq->trans_start = jiffies;
4177}
4178
4179/**
4180 * netif_tx_lock - grab network device transmit lock
4181 * @dev: network device
4182 *
4183 * Get network device transmit lock
4184 */
4185static inline void netif_tx_lock(struct net_device *dev)
4186{
4187 unsigned int i;
4188 int cpu;
4189
4190 spin_lock(&dev->tx_global_lock);
4191 cpu = smp_processor_id();
4192 for (i = 0; i < dev->num_tx_queues; i++) {
4193 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4194
4195 /* We are the only thread of execution doing a
4196 * freeze, but we have to grab the _xmit_lock in
4197 * order to synchronize with threads which are in
4198 * the ->hard_start_xmit() handler and already
4199 * checked the frozen bit.
4200 */
4201 __netif_tx_lock(txq, cpu);
4202 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4203 __netif_tx_unlock(txq);
4204 }
4205}
4206
4207static inline void netif_tx_lock_bh(struct net_device *dev)
4208{
4209 local_bh_disable();
4210 netif_tx_lock(dev);
4211}
4212
4213static inline void netif_tx_unlock(struct net_device *dev)
4214{
4215 unsigned int i;
4216
4217 for (i = 0; i < dev->num_tx_queues; i++) {
4218 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4219
4220 /* No need to grab the _xmit_lock here. If the
4221 * queue is not stopped for another reason, we
4222 * force a schedule.
4223 */
4224 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4225 netif_schedule_queue(txq);
4226 }
4227 spin_unlock(&dev->tx_global_lock);
4228}
4229
4230static inline void netif_tx_unlock_bh(struct net_device *dev)
4231{
4232 netif_tx_unlock(dev);
4233 local_bh_enable();
4234}
4235
4236#define HARD_TX_LOCK(dev, txq, cpu) { \
4237 if ((dev->features & NETIF_F_LLTX) == 0) { \
4238 __netif_tx_lock(txq, cpu); \
4239 } else { \
4240 __netif_tx_acquire(txq); \
4241 } \
4242}
4243
4244#define HARD_TX_TRYLOCK(dev, txq) \
4245 (((dev->features & NETIF_F_LLTX) == 0) ? \
4246 __netif_tx_trylock(txq) : \
4247 __netif_tx_acquire(txq))
4248
4249#define HARD_TX_UNLOCK(dev, txq) { \
4250 if ((dev->features & NETIF_F_LLTX) == 0) { \
4251 __netif_tx_unlock(txq); \
4252 } else { \
4253 __netif_tx_release(txq); \
4254 } \
4255}
4256
4257static inline void netif_tx_disable(struct net_device *dev)
4258{
4259 unsigned int i;
4260 int cpu;
4261
4262 local_bh_disable();
4263 cpu = smp_processor_id();
4264 for (i = 0; i < dev->num_tx_queues; i++) {
4265 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4266
4267 __netif_tx_lock(txq, cpu);
4268 netif_tx_stop_queue(txq);
4269 __netif_tx_unlock(txq);
4270 }
4271 local_bh_enable();
4272}
4273
4274static inline void netif_addr_lock(struct net_device *dev)
4275{
4276 unsigned char nest_level = 0;
4277
4278#ifdef CONFIG_LOCKDEP
4279 nest_level = dev->nested_level;
4280#endif
4281 spin_lock_nested(&dev->addr_list_lock, nest_level);
4282}
4283
4284static inline void netif_addr_lock_bh(struct net_device *dev)
4285{
4286 unsigned char nest_level = 0;
4287
4288#ifdef CONFIG_LOCKDEP
4289 nest_level = dev->nested_level;
4290#endif
4291 local_bh_disable();
4292 spin_lock_nested(&dev->addr_list_lock, nest_level);
4293}
4294
4295static inline void netif_addr_unlock(struct net_device *dev)
4296{
4297 spin_unlock(&dev->addr_list_lock);
4298}
4299
4300static inline void netif_addr_unlock_bh(struct net_device *dev)
4301{
4302 spin_unlock_bh(&dev->addr_list_lock);
4303}
4304
4305/*
4306 * dev_addrs walker. Should be used only for read access. Call with
4307 * rcu_read_lock held.
4308 */
4309#define for_each_dev_addr(dev, ha) \
4310 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4311
4312/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4313
4314void ether_setup(struct net_device *dev);
4315
4316/* Support for loadable net-drivers */
4317struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4318 unsigned char name_assign_type,
4319 void (*setup)(struct net_device *),
4320 unsigned int txqs, unsigned int rxqs);
4321#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4322 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4323
4324#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4325 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4326 count)
4327
4328int register_netdev(struct net_device *dev);
4329void unregister_netdev(struct net_device *dev);
4330
4331int devm_register_netdev(struct device *dev, struct net_device *ndev);
4332
4333/* General hardware address lists handling functions */
4334int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4335 struct netdev_hw_addr_list *from_list, int addr_len);
4336void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4337 struct netdev_hw_addr_list *from_list, int addr_len);
4338int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4339 struct net_device *dev,
4340 int (*sync)(struct net_device *, const unsigned char *),
4341 int (*unsync)(struct net_device *,
4342 const unsigned char *));
4343int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4344 struct net_device *dev,
4345 int (*sync)(struct net_device *,
4346 const unsigned char *, int),
4347 int (*unsync)(struct net_device *,
4348 const unsigned char *, int));
4349void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4350 struct net_device *dev,
4351 int (*unsync)(struct net_device *,
4352 const unsigned char *, int));
4353void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4354 struct net_device *dev,
4355 int (*unsync)(struct net_device *,
4356 const unsigned char *));
4357void __hw_addr_init(struct netdev_hw_addr_list *list);
4358
4359/* Functions used for device addresses handling */
4360int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4361 unsigned char addr_type);
4362int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4363 unsigned char addr_type);
4364void dev_addr_flush(struct net_device *dev);
4365int dev_addr_init(struct net_device *dev);
4366
4367/* Functions used for unicast addresses handling */
4368int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4369int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4370int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4371int dev_uc_sync(struct net_device *to, struct net_device *from);
4372int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4373void dev_uc_unsync(struct net_device *to, struct net_device *from);
4374void dev_uc_flush(struct net_device *dev);
4375void dev_uc_init(struct net_device *dev);
4376
4377/**
4378 * __dev_uc_sync - Synchonize device's unicast list
4379 * @dev: device to sync
4380 * @sync: function to call if address should be added
4381 * @unsync: function to call if address should be removed
4382 *
4383 * Add newly added addresses to the interface, and release
4384 * addresses that have been deleted.
4385 */
4386static inline int __dev_uc_sync(struct net_device *dev,
4387 int (*sync)(struct net_device *,
4388 const unsigned char *),
4389 int (*unsync)(struct net_device *,
4390 const unsigned char *))
4391{
4392 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4393}
4394
4395/**
4396 * __dev_uc_unsync - Remove synchronized addresses from device
4397 * @dev: device to sync
4398 * @unsync: function to call if address should be removed
4399 *
4400 * Remove all addresses that were added to the device by dev_uc_sync().
4401 */
4402static inline void __dev_uc_unsync(struct net_device *dev,
4403 int (*unsync)(struct net_device *,
4404 const unsigned char *))
4405{
4406 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4407}
4408
4409/* Functions used for multicast addresses handling */
4410int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4411int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4412int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4413int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4414int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4415int dev_mc_sync(struct net_device *to, struct net_device *from);
4416int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4417void dev_mc_unsync(struct net_device *to, struct net_device *from);
4418void dev_mc_flush(struct net_device *dev);
4419void dev_mc_init(struct net_device *dev);
4420
4421/**
4422 * __dev_mc_sync - Synchonize device's multicast list
4423 * @dev: device to sync
4424 * @sync: function to call if address should be added
4425 * @unsync: function to call if address should be removed
4426 *
4427 * Add newly added addresses to the interface, and release
4428 * addresses that have been deleted.
4429 */
4430static inline int __dev_mc_sync(struct net_device *dev,
4431 int (*sync)(struct net_device *,
4432 const unsigned char *),
4433 int (*unsync)(struct net_device *,
4434 const unsigned char *))
4435{
4436 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4437}
4438
4439/**
4440 * __dev_mc_unsync - Remove synchronized addresses from device
4441 * @dev: device to sync
4442 * @unsync: function to call if address should be removed
4443 *
4444 * Remove all addresses that were added to the device by dev_mc_sync().
4445 */
4446static inline void __dev_mc_unsync(struct net_device *dev,
4447 int (*unsync)(struct net_device *,
4448 const unsigned char *))
4449{
4450 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4451}
4452
4453/* Functions used for secondary unicast and multicast support */
4454void dev_set_rx_mode(struct net_device *dev);
4455void __dev_set_rx_mode(struct net_device *dev);
4456int dev_set_promiscuity(struct net_device *dev, int inc);
4457int dev_set_allmulti(struct net_device *dev, int inc);
4458void netdev_state_change(struct net_device *dev);
4459void netdev_notify_peers(struct net_device *dev);
4460void netdev_features_change(struct net_device *dev);
4461/* Load a device via the kmod */
4462void dev_load(struct net *net, const char *name);
4463struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4464 struct rtnl_link_stats64 *storage);
4465void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4466 const struct net_device_stats *netdev_stats);
4467
4468extern int netdev_max_backlog;
4469extern int netdev_tstamp_prequeue;
4470extern int weight_p;
4471extern int dev_weight_rx_bias;
4472extern int dev_weight_tx_bias;
4473extern int dev_rx_weight;
4474extern int dev_tx_weight;
4475extern int gro_normal_batch;
4476
4477enum {
4478 NESTED_SYNC_IMM_BIT,
4479 NESTED_SYNC_TODO_BIT,
4480};
4481
4482#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4483#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4484
4485#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4486#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4487
4488struct netdev_nested_priv {
4489 unsigned char flags;
4490 void *data;
4491};
4492
4493bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4494struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4495 struct list_head **iter);
4496struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4497 struct list_head **iter);
4498
4499#ifdef CONFIG_LOCKDEP
4500static LIST_HEAD(net_unlink_list);
4501
4502static inline void net_unlink_todo(struct net_device *dev)
4503{
4504 if (list_empty(&dev->unlink_list))
4505 list_add_tail(&dev->unlink_list, &net_unlink_list);
4506}
4507#endif
4508
4509/* iterate through upper list, must be called under RCU read lock */
4510#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4511 for (iter = &(dev)->adj_list.upper, \
4512 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4513 updev; \
4514 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4515
4516int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4517 int (*fn)(struct net_device *upper_dev,
4518 struct netdev_nested_priv *priv),
4519 struct netdev_nested_priv *priv);
4520
4521bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4522 struct net_device *upper_dev);
4523
4524bool netdev_has_any_upper_dev(struct net_device *dev);
4525
4526void *netdev_lower_get_next_private(struct net_device *dev,
4527 struct list_head **iter);
4528void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4529 struct list_head **iter);
4530
4531#define netdev_for_each_lower_private(dev, priv, iter) \
4532 for (iter = (dev)->adj_list.lower.next, \
4533 priv = netdev_lower_get_next_private(dev, &(iter)); \
4534 priv; \
4535 priv = netdev_lower_get_next_private(dev, &(iter)))
4536
4537#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4538 for (iter = &(dev)->adj_list.lower, \
4539 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4540 priv; \
4541 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4542
4543void *netdev_lower_get_next(struct net_device *dev,
4544 struct list_head **iter);
4545
4546#define netdev_for_each_lower_dev(dev, ldev, iter) \
4547 for (iter = (dev)->adj_list.lower.next, \
4548 ldev = netdev_lower_get_next(dev, &(iter)); \
4549 ldev; \
4550 ldev = netdev_lower_get_next(dev, &(iter)))
4551
4552struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4553 struct list_head **iter);
4554int netdev_walk_all_lower_dev(struct net_device *dev,
4555 int (*fn)(struct net_device *lower_dev,
4556 struct netdev_nested_priv *priv),
4557 struct netdev_nested_priv *priv);
4558int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4559 int (*fn)(struct net_device *lower_dev,
4560 struct netdev_nested_priv *priv),
4561 struct netdev_nested_priv *priv);
4562
4563void *netdev_adjacent_get_private(struct list_head *adj_list);
4564void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4565struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4566struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4567int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4568 struct netlink_ext_ack *extack);
4569int netdev_master_upper_dev_link(struct net_device *dev,
4570 struct net_device *upper_dev,
4571 void *upper_priv, void *upper_info,
4572 struct netlink_ext_ack *extack);
4573void netdev_upper_dev_unlink(struct net_device *dev,
4574 struct net_device *upper_dev);
4575int netdev_adjacent_change_prepare(struct net_device *old_dev,
4576 struct net_device *new_dev,
4577 struct net_device *dev,
4578 struct netlink_ext_ack *extack);
4579void netdev_adjacent_change_commit(struct net_device *old_dev,
4580 struct net_device *new_dev,
4581 struct net_device *dev);
4582void netdev_adjacent_change_abort(struct net_device *old_dev,
4583 struct net_device *new_dev,
4584 struct net_device *dev);
4585void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4586void *netdev_lower_dev_get_private(struct net_device *dev,
4587 struct net_device *lower_dev);
4588void netdev_lower_state_changed(struct net_device *lower_dev,
4589 void *lower_state_info);
4590
4591/* RSS keys are 40 or 52 bytes long */
4592#define NETDEV_RSS_KEY_LEN 52
4593extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4594void netdev_rss_key_fill(void *buffer, size_t len);
4595
4596int skb_checksum_help(struct sk_buff *skb);
4597int skb_crc32c_csum_help(struct sk_buff *skb);
4598int skb_csum_hwoffload_help(struct sk_buff *skb,
4599 const netdev_features_t features);
4600
4601struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4602 netdev_features_t features, bool tx_path);
4603struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4604 netdev_features_t features);
4605
4606struct netdev_bonding_info {
4607 ifslave slave;
4608 ifbond master;
4609};
4610
4611struct netdev_notifier_bonding_info {
4612 struct netdev_notifier_info info; /* must be first */
4613 struct netdev_bonding_info bonding_info;
4614};
4615
4616void netdev_bonding_info_change(struct net_device *dev,
4617 struct netdev_bonding_info *bonding_info);
4618
4619#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4620void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4621#else
4622static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4623 const void *data)
4624{
4625}
4626#endif
4627
4628static inline
4629struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4630{
4631 return __skb_gso_segment(skb, features, true);
4632}
4633__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4634
4635static inline bool can_checksum_protocol(netdev_features_t features,
4636 __be16 protocol)
4637{
4638 if (protocol == htons(ETH_P_FCOE))
4639 return !!(features & NETIF_F_FCOE_CRC);
4640
4641 /* Assume this is an IP checksum (not SCTP CRC) */
4642
4643 if (features & NETIF_F_HW_CSUM) {
4644 /* Can checksum everything */
4645 return true;
4646 }
4647
4648 switch (protocol) {
4649 case htons(ETH_P_IP):
4650 return !!(features & NETIF_F_IP_CSUM);
4651 case htons(ETH_P_IPV6):
4652 return !!(features & NETIF_F_IPV6_CSUM);
4653 default:
4654 return false;
4655 }
4656}
4657
4658#ifdef CONFIG_BUG
4659void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4660#else
4661static inline void netdev_rx_csum_fault(struct net_device *dev,
4662 struct sk_buff *skb)
4663{
4664}
4665#endif
4666/* rx skb timestamps */
4667void net_enable_timestamp(void);
4668void net_disable_timestamp(void);
4669
4670#ifdef CONFIG_PROC_FS
4671int __init dev_proc_init(void);
4672#else
4673#define dev_proc_init() 0
4674#endif
4675
4676static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4677 struct sk_buff *skb, struct net_device *dev,
4678 bool more)
4679{
4680 __this_cpu_write(softnet_data.xmit.more, more);
4681 return ops->ndo_start_xmit(skb, dev);
4682}
4683
4684static inline bool netdev_xmit_more(void)
4685{
4686 return __this_cpu_read(softnet_data.xmit.more);
4687}
4688
4689static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4690 struct netdev_queue *txq, bool more)
4691{
4692 const struct net_device_ops *ops = dev->netdev_ops;
4693 netdev_tx_t rc;
4694
4695 rc = __netdev_start_xmit(ops, skb, dev, more);
4696 if (rc == NETDEV_TX_OK)
4697 txq_trans_update(txq);
4698
4699 return rc;
4700}
4701
4702int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4703 const void *ns);
4704void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4705 const void *ns);
4706
4707static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4708{
4709 return netdev_class_create_file_ns(class_attr, NULL);
4710}
4711
4712static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4713{
4714 netdev_class_remove_file_ns(class_attr, NULL);
4715}
4716
4717extern const struct kobj_ns_type_operations net_ns_type_operations;
4718
4719const char *netdev_drivername(const struct net_device *dev);
4720
4721void linkwatch_run_queue(void);
4722
4723static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4724 netdev_features_t f2)
4725{
4726 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4727 if (f1 & NETIF_F_HW_CSUM)
4728 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4729 else
4730 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4731 }
4732
4733 return f1 & f2;
4734}
4735
4736static inline netdev_features_t netdev_get_wanted_features(
4737 struct net_device *dev)
4738{
4739 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4740}
4741netdev_features_t netdev_increment_features(netdev_features_t all,
4742 netdev_features_t one, netdev_features_t mask);
4743
4744/* Allow TSO being used on stacked device :
4745 * Performing the GSO segmentation before last device
4746 * is a performance improvement.
4747 */
4748static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4749 netdev_features_t mask)
4750{
4751 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4752}
4753
4754int __netdev_update_features(struct net_device *dev);
4755void netdev_update_features(struct net_device *dev);
4756void netdev_change_features(struct net_device *dev);
4757
4758void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4759 struct net_device *dev);
4760
4761netdev_features_t passthru_features_check(struct sk_buff *skb,
4762 struct net_device *dev,
4763 netdev_features_t features);
4764netdev_features_t netif_skb_features(struct sk_buff *skb);
4765
4766static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4767{
4768 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4769
4770 /* check flags correspondence */
4771 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4772 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4773 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4774 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4775 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4776 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4777 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4778 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4779 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4780 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4781 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4782 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4783 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4784 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4785 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4786 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4787 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4788 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4789 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4790
4791 return (features & feature) == feature;
4792}
4793
4794static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4795{
4796 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4797 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4798}
4799
4800static inline bool netif_needs_gso(struct sk_buff *skb,
4801 netdev_features_t features)
4802{
4803 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4804 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4805 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4806}
4807
4808static inline void netif_set_gso_max_size(struct net_device *dev,
4809 unsigned int size)
4810{
4811 dev->gso_max_size = size;
4812}
4813
4814static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4815 int pulled_hlen, u16 mac_offset,
4816 int mac_len)
4817{
4818 skb->protocol = protocol;
4819 skb->encapsulation = 1;
4820 skb_push(skb, pulled_hlen);
4821 skb_reset_transport_header(skb);
4822 skb->mac_header = mac_offset;
4823 skb->network_header = skb->mac_header + mac_len;
4824 skb->mac_len = mac_len;
4825}
4826
4827static inline bool netif_is_macsec(const struct net_device *dev)
4828{
4829 return dev->priv_flags & IFF_MACSEC;
4830}
4831
4832static inline bool netif_is_macvlan(const struct net_device *dev)
4833{
4834 return dev->priv_flags & IFF_MACVLAN;
4835}
4836
4837static inline bool netif_is_macvlan_port(const struct net_device *dev)
4838{
4839 return dev->priv_flags & IFF_MACVLAN_PORT;
4840}
4841
4842static inline bool netif_is_bond_master(const struct net_device *dev)
4843{
4844 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4845}
4846
4847static inline bool netif_is_bond_slave(const struct net_device *dev)
4848{
4849 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4850}
4851
4852static inline bool netif_supports_nofcs(struct net_device *dev)
4853{
4854 return dev->priv_flags & IFF_SUPP_NOFCS;
4855}
4856
4857static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4858{
4859 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4860}
4861
4862static inline bool netif_is_l3_master(const struct net_device *dev)
4863{
4864 return dev->priv_flags & IFF_L3MDEV_MASTER;
4865}
4866
4867static inline bool netif_is_l3_slave(const struct net_device *dev)
4868{
4869 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4870}
4871
4872static inline bool netif_is_bridge_master(const struct net_device *dev)
4873{
4874 return dev->priv_flags & IFF_EBRIDGE;
4875}
4876
4877static inline bool netif_is_bridge_port(const struct net_device *dev)
4878{
4879 return dev->priv_flags & IFF_BRIDGE_PORT;
4880}
4881
4882static inline bool netif_is_ovs_master(const struct net_device *dev)
4883{
4884 return dev->priv_flags & IFF_OPENVSWITCH;
4885}
4886
4887static inline bool netif_is_ovs_port(const struct net_device *dev)
4888{
4889 return dev->priv_flags & IFF_OVS_DATAPATH;
4890}
4891
4892static inline bool netif_is_any_bridge_port(const struct net_device *dev)
4893{
4894 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
4895}
4896
4897static inline bool netif_is_team_master(const struct net_device *dev)
4898{
4899 return dev->priv_flags & IFF_TEAM;
4900}
4901
4902static inline bool netif_is_team_port(const struct net_device *dev)
4903{
4904 return dev->priv_flags & IFF_TEAM_PORT;
4905}
4906
4907static inline bool netif_is_lag_master(const struct net_device *dev)
4908{
4909 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4910}
4911
4912static inline bool netif_is_lag_port(const struct net_device *dev)
4913{
4914 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4915}
4916
4917static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4918{
4919 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4920}
4921
4922static inline bool netif_is_failover(const struct net_device *dev)
4923{
4924 return dev->priv_flags & IFF_FAILOVER;
4925}
4926
4927static inline bool netif_is_failover_slave(const struct net_device *dev)
4928{
4929 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4930}
4931
4932/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4933static inline void netif_keep_dst(struct net_device *dev)
4934{
4935 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4936}
4937
4938/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4939static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4940{
4941 /* TODO: reserve and use an additional IFF bit, if we get more users */
4942 return dev->priv_flags & IFF_MACSEC;
4943}
4944
4945extern struct pernet_operations __net_initdata loopback_net_ops;
4946
4947/* Logging, debugging and troubleshooting/diagnostic helpers. */
4948
4949/* netdev_printk helpers, similar to dev_printk */
4950
4951static inline const char *netdev_name(const struct net_device *dev)
4952{
4953 if (!dev->name[0] || strchr(dev->name, '%'))
4954 return "(unnamed net_device)";
4955 return dev->name;
4956}
4957
4958static inline bool netdev_unregistering(const struct net_device *dev)
4959{
4960 return dev->reg_state == NETREG_UNREGISTERING;
4961}
4962
4963static inline const char *netdev_reg_state(const struct net_device *dev)
4964{
4965 switch (dev->reg_state) {
4966 case NETREG_UNINITIALIZED: return " (uninitialized)";
4967 case NETREG_REGISTERED: return "";
4968 case NETREG_UNREGISTERING: return " (unregistering)";
4969 case NETREG_UNREGISTERED: return " (unregistered)";
4970 case NETREG_RELEASED: return " (released)";
4971 case NETREG_DUMMY: return " (dummy)";
4972 }
4973
4974 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4975 return " (unknown)";
4976}
4977
4978__printf(3, 4) __cold
4979void netdev_printk(const char *level, const struct net_device *dev,
4980 const char *format, ...);
4981__printf(2, 3) __cold
4982void netdev_emerg(const struct net_device *dev, const char *format, ...);
4983__printf(2, 3) __cold
4984void netdev_alert(const struct net_device *dev, const char *format, ...);
4985__printf(2, 3) __cold
4986void netdev_crit(const struct net_device *dev, const char *format, ...);
4987__printf(2, 3) __cold
4988void netdev_err(const struct net_device *dev, const char *format, ...);
4989__printf(2, 3) __cold
4990void netdev_warn(const struct net_device *dev, const char *format, ...);
4991__printf(2, 3) __cold
4992void netdev_notice(const struct net_device *dev, const char *format, ...);
4993__printf(2, 3) __cold
4994void netdev_info(const struct net_device *dev, const char *format, ...);
4995
4996#define netdev_level_once(level, dev, fmt, ...) \
4997do { \
4998 static bool __print_once __read_mostly; \
4999 \
5000 if (!__print_once) { \
5001 __print_once = true; \
5002 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5003 } \
5004} while (0)
5005
5006#define netdev_emerg_once(dev, fmt, ...) \
5007 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5008#define netdev_alert_once(dev, fmt, ...) \
5009 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5010#define netdev_crit_once(dev, fmt, ...) \
5011 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5012#define netdev_err_once(dev, fmt, ...) \
5013 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5014#define netdev_warn_once(dev, fmt, ...) \
5015 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5016#define netdev_notice_once(dev, fmt, ...) \
5017 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5018#define netdev_info_once(dev, fmt, ...) \
5019 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5020
5021#define MODULE_ALIAS_NETDEV(device) \
5022 MODULE_ALIAS("netdev-" device)
5023
5024#if defined(CONFIG_DYNAMIC_DEBUG) || \
5025 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5026#define netdev_dbg(__dev, format, args...) \
5027do { \
5028 dynamic_netdev_dbg(__dev, format, ##args); \
5029} while (0)
5030#elif defined(DEBUG)
5031#define netdev_dbg(__dev, format, args...) \
5032 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5033#else
5034#define netdev_dbg(__dev, format, args...) \
5035({ \
5036 if (0) \
5037 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5038})
5039#endif
5040
5041#if defined(VERBOSE_DEBUG)
5042#define netdev_vdbg netdev_dbg
5043#else
5044
5045#define netdev_vdbg(dev, format, args...) \
5046({ \
5047 if (0) \
5048 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5049 0; \
5050})
5051#endif
5052
5053/*
5054 * netdev_WARN() acts like dev_printk(), but with the key difference
5055 * of using a WARN/WARN_ON to get the message out, including the
5056 * file/line information and a backtrace.
5057 */
5058#define netdev_WARN(dev, format, args...) \
5059 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5060 netdev_reg_state(dev), ##args)
5061
5062#define netdev_WARN_ONCE(dev, format, args...) \
5063 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5064 netdev_reg_state(dev), ##args)
5065
5066/* netif printk helpers, similar to netdev_printk */
5067
5068#define netif_printk(priv, type, level, dev, fmt, args...) \
5069do { \
5070 if (netif_msg_##type(priv)) \
5071 netdev_printk(level, (dev), fmt, ##args); \
5072} while (0)
5073
5074#define netif_level(level, priv, type, dev, fmt, args...) \
5075do { \
5076 if (netif_msg_##type(priv)) \
5077 netdev_##level(dev, fmt, ##args); \
5078} while (0)
5079
5080#define netif_emerg(priv, type, dev, fmt, args...) \
5081 netif_level(emerg, priv, type, dev, fmt, ##args)
5082#define netif_alert(priv, type, dev, fmt, args...) \
5083 netif_level(alert, priv, type, dev, fmt, ##args)
5084#define netif_crit(priv, type, dev, fmt, args...) \
5085 netif_level(crit, priv, type, dev, fmt, ##args)
5086#define netif_err(priv, type, dev, fmt, args...) \
5087 netif_level(err, priv, type, dev, fmt, ##args)
5088#define netif_warn(priv, type, dev, fmt, args...) \
5089 netif_level(warn, priv, type, dev, fmt, ##args)
5090#define netif_notice(priv, type, dev, fmt, args...) \
5091 netif_level(notice, priv, type, dev, fmt, ##args)
5092#define netif_info(priv, type, dev, fmt, args...) \
5093 netif_level(info, priv, type, dev, fmt, ##args)
5094
5095#if defined(CONFIG_DYNAMIC_DEBUG) || \
5096 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5097#define netif_dbg(priv, type, netdev, format, args...) \
5098do { \
5099 if (netif_msg_##type(priv)) \
5100 dynamic_netdev_dbg(netdev, format, ##args); \
5101} while (0)
5102#elif defined(DEBUG)
5103#define netif_dbg(priv, type, dev, format, args...) \
5104 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5105#else
5106#define netif_dbg(priv, type, dev, format, args...) \
5107({ \
5108 if (0) \
5109 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5110 0; \
5111})
5112#endif
5113
5114/* if @cond then downgrade to debug, else print at @level */
5115#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5116 do { \
5117 if (cond) \
5118 netif_dbg(priv, type, netdev, fmt, ##args); \
5119 else \
5120 netif_ ## level(priv, type, netdev, fmt, ##args); \
5121 } while (0)
5122
5123#if defined(VERBOSE_DEBUG)
5124#define netif_vdbg netif_dbg
5125#else
5126#define netif_vdbg(priv, type, dev, format, args...) \
5127({ \
5128 if (0) \
5129 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5130 0; \
5131})
5132#endif
5133
5134/*
5135 * The list of packet types we will receive (as opposed to discard)
5136 * and the routines to invoke.
5137 *
5138 * Why 16. Because with 16 the only overlap we get on a hash of the
5139 * low nibble of the protocol value is RARP/SNAP/X.25.
5140 *
5141 * 0800 IP
5142 * 0001 802.3
5143 * 0002 AX.25
5144 * 0004 802.2
5145 * 8035 RARP
5146 * 0005 SNAP
5147 * 0805 X.25
5148 * 0806 ARP
5149 * 8137 IPX
5150 * 0009 Localtalk
5151 * 86DD IPv6
5152 */
5153#define PTYPE_HASH_SIZE (16)
5154#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5155
5156extern struct net_device *blackhole_netdev;
5157
5158#endif /* _LINUX_NETDEVICE_H */
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
24#include <linux/timer.h>
25#include <linux/bug.h>
26#include <linux/delay.h>
27#include <linux/atomic.h>
28#include <linux/prefetch.h>
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31#include <asm/local.h>
32
33#include <linux/percpu.h>
34#include <linux/rculist.h>
35#include <linux/workqueue.h>
36#include <linux/dynamic_queue_limits.h>
37
38#include <net/net_namespace.h>
39#ifdef CONFIG_DCB
40#include <net/dcbnl.h>
41#endif
42#include <net/netprio_cgroup.h>
43#include <linux/netdev_features.h>
44#include <linux/neighbour.h>
45#include <linux/netdevice_xmit.h>
46#include <uapi/linux/netdevice.h>
47#include <uapi/linux/if_bonding.h>
48#include <uapi/linux/pkt_cls.h>
49#include <uapi/linux/netdev.h>
50#include <linux/hashtable.h>
51#include <linux/rbtree.h>
52#include <net/net_trackers.h>
53#include <net/net_debug.h>
54#include <net/dropreason-core.h>
55#include <net/neighbour_tables.h>
56
57struct netpoll_info;
58struct device;
59struct ethtool_ops;
60struct kernel_hwtstamp_config;
61struct phy_device;
62struct dsa_port;
63struct ip_tunnel_parm_kern;
64struct macsec_context;
65struct macsec_ops;
66struct netdev_name_node;
67struct sd_flow_limit;
68struct sfp_bus;
69/* 802.11 specific */
70struct wireless_dev;
71/* 802.15.4 specific */
72struct wpan_dev;
73struct mpls_dev;
74/* UDP Tunnel offloads */
75struct udp_tunnel_info;
76struct udp_tunnel_nic_info;
77struct udp_tunnel_nic;
78struct bpf_prog;
79struct xdp_buff;
80struct xdp_frame;
81struct xdp_metadata_ops;
82struct xdp_md;
83struct ethtool_netdev_state;
84struct phy_link_topology;
85
86typedef u32 xdp_features_t;
87
88void synchronize_net(void);
89void netdev_set_default_ethtool_ops(struct net_device *dev,
90 const struct ethtool_ops *ops);
91void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
92
93/* Backlog congestion levels */
94#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
95#define NET_RX_DROP 1 /* packet dropped */
96
97#define MAX_NEST_DEV 8
98
99/*
100 * Transmit return codes: transmit return codes originate from three different
101 * namespaces:
102 *
103 * - qdisc return codes
104 * - driver transmit return codes
105 * - errno values
106 *
107 * Drivers are allowed to return any one of those in their hard_start_xmit()
108 * function. Real network devices commonly used with qdiscs should only return
109 * the driver transmit return codes though - when qdiscs are used, the actual
110 * transmission happens asynchronously, so the value is not propagated to
111 * higher layers. Virtual network devices transmit synchronously; in this case
112 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
113 * others are propagated to higher layers.
114 */
115
116/* qdisc ->enqueue() return codes. */
117#define NET_XMIT_SUCCESS 0x00
118#define NET_XMIT_DROP 0x01 /* skb dropped */
119#define NET_XMIT_CN 0x02 /* congestion notification */
120#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
121
122/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
123 * indicates that the device will soon be dropping packets, or already drops
124 * some packets of the same priority; prompting us to send less aggressively. */
125#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
126#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
127
128/* Driver transmit return codes */
129#define NETDEV_TX_MASK 0xf0
130
131enum netdev_tx {
132 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
133 NETDEV_TX_OK = 0x00, /* driver took care of packet */
134 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
135};
136typedef enum netdev_tx netdev_tx_t;
137
138/*
139 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
140 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
141 */
142static inline bool dev_xmit_complete(int rc)
143{
144 /*
145 * Positive cases with an skb consumed by a driver:
146 * - successful transmission (rc == NETDEV_TX_OK)
147 * - error while transmitting (rc < 0)
148 * - error while queueing to a different device (rc & NET_XMIT_MASK)
149 */
150 if (likely(rc < NET_XMIT_MASK))
151 return true;
152
153 return false;
154}
155
156/*
157 * Compute the worst-case header length according to the protocols
158 * used.
159 */
160
161#if defined(CONFIG_HYPERV_NET)
162# define LL_MAX_HEADER 128
163#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
164# if defined(CONFIG_MAC80211_MESH)
165# define LL_MAX_HEADER 128
166# else
167# define LL_MAX_HEADER 96
168# endif
169#else
170# define LL_MAX_HEADER 32
171#endif
172
173#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
174 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
175#define MAX_HEADER LL_MAX_HEADER
176#else
177#define MAX_HEADER (LL_MAX_HEADER + 48)
178#endif
179
180/*
181 * Old network device statistics. Fields are native words
182 * (unsigned long) so they can be read and written atomically.
183 */
184
185#define NET_DEV_STAT(FIELD) \
186 union { \
187 unsigned long FIELD; \
188 atomic_long_t __##FIELD; \
189 }
190
191struct net_device_stats {
192 NET_DEV_STAT(rx_packets);
193 NET_DEV_STAT(tx_packets);
194 NET_DEV_STAT(rx_bytes);
195 NET_DEV_STAT(tx_bytes);
196 NET_DEV_STAT(rx_errors);
197 NET_DEV_STAT(tx_errors);
198 NET_DEV_STAT(rx_dropped);
199 NET_DEV_STAT(tx_dropped);
200 NET_DEV_STAT(multicast);
201 NET_DEV_STAT(collisions);
202 NET_DEV_STAT(rx_length_errors);
203 NET_DEV_STAT(rx_over_errors);
204 NET_DEV_STAT(rx_crc_errors);
205 NET_DEV_STAT(rx_frame_errors);
206 NET_DEV_STAT(rx_fifo_errors);
207 NET_DEV_STAT(rx_missed_errors);
208 NET_DEV_STAT(tx_aborted_errors);
209 NET_DEV_STAT(tx_carrier_errors);
210 NET_DEV_STAT(tx_fifo_errors);
211 NET_DEV_STAT(tx_heartbeat_errors);
212 NET_DEV_STAT(tx_window_errors);
213 NET_DEV_STAT(rx_compressed);
214 NET_DEV_STAT(tx_compressed);
215};
216#undef NET_DEV_STAT
217
218/* per-cpu stats, allocated on demand.
219 * Try to fit them in a single cache line, for dev_get_stats() sake.
220 */
221struct net_device_core_stats {
222 unsigned long rx_dropped;
223 unsigned long tx_dropped;
224 unsigned long rx_nohandler;
225 unsigned long rx_otherhost_dropped;
226} __aligned(4 * sizeof(unsigned long));
227
228#include <linux/cache.h>
229#include <linux/skbuff.h>
230
231struct neighbour;
232struct neigh_parms;
233struct sk_buff;
234
235struct netdev_hw_addr {
236 struct list_head list;
237 struct rb_node node;
238 unsigned char addr[MAX_ADDR_LEN];
239 unsigned char type;
240#define NETDEV_HW_ADDR_T_LAN 1
241#define NETDEV_HW_ADDR_T_SAN 2
242#define NETDEV_HW_ADDR_T_UNICAST 3
243#define NETDEV_HW_ADDR_T_MULTICAST 4
244 bool global_use;
245 int sync_cnt;
246 int refcount;
247 int synced;
248 struct rcu_head rcu_head;
249};
250
251struct netdev_hw_addr_list {
252 struct list_head list;
253 int count;
254
255 /* Auxiliary tree for faster lookup on addition and deletion */
256 struct rb_root tree;
257};
258
259#define netdev_hw_addr_list_count(l) ((l)->count)
260#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
261#define netdev_hw_addr_list_for_each(ha, l) \
262 list_for_each_entry(ha, &(l)->list, list)
263
264#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
265#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
266#define netdev_for_each_uc_addr(ha, dev) \
267 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
268#define netdev_for_each_synced_uc_addr(_ha, _dev) \
269 netdev_for_each_uc_addr((_ha), (_dev)) \
270 if ((_ha)->sync_cnt)
271
272#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
273#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
274#define netdev_for_each_mc_addr(ha, dev) \
275 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
276#define netdev_for_each_synced_mc_addr(_ha, _dev) \
277 netdev_for_each_mc_addr((_ha), (_dev)) \
278 if ((_ha)->sync_cnt)
279
280struct hh_cache {
281 unsigned int hh_len;
282 seqlock_t hh_lock;
283
284 /* cached hardware header; allow for machine alignment needs. */
285#define HH_DATA_MOD 16
286#define HH_DATA_OFF(__len) \
287 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
288#define HH_DATA_ALIGN(__len) \
289 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
290 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
291};
292
293/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
294 * Alternative is:
295 * dev->hard_header_len ? (dev->hard_header_len +
296 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
297 *
298 * We could use other alignment values, but we must maintain the
299 * relationship HH alignment <= LL alignment.
300 */
301#define LL_RESERVED_SPACE(dev) \
302 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
303 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
304#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
305 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
306 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
307
308struct header_ops {
309 int (*create) (struct sk_buff *skb, struct net_device *dev,
310 unsigned short type, const void *daddr,
311 const void *saddr, unsigned int len);
312 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
313 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
314 void (*cache_update)(struct hh_cache *hh,
315 const struct net_device *dev,
316 const unsigned char *haddr);
317 bool (*validate)(const char *ll_header, unsigned int len);
318 __be16 (*parse_protocol)(const struct sk_buff *skb);
319};
320
321/* These flag bits are private to the generic network queueing
322 * layer; they may not be explicitly referenced by any other
323 * code.
324 */
325
326enum netdev_state_t {
327 __LINK_STATE_START,
328 __LINK_STATE_PRESENT,
329 __LINK_STATE_NOCARRIER,
330 __LINK_STATE_LINKWATCH_PENDING,
331 __LINK_STATE_DORMANT,
332 __LINK_STATE_TESTING,
333};
334
335struct gro_list {
336 struct list_head list;
337 int count;
338};
339
340/*
341 * size of gro hash buckets, must less than bit number of
342 * napi_struct::gro_bitmask
343 */
344#define GRO_HASH_BUCKETS 8
345
346/*
347 * Structure for per-NAPI config
348 */
349struct napi_config {
350 u64 gro_flush_timeout;
351 u64 irq_suspend_timeout;
352 u32 defer_hard_irqs;
353 unsigned int napi_id;
354};
355
356/*
357 * Structure for NAPI scheduling similar to tasklet but with weighting
358 */
359struct napi_struct {
360 /* The poll_list must only be managed by the entity which
361 * changes the state of the NAPI_STATE_SCHED bit. This means
362 * whoever atomically sets that bit can add this napi_struct
363 * to the per-CPU poll_list, and whoever clears that bit
364 * can remove from the list right before clearing the bit.
365 */
366 struct list_head poll_list;
367
368 unsigned long state;
369 int weight;
370 u32 defer_hard_irqs_count;
371 unsigned long gro_bitmask;
372 int (*poll)(struct napi_struct *, int);
373#ifdef CONFIG_NETPOLL
374 /* CPU actively polling if netpoll is configured */
375 int poll_owner;
376#endif
377 /* CPU on which NAPI has been scheduled for processing */
378 int list_owner;
379 struct net_device *dev;
380 struct gro_list gro_hash[GRO_HASH_BUCKETS];
381 struct sk_buff *skb;
382 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
383 int rx_count; /* length of rx_list */
384 unsigned int napi_id;
385 struct hrtimer timer;
386 struct task_struct *thread;
387 unsigned long gro_flush_timeout;
388 unsigned long irq_suspend_timeout;
389 u32 defer_hard_irqs;
390 /* control-path-only fields follow */
391 struct list_head dev_list;
392 struct hlist_node napi_hash_node;
393 int irq;
394 int index;
395 struct napi_config *config;
396};
397
398enum {
399 NAPI_STATE_SCHED, /* Poll is scheduled */
400 NAPI_STATE_MISSED, /* reschedule a napi */
401 NAPI_STATE_DISABLE, /* Disable pending */
402 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
403 NAPI_STATE_LISTED, /* NAPI added to system lists */
404 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
405 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
406 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
407 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
408 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
409};
410
411enum {
412 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
413 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
414 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
415 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
416 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
417 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
418 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
419 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
420 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
421 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
422};
423
424enum gro_result {
425 GRO_MERGED,
426 GRO_MERGED_FREE,
427 GRO_HELD,
428 GRO_NORMAL,
429 GRO_CONSUMED,
430};
431typedef enum gro_result gro_result_t;
432
433/*
434 * enum rx_handler_result - Possible return values for rx_handlers.
435 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
436 * further.
437 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
438 * case skb->dev was changed by rx_handler.
439 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
440 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
441 *
442 * rx_handlers are functions called from inside __netif_receive_skb(), to do
443 * special processing of the skb, prior to delivery to protocol handlers.
444 *
445 * Currently, a net_device can only have a single rx_handler registered. Trying
446 * to register a second rx_handler will return -EBUSY.
447 *
448 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
449 * To unregister a rx_handler on a net_device, use
450 * netdev_rx_handler_unregister().
451 *
452 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
453 * do with the skb.
454 *
455 * If the rx_handler consumed the skb in some way, it should return
456 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
457 * the skb to be delivered in some other way.
458 *
459 * If the rx_handler changed skb->dev, to divert the skb to another
460 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
461 * new device will be called if it exists.
462 *
463 * If the rx_handler decides the skb should be ignored, it should return
464 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
465 * are registered on exact device (ptype->dev == skb->dev).
466 *
467 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
468 * delivered, it should return RX_HANDLER_PASS.
469 *
470 * A device without a registered rx_handler will behave as if rx_handler
471 * returned RX_HANDLER_PASS.
472 */
473
474enum rx_handler_result {
475 RX_HANDLER_CONSUMED,
476 RX_HANDLER_ANOTHER,
477 RX_HANDLER_EXACT,
478 RX_HANDLER_PASS,
479};
480typedef enum rx_handler_result rx_handler_result_t;
481typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
482
483void __napi_schedule(struct napi_struct *n);
484void __napi_schedule_irqoff(struct napi_struct *n);
485
486static inline bool napi_disable_pending(struct napi_struct *n)
487{
488 return test_bit(NAPI_STATE_DISABLE, &n->state);
489}
490
491static inline bool napi_prefer_busy_poll(struct napi_struct *n)
492{
493 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
494}
495
496/**
497 * napi_is_scheduled - test if NAPI is scheduled
498 * @n: NAPI context
499 *
500 * This check is "best-effort". With no locking implemented,
501 * a NAPI can be scheduled or terminate right after this check
502 * and produce not precise results.
503 *
504 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
505 * should not be used normally and napi_schedule should be
506 * used instead.
507 *
508 * Use only if the driver really needs to check if a NAPI
509 * is scheduled for example in the context of delayed timer
510 * that can be skipped if a NAPI is already scheduled.
511 *
512 * Return True if NAPI is scheduled, False otherwise.
513 */
514static inline bool napi_is_scheduled(struct napi_struct *n)
515{
516 return test_bit(NAPI_STATE_SCHED, &n->state);
517}
518
519bool napi_schedule_prep(struct napi_struct *n);
520
521/**
522 * napi_schedule - schedule NAPI poll
523 * @n: NAPI context
524 *
525 * Schedule NAPI poll routine to be called if it is not already
526 * running.
527 * Return true if we schedule a NAPI or false if not.
528 * Refer to napi_schedule_prep() for additional reason on why
529 * a NAPI might not be scheduled.
530 */
531static inline bool napi_schedule(struct napi_struct *n)
532{
533 if (napi_schedule_prep(n)) {
534 __napi_schedule(n);
535 return true;
536 }
537
538 return false;
539}
540
541/**
542 * napi_schedule_irqoff - schedule NAPI poll
543 * @n: NAPI context
544 *
545 * Variant of napi_schedule(), assuming hard irqs are masked.
546 */
547static inline void napi_schedule_irqoff(struct napi_struct *n)
548{
549 if (napi_schedule_prep(n))
550 __napi_schedule_irqoff(n);
551}
552
553/**
554 * napi_complete_done - NAPI processing complete
555 * @n: NAPI context
556 * @work_done: number of packets processed
557 *
558 * Mark NAPI processing as complete. Should only be called if poll budget
559 * has not been completely consumed.
560 * Prefer over napi_complete().
561 * Return false if device should avoid rearming interrupts.
562 */
563bool napi_complete_done(struct napi_struct *n, int work_done);
564
565static inline bool napi_complete(struct napi_struct *n)
566{
567 return napi_complete_done(n, 0);
568}
569
570int dev_set_threaded(struct net_device *dev, bool threaded);
571
572/**
573 * napi_disable - prevent NAPI from scheduling
574 * @n: NAPI context
575 *
576 * Stop NAPI from being scheduled on this context.
577 * Waits till any outstanding processing completes.
578 */
579void napi_disable(struct napi_struct *n);
580
581void napi_enable(struct napi_struct *n);
582
583/**
584 * napi_synchronize - wait until NAPI is not running
585 * @n: NAPI context
586 *
587 * Wait until NAPI is done being scheduled on this context.
588 * Waits till any outstanding processing completes but
589 * does not disable future activations.
590 */
591static inline void napi_synchronize(const struct napi_struct *n)
592{
593 if (IS_ENABLED(CONFIG_SMP))
594 while (test_bit(NAPI_STATE_SCHED, &n->state))
595 msleep(1);
596 else
597 barrier();
598}
599
600/**
601 * napi_if_scheduled_mark_missed - if napi is running, set the
602 * NAPIF_STATE_MISSED
603 * @n: NAPI context
604 *
605 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
606 * NAPI is scheduled.
607 **/
608static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
609{
610 unsigned long val, new;
611
612 val = READ_ONCE(n->state);
613 do {
614 if (val & NAPIF_STATE_DISABLE)
615 return true;
616
617 if (!(val & NAPIF_STATE_SCHED))
618 return false;
619
620 new = val | NAPIF_STATE_MISSED;
621 } while (!try_cmpxchg(&n->state, &val, new));
622
623 return true;
624}
625
626enum netdev_queue_state_t {
627 __QUEUE_STATE_DRV_XOFF,
628 __QUEUE_STATE_STACK_XOFF,
629 __QUEUE_STATE_FROZEN,
630};
631
632#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
633#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
634#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
635
636#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
637#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
638 QUEUE_STATE_FROZEN)
639#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
640 QUEUE_STATE_FROZEN)
641
642/*
643 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
644 * netif_tx_* functions below are used to manipulate this flag. The
645 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
646 * queue independently. The netif_xmit_*stopped functions below are called
647 * to check if the queue has been stopped by the driver or stack (either
648 * of the XOFF bits are set in the state). Drivers should not need to call
649 * netif_xmit*stopped functions, they should only be using netif_tx_*.
650 */
651
652struct netdev_queue {
653/*
654 * read-mostly part
655 */
656 struct net_device *dev;
657 netdevice_tracker dev_tracker;
658
659 struct Qdisc __rcu *qdisc;
660 struct Qdisc __rcu *qdisc_sleeping;
661#ifdef CONFIG_SYSFS
662 struct kobject kobj;
663#endif
664 unsigned long tx_maxrate;
665 /*
666 * Number of TX timeouts for this queue
667 * (/sys/class/net/DEV/Q/trans_timeout)
668 */
669 atomic_long_t trans_timeout;
670
671 /* Subordinate device that the queue has been assigned to */
672 struct net_device *sb_dev;
673#ifdef CONFIG_XDP_SOCKETS
674 struct xsk_buff_pool *pool;
675#endif
676
677/*
678 * write-mostly part
679 */
680#ifdef CONFIG_BQL
681 struct dql dql;
682#endif
683 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
684 int xmit_lock_owner;
685 /*
686 * Time (in jiffies) of last Tx
687 */
688 unsigned long trans_start;
689
690 unsigned long state;
691
692/*
693 * slow- / control-path part
694 */
695 /* NAPI instance for the queue
696 * Readers and writers must hold RTNL
697 */
698 struct napi_struct *napi;
699
700#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
701 int numa_node;
702#endif
703} ____cacheline_aligned_in_smp;
704
705extern int sysctl_fb_tunnels_only_for_init_net;
706extern int sysctl_devconf_inherit_init_net;
707
708/*
709 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
710 * == 1 : For initns only
711 * == 2 : For none.
712 */
713static inline bool net_has_fallback_tunnels(const struct net *net)
714{
715#if IS_ENABLED(CONFIG_SYSCTL)
716 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
717
718 return !fb_tunnels_only_for_init_net ||
719 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
720#else
721 return true;
722#endif
723}
724
725static inline int net_inherit_devconf(void)
726{
727#if IS_ENABLED(CONFIG_SYSCTL)
728 return READ_ONCE(sysctl_devconf_inherit_init_net);
729#else
730 return 0;
731#endif
732}
733
734static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
735{
736#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
737 return q->numa_node;
738#else
739 return NUMA_NO_NODE;
740#endif
741}
742
743static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
744{
745#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
746 q->numa_node = node;
747#endif
748}
749
750#ifdef CONFIG_RFS_ACCEL
751bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
752 u16 filter_id);
753#endif
754
755/* XPS map type and offset of the xps map within net_device->xps_maps[]. */
756enum xps_map_type {
757 XPS_CPUS = 0,
758 XPS_RXQS,
759 XPS_MAPS_MAX,
760};
761
762#ifdef CONFIG_XPS
763/*
764 * This structure holds an XPS map which can be of variable length. The
765 * map is an array of queues.
766 */
767struct xps_map {
768 unsigned int len;
769 unsigned int alloc_len;
770 struct rcu_head rcu;
771 u16 queues[];
772};
773#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
774#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
775 - sizeof(struct xps_map)) / sizeof(u16))
776
777/*
778 * This structure holds all XPS maps for device. Maps are indexed by CPU.
779 *
780 * We keep track of the number of cpus/rxqs used when the struct is allocated,
781 * in nr_ids. This will help not accessing out-of-bound memory.
782 *
783 * We keep track of the number of traffic classes used when the struct is
784 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
785 * not crossing its upper bound, as the original dev->num_tc can be updated in
786 * the meantime.
787 */
788struct xps_dev_maps {
789 struct rcu_head rcu;
790 unsigned int nr_ids;
791 s16 num_tc;
792 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
793};
794
795#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
796 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
797
798#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
799 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
800
801#endif /* CONFIG_XPS */
802
803#define TC_MAX_QUEUE 16
804#define TC_BITMASK 15
805/* HW offloaded queuing disciplines txq count and offset maps */
806struct netdev_tc_txq {
807 u16 count;
808 u16 offset;
809};
810
811#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
812/*
813 * This structure is to hold information about the device
814 * configured to run FCoE protocol stack.
815 */
816struct netdev_fcoe_hbainfo {
817 char manufacturer[64];
818 char serial_number[64];
819 char hardware_version[64];
820 char driver_version[64];
821 char optionrom_version[64];
822 char firmware_version[64];
823 char model[256];
824 char model_description[256];
825};
826#endif
827
828#define MAX_PHYS_ITEM_ID_LEN 32
829
830/* This structure holds a unique identifier to identify some
831 * physical item (port for example) used by a netdevice.
832 */
833struct netdev_phys_item_id {
834 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
835 unsigned char id_len;
836};
837
838static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
839 struct netdev_phys_item_id *b)
840{
841 return a->id_len == b->id_len &&
842 memcmp(a->id, b->id, a->id_len) == 0;
843}
844
845typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
846 struct sk_buff *skb,
847 struct net_device *sb_dev);
848
849enum net_device_path_type {
850 DEV_PATH_ETHERNET = 0,
851 DEV_PATH_VLAN,
852 DEV_PATH_BRIDGE,
853 DEV_PATH_PPPOE,
854 DEV_PATH_DSA,
855 DEV_PATH_MTK_WDMA,
856};
857
858struct net_device_path {
859 enum net_device_path_type type;
860 const struct net_device *dev;
861 union {
862 struct {
863 u16 id;
864 __be16 proto;
865 u8 h_dest[ETH_ALEN];
866 } encap;
867 struct {
868 enum {
869 DEV_PATH_BR_VLAN_KEEP,
870 DEV_PATH_BR_VLAN_TAG,
871 DEV_PATH_BR_VLAN_UNTAG,
872 DEV_PATH_BR_VLAN_UNTAG_HW,
873 } vlan_mode;
874 u16 vlan_id;
875 __be16 vlan_proto;
876 } bridge;
877 struct {
878 int port;
879 u16 proto;
880 } dsa;
881 struct {
882 u8 wdma_idx;
883 u8 queue;
884 u16 wcid;
885 u8 bss;
886 u8 amsdu;
887 } mtk_wdma;
888 };
889};
890
891#define NET_DEVICE_PATH_STACK_MAX 5
892#define NET_DEVICE_PATH_VLAN_MAX 2
893
894struct net_device_path_stack {
895 int num_paths;
896 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
897};
898
899struct net_device_path_ctx {
900 const struct net_device *dev;
901 u8 daddr[ETH_ALEN];
902
903 int num_vlans;
904 struct {
905 u16 id;
906 __be16 proto;
907 } vlan[NET_DEVICE_PATH_VLAN_MAX];
908};
909
910enum tc_setup_type {
911 TC_QUERY_CAPS,
912 TC_SETUP_QDISC_MQPRIO,
913 TC_SETUP_CLSU32,
914 TC_SETUP_CLSFLOWER,
915 TC_SETUP_CLSMATCHALL,
916 TC_SETUP_CLSBPF,
917 TC_SETUP_BLOCK,
918 TC_SETUP_QDISC_CBS,
919 TC_SETUP_QDISC_RED,
920 TC_SETUP_QDISC_PRIO,
921 TC_SETUP_QDISC_MQ,
922 TC_SETUP_QDISC_ETF,
923 TC_SETUP_ROOT_QDISC,
924 TC_SETUP_QDISC_GRED,
925 TC_SETUP_QDISC_TAPRIO,
926 TC_SETUP_FT,
927 TC_SETUP_QDISC_ETS,
928 TC_SETUP_QDISC_TBF,
929 TC_SETUP_QDISC_FIFO,
930 TC_SETUP_QDISC_HTB,
931 TC_SETUP_ACT,
932};
933
934/* These structures hold the attributes of bpf state that are being passed
935 * to the netdevice through the bpf op.
936 */
937enum bpf_netdev_command {
938 /* Set or clear a bpf program used in the earliest stages of packet
939 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
940 * is responsible for calling bpf_prog_put on any old progs that are
941 * stored. In case of error, the callee need not release the new prog
942 * reference, but on success it takes ownership and must bpf_prog_put
943 * when it is no longer used.
944 */
945 XDP_SETUP_PROG,
946 XDP_SETUP_PROG_HW,
947 /* BPF program for offload callbacks, invoked at program load time. */
948 BPF_OFFLOAD_MAP_ALLOC,
949 BPF_OFFLOAD_MAP_FREE,
950 XDP_SETUP_XSK_POOL,
951};
952
953struct bpf_prog_offload_ops;
954struct netlink_ext_ack;
955struct xdp_umem;
956struct xdp_dev_bulk_queue;
957struct bpf_xdp_link;
958
959enum bpf_xdp_mode {
960 XDP_MODE_SKB = 0,
961 XDP_MODE_DRV = 1,
962 XDP_MODE_HW = 2,
963 __MAX_XDP_MODE
964};
965
966struct bpf_xdp_entity {
967 struct bpf_prog *prog;
968 struct bpf_xdp_link *link;
969};
970
971struct netdev_bpf {
972 enum bpf_netdev_command command;
973 union {
974 /* XDP_SETUP_PROG */
975 struct {
976 u32 flags;
977 struct bpf_prog *prog;
978 struct netlink_ext_ack *extack;
979 };
980 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
981 struct {
982 struct bpf_offloaded_map *offmap;
983 };
984 /* XDP_SETUP_XSK_POOL */
985 struct {
986 struct xsk_buff_pool *pool;
987 u16 queue_id;
988 } xsk;
989 };
990};
991
992/* Flags for ndo_xsk_wakeup. */
993#define XDP_WAKEUP_RX (1 << 0)
994#define XDP_WAKEUP_TX (1 << 1)
995
996#ifdef CONFIG_XFRM_OFFLOAD
997struct xfrmdev_ops {
998 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
999 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1000 void (*xdo_dev_state_free) (struct xfrm_state *x);
1001 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1002 struct xfrm_state *x);
1003 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1004 void (*xdo_dev_state_update_stats) (struct xfrm_state *x);
1005 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1006 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1007 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1008};
1009#endif
1010
1011struct dev_ifalias {
1012 struct rcu_head rcuhead;
1013 char ifalias[];
1014};
1015
1016struct devlink;
1017struct tlsdev_ops;
1018
1019struct netdev_net_notifier {
1020 struct list_head list;
1021 struct notifier_block *nb;
1022};
1023
1024/*
1025 * This structure defines the management hooks for network devices.
1026 * The following hooks can be defined; unless noted otherwise, they are
1027 * optional and can be filled with a null pointer.
1028 *
1029 * int (*ndo_init)(struct net_device *dev);
1030 * This function is called once when a network device is registered.
1031 * The network device can use this for any late stage initialization
1032 * or semantic validation. It can fail with an error code which will
1033 * be propagated back to register_netdev.
1034 *
1035 * void (*ndo_uninit)(struct net_device *dev);
1036 * This function is called when device is unregistered or when registration
1037 * fails. It is not called if init fails.
1038 *
1039 * int (*ndo_open)(struct net_device *dev);
1040 * This function is called when a network device transitions to the up
1041 * state.
1042 *
1043 * int (*ndo_stop)(struct net_device *dev);
1044 * This function is called when a network device transitions to the down
1045 * state.
1046 *
1047 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1048 * struct net_device *dev);
1049 * Called when a packet needs to be transmitted.
1050 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1051 * the queue before that can happen; it's for obsolete devices and weird
1052 * corner cases, but the stack really does a non-trivial amount
1053 * of useless work if you return NETDEV_TX_BUSY.
1054 * Required; cannot be NULL.
1055 *
1056 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1057 * struct net_device *dev
1058 * netdev_features_t features);
1059 * Called by core transmit path to determine if device is capable of
1060 * performing offload operations on a given packet. This is to give
1061 * the device an opportunity to implement any restrictions that cannot
1062 * be otherwise expressed by feature flags. The check is called with
1063 * the set of features that the stack has calculated and it returns
1064 * those the driver believes to be appropriate.
1065 *
1066 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1067 * struct net_device *sb_dev);
1068 * Called to decide which queue to use when device supports multiple
1069 * transmit queues.
1070 *
1071 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1072 * This function is called to allow device receiver to make
1073 * changes to configuration when multicast or promiscuous is enabled.
1074 *
1075 * void (*ndo_set_rx_mode)(struct net_device *dev);
1076 * This function is called device changes address list filtering.
1077 * If driver handles unicast address filtering, it should set
1078 * IFF_UNICAST_FLT in its priv_flags.
1079 *
1080 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1081 * This function is called when the Media Access Control address
1082 * needs to be changed. If this interface is not defined, the
1083 * MAC address can not be changed.
1084 *
1085 * int (*ndo_validate_addr)(struct net_device *dev);
1086 * Test if Media Access Control address is valid for the device.
1087 *
1088 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1089 * Old-style ioctl entry point. This is used internally by the
1090 * appletalk and ieee802154 subsystems but is no longer called by
1091 * the device ioctl handler.
1092 *
1093 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1094 * Used by the bonding driver for its device specific ioctls:
1095 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1096 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1097 *
1098 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1099 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1100 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1101 *
1102 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1103 * Used to set network devices bus interface parameters. This interface
1104 * is retained for legacy reasons; new devices should use the bus
1105 * interface (PCI) for low level management.
1106 *
1107 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1108 * Called when a user wants to change the Maximum Transfer Unit
1109 * of a device.
1110 *
1111 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1112 * Callback used when the transmitter has not made any progress
1113 * for dev->watchdog ticks.
1114 *
1115 * void (*ndo_get_stats64)(struct net_device *dev,
1116 * struct rtnl_link_stats64 *storage);
1117 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1118 * Called when a user wants to get the network device usage
1119 * statistics. Drivers must do one of the following:
1120 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1121 * rtnl_link_stats64 structure passed by the caller.
1122 * 2. Define @ndo_get_stats to update a net_device_stats structure
1123 * (which should normally be dev->stats) and return a pointer to
1124 * it. The structure may be changed asynchronously only if each
1125 * field is written atomically.
1126 * 3. Update dev->stats asynchronously and atomically, and define
1127 * neither operation.
1128 *
1129 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1130 * Return true if this device supports offload stats of this attr_id.
1131 *
1132 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1133 * void *attr_data)
1134 * Get statistics for offload operations by attr_id. Write it into the
1135 * attr_data pointer.
1136 *
1137 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1138 * If device supports VLAN filtering this function is called when a
1139 * VLAN id is registered.
1140 *
1141 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1142 * If device supports VLAN filtering this function is called when a
1143 * VLAN id is unregistered.
1144 *
1145 * void (*ndo_poll_controller)(struct net_device *dev);
1146 *
1147 * SR-IOV management functions.
1148 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1149 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1150 * u8 qos, __be16 proto);
1151 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1152 * int max_tx_rate);
1153 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1154 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1155 * int (*ndo_get_vf_config)(struct net_device *dev,
1156 * int vf, struct ifla_vf_info *ivf);
1157 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1158 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1159 * struct nlattr *port[]);
1160 *
1161 * Enable or disable the VF ability to query its RSS Redirection Table and
1162 * Hash Key. This is needed since on some devices VF share this information
1163 * with PF and querying it may introduce a theoretical security risk.
1164 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1165 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1166 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1167 * void *type_data);
1168 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1169 * This is always called from the stack with the rtnl lock held and netif
1170 * tx queues stopped. This allows the netdevice to perform queue
1171 * management safely.
1172 *
1173 * Fiber Channel over Ethernet (FCoE) offload functions.
1174 * int (*ndo_fcoe_enable)(struct net_device *dev);
1175 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1176 * so the underlying device can perform whatever needed configuration or
1177 * initialization to support acceleration of FCoE traffic.
1178 *
1179 * int (*ndo_fcoe_disable)(struct net_device *dev);
1180 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1181 * so the underlying device can perform whatever needed clean-ups to
1182 * stop supporting acceleration of FCoE traffic.
1183 *
1184 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1185 * struct scatterlist *sgl, unsigned int sgc);
1186 * Called when the FCoE Initiator wants to initialize an I/O that
1187 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1188 * perform necessary setup and returns 1 to indicate the device is set up
1189 * successfully to perform DDP on this I/O, otherwise this returns 0.
1190 *
1191 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1192 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1193 * indicated by the FC exchange id 'xid', so the underlying device can
1194 * clean up and reuse resources for later DDP requests.
1195 *
1196 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1197 * struct scatterlist *sgl, unsigned int sgc);
1198 * Called when the FCoE Target wants to initialize an I/O that
1199 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1200 * perform necessary setup and returns 1 to indicate the device is set up
1201 * successfully to perform DDP on this I/O, otherwise this returns 0.
1202 *
1203 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1204 * struct netdev_fcoe_hbainfo *hbainfo);
1205 * Called when the FCoE Protocol stack wants information on the underlying
1206 * device. This information is utilized by the FCoE protocol stack to
1207 * register attributes with Fiber Channel management service as per the
1208 * FC-GS Fabric Device Management Information(FDMI) specification.
1209 *
1210 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1211 * Called when the underlying device wants to override default World Wide
1212 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1213 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1214 * protocol stack to use.
1215 *
1216 * RFS acceleration.
1217 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1218 * u16 rxq_index, u32 flow_id);
1219 * Set hardware filter for RFS. rxq_index is the target queue index;
1220 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1221 * Return the filter ID on success, or a negative error code.
1222 *
1223 * Slave management functions (for bridge, bonding, etc).
1224 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1225 * Called to make another netdev an underling.
1226 *
1227 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1228 * Called to release previously enslaved netdev.
1229 *
1230 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1231 * struct sk_buff *skb,
1232 * bool all_slaves);
1233 * Get the xmit slave of master device. If all_slaves is true, function
1234 * assume all the slaves can transmit.
1235 *
1236 * Feature/offload setting functions.
1237 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1238 * netdev_features_t features);
1239 * Adjusts the requested feature flags according to device-specific
1240 * constraints, and returns the resulting flags. Must not modify
1241 * the device state.
1242 *
1243 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1244 * Called to update device configuration to new features. Passed
1245 * feature set might be less than what was returned by ndo_fix_features()).
1246 * Must return >0 or -errno if it changed dev->features itself.
1247 *
1248 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1249 * struct net_device *dev,
1250 * const unsigned char *addr, u16 vid, u16 flags,
1251 * bool *notified, struct netlink_ext_ack *extack);
1252 * Adds an FDB entry to dev for addr.
1253 * Callee shall set *notified to true if it sent any appropriate
1254 * notification(s). Otherwise core will send a generic one.
1255 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1256 * struct net_device *dev,
1257 * const unsigned char *addr, u16 vid
1258 * bool *notified, struct netlink_ext_ack *extack);
1259 * Deletes the FDB entry from dev corresponding to addr.
1260 * Callee shall set *notified to true if it sent any appropriate
1261 * notification(s). Otherwise core will send a generic one.
1262 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1263 * struct netlink_ext_ack *extack);
1264 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1265 * struct net_device *dev, struct net_device *filter_dev,
1266 * int *idx)
1267 * Used to add FDB entries to dump requests. Implementers should add
1268 * entries to skb and update idx with the number of entries.
1269 *
1270 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1271 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1272 * Adds an MDB entry to dev.
1273 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1274 * struct netlink_ext_ack *extack);
1275 * Deletes the MDB entry from dev.
1276 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1277 * struct netlink_ext_ack *extack);
1278 * Bulk deletes MDB entries from dev.
1279 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1280 * struct netlink_callback *cb);
1281 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1282 * callback is used by core rtnetlink code.
1283 *
1284 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1285 * u16 flags, struct netlink_ext_ack *extack)
1286 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1287 * struct net_device *dev, u32 filter_mask,
1288 * int nlflags)
1289 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1290 * u16 flags);
1291 *
1292 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1293 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1294 * which do not represent real hardware may define this to allow their
1295 * userspace components to manage their virtual carrier state. Devices
1296 * that determine carrier state from physical hardware properties (eg
1297 * network cables) or protocol-dependent mechanisms (eg
1298 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1299 *
1300 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1301 * struct netdev_phys_item_id *ppid);
1302 * Called to get ID of physical port of this device. If driver does
1303 * not implement this, it is assumed that the hw is not able to have
1304 * multiple net devices on single physical port.
1305 *
1306 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1307 * struct netdev_phys_item_id *ppid)
1308 * Called to get the parent ID of the physical port of this device.
1309 *
1310 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1311 * struct net_device *dev)
1312 * Called by upper layer devices to accelerate switching or other
1313 * station functionality into hardware. 'pdev is the lowerdev
1314 * to use for the offload and 'dev' is the net device that will
1315 * back the offload. Returns a pointer to the private structure
1316 * the upper layer will maintain.
1317 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1318 * Called by upper layer device to delete the station created
1319 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1320 * the station and priv is the structure returned by the add
1321 * operation.
1322 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1323 * int queue_index, u32 maxrate);
1324 * Called when a user wants to set a max-rate limitation of specific
1325 * TX queue.
1326 * int (*ndo_get_iflink)(const struct net_device *dev);
1327 * Called to get the iflink value of this device.
1328 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1329 * This function is used to get egress tunnel information for given skb.
1330 * This is useful for retrieving outer tunnel header parameters while
1331 * sampling packet.
1332 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1333 * This function is used to specify the headroom that the skb must
1334 * consider when allocation skb during packet reception. Setting
1335 * appropriate rx headroom value allows avoiding skb head copy on
1336 * forward. Setting a negative value resets the rx headroom to the
1337 * default value.
1338 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1339 * This function is used to set or query state related to XDP on the
1340 * netdevice and manage BPF offload. See definition of
1341 * enum bpf_netdev_command for details.
1342 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1343 * u32 flags);
1344 * This function is used to submit @n XDP packets for transmit on a
1345 * netdevice. Returns number of frames successfully transmitted, frames
1346 * that got dropped are freed/returned via xdp_return_frame().
1347 * Returns negative number, means general error invoking ndo, meaning
1348 * no frames were xmit'ed and core-caller will free all frames.
1349 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1350 * struct xdp_buff *xdp);
1351 * Get the xmit slave of master device based on the xdp_buff.
1352 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1353 * This function is used to wake up the softirq, ksoftirqd or kthread
1354 * responsible for sending and/or receiving packets on a specific
1355 * queue id bound to an AF_XDP socket. The flags field specifies if
1356 * only RX, only Tx, or both should be woken up using the flags
1357 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1358 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1359 * int cmd);
1360 * Add, change, delete or get information on an IPv4 tunnel.
1361 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1362 * If a device is paired with a peer device, return the peer instance.
1363 * The caller must be under RCU read context.
1364 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1365 * Get the forwarding path to reach the real device from the HW destination address
1366 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1367 * const struct skb_shared_hwtstamps *hwtstamps,
1368 * bool cycles);
1369 * Get hardware timestamp based on normal/adjustable time or free running
1370 * cycle counter. This function is required if physical clock supports a
1371 * free running cycle counter.
1372 *
1373 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1374 * struct kernel_hwtstamp_config *kernel_config);
1375 * Get the currently configured hardware timestamping parameters for the
1376 * NIC device.
1377 *
1378 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1379 * struct kernel_hwtstamp_config *kernel_config,
1380 * struct netlink_ext_ack *extack);
1381 * Change the hardware timestamping parameters for NIC device.
1382 */
1383struct net_device_ops {
1384 int (*ndo_init)(struct net_device *dev);
1385 void (*ndo_uninit)(struct net_device *dev);
1386 int (*ndo_open)(struct net_device *dev);
1387 int (*ndo_stop)(struct net_device *dev);
1388 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1389 struct net_device *dev);
1390 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1391 struct net_device *dev,
1392 netdev_features_t features);
1393 u16 (*ndo_select_queue)(struct net_device *dev,
1394 struct sk_buff *skb,
1395 struct net_device *sb_dev);
1396 void (*ndo_change_rx_flags)(struct net_device *dev,
1397 int flags);
1398 void (*ndo_set_rx_mode)(struct net_device *dev);
1399 int (*ndo_set_mac_address)(struct net_device *dev,
1400 void *addr);
1401 int (*ndo_validate_addr)(struct net_device *dev);
1402 int (*ndo_do_ioctl)(struct net_device *dev,
1403 struct ifreq *ifr, int cmd);
1404 int (*ndo_eth_ioctl)(struct net_device *dev,
1405 struct ifreq *ifr, int cmd);
1406 int (*ndo_siocbond)(struct net_device *dev,
1407 struct ifreq *ifr, int cmd);
1408 int (*ndo_siocwandev)(struct net_device *dev,
1409 struct if_settings *ifs);
1410 int (*ndo_siocdevprivate)(struct net_device *dev,
1411 struct ifreq *ifr,
1412 void __user *data, int cmd);
1413 int (*ndo_set_config)(struct net_device *dev,
1414 struct ifmap *map);
1415 int (*ndo_change_mtu)(struct net_device *dev,
1416 int new_mtu);
1417 int (*ndo_neigh_setup)(struct net_device *dev,
1418 struct neigh_parms *);
1419 void (*ndo_tx_timeout) (struct net_device *dev,
1420 unsigned int txqueue);
1421
1422 void (*ndo_get_stats64)(struct net_device *dev,
1423 struct rtnl_link_stats64 *storage);
1424 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1425 int (*ndo_get_offload_stats)(int attr_id,
1426 const struct net_device *dev,
1427 void *attr_data);
1428 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1429
1430 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1431 __be16 proto, u16 vid);
1432 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1433 __be16 proto, u16 vid);
1434#ifdef CONFIG_NET_POLL_CONTROLLER
1435 void (*ndo_poll_controller)(struct net_device *dev);
1436 int (*ndo_netpoll_setup)(struct net_device *dev);
1437 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1438#endif
1439 int (*ndo_set_vf_mac)(struct net_device *dev,
1440 int queue, u8 *mac);
1441 int (*ndo_set_vf_vlan)(struct net_device *dev,
1442 int queue, u16 vlan,
1443 u8 qos, __be16 proto);
1444 int (*ndo_set_vf_rate)(struct net_device *dev,
1445 int vf, int min_tx_rate,
1446 int max_tx_rate);
1447 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1448 int vf, bool setting);
1449 int (*ndo_set_vf_trust)(struct net_device *dev,
1450 int vf, bool setting);
1451 int (*ndo_get_vf_config)(struct net_device *dev,
1452 int vf,
1453 struct ifla_vf_info *ivf);
1454 int (*ndo_set_vf_link_state)(struct net_device *dev,
1455 int vf, int link_state);
1456 int (*ndo_get_vf_stats)(struct net_device *dev,
1457 int vf,
1458 struct ifla_vf_stats
1459 *vf_stats);
1460 int (*ndo_set_vf_port)(struct net_device *dev,
1461 int vf,
1462 struct nlattr *port[]);
1463 int (*ndo_get_vf_port)(struct net_device *dev,
1464 int vf, struct sk_buff *skb);
1465 int (*ndo_get_vf_guid)(struct net_device *dev,
1466 int vf,
1467 struct ifla_vf_guid *node_guid,
1468 struct ifla_vf_guid *port_guid);
1469 int (*ndo_set_vf_guid)(struct net_device *dev,
1470 int vf, u64 guid,
1471 int guid_type);
1472 int (*ndo_set_vf_rss_query_en)(
1473 struct net_device *dev,
1474 int vf, bool setting);
1475 int (*ndo_setup_tc)(struct net_device *dev,
1476 enum tc_setup_type type,
1477 void *type_data);
1478#if IS_ENABLED(CONFIG_FCOE)
1479 int (*ndo_fcoe_enable)(struct net_device *dev);
1480 int (*ndo_fcoe_disable)(struct net_device *dev);
1481 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1482 u16 xid,
1483 struct scatterlist *sgl,
1484 unsigned int sgc);
1485 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1486 u16 xid);
1487 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1488 u16 xid,
1489 struct scatterlist *sgl,
1490 unsigned int sgc);
1491 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1492 struct netdev_fcoe_hbainfo *hbainfo);
1493#endif
1494
1495#if IS_ENABLED(CONFIG_LIBFCOE)
1496#define NETDEV_FCOE_WWNN 0
1497#define NETDEV_FCOE_WWPN 1
1498 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1499 u64 *wwn, int type);
1500#endif
1501
1502#ifdef CONFIG_RFS_ACCEL
1503 int (*ndo_rx_flow_steer)(struct net_device *dev,
1504 const struct sk_buff *skb,
1505 u16 rxq_index,
1506 u32 flow_id);
1507#endif
1508 int (*ndo_add_slave)(struct net_device *dev,
1509 struct net_device *slave_dev,
1510 struct netlink_ext_ack *extack);
1511 int (*ndo_del_slave)(struct net_device *dev,
1512 struct net_device *slave_dev);
1513 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1514 struct sk_buff *skb,
1515 bool all_slaves);
1516 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1517 struct sock *sk);
1518 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1519 netdev_features_t features);
1520 int (*ndo_set_features)(struct net_device *dev,
1521 netdev_features_t features);
1522 int (*ndo_neigh_construct)(struct net_device *dev,
1523 struct neighbour *n);
1524 void (*ndo_neigh_destroy)(struct net_device *dev,
1525 struct neighbour *n);
1526
1527 int (*ndo_fdb_add)(struct ndmsg *ndm,
1528 struct nlattr *tb[],
1529 struct net_device *dev,
1530 const unsigned char *addr,
1531 u16 vid,
1532 u16 flags,
1533 bool *notified,
1534 struct netlink_ext_ack *extack);
1535 int (*ndo_fdb_del)(struct ndmsg *ndm,
1536 struct nlattr *tb[],
1537 struct net_device *dev,
1538 const unsigned char *addr,
1539 u16 vid,
1540 bool *notified,
1541 struct netlink_ext_ack *extack);
1542 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1543 struct net_device *dev,
1544 struct netlink_ext_ack *extack);
1545 int (*ndo_fdb_dump)(struct sk_buff *skb,
1546 struct netlink_callback *cb,
1547 struct net_device *dev,
1548 struct net_device *filter_dev,
1549 int *idx);
1550 int (*ndo_fdb_get)(struct sk_buff *skb,
1551 struct nlattr *tb[],
1552 struct net_device *dev,
1553 const unsigned char *addr,
1554 u16 vid, u32 portid, u32 seq,
1555 struct netlink_ext_ack *extack);
1556 int (*ndo_mdb_add)(struct net_device *dev,
1557 struct nlattr *tb[],
1558 u16 nlmsg_flags,
1559 struct netlink_ext_ack *extack);
1560 int (*ndo_mdb_del)(struct net_device *dev,
1561 struct nlattr *tb[],
1562 struct netlink_ext_ack *extack);
1563 int (*ndo_mdb_del_bulk)(struct net_device *dev,
1564 struct nlattr *tb[],
1565 struct netlink_ext_ack *extack);
1566 int (*ndo_mdb_dump)(struct net_device *dev,
1567 struct sk_buff *skb,
1568 struct netlink_callback *cb);
1569 int (*ndo_mdb_get)(struct net_device *dev,
1570 struct nlattr *tb[], u32 portid,
1571 u32 seq,
1572 struct netlink_ext_ack *extack);
1573 int (*ndo_bridge_setlink)(struct net_device *dev,
1574 struct nlmsghdr *nlh,
1575 u16 flags,
1576 struct netlink_ext_ack *extack);
1577 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1578 u32 pid, u32 seq,
1579 struct net_device *dev,
1580 u32 filter_mask,
1581 int nlflags);
1582 int (*ndo_bridge_dellink)(struct net_device *dev,
1583 struct nlmsghdr *nlh,
1584 u16 flags);
1585 int (*ndo_change_carrier)(struct net_device *dev,
1586 bool new_carrier);
1587 int (*ndo_get_phys_port_id)(struct net_device *dev,
1588 struct netdev_phys_item_id *ppid);
1589 int (*ndo_get_port_parent_id)(struct net_device *dev,
1590 struct netdev_phys_item_id *ppid);
1591 int (*ndo_get_phys_port_name)(struct net_device *dev,
1592 char *name, size_t len);
1593 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1594 struct net_device *dev);
1595 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1596 void *priv);
1597
1598 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1599 int queue_index,
1600 u32 maxrate);
1601 int (*ndo_get_iflink)(const struct net_device *dev);
1602 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1603 struct sk_buff *skb);
1604 void (*ndo_set_rx_headroom)(struct net_device *dev,
1605 int needed_headroom);
1606 int (*ndo_bpf)(struct net_device *dev,
1607 struct netdev_bpf *bpf);
1608 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1609 struct xdp_frame **xdp,
1610 u32 flags);
1611 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1612 struct xdp_buff *xdp);
1613 int (*ndo_xsk_wakeup)(struct net_device *dev,
1614 u32 queue_id, u32 flags);
1615 int (*ndo_tunnel_ctl)(struct net_device *dev,
1616 struct ip_tunnel_parm_kern *p,
1617 int cmd);
1618 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1619 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1620 struct net_device_path *path);
1621 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1622 const struct skb_shared_hwtstamps *hwtstamps,
1623 bool cycles);
1624 int (*ndo_hwtstamp_get)(struct net_device *dev,
1625 struct kernel_hwtstamp_config *kernel_config);
1626 int (*ndo_hwtstamp_set)(struct net_device *dev,
1627 struct kernel_hwtstamp_config *kernel_config,
1628 struct netlink_ext_ack *extack);
1629
1630#if IS_ENABLED(CONFIG_NET_SHAPER)
1631 /**
1632 * @net_shaper_ops: Device shaping offload operations
1633 * see include/net/net_shapers.h
1634 */
1635 const struct net_shaper_ops *net_shaper_ops;
1636#endif
1637};
1638
1639/**
1640 * enum netdev_priv_flags - &struct net_device priv_flags
1641 *
1642 * These are the &struct net_device, they are only set internally
1643 * by drivers and used in the kernel. These flags are invisible to
1644 * userspace; this means that the order of these flags can change
1645 * during any kernel release.
1646 *
1647 * You should add bitfield booleans after either net_device::priv_flags
1648 * (hotpath) or ::threaded (slowpath) instead of extending these flags.
1649 *
1650 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1651 * @IFF_EBRIDGE: Ethernet bridging device
1652 * @IFF_BONDING: bonding master or slave
1653 * @IFF_ISATAP: ISATAP interface (RFC4214)
1654 * @IFF_WAN_HDLC: WAN HDLC device
1655 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1656 * release skb->dst
1657 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1658 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1659 * @IFF_MACVLAN_PORT: device used as macvlan port
1660 * @IFF_BRIDGE_PORT: device used as bridge port
1661 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1662 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1663 * @IFF_UNICAST_FLT: Supports unicast filtering
1664 * @IFF_TEAM_PORT: device used as team port
1665 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1666 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1667 * change when it's running
1668 * @IFF_MACVLAN: Macvlan device
1669 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1670 * underlying stacked devices
1671 * @IFF_L3MDEV_MASTER: device is an L3 master device
1672 * @IFF_NO_QUEUE: device can run without qdisc attached
1673 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1674 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1675 * @IFF_TEAM: device is a team device
1676 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1677 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1678 * entity (i.e. the master device for bridged veth)
1679 * @IFF_MACSEC: device is a MACsec device
1680 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1681 * @IFF_FAILOVER: device is a failover master device
1682 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1683 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1684 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1685 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1686 * skb_headlen(skb) == 0 (data starts from frag0)
1687 */
1688enum netdev_priv_flags {
1689 IFF_802_1Q_VLAN = 1<<0,
1690 IFF_EBRIDGE = 1<<1,
1691 IFF_BONDING = 1<<2,
1692 IFF_ISATAP = 1<<3,
1693 IFF_WAN_HDLC = 1<<4,
1694 IFF_XMIT_DST_RELEASE = 1<<5,
1695 IFF_DONT_BRIDGE = 1<<6,
1696 IFF_DISABLE_NETPOLL = 1<<7,
1697 IFF_MACVLAN_PORT = 1<<8,
1698 IFF_BRIDGE_PORT = 1<<9,
1699 IFF_OVS_DATAPATH = 1<<10,
1700 IFF_TX_SKB_SHARING = 1<<11,
1701 IFF_UNICAST_FLT = 1<<12,
1702 IFF_TEAM_PORT = 1<<13,
1703 IFF_SUPP_NOFCS = 1<<14,
1704 IFF_LIVE_ADDR_CHANGE = 1<<15,
1705 IFF_MACVLAN = 1<<16,
1706 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1707 IFF_L3MDEV_MASTER = 1<<18,
1708 IFF_NO_QUEUE = 1<<19,
1709 IFF_OPENVSWITCH = 1<<20,
1710 IFF_L3MDEV_SLAVE = 1<<21,
1711 IFF_TEAM = 1<<22,
1712 IFF_RXFH_CONFIGURED = 1<<23,
1713 IFF_PHONY_HEADROOM = 1<<24,
1714 IFF_MACSEC = 1<<25,
1715 IFF_NO_RX_HANDLER = 1<<26,
1716 IFF_FAILOVER = 1<<27,
1717 IFF_FAILOVER_SLAVE = 1<<28,
1718 IFF_L3MDEV_RX_HANDLER = 1<<29,
1719 IFF_NO_ADDRCONF = BIT_ULL(30),
1720 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1721};
1722
1723/* Specifies the type of the struct net_device::ml_priv pointer */
1724enum netdev_ml_priv_type {
1725 ML_PRIV_NONE,
1726 ML_PRIV_CAN,
1727};
1728
1729enum netdev_stat_type {
1730 NETDEV_PCPU_STAT_NONE,
1731 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1732 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1733 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1734};
1735
1736enum netdev_reg_state {
1737 NETREG_UNINITIALIZED = 0,
1738 NETREG_REGISTERED, /* completed register_netdevice */
1739 NETREG_UNREGISTERING, /* called unregister_netdevice */
1740 NETREG_UNREGISTERED, /* completed unregister todo */
1741 NETREG_RELEASED, /* called free_netdev */
1742 NETREG_DUMMY, /* dummy device for NAPI poll */
1743};
1744
1745/**
1746 * struct net_device - The DEVICE structure.
1747 *
1748 * Actually, this whole structure is a big mistake. It mixes I/O
1749 * data with strictly "high-level" data, and it has to know about
1750 * almost every data structure used in the INET module.
1751 *
1752 * @priv_flags: flags invisible to userspace defined as bits, see
1753 * enum netdev_priv_flags for the definitions
1754 * @lltx: device supports lockless Tx. Deprecated for real HW
1755 * drivers. Mainly used by logical interfaces, such as
1756 * bonding and tunnels
1757 *
1758 * @name: This is the first field of the "visible" part of this structure
1759 * (i.e. as seen by users in the "Space.c" file). It is the name
1760 * of the interface.
1761 *
1762 * @name_node: Name hashlist node
1763 * @ifalias: SNMP alias
1764 * @mem_end: Shared memory end
1765 * @mem_start: Shared memory start
1766 * @base_addr: Device I/O address
1767 * @irq: Device IRQ number
1768 *
1769 * @state: Generic network queuing layer state, see netdev_state_t
1770 * @dev_list: The global list of network devices
1771 * @napi_list: List entry used for polling NAPI devices
1772 * @unreg_list: List entry when we are unregistering the
1773 * device; see the function unregister_netdev
1774 * @close_list: List entry used when we are closing the device
1775 * @ptype_all: Device-specific packet handlers for all protocols
1776 * @ptype_specific: Device-specific, protocol-specific packet handlers
1777 *
1778 * @adj_list: Directly linked devices, like slaves for bonding
1779 * @features: Currently active device features
1780 * @hw_features: User-changeable features
1781 *
1782 * @wanted_features: User-requested features
1783 * @vlan_features: Mask of features inheritable by VLAN devices
1784 *
1785 * @hw_enc_features: Mask of features inherited by encapsulating devices
1786 * This field indicates what encapsulation
1787 * offloads the hardware is capable of doing,
1788 * and drivers will need to set them appropriately.
1789 *
1790 * @mpls_features: Mask of features inheritable by MPLS
1791 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1792 *
1793 * @ifindex: interface index
1794 * @group: The group the device belongs to
1795 *
1796 * @stats: Statistics struct, which was left as a legacy, use
1797 * rtnl_link_stats64 instead
1798 *
1799 * @core_stats: core networking counters,
1800 * do not use this in drivers
1801 * @carrier_up_count: Number of times the carrier has been up
1802 * @carrier_down_count: Number of times the carrier has been down
1803 *
1804 * @wireless_handlers: List of functions to handle Wireless Extensions,
1805 * instead of ioctl,
1806 * see <net/iw_handler.h> for details.
1807 *
1808 * @netdev_ops: Includes several pointers to callbacks,
1809 * if one wants to override the ndo_*() functions
1810 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1811 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks.
1812 * @ethtool_ops: Management operations
1813 * @l3mdev_ops: Layer 3 master device operations
1814 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1815 * discovery handling. Necessary for e.g. 6LoWPAN.
1816 * @xfrmdev_ops: Transformation offload operations
1817 * @tlsdev_ops: Transport Layer Security offload operations
1818 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1819 * of Layer 2 headers.
1820 *
1821 * @flags: Interface flags (a la BSD)
1822 * @xdp_features: XDP capability supported by the device
1823 * @gflags: Global flags ( kept as legacy )
1824 * @priv_len: Size of the ->priv flexible array
1825 * @priv: Flexible array containing private data
1826 * @operstate: RFC2863 operstate
1827 * @link_mode: Mapping policy to operstate
1828 * @if_port: Selectable AUI, TP, ...
1829 * @dma: DMA channel
1830 * @mtu: Interface MTU value
1831 * @min_mtu: Interface Minimum MTU value
1832 * @max_mtu: Interface Maximum MTU value
1833 * @type: Interface hardware type
1834 * @hard_header_len: Maximum hardware header length.
1835 * @min_header_len: Minimum hardware header length
1836 *
1837 * @needed_headroom: Extra headroom the hardware may need, but not in all
1838 * cases can this be guaranteed
1839 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1840 * cases can this be guaranteed. Some cases also use
1841 * LL_MAX_HEADER instead to allocate the skb
1842 *
1843 * interface address info:
1844 *
1845 * @perm_addr: Permanent hw address
1846 * @addr_assign_type: Hw address assignment type
1847 * @addr_len: Hardware address length
1848 * @upper_level: Maximum depth level of upper devices.
1849 * @lower_level: Maximum depth level of lower devices.
1850 * @neigh_priv_len: Used in neigh_alloc()
1851 * @dev_id: Used to differentiate devices that share
1852 * the same link layer address
1853 * @dev_port: Used to differentiate devices that share
1854 * the same function
1855 * @addr_list_lock: XXX: need comments on this one
1856 * @name_assign_type: network interface name assignment type
1857 * @uc_promisc: Counter that indicates promiscuous mode
1858 * has been enabled due to the need to listen to
1859 * additional unicast addresses in a device that
1860 * does not implement ndo_set_rx_mode()
1861 * @uc: unicast mac addresses
1862 * @mc: multicast mac addresses
1863 * @dev_addrs: list of device hw addresses
1864 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1865 * @promiscuity: Number of times the NIC is told to work in
1866 * promiscuous mode; if it becomes 0 the NIC will
1867 * exit promiscuous mode
1868 * @allmulti: Counter, enables or disables allmulticast mode
1869 *
1870 * @vlan_info: VLAN info
1871 * @dsa_ptr: dsa specific data
1872 * @tipc_ptr: TIPC specific data
1873 * @atalk_ptr: AppleTalk link
1874 * @ip_ptr: IPv4 specific data
1875 * @ip6_ptr: IPv6 specific data
1876 * @ax25_ptr: AX.25 specific data
1877 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1878 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1879 * device struct
1880 * @mpls_ptr: mpls_dev struct pointer
1881 * @mctp_ptr: MCTP specific data
1882 *
1883 * @dev_addr: Hw address (before bcast,
1884 * because most packets are unicast)
1885 *
1886 * @_rx: Array of RX queues
1887 * @num_rx_queues: Number of RX queues
1888 * allocated at register_netdev() time
1889 * @real_num_rx_queues: Number of RX queues currently active in device
1890 * @xdp_prog: XDP sockets filter program pointer
1891 *
1892 * @rx_handler: handler for received packets
1893 * @rx_handler_data: XXX: need comments on this one
1894 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1895 * @ingress_queue: XXX: need comments on this one
1896 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1897 * @broadcast: hw bcast address
1898 *
1899 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1900 * indexed by RX queue number. Assigned by driver.
1901 * This must only be set if the ndo_rx_flow_steer
1902 * operation is defined
1903 * @index_hlist: Device index hash chain
1904 *
1905 * @_tx: Array of TX queues
1906 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1907 * @real_num_tx_queues: Number of TX queues currently active in device
1908 * @qdisc: Root qdisc from userspace point of view
1909 * @tx_queue_len: Max frames per queue allowed
1910 * @tx_global_lock: XXX: need comments on this one
1911 * @xdp_bulkq: XDP device bulk queue
1912 * @xps_maps: all CPUs/RXQs maps for XPS device
1913 *
1914 * @xps_maps: XXX: need comments on this one
1915 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1916 * @nf_hooks_egress: netfilter hooks executed for egress packets
1917 * @qdisc_hash: qdisc hash table
1918 * @watchdog_timeo: Represents the timeout that is used by
1919 * the watchdog (see dev_watchdog())
1920 * @watchdog_timer: List of timers
1921 *
1922 * @proto_down_reason: reason a netdev interface is held down
1923 * @pcpu_refcnt: Number of references to this device
1924 * @dev_refcnt: Number of references to this device
1925 * @refcnt_tracker: Tracker directory for tracked references to this device
1926 * @todo_list: Delayed register/unregister
1927 * @link_watch_list: XXX: need comments on this one
1928 *
1929 * @reg_state: Register/unregister state machine
1930 * @dismantle: Device is going to be freed
1931 * @rtnl_link_state: This enum represents the phases of creating
1932 * a new link
1933 *
1934 * @needs_free_netdev: Should unregister perform free_netdev?
1935 * @priv_destructor: Called from unregister
1936 * @npinfo: XXX: need comments on this one
1937 * @nd_net: Network namespace this network device is inside
1938 *
1939 * @ml_priv: Mid-layer private
1940 * @ml_priv_type: Mid-layer private type
1941 *
1942 * @pcpu_stat_type: Type of device statistics which the core should
1943 * allocate/free: none, lstats, tstats, dstats. none
1944 * means the driver is handling statistics allocation/
1945 * freeing internally.
1946 * @lstats: Loopback statistics: packets, bytes
1947 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes
1948 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes
1949 *
1950 * @garp_port: GARP
1951 * @mrp_port: MRP
1952 *
1953 * @dm_private: Drop monitor private
1954 *
1955 * @dev: Class/net/name entry
1956 * @sysfs_groups: Space for optional device, statistics and wireless
1957 * sysfs groups
1958 *
1959 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1960 * @rtnl_link_ops: Rtnl_link_ops
1961 * @stat_ops: Optional ops for queue-aware statistics
1962 * @queue_mgmt_ops: Optional ops for queue management
1963 *
1964 * @gso_max_size: Maximum size of generic segmentation offload
1965 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1966 * @gso_max_segs: Maximum number of segments that can be passed to the
1967 * NIC for GSO
1968 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1969 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
1970 * for IPv4.
1971 *
1972 * @dcbnl_ops: Data Center Bridging netlink ops
1973 * @num_tc: Number of traffic classes in the net device
1974 * @tc_to_txq: XXX: need comments on this one
1975 * @prio_tc_map: XXX: need comments on this one
1976 *
1977 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1978 *
1979 * @priomap: XXX: need comments on this one
1980 * @link_topo: Physical link topology tracking attached PHYs
1981 * @phydev: Physical device may attach itself
1982 * for hardware timestamping
1983 * @sfp_bus: attached &struct sfp_bus structure.
1984 *
1985 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1986 *
1987 * @proto_down: protocol port state information can be sent to the
1988 * switch driver and used to set the phys state of the
1989 * switch port.
1990 *
1991 * @threaded: napi threaded mode is enabled
1992 *
1993 * @see_all_hwtstamp_requests: device wants to see calls to
1994 * ndo_hwtstamp_set() for all timestamp requests
1995 * regardless of source, even if those aren't
1996 * HWTSTAMP_SOURCE_NETDEV
1997 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN
1998 * @netns_local: interface can't change network namespaces
1999 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes
2000 *
2001 * @net_notifier_list: List of per-net netdev notifier block
2002 * that follow this device when it is moved
2003 * to another network namespace.
2004 *
2005 * @macsec_ops: MACsec offloading ops
2006 *
2007 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2008 * offload capabilities of the device
2009 * @udp_tunnel_nic: UDP tunnel offload state
2010 * @ethtool: ethtool related state
2011 * @xdp_state: stores info on attached XDP BPF programs
2012 *
2013 * @nested_level: Used as a parameter of spin_lock_nested() of
2014 * dev->addr_list_lock.
2015 * @unlink_list: As netif_addr_lock() can be called recursively,
2016 * keep a list of interfaces to be deleted.
2017 * @gro_max_size: Maximum size of aggregated packet in generic
2018 * receive offload (GRO)
2019 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2020 * receive offload (GRO), for IPv4.
2021 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2022 * zero copy driver
2023 *
2024 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2025 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2026 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2027 * @dev_registered_tracker: tracker for reference held while
2028 * registered
2029 * @offload_xstats_l3: L3 HW stats for this netdevice.
2030 *
2031 * @devlink_port: Pointer to related devlink port structure.
2032 * Assigned by a driver before netdev registration using
2033 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2034 * during the time netdevice is registered.
2035 *
2036 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2037 * where the clock is recovered.
2038 *
2039 * @max_pacing_offload_horizon: max EDT offload horizon in nsec.
2040 * @napi_config: An array of napi_config structures containing per-NAPI
2041 * settings.
2042 * @gro_flush_timeout: timeout for GRO layer in NAPI
2043 * @napi_defer_hard_irqs: If not zero, provides a counter that would
2044 * allow to avoid NIC hard IRQ, on busy queues.
2045 *
2046 * @neighbours: List heads pointing to this device's neighbours'
2047 * dev_list, one per address-family.
2048 *
2049 * FIXME: cleanup struct net_device such that network protocol info
2050 * moves out.
2051 */
2052
2053struct net_device {
2054 /* Cacheline organization can be found documented in
2055 * Documentation/networking/net_cachelines/net_device.rst.
2056 * Please update the document when adding new fields.
2057 */
2058
2059 /* TX read-mostly hotpath */
2060 __cacheline_group_begin(net_device_read_tx);
2061 struct_group(priv_flags_fast,
2062 unsigned long priv_flags:32;
2063 unsigned long lltx:1;
2064 );
2065 const struct net_device_ops *netdev_ops;
2066 const struct header_ops *header_ops;
2067 struct netdev_queue *_tx;
2068 netdev_features_t gso_partial_features;
2069 unsigned int real_num_tx_queues;
2070 unsigned int gso_max_size;
2071 unsigned int gso_ipv4_max_size;
2072 u16 gso_max_segs;
2073 s16 num_tc;
2074 /* Note : dev->mtu is often read without holding a lock.
2075 * Writers usually hold RTNL.
2076 * It is recommended to use READ_ONCE() to annotate the reads,
2077 * and to use WRITE_ONCE() to annotate the writes.
2078 */
2079 unsigned int mtu;
2080 unsigned short needed_headroom;
2081 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2082#ifdef CONFIG_XPS
2083 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2084#endif
2085#ifdef CONFIG_NETFILTER_EGRESS
2086 struct nf_hook_entries __rcu *nf_hooks_egress;
2087#endif
2088#ifdef CONFIG_NET_XGRESS
2089 struct bpf_mprog_entry __rcu *tcx_egress;
2090#endif
2091 __cacheline_group_end(net_device_read_tx);
2092
2093 /* TXRX read-mostly hotpath */
2094 __cacheline_group_begin(net_device_read_txrx);
2095 union {
2096 struct pcpu_lstats __percpu *lstats;
2097 struct pcpu_sw_netstats __percpu *tstats;
2098 struct pcpu_dstats __percpu *dstats;
2099 };
2100 unsigned long state;
2101 unsigned int flags;
2102 unsigned short hard_header_len;
2103 netdev_features_t features;
2104 struct inet6_dev __rcu *ip6_ptr;
2105 __cacheline_group_end(net_device_read_txrx);
2106
2107 /* RX read-mostly hotpath */
2108 __cacheline_group_begin(net_device_read_rx);
2109 struct bpf_prog __rcu *xdp_prog;
2110 struct list_head ptype_specific;
2111 int ifindex;
2112 unsigned int real_num_rx_queues;
2113 struct netdev_rx_queue *_rx;
2114 unsigned int gro_max_size;
2115 unsigned int gro_ipv4_max_size;
2116 rx_handler_func_t __rcu *rx_handler;
2117 void __rcu *rx_handler_data;
2118 possible_net_t nd_net;
2119#ifdef CONFIG_NETPOLL
2120 struct netpoll_info __rcu *npinfo;
2121#endif
2122#ifdef CONFIG_NET_XGRESS
2123 struct bpf_mprog_entry __rcu *tcx_ingress;
2124#endif
2125 __cacheline_group_end(net_device_read_rx);
2126
2127 char name[IFNAMSIZ];
2128 struct netdev_name_node *name_node;
2129 struct dev_ifalias __rcu *ifalias;
2130 /*
2131 * I/O specific fields
2132 * FIXME: Merge these and struct ifmap into one
2133 */
2134 unsigned long mem_end;
2135 unsigned long mem_start;
2136 unsigned long base_addr;
2137
2138 /*
2139 * Some hardware also needs these fields (state,dev_list,
2140 * napi_list,unreg_list,close_list) but they are not
2141 * part of the usual set specified in Space.c.
2142 */
2143
2144
2145 struct list_head dev_list;
2146 struct list_head napi_list;
2147 struct list_head unreg_list;
2148 struct list_head close_list;
2149 struct list_head ptype_all;
2150
2151 struct {
2152 struct list_head upper;
2153 struct list_head lower;
2154 } adj_list;
2155
2156 /* Read-mostly cache-line for fast-path access */
2157 xdp_features_t xdp_features;
2158 const struct xdp_metadata_ops *xdp_metadata_ops;
2159 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2160 unsigned short gflags;
2161
2162 unsigned short needed_tailroom;
2163
2164 netdev_features_t hw_features;
2165 netdev_features_t wanted_features;
2166 netdev_features_t vlan_features;
2167 netdev_features_t hw_enc_features;
2168 netdev_features_t mpls_features;
2169
2170 unsigned int min_mtu;
2171 unsigned int max_mtu;
2172 unsigned short type;
2173 unsigned char min_header_len;
2174 unsigned char name_assign_type;
2175
2176 int group;
2177
2178 struct net_device_stats stats; /* not used by modern drivers */
2179
2180 struct net_device_core_stats __percpu *core_stats;
2181
2182 /* Stats to monitor link on/off, flapping */
2183 atomic_t carrier_up_count;
2184 atomic_t carrier_down_count;
2185
2186#ifdef CONFIG_WIRELESS_EXT
2187 const struct iw_handler_def *wireless_handlers;
2188#endif
2189 const struct ethtool_ops *ethtool_ops;
2190#ifdef CONFIG_NET_L3_MASTER_DEV
2191 const struct l3mdev_ops *l3mdev_ops;
2192#endif
2193#if IS_ENABLED(CONFIG_IPV6)
2194 const struct ndisc_ops *ndisc_ops;
2195#endif
2196
2197#ifdef CONFIG_XFRM_OFFLOAD
2198 const struct xfrmdev_ops *xfrmdev_ops;
2199#endif
2200
2201#if IS_ENABLED(CONFIG_TLS_DEVICE)
2202 const struct tlsdev_ops *tlsdev_ops;
2203#endif
2204
2205 unsigned int operstate;
2206 unsigned char link_mode;
2207
2208 unsigned char if_port;
2209 unsigned char dma;
2210
2211 /* Interface address info. */
2212 unsigned char perm_addr[MAX_ADDR_LEN];
2213 unsigned char addr_assign_type;
2214 unsigned char addr_len;
2215 unsigned char upper_level;
2216 unsigned char lower_level;
2217
2218 unsigned short neigh_priv_len;
2219 unsigned short dev_id;
2220 unsigned short dev_port;
2221 int irq;
2222 u32 priv_len;
2223
2224 spinlock_t addr_list_lock;
2225
2226 struct netdev_hw_addr_list uc;
2227 struct netdev_hw_addr_list mc;
2228 struct netdev_hw_addr_list dev_addrs;
2229
2230#ifdef CONFIG_SYSFS
2231 struct kset *queues_kset;
2232#endif
2233#ifdef CONFIG_LOCKDEP
2234 struct list_head unlink_list;
2235#endif
2236 unsigned int promiscuity;
2237 unsigned int allmulti;
2238 bool uc_promisc;
2239#ifdef CONFIG_LOCKDEP
2240 unsigned char nested_level;
2241#endif
2242
2243
2244 /* Protocol-specific pointers */
2245 struct in_device __rcu *ip_ptr;
2246 /** @fib_nh_head: nexthops associated with this netdev */
2247 struct hlist_head fib_nh_head;
2248
2249#if IS_ENABLED(CONFIG_VLAN_8021Q)
2250 struct vlan_info __rcu *vlan_info;
2251#endif
2252#if IS_ENABLED(CONFIG_NET_DSA)
2253 struct dsa_port *dsa_ptr;
2254#endif
2255#if IS_ENABLED(CONFIG_TIPC)
2256 struct tipc_bearer __rcu *tipc_ptr;
2257#endif
2258#if IS_ENABLED(CONFIG_ATALK)
2259 void *atalk_ptr;
2260#endif
2261#if IS_ENABLED(CONFIG_AX25)
2262 struct ax25_dev __rcu *ax25_ptr;
2263#endif
2264#if IS_ENABLED(CONFIG_CFG80211)
2265 struct wireless_dev *ieee80211_ptr;
2266#endif
2267#if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2268 struct wpan_dev *ieee802154_ptr;
2269#endif
2270#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2271 struct mpls_dev __rcu *mpls_ptr;
2272#endif
2273#if IS_ENABLED(CONFIG_MCTP)
2274 struct mctp_dev __rcu *mctp_ptr;
2275#endif
2276
2277/*
2278 * Cache lines mostly used on receive path (including eth_type_trans())
2279 */
2280 /* Interface address info used in eth_type_trans() */
2281 const unsigned char *dev_addr;
2282
2283 unsigned int num_rx_queues;
2284#define GRO_LEGACY_MAX_SIZE 65536u
2285/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2286 * and shinfo->gso_segs is a 16bit field.
2287 */
2288#define GRO_MAX_SIZE (8 * 65535u)
2289 unsigned int xdp_zc_max_segs;
2290 struct netdev_queue __rcu *ingress_queue;
2291#ifdef CONFIG_NETFILTER_INGRESS
2292 struct nf_hook_entries __rcu *nf_hooks_ingress;
2293#endif
2294
2295 unsigned char broadcast[MAX_ADDR_LEN];
2296#ifdef CONFIG_RFS_ACCEL
2297 struct cpu_rmap *rx_cpu_rmap;
2298#endif
2299 struct hlist_node index_hlist;
2300
2301/*
2302 * Cache lines mostly used on transmit path
2303 */
2304 unsigned int num_tx_queues;
2305 struct Qdisc __rcu *qdisc;
2306 unsigned int tx_queue_len;
2307 spinlock_t tx_global_lock;
2308
2309 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2310
2311#ifdef CONFIG_NET_SCHED
2312 DECLARE_HASHTABLE (qdisc_hash, 4);
2313#endif
2314 /* These may be needed for future network-power-down code. */
2315 struct timer_list watchdog_timer;
2316 int watchdog_timeo;
2317
2318 u32 proto_down_reason;
2319
2320 struct list_head todo_list;
2321
2322#ifdef CONFIG_PCPU_DEV_REFCNT
2323 int __percpu *pcpu_refcnt;
2324#else
2325 refcount_t dev_refcnt;
2326#endif
2327 struct ref_tracker_dir refcnt_tracker;
2328
2329 struct list_head link_watch_list;
2330
2331 u8 reg_state;
2332
2333 bool dismantle;
2334
2335 enum {
2336 RTNL_LINK_INITIALIZED,
2337 RTNL_LINK_INITIALIZING,
2338 } rtnl_link_state:16;
2339
2340 bool needs_free_netdev;
2341 void (*priv_destructor)(struct net_device *dev);
2342
2343 /* mid-layer private */
2344 void *ml_priv;
2345 enum netdev_ml_priv_type ml_priv_type;
2346
2347 enum netdev_stat_type pcpu_stat_type:8;
2348
2349#if IS_ENABLED(CONFIG_GARP)
2350 struct garp_port __rcu *garp_port;
2351#endif
2352#if IS_ENABLED(CONFIG_MRP)
2353 struct mrp_port __rcu *mrp_port;
2354#endif
2355#if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2356 struct dm_hw_stat_delta __rcu *dm_private;
2357#endif
2358 struct device dev;
2359 const struct attribute_group *sysfs_groups[4];
2360 const struct attribute_group *sysfs_rx_queue_group;
2361
2362 const struct rtnl_link_ops *rtnl_link_ops;
2363
2364 const struct netdev_stat_ops *stat_ops;
2365
2366 const struct netdev_queue_mgmt_ops *queue_mgmt_ops;
2367
2368 /* for setting kernel sock attribute on TCP connection setup */
2369#define GSO_MAX_SEGS 65535u
2370#define GSO_LEGACY_MAX_SIZE 65536u
2371/* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2372 * and shinfo->gso_segs is a 16bit field.
2373 */
2374#define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2375
2376#define TSO_LEGACY_MAX_SIZE 65536
2377#define TSO_MAX_SIZE UINT_MAX
2378 unsigned int tso_max_size;
2379#define TSO_MAX_SEGS U16_MAX
2380 u16 tso_max_segs;
2381
2382#ifdef CONFIG_DCB
2383 const struct dcbnl_rtnl_ops *dcbnl_ops;
2384#endif
2385 u8 prio_tc_map[TC_BITMASK + 1];
2386
2387#if IS_ENABLED(CONFIG_FCOE)
2388 unsigned int fcoe_ddp_xid;
2389#endif
2390#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2391 struct netprio_map __rcu *priomap;
2392#endif
2393 struct phy_link_topology *link_topo;
2394 struct phy_device *phydev;
2395 struct sfp_bus *sfp_bus;
2396 struct lock_class_key *qdisc_tx_busylock;
2397 bool proto_down;
2398 bool threaded;
2399
2400 /* priv_flags_slow, ungrouped to save space */
2401 unsigned long see_all_hwtstamp_requests:1;
2402 unsigned long change_proto_down:1;
2403 unsigned long netns_local:1;
2404 unsigned long fcoe_mtu:1;
2405
2406 struct list_head net_notifier_list;
2407
2408#if IS_ENABLED(CONFIG_MACSEC)
2409 /* MACsec management functions */
2410 const struct macsec_ops *macsec_ops;
2411#endif
2412 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2413 struct udp_tunnel_nic *udp_tunnel_nic;
2414
2415 struct ethtool_netdev_state *ethtool;
2416
2417 /* protected by rtnl_lock */
2418 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2419
2420 u8 dev_addr_shadow[MAX_ADDR_LEN];
2421 netdevice_tracker linkwatch_dev_tracker;
2422 netdevice_tracker watchdog_dev_tracker;
2423 netdevice_tracker dev_registered_tracker;
2424 struct rtnl_hw_stats64 *offload_xstats_l3;
2425
2426 struct devlink_port *devlink_port;
2427
2428#if IS_ENABLED(CONFIG_DPLL)
2429 struct dpll_pin __rcu *dpll_pin;
2430#endif
2431#if IS_ENABLED(CONFIG_PAGE_POOL)
2432 /** @page_pools: page pools created for this netdevice */
2433 struct hlist_head page_pools;
2434#endif
2435
2436 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */
2437 struct dim_irq_moder *irq_moder;
2438
2439 u64 max_pacing_offload_horizon;
2440 struct napi_config *napi_config;
2441 unsigned long gro_flush_timeout;
2442 u32 napi_defer_hard_irqs;
2443
2444 /**
2445 * @lock: protects @net_shaper_hierarchy, feel free to use for other
2446 * netdev-scope protection. Ordering: take after rtnl_lock.
2447 */
2448 struct mutex lock;
2449
2450#if IS_ENABLED(CONFIG_NET_SHAPER)
2451 /**
2452 * @net_shaper_hierarchy: data tracking the current shaper status
2453 * see include/net/net_shapers.h
2454 */
2455 struct net_shaper_hierarchy *net_shaper_hierarchy;
2456#endif
2457
2458 struct hlist_head neighbours[NEIGH_NR_TABLES];
2459
2460 u8 priv[] ____cacheline_aligned
2461 __counted_by(priv_len);
2462} ____cacheline_aligned;
2463#define to_net_dev(d) container_of(d, struct net_device, dev)
2464
2465/*
2466 * Driver should use this to assign devlink port instance to a netdevice
2467 * before it registers the netdevice. Therefore devlink_port is static
2468 * during the netdev lifetime after it is registered.
2469 */
2470#define SET_NETDEV_DEVLINK_PORT(dev, port) \
2471({ \
2472 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2473 ((dev)->devlink_port = (port)); \
2474})
2475
2476static inline bool netif_elide_gro(const struct net_device *dev)
2477{
2478 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2479 return true;
2480 return false;
2481}
2482
2483#define NETDEV_ALIGN 32
2484
2485static inline
2486int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2487{
2488 return dev->prio_tc_map[prio & TC_BITMASK];
2489}
2490
2491static inline
2492int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2493{
2494 if (tc >= dev->num_tc)
2495 return -EINVAL;
2496
2497 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2498 return 0;
2499}
2500
2501int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2502void netdev_reset_tc(struct net_device *dev);
2503int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2504int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2505
2506static inline
2507int netdev_get_num_tc(struct net_device *dev)
2508{
2509 return dev->num_tc;
2510}
2511
2512static inline void net_prefetch(void *p)
2513{
2514 prefetch(p);
2515#if L1_CACHE_BYTES < 128
2516 prefetch((u8 *)p + L1_CACHE_BYTES);
2517#endif
2518}
2519
2520static inline void net_prefetchw(void *p)
2521{
2522 prefetchw(p);
2523#if L1_CACHE_BYTES < 128
2524 prefetchw((u8 *)p + L1_CACHE_BYTES);
2525#endif
2526}
2527
2528void netdev_unbind_sb_channel(struct net_device *dev,
2529 struct net_device *sb_dev);
2530int netdev_bind_sb_channel_queue(struct net_device *dev,
2531 struct net_device *sb_dev,
2532 u8 tc, u16 count, u16 offset);
2533int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2534static inline int netdev_get_sb_channel(struct net_device *dev)
2535{
2536 return max_t(int, -dev->num_tc, 0);
2537}
2538
2539static inline
2540struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2541 unsigned int index)
2542{
2543 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2544 return &dev->_tx[index];
2545}
2546
2547static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2548 const struct sk_buff *skb)
2549{
2550 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2551}
2552
2553static inline void netdev_for_each_tx_queue(struct net_device *dev,
2554 void (*f)(struct net_device *,
2555 struct netdev_queue *,
2556 void *),
2557 void *arg)
2558{
2559 unsigned int i;
2560
2561 for (i = 0; i < dev->num_tx_queues; i++)
2562 f(dev, &dev->_tx[i], arg);
2563}
2564
2565#define netdev_lockdep_set_classes(dev) \
2566{ \
2567 static struct lock_class_key qdisc_tx_busylock_key; \
2568 static struct lock_class_key qdisc_xmit_lock_key; \
2569 static struct lock_class_key dev_addr_list_lock_key; \
2570 unsigned int i; \
2571 \
2572 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2573 lockdep_set_class(&(dev)->addr_list_lock, \
2574 &dev_addr_list_lock_key); \
2575 for (i = 0; i < (dev)->num_tx_queues; i++) \
2576 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2577 &qdisc_xmit_lock_key); \
2578}
2579
2580u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2581 struct net_device *sb_dev);
2582struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2583 struct sk_buff *skb,
2584 struct net_device *sb_dev);
2585
2586/* returns the headroom that the master device needs to take in account
2587 * when forwarding to this dev
2588 */
2589static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2590{
2591 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2592}
2593
2594static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2595{
2596 if (dev->netdev_ops->ndo_set_rx_headroom)
2597 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2598}
2599
2600/* set the device rx headroom to the dev's default */
2601static inline void netdev_reset_rx_headroom(struct net_device *dev)
2602{
2603 netdev_set_rx_headroom(dev, -1);
2604}
2605
2606static inline void *netdev_get_ml_priv(struct net_device *dev,
2607 enum netdev_ml_priv_type type)
2608{
2609 if (dev->ml_priv_type != type)
2610 return NULL;
2611
2612 return dev->ml_priv;
2613}
2614
2615static inline void netdev_set_ml_priv(struct net_device *dev,
2616 void *ml_priv,
2617 enum netdev_ml_priv_type type)
2618{
2619 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2620 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2621 dev->ml_priv_type, type);
2622 WARN(!dev->ml_priv_type && dev->ml_priv,
2623 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2624
2625 dev->ml_priv = ml_priv;
2626 dev->ml_priv_type = type;
2627}
2628
2629/*
2630 * Net namespace inlines
2631 */
2632static inline
2633struct net *dev_net(const struct net_device *dev)
2634{
2635 return read_pnet(&dev->nd_net);
2636}
2637
2638static inline
2639struct net *dev_net_rcu(const struct net_device *dev)
2640{
2641 return read_pnet_rcu(&dev->nd_net);
2642}
2643
2644static inline
2645void dev_net_set(struct net_device *dev, struct net *net)
2646{
2647 write_pnet(&dev->nd_net, net);
2648}
2649
2650/**
2651 * netdev_priv - access network device private data
2652 * @dev: network device
2653 *
2654 * Get network device private data
2655 */
2656static inline void *netdev_priv(const struct net_device *dev)
2657{
2658 return (void *)dev->priv;
2659}
2660
2661/* Set the sysfs physical device reference for the network logical device
2662 * if set prior to registration will cause a symlink during initialization.
2663 */
2664#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2665
2666/* Set the sysfs device type for the network logical device to allow
2667 * fine-grained identification of different network device types. For
2668 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2669 */
2670#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2671
2672void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2673 enum netdev_queue_type type,
2674 struct napi_struct *napi);
2675
2676static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2677{
2678 napi->irq = irq;
2679}
2680
2681/* Default NAPI poll() weight
2682 * Device drivers are strongly advised to not use bigger value
2683 */
2684#define NAPI_POLL_WEIGHT 64
2685
2686void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2687 int (*poll)(struct napi_struct *, int), int weight);
2688
2689/**
2690 * netif_napi_add() - initialize a NAPI context
2691 * @dev: network device
2692 * @napi: NAPI context
2693 * @poll: polling function
2694 *
2695 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2696 * *any* of the other NAPI-related functions.
2697 */
2698static inline void
2699netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2700 int (*poll)(struct napi_struct *, int))
2701{
2702 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2703}
2704
2705static inline void
2706netif_napi_add_tx_weight(struct net_device *dev,
2707 struct napi_struct *napi,
2708 int (*poll)(struct napi_struct *, int),
2709 int weight)
2710{
2711 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2712 netif_napi_add_weight(dev, napi, poll, weight);
2713}
2714
2715/**
2716 * netif_napi_add_config - initialize a NAPI context with persistent config
2717 * @dev: network device
2718 * @napi: NAPI context
2719 * @poll: polling function
2720 * @index: the NAPI index
2721 */
2722static inline void
2723netif_napi_add_config(struct net_device *dev, struct napi_struct *napi,
2724 int (*poll)(struct napi_struct *, int), int index)
2725{
2726 napi->index = index;
2727 napi->config = &dev->napi_config[index];
2728 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2729}
2730
2731/**
2732 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2733 * @dev: network device
2734 * @napi: NAPI context
2735 * @poll: polling function
2736 *
2737 * This variant of netif_napi_add() should be used from drivers using NAPI
2738 * to exclusively poll a TX queue.
2739 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2740 */
2741static inline void netif_napi_add_tx(struct net_device *dev,
2742 struct napi_struct *napi,
2743 int (*poll)(struct napi_struct *, int))
2744{
2745 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2746}
2747
2748/**
2749 * __netif_napi_del - remove a NAPI context
2750 * @napi: NAPI context
2751 *
2752 * Warning: caller must observe RCU grace period before freeing memory
2753 * containing @napi. Drivers might want to call this helper to combine
2754 * all the needed RCU grace periods into a single one.
2755 */
2756void __netif_napi_del(struct napi_struct *napi);
2757
2758/**
2759 * netif_napi_del - remove a NAPI context
2760 * @napi: NAPI context
2761 *
2762 * netif_napi_del() removes a NAPI context from the network device NAPI list
2763 */
2764static inline void netif_napi_del(struct napi_struct *napi)
2765{
2766 __netif_napi_del(napi);
2767 synchronize_net();
2768}
2769
2770struct packet_type {
2771 __be16 type; /* This is really htons(ether_type). */
2772 bool ignore_outgoing;
2773 struct net_device *dev; /* NULL is wildcarded here */
2774 netdevice_tracker dev_tracker;
2775 int (*func) (struct sk_buff *,
2776 struct net_device *,
2777 struct packet_type *,
2778 struct net_device *);
2779 void (*list_func) (struct list_head *,
2780 struct packet_type *,
2781 struct net_device *);
2782 bool (*id_match)(struct packet_type *ptype,
2783 struct sock *sk);
2784 struct net *af_packet_net;
2785 void *af_packet_priv;
2786 struct list_head list;
2787};
2788
2789struct offload_callbacks {
2790 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2791 netdev_features_t features);
2792 struct sk_buff *(*gro_receive)(struct list_head *head,
2793 struct sk_buff *skb);
2794 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2795};
2796
2797struct packet_offload {
2798 __be16 type; /* This is really htons(ether_type). */
2799 u16 priority;
2800 struct offload_callbacks callbacks;
2801 struct list_head list;
2802};
2803
2804/* often modified stats are per-CPU, other are shared (netdev->stats) */
2805struct pcpu_sw_netstats {
2806 u64_stats_t rx_packets;
2807 u64_stats_t rx_bytes;
2808 u64_stats_t tx_packets;
2809 u64_stats_t tx_bytes;
2810 struct u64_stats_sync syncp;
2811} __aligned(4 * sizeof(u64));
2812
2813struct pcpu_dstats {
2814 u64_stats_t rx_packets;
2815 u64_stats_t rx_bytes;
2816 u64_stats_t rx_drops;
2817 u64_stats_t tx_packets;
2818 u64_stats_t tx_bytes;
2819 u64_stats_t tx_drops;
2820 struct u64_stats_sync syncp;
2821} __aligned(8 * sizeof(u64));
2822
2823struct pcpu_lstats {
2824 u64_stats_t packets;
2825 u64_stats_t bytes;
2826 struct u64_stats_sync syncp;
2827} __aligned(2 * sizeof(u64));
2828
2829void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2830
2831static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2832{
2833 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2834
2835 u64_stats_update_begin(&tstats->syncp);
2836 u64_stats_add(&tstats->rx_bytes, len);
2837 u64_stats_inc(&tstats->rx_packets);
2838 u64_stats_update_end(&tstats->syncp);
2839}
2840
2841static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2842 unsigned int packets,
2843 unsigned int len)
2844{
2845 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2846
2847 u64_stats_update_begin(&tstats->syncp);
2848 u64_stats_add(&tstats->tx_bytes, len);
2849 u64_stats_add(&tstats->tx_packets, packets);
2850 u64_stats_update_end(&tstats->syncp);
2851}
2852
2853static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2854{
2855 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2856
2857 u64_stats_update_begin(&lstats->syncp);
2858 u64_stats_add(&lstats->bytes, len);
2859 u64_stats_inc(&lstats->packets);
2860 u64_stats_update_end(&lstats->syncp);
2861}
2862
2863#define __netdev_alloc_pcpu_stats(type, gfp) \
2864({ \
2865 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2866 if (pcpu_stats) { \
2867 int __cpu; \
2868 for_each_possible_cpu(__cpu) { \
2869 typeof(type) *stat; \
2870 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2871 u64_stats_init(&stat->syncp); \
2872 } \
2873 } \
2874 pcpu_stats; \
2875})
2876
2877#define netdev_alloc_pcpu_stats(type) \
2878 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2879
2880#define devm_netdev_alloc_pcpu_stats(dev, type) \
2881({ \
2882 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2883 if (pcpu_stats) { \
2884 int __cpu; \
2885 for_each_possible_cpu(__cpu) { \
2886 typeof(type) *stat; \
2887 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2888 u64_stats_init(&stat->syncp); \
2889 } \
2890 } \
2891 pcpu_stats; \
2892})
2893
2894enum netdev_lag_tx_type {
2895 NETDEV_LAG_TX_TYPE_UNKNOWN,
2896 NETDEV_LAG_TX_TYPE_RANDOM,
2897 NETDEV_LAG_TX_TYPE_BROADCAST,
2898 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2899 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2900 NETDEV_LAG_TX_TYPE_HASH,
2901};
2902
2903enum netdev_lag_hash {
2904 NETDEV_LAG_HASH_NONE,
2905 NETDEV_LAG_HASH_L2,
2906 NETDEV_LAG_HASH_L34,
2907 NETDEV_LAG_HASH_L23,
2908 NETDEV_LAG_HASH_E23,
2909 NETDEV_LAG_HASH_E34,
2910 NETDEV_LAG_HASH_VLAN_SRCMAC,
2911 NETDEV_LAG_HASH_UNKNOWN,
2912};
2913
2914struct netdev_lag_upper_info {
2915 enum netdev_lag_tx_type tx_type;
2916 enum netdev_lag_hash hash_type;
2917};
2918
2919struct netdev_lag_lower_state_info {
2920 u8 link_up : 1,
2921 tx_enabled : 1;
2922};
2923
2924#include <linux/notifier.h>
2925
2926/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2927 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2928 * adding new types.
2929 */
2930enum netdev_cmd {
2931 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2932 NETDEV_DOWN,
2933 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2934 detected a hardware crash and restarted
2935 - we can use this eg to kick tcp sessions
2936 once done */
2937 NETDEV_CHANGE, /* Notify device state change */
2938 NETDEV_REGISTER,
2939 NETDEV_UNREGISTER,
2940 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2941 NETDEV_CHANGEADDR, /* notify after the address change */
2942 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2943 NETDEV_GOING_DOWN,
2944 NETDEV_CHANGENAME,
2945 NETDEV_FEAT_CHANGE,
2946 NETDEV_BONDING_FAILOVER,
2947 NETDEV_PRE_UP,
2948 NETDEV_PRE_TYPE_CHANGE,
2949 NETDEV_POST_TYPE_CHANGE,
2950 NETDEV_POST_INIT,
2951 NETDEV_PRE_UNINIT,
2952 NETDEV_RELEASE,
2953 NETDEV_NOTIFY_PEERS,
2954 NETDEV_JOIN,
2955 NETDEV_CHANGEUPPER,
2956 NETDEV_RESEND_IGMP,
2957 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2958 NETDEV_CHANGEINFODATA,
2959 NETDEV_BONDING_INFO,
2960 NETDEV_PRECHANGEUPPER,
2961 NETDEV_CHANGELOWERSTATE,
2962 NETDEV_UDP_TUNNEL_PUSH_INFO,
2963 NETDEV_UDP_TUNNEL_DROP_INFO,
2964 NETDEV_CHANGE_TX_QUEUE_LEN,
2965 NETDEV_CVLAN_FILTER_PUSH_INFO,
2966 NETDEV_CVLAN_FILTER_DROP_INFO,
2967 NETDEV_SVLAN_FILTER_PUSH_INFO,
2968 NETDEV_SVLAN_FILTER_DROP_INFO,
2969 NETDEV_OFFLOAD_XSTATS_ENABLE,
2970 NETDEV_OFFLOAD_XSTATS_DISABLE,
2971 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2972 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2973 NETDEV_XDP_FEAT_CHANGE,
2974};
2975const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2976
2977int register_netdevice_notifier(struct notifier_block *nb);
2978int unregister_netdevice_notifier(struct notifier_block *nb);
2979int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2980int unregister_netdevice_notifier_net(struct net *net,
2981 struct notifier_block *nb);
2982int register_netdevice_notifier_dev_net(struct net_device *dev,
2983 struct notifier_block *nb,
2984 struct netdev_net_notifier *nn);
2985int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2986 struct notifier_block *nb,
2987 struct netdev_net_notifier *nn);
2988
2989struct netdev_notifier_info {
2990 struct net_device *dev;
2991 struct netlink_ext_ack *extack;
2992};
2993
2994struct netdev_notifier_info_ext {
2995 struct netdev_notifier_info info; /* must be first */
2996 union {
2997 u32 mtu;
2998 } ext;
2999};
3000
3001struct netdev_notifier_change_info {
3002 struct netdev_notifier_info info; /* must be first */
3003 unsigned int flags_changed;
3004};
3005
3006struct netdev_notifier_changeupper_info {
3007 struct netdev_notifier_info info; /* must be first */
3008 struct net_device *upper_dev; /* new upper dev */
3009 bool master; /* is upper dev master */
3010 bool linking; /* is the notification for link or unlink */
3011 void *upper_info; /* upper dev info */
3012};
3013
3014struct netdev_notifier_changelowerstate_info {
3015 struct netdev_notifier_info info; /* must be first */
3016 void *lower_state_info; /* is lower dev state */
3017};
3018
3019struct netdev_notifier_pre_changeaddr_info {
3020 struct netdev_notifier_info info; /* must be first */
3021 const unsigned char *dev_addr;
3022};
3023
3024enum netdev_offload_xstats_type {
3025 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
3026};
3027
3028struct netdev_notifier_offload_xstats_info {
3029 struct netdev_notifier_info info; /* must be first */
3030 enum netdev_offload_xstats_type type;
3031
3032 union {
3033 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
3034 struct netdev_notifier_offload_xstats_rd *report_delta;
3035 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
3036 struct netdev_notifier_offload_xstats_ru *report_used;
3037 };
3038};
3039
3040int netdev_offload_xstats_enable(struct net_device *dev,
3041 enum netdev_offload_xstats_type type,
3042 struct netlink_ext_ack *extack);
3043int netdev_offload_xstats_disable(struct net_device *dev,
3044 enum netdev_offload_xstats_type type);
3045bool netdev_offload_xstats_enabled(const struct net_device *dev,
3046 enum netdev_offload_xstats_type type);
3047int netdev_offload_xstats_get(struct net_device *dev,
3048 enum netdev_offload_xstats_type type,
3049 struct rtnl_hw_stats64 *stats, bool *used,
3050 struct netlink_ext_ack *extack);
3051void
3052netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3053 const struct rtnl_hw_stats64 *stats);
3054void
3055netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3056void netdev_offload_xstats_push_delta(struct net_device *dev,
3057 enum netdev_offload_xstats_type type,
3058 const struct rtnl_hw_stats64 *stats);
3059
3060static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3061 struct net_device *dev)
3062{
3063 info->dev = dev;
3064 info->extack = NULL;
3065}
3066
3067static inline struct net_device *
3068netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3069{
3070 return info->dev;
3071}
3072
3073static inline struct netlink_ext_ack *
3074netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3075{
3076 return info->extack;
3077}
3078
3079int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3080int call_netdevice_notifiers_info(unsigned long val,
3081 struct netdev_notifier_info *info);
3082
3083#define for_each_netdev(net, d) \
3084 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3085#define for_each_netdev_reverse(net, d) \
3086 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3087#define for_each_netdev_rcu(net, d) \
3088 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3089#define for_each_netdev_safe(net, d, n) \
3090 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3091#define for_each_netdev_continue(net, d) \
3092 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3093#define for_each_netdev_continue_reverse(net, d) \
3094 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3095 dev_list)
3096#define for_each_netdev_continue_rcu(net, d) \
3097 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3098#define for_each_netdev_in_bond_rcu(bond, slave) \
3099 for_each_netdev_rcu(&init_net, slave) \
3100 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3101#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3102
3103#define for_each_netdev_dump(net, d, ifindex) \
3104 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \
3105 ULONG_MAX, XA_PRESENT)); ifindex++)
3106
3107static inline struct net_device *next_net_device(struct net_device *dev)
3108{
3109 struct list_head *lh;
3110 struct net *net;
3111
3112 net = dev_net(dev);
3113 lh = dev->dev_list.next;
3114 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3115}
3116
3117static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3118{
3119 struct list_head *lh;
3120 struct net *net;
3121
3122 net = dev_net(dev);
3123 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3124 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3125}
3126
3127static inline struct net_device *first_net_device(struct net *net)
3128{
3129 return list_empty(&net->dev_base_head) ? NULL :
3130 net_device_entry(net->dev_base_head.next);
3131}
3132
3133static inline struct net_device *first_net_device_rcu(struct net *net)
3134{
3135 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3136
3137 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3138}
3139
3140int netdev_boot_setup_check(struct net_device *dev);
3141struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type,
3142 const char *hwaddr);
3143struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3144 const char *hwaddr);
3145struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3146void dev_add_pack(struct packet_type *pt);
3147void dev_remove_pack(struct packet_type *pt);
3148void __dev_remove_pack(struct packet_type *pt);
3149void dev_add_offload(struct packet_offload *po);
3150void dev_remove_offload(struct packet_offload *po);
3151
3152int dev_get_iflink(const struct net_device *dev);
3153int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3154int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3155 struct net_device_path_stack *stack);
3156struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3157 unsigned short mask);
3158struct net_device *dev_get_by_name(struct net *net, const char *name);
3159struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3160struct net_device *__dev_get_by_name(struct net *net, const char *name);
3161bool netdev_name_in_use(struct net *net, const char *name);
3162int dev_alloc_name(struct net_device *dev, const char *name);
3163int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3164void dev_close(struct net_device *dev);
3165void dev_close_many(struct list_head *head, bool unlink);
3166void dev_disable_lro(struct net_device *dev);
3167int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3168u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3169 struct net_device *sb_dev);
3170
3171int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3172int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3173
3174static inline int dev_queue_xmit(struct sk_buff *skb)
3175{
3176 return __dev_queue_xmit(skb, NULL);
3177}
3178
3179static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3180 struct net_device *sb_dev)
3181{
3182 return __dev_queue_xmit(skb, sb_dev);
3183}
3184
3185static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3186{
3187 int ret;
3188
3189 ret = __dev_direct_xmit(skb, queue_id);
3190 if (!dev_xmit_complete(ret))
3191 kfree_skb(skb);
3192 return ret;
3193}
3194
3195int register_netdevice(struct net_device *dev);
3196void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3197void unregister_netdevice_many(struct list_head *head);
3198static inline void unregister_netdevice(struct net_device *dev)
3199{
3200 unregister_netdevice_queue(dev, NULL);
3201}
3202
3203int netdev_refcnt_read(const struct net_device *dev);
3204void free_netdev(struct net_device *dev);
3205void init_dummy_netdev(struct net_device *dev);
3206
3207struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3208 struct sk_buff *skb,
3209 bool all_slaves);
3210struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3211 struct sock *sk);
3212struct net_device *dev_get_by_index(struct net *net, int ifindex);
3213struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3214struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3215 netdevice_tracker *tracker, gfp_t gfp);
3216struct net_device *netdev_get_by_name(struct net *net, const char *name,
3217 netdevice_tracker *tracker, gfp_t gfp);
3218struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3219struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3220void netdev_copy_name(struct net_device *dev, char *name);
3221
3222static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3223 unsigned short type,
3224 const void *daddr, const void *saddr,
3225 unsigned int len)
3226{
3227 if (!dev->header_ops || !dev->header_ops->create)
3228 return 0;
3229
3230 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3231}
3232
3233static inline int dev_parse_header(const struct sk_buff *skb,
3234 unsigned char *haddr)
3235{
3236 const struct net_device *dev = skb->dev;
3237
3238 if (!dev->header_ops || !dev->header_ops->parse)
3239 return 0;
3240 return dev->header_ops->parse(skb, haddr);
3241}
3242
3243static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3244{
3245 const struct net_device *dev = skb->dev;
3246
3247 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3248 return 0;
3249 return dev->header_ops->parse_protocol(skb);
3250}
3251
3252/* ll_header must have at least hard_header_len allocated */
3253static inline bool dev_validate_header(const struct net_device *dev,
3254 char *ll_header, int len)
3255{
3256 if (likely(len >= dev->hard_header_len))
3257 return true;
3258 if (len < dev->min_header_len)
3259 return false;
3260
3261 if (capable(CAP_SYS_RAWIO)) {
3262 memset(ll_header + len, 0, dev->hard_header_len - len);
3263 return true;
3264 }
3265
3266 if (dev->header_ops && dev->header_ops->validate)
3267 return dev->header_ops->validate(ll_header, len);
3268
3269 return false;
3270}
3271
3272static inline bool dev_has_header(const struct net_device *dev)
3273{
3274 return dev->header_ops && dev->header_ops->create;
3275}
3276
3277/*
3278 * Incoming packets are placed on per-CPU queues
3279 */
3280struct softnet_data {
3281 struct list_head poll_list;
3282 struct sk_buff_head process_queue;
3283 local_lock_t process_queue_bh_lock;
3284
3285 /* stats */
3286 unsigned int processed;
3287 unsigned int time_squeeze;
3288#ifdef CONFIG_RPS
3289 struct softnet_data *rps_ipi_list;
3290#endif
3291
3292 unsigned int received_rps;
3293 bool in_net_rx_action;
3294 bool in_napi_threaded_poll;
3295
3296#ifdef CONFIG_NET_FLOW_LIMIT
3297 struct sd_flow_limit __rcu *flow_limit;
3298#endif
3299 struct Qdisc *output_queue;
3300 struct Qdisc **output_queue_tailp;
3301 struct sk_buff *completion_queue;
3302#ifdef CONFIG_XFRM_OFFLOAD
3303 struct sk_buff_head xfrm_backlog;
3304#endif
3305 /* written and read only by owning cpu: */
3306 struct netdev_xmit xmit;
3307#ifdef CONFIG_RPS
3308 /* input_queue_head should be written by cpu owning this struct,
3309 * and only read by other cpus. Worth using a cache line.
3310 */
3311 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3312
3313 /* Elements below can be accessed between CPUs for RPS/RFS */
3314 call_single_data_t csd ____cacheline_aligned_in_smp;
3315 struct softnet_data *rps_ipi_next;
3316 unsigned int cpu;
3317 unsigned int input_queue_tail;
3318#endif
3319 struct sk_buff_head input_pkt_queue;
3320 struct napi_struct backlog;
3321
3322 atomic_t dropped ____cacheline_aligned_in_smp;
3323
3324 /* Another possibly contended cache line */
3325 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3326 int defer_count;
3327 int defer_ipi_scheduled;
3328 struct sk_buff *defer_list;
3329 call_single_data_t defer_csd;
3330};
3331
3332DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3333
3334#ifndef CONFIG_PREEMPT_RT
3335static inline int dev_recursion_level(void)
3336{
3337 return this_cpu_read(softnet_data.xmit.recursion);
3338}
3339#else
3340static inline int dev_recursion_level(void)
3341{
3342 return current->net_xmit.recursion;
3343}
3344
3345#endif
3346
3347void __netif_schedule(struct Qdisc *q);
3348void netif_schedule_queue(struct netdev_queue *txq);
3349
3350static inline void netif_tx_schedule_all(struct net_device *dev)
3351{
3352 unsigned int i;
3353
3354 for (i = 0; i < dev->num_tx_queues; i++)
3355 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3356}
3357
3358static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3359{
3360 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3361}
3362
3363/**
3364 * netif_start_queue - allow transmit
3365 * @dev: network device
3366 *
3367 * Allow upper layers to call the device hard_start_xmit routine.
3368 */
3369static inline void netif_start_queue(struct net_device *dev)
3370{
3371 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3372}
3373
3374static inline void netif_tx_start_all_queues(struct net_device *dev)
3375{
3376 unsigned int i;
3377
3378 for (i = 0; i < dev->num_tx_queues; i++) {
3379 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3380 netif_tx_start_queue(txq);
3381 }
3382}
3383
3384void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3385
3386/**
3387 * netif_wake_queue - restart transmit
3388 * @dev: network device
3389 *
3390 * Allow upper layers to call the device hard_start_xmit routine.
3391 * Used for flow control when transmit resources are available.
3392 */
3393static inline void netif_wake_queue(struct net_device *dev)
3394{
3395 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3396}
3397
3398static inline void netif_tx_wake_all_queues(struct net_device *dev)
3399{
3400 unsigned int i;
3401
3402 for (i = 0; i < dev->num_tx_queues; i++) {
3403 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3404 netif_tx_wake_queue(txq);
3405 }
3406}
3407
3408static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3409{
3410 /* Paired with READ_ONCE() from dev_watchdog() */
3411 WRITE_ONCE(dev_queue->trans_start, jiffies);
3412
3413 /* This barrier is paired with smp_mb() from dev_watchdog() */
3414 smp_mb__before_atomic();
3415
3416 /* Must be an atomic op see netif_txq_try_stop() */
3417 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3418}
3419
3420/**
3421 * netif_stop_queue - stop transmitted packets
3422 * @dev: network device
3423 *
3424 * Stop upper layers calling the device hard_start_xmit routine.
3425 * Used for flow control when transmit resources are unavailable.
3426 */
3427static inline void netif_stop_queue(struct net_device *dev)
3428{
3429 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3430}
3431
3432void netif_tx_stop_all_queues(struct net_device *dev);
3433
3434static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3435{
3436 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3437}
3438
3439/**
3440 * netif_queue_stopped - test if transmit queue is flowblocked
3441 * @dev: network device
3442 *
3443 * Test if transmit queue on device is currently unable to send.
3444 */
3445static inline bool netif_queue_stopped(const struct net_device *dev)
3446{
3447 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3448}
3449
3450static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3451{
3452 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3453}
3454
3455static inline bool
3456netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3457{
3458 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3459}
3460
3461static inline bool
3462netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3463{
3464 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3465}
3466
3467/**
3468 * netdev_queue_set_dql_min_limit - set dql minimum limit
3469 * @dev_queue: pointer to transmit queue
3470 * @min_limit: dql minimum limit
3471 *
3472 * Forces xmit_more() to return true until the minimum threshold
3473 * defined by @min_limit is reached (or until the tx queue is
3474 * empty). Warning: to be use with care, misuse will impact the
3475 * latency.
3476 */
3477static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3478 unsigned int min_limit)
3479{
3480#ifdef CONFIG_BQL
3481 dev_queue->dql.min_limit = min_limit;
3482#endif
3483}
3484
3485static inline int netdev_queue_dql_avail(const struct netdev_queue *txq)
3486{
3487#ifdef CONFIG_BQL
3488 /* Non-BQL migrated drivers will return 0, too. */
3489 return dql_avail(&txq->dql);
3490#else
3491 return 0;
3492#endif
3493}
3494
3495/**
3496 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3497 * @dev_queue: pointer to transmit queue
3498 *
3499 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3500 * to give appropriate hint to the CPU.
3501 */
3502static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3503{
3504#ifdef CONFIG_BQL
3505 prefetchw(&dev_queue->dql.num_queued);
3506#endif
3507}
3508
3509/**
3510 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3511 * @dev_queue: pointer to transmit queue
3512 *
3513 * BQL enabled drivers might use this helper in their TX completion path,
3514 * to give appropriate hint to the CPU.
3515 */
3516static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3517{
3518#ifdef CONFIG_BQL
3519 prefetchw(&dev_queue->dql.limit);
3520#endif
3521}
3522
3523/**
3524 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3525 * @dev_queue: network device queue
3526 * @bytes: number of bytes queued to the device queue
3527 *
3528 * Report the number of bytes queued for sending/completion to the network
3529 * device hardware queue. @bytes should be a good approximation and should
3530 * exactly match netdev_completed_queue() @bytes.
3531 * This is typically called once per packet, from ndo_start_xmit().
3532 */
3533static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3534 unsigned int bytes)
3535{
3536#ifdef CONFIG_BQL
3537 dql_queued(&dev_queue->dql, bytes);
3538
3539 if (likely(dql_avail(&dev_queue->dql) >= 0))
3540 return;
3541
3542 /* Paired with READ_ONCE() from dev_watchdog() */
3543 WRITE_ONCE(dev_queue->trans_start, jiffies);
3544
3545 /* This barrier is paired with smp_mb() from dev_watchdog() */
3546 smp_mb__before_atomic();
3547
3548 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3549
3550 /*
3551 * The XOFF flag must be set before checking the dql_avail below,
3552 * because in netdev_tx_completed_queue we update the dql_completed
3553 * before checking the XOFF flag.
3554 */
3555 smp_mb__after_atomic();
3556
3557 /* check again in case another CPU has just made room avail */
3558 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3559 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3560#endif
3561}
3562
3563/* Variant of netdev_tx_sent_queue() for drivers that are aware
3564 * that they should not test BQL status themselves.
3565 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3566 * skb of a batch.
3567 * Returns true if the doorbell must be used to kick the NIC.
3568 */
3569static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3570 unsigned int bytes,
3571 bool xmit_more)
3572{
3573 if (xmit_more) {
3574#ifdef CONFIG_BQL
3575 dql_queued(&dev_queue->dql, bytes);
3576#endif
3577 return netif_tx_queue_stopped(dev_queue);
3578 }
3579 netdev_tx_sent_queue(dev_queue, bytes);
3580 return true;
3581}
3582
3583/**
3584 * netdev_sent_queue - report the number of bytes queued to hardware
3585 * @dev: network device
3586 * @bytes: number of bytes queued to the hardware device queue
3587 *
3588 * Report the number of bytes queued for sending/completion to the network
3589 * device hardware queue#0. @bytes should be a good approximation and should
3590 * exactly match netdev_completed_queue() @bytes.
3591 * This is typically called once per packet, from ndo_start_xmit().
3592 */
3593static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3594{
3595 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3596}
3597
3598static inline bool __netdev_sent_queue(struct net_device *dev,
3599 unsigned int bytes,
3600 bool xmit_more)
3601{
3602 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3603 xmit_more);
3604}
3605
3606/**
3607 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3608 * @dev_queue: network device queue
3609 * @pkts: number of packets (currently ignored)
3610 * @bytes: number of bytes dequeued from the device queue
3611 *
3612 * Must be called at most once per TX completion round (and not per
3613 * individual packet), so that BQL can adjust its limits appropriately.
3614 */
3615static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3616 unsigned int pkts, unsigned int bytes)
3617{
3618#ifdef CONFIG_BQL
3619 if (unlikely(!bytes))
3620 return;
3621
3622 dql_completed(&dev_queue->dql, bytes);
3623
3624 /*
3625 * Without the memory barrier there is a small possibility that
3626 * netdev_tx_sent_queue will miss the update and cause the queue to
3627 * be stopped forever
3628 */
3629 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3630
3631 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3632 return;
3633
3634 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3635 netif_schedule_queue(dev_queue);
3636#endif
3637}
3638
3639/**
3640 * netdev_completed_queue - report bytes and packets completed by device
3641 * @dev: network device
3642 * @pkts: actual number of packets sent over the medium
3643 * @bytes: actual number of bytes sent over the medium
3644 *
3645 * Report the number of bytes and packets transmitted by the network device
3646 * hardware queue over the physical medium, @bytes must exactly match the
3647 * @bytes amount passed to netdev_sent_queue()
3648 */
3649static inline void netdev_completed_queue(struct net_device *dev,
3650 unsigned int pkts, unsigned int bytes)
3651{
3652 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3653}
3654
3655static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3656{
3657#ifdef CONFIG_BQL
3658 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3659 dql_reset(&q->dql);
3660#endif
3661}
3662
3663/**
3664 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue
3665 * @dev: network device
3666 * @qid: stack index of the queue to reset
3667 */
3668static inline void netdev_tx_reset_subqueue(const struct net_device *dev,
3669 u32 qid)
3670{
3671 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid));
3672}
3673
3674/**
3675 * netdev_reset_queue - reset the packets and bytes count of a network device
3676 * @dev_queue: network device
3677 *
3678 * Reset the bytes and packet count of a network device and clear the
3679 * software flow control OFF bit for this network device
3680 */
3681static inline void netdev_reset_queue(struct net_device *dev_queue)
3682{
3683 netdev_tx_reset_subqueue(dev_queue, 0);
3684}
3685
3686/**
3687 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3688 * @dev: network device
3689 * @queue_index: given tx queue index
3690 *
3691 * Returns 0 if given tx queue index >= number of device tx queues,
3692 * otherwise returns the originally passed tx queue index.
3693 */
3694static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3695{
3696 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3697 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3698 dev->name, queue_index,
3699 dev->real_num_tx_queues);
3700 return 0;
3701 }
3702
3703 return queue_index;
3704}
3705
3706/**
3707 * netif_running - test if up
3708 * @dev: network device
3709 *
3710 * Test if the device has been brought up.
3711 */
3712static inline bool netif_running(const struct net_device *dev)
3713{
3714 return test_bit(__LINK_STATE_START, &dev->state);
3715}
3716
3717/*
3718 * Routines to manage the subqueues on a device. We only need start,
3719 * stop, and a check if it's stopped. All other device management is
3720 * done at the overall netdevice level.
3721 * Also test the device if we're multiqueue.
3722 */
3723
3724/**
3725 * netif_start_subqueue - allow sending packets on subqueue
3726 * @dev: network device
3727 * @queue_index: sub queue index
3728 *
3729 * Start individual transmit queue of a device with multiple transmit queues.
3730 */
3731static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3732{
3733 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3734
3735 netif_tx_start_queue(txq);
3736}
3737
3738/**
3739 * netif_stop_subqueue - stop sending packets on subqueue
3740 * @dev: network device
3741 * @queue_index: sub queue index
3742 *
3743 * Stop individual transmit queue of a device with multiple transmit queues.
3744 */
3745static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3746{
3747 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3748 netif_tx_stop_queue(txq);
3749}
3750
3751/**
3752 * __netif_subqueue_stopped - test status of subqueue
3753 * @dev: network device
3754 * @queue_index: sub queue index
3755 *
3756 * Check individual transmit queue of a device with multiple transmit queues.
3757 */
3758static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3759 u16 queue_index)
3760{
3761 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3762
3763 return netif_tx_queue_stopped(txq);
3764}
3765
3766/**
3767 * netif_subqueue_stopped - test status of subqueue
3768 * @dev: network device
3769 * @skb: sub queue buffer pointer
3770 *
3771 * Check individual transmit queue of a device with multiple transmit queues.
3772 */
3773static inline bool netif_subqueue_stopped(const struct net_device *dev,
3774 struct sk_buff *skb)
3775{
3776 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3777}
3778
3779/**
3780 * netif_wake_subqueue - allow sending packets on subqueue
3781 * @dev: network device
3782 * @queue_index: sub queue index
3783 *
3784 * Resume individual transmit queue of a device with multiple transmit queues.
3785 */
3786static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3787{
3788 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3789
3790 netif_tx_wake_queue(txq);
3791}
3792
3793#ifdef CONFIG_XPS
3794int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3795 u16 index);
3796int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3797 u16 index, enum xps_map_type type);
3798
3799/**
3800 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3801 * @j: CPU/Rx queue index
3802 * @mask: bitmask of all cpus/rx queues
3803 * @nr_bits: number of bits in the bitmask
3804 *
3805 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3806 */
3807static inline bool netif_attr_test_mask(unsigned long j,
3808 const unsigned long *mask,
3809 unsigned int nr_bits)
3810{
3811 cpu_max_bits_warn(j, nr_bits);
3812 return test_bit(j, mask);
3813}
3814
3815/**
3816 * netif_attr_test_online - Test for online CPU/Rx queue
3817 * @j: CPU/Rx queue index
3818 * @online_mask: bitmask for CPUs/Rx queues that are online
3819 * @nr_bits: number of bits in the bitmask
3820 *
3821 * Returns true if a CPU/Rx queue is online.
3822 */
3823static inline bool netif_attr_test_online(unsigned long j,
3824 const unsigned long *online_mask,
3825 unsigned int nr_bits)
3826{
3827 cpu_max_bits_warn(j, nr_bits);
3828
3829 if (online_mask)
3830 return test_bit(j, online_mask);
3831
3832 return (j < nr_bits);
3833}
3834
3835/**
3836 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3837 * @n: CPU/Rx queue index
3838 * @srcp: the cpumask/Rx queue mask pointer
3839 * @nr_bits: number of bits in the bitmask
3840 *
3841 * Returns >= nr_bits if no further CPUs/Rx queues set.
3842 */
3843static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3844 unsigned int nr_bits)
3845{
3846 /* -1 is a legal arg here. */
3847 if (n != -1)
3848 cpu_max_bits_warn(n, nr_bits);
3849
3850 if (srcp)
3851 return find_next_bit(srcp, nr_bits, n + 1);
3852
3853 return n + 1;
3854}
3855
3856/**
3857 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3858 * @n: CPU/Rx queue index
3859 * @src1p: the first CPUs/Rx queues mask pointer
3860 * @src2p: the second CPUs/Rx queues mask pointer
3861 * @nr_bits: number of bits in the bitmask
3862 *
3863 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3864 */
3865static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3866 const unsigned long *src2p,
3867 unsigned int nr_bits)
3868{
3869 /* -1 is a legal arg here. */
3870 if (n != -1)
3871 cpu_max_bits_warn(n, nr_bits);
3872
3873 if (src1p && src2p)
3874 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3875 else if (src1p)
3876 return find_next_bit(src1p, nr_bits, n + 1);
3877 else if (src2p)
3878 return find_next_bit(src2p, nr_bits, n + 1);
3879
3880 return n + 1;
3881}
3882#else
3883static inline int netif_set_xps_queue(struct net_device *dev,
3884 const struct cpumask *mask,
3885 u16 index)
3886{
3887 return 0;
3888}
3889
3890static inline int __netif_set_xps_queue(struct net_device *dev,
3891 const unsigned long *mask,
3892 u16 index, enum xps_map_type type)
3893{
3894 return 0;
3895}
3896#endif
3897
3898/**
3899 * netif_is_multiqueue - test if device has multiple transmit queues
3900 * @dev: network device
3901 *
3902 * Check if device has multiple transmit queues
3903 */
3904static inline bool netif_is_multiqueue(const struct net_device *dev)
3905{
3906 return dev->num_tx_queues > 1;
3907}
3908
3909int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3910
3911#ifdef CONFIG_SYSFS
3912int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3913#else
3914static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3915 unsigned int rxqs)
3916{
3917 dev->real_num_rx_queues = rxqs;
3918 return 0;
3919}
3920#endif
3921int netif_set_real_num_queues(struct net_device *dev,
3922 unsigned int txq, unsigned int rxq);
3923
3924int netif_get_num_default_rss_queues(void);
3925
3926void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3927void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3928
3929/*
3930 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3931 * interrupt context or with hardware interrupts being disabled.
3932 * (in_hardirq() || irqs_disabled())
3933 *
3934 * We provide four helpers that can be used in following contexts :
3935 *
3936 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3937 * replacing kfree_skb(skb)
3938 *
3939 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3940 * Typically used in place of consume_skb(skb) in TX completion path
3941 *
3942 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3943 * replacing kfree_skb(skb)
3944 *
3945 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3946 * and consumed a packet. Used in place of consume_skb(skb)
3947 */
3948static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3949{
3950 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3951}
3952
3953static inline void dev_consume_skb_irq(struct sk_buff *skb)
3954{
3955 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3956}
3957
3958static inline void dev_kfree_skb_any(struct sk_buff *skb)
3959{
3960 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3961}
3962
3963static inline void dev_consume_skb_any(struct sk_buff *skb)
3964{
3965 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3966}
3967
3968u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3969 struct bpf_prog *xdp_prog);
3970void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3971int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb);
3972int netif_rx(struct sk_buff *skb);
3973int __netif_rx(struct sk_buff *skb);
3974
3975int netif_receive_skb(struct sk_buff *skb);
3976int netif_receive_skb_core(struct sk_buff *skb);
3977void netif_receive_skb_list_internal(struct list_head *head);
3978void netif_receive_skb_list(struct list_head *head);
3979gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3980void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3981struct sk_buff *napi_get_frags(struct napi_struct *napi);
3982void napi_get_frags_check(struct napi_struct *napi);
3983gro_result_t napi_gro_frags(struct napi_struct *napi);
3984
3985static inline void napi_free_frags(struct napi_struct *napi)
3986{
3987 kfree_skb(napi->skb);
3988 napi->skb = NULL;
3989}
3990
3991bool netdev_is_rx_handler_busy(struct net_device *dev);
3992int netdev_rx_handler_register(struct net_device *dev,
3993 rx_handler_func_t *rx_handler,
3994 void *rx_handler_data);
3995void netdev_rx_handler_unregister(struct net_device *dev);
3996
3997bool dev_valid_name(const char *name);
3998static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3999{
4000 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
4001}
4002int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
4003int put_user_ifreq(struct ifreq *ifr, void __user *arg);
4004int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4005 void __user *data, bool *need_copyout);
4006int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4007int generic_hwtstamp_get_lower(struct net_device *dev,
4008 struct kernel_hwtstamp_config *kernel_cfg);
4009int generic_hwtstamp_set_lower(struct net_device *dev,
4010 struct kernel_hwtstamp_config *kernel_cfg,
4011 struct netlink_ext_ack *extack);
4012int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4013unsigned int dev_get_flags(const struct net_device *);
4014int __dev_change_flags(struct net_device *dev, unsigned int flags,
4015 struct netlink_ext_ack *extack);
4016int dev_change_flags(struct net_device *dev, unsigned int flags,
4017 struct netlink_ext_ack *extack);
4018int dev_set_alias(struct net_device *, const char *, size_t);
4019int dev_get_alias(const struct net_device *, char *, size_t);
4020int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4021 const char *pat, int new_ifindex);
4022static inline
4023int dev_change_net_namespace(struct net_device *dev, struct net *net,
4024 const char *pat)
4025{
4026 return __dev_change_net_namespace(dev, net, pat, 0);
4027}
4028int __dev_set_mtu(struct net_device *, int);
4029int dev_set_mtu(struct net_device *, int);
4030int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4031 struct netlink_ext_ack *extack);
4032int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4033 struct netlink_ext_ack *extack);
4034int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4035 struct netlink_ext_ack *extack);
4036int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4037int dev_get_port_parent_id(struct net_device *dev,
4038 struct netdev_phys_item_id *ppid, bool recurse);
4039bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4040
4041struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4042struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4043 struct netdev_queue *txq, int *ret);
4044
4045int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4046u8 dev_xdp_prog_count(struct net_device *dev);
4047int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf);
4048u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4049
4050u32 dev_get_min_mp_channel_count(const struct net_device *dev);
4051
4052int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4053int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4054int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4055bool is_skb_forwardable(const struct net_device *dev,
4056 const struct sk_buff *skb);
4057
4058static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4059 const struct sk_buff *skb,
4060 const bool check_mtu)
4061{
4062 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4063 unsigned int len;
4064
4065 if (!(dev->flags & IFF_UP))
4066 return false;
4067
4068 if (!check_mtu)
4069 return true;
4070
4071 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4072 if (skb->len <= len)
4073 return true;
4074
4075 /* if TSO is enabled, we don't care about the length as the packet
4076 * could be forwarded without being segmented before
4077 */
4078 if (skb_is_gso(skb))
4079 return true;
4080
4081 return false;
4082}
4083
4084void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4085
4086#define DEV_CORE_STATS_INC(FIELD) \
4087static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4088{ \
4089 netdev_core_stats_inc(dev, \
4090 offsetof(struct net_device_core_stats, FIELD)); \
4091}
4092DEV_CORE_STATS_INC(rx_dropped)
4093DEV_CORE_STATS_INC(tx_dropped)
4094DEV_CORE_STATS_INC(rx_nohandler)
4095DEV_CORE_STATS_INC(rx_otherhost_dropped)
4096#undef DEV_CORE_STATS_INC
4097
4098static __always_inline int ____dev_forward_skb(struct net_device *dev,
4099 struct sk_buff *skb,
4100 const bool check_mtu)
4101{
4102 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4103 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4104 dev_core_stats_rx_dropped_inc(dev);
4105 kfree_skb(skb);
4106 return NET_RX_DROP;
4107 }
4108
4109 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4110 skb->priority = 0;
4111 return 0;
4112}
4113
4114bool dev_nit_active(struct net_device *dev);
4115void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4116
4117static inline void __dev_put(struct net_device *dev)
4118{
4119 if (dev) {
4120#ifdef CONFIG_PCPU_DEV_REFCNT
4121 this_cpu_dec(*dev->pcpu_refcnt);
4122#else
4123 refcount_dec(&dev->dev_refcnt);
4124#endif
4125 }
4126}
4127
4128static inline void __dev_hold(struct net_device *dev)
4129{
4130 if (dev) {
4131#ifdef CONFIG_PCPU_DEV_REFCNT
4132 this_cpu_inc(*dev->pcpu_refcnt);
4133#else
4134 refcount_inc(&dev->dev_refcnt);
4135#endif
4136 }
4137}
4138
4139static inline void __netdev_tracker_alloc(struct net_device *dev,
4140 netdevice_tracker *tracker,
4141 gfp_t gfp)
4142{
4143#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4144 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4145#endif
4146}
4147
4148/* netdev_tracker_alloc() can upgrade a prior untracked reference
4149 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4150 */
4151static inline void netdev_tracker_alloc(struct net_device *dev,
4152 netdevice_tracker *tracker, gfp_t gfp)
4153{
4154#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4155 refcount_dec(&dev->refcnt_tracker.no_tracker);
4156 __netdev_tracker_alloc(dev, tracker, gfp);
4157#endif
4158}
4159
4160static inline void netdev_tracker_free(struct net_device *dev,
4161 netdevice_tracker *tracker)
4162{
4163#ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4164 ref_tracker_free(&dev->refcnt_tracker, tracker);
4165#endif
4166}
4167
4168static inline void netdev_hold(struct net_device *dev,
4169 netdevice_tracker *tracker, gfp_t gfp)
4170{
4171 if (dev) {
4172 __dev_hold(dev);
4173 __netdev_tracker_alloc(dev, tracker, gfp);
4174 }
4175}
4176
4177static inline void netdev_put(struct net_device *dev,
4178 netdevice_tracker *tracker)
4179{
4180 if (dev) {
4181 netdev_tracker_free(dev, tracker);
4182 __dev_put(dev);
4183 }
4184}
4185
4186/**
4187 * dev_hold - get reference to device
4188 * @dev: network device
4189 *
4190 * Hold reference to device to keep it from being freed.
4191 * Try using netdev_hold() instead.
4192 */
4193static inline void dev_hold(struct net_device *dev)
4194{
4195 netdev_hold(dev, NULL, GFP_ATOMIC);
4196}
4197
4198/**
4199 * dev_put - release reference to device
4200 * @dev: network device
4201 *
4202 * Release reference to device to allow it to be freed.
4203 * Try using netdev_put() instead.
4204 */
4205static inline void dev_put(struct net_device *dev)
4206{
4207 netdev_put(dev, NULL);
4208}
4209
4210DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T))
4211
4212static inline void netdev_ref_replace(struct net_device *odev,
4213 struct net_device *ndev,
4214 netdevice_tracker *tracker,
4215 gfp_t gfp)
4216{
4217 if (odev)
4218 netdev_tracker_free(odev, tracker);
4219
4220 __dev_hold(ndev);
4221 __dev_put(odev);
4222
4223 if (ndev)
4224 __netdev_tracker_alloc(ndev, tracker, gfp);
4225}
4226
4227/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4228 * and _off may be called from IRQ context, but it is caller
4229 * who is responsible for serialization of these calls.
4230 *
4231 * The name carrier is inappropriate, these functions should really be
4232 * called netif_lowerlayer_*() because they represent the state of any
4233 * kind of lower layer not just hardware media.
4234 */
4235void linkwatch_fire_event(struct net_device *dev);
4236
4237/**
4238 * linkwatch_sync_dev - sync linkwatch for the given device
4239 * @dev: network device to sync linkwatch for
4240 *
4241 * Sync linkwatch for the given device, removing it from the
4242 * pending work list (if queued).
4243 */
4244void linkwatch_sync_dev(struct net_device *dev);
4245
4246/**
4247 * netif_carrier_ok - test if carrier present
4248 * @dev: network device
4249 *
4250 * Check if carrier is present on device
4251 */
4252static inline bool netif_carrier_ok(const struct net_device *dev)
4253{
4254 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4255}
4256
4257unsigned long dev_trans_start(struct net_device *dev);
4258
4259void __netdev_watchdog_up(struct net_device *dev);
4260
4261void netif_carrier_on(struct net_device *dev);
4262void netif_carrier_off(struct net_device *dev);
4263void netif_carrier_event(struct net_device *dev);
4264
4265/**
4266 * netif_dormant_on - mark device as dormant.
4267 * @dev: network device
4268 *
4269 * Mark device as dormant (as per RFC2863).
4270 *
4271 * The dormant state indicates that the relevant interface is not
4272 * actually in a condition to pass packets (i.e., it is not 'up') but is
4273 * in a "pending" state, waiting for some external event. For "on-
4274 * demand" interfaces, this new state identifies the situation where the
4275 * interface is waiting for events to place it in the up state.
4276 */
4277static inline void netif_dormant_on(struct net_device *dev)
4278{
4279 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4280 linkwatch_fire_event(dev);
4281}
4282
4283/**
4284 * netif_dormant_off - set device as not dormant.
4285 * @dev: network device
4286 *
4287 * Device is not in dormant state.
4288 */
4289static inline void netif_dormant_off(struct net_device *dev)
4290{
4291 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4292 linkwatch_fire_event(dev);
4293}
4294
4295/**
4296 * netif_dormant - test if device is dormant
4297 * @dev: network device
4298 *
4299 * Check if device is dormant.
4300 */
4301static inline bool netif_dormant(const struct net_device *dev)
4302{
4303 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4304}
4305
4306
4307/**
4308 * netif_testing_on - mark device as under test.
4309 * @dev: network device
4310 *
4311 * Mark device as under test (as per RFC2863).
4312 *
4313 * The testing state indicates that some test(s) must be performed on
4314 * the interface. After completion, of the test, the interface state
4315 * will change to up, dormant, or down, as appropriate.
4316 */
4317static inline void netif_testing_on(struct net_device *dev)
4318{
4319 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4320 linkwatch_fire_event(dev);
4321}
4322
4323/**
4324 * netif_testing_off - set device as not under test.
4325 * @dev: network device
4326 *
4327 * Device is not in testing state.
4328 */
4329static inline void netif_testing_off(struct net_device *dev)
4330{
4331 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4332 linkwatch_fire_event(dev);
4333}
4334
4335/**
4336 * netif_testing - test if device is under test
4337 * @dev: network device
4338 *
4339 * Check if device is under test
4340 */
4341static inline bool netif_testing(const struct net_device *dev)
4342{
4343 return test_bit(__LINK_STATE_TESTING, &dev->state);
4344}
4345
4346
4347/**
4348 * netif_oper_up - test if device is operational
4349 * @dev: network device
4350 *
4351 * Check if carrier is operational
4352 */
4353static inline bool netif_oper_up(const struct net_device *dev)
4354{
4355 unsigned int operstate = READ_ONCE(dev->operstate);
4356
4357 return operstate == IF_OPER_UP ||
4358 operstate == IF_OPER_UNKNOWN /* backward compat */;
4359}
4360
4361/**
4362 * netif_device_present - is device available or removed
4363 * @dev: network device
4364 *
4365 * Check if device has not been removed from system.
4366 */
4367static inline bool netif_device_present(const struct net_device *dev)
4368{
4369 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4370}
4371
4372void netif_device_detach(struct net_device *dev);
4373
4374void netif_device_attach(struct net_device *dev);
4375
4376/*
4377 * Network interface message level settings
4378 */
4379
4380enum {
4381 NETIF_MSG_DRV_BIT,
4382 NETIF_MSG_PROBE_BIT,
4383 NETIF_MSG_LINK_BIT,
4384 NETIF_MSG_TIMER_BIT,
4385 NETIF_MSG_IFDOWN_BIT,
4386 NETIF_MSG_IFUP_BIT,
4387 NETIF_MSG_RX_ERR_BIT,
4388 NETIF_MSG_TX_ERR_BIT,
4389 NETIF_MSG_TX_QUEUED_BIT,
4390 NETIF_MSG_INTR_BIT,
4391 NETIF_MSG_TX_DONE_BIT,
4392 NETIF_MSG_RX_STATUS_BIT,
4393 NETIF_MSG_PKTDATA_BIT,
4394 NETIF_MSG_HW_BIT,
4395 NETIF_MSG_WOL_BIT,
4396
4397 /* When you add a new bit above, update netif_msg_class_names array
4398 * in net/ethtool/common.c
4399 */
4400 NETIF_MSG_CLASS_COUNT,
4401};
4402/* Both ethtool_ops interface and internal driver implementation use u32 */
4403static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4404
4405#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4406#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4407
4408#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4409#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4410#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4411#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4412#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4413#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4414#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4415#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4416#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4417#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4418#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4419#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4420#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4421#define NETIF_MSG_HW __NETIF_MSG(HW)
4422#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4423
4424#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4425#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4426#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4427#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4428#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4429#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4430#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4431#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4432#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4433#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4434#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4435#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4436#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4437#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4438#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4439
4440static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4441{
4442 /* use default */
4443 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4444 return default_msg_enable_bits;
4445 if (debug_value == 0) /* no output */
4446 return 0;
4447 /* set low N bits */
4448 return (1U << debug_value) - 1;
4449}
4450
4451static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4452{
4453 spin_lock(&txq->_xmit_lock);
4454 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4455 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4456}
4457
4458static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4459{
4460 __acquire(&txq->_xmit_lock);
4461 return true;
4462}
4463
4464static inline void __netif_tx_release(struct netdev_queue *txq)
4465{
4466 __release(&txq->_xmit_lock);
4467}
4468
4469static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4470{
4471 spin_lock_bh(&txq->_xmit_lock);
4472 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4473 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4474}
4475
4476static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4477{
4478 bool ok = spin_trylock(&txq->_xmit_lock);
4479
4480 if (likely(ok)) {
4481 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4482 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4483 }
4484 return ok;
4485}
4486
4487static inline void __netif_tx_unlock(struct netdev_queue *txq)
4488{
4489 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4490 WRITE_ONCE(txq->xmit_lock_owner, -1);
4491 spin_unlock(&txq->_xmit_lock);
4492}
4493
4494static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4495{
4496 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4497 WRITE_ONCE(txq->xmit_lock_owner, -1);
4498 spin_unlock_bh(&txq->_xmit_lock);
4499}
4500
4501/*
4502 * txq->trans_start can be read locklessly from dev_watchdog()
4503 */
4504static inline void txq_trans_update(struct netdev_queue *txq)
4505{
4506 if (txq->xmit_lock_owner != -1)
4507 WRITE_ONCE(txq->trans_start, jiffies);
4508}
4509
4510static inline void txq_trans_cond_update(struct netdev_queue *txq)
4511{
4512 unsigned long now = jiffies;
4513
4514 if (READ_ONCE(txq->trans_start) != now)
4515 WRITE_ONCE(txq->trans_start, now);
4516}
4517
4518/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4519static inline void netif_trans_update(struct net_device *dev)
4520{
4521 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4522
4523 txq_trans_cond_update(txq);
4524}
4525
4526/**
4527 * netif_tx_lock - grab network device transmit lock
4528 * @dev: network device
4529 *
4530 * Get network device transmit lock
4531 */
4532void netif_tx_lock(struct net_device *dev);
4533
4534static inline void netif_tx_lock_bh(struct net_device *dev)
4535{
4536 local_bh_disable();
4537 netif_tx_lock(dev);
4538}
4539
4540void netif_tx_unlock(struct net_device *dev);
4541
4542static inline void netif_tx_unlock_bh(struct net_device *dev)
4543{
4544 netif_tx_unlock(dev);
4545 local_bh_enable();
4546}
4547
4548#define HARD_TX_LOCK(dev, txq, cpu) { \
4549 if (!(dev)->lltx) { \
4550 __netif_tx_lock(txq, cpu); \
4551 } else { \
4552 __netif_tx_acquire(txq); \
4553 } \
4554}
4555
4556#define HARD_TX_TRYLOCK(dev, txq) \
4557 (!(dev)->lltx ? \
4558 __netif_tx_trylock(txq) : \
4559 __netif_tx_acquire(txq))
4560
4561#define HARD_TX_UNLOCK(dev, txq) { \
4562 if (!(dev)->lltx) { \
4563 __netif_tx_unlock(txq); \
4564 } else { \
4565 __netif_tx_release(txq); \
4566 } \
4567}
4568
4569static inline void netif_tx_disable(struct net_device *dev)
4570{
4571 unsigned int i;
4572 int cpu;
4573
4574 local_bh_disable();
4575 cpu = smp_processor_id();
4576 spin_lock(&dev->tx_global_lock);
4577 for (i = 0; i < dev->num_tx_queues; i++) {
4578 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4579
4580 __netif_tx_lock(txq, cpu);
4581 netif_tx_stop_queue(txq);
4582 __netif_tx_unlock(txq);
4583 }
4584 spin_unlock(&dev->tx_global_lock);
4585 local_bh_enable();
4586}
4587
4588static inline void netif_addr_lock(struct net_device *dev)
4589{
4590 unsigned char nest_level = 0;
4591
4592#ifdef CONFIG_LOCKDEP
4593 nest_level = dev->nested_level;
4594#endif
4595 spin_lock_nested(&dev->addr_list_lock, nest_level);
4596}
4597
4598static inline void netif_addr_lock_bh(struct net_device *dev)
4599{
4600 unsigned char nest_level = 0;
4601
4602#ifdef CONFIG_LOCKDEP
4603 nest_level = dev->nested_level;
4604#endif
4605 local_bh_disable();
4606 spin_lock_nested(&dev->addr_list_lock, nest_level);
4607}
4608
4609static inline void netif_addr_unlock(struct net_device *dev)
4610{
4611 spin_unlock(&dev->addr_list_lock);
4612}
4613
4614static inline void netif_addr_unlock_bh(struct net_device *dev)
4615{
4616 spin_unlock_bh(&dev->addr_list_lock);
4617}
4618
4619/*
4620 * dev_addrs walker. Should be used only for read access. Call with
4621 * rcu_read_lock held.
4622 */
4623#define for_each_dev_addr(dev, ha) \
4624 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4625
4626/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4627
4628void ether_setup(struct net_device *dev);
4629
4630/* Allocate dummy net_device */
4631struct net_device *alloc_netdev_dummy(int sizeof_priv);
4632
4633/* Support for loadable net-drivers */
4634struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4635 unsigned char name_assign_type,
4636 void (*setup)(struct net_device *),
4637 unsigned int txqs, unsigned int rxqs);
4638#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4639 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4640
4641#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4642 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4643 count)
4644
4645int register_netdev(struct net_device *dev);
4646void unregister_netdev(struct net_device *dev);
4647
4648int devm_register_netdev(struct device *dev, struct net_device *ndev);
4649
4650/* General hardware address lists handling functions */
4651int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4652 struct netdev_hw_addr_list *from_list, int addr_len);
4653void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4654 struct netdev_hw_addr_list *from_list, int addr_len);
4655int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4656 struct net_device *dev,
4657 int (*sync)(struct net_device *, const unsigned char *),
4658 int (*unsync)(struct net_device *,
4659 const unsigned char *));
4660int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4661 struct net_device *dev,
4662 int (*sync)(struct net_device *,
4663 const unsigned char *, int),
4664 int (*unsync)(struct net_device *,
4665 const unsigned char *, int));
4666void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4667 struct net_device *dev,
4668 int (*unsync)(struct net_device *,
4669 const unsigned char *, int));
4670void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4671 struct net_device *dev,
4672 int (*unsync)(struct net_device *,
4673 const unsigned char *));
4674void __hw_addr_init(struct netdev_hw_addr_list *list);
4675
4676/* Functions used for device addresses handling */
4677void dev_addr_mod(struct net_device *dev, unsigned int offset,
4678 const void *addr, size_t len);
4679
4680static inline void
4681__dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4682{
4683 dev_addr_mod(dev, 0, addr, len);
4684}
4685
4686static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4687{
4688 __dev_addr_set(dev, addr, dev->addr_len);
4689}
4690
4691int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4692 unsigned char addr_type);
4693int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4694 unsigned char addr_type);
4695
4696/* Functions used for unicast addresses handling */
4697int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4698int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4699int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4700int dev_uc_sync(struct net_device *to, struct net_device *from);
4701int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4702void dev_uc_unsync(struct net_device *to, struct net_device *from);
4703void dev_uc_flush(struct net_device *dev);
4704void dev_uc_init(struct net_device *dev);
4705
4706/**
4707 * __dev_uc_sync - Synchronize device's unicast list
4708 * @dev: device to sync
4709 * @sync: function to call if address should be added
4710 * @unsync: function to call if address should be removed
4711 *
4712 * Add newly added addresses to the interface, and release
4713 * addresses that have been deleted.
4714 */
4715static inline int __dev_uc_sync(struct net_device *dev,
4716 int (*sync)(struct net_device *,
4717 const unsigned char *),
4718 int (*unsync)(struct net_device *,
4719 const unsigned char *))
4720{
4721 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4722}
4723
4724/**
4725 * __dev_uc_unsync - Remove synchronized addresses from device
4726 * @dev: device to sync
4727 * @unsync: function to call if address should be removed
4728 *
4729 * Remove all addresses that were added to the device by dev_uc_sync().
4730 */
4731static inline void __dev_uc_unsync(struct net_device *dev,
4732 int (*unsync)(struct net_device *,
4733 const unsigned char *))
4734{
4735 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4736}
4737
4738/* Functions used for multicast addresses handling */
4739int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4740int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4741int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4742int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4743int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4744int dev_mc_sync(struct net_device *to, struct net_device *from);
4745int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4746void dev_mc_unsync(struct net_device *to, struct net_device *from);
4747void dev_mc_flush(struct net_device *dev);
4748void dev_mc_init(struct net_device *dev);
4749
4750/**
4751 * __dev_mc_sync - Synchronize device's multicast list
4752 * @dev: device to sync
4753 * @sync: function to call if address should be added
4754 * @unsync: function to call if address should be removed
4755 *
4756 * Add newly added addresses to the interface, and release
4757 * addresses that have been deleted.
4758 */
4759static inline int __dev_mc_sync(struct net_device *dev,
4760 int (*sync)(struct net_device *,
4761 const unsigned char *),
4762 int (*unsync)(struct net_device *,
4763 const unsigned char *))
4764{
4765 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4766}
4767
4768/**
4769 * __dev_mc_unsync - Remove synchronized addresses from device
4770 * @dev: device to sync
4771 * @unsync: function to call if address should be removed
4772 *
4773 * Remove all addresses that were added to the device by dev_mc_sync().
4774 */
4775static inline void __dev_mc_unsync(struct net_device *dev,
4776 int (*unsync)(struct net_device *,
4777 const unsigned char *))
4778{
4779 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4780}
4781
4782/* Functions used for secondary unicast and multicast support */
4783void dev_set_rx_mode(struct net_device *dev);
4784int dev_set_promiscuity(struct net_device *dev, int inc);
4785int dev_set_allmulti(struct net_device *dev, int inc);
4786void netdev_state_change(struct net_device *dev);
4787void __netdev_notify_peers(struct net_device *dev);
4788void netdev_notify_peers(struct net_device *dev);
4789void netdev_features_change(struct net_device *dev);
4790/* Load a device via the kmod */
4791void dev_load(struct net *net, const char *name);
4792struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4793 struct rtnl_link_stats64 *storage);
4794void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4795 const struct net_device_stats *netdev_stats);
4796void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4797 const struct pcpu_sw_netstats __percpu *netstats);
4798void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4799
4800enum {
4801 NESTED_SYNC_IMM_BIT,
4802 NESTED_SYNC_TODO_BIT,
4803};
4804
4805#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4806#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4807
4808#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4809#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4810
4811struct netdev_nested_priv {
4812 unsigned char flags;
4813 void *data;
4814};
4815
4816bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4817struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4818 struct list_head **iter);
4819
4820/* iterate through upper list, must be called under RCU read lock */
4821#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4822 for (iter = &(dev)->adj_list.upper, \
4823 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4824 updev; \
4825 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4826
4827int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4828 int (*fn)(struct net_device *upper_dev,
4829 struct netdev_nested_priv *priv),
4830 struct netdev_nested_priv *priv);
4831
4832bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4833 struct net_device *upper_dev);
4834
4835bool netdev_has_any_upper_dev(struct net_device *dev);
4836
4837void *netdev_lower_get_next_private(struct net_device *dev,
4838 struct list_head **iter);
4839void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4840 struct list_head **iter);
4841
4842#define netdev_for_each_lower_private(dev, priv, iter) \
4843 for (iter = (dev)->adj_list.lower.next, \
4844 priv = netdev_lower_get_next_private(dev, &(iter)); \
4845 priv; \
4846 priv = netdev_lower_get_next_private(dev, &(iter)))
4847
4848#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4849 for (iter = &(dev)->adj_list.lower, \
4850 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4851 priv; \
4852 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4853
4854void *netdev_lower_get_next(struct net_device *dev,
4855 struct list_head **iter);
4856
4857#define netdev_for_each_lower_dev(dev, ldev, iter) \
4858 for (iter = (dev)->adj_list.lower.next, \
4859 ldev = netdev_lower_get_next(dev, &(iter)); \
4860 ldev; \
4861 ldev = netdev_lower_get_next(dev, &(iter)))
4862
4863struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4864 struct list_head **iter);
4865int netdev_walk_all_lower_dev(struct net_device *dev,
4866 int (*fn)(struct net_device *lower_dev,
4867 struct netdev_nested_priv *priv),
4868 struct netdev_nested_priv *priv);
4869int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4870 int (*fn)(struct net_device *lower_dev,
4871 struct netdev_nested_priv *priv),
4872 struct netdev_nested_priv *priv);
4873
4874void *netdev_adjacent_get_private(struct list_head *adj_list);
4875void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4876struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4877struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4878int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4879 struct netlink_ext_ack *extack);
4880int netdev_master_upper_dev_link(struct net_device *dev,
4881 struct net_device *upper_dev,
4882 void *upper_priv, void *upper_info,
4883 struct netlink_ext_ack *extack);
4884void netdev_upper_dev_unlink(struct net_device *dev,
4885 struct net_device *upper_dev);
4886int netdev_adjacent_change_prepare(struct net_device *old_dev,
4887 struct net_device *new_dev,
4888 struct net_device *dev,
4889 struct netlink_ext_ack *extack);
4890void netdev_adjacent_change_commit(struct net_device *old_dev,
4891 struct net_device *new_dev,
4892 struct net_device *dev);
4893void netdev_adjacent_change_abort(struct net_device *old_dev,
4894 struct net_device *new_dev,
4895 struct net_device *dev);
4896void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4897void *netdev_lower_dev_get_private(struct net_device *dev,
4898 struct net_device *lower_dev);
4899void netdev_lower_state_changed(struct net_device *lower_dev,
4900 void *lower_state_info);
4901
4902/* RSS keys are 40 or 52 bytes long */
4903#define NETDEV_RSS_KEY_LEN 52
4904extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4905void netdev_rss_key_fill(void *buffer, size_t len);
4906
4907int skb_checksum_help(struct sk_buff *skb);
4908int skb_crc32c_csum_help(struct sk_buff *skb);
4909int skb_csum_hwoffload_help(struct sk_buff *skb,
4910 const netdev_features_t features);
4911
4912struct netdev_bonding_info {
4913 ifslave slave;
4914 ifbond master;
4915};
4916
4917struct netdev_notifier_bonding_info {
4918 struct netdev_notifier_info info; /* must be first */
4919 struct netdev_bonding_info bonding_info;
4920};
4921
4922void netdev_bonding_info_change(struct net_device *dev,
4923 struct netdev_bonding_info *bonding_info);
4924
4925#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4926void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4927#else
4928static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4929 const void *data)
4930{
4931}
4932#endif
4933
4934__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4935
4936static inline bool can_checksum_protocol(netdev_features_t features,
4937 __be16 protocol)
4938{
4939 if (protocol == htons(ETH_P_FCOE))
4940 return !!(features & NETIF_F_FCOE_CRC);
4941
4942 /* Assume this is an IP checksum (not SCTP CRC) */
4943
4944 if (features & NETIF_F_HW_CSUM) {
4945 /* Can checksum everything */
4946 return true;
4947 }
4948
4949 switch (protocol) {
4950 case htons(ETH_P_IP):
4951 return !!(features & NETIF_F_IP_CSUM);
4952 case htons(ETH_P_IPV6):
4953 return !!(features & NETIF_F_IPV6_CSUM);
4954 default:
4955 return false;
4956 }
4957}
4958
4959#ifdef CONFIG_BUG
4960void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4961#else
4962static inline void netdev_rx_csum_fault(struct net_device *dev,
4963 struct sk_buff *skb)
4964{
4965}
4966#endif
4967/* rx skb timestamps */
4968void net_enable_timestamp(void);
4969void net_disable_timestamp(void);
4970
4971static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4972 const struct skb_shared_hwtstamps *hwtstamps,
4973 bool cycles)
4974{
4975 const struct net_device_ops *ops = dev->netdev_ops;
4976
4977 if (ops->ndo_get_tstamp)
4978 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4979
4980 return hwtstamps->hwtstamp;
4981}
4982
4983#ifndef CONFIG_PREEMPT_RT
4984static inline void netdev_xmit_set_more(bool more)
4985{
4986 __this_cpu_write(softnet_data.xmit.more, more);
4987}
4988
4989static inline bool netdev_xmit_more(void)
4990{
4991 return __this_cpu_read(softnet_data.xmit.more);
4992}
4993#else
4994static inline void netdev_xmit_set_more(bool more)
4995{
4996 current->net_xmit.more = more;
4997}
4998
4999static inline bool netdev_xmit_more(void)
5000{
5001 return current->net_xmit.more;
5002}
5003#endif
5004
5005static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
5006 struct sk_buff *skb, struct net_device *dev,
5007 bool more)
5008{
5009 netdev_xmit_set_more(more);
5010 return ops->ndo_start_xmit(skb, dev);
5011}
5012
5013static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
5014 struct netdev_queue *txq, bool more)
5015{
5016 const struct net_device_ops *ops = dev->netdev_ops;
5017 netdev_tx_t rc;
5018
5019 rc = __netdev_start_xmit(ops, skb, dev, more);
5020 if (rc == NETDEV_TX_OK)
5021 txq_trans_update(txq);
5022
5023 return rc;
5024}
5025
5026int netdev_class_create_file_ns(const struct class_attribute *class_attr,
5027 const void *ns);
5028void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
5029 const void *ns);
5030
5031extern const struct kobj_ns_type_operations net_ns_type_operations;
5032
5033const char *netdev_drivername(const struct net_device *dev);
5034
5035static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
5036 netdev_features_t f2)
5037{
5038 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
5039 if (f1 & NETIF_F_HW_CSUM)
5040 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5041 else
5042 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5043 }
5044
5045 return f1 & f2;
5046}
5047
5048static inline netdev_features_t netdev_get_wanted_features(
5049 struct net_device *dev)
5050{
5051 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5052}
5053netdev_features_t netdev_increment_features(netdev_features_t all,
5054 netdev_features_t one, netdev_features_t mask);
5055
5056/* Allow TSO being used on stacked device :
5057 * Performing the GSO segmentation before last device
5058 * is a performance improvement.
5059 */
5060static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5061 netdev_features_t mask)
5062{
5063 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5064}
5065
5066int __netdev_update_features(struct net_device *dev);
5067void netdev_update_features(struct net_device *dev);
5068void netdev_change_features(struct net_device *dev);
5069
5070void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5071 struct net_device *dev);
5072
5073netdev_features_t passthru_features_check(struct sk_buff *skb,
5074 struct net_device *dev,
5075 netdev_features_t features);
5076netdev_features_t netif_skb_features(struct sk_buff *skb);
5077void skb_warn_bad_offload(const struct sk_buff *skb);
5078
5079static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5080{
5081 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5082
5083 /* check flags correspondence */
5084 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5085 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5086 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5087 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5088 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5089 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5090 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5091 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5092 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5093 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5094 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5095 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5096 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5097 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5098 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5099 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5100 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5101 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5102 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5103
5104 return (features & feature) == feature;
5105}
5106
5107static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5108{
5109 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5110 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5111}
5112
5113static inline bool netif_needs_gso(struct sk_buff *skb,
5114 netdev_features_t features)
5115{
5116 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5117 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5118 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5119}
5120
5121void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5122void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5123void netif_inherit_tso_max(struct net_device *to,
5124 const struct net_device *from);
5125
5126static inline unsigned int
5127netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb)
5128{
5129 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */
5130 return skb->protocol == htons(ETH_P_IPV6) ?
5131 READ_ONCE(dev->gro_max_size) :
5132 READ_ONCE(dev->gro_ipv4_max_size);
5133}
5134
5135static inline unsigned int
5136netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb)
5137{
5138 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */
5139 return skb->protocol == htons(ETH_P_IPV6) ?
5140 READ_ONCE(dev->gso_max_size) :
5141 READ_ONCE(dev->gso_ipv4_max_size);
5142}
5143
5144static inline bool netif_is_macsec(const struct net_device *dev)
5145{
5146 return dev->priv_flags & IFF_MACSEC;
5147}
5148
5149static inline bool netif_is_macvlan(const struct net_device *dev)
5150{
5151 return dev->priv_flags & IFF_MACVLAN;
5152}
5153
5154static inline bool netif_is_macvlan_port(const struct net_device *dev)
5155{
5156 return dev->priv_flags & IFF_MACVLAN_PORT;
5157}
5158
5159static inline bool netif_is_bond_master(const struct net_device *dev)
5160{
5161 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5162}
5163
5164static inline bool netif_is_bond_slave(const struct net_device *dev)
5165{
5166 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5167}
5168
5169static inline bool netif_supports_nofcs(struct net_device *dev)
5170{
5171 return dev->priv_flags & IFF_SUPP_NOFCS;
5172}
5173
5174static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5175{
5176 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5177}
5178
5179static inline bool netif_is_l3_master(const struct net_device *dev)
5180{
5181 return dev->priv_flags & IFF_L3MDEV_MASTER;
5182}
5183
5184static inline bool netif_is_l3_slave(const struct net_device *dev)
5185{
5186 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5187}
5188
5189static inline int dev_sdif(const struct net_device *dev)
5190{
5191#ifdef CONFIG_NET_L3_MASTER_DEV
5192 if (netif_is_l3_slave(dev))
5193 return dev->ifindex;
5194#endif
5195 return 0;
5196}
5197
5198static inline bool netif_is_bridge_master(const struct net_device *dev)
5199{
5200 return dev->priv_flags & IFF_EBRIDGE;
5201}
5202
5203static inline bool netif_is_bridge_port(const struct net_device *dev)
5204{
5205 return dev->priv_flags & IFF_BRIDGE_PORT;
5206}
5207
5208static inline bool netif_is_ovs_master(const struct net_device *dev)
5209{
5210 return dev->priv_flags & IFF_OPENVSWITCH;
5211}
5212
5213static inline bool netif_is_ovs_port(const struct net_device *dev)
5214{
5215 return dev->priv_flags & IFF_OVS_DATAPATH;
5216}
5217
5218static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5219{
5220 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5221}
5222
5223static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5224{
5225 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5226}
5227
5228static inline bool netif_is_team_master(const struct net_device *dev)
5229{
5230 return dev->priv_flags & IFF_TEAM;
5231}
5232
5233static inline bool netif_is_team_port(const struct net_device *dev)
5234{
5235 return dev->priv_flags & IFF_TEAM_PORT;
5236}
5237
5238static inline bool netif_is_lag_master(const struct net_device *dev)
5239{
5240 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5241}
5242
5243static inline bool netif_is_lag_port(const struct net_device *dev)
5244{
5245 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5246}
5247
5248static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5249{
5250 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5251}
5252
5253static inline bool netif_is_failover(const struct net_device *dev)
5254{
5255 return dev->priv_flags & IFF_FAILOVER;
5256}
5257
5258static inline bool netif_is_failover_slave(const struct net_device *dev)
5259{
5260 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5261}
5262
5263/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5264static inline void netif_keep_dst(struct net_device *dev)
5265{
5266 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5267}
5268
5269/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5270static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5271{
5272 /* TODO: reserve and use an additional IFF bit, if we get more users */
5273 return netif_is_macsec(dev);
5274}
5275
5276extern struct pernet_operations __net_initdata loopback_net_ops;
5277
5278/* Logging, debugging and troubleshooting/diagnostic helpers. */
5279
5280/* netdev_printk helpers, similar to dev_printk */
5281
5282static inline const char *netdev_name(const struct net_device *dev)
5283{
5284 if (!dev->name[0] || strchr(dev->name, '%'))
5285 return "(unnamed net_device)";
5286 return dev->name;
5287}
5288
5289static inline const char *netdev_reg_state(const struct net_device *dev)
5290{
5291 u8 reg_state = READ_ONCE(dev->reg_state);
5292
5293 switch (reg_state) {
5294 case NETREG_UNINITIALIZED: return " (uninitialized)";
5295 case NETREG_REGISTERED: return "";
5296 case NETREG_UNREGISTERING: return " (unregistering)";
5297 case NETREG_UNREGISTERED: return " (unregistered)";
5298 case NETREG_RELEASED: return " (released)";
5299 case NETREG_DUMMY: return " (dummy)";
5300 }
5301
5302 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state);
5303 return " (unknown)";
5304}
5305
5306#define MODULE_ALIAS_NETDEV(device) \
5307 MODULE_ALIAS("netdev-" device)
5308
5309/*
5310 * netdev_WARN() acts like dev_printk(), but with the key difference
5311 * of using a WARN/WARN_ON to get the message out, including the
5312 * file/line information and a backtrace.
5313 */
5314#define netdev_WARN(dev, format, args...) \
5315 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5316 netdev_reg_state(dev), ##args)
5317
5318#define netdev_WARN_ONCE(dev, format, args...) \
5319 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5320 netdev_reg_state(dev), ##args)
5321
5322/*
5323 * The list of packet types we will receive (as opposed to discard)
5324 * and the routines to invoke.
5325 *
5326 * Why 16. Because with 16 the only overlap we get on a hash of the
5327 * low nibble of the protocol value is RARP/SNAP/X.25.
5328 *
5329 * 0800 IP
5330 * 0001 802.3
5331 * 0002 AX.25
5332 * 0004 802.2
5333 * 8035 RARP
5334 * 0005 SNAP
5335 * 0805 X.25
5336 * 0806 ARP
5337 * 8137 IPX
5338 * 0009 Localtalk
5339 * 86DD IPv6
5340 */
5341#define PTYPE_HASH_SIZE (16)
5342#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5343
5344extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5345
5346extern struct net_device *blackhole_netdev;
5347
5348/* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5349#define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5350#define DEV_STATS_ADD(DEV, FIELD, VAL) \
5351 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5352#define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5353
5354#endif /* _LINUX_NETDEVICE_H */