netlink: revert broken, broken "2-clause nla_ok()"
[linux-2.6/btrfs-unstable.git] / include / linux / netdevice.h
blob994f7423a74bd622884c3b646f4123d28697b8ad
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.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
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>
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.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
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>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
55 #include <linux/hashtable.h>
57 struct netpoll_info;
58 struct device;
59 struct phy_device;
60 /* 802.11 specific */
61 struct wireless_dev;
62 /* 802.15.4 specific */
63 struct wpan_dev;
64 struct mpls_dev;
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
67 struct bpf_prog;
69 void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
72 /* Backlog congestion levels */
73 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74 #define NET_RX_DROP 1 /* packet dropped */
77 * Transmit return codes: transmit return codes originate from three different
78 * namespaces:
80 * - qdisc return codes
81 * - driver transmit return codes
82 * - errno values
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
90 * others are propagated to higher layers.
93 /* qdisc ->enqueue() return codes. */
94 #define NET_XMIT_SUCCESS 0x00
95 #define NET_XMIT_DROP 0x01 /* skb dropped */
96 #define NET_XMIT_CN 0x02 /* congestion notification */
97 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
99 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
102 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
103 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105 /* Driver transmit return codes */
106 #define NETDEV_TX_MASK 0xf0
108 enum netdev_tx {
109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
113 typedef enum netdev_tx netdev_tx_t;
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 static inline bool dev_xmit_complete(int rc)
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 if (likely(rc < NET_XMIT_MASK))
128 return true;
130 return false;
134 * Compute the worst-case header length according to the protocols
135 * used.
138 #if defined(CONFIG_HYPERV_NET)
139 # define LL_MAX_HEADER 128
140 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
141 # if defined(CONFIG_MAC80211_MESH)
142 # define LL_MAX_HEADER 128
143 # else
144 # define LL_MAX_HEADER 96
145 # endif
146 #else
147 # define LL_MAX_HEADER 32
148 #endif
150 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
152 #define MAX_HEADER LL_MAX_HEADER
153 #else
154 #define MAX_HEADER (LL_MAX_HEADER + 48)
155 #endif
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
162 struct net_device_stats {
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
172 unsigned long collisions;
173 unsigned long rx_length_errors;
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
189 #include <linux/cache.h>
190 #include <linux/skbuff.h>
192 #ifdef CONFIG_RPS
193 #include <linux/static_key.h>
194 extern struct static_key rps_needed;
195 extern struct static_key rfs_needed;
196 #endif
198 struct neighbour;
199 struct neigh_parms;
200 struct sk_buff;
202 struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
205 unsigned char type;
206 #define NETDEV_HW_ADDR_T_LAN 1
207 #define NETDEV_HW_ADDR_T_SAN 2
208 #define NETDEV_HW_ADDR_T_SLAVE 3
209 #define NETDEV_HW_ADDR_T_UNICAST 4
210 #define NETDEV_HW_ADDR_T_MULTICAST 5
211 bool global_use;
212 int sync_cnt;
213 int refcount;
214 int synced;
215 struct rcu_head rcu_head;
218 struct netdev_hw_addr_list {
219 struct list_head list;
220 int count;
223 #define netdev_hw_addr_list_count(l) ((l)->count)
224 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225 #define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
228 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230 #define netdev_for_each_uc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
233 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
235 #define netdev_for_each_mc_addr(ha, dev) \
236 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
238 struct hh_cache {
239 u16 hh_len;
240 u16 __pad;
241 seqlock_t hh_lock;
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
265 struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
268 const void *saddr, unsigned int len);
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
274 bool (*validate)(const char *ll_header, unsigned int len);
277 /* These flag bits are private to the generic network queueing
278 * layer; they may not be explicitly referenced by any other
279 * code.
282 enum netdev_state_t {
283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
285 __LINK_STATE_NOCARRIER,
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
292 * This structure holds boot-time configured netdevice settings. They
293 * are then used in the device probing.
295 struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
299 #define NETDEV_BOOT_SETUP_MAX 8
301 int __init netdev_boot_setup(char *str);
304 * Structure for NAPI scheduling similar to tasklet but with weighting
306 struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-CPU poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
313 struct list_head poll_list;
315 unsigned long state;
316 int weight;
317 unsigned int gro_count;
318 int (*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
320 int poll_owner;
321 #endif
322 struct net_device *dev;
323 struct sk_buff *gro_list;
324 struct sk_buff *skb;
325 struct hrtimer timer;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
331 enum {
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_DISABLE, /* Disable pending */
334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
337 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
340 enum {
341 NAPIF_STATE_SCHED = (1UL << NAPI_STATE_SCHED),
342 NAPIF_STATE_DISABLE = (1UL << NAPI_STATE_DISABLE),
343 NAPIF_STATE_NPSVC = (1UL << NAPI_STATE_NPSVC),
344 NAPIF_STATE_HASHED = (1UL << NAPI_STATE_HASHED),
345 NAPIF_STATE_NO_BUSY_POLL = (1UL << NAPI_STATE_NO_BUSY_POLL),
346 NAPIF_STATE_IN_BUSY_POLL = (1UL << NAPI_STATE_IN_BUSY_POLL),
349 enum gro_result {
350 GRO_MERGED,
351 GRO_MERGED_FREE,
352 GRO_HELD,
353 GRO_NORMAL,
354 GRO_DROP,
356 typedef enum gro_result gro_result_t;
359 * enum rx_handler_result - Possible return values for rx_handlers.
360 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
361 * further.
362 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
363 * case skb->dev was changed by rx_handler.
364 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
365 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
367 * rx_handlers are functions called from inside __netif_receive_skb(), to do
368 * special processing of the skb, prior to delivery to protocol handlers.
370 * Currently, a net_device can only have a single rx_handler registered. Trying
371 * to register a second rx_handler will return -EBUSY.
373 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
374 * To unregister a rx_handler on a net_device, use
375 * netdev_rx_handler_unregister().
377 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
378 * do with the skb.
380 * If the rx_handler consumed the skb in some way, it should return
381 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
382 * the skb to be delivered in some other way.
384 * If the rx_handler changed skb->dev, to divert the skb to another
385 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
386 * new device will be called if it exists.
388 * If the rx_handler decides the skb should be ignored, it should return
389 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
390 * are registered on exact device (ptype->dev == skb->dev).
392 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
393 * delivered, it should return RX_HANDLER_PASS.
395 * A device without a registered rx_handler will behave as if rx_handler
396 * returned RX_HANDLER_PASS.
399 enum rx_handler_result {
400 RX_HANDLER_CONSUMED,
401 RX_HANDLER_ANOTHER,
402 RX_HANDLER_EXACT,
403 RX_HANDLER_PASS,
405 typedef enum rx_handler_result rx_handler_result_t;
406 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
408 void __napi_schedule(struct napi_struct *n);
409 void __napi_schedule_irqoff(struct napi_struct *n);
411 static inline bool napi_disable_pending(struct napi_struct *n)
413 return test_bit(NAPI_STATE_DISABLE, &n->state);
417 * napi_schedule_prep - check if NAPI can be scheduled
418 * @n: NAPI context
420 * Test if NAPI routine is already running, and if not mark
421 * it as running. This is used as a condition variable to
422 * insure only one NAPI poll instance runs. We also make
423 * sure there is no pending NAPI disable.
425 static inline bool napi_schedule_prep(struct napi_struct *n)
427 return !napi_disable_pending(n) &&
428 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
432 * napi_schedule - schedule NAPI poll
433 * @n: NAPI context
435 * Schedule NAPI poll routine to be called if it is not already
436 * running.
438 static inline void napi_schedule(struct napi_struct *n)
440 if (napi_schedule_prep(n))
441 __napi_schedule(n);
445 * napi_schedule_irqoff - schedule NAPI poll
446 * @n: NAPI context
448 * Variant of napi_schedule(), assuming hard irqs are masked.
450 static inline void napi_schedule_irqoff(struct napi_struct *n)
452 if (napi_schedule_prep(n))
453 __napi_schedule_irqoff(n);
456 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
457 static inline bool napi_reschedule(struct napi_struct *napi)
459 if (napi_schedule_prep(napi)) {
460 __napi_schedule(napi);
461 return true;
463 return false;
466 bool __napi_complete(struct napi_struct *n);
467 bool napi_complete_done(struct napi_struct *n, int work_done);
469 * napi_complete - NAPI processing complete
470 * @n: NAPI context
472 * Mark NAPI processing as complete.
473 * Consider using napi_complete_done() instead.
474 * Return false if device should avoid rearming interrupts.
476 static inline bool napi_complete(struct napi_struct *n)
478 return napi_complete_done(n, 0);
482 * napi_hash_del - remove a NAPI from global table
483 * @napi: NAPI context
485 * Warning: caller must observe RCU grace period
486 * before freeing memory containing @napi, if
487 * this function returns true.
488 * Note: core networking stack automatically calls it
489 * from netif_napi_del().
490 * Drivers might want to call this helper to combine all
491 * the needed RCU grace periods into a single one.
493 bool napi_hash_del(struct napi_struct *napi);
496 * napi_disable - prevent NAPI from scheduling
497 * @n: NAPI context
499 * Stop NAPI from being scheduled on this context.
500 * Waits till any outstanding processing completes.
502 void napi_disable(struct napi_struct *n);
505 * napi_enable - enable NAPI scheduling
506 * @n: NAPI context
508 * Resume NAPI from being scheduled on this context.
509 * Must be paired with napi_disable.
511 static inline void napi_enable(struct napi_struct *n)
513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
514 smp_mb__before_atomic();
515 clear_bit(NAPI_STATE_SCHED, &n->state);
516 clear_bit(NAPI_STATE_NPSVC, &n->state);
520 * napi_synchronize - wait until NAPI is not running
521 * @n: NAPI context
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
527 static inline void napi_synchronize(const struct napi_struct *n)
529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
531 msleep(1);
532 else
533 barrier();
536 enum netdev_queue_state_t {
537 __QUEUE_STATE_DRV_XOFF,
538 __QUEUE_STATE_STACK_XOFF,
539 __QUEUE_STATE_FROZEN,
542 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
543 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
544 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
546 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 QUEUE_STATE_FROZEN)
549 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
550 QUEUE_STATE_FROZEN)
553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
554 * netif_tx_* functions below are used to manipulate this flag. The
555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556 * queue independently. The netif_xmit_*stopped functions below are called
557 * to check if the queue has been stopped by the driver or stack (either
558 * of the XOFF bits are set in the state). Drivers should not need to call
559 * netif_xmit*stopped functions, they should only be using netif_tx_*.
562 struct netdev_queue {
564 * read-mostly part
566 struct net_device *dev;
567 struct Qdisc __rcu *qdisc;
568 struct Qdisc *qdisc_sleeping;
569 #ifdef CONFIG_SYSFS
570 struct kobject kobj;
571 #endif
572 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 int numa_node;
574 #endif
575 unsigned long tx_maxrate;
577 * Number of TX timeouts for this queue
578 * (/sys/class/net/DEV/Q/trans_timeout)
580 unsigned long trans_timeout;
582 * write-mostly part
584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
585 int xmit_lock_owner;
587 * Time (in jiffies) of last Tx
589 unsigned long trans_start;
591 unsigned long state;
593 #ifdef CONFIG_BQL
594 struct dql dql;
595 #endif
596 } ____cacheline_aligned_in_smp;
598 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 return q->numa_node;
602 #else
603 return NUMA_NO_NODE;
604 #endif
607 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
609 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 q->numa_node = node;
611 #endif
614 #ifdef CONFIG_RPS
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
619 struct rps_map {
620 unsigned int len;
621 struct rcu_head rcu;
622 u16 cpus[0];
624 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
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.
631 struct rps_dev_flow {
632 u16 cpu;
633 u16 filter;
634 unsigned int last_qtail;
636 #define RPS_NO_FILTER 0xffff
639 * The rps_dev_flow_table structure contains a table of flow mappings.
641 struct rps_dev_flow_table {
642 unsigned int mask;
643 struct rcu_head rcu;
644 struct rps_dev_flow flows[0];
646 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
647 ((_num) * sizeof(struct rps_dev_flow)))
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.
659 struct rps_sock_flow_table {
660 u32 mask;
662 u32 ents[0] ____cacheline_aligned_in_smp;
664 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
666 #define RPS_NO_CPU 0xffff
668 extern u32 rps_cpu_mask;
669 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
671 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
672 u32 hash)
674 if (table && hash) {
675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
678 /* We only give a hint, preemption can change CPU under us */
679 val |= raw_smp_processor_id();
681 if (table->ents[index] != val)
682 table->ents[index] = val;
686 #ifdef CONFIG_RFS_ACCEL
687 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
688 u16 filter_id);
689 #endif
690 #endif /* CONFIG_RPS */
692 /* This structure contains an instance of an RX queue. */
693 struct 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 } ____cacheline_aligned_in_smp;
703 * RX queue sysfs structures and functions.
705 struct rx_queue_attribute {
706 struct attribute attr;
707 ssize_t (*show)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, char *buf);
709 ssize_t (*store)(struct netdev_rx_queue *queue,
710 struct rx_queue_attribute *attr, const char *buf, size_t len);
713 #ifdef CONFIG_XPS
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
718 struct xps_map {
719 unsigned int len;
720 unsigned int alloc_len;
721 struct rcu_head rcu;
722 u16 queues[0];
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))
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
731 struct xps_dev_maps {
732 struct rcu_head rcu;
733 struct xps_map __rcu *cpu_map[0];
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 */
739 #define TC_MAX_QUEUE 16
740 #define TC_BITMASK 15
741 /* HW offloaded queuing disciplines txq count and offset maps */
742 struct netdev_tc_txq {
743 u16 count;
744 u16 offset;
747 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
752 struct 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];
762 #endif
764 #define MAX_PHYS_ITEM_ID_LEN 32
766 /* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
769 struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
771 unsigned char id_len;
774 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
781 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
784 /* These structures hold the attributes of qdisc and classifiers
785 * that are being passed to the netdevice through the setup_tc op.
787 enum {
788 TC_SETUP_MQPRIO,
789 TC_SETUP_CLSU32,
790 TC_SETUP_CLSFLOWER,
791 TC_SETUP_MATCHALL,
792 TC_SETUP_CLSBPF,
795 struct tc_cls_u32_offload;
797 struct tc_to_netdev {
798 unsigned int type;
799 union {
800 u8 tc;
801 struct tc_cls_u32_offload *cls_u32;
802 struct tc_cls_flower_offload *cls_flower;
803 struct tc_cls_matchall_offload *cls_mall;
804 struct tc_cls_bpf_offload *cls_bpf;
806 bool egress_dev;
809 /* These structures hold the attributes of xdp state that are being passed
810 * to the netdevice through the xdp op.
812 enum xdp_netdev_command {
813 /* Set or clear a bpf program used in the earliest stages of packet
814 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
815 * is responsible for calling bpf_prog_put on any old progs that are
816 * stored. In case of error, the callee need not release the new prog
817 * reference, but on success it takes ownership and must bpf_prog_put
818 * when it is no longer used.
820 XDP_SETUP_PROG,
821 /* Check if a bpf program is set on the device. The callee should
822 * return true if a program is currently attached and running.
824 XDP_QUERY_PROG,
827 struct netdev_xdp {
828 enum xdp_netdev_command command;
829 union {
830 /* XDP_SETUP_PROG */
831 struct bpf_prog *prog;
832 /* XDP_QUERY_PROG */
833 bool prog_attached;
838 * This structure defines the management hooks for network devices.
839 * The following hooks can be defined; unless noted otherwise, they are
840 * optional and can be filled with a null pointer.
842 * int (*ndo_init)(struct net_device *dev);
843 * This function is called once when a network device is registered.
844 * The network device can use this for any late stage initialization
845 * or semantic validation. It can fail with an error code which will
846 * be propagated back to register_netdev.
848 * void (*ndo_uninit)(struct net_device *dev);
849 * This function is called when device is unregistered or when registration
850 * fails. It is not called if init fails.
852 * int (*ndo_open)(struct net_device *dev);
853 * This function is called when a network device transitions to the up
854 * state.
856 * int (*ndo_stop)(struct net_device *dev);
857 * This function is called when a network device transitions to the down
858 * state.
860 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
861 * struct net_device *dev);
862 * Called when a packet needs to be transmitted.
863 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
864 * the queue before that can happen; it's for obsolete devices and weird
865 * corner cases, but the stack really does a non-trivial amount
866 * of useless work if you return NETDEV_TX_BUSY.
867 * Required; cannot be NULL.
869 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
870 * netdev_features_t features);
871 * Adjusts the requested feature flags according to device-specific
872 * constraints, and returns the resulting flags. Must not modify
873 * the device state.
875 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
876 * void *accel_priv, select_queue_fallback_t fallback);
877 * Called to decide which queue to use when device supports multiple
878 * transmit queues.
880 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
881 * This function is called to allow device receiver to make
882 * changes to configuration when multicast or promiscuous is enabled.
884 * void (*ndo_set_rx_mode)(struct net_device *dev);
885 * This function is called device changes address list filtering.
886 * If driver handles unicast address filtering, it should set
887 * IFF_UNICAST_FLT in its priv_flags.
889 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
890 * This function is called when the Media Access Control address
891 * needs to be changed. If this interface is not defined, the
892 * MAC address can not be changed.
894 * int (*ndo_validate_addr)(struct net_device *dev);
895 * Test if Media Access Control address is valid for the device.
897 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
898 * Called when a user requests an ioctl which can't be handled by
899 * the generic interface code. If not defined ioctls return
900 * not supported error code.
902 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
903 * Used to set network devices bus interface parameters. This interface
904 * is retained for legacy reasons; new devices should use the bus
905 * interface (PCI) for low level management.
907 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
908 * Called when a user wants to change the Maximum Transfer Unit
909 * of a device. If not defined, any request to change MTU will
910 * will return an error.
912 * void (*ndo_tx_timeout)(struct net_device *dev);
913 * Callback used when the transmitter has not made any progress
914 * for dev->watchdog ticks.
916 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
917 * struct rtnl_link_stats64 *storage);
918 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
919 * Called when a user wants to get the network device usage
920 * statistics. Drivers must do one of the following:
921 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
922 * rtnl_link_stats64 structure passed by the caller.
923 * 2. Define @ndo_get_stats to update a net_device_stats structure
924 * (which should normally be dev->stats) and return a pointer to
925 * it. The structure may be changed asynchronously only if each
926 * field is written atomically.
927 * 3. Update dev->stats asynchronously and atomically, and define
928 * neither operation.
930 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
931 * Return true if this device supports offload stats of this attr_id.
933 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
934 * void *attr_data)
935 * Get statistics for offload operations by attr_id. Write it into the
936 * attr_data pointer.
938 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
939 * If device supports VLAN filtering this function is called when a
940 * VLAN id is registered.
942 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
943 * If device supports VLAN filtering this function is called when a
944 * VLAN id is unregistered.
946 * void (*ndo_poll_controller)(struct net_device *dev);
948 * SR-IOV management functions.
949 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
950 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
951 * u8 qos, __be16 proto);
952 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
953 * int max_tx_rate);
954 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
955 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
956 * int (*ndo_get_vf_config)(struct net_device *dev,
957 * int vf, struct ifla_vf_info *ivf);
958 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
959 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
960 * struct nlattr *port[]);
962 * Enable or disable the VF ability to query its RSS Redirection Table and
963 * Hash Key. This is needed since on some devices VF share this information
964 * with PF and querying it may introduce a theoretical security risk.
965 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
966 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
967 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
968 * Called to setup 'tc' number of traffic classes in the net device. This
969 * is always called from the stack with the rtnl lock held and netif tx
970 * queues stopped. This allows the netdevice to perform queue management
971 * safely.
973 * Fiber Channel over Ethernet (FCoE) offload functions.
974 * int (*ndo_fcoe_enable)(struct net_device *dev);
975 * Called when the FCoE protocol stack wants to start using LLD for FCoE
976 * so the underlying device can perform whatever needed configuration or
977 * initialization to support acceleration of FCoE traffic.
979 * int (*ndo_fcoe_disable)(struct net_device *dev);
980 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
981 * so the underlying device can perform whatever needed clean-ups to
982 * stop supporting acceleration of FCoE traffic.
984 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
985 * struct scatterlist *sgl, unsigned int sgc);
986 * Called when the FCoE Initiator wants to initialize an I/O that
987 * is a possible candidate for Direct Data Placement (DDP). The LLD can
988 * perform necessary setup and returns 1 to indicate the device is set up
989 * successfully to perform DDP on this I/O, otherwise this returns 0.
991 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
992 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
993 * indicated by the FC exchange id 'xid', so the underlying device can
994 * clean up and reuse resources for later DDP requests.
996 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
997 * struct scatterlist *sgl, unsigned int sgc);
998 * Called when the FCoE Target wants to initialize an I/O that
999 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1000 * perform necessary setup and returns 1 to indicate the device is set up
1001 * successfully to perform DDP on this I/O, otherwise this returns 0.
1003 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1004 * struct netdev_fcoe_hbainfo *hbainfo);
1005 * Called when the FCoE Protocol stack wants information on the underlying
1006 * device. This information is utilized by the FCoE protocol stack to
1007 * register attributes with Fiber Channel management service as per the
1008 * FC-GS Fabric Device Management Information(FDMI) specification.
1010 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1011 * Called when the underlying device wants to override default World Wide
1012 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1013 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1014 * protocol stack to use.
1016 * RFS acceleration.
1017 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1018 * u16 rxq_index, u32 flow_id);
1019 * Set hardware filter for RFS. rxq_index is the target queue index;
1020 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1021 * Return the filter ID on success, or a negative error code.
1023 * Slave management functions (for bridge, bonding, etc).
1024 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1025 * Called to make another netdev an underling.
1027 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1028 * Called to release previously enslaved netdev.
1030 * Feature/offload setting functions.
1031 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1032 * Called to update device configuration to new features. Passed
1033 * feature set might be less than what was returned by ndo_fix_features()).
1034 * Must return >0 or -errno if it changed dev->features itself.
1036 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1037 * struct net_device *dev,
1038 * const unsigned char *addr, u16 vid, u16 flags)
1039 * Adds an FDB entry to dev for addr.
1040 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1041 * struct net_device *dev,
1042 * const unsigned char *addr, u16 vid)
1043 * Deletes the FDB entry from dev coresponding to addr.
1044 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1045 * struct net_device *dev, struct net_device *filter_dev,
1046 * int *idx)
1047 * Used to add FDB entries to dump requests. Implementers should add
1048 * entries to skb and update idx with the number of entries.
1050 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1051 * u16 flags)
1052 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1053 * struct net_device *dev, u32 filter_mask,
1054 * int nlflags)
1055 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1056 * u16 flags);
1058 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1059 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1060 * which do not represent real hardware may define this to allow their
1061 * userspace components to manage their virtual carrier state. Devices
1062 * that determine carrier state from physical hardware properties (eg
1063 * network cables) or protocol-dependent mechanisms (eg
1064 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1066 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1067 * struct netdev_phys_item_id *ppid);
1068 * Called to get ID of physical port of this device. If driver does
1069 * not implement this, it is assumed that the hw is not able to have
1070 * multiple net devices on single physical port.
1072 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1073 * struct udp_tunnel_info *ti);
1074 * Called by UDP tunnel to notify a driver about the UDP port and socket
1075 * address family that a UDP tunnel is listnening to. It is called only
1076 * when a new port starts listening. The operation is protected by the
1077 * RTNL.
1079 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1080 * struct udp_tunnel_info *ti);
1081 * Called by UDP tunnel to notify the driver about a UDP port and socket
1082 * address family that the UDP tunnel is not listening to anymore. The
1083 * operation is protected by the RTNL.
1085 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1086 * struct net_device *dev)
1087 * Called by upper layer devices to accelerate switching or other
1088 * station functionality into hardware. 'pdev is the lowerdev
1089 * to use for the offload and 'dev' is the net device that will
1090 * back the offload. Returns a pointer to the private structure
1091 * the upper layer will maintain.
1092 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1093 * Called by upper layer device to delete the station created
1094 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1095 * the station and priv is the structure returned by the add
1096 * operation.
1097 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1098 * struct net_device *dev,
1099 * void *priv);
1100 * Callback to use for xmit over the accelerated station. This
1101 * is used in place of ndo_start_xmit on accelerated net
1102 * devices.
1103 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1104 * struct net_device *dev
1105 * netdev_features_t features);
1106 * Called by core transmit path to determine if device is capable of
1107 * performing offload operations on a given packet. This is to give
1108 * the device an opportunity to implement any restrictions that cannot
1109 * be otherwise expressed by feature flags. The check is called with
1110 * the set of features that the stack has calculated and it returns
1111 * those the driver believes to be appropriate.
1112 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1113 * int queue_index, u32 maxrate);
1114 * Called when a user wants to set a max-rate limitation of specific
1115 * TX queue.
1116 * int (*ndo_get_iflink)(const struct net_device *dev);
1117 * Called to get the iflink value of this device.
1118 * void (*ndo_change_proto_down)(struct net_device *dev,
1119 * bool proto_down);
1120 * This function is used to pass protocol port error state information
1121 * to the switch driver. The switch driver can react to the proto_down
1122 * by doing a phys down on the associated switch port.
1123 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1124 * This function is used to get egress tunnel information for given skb.
1125 * This is useful for retrieving outer tunnel header parameters while
1126 * sampling packet.
1127 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1128 * This function is used to specify the headroom that the skb must
1129 * consider when allocation skb during packet reception. Setting
1130 * appropriate rx headroom value allows avoiding skb head copy on
1131 * forward. Setting a negative value resets the rx headroom to the
1132 * default value.
1133 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1134 * This function is used to set or query state related to XDP on the
1135 * netdevice. See definition of enum xdp_netdev_command for details.
1138 struct net_device_ops {
1139 int (*ndo_init)(struct net_device *dev);
1140 void (*ndo_uninit)(struct net_device *dev);
1141 int (*ndo_open)(struct net_device *dev);
1142 int (*ndo_stop)(struct net_device *dev);
1143 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1144 struct net_device *dev);
1145 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1146 struct net_device *dev,
1147 netdev_features_t features);
1148 u16 (*ndo_select_queue)(struct net_device *dev,
1149 struct sk_buff *skb,
1150 void *accel_priv,
1151 select_queue_fallback_t fallback);
1152 void (*ndo_change_rx_flags)(struct net_device *dev,
1153 int flags);
1154 void (*ndo_set_rx_mode)(struct net_device *dev);
1155 int (*ndo_set_mac_address)(struct net_device *dev,
1156 void *addr);
1157 int (*ndo_validate_addr)(struct net_device *dev);
1158 int (*ndo_do_ioctl)(struct net_device *dev,
1159 struct ifreq *ifr, int cmd);
1160 int (*ndo_set_config)(struct net_device *dev,
1161 struct ifmap *map);
1162 int (*ndo_change_mtu)(struct net_device *dev,
1163 int new_mtu);
1164 int (*ndo_neigh_setup)(struct net_device *dev,
1165 struct neigh_parms *);
1166 void (*ndo_tx_timeout) (struct net_device *dev);
1168 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1169 struct rtnl_link_stats64 *storage);
1170 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1171 int (*ndo_get_offload_stats)(int attr_id,
1172 const struct net_device *dev,
1173 void *attr_data);
1174 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1176 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1177 __be16 proto, u16 vid);
1178 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1179 __be16 proto, u16 vid);
1180 #ifdef CONFIG_NET_POLL_CONTROLLER
1181 void (*ndo_poll_controller)(struct net_device *dev);
1182 int (*ndo_netpoll_setup)(struct net_device *dev,
1183 struct netpoll_info *info);
1184 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1185 #endif
1186 #ifdef CONFIG_NET_RX_BUSY_POLL
1187 int (*ndo_busy_poll)(struct napi_struct *dev);
1188 #endif
1189 int (*ndo_set_vf_mac)(struct net_device *dev,
1190 int queue, u8 *mac);
1191 int (*ndo_set_vf_vlan)(struct net_device *dev,
1192 int queue, u16 vlan,
1193 u8 qos, __be16 proto);
1194 int (*ndo_set_vf_rate)(struct net_device *dev,
1195 int vf, int min_tx_rate,
1196 int max_tx_rate);
1197 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1198 int vf, bool setting);
1199 int (*ndo_set_vf_trust)(struct net_device *dev,
1200 int vf, bool setting);
1201 int (*ndo_get_vf_config)(struct net_device *dev,
1202 int vf,
1203 struct ifla_vf_info *ivf);
1204 int (*ndo_set_vf_link_state)(struct net_device *dev,
1205 int vf, int link_state);
1206 int (*ndo_get_vf_stats)(struct net_device *dev,
1207 int vf,
1208 struct ifla_vf_stats
1209 *vf_stats);
1210 int (*ndo_set_vf_port)(struct net_device *dev,
1211 int vf,
1212 struct nlattr *port[]);
1213 int (*ndo_get_vf_port)(struct net_device *dev,
1214 int vf, struct sk_buff *skb);
1215 int (*ndo_set_vf_guid)(struct net_device *dev,
1216 int vf, u64 guid,
1217 int guid_type);
1218 int (*ndo_set_vf_rss_query_en)(
1219 struct net_device *dev,
1220 int vf, bool setting);
1221 int (*ndo_setup_tc)(struct net_device *dev,
1222 u32 handle,
1223 __be16 protocol,
1224 struct tc_to_netdev *tc);
1225 #if IS_ENABLED(CONFIG_FCOE)
1226 int (*ndo_fcoe_enable)(struct net_device *dev);
1227 int (*ndo_fcoe_disable)(struct net_device *dev);
1228 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1229 u16 xid,
1230 struct scatterlist *sgl,
1231 unsigned int sgc);
1232 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1233 u16 xid);
1234 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1235 u16 xid,
1236 struct scatterlist *sgl,
1237 unsigned int sgc);
1238 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1239 struct netdev_fcoe_hbainfo *hbainfo);
1240 #endif
1242 #if IS_ENABLED(CONFIG_LIBFCOE)
1243 #define NETDEV_FCOE_WWNN 0
1244 #define NETDEV_FCOE_WWPN 1
1245 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1246 u64 *wwn, int type);
1247 #endif
1249 #ifdef CONFIG_RFS_ACCEL
1250 int (*ndo_rx_flow_steer)(struct net_device *dev,
1251 const struct sk_buff *skb,
1252 u16 rxq_index,
1253 u32 flow_id);
1254 #endif
1255 int (*ndo_add_slave)(struct net_device *dev,
1256 struct net_device *slave_dev);
1257 int (*ndo_del_slave)(struct net_device *dev,
1258 struct net_device *slave_dev);
1259 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1260 netdev_features_t features);
1261 int (*ndo_set_features)(struct net_device *dev,
1262 netdev_features_t features);
1263 int (*ndo_neigh_construct)(struct net_device *dev,
1264 struct neighbour *n);
1265 void (*ndo_neigh_destroy)(struct net_device *dev,
1266 struct neighbour *n);
1268 int (*ndo_fdb_add)(struct ndmsg *ndm,
1269 struct nlattr *tb[],
1270 struct net_device *dev,
1271 const unsigned char *addr,
1272 u16 vid,
1273 u16 flags);
1274 int (*ndo_fdb_del)(struct ndmsg *ndm,
1275 struct nlattr *tb[],
1276 struct net_device *dev,
1277 const unsigned char *addr,
1278 u16 vid);
1279 int (*ndo_fdb_dump)(struct sk_buff *skb,
1280 struct netlink_callback *cb,
1281 struct net_device *dev,
1282 struct net_device *filter_dev,
1283 int *idx);
1285 int (*ndo_bridge_setlink)(struct net_device *dev,
1286 struct nlmsghdr *nlh,
1287 u16 flags);
1288 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1289 u32 pid, u32 seq,
1290 struct net_device *dev,
1291 u32 filter_mask,
1292 int nlflags);
1293 int (*ndo_bridge_dellink)(struct net_device *dev,
1294 struct nlmsghdr *nlh,
1295 u16 flags);
1296 int (*ndo_change_carrier)(struct net_device *dev,
1297 bool new_carrier);
1298 int (*ndo_get_phys_port_id)(struct net_device *dev,
1299 struct netdev_phys_item_id *ppid);
1300 int (*ndo_get_phys_port_name)(struct net_device *dev,
1301 char *name, size_t len);
1302 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1303 struct udp_tunnel_info *ti);
1304 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1305 struct udp_tunnel_info *ti);
1306 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1307 struct net_device *dev);
1308 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1309 void *priv);
1311 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1312 struct net_device *dev,
1313 void *priv);
1314 int (*ndo_get_lock_subclass)(struct net_device *dev);
1315 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1316 int queue_index,
1317 u32 maxrate);
1318 int (*ndo_get_iflink)(const struct net_device *dev);
1319 int (*ndo_change_proto_down)(struct net_device *dev,
1320 bool proto_down);
1321 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1322 struct sk_buff *skb);
1323 void (*ndo_set_rx_headroom)(struct net_device *dev,
1324 int needed_headroom);
1325 int (*ndo_xdp)(struct net_device *dev,
1326 struct netdev_xdp *xdp);
1330 * enum net_device_priv_flags - &struct net_device priv_flags
1332 * These are the &struct net_device, they are only set internally
1333 * by drivers and used in the kernel. These flags are invisible to
1334 * userspace; this means that the order of these flags can change
1335 * during any kernel release.
1337 * You should have a pretty good reason to be extending these flags.
1339 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1340 * @IFF_EBRIDGE: Ethernet bridging device
1341 * @IFF_BONDING: bonding master or slave
1342 * @IFF_ISATAP: ISATAP interface (RFC4214)
1343 * @IFF_WAN_HDLC: WAN HDLC device
1344 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1345 * release skb->dst
1346 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1347 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1348 * @IFF_MACVLAN_PORT: device used as macvlan port
1349 * @IFF_BRIDGE_PORT: device used as bridge port
1350 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1351 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1352 * @IFF_UNICAST_FLT: Supports unicast filtering
1353 * @IFF_TEAM_PORT: device used as team port
1354 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1355 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1356 * change when it's running
1357 * @IFF_MACVLAN: Macvlan device
1358 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1359 * underlying stacked devices
1360 * @IFF_IPVLAN_MASTER: IPvlan master device
1361 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1362 * @IFF_L3MDEV_MASTER: device is an L3 master device
1363 * @IFF_NO_QUEUE: device can run without qdisc attached
1364 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1365 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1366 * @IFF_TEAM: device is a team device
1367 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1368 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1369 * entity (i.e. the master device for bridged veth)
1370 * @IFF_MACSEC: device is a MACsec device
1372 enum netdev_priv_flags {
1373 IFF_802_1Q_VLAN = 1<<0,
1374 IFF_EBRIDGE = 1<<1,
1375 IFF_BONDING = 1<<2,
1376 IFF_ISATAP = 1<<3,
1377 IFF_WAN_HDLC = 1<<4,
1378 IFF_XMIT_DST_RELEASE = 1<<5,
1379 IFF_DONT_BRIDGE = 1<<6,
1380 IFF_DISABLE_NETPOLL = 1<<7,
1381 IFF_MACVLAN_PORT = 1<<8,
1382 IFF_BRIDGE_PORT = 1<<9,
1383 IFF_OVS_DATAPATH = 1<<10,
1384 IFF_TX_SKB_SHARING = 1<<11,
1385 IFF_UNICAST_FLT = 1<<12,
1386 IFF_TEAM_PORT = 1<<13,
1387 IFF_SUPP_NOFCS = 1<<14,
1388 IFF_LIVE_ADDR_CHANGE = 1<<15,
1389 IFF_MACVLAN = 1<<16,
1390 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1391 IFF_IPVLAN_MASTER = 1<<18,
1392 IFF_IPVLAN_SLAVE = 1<<19,
1393 IFF_L3MDEV_MASTER = 1<<20,
1394 IFF_NO_QUEUE = 1<<21,
1395 IFF_OPENVSWITCH = 1<<22,
1396 IFF_L3MDEV_SLAVE = 1<<23,
1397 IFF_TEAM = 1<<24,
1398 IFF_RXFH_CONFIGURED = 1<<25,
1399 IFF_PHONY_HEADROOM = 1<<26,
1400 IFF_MACSEC = 1<<27,
1403 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1404 #define IFF_EBRIDGE IFF_EBRIDGE
1405 #define IFF_BONDING IFF_BONDING
1406 #define IFF_ISATAP IFF_ISATAP
1407 #define IFF_WAN_HDLC IFF_WAN_HDLC
1408 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1409 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1410 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1411 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1412 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1413 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1414 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1415 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1416 #define IFF_TEAM_PORT IFF_TEAM_PORT
1417 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1418 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1419 #define IFF_MACVLAN IFF_MACVLAN
1420 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1421 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1422 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1423 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1424 #define IFF_NO_QUEUE IFF_NO_QUEUE
1425 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1426 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1427 #define IFF_TEAM IFF_TEAM
1428 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1429 #define IFF_MACSEC IFF_MACSEC
1432 * struct net_device - The DEVICE structure.
1433 * Actually, this whole structure is a big mistake. It mixes I/O
1434 * data with strictly "high-level" data, and it has to know about
1435 * almost every data structure used in the INET module.
1437 * @name: This is the first field of the "visible" part of this structure
1438 * (i.e. as seen by users in the "Space.c" file). It is the name
1439 * of the interface.
1441 * @name_hlist: Device name hash chain, please keep it close to name[]
1442 * @ifalias: SNMP alias
1443 * @mem_end: Shared memory end
1444 * @mem_start: Shared memory start
1445 * @base_addr: Device I/O address
1446 * @irq: Device IRQ number
1448 * @carrier_changes: Stats to monitor carrier on<->off transitions
1450 * @state: Generic network queuing layer state, see netdev_state_t
1451 * @dev_list: The global list of network devices
1452 * @napi_list: List entry used for polling NAPI devices
1453 * @unreg_list: List entry when we are unregistering the
1454 * device; see the function unregister_netdev
1455 * @close_list: List entry used when we are closing the device
1456 * @ptype_all: Device-specific packet handlers for all protocols
1457 * @ptype_specific: Device-specific, protocol-specific packet handlers
1459 * @adj_list: Directly linked devices, like slaves for bonding
1460 * @features: Currently active device features
1461 * @hw_features: User-changeable features
1463 * @wanted_features: User-requested features
1464 * @vlan_features: Mask of features inheritable by VLAN devices
1466 * @hw_enc_features: Mask of features inherited by encapsulating devices
1467 * This field indicates what encapsulation
1468 * offloads the hardware is capable of doing,
1469 * and drivers will need to set them appropriately.
1471 * @mpls_features: Mask of features inheritable by MPLS
1473 * @ifindex: interface index
1474 * @group: The group the device belongs to
1476 * @stats: Statistics struct, which was left as a legacy, use
1477 * rtnl_link_stats64 instead
1479 * @rx_dropped: Dropped packets by core network,
1480 * do not use this in drivers
1481 * @tx_dropped: Dropped packets by core network,
1482 * do not use this in drivers
1483 * @rx_nohandler: nohandler dropped packets by core network on
1484 * inactive devices, do not use this in drivers
1486 * @wireless_handlers: List of functions to handle Wireless Extensions,
1487 * instead of ioctl,
1488 * see <net/iw_handler.h> for details.
1489 * @wireless_data: Instance data managed by the core of wireless extensions
1491 * @netdev_ops: Includes several pointers to callbacks,
1492 * if one wants to override the ndo_*() functions
1493 * @ethtool_ops: Management operations
1494 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1495 * discovery handling. Necessary for e.g. 6LoWPAN.
1496 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1497 * of Layer 2 headers.
1499 * @flags: Interface flags (a la BSD)
1500 * @priv_flags: Like 'flags' but invisible to userspace,
1501 * see if.h for the definitions
1502 * @gflags: Global flags ( kept as legacy )
1503 * @padded: How much padding added by alloc_netdev()
1504 * @operstate: RFC2863 operstate
1505 * @link_mode: Mapping policy to operstate
1506 * @if_port: Selectable AUI, TP, ...
1507 * @dma: DMA channel
1508 * @mtu: Interface MTU value
1509 * @min_mtu: Interface Minimum MTU value
1510 * @max_mtu: Interface Maximum MTU value
1511 * @type: Interface hardware type
1512 * @hard_header_len: Maximum hardware header length.
1514 * @needed_headroom: Extra headroom the hardware may need, but not in all
1515 * cases can this be guaranteed
1516 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1517 * cases can this be guaranteed. Some cases also use
1518 * LL_MAX_HEADER instead to allocate the skb
1520 * interface address info:
1522 * @perm_addr: Permanent hw address
1523 * @addr_assign_type: Hw address assignment type
1524 * @addr_len: Hardware address length
1525 * @neigh_priv_len: Used in neigh_alloc()
1526 * @dev_id: Used to differentiate devices that share
1527 * the same link layer address
1528 * @dev_port: Used to differentiate devices that share
1529 * the same function
1530 * @addr_list_lock: XXX: need comments on this one
1531 * @uc_promisc: Counter that indicates promiscuous mode
1532 * has been enabled due to the need to listen to
1533 * additional unicast addresses in a device that
1534 * does not implement ndo_set_rx_mode()
1535 * @uc: unicast mac addresses
1536 * @mc: multicast mac addresses
1537 * @dev_addrs: list of device hw addresses
1538 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1539 * @promiscuity: Number of times the NIC is told to work in
1540 * promiscuous mode; if it becomes 0 the NIC will
1541 * exit promiscuous mode
1542 * @allmulti: Counter, enables or disables allmulticast mode
1544 * @vlan_info: VLAN info
1545 * @dsa_ptr: dsa specific data
1546 * @tipc_ptr: TIPC specific data
1547 * @atalk_ptr: AppleTalk link
1548 * @ip_ptr: IPv4 specific data
1549 * @dn_ptr: DECnet specific data
1550 * @ip6_ptr: IPv6 specific data
1551 * @ax25_ptr: AX.25 specific data
1552 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1554 * @last_rx: Time of last Rx
1555 * @dev_addr: Hw address (before bcast,
1556 * because most packets are unicast)
1558 * @_rx: Array of RX queues
1559 * @num_rx_queues: Number of RX queues
1560 * allocated at register_netdev() time
1561 * @real_num_rx_queues: Number of RX queues currently active in device
1563 * @rx_handler: handler for received packets
1564 * @rx_handler_data: XXX: need comments on this one
1565 * @ingress_queue: XXX: need comments on this one
1566 * @broadcast: hw bcast address
1568 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1569 * indexed by RX queue number. Assigned by driver.
1570 * This must only be set if the ndo_rx_flow_steer
1571 * operation is defined
1572 * @index_hlist: Device index hash chain
1574 * @_tx: Array of TX queues
1575 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1576 * @real_num_tx_queues: Number of TX queues currently active in device
1577 * @qdisc: Root qdisc from userspace point of view
1578 * @tx_queue_len: Max frames per queue allowed
1579 * @tx_global_lock: XXX: need comments on this one
1581 * @xps_maps: XXX: need comments on this one
1583 * @watchdog_timeo: Represents the timeout that is used by
1584 * the watchdog (see dev_watchdog())
1585 * @watchdog_timer: List of timers
1587 * @pcpu_refcnt: Number of references to this device
1588 * @todo_list: Delayed register/unregister
1589 * @link_watch_list: XXX: need comments on this one
1591 * @reg_state: Register/unregister state machine
1592 * @dismantle: Device is going to be freed
1593 * @rtnl_link_state: This enum represents the phases of creating
1594 * a new link
1596 * @destructor: Called from unregister,
1597 * can be used to call free_netdev
1598 * @npinfo: XXX: need comments on this one
1599 * @nd_net: Network namespace this network device is inside
1601 * @ml_priv: Mid-layer private
1602 * @lstats: Loopback statistics
1603 * @tstats: Tunnel statistics
1604 * @dstats: Dummy statistics
1605 * @vstats: Virtual ethernet statistics
1607 * @garp_port: GARP
1608 * @mrp_port: MRP
1610 * @dev: Class/net/name entry
1611 * @sysfs_groups: Space for optional device, statistics and wireless
1612 * sysfs groups
1614 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1615 * @rtnl_link_ops: Rtnl_link_ops
1617 * @gso_max_size: Maximum size of generic segmentation offload
1618 * @gso_max_segs: Maximum number of segments that can be passed to the
1619 * NIC for GSO
1621 * @dcbnl_ops: Data Center Bridging netlink ops
1622 * @num_tc: Number of traffic classes in the net device
1623 * @tc_to_txq: XXX: need comments on this one
1624 * @prio_tc_map: XXX: need comments on this one
1626 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1628 * @priomap: XXX: need comments on this one
1629 * @phydev: Physical device may attach itself
1630 * for hardware timestamping
1632 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1633 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1635 * @proto_down: protocol port state information can be sent to the
1636 * switch driver and used to set the phys state of the
1637 * switch port.
1639 * FIXME: cleanup struct net_device such that network protocol info
1640 * moves out.
1643 struct net_device {
1644 char name[IFNAMSIZ];
1645 struct hlist_node name_hlist;
1646 char *ifalias;
1648 * I/O specific fields
1649 * FIXME: Merge these and struct ifmap into one
1651 unsigned long mem_end;
1652 unsigned long mem_start;
1653 unsigned long base_addr;
1654 int irq;
1656 atomic_t carrier_changes;
1659 * Some hardware also needs these fields (state,dev_list,
1660 * napi_list,unreg_list,close_list) but they are not
1661 * part of the usual set specified in Space.c.
1664 unsigned long state;
1666 struct list_head dev_list;
1667 struct list_head napi_list;
1668 struct list_head unreg_list;
1669 struct list_head close_list;
1670 struct list_head ptype_all;
1671 struct list_head ptype_specific;
1673 struct {
1674 struct list_head upper;
1675 struct list_head lower;
1676 } adj_list;
1678 netdev_features_t features;
1679 netdev_features_t hw_features;
1680 netdev_features_t wanted_features;
1681 netdev_features_t vlan_features;
1682 netdev_features_t hw_enc_features;
1683 netdev_features_t mpls_features;
1684 netdev_features_t gso_partial_features;
1686 int ifindex;
1687 int group;
1689 struct net_device_stats stats;
1691 atomic_long_t rx_dropped;
1692 atomic_long_t tx_dropped;
1693 atomic_long_t rx_nohandler;
1695 #ifdef CONFIG_WIRELESS_EXT
1696 const struct iw_handler_def *wireless_handlers;
1697 struct iw_public_data *wireless_data;
1698 #endif
1699 const struct net_device_ops *netdev_ops;
1700 const struct ethtool_ops *ethtool_ops;
1701 #ifdef CONFIG_NET_SWITCHDEV
1702 const struct switchdev_ops *switchdev_ops;
1703 #endif
1704 #ifdef CONFIG_NET_L3_MASTER_DEV
1705 const struct l3mdev_ops *l3mdev_ops;
1706 #endif
1707 #if IS_ENABLED(CONFIG_IPV6)
1708 const struct ndisc_ops *ndisc_ops;
1709 #endif
1711 const struct header_ops *header_ops;
1713 unsigned int flags;
1714 unsigned int priv_flags;
1716 unsigned short gflags;
1717 unsigned short padded;
1719 unsigned char operstate;
1720 unsigned char link_mode;
1722 unsigned char if_port;
1723 unsigned char dma;
1725 unsigned int mtu;
1726 unsigned int min_mtu;
1727 unsigned int max_mtu;
1728 unsigned short type;
1729 unsigned short hard_header_len;
1731 unsigned short needed_headroom;
1732 unsigned short needed_tailroom;
1734 /* Interface address info. */
1735 unsigned char perm_addr[MAX_ADDR_LEN];
1736 unsigned char addr_assign_type;
1737 unsigned char addr_len;
1738 unsigned short neigh_priv_len;
1739 unsigned short dev_id;
1740 unsigned short dev_port;
1741 spinlock_t addr_list_lock;
1742 unsigned char name_assign_type;
1743 bool uc_promisc;
1744 struct netdev_hw_addr_list uc;
1745 struct netdev_hw_addr_list mc;
1746 struct netdev_hw_addr_list dev_addrs;
1748 #ifdef CONFIG_SYSFS
1749 struct kset *queues_kset;
1750 #endif
1751 unsigned int promiscuity;
1752 unsigned int allmulti;
1755 /* Protocol-specific pointers */
1757 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1758 struct vlan_info __rcu *vlan_info;
1759 #endif
1760 #if IS_ENABLED(CONFIG_NET_DSA)
1761 struct dsa_switch_tree *dsa_ptr;
1762 #endif
1763 #if IS_ENABLED(CONFIG_TIPC)
1764 struct tipc_bearer __rcu *tipc_ptr;
1765 #endif
1766 void *atalk_ptr;
1767 struct in_device __rcu *ip_ptr;
1768 struct dn_dev __rcu *dn_ptr;
1769 struct inet6_dev __rcu *ip6_ptr;
1770 void *ax25_ptr;
1771 struct wireless_dev *ieee80211_ptr;
1772 struct wpan_dev *ieee802154_ptr;
1773 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1774 struct mpls_dev __rcu *mpls_ptr;
1775 #endif
1778 * Cache lines mostly used on receive path (including eth_type_trans())
1780 unsigned long last_rx;
1782 /* Interface address info used in eth_type_trans() */
1783 unsigned char *dev_addr;
1785 #ifdef CONFIG_SYSFS
1786 struct netdev_rx_queue *_rx;
1788 unsigned int num_rx_queues;
1789 unsigned int real_num_rx_queues;
1790 #endif
1792 unsigned long gro_flush_timeout;
1793 rx_handler_func_t __rcu *rx_handler;
1794 void __rcu *rx_handler_data;
1796 #ifdef CONFIG_NET_CLS_ACT
1797 struct tcf_proto __rcu *ingress_cl_list;
1798 #endif
1799 struct netdev_queue __rcu *ingress_queue;
1800 #ifdef CONFIG_NETFILTER_INGRESS
1801 struct nf_hook_entry __rcu *nf_hooks_ingress;
1802 #endif
1804 unsigned char broadcast[MAX_ADDR_LEN];
1805 #ifdef CONFIG_RFS_ACCEL
1806 struct cpu_rmap *rx_cpu_rmap;
1807 #endif
1808 struct hlist_node index_hlist;
1811 * Cache lines mostly used on transmit path
1813 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1814 unsigned int num_tx_queues;
1815 unsigned int real_num_tx_queues;
1816 struct Qdisc *qdisc;
1817 #ifdef CONFIG_NET_SCHED
1818 DECLARE_HASHTABLE (qdisc_hash, 4);
1819 #endif
1820 unsigned long tx_queue_len;
1821 spinlock_t tx_global_lock;
1822 int watchdog_timeo;
1824 #ifdef CONFIG_XPS
1825 struct xps_dev_maps __rcu *xps_maps;
1826 #endif
1827 #ifdef CONFIG_NET_CLS_ACT
1828 struct tcf_proto __rcu *egress_cl_list;
1829 #endif
1831 /* These may be needed for future network-power-down code. */
1832 struct timer_list watchdog_timer;
1834 int __percpu *pcpu_refcnt;
1835 struct list_head todo_list;
1837 struct list_head link_watch_list;
1839 enum { NETREG_UNINITIALIZED=0,
1840 NETREG_REGISTERED, /* completed register_netdevice */
1841 NETREG_UNREGISTERING, /* called unregister_netdevice */
1842 NETREG_UNREGISTERED, /* completed unregister todo */
1843 NETREG_RELEASED, /* called free_netdev */
1844 NETREG_DUMMY, /* dummy device for NAPI poll */
1845 } reg_state:8;
1847 bool dismantle;
1849 enum {
1850 RTNL_LINK_INITIALIZED,
1851 RTNL_LINK_INITIALIZING,
1852 } rtnl_link_state:16;
1854 void (*destructor)(struct net_device *dev);
1856 #ifdef CONFIG_NETPOLL
1857 struct netpoll_info __rcu *npinfo;
1858 #endif
1860 possible_net_t nd_net;
1862 /* mid-layer private */
1863 union {
1864 void *ml_priv;
1865 struct pcpu_lstats __percpu *lstats;
1866 struct pcpu_sw_netstats __percpu *tstats;
1867 struct pcpu_dstats __percpu *dstats;
1868 struct pcpu_vstats __percpu *vstats;
1871 struct garp_port __rcu *garp_port;
1872 struct mrp_port __rcu *mrp_port;
1874 struct device dev;
1875 const struct attribute_group *sysfs_groups[4];
1876 const struct attribute_group *sysfs_rx_queue_group;
1878 const struct rtnl_link_ops *rtnl_link_ops;
1880 /* for setting kernel sock attribute on TCP connection setup */
1881 #define GSO_MAX_SIZE 65536
1882 unsigned int gso_max_size;
1883 #define GSO_MAX_SEGS 65535
1884 u16 gso_max_segs;
1886 #ifdef CONFIG_DCB
1887 const struct dcbnl_rtnl_ops *dcbnl_ops;
1888 #endif
1889 u8 num_tc;
1890 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1891 u8 prio_tc_map[TC_BITMASK + 1];
1893 #if IS_ENABLED(CONFIG_FCOE)
1894 unsigned int fcoe_ddp_xid;
1895 #endif
1896 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1897 struct netprio_map __rcu *priomap;
1898 #endif
1899 struct phy_device *phydev;
1900 struct lock_class_key *qdisc_tx_busylock;
1901 struct lock_class_key *qdisc_running_key;
1902 bool proto_down;
1904 #define to_net_dev(d) container_of(d, struct net_device, dev)
1906 #define NETDEV_ALIGN 32
1908 static inline
1909 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1911 return dev->prio_tc_map[prio & TC_BITMASK];
1914 static inline
1915 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1917 if (tc >= dev->num_tc)
1918 return -EINVAL;
1920 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1921 return 0;
1924 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1925 void netdev_reset_tc(struct net_device *dev);
1926 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1927 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1929 static inline
1930 int netdev_get_num_tc(struct net_device *dev)
1932 return dev->num_tc;
1935 static inline
1936 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1937 unsigned int index)
1939 return &dev->_tx[index];
1942 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1943 const struct sk_buff *skb)
1945 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1948 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1949 void (*f)(struct net_device *,
1950 struct netdev_queue *,
1951 void *),
1952 void *arg)
1954 unsigned int i;
1956 for (i = 0; i < dev->num_tx_queues; i++)
1957 f(dev, &dev->_tx[i], arg);
1960 #define netdev_lockdep_set_classes(dev) \
1962 static struct lock_class_key qdisc_tx_busylock_key; \
1963 static struct lock_class_key qdisc_running_key; \
1964 static struct lock_class_key qdisc_xmit_lock_key; \
1965 static struct lock_class_key dev_addr_list_lock_key; \
1966 unsigned int i; \
1968 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1969 (dev)->qdisc_running_key = &qdisc_running_key; \
1970 lockdep_set_class(&(dev)->addr_list_lock, \
1971 &dev_addr_list_lock_key); \
1972 for (i = 0; i < (dev)->num_tx_queues; i++) \
1973 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1974 &qdisc_xmit_lock_key); \
1977 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1978 struct sk_buff *skb,
1979 void *accel_priv);
1981 /* returns the headroom that the master device needs to take in account
1982 * when forwarding to this dev
1984 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1986 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1989 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1991 if (dev->netdev_ops->ndo_set_rx_headroom)
1992 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1995 /* set the device rx headroom to the dev's default */
1996 static inline void netdev_reset_rx_headroom(struct net_device *dev)
1998 netdev_set_rx_headroom(dev, -1);
2002 * Net namespace inlines
2004 static inline
2005 struct net *dev_net(const struct net_device *dev)
2007 return read_pnet(&dev->nd_net);
2010 static inline
2011 void dev_net_set(struct net_device *dev, struct net *net)
2013 write_pnet(&dev->nd_net, net);
2016 static inline bool netdev_uses_dsa(struct net_device *dev)
2018 #if IS_ENABLED(CONFIG_NET_DSA)
2019 if (dev->dsa_ptr != NULL)
2020 return dsa_uses_tagged_protocol(dev->dsa_ptr);
2021 #endif
2022 return false;
2026 * netdev_priv - access network device private data
2027 * @dev: network device
2029 * Get network device private data
2031 static inline void *netdev_priv(const struct net_device *dev)
2033 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2036 /* Set the sysfs physical device reference for the network logical device
2037 * if set prior to registration will cause a symlink during initialization.
2039 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2041 /* Set the sysfs device type for the network logical device to allow
2042 * fine-grained identification of different network device types. For
2043 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2045 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2047 /* Default NAPI poll() weight
2048 * Device drivers are strongly advised to not use bigger value
2050 #define NAPI_POLL_WEIGHT 64
2053 * netif_napi_add - initialize a NAPI context
2054 * @dev: network device
2055 * @napi: NAPI context
2056 * @poll: polling function
2057 * @weight: default weight
2059 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2060 * *any* of the other NAPI-related functions.
2062 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2063 int (*poll)(struct napi_struct *, int), int weight);
2066 * netif_tx_napi_add - initialize a NAPI context
2067 * @dev: network device
2068 * @napi: NAPI context
2069 * @poll: polling function
2070 * @weight: default weight
2072 * This variant of netif_napi_add() should be used from drivers using NAPI
2073 * to exclusively poll a TX queue.
2074 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2076 static inline void netif_tx_napi_add(struct net_device *dev,
2077 struct napi_struct *napi,
2078 int (*poll)(struct napi_struct *, int),
2079 int weight)
2081 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2082 netif_napi_add(dev, napi, poll, weight);
2086 * netif_napi_del - remove a NAPI context
2087 * @napi: NAPI context
2089 * netif_napi_del() removes a NAPI context from the network device NAPI list
2091 void netif_napi_del(struct napi_struct *napi);
2093 struct napi_gro_cb {
2094 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2095 void *frag0;
2097 /* Length of frag0. */
2098 unsigned int frag0_len;
2100 /* This indicates where we are processing relative to skb->data. */
2101 int data_offset;
2103 /* This is non-zero if the packet cannot be merged with the new skb. */
2104 u16 flush;
2106 /* Save the IP ID here and check when we get to the transport layer */
2107 u16 flush_id;
2109 /* Number of segments aggregated. */
2110 u16 count;
2112 /* Start offset for remote checksum offload */
2113 u16 gro_remcsum_start;
2115 /* jiffies when first packet was created/queued */
2116 unsigned long age;
2118 /* Used in ipv6_gro_receive() and foo-over-udp */
2119 u16 proto;
2121 /* This is non-zero if the packet may be of the same flow. */
2122 u8 same_flow:1;
2124 /* Used in tunnel GRO receive */
2125 u8 encap_mark:1;
2127 /* GRO checksum is valid */
2128 u8 csum_valid:1;
2130 /* Number of checksums via CHECKSUM_UNNECESSARY */
2131 u8 csum_cnt:3;
2133 /* Free the skb? */
2134 u8 free:2;
2135 #define NAPI_GRO_FREE 1
2136 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2138 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2139 u8 is_ipv6:1;
2141 /* Used in GRE, set in fou/gue_gro_receive */
2142 u8 is_fou:1;
2144 /* Used to determine if flush_id can be ignored */
2145 u8 is_atomic:1;
2147 /* Number of gro_receive callbacks this packet already went through */
2148 u8 recursion_counter:4;
2150 /* 1 bit hole */
2152 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2153 __wsum csum;
2155 /* used in skb_gro_receive() slow path */
2156 struct sk_buff *last;
2159 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2161 #define GRO_RECURSION_LIMIT 15
2162 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2164 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2167 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2168 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2169 struct sk_buff **head,
2170 struct sk_buff *skb)
2172 if (unlikely(gro_recursion_inc_test(skb))) {
2173 NAPI_GRO_CB(skb)->flush |= 1;
2174 return NULL;
2177 return cb(head, skb);
2180 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2181 struct sk_buff *);
2182 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2183 struct sock *sk,
2184 struct sk_buff **head,
2185 struct sk_buff *skb)
2187 if (unlikely(gro_recursion_inc_test(skb))) {
2188 NAPI_GRO_CB(skb)->flush |= 1;
2189 return NULL;
2192 return cb(sk, head, skb);
2195 struct packet_type {
2196 __be16 type; /* This is really htons(ether_type). */
2197 struct net_device *dev; /* NULL is wildcarded here */
2198 int (*func) (struct sk_buff *,
2199 struct net_device *,
2200 struct packet_type *,
2201 struct net_device *);
2202 bool (*id_match)(struct packet_type *ptype,
2203 struct sock *sk);
2204 void *af_packet_priv;
2205 struct list_head list;
2208 struct offload_callbacks {
2209 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2210 netdev_features_t features);
2211 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2212 struct sk_buff *skb);
2213 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2216 struct packet_offload {
2217 __be16 type; /* This is really htons(ether_type). */
2218 u16 priority;
2219 struct offload_callbacks callbacks;
2220 struct list_head list;
2223 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2224 struct pcpu_sw_netstats {
2225 u64 rx_packets;
2226 u64 rx_bytes;
2227 u64 tx_packets;
2228 u64 tx_bytes;
2229 struct u64_stats_sync syncp;
2232 #define __netdev_alloc_pcpu_stats(type, gfp) \
2233 ({ \
2234 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2235 if (pcpu_stats) { \
2236 int __cpu; \
2237 for_each_possible_cpu(__cpu) { \
2238 typeof(type) *stat; \
2239 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2240 u64_stats_init(&stat->syncp); \
2243 pcpu_stats; \
2246 #define netdev_alloc_pcpu_stats(type) \
2247 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2249 enum netdev_lag_tx_type {
2250 NETDEV_LAG_TX_TYPE_UNKNOWN,
2251 NETDEV_LAG_TX_TYPE_RANDOM,
2252 NETDEV_LAG_TX_TYPE_BROADCAST,
2253 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2254 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2255 NETDEV_LAG_TX_TYPE_HASH,
2258 struct netdev_lag_upper_info {
2259 enum netdev_lag_tx_type tx_type;
2262 struct netdev_lag_lower_state_info {
2263 u8 link_up : 1,
2264 tx_enabled : 1;
2267 #include <linux/notifier.h>
2269 /* netdevice notifier chain. Please remember to update the rtnetlink
2270 * notification exclusion list in rtnetlink_event() when adding new
2271 * types.
2273 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2274 #define NETDEV_DOWN 0x0002
2275 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2276 detected a hardware crash and restarted
2277 - we can use this eg to kick tcp sessions
2278 once done */
2279 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2280 #define NETDEV_REGISTER 0x0005
2281 #define NETDEV_UNREGISTER 0x0006
2282 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2283 #define NETDEV_CHANGEADDR 0x0008
2284 #define NETDEV_GOING_DOWN 0x0009
2285 #define NETDEV_CHANGENAME 0x000A
2286 #define NETDEV_FEAT_CHANGE 0x000B
2287 #define NETDEV_BONDING_FAILOVER 0x000C
2288 #define NETDEV_PRE_UP 0x000D
2289 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2290 #define NETDEV_POST_TYPE_CHANGE 0x000F
2291 #define NETDEV_POST_INIT 0x0010
2292 #define NETDEV_UNREGISTER_FINAL 0x0011
2293 #define NETDEV_RELEASE 0x0012
2294 #define NETDEV_NOTIFY_PEERS 0x0013
2295 #define NETDEV_JOIN 0x0014
2296 #define NETDEV_CHANGEUPPER 0x0015
2297 #define NETDEV_RESEND_IGMP 0x0016
2298 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2299 #define NETDEV_CHANGEINFODATA 0x0018
2300 #define NETDEV_BONDING_INFO 0x0019
2301 #define NETDEV_PRECHANGEUPPER 0x001A
2302 #define NETDEV_CHANGELOWERSTATE 0x001B
2303 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2304 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2306 int register_netdevice_notifier(struct notifier_block *nb);
2307 int unregister_netdevice_notifier(struct notifier_block *nb);
2309 struct netdev_notifier_info {
2310 struct net_device *dev;
2313 struct netdev_notifier_change_info {
2314 struct netdev_notifier_info info; /* must be first */
2315 unsigned int flags_changed;
2318 struct netdev_notifier_changeupper_info {
2319 struct netdev_notifier_info info; /* must be first */
2320 struct net_device *upper_dev; /* new upper dev */
2321 bool master; /* is upper dev master */
2322 bool linking; /* is the notification for link or unlink */
2323 void *upper_info; /* upper dev info */
2326 struct netdev_notifier_changelowerstate_info {
2327 struct netdev_notifier_info info; /* must be first */
2328 void *lower_state_info; /* is lower dev state */
2331 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2332 struct net_device *dev)
2334 info->dev = dev;
2337 static inline struct net_device *
2338 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2340 return info->dev;
2343 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2346 extern rwlock_t dev_base_lock; /* Device list lock */
2348 #define for_each_netdev(net, d) \
2349 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2350 #define for_each_netdev_reverse(net, d) \
2351 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2352 #define for_each_netdev_rcu(net, d) \
2353 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2354 #define for_each_netdev_safe(net, d, n) \
2355 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2356 #define for_each_netdev_continue(net, d) \
2357 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2358 #define for_each_netdev_continue_rcu(net, d) \
2359 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2360 #define for_each_netdev_in_bond_rcu(bond, slave) \
2361 for_each_netdev_rcu(&init_net, slave) \
2362 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2363 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2365 static inline struct net_device *next_net_device(struct net_device *dev)
2367 struct list_head *lh;
2368 struct net *net;
2370 net = dev_net(dev);
2371 lh = dev->dev_list.next;
2372 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2375 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2377 struct list_head *lh;
2378 struct net *net;
2380 net = dev_net(dev);
2381 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2382 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2385 static inline struct net_device *first_net_device(struct net *net)
2387 return list_empty(&net->dev_base_head) ? NULL :
2388 net_device_entry(net->dev_base_head.next);
2391 static inline struct net_device *first_net_device_rcu(struct net *net)
2393 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2395 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2398 int netdev_boot_setup_check(struct net_device *dev);
2399 unsigned long netdev_boot_base(const char *prefix, int unit);
2400 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2401 const char *hwaddr);
2402 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2403 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2404 void dev_add_pack(struct packet_type *pt);
2405 void dev_remove_pack(struct packet_type *pt);
2406 void __dev_remove_pack(struct packet_type *pt);
2407 void dev_add_offload(struct packet_offload *po);
2408 void dev_remove_offload(struct packet_offload *po);
2410 int dev_get_iflink(const struct net_device *dev);
2411 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2412 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2413 unsigned short mask);
2414 struct net_device *dev_get_by_name(struct net *net, const char *name);
2415 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2416 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2417 int dev_alloc_name(struct net_device *dev, const char *name);
2418 int dev_open(struct net_device *dev);
2419 int dev_close(struct net_device *dev);
2420 int dev_close_many(struct list_head *head, bool unlink);
2421 void dev_disable_lro(struct net_device *dev);
2422 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2423 int dev_queue_xmit(struct sk_buff *skb);
2424 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2425 int register_netdevice(struct net_device *dev);
2426 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2427 void unregister_netdevice_many(struct list_head *head);
2428 static inline void unregister_netdevice(struct net_device *dev)
2430 unregister_netdevice_queue(dev, NULL);
2433 int netdev_refcnt_read(const struct net_device *dev);
2434 void free_netdev(struct net_device *dev);
2435 void netdev_freemem(struct net_device *dev);
2436 void synchronize_net(void);
2437 int init_dummy_netdev(struct net_device *dev);
2439 DECLARE_PER_CPU(int, xmit_recursion);
2440 #define XMIT_RECURSION_LIMIT 10
2442 static inline int dev_recursion_level(void)
2444 return this_cpu_read(xmit_recursion);
2447 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2448 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2449 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2450 int netdev_get_name(struct net *net, char *name, int ifindex);
2451 int dev_restart(struct net_device *dev);
2452 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2454 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2456 return NAPI_GRO_CB(skb)->data_offset;
2459 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2461 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2464 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2466 NAPI_GRO_CB(skb)->data_offset += len;
2469 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2470 unsigned int offset)
2472 return NAPI_GRO_CB(skb)->frag0 + offset;
2475 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2477 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2480 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2481 unsigned int offset)
2483 if (!pskb_may_pull(skb, hlen))
2484 return NULL;
2486 NAPI_GRO_CB(skb)->frag0 = NULL;
2487 NAPI_GRO_CB(skb)->frag0_len = 0;
2488 return skb->data + offset;
2491 static inline void *skb_gro_network_header(struct sk_buff *skb)
2493 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2494 skb_network_offset(skb);
2497 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2498 const void *start, unsigned int len)
2500 if (NAPI_GRO_CB(skb)->csum_valid)
2501 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2502 csum_partial(start, len, 0));
2505 /* GRO checksum functions. These are logical equivalents of the normal
2506 * checksum functions (in skbuff.h) except that they operate on the GRO
2507 * offsets and fields in sk_buff.
2510 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2512 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2514 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2517 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2518 bool zero_okay,
2519 __sum16 check)
2521 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2522 skb_checksum_start_offset(skb) <
2523 skb_gro_offset(skb)) &&
2524 !skb_at_gro_remcsum_start(skb) &&
2525 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2526 (!zero_okay || check));
2529 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2530 __wsum psum)
2532 if (NAPI_GRO_CB(skb)->csum_valid &&
2533 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2534 return 0;
2536 NAPI_GRO_CB(skb)->csum = psum;
2538 return __skb_gro_checksum_complete(skb);
2541 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2543 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2544 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2545 NAPI_GRO_CB(skb)->csum_cnt--;
2546 } else {
2547 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2548 * verified a new top level checksum or an encapsulated one
2549 * during GRO. This saves work if we fallback to normal path.
2551 __skb_incr_checksum_unnecessary(skb);
2555 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2556 compute_pseudo) \
2557 ({ \
2558 __sum16 __ret = 0; \
2559 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2560 __ret = __skb_gro_checksum_validate_complete(skb, \
2561 compute_pseudo(skb, proto)); \
2562 if (__ret) \
2563 __skb_mark_checksum_bad(skb); \
2564 else \
2565 skb_gro_incr_csum_unnecessary(skb); \
2566 __ret; \
2569 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2570 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2572 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2573 compute_pseudo) \
2574 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2576 #define skb_gro_checksum_simple_validate(skb) \
2577 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2579 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2581 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2582 !NAPI_GRO_CB(skb)->csum_valid);
2585 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2586 __sum16 check, __wsum pseudo)
2588 NAPI_GRO_CB(skb)->csum = ~pseudo;
2589 NAPI_GRO_CB(skb)->csum_valid = 1;
2592 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2593 do { \
2594 if (__skb_gro_checksum_convert_check(skb)) \
2595 __skb_gro_checksum_convert(skb, check, \
2596 compute_pseudo(skb, proto)); \
2597 } while (0)
2599 struct gro_remcsum {
2600 int offset;
2601 __wsum delta;
2604 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2606 grc->offset = 0;
2607 grc->delta = 0;
2610 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2611 unsigned int off, size_t hdrlen,
2612 int start, int offset,
2613 struct gro_remcsum *grc,
2614 bool nopartial)
2616 __wsum delta;
2617 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2619 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2621 if (!nopartial) {
2622 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2623 return ptr;
2626 ptr = skb_gro_header_fast(skb, off);
2627 if (skb_gro_header_hard(skb, off + plen)) {
2628 ptr = skb_gro_header_slow(skb, off + plen, off);
2629 if (!ptr)
2630 return NULL;
2633 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2634 start, offset);
2636 /* Adjust skb->csum since we changed the packet */
2637 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2639 grc->offset = off + hdrlen + offset;
2640 grc->delta = delta;
2642 return ptr;
2645 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2646 struct gro_remcsum *grc)
2648 void *ptr;
2649 size_t plen = grc->offset + sizeof(u16);
2651 if (!grc->delta)
2652 return;
2654 ptr = skb_gro_header_fast(skb, grc->offset);
2655 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2656 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2657 if (!ptr)
2658 return;
2661 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2664 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2665 unsigned short type,
2666 const void *daddr, const void *saddr,
2667 unsigned int len)
2669 if (!dev->header_ops || !dev->header_ops->create)
2670 return 0;
2672 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2675 static inline int dev_parse_header(const struct sk_buff *skb,
2676 unsigned char *haddr)
2678 const struct net_device *dev = skb->dev;
2680 if (!dev->header_ops || !dev->header_ops->parse)
2681 return 0;
2682 return dev->header_ops->parse(skb, haddr);
2685 /* ll_header must have at least hard_header_len allocated */
2686 static inline bool dev_validate_header(const struct net_device *dev,
2687 char *ll_header, int len)
2689 if (likely(len >= dev->hard_header_len))
2690 return true;
2692 if (capable(CAP_SYS_RAWIO)) {
2693 memset(ll_header + len, 0, dev->hard_header_len - len);
2694 return true;
2697 if (dev->header_ops && dev->header_ops->validate)
2698 return dev->header_ops->validate(ll_header, len);
2700 return false;
2703 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2704 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2705 static inline int unregister_gifconf(unsigned int family)
2707 return register_gifconf(family, NULL);
2710 #ifdef CONFIG_NET_FLOW_LIMIT
2711 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2712 struct sd_flow_limit {
2713 u64 count;
2714 unsigned int num_buckets;
2715 unsigned int history_head;
2716 u16 history[FLOW_LIMIT_HISTORY];
2717 u8 buckets[];
2720 extern int netdev_flow_limit_table_len;
2721 #endif /* CONFIG_NET_FLOW_LIMIT */
2724 * Incoming packets are placed on per-CPU queues
2726 struct softnet_data {
2727 struct list_head poll_list;
2728 struct sk_buff_head process_queue;
2730 /* stats */
2731 unsigned int processed;
2732 unsigned int time_squeeze;
2733 unsigned int received_rps;
2734 #ifdef CONFIG_RPS
2735 struct softnet_data *rps_ipi_list;
2736 #endif
2737 #ifdef CONFIG_NET_FLOW_LIMIT
2738 struct sd_flow_limit __rcu *flow_limit;
2739 #endif
2740 struct Qdisc *output_queue;
2741 struct Qdisc **output_queue_tailp;
2742 struct sk_buff *completion_queue;
2744 #ifdef CONFIG_RPS
2745 /* input_queue_head should be written by cpu owning this struct,
2746 * and only read by other cpus. Worth using a cache line.
2748 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2750 /* Elements below can be accessed between CPUs for RPS/RFS */
2751 struct call_single_data csd ____cacheline_aligned_in_smp;
2752 struct softnet_data *rps_ipi_next;
2753 unsigned int cpu;
2754 unsigned int input_queue_tail;
2755 #endif
2756 unsigned int dropped;
2757 struct sk_buff_head input_pkt_queue;
2758 struct napi_struct backlog;
2762 static inline void input_queue_head_incr(struct softnet_data *sd)
2764 #ifdef CONFIG_RPS
2765 sd->input_queue_head++;
2766 #endif
2769 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2770 unsigned int *qtail)
2772 #ifdef CONFIG_RPS
2773 *qtail = ++sd->input_queue_tail;
2774 #endif
2777 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2779 void __netif_schedule(struct Qdisc *q);
2780 void netif_schedule_queue(struct netdev_queue *txq);
2782 static inline void netif_tx_schedule_all(struct net_device *dev)
2784 unsigned int i;
2786 for (i = 0; i < dev->num_tx_queues; i++)
2787 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2790 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2792 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2796 * netif_start_queue - allow transmit
2797 * @dev: network device
2799 * Allow upper layers to call the device hard_start_xmit routine.
2801 static inline void netif_start_queue(struct net_device *dev)
2803 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2806 static inline void netif_tx_start_all_queues(struct net_device *dev)
2808 unsigned int i;
2810 for (i = 0; i < dev->num_tx_queues; i++) {
2811 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2812 netif_tx_start_queue(txq);
2816 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2819 * netif_wake_queue - restart transmit
2820 * @dev: network device
2822 * Allow upper layers to call the device hard_start_xmit routine.
2823 * Used for flow control when transmit resources are available.
2825 static inline void netif_wake_queue(struct net_device *dev)
2827 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2830 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2832 unsigned int i;
2834 for (i = 0; i < dev->num_tx_queues; i++) {
2835 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2836 netif_tx_wake_queue(txq);
2840 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2842 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2846 * netif_stop_queue - stop transmitted packets
2847 * @dev: network device
2849 * Stop upper layers calling the device hard_start_xmit routine.
2850 * Used for flow control when transmit resources are unavailable.
2852 static inline void netif_stop_queue(struct net_device *dev)
2854 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2857 void netif_tx_stop_all_queues(struct net_device *dev);
2859 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2861 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2865 * netif_queue_stopped - test if transmit queue is flowblocked
2866 * @dev: network device
2868 * Test if transmit queue on device is currently unable to send.
2870 static inline bool netif_queue_stopped(const struct net_device *dev)
2872 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2875 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2877 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2880 static inline bool
2881 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2883 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2886 static inline bool
2887 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2889 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2893 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2894 * @dev_queue: pointer to transmit queue
2896 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2897 * to give appropriate hint to the CPU.
2899 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2901 #ifdef CONFIG_BQL
2902 prefetchw(&dev_queue->dql.num_queued);
2903 #endif
2907 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2908 * @dev_queue: pointer to transmit queue
2910 * BQL enabled drivers might use this helper in their TX completion path,
2911 * to give appropriate hint to the CPU.
2913 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2915 #ifdef CONFIG_BQL
2916 prefetchw(&dev_queue->dql.limit);
2917 #endif
2920 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2921 unsigned int bytes)
2923 #ifdef CONFIG_BQL
2924 dql_queued(&dev_queue->dql, bytes);
2926 if (likely(dql_avail(&dev_queue->dql) >= 0))
2927 return;
2929 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2932 * The XOFF flag must be set before checking the dql_avail below,
2933 * because in netdev_tx_completed_queue we update the dql_completed
2934 * before checking the XOFF flag.
2936 smp_mb();
2938 /* check again in case another CPU has just made room avail */
2939 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2940 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2941 #endif
2945 * netdev_sent_queue - report the number of bytes queued to hardware
2946 * @dev: network device
2947 * @bytes: number of bytes queued to the hardware device queue
2949 * Report the number of bytes queued for sending/completion to the network
2950 * device hardware queue. @bytes should be a good approximation and should
2951 * exactly match netdev_completed_queue() @bytes
2953 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2955 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2958 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2959 unsigned int pkts, unsigned int bytes)
2961 #ifdef CONFIG_BQL
2962 if (unlikely(!bytes))
2963 return;
2965 dql_completed(&dev_queue->dql, bytes);
2968 * Without the memory barrier there is a small possiblity that
2969 * netdev_tx_sent_queue will miss the update and cause the queue to
2970 * be stopped forever
2972 smp_mb();
2974 if (dql_avail(&dev_queue->dql) < 0)
2975 return;
2977 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2978 netif_schedule_queue(dev_queue);
2979 #endif
2983 * netdev_completed_queue - report bytes and packets completed by device
2984 * @dev: network device
2985 * @pkts: actual number of packets sent over the medium
2986 * @bytes: actual number of bytes sent over the medium
2988 * Report the number of bytes and packets transmitted by the network device
2989 * hardware queue over the physical medium, @bytes must exactly match the
2990 * @bytes amount passed to netdev_sent_queue()
2992 static inline void netdev_completed_queue(struct net_device *dev,
2993 unsigned int pkts, unsigned int bytes)
2995 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2998 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3000 #ifdef CONFIG_BQL
3001 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3002 dql_reset(&q->dql);
3003 #endif
3007 * netdev_reset_queue - reset the packets and bytes count of a network device
3008 * @dev_queue: network device
3010 * Reset the bytes and packet count of a network device and clear the
3011 * software flow control OFF bit for this network device
3013 static inline void netdev_reset_queue(struct net_device *dev_queue)
3015 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3019 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3020 * @dev: network device
3021 * @queue_index: given tx queue index
3023 * Returns 0 if given tx queue index >= number of device tx queues,
3024 * otherwise returns the originally passed tx queue index.
3026 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3028 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3029 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3030 dev->name, queue_index,
3031 dev->real_num_tx_queues);
3032 return 0;
3035 return queue_index;
3039 * netif_running - test if up
3040 * @dev: network device
3042 * Test if the device has been brought up.
3044 static inline bool netif_running(const struct net_device *dev)
3046 return test_bit(__LINK_STATE_START, &dev->state);
3050 * Routines to manage the subqueues on a device. We only need start,
3051 * stop, and a check if it's stopped. All other device management is
3052 * done at the overall netdevice level.
3053 * Also test the device if we're multiqueue.
3057 * netif_start_subqueue - allow sending packets on subqueue
3058 * @dev: network device
3059 * @queue_index: sub queue index
3061 * Start individual transmit queue of a device with multiple transmit queues.
3063 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3065 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3067 netif_tx_start_queue(txq);
3071 * netif_stop_subqueue - stop sending packets on subqueue
3072 * @dev: network device
3073 * @queue_index: sub queue index
3075 * Stop individual transmit queue of a device with multiple transmit queues.
3077 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3079 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3080 netif_tx_stop_queue(txq);
3084 * netif_subqueue_stopped - test status of subqueue
3085 * @dev: network device
3086 * @queue_index: sub queue index
3088 * Check individual transmit queue of a device with multiple transmit queues.
3090 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3091 u16 queue_index)
3093 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3095 return netif_tx_queue_stopped(txq);
3098 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3099 struct sk_buff *skb)
3101 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3104 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3106 #ifdef CONFIG_XPS
3107 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3108 u16 index);
3109 #else
3110 static inline int netif_set_xps_queue(struct net_device *dev,
3111 const struct cpumask *mask,
3112 u16 index)
3114 return 0;
3116 #endif
3118 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3119 unsigned int num_tx_queues);
3122 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3123 * as a distribution range limit for the returned value.
3125 static inline u16 skb_tx_hash(const struct net_device *dev,
3126 struct sk_buff *skb)
3128 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3132 * netif_is_multiqueue - test if device has multiple transmit queues
3133 * @dev: network device
3135 * Check if device has multiple transmit queues
3137 static inline bool netif_is_multiqueue(const struct net_device *dev)
3139 return dev->num_tx_queues > 1;
3142 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3144 #ifdef CONFIG_SYSFS
3145 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3146 #else
3147 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3148 unsigned int rxq)
3150 return 0;
3152 #endif
3154 #ifdef CONFIG_SYSFS
3155 static inline unsigned int get_netdev_rx_queue_index(
3156 struct netdev_rx_queue *queue)
3158 struct net_device *dev = queue->dev;
3159 int index = queue - dev->_rx;
3161 BUG_ON(index >= dev->num_rx_queues);
3162 return index;
3164 #endif
3166 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3167 int netif_get_num_default_rss_queues(void);
3169 enum skb_free_reason {
3170 SKB_REASON_CONSUMED,
3171 SKB_REASON_DROPPED,
3174 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3175 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3178 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3179 * interrupt context or with hardware interrupts being disabled.
3180 * (in_irq() || irqs_disabled())
3182 * We provide four helpers that can be used in following contexts :
3184 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3185 * replacing kfree_skb(skb)
3187 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3188 * Typically used in place of consume_skb(skb) in TX completion path
3190 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3191 * replacing kfree_skb(skb)
3193 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3194 * and consumed a packet. Used in place of consume_skb(skb)
3196 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3198 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3201 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3203 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3206 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3208 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3211 static inline void dev_consume_skb_any(struct sk_buff *skb)
3213 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3216 int netif_rx(struct sk_buff *skb);
3217 int netif_rx_ni(struct sk_buff *skb);
3218 int netif_receive_skb(struct sk_buff *skb);
3219 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3220 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3221 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3222 gro_result_t napi_gro_frags(struct napi_struct *napi);
3223 struct packet_offload *gro_find_receive_by_type(__be16 type);
3224 struct packet_offload *gro_find_complete_by_type(__be16 type);
3226 static inline void napi_free_frags(struct napi_struct *napi)
3228 kfree_skb(napi->skb);
3229 napi->skb = NULL;
3232 bool netdev_is_rx_handler_busy(struct net_device *dev);
3233 int netdev_rx_handler_register(struct net_device *dev,
3234 rx_handler_func_t *rx_handler,
3235 void *rx_handler_data);
3236 void netdev_rx_handler_unregister(struct net_device *dev);
3238 bool dev_valid_name(const char *name);
3239 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3240 int dev_ethtool(struct net *net, struct ifreq *);
3241 unsigned int dev_get_flags(const struct net_device *);
3242 int __dev_change_flags(struct net_device *, unsigned int flags);
3243 int dev_change_flags(struct net_device *, unsigned int);
3244 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3245 unsigned int gchanges);
3246 int dev_change_name(struct net_device *, const char *);
3247 int dev_set_alias(struct net_device *, const char *, size_t);
3248 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3249 int dev_set_mtu(struct net_device *, int);
3250 void dev_set_group(struct net_device *, int);
3251 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3252 int dev_change_carrier(struct net_device *, bool new_carrier);
3253 int dev_get_phys_port_id(struct net_device *dev,
3254 struct netdev_phys_item_id *ppid);
3255 int dev_get_phys_port_name(struct net_device *dev,
3256 char *name, size_t len);
3257 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3258 int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
3259 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3260 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3261 struct netdev_queue *txq, int *ret);
3262 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3263 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3264 bool is_skb_forwardable(const struct net_device *dev,
3265 const struct sk_buff *skb);
3267 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3268 struct sk_buff *skb)
3270 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3271 unlikely(!is_skb_forwardable(dev, skb))) {
3272 atomic_long_inc(&dev->rx_dropped);
3273 kfree_skb(skb);
3274 return NET_RX_DROP;
3277 skb_scrub_packet(skb, true);
3278 skb->priority = 0;
3279 return 0;
3282 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3284 extern int netdev_budget;
3286 /* Called by rtnetlink.c:rtnl_unlock() */
3287 void netdev_run_todo(void);
3290 * dev_put - release reference to device
3291 * @dev: network device
3293 * Release reference to device to allow it to be freed.
3295 static inline void dev_put(struct net_device *dev)
3297 this_cpu_dec(*dev->pcpu_refcnt);
3301 * dev_hold - get reference to device
3302 * @dev: network device
3304 * Hold reference to device to keep it from being freed.
3306 static inline void dev_hold(struct net_device *dev)
3308 this_cpu_inc(*dev->pcpu_refcnt);
3311 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3312 * and _off may be called from IRQ context, but it is caller
3313 * who is responsible for serialization of these calls.
3315 * The name carrier is inappropriate, these functions should really be
3316 * called netif_lowerlayer_*() because they represent the state of any
3317 * kind of lower layer not just hardware media.
3320 void linkwatch_init_dev(struct net_device *dev);
3321 void linkwatch_fire_event(struct net_device *dev);
3322 void linkwatch_forget_dev(struct net_device *dev);
3325 * netif_carrier_ok - test if carrier present
3326 * @dev: network device
3328 * Check if carrier is present on device
3330 static inline bool netif_carrier_ok(const struct net_device *dev)
3332 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3335 unsigned long dev_trans_start(struct net_device *dev);
3337 void __netdev_watchdog_up(struct net_device *dev);
3339 void netif_carrier_on(struct net_device *dev);
3341 void netif_carrier_off(struct net_device *dev);
3344 * netif_dormant_on - mark device as dormant.
3345 * @dev: network device
3347 * Mark device as dormant (as per RFC2863).
3349 * The dormant state indicates that the relevant interface is not
3350 * actually in a condition to pass packets (i.e., it is not 'up') but is
3351 * in a "pending" state, waiting for some external event. For "on-
3352 * demand" interfaces, this new state identifies the situation where the
3353 * interface is waiting for events to place it in the up state.
3355 static inline void netif_dormant_on(struct net_device *dev)
3357 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3358 linkwatch_fire_event(dev);
3362 * netif_dormant_off - set device as not dormant.
3363 * @dev: network device
3365 * Device is not in dormant state.
3367 static inline void netif_dormant_off(struct net_device *dev)
3369 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3370 linkwatch_fire_event(dev);
3374 * netif_dormant - test if carrier present
3375 * @dev: network device
3377 * Check if carrier is present on device
3379 static inline bool netif_dormant(const struct net_device *dev)
3381 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3386 * netif_oper_up - test if device is operational
3387 * @dev: network device
3389 * Check if carrier is operational
3391 static inline bool netif_oper_up(const struct net_device *dev)
3393 return (dev->operstate == IF_OPER_UP ||
3394 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3398 * netif_device_present - is device available or removed
3399 * @dev: network device
3401 * Check if device has not been removed from system.
3403 static inline bool netif_device_present(struct net_device *dev)
3405 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3408 void netif_device_detach(struct net_device *dev);
3410 void netif_device_attach(struct net_device *dev);
3413 * Network interface message level settings
3416 enum {
3417 NETIF_MSG_DRV = 0x0001,
3418 NETIF_MSG_PROBE = 0x0002,
3419 NETIF_MSG_LINK = 0x0004,
3420 NETIF_MSG_TIMER = 0x0008,
3421 NETIF_MSG_IFDOWN = 0x0010,
3422 NETIF_MSG_IFUP = 0x0020,
3423 NETIF_MSG_RX_ERR = 0x0040,
3424 NETIF_MSG_TX_ERR = 0x0080,
3425 NETIF_MSG_TX_QUEUED = 0x0100,
3426 NETIF_MSG_INTR = 0x0200,
3427 NETIF_MSG_TX_DONE = 0x0400,
3428 NETIF_MSG_RX_STATUS = 0x0800,
3429 NETIF_MSG_PKTDATA = 0x1000,
3430 NETIF_MSG_HW = 0x2000,
3431 NETIF_MSG_WOL = 0x4000,
3434 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3435 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3436 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3437 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3438 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3439 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3440 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3441 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3442 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3443 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3444 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3445 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3446 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3447 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3448 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3450 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3452 /* use default */
3453 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3454 return default_msg_enable_bits;
3455 if (debug_value == 0) /* no output */
3456 return 0;
3457 /* set low N bits */
3458 return (1 << debug_value) - 1;
3461 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3463 spin_lock(&txq->_xmit_lock);
3464 txq->xmit_lock_owner = cpu;
3467 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3469 __acquire(&txq->_xmit_lock);
3470 return true;
3473 static inline void __netif_tx_release(struct netdev_queue *txq)
3475 __release(&txq->_xmit_lock);
3478 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3480 spin_lock_bh(&txq->_xmit_lock);
3481 txq->xmit_lock_owner = smp_processor_id();
3484 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3486 bool ok = spin_trylock(&txq->_xmit_lock);
3487 if (likely(ok))
3488 txq->xmit_lock_owner = smp_processor_id();
3489 return ok;
3492 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3494 txq->xmit_lock_owner = -1;
3495 spin_unlock(&txq->_xmit_lock);
3498 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3500 txq->xmit_lock_owner = -1;
3501 spin_unlock_bh(&txq->_xmit_lock);
3504 static inline void txq_trans_update(struct netdev_queue *txq)
3506 if (txq->xmit_lock_owner != -1)
3507 txq->trans_start = jiffies;
3510 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3511 static inline void netif_trans_update(struct net_device *dev)
3513 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3515 if (txq->trans_start != jiffies)
3516 txq->trans_start = jiffies;
3520 * netif_tx_lock - grab network device transmit lock
3521 * @dev: network device
3523 * Get network device transmit lock
3525 static inline void netif_tx_lock(struct net_device *dev)
3527 unsigned int i;
3528 int cpu;
3530 spin_lock(&dev->tx_global_lock);
3531 cpu = smp_processor_id();
3532 for (i = 0; i < dev->num_tx_queues; i++) {
3533 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3535 /* We are the only thread of execution doing a
3536 * freeze, but we have to grab the _xmit_lock in
3537 * order to synchronize with threads which are in
3538 * the ->hard_start_xmit() handler and already
3539 * checked the frozen bit.
3541 __netif_tx_lock(txq, cpu);
3542 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3543 __netif_tx_unlock(txq);
3547 static inline void netif_tx_lock_bh(struct net_device *dev)
3549 local_bh_disable();
3550 netif_tx_lock(dev);
3553 static inline void netif_tx_unlock(struct net_device *dev)
3555 unsigned int i;
3557 for (i = 0; i < dev->num_tx_queues; i++) {
3558 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3560 /* No need to grab the _xmit_lock here. If the
3561 * queue is not stopped for another reason, we
3562 * force a schedule.
3564 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3565 netif_schedule_queue(txq);
3567 spin_unlock(&dev->tx_global_lock);
3570 static inline void netif_tx_unlock_bh(struct net_device *dev)
3572 netif_tx_unlock(dev);
3573 local_bh_enable();
3576 #define HARD_TX_LOCK(dev, txq, cpu) { \
3577 if ((dev->features & NETIF_F_LLTX) == 0) { \
3578 __netif_tx_lock(txq, cpu); \
3579 } else { \
3580 __netif_tx_acquire(txq); \
3584 #define HARD_TX_TRYLOCK(dev, txq) \
3585 (((dev->features & NETIF_F_LLTX) == 0) ? \
3586 __netif_tx_trylock(txq) : \
3587 __netif_tx_acquire(txq))
3589 #define HARD_TX_UNLOCK(dev, txq) { \
3590 if ((dev->features & NETIF_F_LLTX) == 0) { \
3591 __netif_tx_unlock(txq); \
3592 } else { \
3593 __netif_tx_release(txq); \
3597 static inline void netif_tx_disable(struct net_device *dev)
3599 unsigned int i;
3600 int cpu;
3602 local_bh_disable();
3603 cpu = smp_processor_id();
3604 for (i = 0; i < dev->num_tx_queues; i++) {
3605 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3607 __netif_tx_lock(txq, cpu);
3608 netif_tx_stop_queue(txq);
3609 __netif_tx_unlock(txq);
3611 local_bh_enable();
3614 static inline void netif_addr_lock(struct net_device *dev)
3616 spin_lock(&dev->addr_list_lock);
3619 static inline void netif_addr_lock_nested(struct net_device *dev)
3621 int subclass = SINGLE_DEPTH_NESTING;
3623 if (dev->netdev_ops->ndo_get_lock_subclass)
3624 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3626 spin_lock_nested(&dev->addr_list_lock, subclass);
3629 static inline void netif_addr_lock_bh(struct net_device *dev)
3631 spin_lock_bh(&dev->addr_list_lock);
3634 static inline void netif_addr_unlock(struct net_device *dev)
3636 spin_unlock(&dev->addr_list_lock);
3639 static inline void netif_addr_unlock_bh(struct net_device *dev)
3641 spin_unlock_bh(&dev->addr_list_lock);
3645 * dev_addrs walker. Should be used only for read access. Call with
3646 * rcu_read_lock held.
3648 #define for_each_dev_addr(dev, ha) \
3649 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3651 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3653 void ether_setup(struct net_device *dev);
3655 /* Support for loadable net-drivers */
3656 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3657 unsigned char name_assign_type,
3658 void (*setup)(struct net_device *),
3659 unsigned int txqs, unsigned int rxqs);
3660 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3661 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3663 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3664 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3665 count)
3667 int register_netdev(struct net_device *dev);
3668 void unregister_netdev(struct net_device *dev);
3670 /* General hardware address lists handling functions */
3671 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3672 struct netdev_hw_addr_list *from_list, int addr_len);
3673 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3674 struct netdev_hw_addr_list *from_list, int addr_len);
3675 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3676 struct net_device *dev,
3677 int (*sync)(struct net_device *, const unsigned char *),
3678 int (*unsync)(struct net_device *,
3679 const unsigned char *));
3680 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3681 struct net_device *dev,
3682 int (*unsync)(struct net_device *,
3683 const unsigned char *));
3684 void __hw_addr_init(struct netdev_hw_addr_list *list);
3686 /* Functions used for device addresses handling */
3687 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3688 unsigned char addr_type);
3689 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3690 unsigned char addr_type);
3691 void dev_addr_flush(struct net_device *dev);
3692 int dev_addr_init(struct net_device *dev);
3694 /* Functions used for unicast addresses handling */
3695 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3696 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3697 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3698 int dev_uc_sync(struct net_device *to, struct net_device *from);
3699 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3700 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3701 void dev_uc_flush(struct net_device *dev);
3702 void dev_uc_init(struct net_device *dev);
3705 * __dev_uc_sync - Synchonize device's unicast list
3706 * @dev: device to sync
3707 * @sync: function to call if address should be added
3708 * @unsync: function to call if address should be removed
3710 * Add newly added addresses to the interface, and release
3711 * addresses that have been deleted.
3713 static inline int __dev_uc_sync(struct net_device *dev,
3714 int (*sync)(struct net_device *,
3715 const unsigned char *),
3716 int (*unsync)(struct net_device *,
3717 const unsigned char *))
3719 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3723 * __dev_uc_unsync - Remove synchronized addresses from device
3724 * @dev: device to sync
3725 * @unsync: function to call if address should be removed
3727 * Remove all addresses that were added to the device by dev_uc_sync().
3729 static inline void __dev_uc_unsync(struct net_device *dev,
3730 int (*unsync)(struct net_device *,
3731 const unsigned char *))
3733 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3736 /* Functions used for multicast addresses handling */
3737 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3738 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3739 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3740 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3741 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3742 int dev_mc_sync(struct net_device *to, struct net_device *from);
3743 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3744 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3745 void dev_mc_flush(struct net_device *dev);
3746 void dev_mc_init(struct net_device *dev);
3749 * __dev_mc_sync - Synchonize device's multicast list
3750 * @dev: device to sync
3751 * @sync: function to call if address should be added
3752 * @unsync: function to call if address should be removed
3754 * Add newly added addresses to the interface, and release
3755 * addresses that have been deleted.
3757 static inline int __dev_mc_sync(struct net_device *dev,
3758 int (*sync)(struct net_device *,
3759 const unsigned char *),
3760 int (*unsync)(struct net_device *,
3761 const unsigned char *))
3763 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3767 * __dev_mc_unsync - Remove synchronized addresses from device
3768 * @dev: device to sync
3769 * @unsync: function to call if address should be removed
3771 * Remove all addresses that were added to the device by dev_mc_sync().
3773 static inline void __dev_mc_unsync(struct net_device *dev,
3774 int (*unsync)(struct net_device *,
3775 const unsigned char *))
3777 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3780 /* Functions used for secondary unicast and multicast support */
3781 void dev_set_rx_mode(struct net_device *dev);
3782 void __dev_set_rx_mode(struct net_device *dev);
3783 int dev_set_promiscuity(struct net_device *dev, int inc);
3784 int dev_set_allmulti(struct net_device *dev, int inc);
3785 void netdev_state_change(struct net_device *dev);
3786 void netdev_notify_peers(struct net_device *dev);
3787 void netdev_features_change(struct net_device *dev);
3788 /* Load a device via the kmod */
3789 void dev_load(struct net *net, const char *name);
3790 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3791 struct rtnl_link_stats64 *storage);
3792 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3793 const struct net_device_stats *netdev_stats);
3795 extern int netdev_max_backlog;
3796 extern int netdev_tstamp_prequeue;
3797 extern int weight_p;
3799 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3800 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3801 struct list_head **iter);
3802 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3803 struct list_head **iter);
3805 /* iterate through upper list, must be called under RCU read lock */
3806 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3807 for (iter = &(dev)->adj_list.upper, \
3808 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3809 updev; \
3810 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3812 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3813 int (*fn)(struct net_device *upper_dev,
3814 void *data),
3815 void *data);
3817 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3818 struct net_device *upper_dev);
3820 void *netdev_lower_get_next_private(struct net_device *dev,
3821 struct list_head **iter);
3822 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3823 struct list_head **iter);
3825 #define netdev_for_each_lower_private(dev, priv, iter) \
3826 for (iter = (dev)->adj_list.lower.next, \
3827 priv = netdev_lower_get_next_private(dev, &(iter)); \
3828 priv; \
3829 priv = netdev_lower_get_next_private(dev, &(iter)))
3831 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3832 for (iter = &(dev)->adj_list.lower, \
3833 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3834 priv; \
3835 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3837 void *netdev_lower_get_next(struct net_device *dev,
3838 struct list_head **iter);
3840 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3841 for (iter = (dev)->adj_list.lower.next, \
3842 ldev = netdev_lower_get_next(dev, &(iter)); \
3843 ldev; \
3844 ldev = netdev_lower_get_next(dev, &(iter)))
3846 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3847 struct list_head **iter);
3848 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3849 struct list_head **iter);
3851 int netdev_walk_all_lower_dev(struct net_device *dev,
3852 int (*fn)(struct net_device *lower_dev,
3853 void *data),
3854 void *data);
3855 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3856 int (*fn)(struct net_device *lower_dev,
3857 void *data),
3858 void *data);
3860 void *netdev_adjacent_get_private(struct list_head *adj_list);
3861 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3862 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3863 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3864 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3865 int netdev_master_upper_dev_link(struct net_device *dev,
3866 struct net_device *upper_dev,
3867 void *upper_priv, void *upper_info);
3868 void netdev_upper_dev_unlink(struct net_device *dev,
3869 struct net_device *upper_dev);
3870 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3871 void *netdev_lower_dev_get_private(struct net_device *dev,
3872 struct net_device *lower_dev);
3873 void netdev_lower_state_changed(struct net_device *lower_dev,
3874 void *lower_state_info);
3875 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3876 struct neighbour *n);
3877 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3878 struct neighbour *n);
3880 /* RSS keys are 40 or 52 bytes long */
3881 #define NETDEV_RSS_KEY_LEN 52
3882 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3883 void netdev_rss_key_fill(void *buffer, size_t len);
3885 int dev_get_nest_level(struct net_device *dev);
3886 int skb_checksum_help(struct sk_buff *skb);
3887 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3888 netdev_features_t features, bool tx_path);
3889 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3890 netdev_features_t features);
3892 struct netdev_bonding_info {
3893 ifslave slave;
3894 ifbond master;
3897 struct netdev_notifier_bonding_info {
3898 struct netdev_notifier_info info; /* must be first */
3899 struct netdev_bonding_info bonding_info;
3902 void netdev_bonding_info_change(struct net_device *dev,
3903 struct netdev_bonding_info *bonding_info);
3905 static inline
3906 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3908 return __skb_gso_segment(skb, features, true);
3910 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3912 static inline bool can_checksum_protocol(netdev_features_t features,
3913 __be16 protocol)
3915 if (protocol == htons(ETH_P_FCOE))
3916 return !!(features & NETIF_F_FCOE_CRC);
3918 /* Assume this is an IP checksum (not SCTP CRC) */
3920 if (features & NETIF_F_HW_CSUM) {
3921 /* Can checksum everything */
3922 return true;
3925 switch (protocol) {
3926 case htons(ETH_P_IP):
3927 return !!(features & NETIF_F_IP_CSUM);
3928 case htons(ETH_P_IPV6):
3929 return !!(features & NETIF_F_IPV6_CSUM);
3930 default:
3931 return false;
3935 #ifdef CONFIG_BUG
3936 void netdev_rx_csum_fault(struct net_device *dev);
3937 #else
3938 static inline void netdev_rx_csum_fault(struct net_device *dev)
3941 #endif
3942 /* rx skb timestamps */
3943 void net_enable_timestamp(void);
3944 void net_disable_timestamp(void);
3946 #ifdef CONFIG_PROC_FS
3947 int __init dev_proc_init(void);
3948 #else
3949 #define dev_proc_init() 0
3950 #endif
3952 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3953 struct sk_buff *skb, struct net_device *dev,
3954 bool more)
3956 skb->xmit_more = more ? 1 : 0;
3957 return ops->ndo_start_xmit(skb, dev);
3960 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3961 struct netdev_queue *txq, bool more)
3963 const struct net_device_ops *ops = dev->netdev_ops;
3964 int rc;
3966 rc = __netdev_start_xmit(ops, skb, dev, more);
3967 if (rc == NETDEV_TX_OK)
3968 txq_trans_update(txq);
3970 return rc;
3973 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3974 const void *ns);
3975 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3976 const void *ns);
3978 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3980 return netdev_class_create_file_ns(class_attr, NULL);
3983 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3985 netdev_class_remove_file_ns(class_attr, NULL);
3988 extern struct kobj_ns_type_operations net_ns_type_operations;
3990 const char *netdev_drivername(const struct net_device *dev);
3992 void linkwatch_run_queue(void);
3994 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3995 netdev_features_t f2)
3997 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3998 if (f1 & NETIF_F_HW_CSUM)
3999 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4000 else
4001 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4004 return f1 & f2;
4007 static inline netdev_features_t netdev_get_wanted_features(
4008 struct net_device *dev)
4010 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4012 netdev_features_t netdev_increment_features(netdev_features_t all,
4013 netdev_features_t one, netdev_features_t mask);
4015 /* Allow TSO being used on stacked device :
4016 * Performing the GSO segmentation before last device
4017 * is a performance improvement.
4019 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4020 netdev_features_t mask)
4022 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4025 int __netdev_update_features(struct net_device *dev);
4026 void netdev_update_features(struct net_device *dev);
4027 void netdev_change_features(struct net_device *dev);
4029 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4030 struct net_device *dev);
4032 netdev_features_t passthru_features_check(struct sk_buff *skb,
4033 struct net_device *dev,
4034 netdev_features_t features);
4035 netdev_features_t netif_skb_features(struct sk_buff *skb);
4037 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4039 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4041 /* check flags correspondence */
4042 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4043 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4044 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4045 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4046 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4047 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4048 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4049 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4050 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4051 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4052 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4053 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4054 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4055 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4056 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4057 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4059 return (features & feature) == feature;
4062 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4064 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4065 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4068 static inline bool netif_needs_gso(struct sk_buff *skb,
4069 netdev_features_t features)
4071 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4072 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4073 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4076 static inline void netif_set_gso_max_size(struct net_device *dev,
4077 unsigned int size)
4079 dev->gso_max_size = size;
4082 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4083 int pulled_hlen, u16 mac_offset,
4084 int mac_len)
4086 skb->protocol = protocol;
4087 skb->encapsulation = 1;
4088 skb_push(skb, pulled_hlen);
4089 skb_reset_transport_header(skb);
4090 skb->mac_header = mac_offset;
4091 skb->network_header = skb->mac_header + mac_len;
4092 skb->mac_len = mac_len;
4095 static inline bool netif_is_macsec(const struct net_device *dev)
4097 return dev->priv_flags & IFF_MACSEC;
4100 static inline bool netif_is_macvlan(const struct net_device *dev)
4102 return dev->priv_flags & IFF_MACVLAN;
4105 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4107 return dev->priv_flags & IFF_MACVLAN_PORT;
4110 static inline bool netif_is_ipvlan(const struct net_device *dev)
4112 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4115 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4117 return dev->priv_flags & IFF_IPVLAN_MASTER;
4120 static inline bool netif_is_bond_master(const struct net_device *dev)
4122 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4125 static inline bool netif_is_bond_slave(const struct net_device *dev)
4127 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4130 static inline bool netif_supports_nofcs(struct net_device *dev)
4132 return dev->priv_flags & IFF_SUPP_NOFCS;
4135 static inline bool netif_is_l3_master(const struct net_device *dev)
4137 return dev->priv_flags & IFF_L3MDEV_MASTER;
4140 static inline bool netif_is_l3_slave(const struct net_device *dev)
4142 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4145 static inline bool netif_is_bridge_master(const struct net_device *dev)
4147 return dev->priv_flags & IFF_EBRIDGE;
4150 static inline bool netif_is_bridge_port(const struct net_device *dev)
4152 return dev->priv_flags & IFF_BRIDGE_PORT;
4155 static inline bool netif_is_ovs_master(const struct net_device *dev)
4157 return dev->priv_flags & IFF_OPENVSWITCH;
4160 static inline bool netif_is_team_master(const struct net_device *dev)
4162 return dev->priv_flags & IFF_TEAM;
4165 static inline bool netif_is_team_port(const struct net_device *dev)
4167 return dev->priv_flags & IFF_TEAM_PORT;
4170 static inline bool netif_is_lag_master(const struct net_device *dev)
4172 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4175 static inline bool netif_is_lag_port(const struct net_device *dev)
4177 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4180 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4182 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4185 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4186 static inline void netif_keep_dst(struct net_device *dev)
4188 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4191 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4192 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4194 /* TODO: reserve and use an additional IFF bit, if we get more users */
4195 return dev->priv_flags & IFF_MACSEC;
4198 extern struct pernet_operations __net_initdata loopback_net_ops;
4200 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4202 /* netdev_printk helpers, similar to dev_printk */
4204 static inline const char *netdev_name(const struct net_device *dev)
4206 if (!dev->name[0] || strchr(dev->name, '%'))
4207 return "(unnamed net_device)";
4208 return dev->name;
4211 static inline const char *netdev_reg_state(const struct net_device *dev)
4213 switch (dev->reg_state) {
4214 case NETREG_UNINITIALIZED: return " (uninitialized)";
4215 case NETREG_REGISTERED: return "";
4216 case NETREG_UNREGISTERING: return " (unregistering)";
4217 case NETREG_UNREGISTERED: return " (unregistered)";
4218 case NETREG_RELEASED: return " (released)";
4219 case NETREG_DUMMY: return " (dummy)";
4222 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4223 return " (unknown)";
4226 __printf(3, 4)
4227 void netdev_printk(const char *level, const struct net_device *dev,
4228 const char *format, ...);
4229 __printf(2, 3)
4230 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4231 __printf(2, 3)
4232 void netdev_alert(const struct net_device *dev, const char *format, ...);
4233 __printf(2, 3)
4234 void netdev_crit(const struct net_device *dev, const char *format, ...);
4235 __printf(2, 3)
4236 void netdev_err(const struct net_device *dev, const char *format, ...);
4237 __printf(2, 3)
4238 void netdev_warn(const struct net_device *dev, const char *format, ...);
4239 __printf(2, 3)
4240 void netdev_notice(const struct net_device *dev, const char *format, ...);
4241 __printf(2, 3)
4242 void netdev_info(const struct net_device *dev, const char *format, ...);
4244 #define MODULE_ALIAS_NETDEV(device) \
4245 MODULE_ALIAS("netdev-" device)
4247 #if defined(CONFIG_DYNAMIC_DEBUG)
4248 #define netdev_dbg(__dev, format, args...) \
4249 do { \
4250 dynamic_netdev_dbg(__dev, format, ##args); \
4251 } while (0)
4252 #elif defined(DEBUG)
4253 #define netdev_dbg(__dev, format, args...) \
4254 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4255 #else
4256 #define netdev_dbg(__dev, format, args...) \
4257 ({ \
4258 if (0) \
4259 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4261 #endif
4263 #if defined(VERBOSE_DEBUG)
4264 #define netdev_vdbg netdev_dbg
4265 #else
4267 #define netdev_vdbg(dev, format, args...) \
4268 ({ \
4269 if (0) \
4270 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4271 0; \
4273 #endif
4276 * netdev_WARN() acts like dev_printk(), but with the key difference
4277 * of using a WARN/WARN_ON to get the message out, including the
4278 * file/line information and a backtrace.
4280 #define netdev_WARN(dev, format, args...) \
4281 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4282 netdev_reg_state(dev), ##args)
4284 /* netif printk helpers, similar to netdev_printk */
4286 #define netif_printk(priv, type, level, dev, fmt, args...) \
4287 do { \
4288 if (netif_msg_##type(priv)) \
4289 netdev_printk(level, (dev), fmt, ##args); \
4290 } while (0)
4292 #define netif_level(level, priv, type, dev, fmt, args...) \
4293 do { \
4294 if (netif_msg_##type(priv)) \
4295 netdev_##level(dev, fmt, ##args); \
4296 } while (0)
4298 #define netif_emerg(priv, type, dev, fmt, args...) \
4299 netif_level(emerg, priv, type, dev, fmt, ##args)
4300 #define netif_alert(priv, type, dev, fmt, args...) \
4301 netif_level(alert, priv, type, dev, fmt, ##args)
4302 #define netif_crit(priv, type, dev, fmt, args...) \
4303 netif_level(crit, priv, type, dev, fmt, ##args)
4304 #define netif_err(priv, type, dev, fmt, args...) \
4305 netif_level(err, priv, type, dev, fmt, ##args)
4306 #define netif_warn(priv, type, dev, fmt, args...) \
4307 netif_level(warn, priv, type, dev, fmt, ##args)
4308 #define netif_notice(priv, type, dev, fmt, args...) \
4309 netif_level(notice, priv, type, dev, fmt, ##args)
4310 #define netif_info(priv, type, dev, fmt, args...) \
4311 netif_level(info, priv, type, dev, fmt, ##args)
4313 #if defined(CONFIG_DYNAMIC_DEBUG)
4314 #define netif_dbg(priv, type, netdev, format, args...) \
4315 do { \
4316 if (netif_msg_##type(priv)) \
4317 dynamic_netdev_dbg(netdev, format, ##args); \
4318 } while (0)
4319 #elif defined(DEBUG)
4320 #define netif_dbg(priv, type, dev, format, args...) \
4321 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4322 #else
4323 #define netif_dbg(priv, type, dev, format, args...) \
4324 ({ \
4325 if (0) \
4326 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4327 0; \
4329 #endif
4331 #if defined(VERBOSE_DEBUG)
4332 #define netif_vdbg netif_dbg
4333 #else
4334 #define netif_vdbg(priv, type, dev, format, args...) \
4335 ({ \
4336 if (0) \
4337 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4338 0; \
4340 #endif
4343 * The list of packet types we will receive (as opposed to discard)
4344 * and the routines to invoke.
4346 * Why 16. Because with 16 the only overlap we get on a hash of the
4347 * low nibble of the protocol value is RARP/SNAP/X.25.
4349 * NOTE: That is no longer true with the addition of VLAN tags. Not
4350 * sure which should go first, but I bet it won't make much
4351 * difference if we are running VLANs. The good news is that
4352 * this protocol won't be in the list unless compiled in, so
4353 * the average user (w/out VLANs) will not be adversely affected.
4354 * --BLG
4356 * 0800 IP
4357 * 8100 802.1Q VLAN
4358 * 0001 802.3
4359 * 0002 AX.25
4360 * 0004 802.2
4361 * 8035 RARP
4362 * 0005 SNAP
4363 * 0805 X.25
4364 * 0806 ARP
4365 * 8137 IPX
4366 * 0009 Localtalk
4367 * 86DD IPv6
4369 #define PTYPE_HASH_SIZE (16)
4370 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4372 #endif /* _LINUX_NETDEVICE_H */