Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
28#include <linux/timer.h>
29#include <linux/bug.h>
30#include <linux/delay.h>
31#include <linux/atomic.h>
32#include <linux/prefetch.h>
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
36#include <linux/percpu.h>
37#include <linux/rculist.h>
38#include <linux/workqueue.h>
39#include <linux/dynamic_queue_limits.h>
40
41#include <linux/ethtool.h>
42#include <net/net_namespace.h>
43#ifdef CONFIG_DCB
44#include <net/dcbnl.h>
45#endif
46#include <net/netprio_cgroup.h>
47#include <net/xdp.h>
48
49#include <linux/netdev_features.h>
50#include <linux/neighbour.h>
51#include <uapi/linux/netdevice.h>
52#include <uapi/linux/if_bonding.h>
53#include <uapi/linux/pkt_cls.h>
54#include <linux/hashtable.h>
55
56struct netpoll_info;
57struct device;
58struct phy_device;
59struct dsa_port;
60
61struct sfp_bus;
62/* 802.11 specific */
63struct wireless_dev;
64/* 802.15.4 specific */
65struct wpan_dev;
66struct mpls_dev;
67/* UDP Tunnel offloads */
68struct udp_tunnel_info;
69struct bpf_prog;
70struct xdp_buff;
71
72void netdev_set_default_ethtool_ops(struct net_device *dev,
73 const struct ethtool_ops *ops);
74
75/* Backlog congestion levels */
76#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
77#define NET_RX_DROP 1 /* packet dropped */
78
79/*
80 * Transmit return codes: transmit return codes originate from three different
81 * namespaces:
82 *
83 * - qdisc return codes
84 * - driver transmit return codes
85 * - errno values
86 *
87 * Drivers are allowed to return any one of those in their hard_start_xmit()
88 * function. Real network devices commonly used with qdiscs should only return
89 * the driver transmit return codes though - when qdiscs are used, the actual
90 * transmission happens asynchronously, so the value is not propagated to
91 * higher layers. Virtual network devices transmit synchronously; in this case
92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
93 * others are propagated to higher layers.
94 */
95
96/* qdisc ->enqueue() return codes. */
97#define NET_XMIT_SUCCESS 0x00
98#define NET_XMIT_DROP 0x01 /* skb dropped */
99#define NET_XMIT_CN 0x02 /* congestion notification */
100#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
101
102/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
103 * indicates that the device will soon be dropping packets, or already drops
104 * some packets of the same priority; prompting us to send less aggressively. */
105#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
106#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
107
108/* Driver transmit return codes */
109#define NETDEV_TX_MASK 0xf0
110
111enum netdev_tx {
112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
113 NETDEV_TX_OK = 0x00, /* driver took care of packet */
114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
115};
116typedef enum netdev_tx netdev_tx_t;
117
118/*
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121 */
122static inline bool dev_xmit_complete(int rc)
123{
124 /*
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 */
130 if (likely(rc < NET_XMIT_MASK))
131 return true;
132
133 return false;
134}
135
136/*
137 * Compute the worst-case header length according to the protocols
138 * used.
139 */
140
141#if defined(CONFIG_HYPERV_NET)
142# define LL_MAX_HEADER 128
143#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
144# if defined(CONFIG_MAC80211_MESH)
145# define LL_MAX_HEADER 128
146# else
147# define LL_MAX_HEADER 96
148# endif
149#else
150# define LL_MAX_HEADER 32
151#endif
152
153#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
155#define MAX_HEADER LL_MAX_HEADER
156#else
157#define MAX_HEADER (LL_MAX_HEADER + 48)
158#endif
159
160/*
161 * Old network device statistics. Fields are native words
162 * (unsigned long) so they can be read and written atomically.
163 */
164
165struct net_device_stats {
166 unsigned long rx_packets;
167 unsigned long tx_packets;
168 unsigned long rx_bytes;
169 unsigned long tx_bytes;
170 unsigned long rx_errors;
171 unsigned long tx_errors;
172 unsigned long rx_dropped;
173 unsigned long tx_dropped;
174 unsigned long multicast;
175 unsigned long collisions;
176 unsigned long rx_length_errors;
177 unsigned long rx_over_errors;
178 unsigned long rx_crc_errors;
179 unsigned long rx_frame_errors;
180 unsigned long rx_fifo_errors;
181 unsigned long rx_missed_errors;
182 unsigned long tx_aborted_errors;
183 unsigned long tx_carrier_errors;
184 unsigned long tx_fifo_errors;
185 unsigned long tx_heartbeat_errors;
186 unsigned long tx_window_errors;
187 unsigned long rx_compressed;
188 unsigned long tx_compressed;
189};
190
191
192#include <linux/cache.h>
193#include <linux/skbuff.h>
194
195#ifdef CONFIG_RPS
196#include <linux/static_key.h>
197extern struct static_key_false rps_needed;
198extern struct static_key_false rfs_needed;
199#endif
200
201struct neighbour;
202struct neigh_parms;
203struct sk_buff;
204
205struct netdev_hw_addr {
206 struct list_head list;
207 unsigned char addr[MAX_ADDR_LEN];
208 unsigned char type;
209#define NETDEV_HW_ADDR_T_LAN 1
210#define NETDEV_HW_ADDR_T_SAN 2
211#define NETDEV_HW_ADDR_T_SLAVE 3
212#define NETDEV_HW_ADDR_T_UNICAST 4
213#define NETDEV_HW_ADDR_T_MULTICAST 5
214 bool global_use;
215 int sync_cnt;
216 int refcount;
217 int synced;
218 struct rcu_head rcu_head;
219};
220
221struct netdev_hw_addr_list {
222 struct list_head list;
223 int count;
224};
225
226#define netdev_hw_addr_list_count(l) ((l)->count)
227#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
228#define netdev_hw_addr_list_for_each(ha, l) \
229 list_for_each_entry(ha, &(l)->list, list)
230
231#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
232#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
233#define netdev_for_each_uc_addr(ha, dev) \
234 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
235
236#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
237#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
238#define netdev_for_each_mc_addr(ha, dev) \
239 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
240
241struct hh_cache {
242 unsigned int hh_len;
243 seqlock_t hh_lock;
244
245 /* cached hardware header; allow for machine alignment needs. */
246#define HH_DATA_MOD 16
247#define HH_DATA_OFF(__len) \
248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
249#define HH_DATA_ALIGN(__len) \
250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
252};
253
254/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
255 * Alternative is:
256 * dev->hard_header_len ? (dev->hard_header_len +
257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
258 *
259 * We could use other alignment values, but we must maintain the
260 * relationship HH alignment <= LL alignment.
261 */
262#define LL_RESERVED_SPACE(dev) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
266
267struct header_ops {
268 int (*create) (struct sk_buff *skb, struct net_device *dev,
269 unsigned short type, const void *daddr,
270 const void *saddr, unsigned int len);
271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
273 void (*cache_update)(struct hh_cache *hh,
274 const struct net_device *dev,
275 const unsigned char *haddr);
276 bool (*validate)(const char *ll_header, unsigned int len);
277 __be16 (*parse_protocol)(const struct sk_buff *skb);
278};
279
280/* These flag bits are private to the generic network queueing
281 * layer; they may not be explicitly referenced by any other
282 * code.
283 */
284
285enum netdev_state_t {
286 __LINK_STATE_START,
287 __LINK_STATE_PRESENT,
288 __LINK_STATE_NOCARRIER,
289 __LINK_STATE_LINKWATCH_PENDING,
290 __LINK_STATE_DORMANT,
291};
292
293
294/*
295 * This structure holds boot-time configured netdevice settings. They
296 * are then used in the device probing.
297 */
298struct netdev_boot_setup {
299 char name[IFNAMSIZ];
300 struct ifmap map;
301};
302#define NETDEV_BOOT_SETUP_MAX 8
303
304int __init netdev_boot_setup(char *str);
305
306struct gro_list {
307 struct list_head list;
308 int count;
309};
310
311/*
312 * size of gro hash buckets, must less than bit number of
313 * napi_struct::gro_bitmask
314 */
315#define GRO_HASH_BUCKETS 8
316
317/*
318 * Structure for NAPI scheduling similar to tasklet but with weighting
319 */
320struct napi_struct {
321 /* The poll_list must only be managed by the entity which
322 * changes the state of the NAPI_STATE_SCHED bit. This means
323 * whoever atomically sets that bit can add this napi_struct
324 * to the per-CPU poll_list, and whoever clears that bit
325 * can remove from the list right before clearing the bit.
326 */
327 struct list_head poll_list;
328
329 unsigned long state;
330 int weight;
331 unsigned long gro_bitmask;
332 int (*poll)(struct napi_struct *, int);
333#ifdef CONFIG_NETPOLL
334 int poll_owner;
335#endif
336 struct net_device *dev;
337 struct gro_list gro_hash[GRO_HASH_BUCKETS];
338 struct sk_buff *skb;
339 struct hrtimer timer;
340 struct list_head dev_list;
341 struct hlist_node napi_hash_node;
342 unsigned int napi_id;
343};
344
345enum {
346 NAPI_STATE_SCHED, /* Poll is scheduled */
347 NAPI_STATE_MISSED, /* reschedule a napi */
348 NAPI_STATE_DISABLE, /* Disable pending */
349 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
350 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
351 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
352 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
353};
354
355enum {
356 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
357 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
358 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
359 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
360 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
361 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
362 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
363};
364
365enum gro_result {
366 GRO_MERGED,
367 GRO_MERGED_FREE,
368 GRO_HELD,
369 GRO_NORMAL,
370 GRO_DROP,
371 GRO_CONSUMED,
372};
373typedef enum gro_result gro_result_t;
374
375/*
376 * enum rx_handler_result - Possible return values for rx_handlers.
377 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
378 * further.
379 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
380 * case skb->dev was changed by rx_handler.
381 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
382 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
383 *
384 * rx_handlers are functions called from inside __netif_receive_skb(), to do
385 * special processing of the skb, prior to delivery to protocol handlers.
386 *
387 * Currently, a net_device can only have a single rx_handler registered. Trying
388 * to register a second rx_handler will return -EBUSY.
389 *
390 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
391 * To unregister a rx_handler on a net_device, use
392 * netdev_rx_handler_unregister().
393 *
394 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
395 * do with the skb.
396 *
397 * If the rx_handler consumed the skb in some way, it should return
398 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
399 * the skb to be delivered in some other way.
400 *
401 * If the rx_handler changed skb->dev, to divert the skb to another
402 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
403 * new device will be called if it exists.
404 *
405 * If the rx_handler decides the skb should be ignored, it should return
406 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
407 * are registered on exact device (ptype->dev == skb->dev).
408 *
409 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
410 * delivered, it should return RX_HANDLER_PASS.
411 *
412 * A device without a registered rx_handler will behave as if rx_handler
413 * returned RX_HANDLER_PASS.
414 */
415
416enum rx_handler_result {
417 RX_HANDLER_CONSUMED,
418 RX_HANDLER_ANOTHER,
419 RX_HANDLER_EXACT,
420 RX_HANDLER_PASS,
421};
422typedef enum rx_handler_result rx_handler_result_t;
423typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
424
425void __napi_schedule(struct napi_struct *n);
426void __napi_schedule_irqoff(struct napi_struct *n);
427
428static inline bool napi_disable_pending(struct napi_struct *n)
429{
430 return test_bit(NAPI_STATE_DISABLE, &n->state);
431}
432
433bool napi_schedule_prep(struct napi_struct *n);
434
435/**
436 * napi_schedule - schedule NAPI poll
437 * @n: NAPI context
438 *
439 * Schedule NAPI poll routine to be called if it is not already
440 * running.
441 */
442static inline void napi_schedule(struct napi_struct *n)
443{
444 if (napi_schedule_prep(n))
445 __napi_schedule(n);
446}
447
448/**
449 * napi_schedule_irqoff - schedule NAPI poll
450 * @n: NAPI context
451 *
452 * Variant of napi_schedule(), assuming hard irqs are masked.
453 */
454static inline void napi_schedule_irqoff(struct napi_struct *n)
455{
456 if (napi_schedule_prep(n))
457 __napi_schedule_irqoff(n);
458}
459
460/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
461static inline bool napi_reschedule(struct napi_struct *napi)
462{
463 if (napi_schedule_prep(napi)) {
464 __napi_schedule(napi);
465 return true;
466 }
467 return false;
468}
469
470bool napi_complete_done(struct napi_struct *n, int work_done);
471/**
472 * napi_complete - NAPI processing complete
473 * @n: NAPI context
474 *
475 * Mark NAPI processing as complete.
476 * Consider using napi_complete_done() instead.
477 * Return false if device should avoid rearming interrupts.
478 */
479static inline bool napi_complete(struct napi_struct *n)
480{
481 return napi_complete_done(n, 0);
482}
483
484/**
485 * napi_hash_del - remove a NAPI from global table
486 * @napi: NAPI context
487 *
488 * Warning: caller must observe RCU grace period
489 * before freeing memory containing @napi, if
490 * this function returns true.
491 * Note: core networking stack automatically calls it
492 * from netif_napi_del().
493 * Drivers might want to call this helper to combine all
494 * the needed RCU grace periods into a single one.
495 */
496bool napi_hash_del(struct napi_struct *napi);
497
498/**
499 * napi_disable - prevent NAPI from scheduling
500 * @n: NAPI context
501 *
502 * Stop NAPI from being scheduled on this context.
503 * Waits till any outstanding processing completes.
504 */
505void napi_disable(struct napi_struct *n);
506
507/**
508 * napi_enable - enable NAPI scheduling
509 * @n: NAPI context
510 *
511 * Resume NAPI from being scheduled on this context.
512 * Must be paired with napi_disable.
513 */
514static inline void napi_enable(struct napi_struct *n)
515{
516 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
517 smp_mb__before_atomic();
518 clear_bit(NAPI_STATE_SCHED, &n->state);
519 clear_bit(NAPI_STATE_NPSVC, &n->state);
520}
521
522/**
523 * napi_synchronize - wait until NAPI is not running
524 * @n: NAPI context
525 *
526 * Wait until NAPI is done being scheduled on this context.
527 * Waits till any outstanding processing completes but
528 * does not disable future activations.
529 */
530static inline void napi_synchronize(const struct napi_struct *n)
531{
532 if (IS_ENABLED(CONFIG_SMP))
533 while (test_bit(NAPI_STATE_SCHED, &n->state))
534 msleep(1);
535 else
536 barrier();
537}
538
539/**
540 * napi_if_scheduled_mark_missed - if napi is running, set the
541 * NAPIF_STATE_MISSED
542 * @n: NAPI context
543 *
544 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
545 * NAPI is scheduled.
546 **/
547static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
548{
549 unsigned long val, new;
550
551 do {
552 val = READ_ONCE(n->state);
553 if (val & NAPIF_STATE_DISABLE)
554 return true;
555
556 if (!(val & NAPIF_STATE_SCHED))
557 return false;
558
559 new = val | NAPIF_STATE_MISSED;
560 } while (cmpxchg(&n->state, val, new) != val);
561
562 return true;
563}
564
565enum netdev_queue_state_t {
566 __QUEUE_STATE_DRV_XOFF,
567 __QUEUE_STATE_STACK_XOFF,
568 __QUEUE_STATE_FROZEN,
569};
570
571#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
572#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
573#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
574
575#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
576#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
577 QUEUE_STATE_FROZEN)
578#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
579 QUEUE_STATE_FROZEN)
580
581/*
582 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
583 * netif_tx_* functions below are used to manipulate this flag. The
584 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
585 * queue independently. The netif_xmit_*stopped functions below are called
586 * to check if the queue has been stopped by the driver or stack (either
587 * of the XOFF bits are set in the state). Drivers should not need to call
588 * netif_xmit*stopped functions, they should only be using netif_tx_*.
589 */
590
591struct netdev_queue {
592/*
593 * read-mostly part
594 */
595 struct net_device *dev;
596 struct Qdisc __rcu *qdisc;
597 struct Qdisc *qdisc_sleeping;
598#ifdef CONFIG_SYSFS
599 struct kobject kobj;
600#endif
601#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
602 int numa_node;
603#endif
604 unsigned long tx_maxrate;
605 /*
606 * Number of TX timeouts for this queue
607 * (/sys/class/net/DEV/Q/trans_timeout)
608 */
609 unsigned long trans_timeout;
610
611 /* Subordinate device that the queue has been assigned to */
612 struct net_device *sb_dev;
613#ifdef CONFIG_XDP_SOCKETS
614 struct xdp_umem *umem;
615#endif
616/*
617 * write-mostly part
618 */
619 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
620 int xmit_lock_owner;
621 /*
622 * Time (in jiffies) of last Tx
623 */
624 unsigned long trans_start;
625
626 unsigned long state;
627
628#ifdef CONFIG_BQL
629 struct dql dql;
630#endif
631} ____cacheline_aligned_in_smp;
632
633extern int sysctl_fb_tunnels_only_for_init_net;
634extern int sysctl_devconf_inherit_init_net;
635
636static inline bool net_has_fallback_tunnels(const struct net *net)
637{
638 return net == &init_net ||
639 !IS_ENABLED(CONFIG_SYSCTL) ||
640 !sysctl_fb_tunnels_only_for_init_net;
641}
642
643static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
644{
645#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
646 return q->numa_node;
647#else
648 return NUMA_NO_NODE;
649#endif
650}
651
652static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
653{
654#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
655 q->numa_node = node;
656#endif
657}
658
659#ifdef CONFIG_RPS
660/*
661 * This structure holds an RPS map which can be of variable length. The
662 * map is an array of CPUs.
663 */
664struct rps_map {
665 unsigned int len;
666 struct rcu_head rcu;
667 u16 cpus[0];
668};
669#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
670
671/*
672 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
673 * tail pointer for that CPU's input queue at the time of last enqueue, and
674 * a hardware filter index.
675 */
676struct rps_dev_flow {
677 u16 cpu;
678 u16 filter;
679 unsigned int last_qtail;
680};
681#define RPS_NO_FILTER 0xffff
682
683/*
684 * The rps_dev_flow_table structure contains a table of flow mappings.
685 */
686struct rps_dev_flow_table {
687 unsigned int mask;
688 struct rcu_head rcu;
689 struct rps_dev_flow flows[0];
690};
691#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
692 ((_num) * sizeof(struct rps_dev_flow)))
693
694/*
695 * The rps_sock_flow_table contains mappings of flows to the last CPU
696 * on which they were processed by the application (set in recvmsg).
697 * Each entry is a 32bit value. Upper part is the high-order bits
698 * of flow hash, lower part is CPU number.
699 * rps_cpu_mask is used to partition the space, depending on number of
700 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
701 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
702 * meaning we use 32-6=26 bits for the hash.
703 */
704struct rps_sock_flow_table {
705 u32 mask;
706
707 u32 ents[0] ____cacheline_aligned_in_smp;
708};
709#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
710
711#define RPS_NO_CPU 0xffff
712
713extern u32 rps_cpu_mask;
714extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
715
716static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
717 u32 hash)
718{
719 if (table && hash) {
720 unsigned int index = hash & table->mask;
721 u32 val = hash & ~rps_cpu_mask;
722
723 /* We only give a hint, preemption can change CPU under us */
724 val |= raw_smp_processor_id();
725
726 if (table->ents[index] != val)
727 table->ents[index] = val;
728 }
729}
730
731#ifdef CONFIG_RFS_ACCEL
732bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
733 u16 filter_id);
734#endif
735#endif /* CONFIG_RPS */
736
737/* This structure contains an instance of an RX queue. */
738struct netdev_rx_queue {
739#ifdef CONFIG_RPS
740 struct rps_map __rcu *rps_map;
741 struct rps_dev_flow_table __rcu *rps_flow_table;
742#endif
743 struct kobject kobj;
744 struct net_device *dev;
745 struct xdp_rxq_info xdp_rxq;
746#ifdef CONFIG_XDP_SOCKETS
747 struct xdp_umem *umem;
748#endif
749} ____cacheline_aligned_in_smp;
750
751/*
752 * RX queue sysfs structures and functions.
753 */
754struct rx_queue_attribute {
755 struct attribute attr;
756 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
757 ssize_t (*store)(struct netdev_rx_queue *queue,
758 const char *buf, size_t len);
759};
760
761#ifdef CONFIG_XPS
762/*
763 * This structure holds an XPS map which can be of variable length. The
764 * map is an array of queues.
765 */
766struct xps_map {
767 unsigned int len;
768 unsigned int alloc_len;
769 struct rcu_head rcu;
770 u16 queues[0];
771};
772#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
773#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
774 - sizeof(struct xps_map)) / sizeof(u16))
775
776/*
777 * This structure holds all XPS maps for device. Maps are indexed by CPU.
778 */
779struct xps_dev_maps {
780 struct rcu_head rcu;
781 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
782};
783
784#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
785 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
786
787#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
788 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
789
790#endif /* CONFIG_XPS */
791
792#define TC_MAX_QUEUE 16
793#define TC_BITMASK 15
794/* HW offloaded queuing disciplines txq count and offset maps */
795struct netdev_tc_txq {
796 u16 count;
797 u16 offset;
798};
799
800#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
801/*
802 * This structure is to hold information about the device
803 * configured to run FCoE protocol stack.
804 */
805struct netdev_fcoe_hbainfo {
806 char manufacturer[64];
807 char serial_number[64];
808 char hardware_version[64];
809 char driver_version[64];
810 char optionrom_version[64];
811 char firmware_version[64];
812 char model[256];
813 char model_description[256];
814};
815#endif
816
817#define MAX_PHYS_ITEM_ID_LEN 32
818
819/* This structure holds a unique identifier to identify some
820 * physical item (port for example) used by a netdevice.
821 */
822struct netdev_phys_item_id {
823 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
824 unsigned char id_len;
825};
826
827static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
828 struct netdev_phys_item_id *b)
829{
830 return a->id_len == b->id_len &&
831 memcmp(a->id, b->id, a->id_len) == 0;
832}
833
834typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
835 struct sk_buff *skb,
836 struct net_device *sb_dev);
837
838enum tc_setup_type {
839 TC_SETUP_QDISC_MQPRIO,
840 TC_SETUP_CLSU32,
841 TC_SETUP_CLSFLOWER,
842 TC_SETUP_CLSMATCHALL,
843 TC_SETUP_CLSBPF,
844 TC_SETUP_BLOCK,
845 TC_SETUP_QDISC_CBS,
846 TC_SETUP_QDISC_RED,
847 TC_SETUP_QDISC_PRIO,
848 TC_SETUP_QDISC_MQ,
849 TC_SETUP_QDISC_ETF,
850 TC_SETUP_ROOT_QDISC,
851 TC_SETUP_QDISC_GRED,
852};
853
854/* These structures hold the attributes of bpf state that are being passed
855 * to the netdevice through the bpf op.
856 */
857enum bpf_netdev_command {
858 /* Set or clear a bpf program used in the earliest stages of packet
859 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
860 * is responsible for calling bpf_prog_put on any old progs that are
861 * stored. In case of error, the callee need not release the new prog
862 * reference, but on success it takes ownership and must bpf_prog_put
863 * when it is no longer used.
864 */
865 XDP_SETUP_PROG,
866 XDP_SETUP_PROG_HW,
867 XDP_QUERY_PROG,
868 XDP_QUERY_PROG_HW,
869 /* BPF program for offload callbacks, invoked at program load time. */
870 BPF_OFFLOAD_MAP_ALLOC,
871 BPF_OFFLOAD_MAP_FREE,
872 XDP_SETUP_XSK_UMEM,
873};
874
875struct bpf_prog_offload_ops;
876struct netlink_ext_ack;
877struct xdp_umem;
878
879struct netdev_bpf {
880 enum bpf_netdev_command command;
881 union {
882 /* XDP_SETUP_PROG */
883 struct {
884 u32 flags;
885 struct bpf_prog *prog;
886 struct netlink_ext_ack *extack;
887 };
888 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
889 struct {
890 u32 prog_id;
891 /* flags with which program was installed */
892 u32 prog_flags;
893 };
894 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
895 struct {
896 struct bpf_offloaded_map *offmap;
897 };
898 /* XDP_SETUP_XSK_UMEM */
899 struct {
900 struct xdp_umem *umem;
901 u16 queue_id;
902 } xsk;
903 };
904};
905
906#ifdef CONFIG_XFRM_OFFLOAD
907struct xfrmdev_ops {
908 int (*xdo_dev_state_add) (struct xfrm_state *x);
909 void (*xdo_dev_state_delete) (struct xfrm_state *x);
910 void (*xdo_dev_state_free) (struct xfrm_state *x);
911 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
912 struct xfrm_state *x);
913 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
914};
915#endif
916
917struct dev_ifalias {
918 struct rcu_head rcuhead;
919 char ifalias[];
920};
921
922struct devlink;
923struct tlsdev_ops;
924
925/*
926 * This structure defines the management hooks for network devices.
927 * The following hooks can be defined; unless noted otherwise, they are
928 * optional and can be filled with a null pointer.
929 *
930 * int (*ndo_init)(struct net_device *dev);
931 * This function is called once when a network device is registered.
932 * The network device can use this for any late stage initialization
933 * or semantic validation. It can fail with an error code which will
934 * be propagated back to register_netdev.
935 *
936 * void (*ndo_uninit)(struct net_device *dev);
937 * This function is called when device is unregistered or when registration
938 * fails. It is not called if init fails.
939 *
940 * int (*ndo_open)(struct net_device *dev);
941 * This function is called when a network device transitions to the up
942 * state.
943 *
944 * int (*ndo_stop)(struct net_device *dev);
945 * This function is called when a network device transitions to the down
946 * state.
947 *
948 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
949 * struct net_device *dev);
950 * Called when a packet needs to be transmitted.
951 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
952 * the queue before that can happen; it's for obsolete devices and weird
953 * corner cases, but the stack really does a non-trivial amount
954 * of useless work if you return NETDEV_TX_BUSY.
955 * Required; cannot be NULL.
956 *
957 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
958 * struct net_device *dev
959 * netdev_features_t features);
960 * Called by core transmit path to determine if device is capable of
961 * performing offload operations on a given packet. This is to give
962 * the device an opportunity to implement any restrictions that cannot
963 * be otherwise expressed by feature flags. The check is called with
964 * the set of features that the stack has calculated and it returns
965 * those the driver believes to be appropriate.
966 *
967 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
968 * struct net_device *sb_dev);
969 * Called to decide which queue to use when device supports multiple
970 * transmit queues.
971 *
972 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
973 * This function is called to allow device receiver to make
974 * changes to configuration when multicast or promiscuous is enabled.
975 *
976 * void (*ndo_set_rx_mode)(struct net_device *dev);
977 * This function is called device changes address list filtering.
978 * If driver handles unicast address filtering, it should set
979 * IFF_UNICAST_FLT in its priv_flags.
980 *
981 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
982 * This function is called when the Media Access Control address
983 * needs to be changed. If this interface is not defined, the
984 * MAC address can not be changed.
985 *
986 * int (*ndo_validate_addr)(struct net_device *dev);
987 * Test if Media Access Control address is valid for the device.
988 *
989 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
990 * Called when a user requests an ioctl which can't be handled by
991 * the generic interface code. If not defined ioctls return
992 * not supported error code.
993 *
994 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
995 * Used to set network devices bus interface parameters. This interface
996 * is retained for legacy reasons; new devices should use the bus
997 * interface (PCI) for low level management.
998 *
999 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1000 * Called when a user wants to change the Maximum Transfer Unit
1001 * of a device.
1002 *
1003 * void (*ndo_tx_timeout)(struct net_device *dev);
1004 * Callback used when the transmitter has not made any progress
1005 * for dev->watchdog ticks.
1006 *
1007 * void (*ndo_get_stats64)(struct net_device *dev,
1008 * struct rtnl_link_stats64 *storage);
1009 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1010 * Called when a user wants to get the network device usage
1011 * statistics. Drivers must do one of the following:
1012 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1013 * rtnl_link_stats64 structure passed by the caller.
1014 * 2. Define @ndo_get_stats to update a net_device_stats structure
1015 * (which should normally be dev->stats) and return a pointer to
1016 * it. The structure may be changed asynchronously only if each
1017 * field is written atomically.
1018 * 3. Update dev->stats asynchronously and atomically, and define
1019 * neither operation.
1020 *
1021 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1022 * Return true if this device supports offload stats of this attr_id.
1023 *
1024 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1025 * void *attr_data)
1026 * Get statistics for offload operations by attr_id. Write it into the
1027 * attr_data pointer.
1028 *
1029 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1030 * If device supports VLAN filtering this function is called when a
1031 * VLAN id is registered.
1032 *
1033 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1034 * If device supports VLAN filtering this function is called when a
1035 * VLAN id is unregistered.
1036 *
1037 * void (*ndo_poll_controller)(struct net_device *dev);
1038 *
1039 * SR-IOV management functions.
1040 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1041 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1042 * u8 qos, __be16 proto);
1043 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1044 * int max_tx_rate);
1045 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1046 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1047 * int (*ndo_get_vf_config)(struct net_device *dev,
1048 * int vf, struct ifla_vf_info *ivf);
1049 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1050 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1051 * struct nlattr *port[]);
1052 *
1053 * Enable or disable the VF ability to query its RSS Redirection Table and
1054 * Hash Key. This is needed since on some devices VF share this information
1055 * with PF and querying it may introduce a theoretical security risk.
1056 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1057 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1058 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1059 * void *type_data);
1060 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1061 * This is always called from the stack with the rtnl lock held and netif
1062 * tx queues stopped. This allows the netdevice to perform queue
1063 * management safely.
1064 *
1065 * Fiber Channel over Ethernet (FCoE) offload functions.
1066 * int (*ndo_fcoe_enable)(struct net_device *dev);
1067 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1068 * so the underlying device can perform whatever needed configuration or
1069 * initialization to support acceleration of FCoE traffic.
1070 *
1071 * int (*ndo_fcoe_disable)(struct net_device *dev);
1072 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1073 * so the underlying device can perform whatever needed clean-ups to
1074 * stop supporting acceleration of FCoE traffic.
1075 *
1076 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1077 * struct scatterlist *sgl, unsigned int sgc);
1078 * Called when the FCoE Initiator wants to initialize an I/O that
1079 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1080 * perform necessary setup and returns 1 to indicate the device is set up
1081 * successfully to perform DDP on this I/O, otherwise this returns 0.
1082 *
1083 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1084 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1085 * indicated by the FC exchange id 'xid', so the underlying device can
1086 * clean up and reuse resources for later DDP requests.
1087 *
1088 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1089 * struct scatterlist *sgl, unsigned int sgc);
1090 * Called when the FCoE Target wants to initialize an I/O that
1091 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1092 * perform necessary setup and returns 1 to indicate the device is set up
1093 * successfully to perform DDP on this I/O, otherwise this returns 0.
1094 *
1095 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1096 * struct netdev_fcoe_hbainfo *hbainfo);
1097 * Called when the FCoE Protocol stack wants information on the underlying
1098 * device. This information is utilized by the FCoE protocol stack to
1099 * register attributes with Fiber Channel management service as per the
1100 * FC-GS Fabric Device Management Information(FDMI) specification.
1101 *
1102 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1103 * Called when the underlying device wants to override default World Wide
1104 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1105 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1106 * protocol stack to use.
1107 *
1108 * RFS acceleration.
1109 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1110 * u16 rxq_index, u32 flow_id);
1111 * Set hardware filter for RFS. rxq_index is the target queue index;
1112 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1113 * Return the filter ID on success, or a negative error code.
1114 *
1115 * Slave management functions (for bridge, bonding, etc).
1116 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1117 * Called to make another netdev an underling.
1118 *
1119 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1120 * Called to release previously enslaved netdev.
1121 *
1122 * Feature/offload setting functions.
1123 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1124 * netdev_features_t features);
1125 * Adjusts the requested feature flags according to device-specific
1126 * constraints, and returns the resulting flags. Must not modify
1127 * the device state.
1128 *
1129 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1130 * Called to update device configuration to new features. Passed
1131 * feature set might be less than what was returned by ndo_fix_features()).
1132 * Must return >0 or -errno if it changed dev->features itself.
1133 *
1134 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1135 * struct net_device *dev,
1136 * const unsigned char *addr, u16 vid, u16 flags,
1137 * struct netlink_ext_ack *extack);
1138 * Adds an FDB entry to dev for addr.
1139 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1140 * struct net_device *dev,
1141 * const unsigned char *addr, u16 vid)
1142 * Deletes the FDB entry from dev coresponding to addr.
1143 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1144 * struct net_device *dev, struct net_device *filter_dev,
1145 * int *idx)
1146 * Used to add FDB entries to dump requests. Implementers should add
1147 * entries to skb and update idx with the number of entries.
1148 *
1149 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1150 * u16 flags, struct netlink_ext_ack *extack)
1151 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1152 * struct net_device *dev, u32 filter_mask,
1153 * int nlflags)
1154 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1155 * u16 flags);
1156 *
1157 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1158 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1159 * which do not represent real hardware may define this to allow their
1160 * userspace components to manage their virtual carrier state. Devices
1161 * that determine carrier state from physical hardware properties (eg
1162 * network cables) or protocol-dependent mechanisms (eg
1163 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1164 *
1165 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1166 * struct netdev_phys_item_id *ppid);
1167 * Called to get ID of physical port of this device. If driver does
1168 * not implement this, it is assumed that the hw is not able to have
1169 * multiple net devices on single physical port.
1170 *
1171 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1172 * struct netdev_phys_item_id *ppid)
1173 * Called to get the parent ID of the physical port of this device.
1174 *
1175 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1176 * struct udp_tunnel_info *ti);
1177 * Called by UDP tunnel to notify a driver about the UDP port and socket
1178 * address family that a UDP tunnel is listnening to. It is called only
1179 * when a new port starts listening. The operation is protected by the
1180 * RTNL.
1181 *
1182 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1183 * struct udp_tunnel_info *ti);
1184 * Called by UDP tunnel to notify the driver about a UDP port and socket
1185 * address family that the UDP tunnel is not listening to anymore. The
1186 * operation is protected by the RTNL.
1187 *
1188 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1189 * struct net_device *dev)
1190 * Called by upper layer devices to accelerate switching or other
1191 * station functionality into hardware. 'pdev is the lowerdev
1192 * to use for the offload and 'dev' is the net device that will
1193 * back the offload. Returns a pointer to the private structure
1194 * the upper layer will maintain.
1195 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1196 * Called by upper layer device to delete the station created
1197 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1198 * the station and priv is the structure returned by the add
1199 * operation.
1200 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1201 * int queue_index, u32 maxrate);
1202 * Called when a user wants to set a max-rate limitation of specific
1203 * TX queue.
1204 * int (*ndo_get_iflink)(const struct net_device *dev);
1205 * Called to get the iflink value of this device.
1206 * void (*ndo_change_proto_down)(struct net_device *dev,
1207 * bool proto_down);
1208 * This function is used to pass protocol port error state information
1209 * to the switch driver. The switch driver can react to the proto_down
1210 * by doing a phys down on the associated switch port.
1211 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1212 * This function is used to get egress tunnel information for given skb.
1213 * This is useful for retrieving outer tunnel header parameters while
1214 * sampling packet.
1215 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1216 * This function is used to specify the headroom that the skb must
1217 * consider when allocation skb during packet reception. Setting
1218 * appropriate rx headroom value allows avoiding skb head copy on
1219 * forward. Setting a negative value resets the rx headroom to the
1220 * default value.
1221 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1222 * This function is used to set or query state related to XDP on the
1223 * netdevice and manage BPF offload. See definition of
1224 * enum bpf_netdev_command for details.
1225 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1226 * u32 flags);
1227 * This function is used to submit @n XDP packets for transmit on a
1228 * netdevice. Returns number of frames successfully transmitted, frames
1229 * that got dropped are freed/returned via xdp_return_frame().
1230 * Returns negative number, means general error invoking ndo, meaning
1231 * no frames were xmit'ed and core-caller will free all frames.
1232 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1233 * Get devlink port instance associated with a given netdev.
1234 * Called with a reference on the netdevice and devlink locks only,
1235 * rtnl_lock is not held.
1236 */
1237struct net_device_ops {
1238 int (*ndo_init)(struct net_device *dev);
1239 void (*ndo_uninit)(struct net_device *dev);
1240 int (*ndo_open)(struct net_device *dev);
1241 int (*ndo_stop)(struct net_device *dev);
1242 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1243 struct net_device *dev);
1244 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1245 struct net_device *dev,
1246 netdev_features_t features);
1247 u16 (*ndo_select_queue)(struct net_device *dev,
1248 struct sk_buff *skb,
1249 struct net_device *sb_dev);
1250 void (*ndo_change_rx_flags)(struct net_device *dev,
1251 int flags);
1252 void (*ndo_set_rx_mode)(struct net_device *dev);
1253 int (*ndo_set_mac_address)(struct net_device *dev,
1254 void *addr);
1255 int (*ndo_validate_addr)(struct net_device *dev);
1256 int (*ndo_do_ioctl)(struct net_device *dev,
1257 struct ifreq *ifr, int cmd);
1258 int (*ndo_set_config)(struct net_device *dev,
1259 struct ifmap *map);
1260 int (*ndo_change_mtu)(struct net_device *dev,
1261 int new_mtu);
1262 int (*ndo_neigh_setup)(struct net_device *dev,
1263 struct neigh_parms *);
1264 void (*ndo_tx_timeout) (struct net_device *dev);
1265
1266 void (*ndo_get_stats64)(struct net_device *dev,
1267 struct rtnl_link_stats64 *storage);
1268 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1269 int (*ndo_get_offload_stats)(int attr_id,
1270 const struct net_device *dev,
1271 void *attr_data);
1272 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1273
1274 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1275 __be16 proto, u16 vid);
1276 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1277 __be16 proto, u16 vid);
1278#ifdef CONFIG_NET_POLL_CONTROLLER
1279 void (*ndo_poll_controller)(struct net_device *dev);
1280 int (*ndo_netpoll_setup)(struct net_device *dev,
1281 struct netpoll_info *info);
1282 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1283#endif
1284 int (*ndo_set_vf_mac)(struct net_device *dev,
1285 int queue, u8 *mac);
1286 int (*ndo_set_vf_vlan)(struct net_device *dev,
1287 int queue, u16 vlan,
1288 u8 qos, __be16 proto);
1289 int (*ndo_set_vf_rate)(struct net_device *dev,
1290 int vf, int min_tx_rate,
1291 int max_tx_rate);
1292 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1293 int vf, bool setting);
1294 int (*ndo_set_vf_trust)(struct net_device *dev,
1295 int vf, bool setting);
1296 int (*ndo_get_vf_config)(struct net_device *dev,
1297 int vf,
1298 struct ifla_vf_info *ivf);
1299 int (*ndo_set_vf_link_state)(struct net_device *dev,
1300 int vf, int link_state);
1301 int (*ndo_get_vf_stats)(struct net_device *dev,
1302 int vf,
1303 struct ifla_vf_stats
1304 *vf_stats);
1305 int (*ndo_set_vf_port)(struct net_device *dev,
1306 int vf,
1307 struct nlattr *port[]);
1308 int (*ndo_get_vf_port)(struct net_device *dev,
1309 int vf, struct sk_buff *skb);
1310 int (*ndo_set_vf_guid)(struct net_device *dev,
1311 int vf, u64 guid,
1312 int guid_type);
1313 int (*ndo_set_vf_rss_query_en)(
1314 struct net_device *dev,
1315 int vf, bool setting);
1316 int (*ndo_setup_tc)(struct net_device *dev,
1317 enum tc_setup_type type,
1318 void *type_data);
1319#if IS_ENABLED(CONFIG_FCOE)
1320 int (*ndo_fcoe_enable)(struct net_device *dev);
1321 int (*ndo_fcoe_disable)(struct net_device *dev);
1322 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1323 u16 xid,
1324 struct scatterlist *sgl,
1325 unsigned int sgc);
1326 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1327 u16 xid);
1328 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1329 u16 xid,
1330 struct scatterlist *sgl,
1331 unsigned int sgc);
1332 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1333 struct netdev_fcoe_hbainfo *hbainfo);
1334#endif
1335
1336#if IS_ENABLED(CONFIG_LIBFCOE)
1337#define NETDEV_FCOE_WWNN 0
1338#define NETDEV_FCOE_WWPN 1
1339 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1340 u64 *wwn, int type);
1341#endif
1342
1343#ifdef CONFIG_RFS_ACCEL
1344 int (*ndo_rx_flow_steer)(struct net_device *dev,
1345 const struct sk_buff *skb,
1346 u16 rxq_index,
1347 u32 flow_id);
1348#endif
1349 int (*ndo_add_slave)(struct net_device *dev,
1350 struct net_device *slave_dev,
1351 struct netlink_ext_ack *extack);
1352 int (*ndo_del_slave)(struct net_device *dev,
1353 struct net_device *slave_dev);
1354 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1355 netdev_features_t features);
1356 int (*ndo_set_features)(struct net_device *dev,
1357 netdev_features_t features);
1358 int (*ndo_neigh_construct)(struct net_device *dev,
1359 struct neighbour *n);
1360 void (*ndo_neigh_destroy)(struct net_device *dev,
1361 struct neighbour *n);
1362
1363 int (*ndo_fdb_add)(struct ndmsg *ndm,
1364 struct nlattr *tb[],
1365 struct net_device *dev,
1366 const unsigned char *addr,
1367 u16 vid,
1368 u16 flags,
1369 struct netlink_ext_ack *extack);
1370 int (*ndo_fdb_del)(struct ndmsg *ndm,
1371 struct nlattr *tb[],
1372 struct net_device *dev,
1373 const unsigned char *addr,
1374 u16 vid);
1375 int (*ndo_fdb_dump)(struct sk_buff *skb,
1376 struct netlink_callback *cb,
1377 struct net_device *dev,
1378 struct net_device *filter_dev,
1379 int *idx);
1380 int (*ndo_fdb_get)(struct sk_buff *skb,
1381 struct nlattr *tb[],
1382 struct net_device *dev,
1383 const unsigned char *addr,
1384 u16 vid, u32 portid, u32 seq,
1385 struct netlink_ext_ack *extack);
1386 int (*ndo_bridge_setlink)(struct net_device *dev,
1387 struct nlmsghdr *nlh,
1388 u16 flags,
1389 struct netlink_ext_ack *extack);
1390 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1391 u32 pid, u32 seq,
1392 struct net_device *dev,
1393 u32 filter_mask,
1394 int nlflags);
1395 int (*ndo_bridge_dellink)(struct net_device *dev,
1396 struct nlmsghdr *nlh,
1397 u16 flags);
1398 int (*ndo_change_carrier)(struct net_device *dev,
1399 bool new_carrier);
1400 int (*ndo_get_phys_port_id)(struct net_device *dev,
1401 struct netdev_phys_item_id *ppid);
1402 int (*ndo_get_port_parent_id)(struct net_device *dev,
1403 struct netdev_phys_item_id *ppid);
1404 int (*ndo_get_phys_port_name)(struct net_device *dev,
1405 char *name, size_t len);
1406 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1407 struct udp_tunnel_info *ti);
1408 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1409 struct udp_tunnel_info *ti);
1410 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1411 struct net_device *dev);
1412 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1413 void *priv);
1414
1415 int (*ndo_get_lock_subclass)(struct net_device *dev);
1416 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1417 int queue_index,
1418 u32 maxrate);
1419 int (*ndo_get_iflink)(const struct net_device *dev);
1420 int (*ndo_change_proto_down)(struct net_device *dev,
1421 bool proto_down);
1422 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1423 struct sk_buff *skb);
1424 void (*ndo_set_rx_headroom)(struct net_device *dev,
1425 int needed_headroom);
1426 int (*ndo_bpf)(struct net_device *dev,
1427 struct netdev_bpf *bpf);
1428 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1429 struct xdp_frame **xdp,
1430 u32 flags);
1431 int (*ndo_xsk_async_xmit)(struct net_device *dev,
1432 u32 queue_id);
1433 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1434};
1435
1436/**
1437 * enum net_device_priv_flags - &struct net_device priv_flags
1438 *
1439 * These are the &struct net_device, they are only set internally
1440 * by drivers and used in the kernel. These flags are invisible to
1441 * userspace; this means that the order of these flags can change
1442 * during any kernel release.
1443 *
1444 * You should have a pretty good reason to be extending these flags.
1445 *
1446 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1447 * @IFF_EBRIDGE: Ethernet bridging device
1448 * @IFF_BONDING: bonding master or slave
1449 * @IFF_ISATAP: ISATAP interface (RFC4214)
1450 * @IFF_WAN_HDLC: WAN HDLC device
1451 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1452 * release skb->dst
1453 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1454 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1455 * @IFF_MACVLAN_PORT: device used as macvlan port
1456 * @IFF_BRIDGE_PORT: device used as bridge port
1457 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1458 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1459 * @IFF_UNICAST_FLT: Supports unicast filtering
1460 * @IFF_TEAM_PORT: device used as team port
1461 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1462 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1463 * change when it's running
1464 * @IFF_MACVLAN: Macvlan device
1465 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1466 * underlying stacked devices
1467 * @IFF_L3MDEV_MASTER: device is an L3 master device
1468 * @IFF_NO_QUEUE: device can run without qdisc attached
1469 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1470 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1471 * @IFF_TEAM: device is a team device
1472 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1473 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1474 * entity (i.e. the master device for bridged veth)
1475 * @IFF_MACSEC: device is a MACsec device
1476 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1477 * @IFF_FAILOVER: device is a failover master device
1478 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1479 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1480 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1481 */
1482enum netdev_priv_flags {
1483 IFF_802_1Q_VLAN = 1<<0,
1484 IFF_EBRIDGE = 1<<1,
1485 IFF_BONDING = 1<<2,
1486 IFF_ISATAP = 1<<3,
1487 IFF_WAN_HDLC = 1<<4,
1488 IFF_XMIT_DST_RELEASE = 1<<5,
1489 IFF_DONT_BRIDGE = 1<<6,
1490 IFF_DISABLE_NETPOLL = 1<<7,
1491 IFF_MACVLAN_PORT = 1<<8,
1492 IFF_BRIDGE_PORT = 1<<9,
1493 IFF_OVS_DATAPATH = 1<<10,
1494 IFF_TX_SKB_SHARING = 1<<11,
1495 IFF_UNICAST_FLT = 1<<12,
1496 IFF_TEAM_PORT = 1<<13,
1497 IFF_SUPP_NOFCS = 1<<14,
1498 IFF_LIVE_ADDR_CHANGE = 1<<15,
1499 IFF_MACVLAN = 1<<16,
1500 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1501 IFF_L3MDEV_MASTER = 1<<18,
1502 IFF_NO_QUEUE = 1<<19,
1503 IFF_OPENVSWITCH = 1<<20,
1504 IFF_L3MDEV_SLAVE = 1<<21,
1505 IFF_TEAM = 1<<22,
1506 IFF_RXFH_CONFIGURED = 1<<23,
1507 IFF_PHONY_HEADROOM = 1<<24,
1508 IFF_MACSEC = 1<<25,
1509 IFF_NO_RX_HANDLER = 1<<26,
1510 IFF_FAILOVER = 1<<27,
1511 IFF_FAILOVER_SLAVE = 1<<28,
1512 IFF_L3MDEV_RX_HANDLER = 1<<29,
1513 IFF_LIVE_RENAME_OK = 1<<30,
1514};
1515
1516#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1517#define IFF_EBRIDGE IFF_EBRIDGE
1518#define IFF_BONDING IFF_BONDING
1519#define IFF_ISATAP IFF_ISATAP
1520#define IFF_WAN_HDLC IFF_WAN_HDLC
1521#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1522#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1523#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1524#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1525#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1526#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1527#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1528#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1529#define IFF_TEAM_PORT IFF_TEAM_PORT
1530#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1531#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1532#define IFF_MACVLAN IFF_MACVLAN
1533#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1534#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1535#define IFF_NO_QUEUE IFF_NO_QUEUE
1536#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1537#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1538#define IFF_TEAM IFF_TEAM
1539#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1540#define IFF_MACSEC IFF_MACSEC
1541#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1542#define IFF_FAILOVER IFF_FAILOVER
1543#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1544#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1545#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1546
1547/**
1548 * struct net_device - The DEVICE structure.
1549 *
1550 * Actually, this whole structure is a big mistake. It mixes I/O
1551 * data with strictly "high-level" data, and it has to know about
1552 * almost every data structure used in the INET module.
1553 *
1554 * @name: This is the first field of the "visible" part of this structure
1555 * (i.e. as seen by users in the "Space.c" file). It is the name
1556 * of the interface.
1557 *
1558 * @name_hlist: Device name hash chain, please keep it close to name[]
1559 * @ifalias: SNMP alias
1560 * @mem_end: Shared memory end
1561 * @mem_start: Shared memory start
1562 * @base_addr: Device I/O address
1563 * @irq: Device IRQ number
1564 *
1565 * @state: Generic network queuing layer state, see netdev_state_t
1566 * @dev_list: The global list of network devices
1567 * @napi_list: List entry used for polling NAPI devices
1568 * @unreg_list: List entry when we are unregistering the
1569 * device; see the function unregister_netdev
1570 * @close_list: List entry used when we are closing the device
1571 * @ptype_all: Device-specific packet handlers for all protocols
1572 * @ptype_specific: Device-specific, protocol-specific packet handlers
1573 *
1574 * @adj_list: Directly linked devices, like slaves for bonding
1575 * @features: Currently active device features
1576 * @hw_features: User-changeable features
1577 *
1578 * @wanted_features: User-requested features
1579 * @vlan_features: Mask of features inheritable by VLAN devices
1580 *
1581 * @hw_enc_features: Mask of features inherited by encapsulating devices
1582 * This field indicates what encapsulation
1583 * offloads the hardware is capable of doing,
1584 * and drivers will need to set them appropriately.
1585 *
1586 * @mpls_features: Mask of features inheritable by MPLS
1587 *
1588 * @ifindex: interface index
1589 * @group: The group the device belongs to
1590 *
1591 * @stats: Statistics struct, which was left as a legacy, use
1592 * rtnl_link_stats64 instead
1593 *
1594 * @rx_dropped: Dropped packets by core network,
1595 * do not use this in drivers
1596 * @tx_dropped: Dropped packets by core network,
1597 * do not use this in drivers
1598 * @rx_nohandler: nohandler dropped packets by core network on
1599 * inactive devices, do not use this in drivers
1600 * @carrier_up_count: Number of times the carrier has been up
1601 * @carrier_down_count: Number of times the carrier has been down
1602 *
1603 * @wireless_handlers: List of functions to handle Wireless Extensions,
1604 * instead of ioctl,
1605 * see <net/iw_handler.h> for details.
1606 * @wireless_data: Instance data managed by the core of wireless extensions
1607 *
1608 * @netdev_ops: Includes several pointers to callbacks,
1609 * if one wants to override the ndo_*() functions
1610 * @ethtool_ops: Management operations
1611 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1612 * discovery handling. Necessary for e.g. 6LoWPAN.
1613 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1614 * of Layer 2 headers.
1615 *
1616 * @flags: Interface flags (a la BSD)
1617 * @priv_flags: Like 'flags' but invisible to userspace,
1618 * see if.h for the definitions
1619 * @gflags: Global flags ( kept as legacy )
1620 * @padded: How much padding added by alloc_netdev()
1621 * @operstate: RFC2863 operstate
1622 * @link_mode: Mapping policy to operstate
1623 * @if_port: Selectable AUI, TP, ...
1624 * @dma: DMA channel
1625 * @mtu: Interface MTU value
1626 * @min_mtu: Interface Minimum MTU value
1627 * @max_mtu: Interface Maximum MTU value
1628 * @type: Interface hardware type
1629 * @hard_header_len: Maximum hardware header length.
1630 * @min_header_len: Minimum hardware header length
1631 *
1632 * @needed_headroom: Extra headroom the hardware may need, but not in all
1633 * cases can this be guaranteed
1634 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1635 * cases can this be guaranteed. Some cases also use
1636 * LL_MAX_HEADER instead to allocate the skb
1637 *
1638 * interface address info:
1639 *
1640 * @perm_addr: Permanent hw address
1641 * @addr_assign_type: Hw address assignment type
1642 * @addr_len: Hardware address length
1643 * @neigh_priv_len: Used in neigh_alloc()
1644 * @dev_id: Used to differentiate devices that share
1645 * the same link layer address
1646 * @dev_port: Used to differentiate devices that share
1647 * the same function
1648 * @addr_list_lock: XXX: need comments on this one
1649 * @uc_promisc: Counter that indicates promiscuous mode
1650 * has been enabled due to the need to listen to
1651 * additional unicast addresses in a device that
1652 * does not implement ndo_set_rx_mode()
1653 * @uc: unicast mac addresses
1654 * @mc: multicast mac addresses
1655 * @dev_addrs: list of device hw addresses
1656 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1657 * @promiscuity: Number of times the NIC is told to work in
1658 * promiscuous mode; if it becomes 0 the NIC will
1659 * exit promiscuous mode
1660 * @allmulti: Counter, enables or disables allmulticast mode
1661 *
1662 * @vlan_info: VLAN info
1663 * @dsa_ptr: dsa specific data
1664 * @tipc_ptr: TIPC specific data
1665 * @atalk_ptr: AppleTalk link
1666 * @ip_ptr: IPv4 specific data
1667 * @dn_ptr: DECnet specific data
1668 * @ip6_ptr: IPv6 specific data
1669 * @ax25_ptr: AX.25 specific data
1670 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1671 *
1672 * @dev_addr: Hw address (before bcast,
1673 * because most packets are unicast)
1674 *
1675 * @_rx: Array of RX queues
1676 * @num_rx_queues: Number of RX queues
1677 * allocated at register_netdev() time
1678 * @real_num_rx_queues: Number of RX queues currently active in device
1679 *
1680 * @rx_handler: handler for received packets
1681 * @rx_handler_data: XXX: need comments on this one
1682 * @miniq_ingress: ingress/clsact qdisc specific data for
1683 * ingress processing
1684 * @ingress_queue: XXX: need comments on this one
1685 * @broadcast: hw bcast address
1686 *
1687 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1688 * indexed by RX queue number. Assigned by driver.
1689 * This must only be set if the ndo_rx_flow_steer
1690 * operation is defined
1691 * @index_hlist: Device index hash chain
1692 *
1693 * @_tx: Array of TX queues
1694 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1695 * @real_num_tx_queues: Number of TX queues currently active in device
1696 * @qdisc: Root qdisc from userspace point of view
1697 * @tx_queue_len: Max frames per queue allowed
1698 * @tx_global_lock: XXX: need comments on this one
1699 *
1700 * @xps_maps: XXX: need comments on this one
1701 * @miniq_egress: clsact qdisc specific data for
1702 * egress processing
1703 * @watchdog_timeo: Represents the timeout that is used by
1704 * the watchdog (see dev_watchdog())
1705 * @watchdog_timer: List of timers
1706 *
1707 * @pcpu_refcnt: Number of references to this device
1708 * @todo_list: Delayed register/unregister
1709 * @link_watch_list: XXX: need comments on this one
1710 *
1711 * @reg_state: Register/unregister state machine
1712 * @dismantle: Device is going to be freed
1713 * @rtnl_link_state: This enum represents the phases of creating
1714 * a new link
1715 *
1716 * @needs_free_netdev: Should unregister perform free_netdev?
1717 * @priv_destructor: Called from unregister
1718 * @npinfo: XXX: need comments on this one
1719 * @nd_net: Network namespace this network device is inside
1720 *
1721 * @ml_priv: Mid-layer private
1722 * @lstats: Loopback statistics
1723 * @tstats: Tunnel statistics
1724 * @dstats: Dummy statistics
1725 * @vstats: Virtual ethernet statistics
1726 *
1727 * @garp_port: GARP
1728 * @mrp_port: MRP
1729 *
1730 * @dev: Class/net/name entry
1731 * @sysfs_groups: Space for optional device, statistics and wireless
1732 * sysfs groups
1733 *
1734 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1735 * @rtnl_link_ops: Rtnl_link_ops
1736 *
1737 * @gso_max_size: Maximum size of generic segmentation offload
1738 * @gso_max_segs: Maximum number of segments that can be passed to the
1739 * NIC for GSO
1740 *
1741 * @dcbnl_ops: Data Center Bridging netlink ops
1742 * @num_tc: Number of traffic classes in the net device
1743 * @tc_to_txq: XXX: need comments on this one
1744 * @prio_tc_map: XXX: need comments on this one
1745 *
1746 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1747 *
1748 * @priomap: XXX: need comments on this one
1749 * @phydev: Physical device may attach itself
1750 * for hardware timestamping
1751 * @sfp_bus: attached &struct sfp_bus structure.
1752 *
1753 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1754 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1755 *
1756 * @proto_down: protocol port state information can be sent to the
1757 * switch driver and used to set the phys state of the
1758 * switch port.
1759 *
1760 * @wol_enabled: Wake-on-LAN is enabled
1761 *
1762 * FIXME: cleanup struct net_device such that network protocol info
1763 * moves out.
1764 */
1765
1766struct net_device {
1767 char name[IFNAMSIZ];
1768 struct hlist_node name_hlist;
1769 struct dev_ifalias __rcu *ifalias;
1770 /*
1771 * I/O specific fields
1772 * FIXME: Merge these and struct ifmap into one
1773 */
1774 unsigned long mem_end;
1775 unsigned long mem_start;
1776 unsigned long base_addr;
1777 int irq;
1778
1779 /*
1780 * Some hardware also needs these fields (state,dev_list,
1781 * napi_list,unreg_list,close_list) but they are not
1782 * part of the usual set specified in Space.c.
1783 */
1784
1785 unsigned long state;
1786
1787 struct list_head dev_list;
1788 struct list_head napi_list;
1789 struct list_head unreg_list;
1790 struct list_head close_list;
1791 struct list_head ptype_all;
1792 struct list_head ptype_specific;
1793
1794 struct {
1795 struct list_head upper;
1796 struct list_head lower;
1797 } adj_list;
1798
1799 netdev_features_t features;
1800 netdev_features_t hw_features;
1801 netdev_features_t wanted_features;
1802 netdev_features_t vlan_features;
1803 netdev_features_t hw_enc_features;
1804 netdev_features_t mpls_features;
1805 netdev_features_t gso_partial_features;
1806
1807 int ifindex;
1808 int group;
1809
1810 struct net_device_stats stats;
1811
1812 atomic_long_t rx_dropped;
1813 atomic_long_t tx_dropped;
1814 atomic_long_t rx_nohandler;
1815
1816 /* Stats to monitor link on/off, flapping */
1817 atomic_t carrier_up_count;
1818 atomic_t carrier_down_count;
1819
1820#ifdef CONFIG_WIRELESS_EXT
1821 const struct iw_handler_def *wireless_handlers;
1822 struct iw_public_data *wireless_data;
1823#endif
1824 const struct net_device_ops *netdev_ops;
1825 const struct ethtool_ops *ethtool_ops;
1826#ifdef CONFIG_NET_L3_MASTER_DEV
1827 const struct l3mdev_ops *l3mdev_ops;
1828#endif
1829#if IS_ENABLED(CONFIG_IPV6)
1830 const struct ndisc_ops *ndisc_ops;
1831#endif
1832
1833#ifdef CONFIG_XFRM_OFFLOAD
1834 const struct xfrmdev_ops *xfrmdev_ops;
1835#endif
1836
1837#if IS_ENABLED(CONFIG_TLS_DEVICE)
1838 const struct tlsdev_ops *tlsdev_ops;
1839#endif
1840
1841 const struct header_ops *header_ops;
1842
1843 unsigned int flags;
1844 unsigned int priv_flags;
1845
1846 unsigned short gflags;
1847 unsigned short padded;
1848
1849 unsigned char operstate;
1850 unsigned char link_mode;
1851
1852 unsigned char if_port;
1853 unsigned char dma;
1854
1855 unsigned int mtu;
1856 unsigned int min_mtu;
1857 unsigned int max_mtu;
1858 unsigned short type;
1859 unsigned short hard_header_len;
1860 unsigned char min_header_len;
1861
1862 unsigned short needed_headroom;
1863 unsigned short needed_tailroom;
1864
1865 /* Interface address info. */
1866 unsigned char perm_addr[MAX_ADDR_LEN];
1867 unsigned char addr_assign_type;
1868 unsigned char addr_len;
1869 unsigned short neigh_priv_len;
1870 unsigned short dev_id;
1871 unsigned short dev_port;
1872 spinlock_t addr_list_lock;
1873 unsigned char name_assign_type;
1874 bool uc_promisc;
1875 struct netdev_hw_addr_list uc;
1876 struct netdev_hw_addr_list mc;
1877 struct netdev_hw_addr_list dev_addrs;
1878
1879#ifdef CONFIG_SYSFS
1880 struct kset *queues_kset;
1881#endif
1882 unsigned int promiscuity;
1883 unsigned int allmulti;
1884
1885
1886 /* Protocol-specific pointers */
1887
1888#if IS_ENABLED(CONFIG_VLAN_8021Q)
1889 struct vlan_info __rcu *vlan_info;
1890#endif
1891#if IS_ENABLED(CONFIG_NET_DSA)
1892 struct dsa_port *dsa_ptr;
1893#endif
1894#if IS_ENABLED(CONFIG_TIPC)
1895 struct tipc_bearer __rcu *tipc_ptr;
1896#endif
1897#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1898 void *atalk_ptr;
1899#endif
1900 struct in_device __rcu *ip_ptr;
1901#if IS_ENABLED(CONFIG_DECNET)
1902 struct dn_dev __rcu *dn_ptr;
1903#endif
1904 struct inet6_dev __rcu *ip6_ptr;
1905#if IS_ENABLED(CONFIG_AX25)
1906 void *ax25_ptr;
1907#endif
1908 struct wireless_dev *ieee80211_ptr;
1909 struct wpan_dev *ieee802154_ptr;
1910#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1911 struct mpls_dev __rcu *mpls_ptr;
1912#endif
1913
1914/*
1915 * Cache lines mostly used on receive path (including eth_type_trans())
1916 */
1917 /* Interface address info used in eth_type_trans() */
1918 unsigned char *dev_addr;
1919
1920 struct netdev_rx_queue *_rx;
1921 unsigned int num_rx_queues;
1922 unsigned int real_num_rx_queues;
1923
1924 struct bpf_prog __rcu *xdp_prog;
1925 unsigned long gro_flush_timeout;
1926 rx_handler_func_t __rcu *rx_handler;
1927 void __rcu *rx_handler_data;
1928
1929#ifdef CONFIG_NET_CLS_ACT
1930 struct mini_Qdisc __rcu *miniq_ingress;
1931#endif
1932 struct netdev_queue __rcu *ingress_queue;
1933#ifdef CONFIG_NETFILTER_INGRESS
1934 struct nf_hook_entries __rcu *nf_hooks_ingress;
1935#endif
1936
1937 unsigned char broadcast[MAX_ADDR_LEN];
1938#ifdef CONFIG_RFS_ACCEL
1939 struct cpu_rmap *rx_cpu_rmap;
1940#endif
1941 struct hlist_node index_hlist;
1942
1943/*
1944 * Cache lines mostly used on transmit path
1945 */
1946 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1947 unsigned int num_tx_queues;
1948 unsigned int real_num_tx_queues;
1949 struct Qdisc *qdisc;
1950#ifdef CONFIG_NET_SCHED
1951 DECLARE_HASHTABLE (qdisc_hash, 4);
1952#endif
1953 unsigned int tx_queue_len;
1954 spinlock_t tx_global_lock;
1955 int watchdog_timeo;
1956
1957#ifdef CONFIG_XPS
1958 struct xps_dev_maps __rcu *xps_cpus_map;
1959 struct xps_dev_maps __rcu *xps_rxqs_map;
1960#endif
1961#ifdef CONFIG_NET_CLS_ACT
1962 struct mini_Qdisc __rcu *miniq_egress;
1963#endif
1964
1965 /* These may be needed for future network-power-down code. */
1966 struct timer_list watchdog_timer;
1967
1968 int __percpu *pcpu_refcnt;
1969 struct list_head todo_list;
1970
1971 struct list_head link_watch_list;
1972
1973 enum { NETREG_UNINITIALIZED=0,
1974 NETREG_REGISTERED, /* completed register_netdevice */
1975 NETREG_UNREGISTERING, /* called unregister_netdevice */
1976 NETREG_UNREGISTERED, /* completed unregister todo */
1977 NETREG_RELEASED, /* called free_netdev */
1978 NETREG_DUMMY, /* dummy device for NAPI poll */
1979 } reg_state:8;
1980
1981 bool dismantle;
1982
1983 enum {
1984 RTNL_LINK_INITIALIZED,
1985 RTNL_LINK_INITIALIZING,
1986 } rtnl_link_state:16;
1987
1988 bool needs_free_netdev;
1989 void (*priv_destructor)(struct net_device *dev);
1990
1991#ifdef CONFIG_NETPOLL
1992 struct netpoll_info __rcu *npinfo;
1993#endif
1994
1995 possible_net_t nd_net;
1996
1997 /* mid-layer private */
1998 union {
1999 void *ml_priv;
2000 struct pcpu_lstats __percpu *lstats;
2001 struct pcpu_sw_netstats __percpu *tstats;
2002 struct pcpu_dstats __percpu *dstats;
2003 };
2004
2005#if IS_ENABLED(CONFIG_GARP)
2006 struct garp_port __rcu *garp_port;
2007#endif
2008#if IS_ENABLED(CONFIG_MRP)
2009 struct mrp_port __rcu *mrp_port;
2010#endif
2011
2012 struct device dev;
2013 const struct attribute_group *sysfs_groups[4];
2014 const struct attribute_group *sysfs_rx_queue_group;
2015
2016 const struct rtnl_link_ops *rtnl_link_ops;
2017
2018 /* for setting kernel sock attribute on TCP connection setup */
2019#define GSO_MAX_SIZE 65536
2020 unsigned int gso_max_size;
2021#define GSO_MAX_SEGS 65535
2022 u16 gso_max_segs;
2023
2024#ifdef CONFIG_DCB
2025 const struct dcbnl_rtnl_ops *dcbnl_ops;
2026#endif
2027 s16 num_tc;
2028 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2029 u8 prio_tc_map[TC_BITMASK + 1];
2030
2031#if IS_ENABLED(CONFIG_FCOE)
2032 unsigned int fcoe_ddp_xid;
2033#endif
2034#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2035 struct netprio_map __rcu *priomap;
2036#endif
2037 struct phy_device *phydev;
2038 struct sfp_bus *sfp_bus;
2039 struct lock_class_key *qdisc_tx_busylock;
2040 struct lock_class_key *qdisc_running_key;
2041 bool proto_down;
2042 unsigned wol_enabled:1;
2043};
2044#define to_net_dev(d) container_of(d, struct net_device, dev)
2045
2046static inline bool netif_elide_gro(const struct net_device *dev)
2047{
2048 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2049 return true;
2050 return false;
2051}
2052
2053#define NETDEV_ALIGN 32
2054
2055static inline
2056int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2057{
2058 return dev->prio_tc_map[prio & TC_BITMASK];
2059}
2060
2061static inline
2062int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2063{
2064 if (tc >= dev->num_tc)
2065 return -EINVAL;
2066
2067 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2068 return 0;
2069}
2070
2071int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2072void netdev_reset_tc(struct net_device *dev);
2073int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2074int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2075
2076static inline
2077int netdev_get_num_tc(struct net_device *dev)
2078{
2079 return dev->num_tc;
2080}
2081
2082void netdev_unbind_sb_channel(struct net_device *dev,
2083 struct net_device *sb_dev);
2084int netdev_bind_sb_channel_queue(struct net_device *dev,
2085 struct net_device *sb_dev,
2086 u8 tc, u16 count, u16 offset);
2087int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2088static inline int netdev_get_sb_channel(struct net_device *dev)
2089{
2090 return max_t(int, -dev->num_tc, 0);
2091}
2092
2093static inline
2094struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2095 unsigned int index)
2096{
2097 return &dev->_tx[index];
2098}
2099
2100static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2101 const struct sk_buff *skb)
2102{
2103 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2104}
2105
2106static inline void netdev_for_each_tx_queue(struct net_device *dev,
2107 void (*f)(struct net_device *,
2108 struct netdev_queue *,
2109 void *),
2110 void *arg)
2111{
2112 unsigned int i;
2113
2114 for (i = 0; i < dev->num_tx_queues; i++)
2115 f(dev, &dev->_tx[i], arg);
2116}
2117
2118#define netdev_lockdep_set_classes(dev) \
2119{ \
2120 static struct lock_class_key qdisc_tx_busylock_key; \
2121 static struct lock_class_key qdisc_running_key; \
2122 static struct lock_class_key qdisc_xmit_lock_key; \
2123 static struct lock_class_key dev_addr_list_lock_key; \
2124 unsigned int i; \
2125 \
2126 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2127 (dev)->qdisc_running_key = &qdisc_running_key; \
2128 lockdep_set_class(&(dev)->addr_list_lock, \
2129 &dev_addr_list_lock_key); \
2130 for (i = 0; i < (dev)->num_tx_queues; i++) \
2131 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2132 &qdisc_xmit_lock_key); \
2133}
2134
2135u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2136 struct net_device *sb_dev);
2137struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2138 struct sk_buff *skb,
2139 struct net_device *sb_dev);
2140
2141/* returns the headroom that the master device needs to take in account
2142 * when forwarding to this dev
2143 */
2144static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2145{
2146 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2147}
2148
2149static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2150{
2151 if (dev->netdev_ops->ndo_set_rx_headroom)
2152 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2153}
2154
2155/* set the device rx headroom to the dev's default */
2156static inline void netdev_reset_rx_headroom(struct net_device *dev)
2157{
2158 netdev_set_rx_headroom(dev, -1);
2159}
2160
2161/*
2162 * Net namespace inlines
2163 */
2164static inline
2165struct net *dev_net(const struct net_device *dev)
2166{
2167 return read_pnet(&dev->nd_net);
2168}
2169
2170static inline
2171void dev_net_set(struct net_device *dev, struct net *net)
2172{
2173 write_pnet(&dev->nd_net, net);
2174}
2175
2176/**
2177 * netdev_priv - access network device private data
2178 * @dev: network device
2179 *
2180 * Get network device private data
2181 */
2182static inline void *netdev_priv(const struct net_device *dev)
2183{
2184 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2185}
2186
2187/* Set the sysfs physical device reference for the network logical device
2188 * if set prior to registration will cause a symlink during initialization.
2189 */
2190#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2191
2192/* Set the sysfs device type for the network logical device to allow
2193 * fine-grained identification of different network device types. For
2194 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2195 */
2196#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2197
2198/* Default NAPI poll() weight
2199 * Device drivers are strongly advised to not use bigger value
2200 */
2201#define NAPI_POLL_WEIGHT 64
2202
2203/**
2204 * netif_napi_add - initialize a NAPI context
2205 * @dev: network device
2206 * @napi: NAPI context
2207 * @poll: polling function
2208 * @weight: default weight
2209 *
2210 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2211 * *any* of the other NAPI-related functions.
2212 */
2213void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2214 int (*poll)(struct napi_struct *, int), int weight);
2215
2216/**
2217 * netif_tx_napi_add - initialize a NAPI context
2218 * @dev: network device
2219 * @napi: NAPI context
2220 * @poll: polling function
2221 * @weight: default weight
2222 *
2223 * This variant of netif_napi_add() should be used from drivers using NAPI
2224 * to exclusively poll a TX queue.
2225 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2226 */
2227static inline void netif_tx_napi_add(struct net_device *dev,
2228 struct napi_struct *napi,
2229 int (*poll)(struct napi_struct *, int),
2230 int weight)
2231{
2232 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2233 netif_napi_add(dev, napi, poll, weight);
2234}
2235
2236/**
2237 * netif_napi_del - remove a NAPI context
2238 * @napi: NAPI context
2239 *
2240 * netif_napi_del() removes a NAPI context from the network device NAPI list
2241 */
2242void netif_napi_del(struct napi_struct *napi);
2243
2244struct napi_gro_cb {
2245 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2246 void *frag0;
2247
2248 /* Length of frag0. */
2249 unsigned int frag0_len;
2250
2251 /* This indicates where we are processing relative to skb->data. */
2252 int data_offset;
2253
2254 /* This is non-zero if the packet cannot be merged with the new skb. */
2255 u16 flush;
2256
2257 /* Save the IP ID here and check when we get to the transport layer */
2258 u16 flush_id;
2259
2260 /* Number of segments aggregated. */
2261 u16 count;
2262
2263 /* Start offset for remote checksum offload */
2264 u16 gro_remcsum_start;
2265
2266 /* jiffies when first packet was created/queued */
2267 unsigned long age;
2268
2269 /* Used in ipv6_gro_receive() and foo-over-udp */
2270 u16 proto;
2271
2272 /* This is non-zero if the packet may be of the same flow. */
2273 u8 same_flow:1;
2274
2275 /* Used in tunnel GRO receive */
2276 u8 encap_mark:1;
2277
2278 /* GRO checksum is valid */
2279 u8 csum_valid:1;
2280
2281 /* Number of checksums via CHECKSUM_UNNECESSARY */
2282 u8 csum_cnt:3;
2283
2284 /* Free the skb? */
2285 u8 free:2;
2286#define NAPI_GRO_FREE 1
2287#define NAPI_GRO_FREE_STOLEN_HEAD 2
2288
2289 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2290 u8 is_ipv6:1;
2291
2292 /* Used in GRE, set in fou/gue_gro_receive */
2293 u8 is_fou:1;
2294
2295 /* Used to determine if flush_id can be ignored */
2296 u8 is_atomic:1;
2297
2298 /* Number of gro_receive callbacks this packet already went through */
2299 u8 recursion_counter:4;
2300
2301 /* 1 bit hole */
2302
2303 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2304 __wsum csum;
2305
2306 /* used in skb_gro_receive() slow path */
2307 struct sk_buff *last;
2308};
2309
2310#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2311
2312#define GRO_RECURSION_LIMIT 15
2313static inline int gro_recursion_inc_test(struct sk_buff *skb)
2314{
2315 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2316}
2317
2318typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2319static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2320 struct list_head *head,
2321 struct sk_buff *skb)
2322{
2323 if (unlikely(gro_recursion_inc_test(skb))) {
2324 NAPI_GRO_CB(skb)->flush |= 1;
2325 return NULL;
2326 }
2327
2328 return cb(head, skb);
2329}
2330
2331typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2332 struct sk_buff *);
2333static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2334 struct sock *sk,
2335 struct list_head *head,
2336 struct sk_buff *skb)
2337{
2338 if (unlikely(gro_recursion_inc_test(skb))) {
2339 NAPI_GRO_CB(skb)->flush |= 1;
2340 return NULL;
2341 }
2342
2343 return cb(sk, head, skb);
2344}
2345
2346struct packet_type {
2347 __be16 type; /* This is really htons(ether_type). */
2348 bool ignore_outgoing;
2349 struct net_device *dev; /* NULL is wildcarded here */
2350 int (*func) (struct sk_buff *,
2351 struct net_device *,
2352 struct packet_type *,
2353 struct net_device *);
2354 void (*list_func) (struct list_head *,
2355 struct packet_type *,
2356 struct net_device *);
2357 bool (*id_match)(struct packet_type *ptype,
2358 struct sock *sk);
2359 void *af_packet_priv;
2360 struct list_head list;
2361};
2362
2363struct offload_callbacks {
2364 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2365 netdev_features_t features);
2366 struct sk_buff *(*gro_receive)(struct list_head *head,
2367 struct sk_buff *skb);
2368 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2369};
2370
2371struct packet_offload {
2372 __be16 type; /* This is really htons(ether_type). */
2373 u16 priority;
2374 struct offload_callbacks callbacks;
2375 struct list_head list;
2376};
2377
2378/* often modified stats are per-CPU, other are shared (netdev->stats) */
2379struct pcpu_sw_netstats {
2380 u64 rx_packets;
2381 u64 rx_bytes;
2382 u64 tx_packets;
2383 u64 tx_bytes;
2384 struct u64_stats_sync syncp;
2385} __aligned(4 * sizeof(u64));
2386
2387struct pcpu_lstats {
2388 u64 packets;
2389 u64 bytes;
2390 struct u64_stats_sync syncp;
2391} __aligned(2 * sizeof(u64));
2392
2393#define __netdev_alloc_pcpu_stats(type, gfp) \
2394({ \
2395 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2396 if (pcpu_stats) { \
2397 int __cpu; \
2398 for_each_possible_cpu(__cpu) { \
2399 typeof(type) *stat; \
2400 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2401 u64_stats_init(&stat->syncp); \
2402 } \
2403 } \
2404 pcpu_stats; \
2405})
2406
2407#define netdev_alloc_pcpu_stats(type) \
2408 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2409
2410enum netdev_lag_tx_type {
2411 NETDEV_LAG_TX_TYPE_UNKNOWN,
2412 NETDEV_LAG_TX_TYPE_RANDOM,
2413 NETDEV_LAG_TX_TYPE_BROADCAST,
2414 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2415 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2416 NETDEV_LAG_TX_TYPE_HASH,
2417};
2418
2419enum netdev_lag_hash {
2420 NETDEV_LAG_HASH_NONE,
2421 NETDEV_LAG_HASH_L2,
2422 NETDEV_LAG_HASH_L34,
2423 NETDEV_LAG_HASH_L23,
2424 NETDEV_LAG_HASH_E23,
2425 NETDEV_LAG_HASH_E34,
2426 NETDEV_LAG_HASH_UNKNOWN,
2427};
2428
2429struct netdev_lag_upper_info {
2430 enum netdev_lag_tx_type tx_type;
2431 enum netdev_lag_hash hash_type;
2432};
2433
2434struct netdev_lag_lower_state_info {
2435 u8 link_up : 1,
2436 tx_enabled : 1;
2437};
2438
2439#include <linux/notifier.h>
2440
2441/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2442 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2443 * adding new types.
2444 */
2445enum netdev_cmd {
2446 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2447 NETDEV_DOWN,
2448 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2449 detected a hardware crash and restarted
2450 - we can use this eg to kick tcp sessions
2451 once done */
2452 NETDEV_CHANGE, /* Notify device state change */
2453 NETDEV_REGISTER,
2454 NETDEV_UNREGISTER,
2455 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2456 NETDEV_CHANGEADDR, /* notify after the address change */
2457 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2458 NETDEV_GOING_DOWN,
2459 NETDEV_CHANGENAME,
2460 NETDEV_FEAT_CHANGE,
2461 NETDEV_BONDING_FAILOVER,
2462 NETDEV_PRE_UP,
2463 NETDEV_PRE_TYPE_CHANGE,
2464 NETDEV_POST_TYPE_CHANGE,
2465 NETDEV_POST_INIT,
2466 NETDEV_RELEASE,
2467 NETDEV_NOTIFY_PEERS,
2468 NETDEV_JOIN,
2469 NETDEV_CHANGEUPPER,
2470 NETDEV_RESEND_IGMP,
2471 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2472 NETDEV_CHANGEINFODATA,
2473 NETDEV_BONDING_INFO,
2474 NETDEV_PRECHANGEUPPER,
2475 NETDEV_CHANGELOWERSTATE,
2476 NETDEV_UDP_TUNNEL_PUSH_INFO,
2477 NETDEV_UDP_TUNNEL_DROP_INFO,
2478 NETDEV_CHANGE_TX_QUEUE_LEN,
2479 NETDEV_CVLAN_FILTER_PUSH_INFO,
2480 NETDEV_CVLAN_FILTER_DROP_INFO,
2481 NETDEV_SVLAN_FILTER_PUSH_INFO,
2482 NETDEV_SVLAN_FILTER_DROP_INFO,
2483};
2484const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2485
2486int register_netdevice_notifier(struct notifier_block *nb);
2487int unregister_netdevice_notifier(struct notifier_block *nb);
2488
2489struct netdev_notifier_info {
2490 struct net_device *dev;
2491 struct netlink_ext_ack *extack;
2492};
2493
2494struct netdev_notifier_info_ext {
2495 struct netdev_notifier_info info; /* must be first */
2496 union {
2497 u32 mtu;
2498 } ext;
2499};
2500
2501struct netdev_notifier_change_info {
2502 struct netdev_notifier_info info; /* must be first */
2503 unsigned int flags_changed;
2504};
2505
2506struct netdev_notifier_changeupper_info {
2507 struct netdev_notifier_info info; /* must be first */
2508 struct net_device *upper_dev; /* new upper dev */
2509 bool master; /* is upper dev master */
2510 bool linking; /* is the notification for link or unlink */
2511 void *upper_info; /* upper dev info */
2512};
2513
2514struct netdev_notifier_changelowerstate_info {
2515 struct netdev_notifier_info info; /* must be first */
2516 void *lower_state_info; /* is lower dev state */
2517};
2518
2519struct netdev_notifier_pre_changeaddr_info {
2520 struct netdev_notifier_info info; /* must be first */
2521 const unsigned char *dev_addr;
2522};
2523
2524static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2525 struct net_device *dev)
2526{
2527 info->dev = dev;
2528 info->extack = NULL;
2529}
2530
2531static inline struct net_device *
2532netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2533{
2534 return info->dev;
2535}
2536
2537static inline struct netlink_ext_ack *
2538netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2539{
2540 return info->extack;
2541}
2542
2543int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2544
2545
2546extern rwlock_t dev_base_lock; /* Device list lock */
2547
2548#define for_each_netdev(net, d) \
2549 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2550#define for_each_netdev_reverse(net, d) \
2551 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2552#define for_each_netdev_rcu(net, d) \
2553 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2554#define for_each_netdev_safe(net, d, n) \
2555 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2556#define for_each_netdev_continue(net, d) \
2557 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2558#define for_each_netdev_continue_rcu(net, d) \
2559 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2560#define for_each_netdev_in_bond_rcu(bond, slave) \
2561 for_each_netdev_rcu(&init_net, slave) \
2562 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2563#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2564
2565static inline struct net_device *next_net_device(struct net_device *dev)
2566{
2567 struct list_head *lh;
2568 struct net *net;
2569
2570 net = dev_net(dev);
2571 lh = dev->dev_list.next;
2572 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2573}
2574
2575static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2576{
2577 struct list_head *lh;
2578 struct net *net;
2579
2580 net = dev_net(dev);
2581 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2582 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2583}
2584
2585static inline struct net_device *first_net_device(struct net *net)
2586{
2587 return list_empty(&net->dev_base_head) ? NULL :
2588 net_device_entry(net->dev_base_head.next);
2589}
2590
2591static inline struct net_device *first_net_device_rcu(struct net *net)
2592{
2593 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2594
2595 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2596}
2597
2598int netdev_boot_setup_check(struct net_device *dev);
2599unsigned long netdev_boot_base(const char *prefix, int unit);
2600struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2601 const char *hwaddr);
2602struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2603struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2604void dev_add_pack(struct packet_type *pt);
2605void dev_remove_pack(struct packet_type *pt);
2606void __dev_remove_pack(struct packet_type *pt);
2607void dev_add_offload(struct packet_offload *po);
2608void dev_remove_offload(struct packet_offload *po);
2609
2610int dev_get_iflink(const struct net_device *dev);
2611int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2612struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2613 unsigned short mask);
2614struct net_device *dev_get_by_name(struct net *net, const char *name);
2615struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2616struct net_device *__dev_get_by_name(struct net *net, const char *name);
2617int dev_alloc_name(struct net_device *dev, const char *name);
2618int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2619void dev_close(struct net_device *dev);
2620void dev_close_many(struct list_head *head, bool unlink);
2621void dev_disable_lro(struct net_device *dev);
2622int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2623u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2624 struct net_device *sb_dev);
2625u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2626 struct net_device *sb_dev);
2627int dev_queue_xmit(struct sk_buff *skb);
2628int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2629int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2630int register_netdevice(struct net_device *dev);
2631void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2632void unregister_netdevice_many(struct list_head *head);
2633static inline void unregister_netdevice(struct net_device *dev)
2634{
2635 unregister_netdevice_queue(dev, NULL);
2636}
2637
2638int netdev_refcnt_read(const struct net_device *dev);
2639void free_netdev(struct net_device *dev);
2640void netdev_freemem(struct net_device *dev);
2641void synchronize_net(void);
2642int init_dummy_netdev(struct net_device *dev);
2643
2644struct net_device *dev_get_by_index(struct net *net, int ifindex);
2645struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2646struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2647struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2648int netdev_get_name(struct net *net, char *name, int ifindex);
2649int dev_restart(struct net_device *dev);
2650int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2651
2652static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2653{
2654 return NAPI_GRO_CB(skb)->data_offset;
2655}
2656
2657static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2658{
2659 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2660}
2661
2662static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2663{
2664 NAPI_GRO_CB(skb)->data_offset += len;
2665}
2666
2667static inline void *skb_gro_header_fast(struct sk_buff *skb,
2668 unsigned int offset)
2669{
2670 return NAPI_GRO_CB(skb)->frag0 + offset;
2671}
2672
2673static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2674{
2675 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2676}
2677
2678static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2679{
2680 NAPI_GRO_CB(skb)->frag0 = NULL;
2681 NAPI_GRO_CB(skb)->frag0_len = 0;
2682}
2683
2684static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2685 unsigned int offset)
2686{
2687 if (!pskb_may_pull(skb, hlen))
2688 return NULL;
2689
2690 skb_gro_frag0_invalidate(skb);
2691 return skb->data + offset;
2692}
2693
2694static inline void *skb_gro_network_header(struct sk_buff *skb)
2695{
2696 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2697 skb_network_offset(skb);
2698}
2699
2700static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2701 const void *start, unsigned int len)
2702{
2703 if (NAPI_GRO_CB(skb)->csum_valid)
2704 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2705 csum_partial(start, len, 0));
2706}
2707
2708/* GRO checksum functions. These are logical equivalents of the normal
2709 * checksum functions (in skbuff.h) except that they operate on the GRO
2710 * offsets and fields in sk_buff.
2711 */
2712
2713__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2714
2715static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2716{
2717 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2718}
2719
2720static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2721 bool zero_okay,
2722 __sum16 check)
2723{
2724 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2725 skb_checksum_start_offset(skb) <
2726 skb_gro_offset(skb)) &&
2727 !skb_at_gro_remcsum_start(skb) &&
2728 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2729 (!zero_okay || check));
2730}
2731
2732static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2733 __wsum psum)
2734{
2735 if (NAPI_GRO_CB(skb)->csum_valid &&
2736 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2737 return 0;
2738
2739 NAPI_GRO_CB(skb)->csum = psum;
2740
2741 return __skb_gro_checksum_complete(skb);
2742}
2743
2744static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2745{
2746 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2747 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2748 NAPI_GRO_CB(skb)->csum_cnt--;
2749 } else {
2750 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2751 * verified a new top level checksum or an encapsulated one
2752 * during GRO. This saves work if we fallback to normal path.
2753 */
2754 __skb_incr_checksum_unnecessary(skb);
2755 }
2756}
2757
2758#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2759 compute_pseudo) \
2760({ \
2761 __sum16 __ret = 0; \
2762 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2763 __ret = __skb_gro_checksum_validate_complete(skb, \
2764 compute_pseudo(skb, proto)); \
2765 if (!__ret) \
2766 skb_gro_incr_csum_unnecessary(skb); \
2767 __ret; \
2768})
2769
2770#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2771 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2772
2773#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2774 compute_pseudo) \
2775 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2776
2777#define skb_gro_checksum_simple_validate(skb) \
2778 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2779
2780static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2781{
2782 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2783 !NAPI_GRO_CB(skb)->csum_valid);
2784}
2785
2786static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2787 __sum16 check, __wsum pseudo)
2788{
2789 NAPI_GRO_CB(skb)->csum = ~pseudo;
2790 NAPI_GRO_CB(skb)->csum_valid = 1;
2791}
2792
2793#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2794do { \
2795 if (__skb_gro_checksum_convert_check(skb)) \
2796 __skb_gro_checksum_convert(skb, check, \
2797 compute_pseudo(skb, proto)); \
2798} while (0)
2799
2800struct gro_remcsum {
2801 int offset;
2802 __wsum delta;
2803};
2804
2805static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2806{
2807 grc->offset = 0;
2808 grc->delta = 0;
2809}
2810
2811static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2812 unsigned int off, size_t hdrlen,
2813 int start, int offset,
2814 struct gro_remcsum *grc,
2815 bool nopartial)
2816{
2817 __wsum delta;
2818 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2819
2820 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2821
2822 if (!nopartial) {
2823 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2824 return ptr;
2825 }
2826
2827 ptr = skb_gro_header_fast(skb, off);
2828 if (skb_gro_header_hard(skb, off + plen)) {
2829 ptr = skb_gro_header_slow(skb, off + plen, off);
2830 if (!ptr)
2831 return NULL;
2832 }
2833
2834 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2835 start, offset);
2836
2837 /* Adjust skb->csum since we changed the packet */
2838 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2839
2840 grc->offset = off + hdrlen + offset;
2841 grc->delta = delta;
2842
2843 return ptr;
2844}
2845
2846static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2847 struct gro_remcsum *grc)
2848{
2849 void *ptr;
2850 size_t plen = grc->offset + sizeof(u16);
2851
2852 if (!grc->delta)
2853 return;
2854
2855 ptr = skb_gro_header_fast(skb, grc->offset);
2856 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2857 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2858 if (!ptr)
2859 return;
2860 }
2861
2862 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2863}
2864
2865#ifdef CONFIG_XFRM_OFFLOAD
2866static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2867{
2868 if (PTR_ERR(pp) != -EINPROGRESS)
2869 NAPI_GRO_CB(skb)->flush |= flush;
2870}
2871static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2872 struct sk_buff *pp,
2873 int flush,
2874 struct gro_remcsum *grc)
2875{
2876 if (PTR_ERR(pp) != -EINPROGRESS) {
2877 NAPI_GRO_CB(skb)->flush |= flush;
2878 skb_gro_remcsum_cleanup(skb, grc);
2879 skb->remcsum_offload = 0;
2880 }
2881}
2882#else
2883static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2884{
2885 NAPI_GRO_CB(skb)->flush |= flush;
2886}
2887static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2888 struct sk_buff *pp,
2889 int flush,
2890 struct gro_remcsum *grc)
2891{
2892 NAPI_GRO_CB(skb)->flush |= flush;
2893 skb_gro_remcsum_cleanup(skb, grc);
2894 skb->remcsum_offload = 0;
2895}
2896#endif
2897
2898static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2899 unsigned short type,
2900 const void *daddr, const void *saddr,
2901 unsigned int len)
2902{
2903 if (!dev->header_ops || !dev->header_ops->create)
2904 return 0;
2905
2906 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2907}
2908
2909static inline int dev_parse_header(const struct sk_buff *skb,
2910 unsigned char *haddr)
2911{
2912 const struct net_device *dev = skb->dev;
2913
2914 if (!dev->header_ops || !dev->header_ops->parse)
2915 return 0;
2916 return dev->header_ops->parse(skb, haddr);
2917}
2918
2919static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2920{
2921 const struct net_device *dev = skb->dev;
2922
2923 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2924 return 0;
2925 return dev->header_ops->parse_protocol(skb);
2926}
2927
2928/* ll_header must have at least hard_header_len allocated */
2929static inline bool dev_validate_header(const struct net_device *dev,
2930 char *ll_header, int len)
2931{
2932 if (likely(len >= dev->hard_header_len))
2933 return true;
2934 if (len < dev->min_header_len)
2935 return false;
2936
2937 if (capable(CAP_SYS_RAWIO)) {
2938 memset(ll_header + len, 0, dev->hard_header_len - len);
2939 return true;
2940 }
2941
2942 if (dev->header_ops && dev->header_ops->validate)
2943 return dev->header_ops->validate(ll_header, len);
2944
2945 return false;
2946}
2947
2948typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2949 int len, int size);
2950int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2951static inline int unregister_gifconf(unsigned int family)
2952{
2953 return register_gifconf(family, NULL);
2954}
2955
2956#ifdef CONFIG_NET_FLOW_LIMIT
2957#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2958struct sd_flow_limit {
2959 u64 count;
2960 unsigned int num_buckets;
2961 unsigned int history_head;
2962 u16 history[FLOW_LIMIT_HISTORY];
2963 u8 buckets[];
2964};
2965
2966extern int netdev_flow_limit_table_len;
2967#endif /* CONFIG_NET_FLOW_LIMIT */
2968
2969/*
2970 * Incoming packets are placed on per-CPU queues
2971 */
2972struct softnet_data {
2973 struct list_head poll_list;
2974 struct sk_buff_head process_queue;
2975
2976 /* stats */
2977 unsigned int processed;
2978 unsigned int time_squeeze;
2979 unsigned int received_rps;
2980#ifdef CONFIG_RPS
2981 struct softnet_data *rps_ipi_list;
2982#endif
2983#ifdef CONFIG_NET_FLOW_LIMIT
2984 struct sd_flow_limit __rcu *flow_limit;
2985#endif
2986 struct Qdisc *output_queue;
2987 struct Qdisc **output_queue_tailp;
2988 struct sk_buff *completion_queue;
2989#ifdef CONFIG_XFRM_OFFLOAD
2990 struct sk_buff_head xfrm_backlog;
2991#endif
2992 /* written and read only by owning cpu: */
2993 struct {
2994 u16 recursion;
2995 u8 more;
2996 } xmit;
2997#ifdef CONFIG_RPS
2998 /* input_queue_head should be written by cpu owning this struct,
2999 * and only read by other cpus. Worth using a cache line.
3000 */
3001 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3002
3003 /* Elements below can be accessed between CPUs for RPS/RFS */
3004 call_single_data_t csd ____cacheline_aligned_in_smp;
3005 struct softnet_data *rps_ipi_next;
3006 unsigned int cpu;
3007 unsigned int input_queue_tail;
3008#endif
3009 unsigned int dropped;
3010 struct sk_buff_head input_pkt_queue;
3011 struct napi_struct backlog;
3012
3013};
3014
3015static inline void input_queue_head_incr(struct softnet_data *sd)
3016{
3017#ifdef CONFIG_RPS
3018 sd->input_queue_head++;
3019#endif
3020}
3021
3022static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3023 unsigned int *qtail)
3024{
3025#ifdef CONFIG_RPS
3026 *qtail = ++sd->input_queue_tail;
3027#endif
3028}
3029
3030DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3031
3032static inline int dev_recursion_level(void)
3033{
3034 return this_cpu_read(softnet_data.xmit.recursion);
3035}
3036
3037#define XMIT_RECURSION_LIMIT 10
3038static inline bool dev_xmit_recursion(void)
3039{
3040 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3041 XMIT_RECURSION_LIMIT);
3042}
3043
3044static inline void dev_xmit_recursion_inc(void)
3045{
3046 __this_cpu_inc(softnet_data.xmit.recursion);
3047}
3048
3049static inline void dev_xmit_recursion_dec(void)
3050{
3051 __this_cpu_dec(softnet_data.xmit.recursion);
3052}
3053
3054void __netif_schedule(struct Qdisc *q);
3055void netif_schedule_queue(struct netdev_queue *txq);
3056
3057static inline void netif_tx_schedule_all(struct net_device *dev)
3058{
3059 unsigned int i;
3060
3061 for (i = 0; i < dev->num_tx_queues; i++)
3062 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3063}
3064
3065static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3066{
3067 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3068}
3069
3070/**
3071 * netif_start_queue - allow transmit
3072 * @dev: network device
3073 *
3074 * Allow upper layers to call the device hard_start_xmit routine.
3075 */
3076static inline void netif_start_queue(struct net_device *dev)
3077{
3078 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3079}
3080
3081static inline void netif_tx_start_all_queues(struct net_device *dev)
3082{
3083 unsigned int i;
3084
3085 for (i = 0; i < dev->num_tx_queues; i++) {
3086 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3087 netif_tx_start_queue(txq);
3088 }
3089}
3090
3091void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3092
3093/**
3094 * netif_wake_queue - restart transmit
3095 * @dev: network device
3096 *
3097 * Allow upper layers to call the device hard_start_xmit routine.
3098 * Used for flow control when transmit resources are available.
3099 */
3100static inline void netif_wake_queue(struct net_device *dev)
3101{
3102 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3103}
3104
3105static inline void netif_tx_wake_all_queues(struct net_device *dev)
3106{
3107 unsigned int i;
3108
3109 for (i = 0; i < dev->num_tx_queues; i++) {
3110 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3111 netif_tx_wake_queue(txq);
3112 }
3113}
3114
3115static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3116{
3117 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3118}
3119
3120/**
3121 * netif_stop_queue - stop transmitted packets
3122 * @dev: network device
3123 *
3124 * Stop upper layers calling the device hard_start_xmit routine.
3125 * Used for flow control when transmit resources are unavailable.
3126 */
3127static inline void netif_stop_queue(struct net_device *dev)
3128{
3129 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3130}
3131
3132void netif_tx_stop_all_queues(struct net_device *dev);
3133
3134static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3135{
3136 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3137}
3138
3139/**
3140 * netif_queue_stopped - test if transmit queue is flowblocked
3141 * @dev: network device
3142 *
3143 * Test if transmit queue on device is currently unable to send.
3144 */
3145static inline bool netif_queue_stopped(const struct net_device *dev)
3146{
3147 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3148}
3149
3150static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3151{
3152 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3153}
3154
3155static inline bool
3156netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3157{
3158 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3159}
3160
3161static inline bool
3162netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3163{
3164 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3165}
3166
3167/**
3168 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3169 * @dev_queue: pointer to transmit queue
3170 *
3171 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3172 * to give appropriate hint to the CPU.
3173 */
3174static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3175{
3176#ifdef CONFIG_BQL
3177 prefetchw(&dev_queue->dql.num_queued);
3178#endif
3179}
3180
3181/**
3182 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3183 * @dev_queue: pointer to transmit queue
3184 *
3185 * BQL enabled drivers might use this helper in their TX completion path,
3186 * to give appropriate hint to the CPU.
3187 */
3188static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3189{
3190#ifdef CONFIG_BQL
3191 prefetchw(&dev_queue->dql.limit);
3192#endif
3193}
3194
3195static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3196 unsigned int bytes)
3197{
3198#ifdef CONFIG_BQL
3199 dql_queued(&dev_queue->dql, bytes);
3200
3201 if (likely(dql_avail(&dev_queue->dql) >= 0))
3202 return;
3203
3204 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3205
3206 /*
3207 * The XOFF flag must be set before checking the dql_avail below,
3208 * because in netdev_tx_completed_queue we update the dql_completed
3209 * before checking the XOFF flag.
3210 */
3211 smp_mb();
3212
3213 /* check again in case another CPU has just made room avail */
3214 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3215 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3216#endif
3217}
3218
3219/* Variant of netdev_tx_sent_queue() for drivers that are aware
3220 * that they should not test BQL status themselves.
3221 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3222 * skb of a batch.
3223 * Returns true if the doorbell must be used to kick the NIC.
3224 */
3225static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3226 unsigned int bytes,
3227 bool xmit_more)
3228{
3229 if (xmit_more) {
3230#ifdef CONFIG_BQL
3231 dql_queued(&dev_queue->dql, bytes);
3232#endif
3233 return netif_tx_queue_stopped(dev_queue);
3234 }
3235 netdev_tx_sent_queue(dev_queue, bytes);
3236 return true;
3237}
3238
3239/**
3240 * netdev_sent_queue - report the number of bytes queued to hardware
3241 * @dev: network device
3242 * @bytes: number of bytes queued to the hardware device queue
3243 *
3244 * Report the number of bytes queued for sending/completion to the network
3245 * device hardware queue. @bytes should be a good approximation and should
3246 * exactly match netdev_completed_queue() @bytes
3247 */
3248static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3249{
3250 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3251}
3252
3253static inline bool __netdev_sent_queue(struct net_device *dev,
3254 unsigned int bytes,
3255 bool xmit_more)
3256{
3257 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3258 xmit_more);
3259}
3260
3261static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3262 unsigned int pkts, unsigned int bytes)
3263{
3264#ifdef CONFIG_BQL
3265 if (unlikely(!bytes))
3266 return;
3267
3268 dql_completed(&dev_queue->dql, bytes);
3269
3270 /*
3271 * Without the memory barrier there is a small possiblity that
3272 * netdev_tx_sent_queue will miss the update and cause the queue to
3273 * be stopped forever
3274 */
3275 smp_mb();
3276
3277 if (dql_avail(&dev_queue->dql) < 0)
3278 return;
3279
3280 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3281 netif_schedule_queue(dev_queue);
3282#endif
3283}
3284
3285/**
3286 * netdev_completed_queue - report bytes and packets completed by device
3287 * @dev: network device
3288 * @pkts: actual number of packets sent over the medium
3289 * @bytes: actual number of bytes sent over the medium
3290 *
3291 * Report the number of bytes and packets transmitted by the network device
3292 * hardware queue over the physical medium, @bytes must exactly match the
3293 * @bytes amount passed to netdev_sent_queue()
3294 */
3295static inline void netdev_completed_queue(struct net_device *dev,
3296 unsigned int pkts, unsigned int bytes)
3297{
3298 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3299}
3300
3301static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3302{
3303#ifdef CONFIG_BQL
3304 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3305 dql_reset(&q->dql);
3306#endif
3307}
3308
3309/**
3310 * netdev_reset_queue - reset the packets and bytes count of a network device
3311 * @dev_queue: network device
3312 *
3313 * Reset the bytes and packet count of a network device and clear the
3314 * software flow control OFF bit for this network device
3315 */
3316static inline void netdev_reset_queue(struct net_device *dev_queue)
3317{
3318 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3319}
3320
3321/**
3322 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3323 * @dev: network device
3324 * @queue_index: given tx queue index
3325 *
3326 * Returns 0 if given tx queue index >= number of device tx queues,
3327 * otherwise returns the originally passed tx queue index.
3328 */
3329static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3330{
3331 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3332 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3333 dev->name, queue_index,
3334 dev->real_num_tx_queues);
3335 return 0;
3336 }
3337
3338 return queue_index;
3339}
3340
3341/**
3342 * netif_running - test if up
3343 * @dev: network device
3344 *
3345 * Test if the device has been brought up.
3346 */
3347static inline bool netif_running(const struct net_device *dev)
3348{
3349 return test_bit(__LINK_STATE_START, &dev->state);
3350}
3351
3352/*
3353 * Routines to manage the subqueues on a device. We only need start,
3354 * stop, and a check if it's stopped. All other device management is
3355 * done at the overall netdevice level.
3356 * Also test the device if we're multiqueue.
3357 */
3358
3359/**
3360 * netif_start_subqueue - allow sending packets on subqueue
3361 * @dev: network device
3362 * @queue_index: sub queue index
3363 *
3364 * Start individual transmit queue of a device with multiple transmit queues.
3365 */
3366static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3367{
3368 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3369
3370 netif_tx_start_queue(txq);
3371}
3372
3373/**
3374 * netif_stop_subqueue - stop sending packets on subqueue
3375 * @dev: network device
3376 * @queue_index: sub queue index
3377 *
3378 * Stop individual transmit queue of a device with multiple transmit queues.
3379 */
3380static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3381{
3382 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3383 netif_tx_stop_queue(txq);
3384}
3385
3386/**
3387 * netif_subqueue_stopped - test status of subqueue
3388 * @dev: network device
3389 * @queue_index: sub queue index
3390 *
3391 * Check individual transmit queue of a device with multiple transmit queues.
3392 */
3393static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3394 u16 queue_index)
3395{
3396 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3397
3398 return netif_tx_queue_stopped(txq);
3399}
3400
3401static inline bool netif_subqueue_stopped(const struct net_device *dev,
3402 struct sk_buff *skb)
3403{
3404 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3405}
3406
3407/**
3408 * netif_wake_subqueue - allow sending packets on subqueue
3409 * @dev: network device
3410 * @queue_index: sub queue index
3411 *
3412 * Resume individual transmit queue of a device with multiple transmit queues.
3413 */
3414static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3415{
3416 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3417
3418 netif_tx_wake_queue(txq);
3419}
3420
3421#ifdef CONFIG_XPS
3422int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3423 u16 index);
3424int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3425 u16 index, bool is_rxqs_map);
3426
3427/**
3428 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3429 * @j: CPU/Rx queue index
3430 * @mask: bitmask of all cpus/rx queues
3431 * @nr_bits: number of bits in the bitmask
3432 *
3433 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3434 */
3435static inline bool netif_attr_test_mask(unsigned long j,
3436 const unsigned long *mask,
3437 unsigned int nr_bits)
3438{
3439 cpu_max_bits_warn(j, nr_bits);
3440 return test_bit(j, mask);
3441}
3442
3443/**
3444 * netif_attr_test_online - Test for online CPU/Rx queue
3445 * @j: CPU/Rx queue index
3446 * @online_mask: bitmask for CPUs/Rx queues that are online
3447 * @nr_bits: number of bits in the bitmask
3448 *
3449 * Returns true if a CPU/Rx queue is online.
3450 */
3451static inline bool netif_attr_test_online(unsigned long j,
3452 const unsigned long *online_mask,
3453 unsigned int nr_bits)
3454{
3455 cpu_max_bits_warn(j, nr_bits);
3456
3457 if (online_mask)
3458 return test_bit(j, online_mask);
3459
3460 return (j < nr_bits);
3461}
3462
3463/**
3464 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3465 * @n: CPU/Rx queue index
3466 * @srcp: the cpumask/Rx queue mask pointer
3467 * @nr_bits: number of bits in the bitmask
3468 *
3469 * Returns >= nr_bits if no further CPUs/Rx queues set.
3470 */
3471static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3472 unsigned int nr_bits)
3473{
3474 /* -1 is a legal arg here. */
3475 if (n != -1)
3476 cpu_max_bits_warn(n, nr_bits);
3477
3478 if (srcp)
3479 return find_next_bit(srcp, nr_bits, n + 1);
3480
3481 return n + 1;
3482}
3483
3484/**
3485 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3486 * @n: CPU/Rx queue index
3487 * @src1p: the first CPUs/Rx queues mask pointer
3488 * @src2p: the second CPUs/Rx queues mask pointer
3489 * @nr_bits: number of bits in the bitmask
3490 *
3491 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3492 */
3493static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3494 const unsigned long *src2p,
3495 unsigned int nr_bits)
3496{
3497 /* -1 is a legal arg here. */
3498 if (n != -1)
3499 cpu_max_bits_warn(n, nr_bits);
3500
3501 if (src1p && src2p)
3502 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3503 else if (src1p)
3504 return find_next_bit(src1p, nr_bits, n + 1);
3505 else if (src2p)
3506 return find_next_bit(src2p, nr_bits, n + 1);
3507
3508 return n + 1;
3509}
3510#else
3511static inline int netif_set_xps_queue(struct net_device *dev,
3512 const struct cpumask *mask,
3513 u16 index)
3514{
3515 return 0;
3516}
3517
3518static inline int __netif_set_xps_queue(struct net_device *dev,
3519 const unsigned long *mask,
3520 u16 index, bool is_rxqs_map)
3521{
3522 return 0;
3523}
3524#endif
3525
3526/**
3527 * netif_is_multiqueue - test if device has multiple transmit queues
3528 * @dev: network device
3529 *
3530 * Check if device has multiple transmit queues
3531 */
3532static inline bool netif_is_multiqueue(const struct net_device *dev)
3533{
3534 return dev->num_tx_queues > 1;
3535}
3536
3537int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3538
3539#ifdef CONFIG_SYSFS
3540int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3541#else
3542static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3543 unsigned int rxqs)
3544{
3545 dev->real_num_rx_queues = rxqs;
3546 return 0;
3547}
3548#endif
3549
3550static inline struct netdev_rx_queue *
3551__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3552{
3553 return dev->_rx + rxq;
3554}
3555
3556#ifdef CONFIG_SYSFS
3557static inline unsigned int get_netdev_rx_queue_index(
3558 struct netdev_rx_queue *queue)
3559{
3560 struct net_device *dev = queue->dev;
3561 int index = queue - dev->_rx;
3562
3563 BUG_ON(index >= dev->num_rx_queues);
3564 return index;
3565}
3566#endif
3567
3568#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3569int netif_get_num_default_rss_queues(void);
3570
3571enum skb_free_reason {
3572 SKB_REASON_CONSUMED,
3573 SKB_REASON_DROPPED,
3574};
3575
3576void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3577void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3578
3579/*
3580 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3581 * interrupt context or with hardware interrupts being disabled.
3582 * (in_irq() || irqs_disabled())
3583 *
3584 * We provide four helpers that can be used in following contexts :
3585 *
3586 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3587 * replacing kfree_skb(skb)
3588 *
3589 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3590 * Typically used in place of consume_skb(skb) in TX completion path
3591 *
3592 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3593 * replacing kfree_skb(skb)
3594 *
3595 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3596 * and consumed a packet. Used in place of consume_skb(skb)
3597 */
3598static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3599{
3600 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3601}
3602
3603static inline void dev_consume_skb_irq(struct sk_buff *skb)
3604{
3605 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3606}
3607
3608static inline void dev_kfree_skb_any(struct sk_buff *skb)
3609{
3610 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3611}
3612
3613static inline void dev_consume_skb_any(struct sk_buff *skb)
3614{
3615 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3616}
3617
3618void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3619int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3620int netif_rx(struct sk_buff *skb);
3621int netif_rx_ni(struct sk_buff *skb);
3622int netif_receive_skb(struct sk_buff *skb);
3623int netif_receive_skb_core(struct sk_buff *skb);
3624void netif_receive_skb_list(struct list_head *head);
3625gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3626void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3627struct sk_buff *napi_get_frags(struct napi_struct *napi);
3628gro_result_t napi_gro_frags(struct napi_struct *napi);
3629struct packet_offload *gro_find_receive_by_type(__be16 type);
3630struct packet_offload *gro_find_complete_by_type(__be16 type);
3631
3632static inline void napi_free_frags(struct napi_struct *napi)
3633{
3634 kfree_skb(napi->skb);
3635 napi->skb = NULL;
3636}
3637
3638bool netdev_is_rx_handler_busy(struct net_device *dev);
3639int netdev_rx_handler_register(struct net_device *dev,
3640 rx_handler_func_t *rx_handler,
3641 void *rx_handler_data);
3642void netdev_rx_handler_unregister(struct net_device *dev);
3643
3644bool dev_valid_name(const char *name);
3645int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3646 bool *need_copyout);
3647int dev_ifconf(struct net *net, struct ifconf *, int);
3648int dev_ethtool(struct net *net, struct ifreq *);
3649unsigned int dev_get_flags(const struct net_device *);
3650int __dev_change_flags(struct net_device *dev, unsigned int flags,
3651 struct netlink_ext_ack *extack);
3652int dev_change_flags(struct net_device *dev, unsigned int flags,
3653 struct netlink_ext_ack *extack);
3654void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3655 unsigned int gchanges);
3656int dev_change_name(struct net_device *, const char *);
3657int dev_set_alias(struct net_device *, const char *, size_t);
3658int dev_get_alias(const struct net_device *, char *, size_t);
3659int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3660int __dev_set_mtu(struct net_device *, int);
3661int dev_set_mtu_ext(struct net_device *dev, int mtu,
3662 struct netlink_ext_ack *extack);
3663int dev_set_mtu(struct net_device *, int);
3664int dev_change_tx_queue_len(struct net_device *, unsigned long);
3665void dev_set_group(struct net_device *, int);
3666int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3667 struct netlink_ext_ack *extack);
3668int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3669 struct netlink_ext_ack *extack);
3670int dev_change_carrier(struct net_device *, bool new_carrier);
3671int dev_get_phys_port_id(struct net_device *dev,
3672 struct netdev_phys_item_id *ppid);
3673int dev_get_phys_port_name(struct net_device *dev,
3674 char *name, size_t len);
3675int dev_get_port_parent_id(struct net_device *dev,
3676 struct netdev_phys_item_id *ppid, bool recurse);
3677bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3678int dev_change_proto_down(struct net_device *dev, bool proto_down);
3679int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3680struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3681struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3682 struct netdev_queue *txq, int *ret);
3683
3684typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3685int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3686 int fd, u32 flags);
3687u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3688 enum bpf_netdev_command cmd);
3689int xdp_umem_query(struct net_device *dev, u16 queue_id);
3690
3691int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3692int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3693bool is_skb_forwardable(const struct net_device *dev,
3694 const struct sk_buff *skb);
3695
3696static __always_inline int ____dev_forward_skb(struct net_device *dev,
3697 struct sk_buff *skb)
3698{
3699 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3700 unlikely(!is_skb_forwardable(dev, skb))) {
3701 atomic_long_inc(&dev->rx_dropped);
3702 kfree_skb(skb);
3703 return NET_RX_DROP;
3704 }
3705
3706 skb_scrub_packet(skb, true);
3707 skb->priority = 0;
3708 return 0;
3709}
3710
3711bool dev_nit_active(struct net_device *dev);
3712void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3713
3714extern int netdev_budget;
3715extern unsigned int netdev_budget_usecs;
3716
3717/* Called by rtnetlink.c:rtnl_unlock() */
3718void netdev_run_todo(void);
3719
3720/**
3721 * dev_put - release reference to device
3722 * @dev: network device
3723 *
3724 * Release reference to device to allow it to be freed.
3725 */
3726static inline void dev_put(struct net_device *dev)
3727{
3728 this_cpu_dec(*dev->pcpu_refcnt);
3729}
3730
3731/**
3732 * dev_hold - get reference to device
3733 * @dev: network device
3734 *
3735 * Hold reference to device to keep it from being freed.
3736 */
3737static inline void dev_hold(struct net_device *dev)
3738{
3739 this_cpu_inc(*dev->pcpu_refcnt);
3740}
3741
3742/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3743 * and _off may be called from IRQ context, but it is caller
3744 * who is responsible for serialization of these calls.
3745 *
3746 * The name carrier is inappropriate, these functions should really be
3747 * called netif_lowerlayer_*() because they represent the state of any
3748 * kind of lower layer not just hardware media.
3749 */
3750
3751void linkwatch_init_dev(struct net_device *dev);
3752void linkwatch_fire_event(struct net_device *dev);
3753void linkwatch_forget_dev(struct net_device *dev);
3754
3755/**
3756 * netif_carrier_ok - test if carrier present
3757 * @dev: network device
3758 *
3759 * Check if carrier is present on device
3760 */
3761static inline bool netif_carrier_ok(const struct net_device *dev)
3762{
3763 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3764}
3765
3766unsigned long dev_trans_start(struct net_device *dev);
3767
3768void __netdev_watchdog_up(struct net_device *dev);
3769
3770void netif_carrier_on(struct net_device *dev);
3771
3772void netif_carrier_off(struct net_device *dev);
3773
3774/**
3775 * netif_dormant_on - mark device as dormant.
3776 * @dev: network device
3777 *
3778 * Mark device as dormant (as per RFC2863).
3779 *
3780 * The dormant state indicates that the relevant interface is not
3781 * actually in a condition to pass packets (i.e., it is not 'up') but is
3782 * in a "pending" state, waiting for some external event. For "on-
3783 * demand" interfaces, this new state identifies the situation where the
3784 * interface is waiting for events to place it in the up state.
3785 */
3786static inline void netif_dormant_on(struct net_device *dev)
3787{
3788 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3789 linkwatch_fire_event(dev);
3790}
3791
3792/**
3793 * netif_dormant_off - set device as not dormant.
3794 * @dev: network device
3795 *
3796 * Device is not in dormant state.
3797 */
3798static inline void netif_dormant_off(struct net_device *dev)
3799{
3800 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3801 linkwatch_fire_event(dev);
3802}
3803
3804/**
3805 * netif_dormant - test if device is dormant
3806 * @dev: network device
3807 *
3808 * Check if device is dormant.
3809 */
3810static inline bool netif_dormant(const struct net_device *dev)
3811{
3812 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3813}
3814
3815
3816/**
3817 * netif_oper_up - test if device is operational
3818 * @dev: network device
3819 *
3820 * Check if carrier is operational
3821 */
3822static inline bool netif_oper_up(const struct net_device *dev)
3823{
3824 return (dev->operstate == IF_OPER_UP ||
3825 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3826}
3827
3828/**
3829 * netif_device_present - is device available or removed
3830 * @dev: network device
3831 *
3832 * Check if device has not been removed from system.
3833 */
3834static inline bool netif_device_present(struct net_device *dev)
3835{
3836 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3837}
3838
3839void netif_device_detach(struct net_device *dev);
3840
3841void netif_device_attach(struct net_device *dev);
3842
3843/*
3844 * Network interface message level settings
3845 */
3846
3847enum {
3848 NETIF_MSG_DRV = 0x0001,
3849 NETIF_MSG_PROBE = 0x0002,
3850 NETIF_MSG_LINK = 0x0004,
3851 NETIF_MSG_TIMER = 0x0008,
3852 NETIF_MSG_IFDOWN = 0x0010,
3853 NETIF_MSG_IFUP = 0x0020,
3854 NETIF_MSG_RX_ERR = 0x0040,
3855 NETIF_MSG_TX_ERR = 0x0080,
3856 NETIF_MSG_TX_QUEUED = 0x0100,
3857 NETIF_MSG_INTR = 0x0200,
3858 NETIF_MSG_TX_DONE = 0x0400,
3859 NETIF_MSG_RX_STATUS = 0x0800,
3860 NETIF_MSG_PKTDATA = 0x1000,
3861 NETIF_MSG_HW = 0x2000,
3862 NETIF_MSG_WOL = 0x4000,
3863};
3864
3865#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3866#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3867#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3868#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3869#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3870#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3871#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3872#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3873#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3874#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3875#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3876#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3877#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3878#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3879#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3880
3881static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3882{
3883 /* use default */
3884 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3885 return default_msg_enable_bits;
3886 if (debug_value == 0) /* no output */
3887 return 0;
3888 /* set low N bits */
3889 return (1U << debug_value) - 1;
3890}
3891
3892static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3893{
3894 spin_lock(&txq->_xmit_lock);
3895 txq->xmit_lock_owner = cpu;
3896}
3897
3898static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3899{
3900 __acquire(&txq->_xmit_lock);
3901 return true;
3902}
3903
3904static inline void __netif_tx_release(struct netdev_queue *txq)
3905{
3906 __release(&txq->_xmit_lock);
3907}
3908
3909static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3910{
3911 spin_lock_bh(&txq->_xmit_lock);
3912 txq->xmit_lock_owner = smp_processor_id();
3913}
3914
3915static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3916{
3917 bool ok = spin_trylock(&txq->_xmit_lock);
3918 if (likely(ok))
3919 txq->xmit_lock_owner = smp_processor_id();
3920 return ok;
3921}
3922
3923static inline void __netif_tx_unlock(struct netdev_queue *txq)
3924{
3925 txq->xmit_lock_owner = -1;
3926 spin_unlock(&txq->_xmit_lock);
3927}
3928
3929static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3930{
3931 txq->xmit_lock_owner = -1;
3932 spin_unlock_bh(&txq->_xmit_lock);
3933}
3934
3935static inline void txq_trans_update(struct netdev_queue *txq)
3936{
3937 if (txq->xmit_lock_owner != -1)
3938 txq->trans_start = jiffies;
3939}
3940
3941/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3942static inline void netif_trans_update(struct net_device *dev)
3943{
3944 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3945
3946 if (txq->trans_start != jiffies)
3947 txq->trans_start = jiffies;
3948}
3949
3950/**
3951 * netif_tx_lock - grab network device transmit lock
3952 * @dev: network device
3953 *
3954 * Get network device transmit lock
3955 */
3956static inline void netif_tx_lock(struct net_device *dev)
3957{
3958 unsigned int i;
3959 int cpu;
3960
3961 spin_lock(&dev->tx_global_lock);
3962 cpu = smp_processor_id();
3963 for (i = 0; i < dev->num_tx_queues; i++) {
3964 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3965
3966 /* We are the only thread of execution doing a
3967 * freeze, but we have to grab the _xmit_lock in
3968 * order to synchronize with threads which are in
3969 * the ->hard_start_xmit() handler and already
3970 * checked the frozen bit.
3971 */
3972 __netif_tx_lock(txq, cpu);
3973 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3974 __netif_tx_unlock(txq);
3975 }
3976}
3977
3978static inline void netif_tx_lock_bh(struct net_device *dev)
3979{
3980 local_bh_disable();
3981 netif_tx_lock(dev);
3982}
3983
3984static inline void netif_tx_unlock(struct net_device *dev)
3985{
3986 unsigned int i;
3987
3988 for (i = 0; i < dev->num_tx_queues; i++) {
3989 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3990
3991 /* No need to grab the _xmit_lock here. If the
3992 * queue is not stopped for another reason, we
3993 * force a schedule.
3994 */
3995 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3996 netif_schedule_queue(txq);
3997 }
3998 spin_unlock(&dev->tx_global_lock);
3999}
4000
4001static inline void netif_tx_unlock_bh(struct net_device *dev)
4002{
4003 netif_tx_unlock(dev);
4004 local_bh_enable();
4005}
4006
4007#define HARD_TX_LOCK(dev, txq, cpu) { \
4008 if ((dev->features & NETIF_F_LLTX) == 0) { \
4009 __netif_tx_lock(txq, cpu); \
4010 } else { \
4011 __netif_tx_acquire(txq); \
4012 } \
4013}
4014
4015#define HARD_TX_TRYLOCK(dev, txq) \
4016 (((dev->features & NETIF_F_LLTX) == 0) ? \
4017 __netif_tx_trylock(txq) : \
4018 __netif_tx_acquire(txq))
4019
4020#define HARD_TX_UNLOCK(dev, txq) { \
4021 if ((dev->features & NETIF_F_LLTX) == 0) { \
4022 __netif_tx_unlock(txq); \
4023 } else { \
4024 __netif_tx_release(txq); \
4025 } \
4026}
4027
4028static inline void netif_tx_disable(struct net_device *dev)
4029{
4030 unsigned int i;
4031 int cpu;
4032
4033 local_bh_disable();
4034 cpu = smp_processor_id();
4035 for (i = 0; i < dev->num_tx_queues; i++) {
4036 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4037
4038 __netif_tx_lock(txq, cpu);
4039 netif_tx_stop_queue(txq);
4040 __netif_tx_unlock(txq);
4041 }
4042 local_bh_enable();
4043}
4044
4045static inline void netif_addr_lock(struct net_device *dev)
4046{
4047 spin_lock(&dev->addr_list_lock);
4048}
4049
4050static inline void netif_addr_lock_nested(struct net_device *dev)
4051{
4052 int subclass = SINGLE_DEPTH_NESTING;
4053
4054 if (dev->netdev_ops->ndo_get_lock_subclass)
4055 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
4056
4057 spin_lock_nested(&dev->addr_list_lock, subclass);
4058}
4059
4060static inline void netif_addr_lock_bh(struct net_device *dev)
4061{
4062 spin_lock_bh(&dev->addr_list_lock);
4063}
4064
4065static inline void netif_addr_unlock(struct net_device *dev)
4066{
4067 spin_unlock(&dev->addr_list_lock);
4068}
4069
4070static inline void netif_addr_unlock_bh(struct net_device *dev)
4071{
4072 spin_unlock_bh(&dev->addr_list_lock);
4073}
4074
4075/*
4076 * dev_addrs walker. Should be used only for read access. Call with
4077 * rcu_read_lock held.
4078 */
4079#define for_each_dev_addr(dev, ha) \
4080 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4081
4082/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4083
4084void ether_setup(struct net_device *dev);
4085
4086/* Support for loadable net-drivers */
4087struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4088 unsigned char name_assign_type,
4089 void (*setup)(struct net_device *),
4090 unsigned int txqs, unsigned int rxqs);
4091int dev_get_valid_name(struct net *net, struct net_device *dev,
4092 const char *name);
4093
4094#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4095 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4096
4097#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4098 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4099 count)
4100
4101int register_netdev(struct net_device *dev);
4102void unregister_netdev(struct net_device *dev);
4103
4104/* General hardware address lists handling functions */
4105int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4106 struct netdev_hw_addr_list *from_list, int addr_len);
4107void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4108 struct netdev_hw_addr_list *from_list, int addr_len);
4109int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4110 struct net_device *dev,
4111 int (*sync)(struct net_device *, const unsigned char *),
4112 int (*unsync)(struct net_device *,
4113 const unsigned char *));
4114int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4115 struct net_device *dev,
4116 int (*sync)(struct net_device *,
4117 const unsigned char *, int),
4118 int (*unsync)(struct net_device *,
4119 const unsigned char *, int));
4120void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4121 struct net_device *dev,
4122 int (*unsync)(struct net_device *,
4123 const unsigned char *, int));
4124void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4125 struct net_device *dev,
4126 int (*unsync)(struct net_device *,
4127 const unsigned char *));
4128void __hw_addr_init(struct netdev_hw_addr_list *list);
4129
4130/* Functions used for device addresses handling */
4131int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4132 unsigned char addr_type);
4133int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4134 unsigned char addr_type);
4135void dev_addr_flush(struct net_device *dev);
4136int dev_addr_init(struct net_device *dev);
4137
4138/* Functions used for unicast addresses handling */
4139int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4140int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4141int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4142int dev_uc_sync(struct net_device *to, struct net_device *from);
4143int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4144void dev_uc_unsync(struct net_device *to, struct net_device *from);
4145void dev_uc_flush(struct net_device *dev);
4146void dev_uc_init(struct net_device *dev);
4147
4148/**
4149 * __dev_uc_sync - Synchonize device's unicast list
4150 * @dev: device to sync
4151 * @sync: function to call if address should be added
4152 * @unsync: function to call if address should be removed
4153 *
4154 * Add newly added addresses to the interface, and release
4155 * addresses that have been deleted.
4156 */
4157static inline int __dev_uc_sync(struct net_device *dev,
4158 int (*sync)(struct net_device *,
4159 const unsigned char *),
4160 int (*unsync)(struct net_device *,
4161 const unsigned char *))
4162{
4163 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4164}
4165
4166/**
4167 * __dev_uc_unsync - Remove synchronized addresses from device
4168 * @dev: device to sync
4169 * @unsync: function to call if address should be removed
4170 *
4171 * Remove all addresses that were added to the device by dev_uc_sync().
4172 */
4173static inline void __dev_uc_unsync(struct net_device *dev,
4174 int (*unsync)(struct net_device *,
4175 const unsigned char *))
4176{
4177 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4178}
4179
4180/* Functions used for multicast addresses handling */
4181int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4182int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4183int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4184int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4185int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4186int dev_mc_sync(struct net_device *to, struct net_device *from);
4187int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4188void dev_mc_unsync(struct net_device *to, struct net_device *from);
4189void dev_mc_flush(struct net_device *dev);
4190void dev_mc_init(struct net_device *dev);
4191
4192/**
4193 * __dev_mc_sync - Synchonize device's multicast list
4194 * @dev: device to sync
4195 * @sync: function to call if address should be added
4196 * @unsync: function to call if address should be removed
4197 *
4198 * Add newly added addresses to the interface, and release
4199 * addresses that have been deleted.
4200 */
4201static inline int __dev_mc_sync(struct net_device *dev,
4202 int (*sync)(struct net_device *,
4203 const unsigned char *),
4204 int (*unsync)(struct net_device *,
4205 const unsigned char *))
4206{
4207 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4208}
4209
4210/**
4211 * __dev_mc_unsync - Remove synchronized addresses from device
4212 * @dev: device to sync
4213 * @unsync: function to call if address should be removed
4214 *
4215 * Remove all addresses that were added to the device by dev_mc_sync().
4216 */
4217static inline void __dev_mc_unsync(struct net_device *dev,
4218 int (*unsync)(struct net_device *,
4219 const unsigned char *))
4220{
4221 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4222}
4223
4224/* Functions used for secondary unicast and multicast support */
4225void dev_set_rx_mode(struct net_device *dev);
4226void __dev_set_rx_mode(struct net_device *dev);
4227int dev_set_promiscuity(struct net_device *dev, int inc);
4228int dev_set_allmulti(struct net_device *dev, int inc);
4229void netdev_state_change(struct net_device *dev);
4230void netdev_notify_peers(struct net_device *dev);
4231void netdev_features_change(struct net_device *dev);
4232/* Load a device via the kmod */
4233void dev_load(struct net *net, const char *name);
4234struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4235 struct rtnl_link_stats64 *storage);
4236void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4237 const struct net_device_stats *netdev_stats);
4238
4239extern int netdev_max_backlog;
4240extern int netdev_tstamp_prequeue;
4241extern int weight_p;
4242extern int dev_weight_rx_bias;
4243extern int dev_weight_tx_bias;
4244extern int dev_rx_weight;
4245extern int dev_tx_weight;
4246
4247bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4248struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4249 struct list_head **iter);
4250struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4251 struct list_head **iter);
4252
4253/* iterate through upper list, must be called under RCU read lock */
4254#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4255 for (iter = &(dev)->adj_list.upper, \
4256 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4257 updev; \
4258 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4259
4260int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4261 int (*fn)(struct net_device *upper_dev,
4262 void *data),
4263 void *data);
4264
4265bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4266 struct net_device *upper_dev);
4267
4268bool netdev_has_any_upper_dev(struct net_device *dev);
4269
4270void *netdev_lower_get_next_private(struct net_device *dev,
4271 struct list_head **iter);
4272void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4273 struct list_head **iter);
4274
4275#define netdev_for_each_lower_private(dev, priv, iter) \
4276 for (iter = (dev)->adj_list.lower.next, \
4277 priv = netdev_lower_get_next_private(dev, &(iter)); \
4278 priv; \
4279 priv = netdev_lower_get_next_private(dev, &(iter)))
4280
4281#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4282 for (iter = &(dev)->adj_list.lower, \
4283 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4284 priv; \
4285 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4286
4287void *netdev_lower_get_next(struct net_device *dev,
4288 struct list_head **iter);
4289
4290#define netdev_for_each_lower_dev(dev, ldev, iter) \
4291 for (iter = (dev)->adj_list.lower.next, \
4292 ldev = netdev_lower_get_next(dev, &(iter)); \
4293 ldev; \
4294 ldev = netdev_lower_get_next(dev, &(iter)))
4295
4296struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4297 struct list_head **iter);
4298struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4299 struct list_head **iter);
4300
4301int netdev_walk_all_lower_dev(struct net_device *dev,
4302 int (*fn)(struct net_device *lower_dev,
4303 void *data),
4304 void *data);
4305int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4306 int (*fn)(struct net_device *lower_dev,
4307 void *data),
4308 void *data);
4309
4310void *netdev_adjacent_get_private(struct list_head *adj_list);
4311void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4312struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4313struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4314int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4315 struct netlink_ext_ack *extack);
4316int netdev_master_upper_dev_link(struct net_device *dev,
4317 struct net_device *upper_dev,
4318 void *upper_priv, void *upper_info,
4319 struct netlink_ext_ack *extack);
4320void netdev_upper_dev_unlink(struct net_device *dev,
4321 struct net_device *upper_dev);
4322void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4323void *netdev_lower_dev_get_private(struct net_device *dev,
4324 struct net_device *lower_dev);
4325void netdev_lower_state_changed(struct net_device *lower_dev,
4326 void *lower_state_info);
4327
4328/* RSS keys are 40 or 52 bytes long */
4329#define NETDEV_RSS_KEY_LEN 52
4330extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4331void netdev_rss_key_fill(void *buffer, size_t len);
4332
4333int dev_get_nest_level(struct net_device *dev);
4334int skb_checksum_help(struct sk_buff *skb);
4335int skb_crc32c_csum_help(struct sk_buff *skb);
4336int skb_csum_hwoffload_help(struct sk_buff *skb,
4337 const netdev_features_t features);
4338
4339struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4340 netdev_features_t features, bool tx_path);
4341struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4342 netdev_features_t features);
4343
4344struct netdev_bonding_info {
4345 ifslave slave;
4346 ifbond master;
4347};
4348
4349struct netdev_notifier_bonding_info {
4350 struct netdev_notifier_info info; /* must be first */
4351 struct netdev_bonding_info bonding_info;
4352};
4353
4354void netdev_bonding_info_change(struct net_device *dev,
4355 struct netdev_bonding_info *bonding_info);
4356
4357static inline
4358struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4359{
4360 return __skb_gso_segment(skb, features, true);
4361}
4362__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4363
4364static inline bool can_checksum_protocol(netdev_features_t features,
4365 __be16 protocol)
4366{
4367 if (protocol == htons(ETH_P_FCOE))
4368 return !!(features & NETIF_F_FCOE_CRC);
4369
4370 /* Assume this is an IP checksum (not SCTP CRC) */
4371
4372 if (features & NETIF_F_HW_CSUM) {
4373 /* Can checksum everything */
4374 return true;
4375 }
4376
4377 switch (protocol) {
4378 case htons(ETH_P_IP):
4379 return !!(features & NETIF_F_IP_CSUM);
4380 case htons(ETH_P_IPV6):
4381 return !!(features & NETIF_F_IPV6_CSUM);
4382 default:
4383 return false;
4384 }
4385}
4386
4387#ifdef CONFIG_BUG
4388void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4389#else
4390static inline void netdev_rx_csum_fault(struct net_device *dev,
4391 struct sk_buff *skb)
4392{
4393}
4394#endif
4395/* rx skb timestamps */
4396void net_enable_timestamp(void);
4397void net_disable_timestamp(void);
4398
4399#ifdef CONFIG_PROC_FS
4400int __init dev_proc_init(void);
4401#else
4402#define dev_proc_init() 0
4403#endif
4404
4405static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4406 struct sk_buff *skb, struct net_device *dev,
4407 bool more)
4408{
4409 __this_cpu_write(softnet_data.xmit.more, more);
4410 return ops->ndo_start_xmit(skb, dev);
4411}
4412
4413static inline bool netdev_xmit_more(void)
4414{
4415 return __this_cpu_read(softnet_data.xmit.more);
4416}
4417
4418static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4419 struct netdev_queue *txq, bool more)
4420{
4421 const struct net_device_ops *ops = dev->netdev_ops;
4422 netdev_tx_t rc;
4423
4424 rc = __netdev_start_xmit(ops, skb, dev, more);
4425 if (rc == NETDEV_TX_OK)
4426 txq_trans_update(txq);
4427
4428 return rc;
4429}
4430
4431int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4432 const void *ns);
4433void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4434 const void *ns);
4435
4436static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4437{
4438 return netdev_class_create_file_ns(class_attr, NULL);
4439}
4440
4441static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4442{
4443 netdev_class_remove_file_ns(class_attr, NULL);
4444}
4445
4446extern const struct kobj_ns_type_operations net_ns_type_operations;
4447
4448const char *netdev_drivername(const struct net_device *dev);
4449
4450void linkwatch_run_queue(void);
4451
4452static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4453 netdev_features_t f2)
4454{
4455 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4456 if (f1 & NETIF_F_HW_CSUM)
4457 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4458 else
4459 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4460 }
4461
4462 return f1 & f2;
4463}
4464
4465static inline netdev_features_t netdev_get_wanted_features(
4466 struct net_device *dev)
4467{
4468 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4469}
4470netdev_features_t netdev_increment_features(netdev_features_t all,
4471 netdev_features_t one, netdev_features_t mask);
4472
4473/* Allow TSO being used on stacked device :
4474 * Performing the GSO segmentation before last device
4475 * is a performance improvement.
4476 */
4477static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4478 netdev_features_t mask)
4479{
4480 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4481}
4482
4483int __netdev_update_features(struct net_device *dev);
4484void netdev_update_features(struct net_device *dev);
4485void netdev_change_features(struct net_device *dev);
4486
4487void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4488 struct net_device *dev);
4489
4490netdev_features_t passthru_features_check(struct sk_buff *skb,
4491 struct net_device *dev,
4492 netdev_features_t features);
4493netdev_features_t netif_skb_features(struct sk_buff *skb);
4494
4495static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4496{
4497 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4498
4499 /* check flags correspondence */
4500 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4501 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4502 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4503 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4504 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4505 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4506 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4507 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4508 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4509 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4510 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4511 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4512 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4513 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4514 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4515 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4516 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4517 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4518
4519 return (features & feature) == feature;
4520}
4521
4522static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4523{
4524 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4525 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4526}
4527
4528static inline bool netif_needs_gso(struct sk_buff *skb,
4529 netdev_features_t features)
4530{
4531 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4532 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4533 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4534}
4535
4536static inline void netif_set_gso_max_size(struct net_device *dev,
4537 unsigned int size)
4538{
4539 dev->gso_max_size = size;
4540}
4541
4542static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4543 int pulled_hlen, u16 mac_offset,
4544 int mac_len)
4545{
4546 skb->protocol = protocol;
4547 skb->encapsulation = 1;
4548 skb_push(skb, pulled_hlen);
4549 skb_reset_transport_header(skb);
4550 skb->mac_header = mac_offset;
4551 skb->network_header = skb->mac_header + mac_len;
4552 skb->mac_len = mac_len;
4553}
4554
4555static inline bool netif_is_macsec(const struct net_device *dev)
4556{
4557 return dev->priv_flags & IFF_MACSEC;
4558}
4559
4560static inline bool netif_is_macvlan(const struct net_device *dev)
4561{
4562 return dev->priv_flags & IFF_MACVLAN;
4563}
4564
4565static inline bool netif_is_macvlan_port(const struct net_device *dev)
4566{
4567 return dev->priv_flags & IFF_MACVLAN_PORT;
4568}
4569
4570static inline bool netif_is_bond_master(const struct net_device *dev)
4571{
4572 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4573}
4574
4575static inline bool netif_is_bond_slave(const struct net_device *dev)
4576{
4577 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4578}
4579
4580static inline bool netif_supports_nofcs(struct net_device *dev)
4581{
4582 return dev->priv_flags & IFF_SUPP_NOFCS;
4583}
4584
4585static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4586{
4587 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4588}
4589
4590static inline bool netif_is_l3_master(const struct net_device *dev)
4591{
4592 return dev->priv_flags & IFF_L3MDEV_MASTER;
4593}
4594
4595static inline bool netif_is_l3_slave(const struct net_device *dev)
4596{
4597 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4598}
4599
4600static inline bool netif_is_bridge_master(const struct net_device *dev)
4601{
4602 return dev->priv_flags & IFF_EBRIDGE;
4603}
4604
4605static inline bool netif_is_bridge_port(const struct net_device *dev)
4606{
4607 return dev->priv_flags & IFF_BRIDGE_PORT;
4608}
4609
4610static inline bool netif_is_ovs_master(const struct net_device *dev)
4611{
4612 return dev->priv_flags & IFF_OPENVSWITCH;
4613}
4614
4615static inline bool netif_is_ovs_port(const struct net_device *dev)
4616{
4617 return dev->priv_flags & IFF_OVS_DATAPATH;
4618}
4619
4620static inline bool netif_is_team_master(const struct net_device *dev)
4621{
4622 return dev->priv_flags & IFF_TEAM;
4623}
4624
4625static inline bool netif_is_team_port(const struct net_device *dev)
4626{
4627 return dev->priv_flags & IFF_TEAM_PORT;
4628}
4629
4630static inline bool netif_is_lag_master(const struct net_device *dev)
4631{
4632 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4633}
4634
4635static inline bool netif_is_lag_port(const struct net_device *dev)
4636{
4637 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4638}
4639
4640static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4641{
4642 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4643}
4644
4645static inline bool netif_is_failover(const struct net_device *dev)
4646{
4647 return dev->priv_flags & IFF_FAILOVER;
4648}
4649
4650static inline bool netif_is_failover_slave(const struct net_device *dev)
4651{
4652 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4653}
4654
4655/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4656static inline void netif_keep_dst(struct net_device *dev)
4657{
4658 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4659}
4660
4661/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4662static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4663{
4664 /* TODO: reserve and use an additional IFF bit, if we get more users */
4665 return dev->priv_flags & IFF_MACSEC;
4666}
4667
4668extern struct pernet_operations __net_initdata loopback_net_ops;
4669
4670/* Logging, debugging and troubleshooting/diagnostic helpers. */
4671
4672/* netdev_printk helpers, similar to dev_printk */
4673
4674static inline const char *netdev_name(const struct net_device *dev)
4675{
4676 if (!dev->name[0] || strchr(dev->name, '%'))
4677 return "(unnamed net_device)";
4678 return dev->name;
4679}
4680
4681static inline bool netdev_unregistering(const struct net_device *dev)
4682{
4683 return dev->reg_state == NETREG_UNREGISTERING;
4684}
4685
4686static inline const char *netdev_reg_state(const struct net_device *dev)
4687{
4688 switch (dev->reg_state) {
4689 case NETREG_UNINITIALIZED: return " (uninitialized)";
4690 case NETREG_REGISTERED: return "";
4691 case NETREG_UNREGISTERING: return " (unregistering)";
4692 case NETREG_UNREGISTERED: return " (unregistered)";
4693 case NETREG_RELEASED: return " (released)";
4694 case NETREG_DUMMY: return " (dummy)";
4695 }
4696
4697 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4698 return " (unknown)";
4699}
4700
4701__printf(3, 4) __cold
4702void netdev_printk(const char *level, const struct net_device *dev,
4703 const char *format, ...);
4704__printf(2, 3) __cold
4705void netdev_emerg(const struct net_device *dev, const char *format, ...);
4706__printf(2, 3) __cold
4707void netdev_alert(const struct net_device *dev, const char *format, ...);
4708__printf(2, 3) __cold
4709void netdev_crit(const struct net_device *dev, const char *format, ...);
4710__printf(2, 3) __cold
4711void netdev_err(const struct net_device *dev, const char *format, ...);
4712__printf(2, 3) __cold
4713void netdev_warn(const struct net_device *dev, const char *format, ...);
4714__printf(2, 3) __cold
4715void netdev_notice(const struct net_device *dev, const char *format, ...);
4716__printf(2, 3) __cold
4717void netdev_info(const struct net_device *dev, const char *format, ...);
4718
4719#define netdev_level_once(level, dev, fmt, ...) \
4720do { \
4721 static bool __print_once __read_mostly; \
4722 \
4723 if (!__print_once) { \
4724 __print_once = true; \
4725 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4726 } \
4727} while (0)
4728
4729#define netdev_emerg_once(dev, fmt, ...) \
4730 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4731#define netdev_alert_once(dev, fmt, ...) \
4732 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4733#define netdev_crit_once(dev, fmt, ...) \
4734 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4735#define netdev_err_once(dev, fmt, ...) \
4736 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4737#define netdev_warn_once(dev, fmt, ...) \
4738 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4739#define netdev_notice_once(dev, fmt, ...) \
4740 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4741#define netdev_info_once(dev, fmt, ...) \
4742 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4743
4744#define MODULE_ALIAS_NETDEV(device) \
4745 MODULE_ALIAS("netdev-" device)
4746
4747#if defined(CONFIG_DYNAMIC_DEBUG)
4748#define netdev_dbg(__dev, format, args...) \
4749do { \
4750 dynamic_netdev_dbg(__dev, format, ##args); \
4751} while (0)
4752#elif defined(DEBUG)
4753#define netdev_dbg(__dev, format, args...) \
4754 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4755#else
4756#define netdev_dbg(__dev, format, args...) \
4757({ \
4758 if (0) \
4759 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4760})
4761#endif
4762
4763#if defined(VERBOSE_DEBUG)
4764#define netdev_vdbg netdev_dbg
4765#else
4766
4767#define netdev_vdbg(dev, format, args...) \
4768({ \
4769 if (0) \
4770 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4771 0; \
4772})
4773#endif
4774
4775/*
4776 * netdev_WARN() acts like dev_printk(), but with the key difference
4777 * of using a WARN/WARN_ON to get the message out, including the
4778 * file/line information and a backtrace.
4779 */
4780#define netdev_WARN(dev, format, args...) \
4781 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4782 netdev_reg_state(dev), ##args)
4783
4784#define netdev_WARN_ONCE(dev, format, args...) \
4785 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4786 netdev_reg_state(dev), ##args)
4787
4788/* netif printk helpers, similar to netdev_printk */
4789
4790#define netif_printk(priv, type, level, dev, fmt, args...) \
4791do { \
4792 if (netif_msg_##type(priv)) \
4793 netdev_printk(level, (dev), fmt, ##args); \
4794} while (0)
4795
4796#define netif_level(level, priv, type, dev, fmt, args...) \
4797do { \
4798 if (netif_msg_##type(priv)) \
4799 netdev_##level(dev, fmt, ##args); \
4800} while (0)
4801
4802#define netif_emerg(priv, type, dev, fmt, args...) \
4803 netif_level(emerg, priv, type, dev, fmt, ##args)
4804#define netif_alert(priv, type, dev, fmt, args...) \
4805 netif_level(alert, priv, type, dev, fmt, ##args)
4806#define netif_crit(priv, type, dev, fmt, args...) \
4807 netif_level(crit, priv, type, dev, fmt, ##args)
4808#define netif_err(priv, type, dev, fmt, args...) \
4809 netif_level(err, priv, type, dev, fmt, ##args)
4810#define netif_warn(priv, type, dev, fmt, args...) \
4811 netif_level(warn, priv, type, dev, fmt, ##args)
4812#define netif_notice(priv, type, dev, fmt, args...) \
4813 netif_level(notice, priv, type, dev, fmt, ##args)
4814#define netif_info(priv, type, dev, fmt, args...) \
4815 netif_level(info, priv, type, dev, fmt, ##args)
4816
4817#if defined(CONFIG_DYNAMIC_DEBUG)
4818#define netif_dbg(priv, type, netdev, format, args...) \
4819do { \
4820 if (netif_msg_##type(priv)) \
4821 dynamic_netdev_dbg(netdev, format, ##args); \
4822} while (0)
4823#elif defined(DEBUG)
4824#define netif_dbg(priv, type, dev, format, args...) \
4825 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4826#else
4827#define netif_dbg(priv, type, dev, format, args...) \
4828({ \
4829 if (0) \
4830 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4831 0; \
4832})
4833#endif
4834
4835/* if @cond then downgrade to debug, else print at @level */
4836#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4837 do { \
4838 if (cond) \
4839 netif_dbg(priv, type, netdev, fmt, ##args); \
4840 else \
4841 netif_ ## level(priv, type, netdev, fmt, ##args); \
4842 } while (0)
4843
4844#if defined(VERBOSE_DEBUG)
4845#define netif_vdbg netif_dbg
4846#else
4847#define netif_vdbg(priv, type, dev, format, args...) \
4848({ \
4849 if (0) \
4850 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4851 0; \
4852})
4853#endif
4854
4855/*
4856 * The list of packet types we will receive (as opposed to discard)
4857 * and the routines to invoke.
4858 *
4859 * Why 16. Because with 16 the only overlap we get on a hash of the
4860 * low nibble of the protocol value is RARP/SNAP/X.25.
4861 *
4862 * 0800 IP
4863 * 0001 802.3
4864 * 0002 AX.25
4865 * 0004 802.2
4866 * 8035 RARP
4867 * 0005 SNAP
4868 * 0805 X.25
4869 * 0806 ARP
4870 * 8137 IPX
4871 * 0009 Localtalk
4872 * 86DD IPv6
4873 */
4874#define PTYPE_HASH_SIZE (16)
4875#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4876
4877#endif /* _LINUX_NETDEVICE_H */