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