Merge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
[linux-2.6.git] / include / linux / netdevice.h
blob7f0ed423a3606f1cc13e09a00d332dc4a45f924b
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/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.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>
54 struct netpoll_info;
55 struct device;
56 struct phy_device;
57 /* 802.11 specific */
58 struct wireless_dev;
59 /* source back-compat hooks */
60 #define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
63 void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct ethtool_ops *ops);
66 /* hardware address assignment types */
67 #define NET_ADDR_PERM 0 /* address is permanent (default) */
68 #define NET_ADDR_RANDOM 1 /* address is generated randomly */
69 #define NET_ADDR_STOLEN 2 /* address is stolen from other device */
70 #define NET_ADDR_SET 3 /* address is set using
71 * dev_set_mac_address() */
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously, in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), all
91 * others are propagated to higher layers.
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK 0xf0
110 enum netdev_tx {
111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
116 typedef enum netdev_tx netdev_tx_t;
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
122 static inline bool dev_xmit_complete(int rc)
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
130 if (likely(rc < NET_XMIT_MASK))
131 return true;
133 return false;
137 * Compute the worst case header length according to the protocols
138 * used.
141 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
144 # else
145 # define LL_MAX_HEADER 96
146 # endif
147 #else
148 # define LL_MAX_HEADER 32
149 #endif
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
154 #else
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
156 #endif
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
163 struct net_device_stats {
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
173 unsigned long collisions;
174 unsigned long rx_length_errors;
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
193 #ifdef CONFIG_RPS
194 #include <linux/static_key.h>
195 extern struct static_key rps_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 (*rebuild)(struct sk_buff *skb);
271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
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 at 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 spinlock_t poll_lock;
321 int poll_owner;
322 #endif
323 struct net_device *dev;
324 struct sk_buff *gro_list;
325 struct sk_buff *skb;
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 */
338 enum gro_result {
339 GRO_MERGED,
340 GRO_MERGED_FREE,
341 GRO_HELD,
342 GRO_NORMAL,
343 GRO_DROP,
345 typedef enum gro_result gro_result_t;
348 * enum rx_handler_result - Possible return values for rx_handlers.
349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
350 * further.
351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
352 * case skb->dev was changed by rx_handler.
353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
354 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
356 * rx_handlers are functions called from inside __netif_receive_skb(), to do
357 * special processing of the skb, prior to delivery to protocol handlers.
359 * Currently, a net_device can only have a single rx_handler registered. Trying
360 * to register a second rx_handler will return -EBUSY.
362 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
363 * To unregister a rx_handler on a net_device, use
364 * netdev_rx_handler_unregister().
366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
367 * do with the skb.
369 * If the rx_handler consumed to skb in some way, it should return
370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
371 * the skb to be delivered in some other ways.
373 * If the rx_handler changed skb->dev, to divert the skb to another
374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
375 * new device will be called if it exists.
377 * If the rx_handler consider the skb should be ignored, it should return
378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
379 * are registered on exact device (ptype->dev == skb->dev).
381 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
382 * delivered, it should return RX_HANDLER_PASS.
384 * A device without a registered rx_handler will behave as if rx_handler
385 * returned RX_HANDLER_PASS.
388 enum rx_handler_result {
389 RX_HANDLER_CONSUMED,
390 RX_HANDLER_ANOTHER,
391 RX_HANDLER_EXACT,
392 RX_HANDLER_PASS,
394 typedef enum rx_handler_result rx_handler_result_t;
395 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
397 void __napi_schedule(struct napi_struct *n);
399 static inline bool napi_disable_pending(struct napi_struct *n)
401 return test_bit(NAPI_STATE_DISABLE, &n->state);
405 * napi_schedule_prep - check if napi can be scheduled
406 * @n: napi context
408 * Test if NAPI routine is already running, and if not mark
409 * it as running. This is used as a condition variable
410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
413 static inline bool napi_schedule_prep(struct napi_struct *n)
415 return !napi_disable_pending(n) &&
416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
420 * napi_schedule - schedule NAPI poll
421 * @n: napi context
423 * Schedule NAPI poll routine to be called if it is not already
424 * running.
426 static inline void napi_schedule(struct napi_struct *n)
428 if (napi_schedule_prep(n))
429 __napi_schedule(n);
432 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
433 static inline bool napi_reschedule(struct napi_struct *napi)
435 if (napi_schedule_prep(napi)) {
436 __napi_schedule(napi);
437 return true;
439 return false;
443 * napi_complete - NAPI processing complete
444 * @n: napi context
446 * Mark NAPI processing as complete.
448 void __napi_complete(struct napi_struct *n);
449 void napi_complete(struct napi_struct *n);
452 * napi_by_id - lookup a NAPI by napi_id
453 * @napi_id: hashed napi_id
455 * lookup @napi_id in napi_hash table
456 * must be called under rcu_read_lock()
458 struct napi_struct *napi_by_id(unsigned int napi_id);
461 * napi_hash_add - add a NAPI to global hashtable
462 * @napi: napi context
464 * generate a new napi_id and store a @napi under it in napi_hash
466 void napi_hash_add(struct napi_struct *napi);
469 * napi_hash_del - remove a NAPI from global table
470 * @napi: napi context
472 * Warning: caller must observe rcu grace period
473 * before freeing memory containing @napi
475 void napi_hash_del(struct napi_struct *napi);
478 * napi_disable - prevent NAPI from scheduling
479 * @n: napi context
481 * Stop NAPI from being scheduled on this context.
482 * Waits till any outstanding processing completes.
484 static inline void napi_disable(struct napi_struct *n)
486 might_sleep();
487 set_bit(NAPI_STATE_DISABLE, &n->state);
488 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
489 msleep(1);
490 clear_bit(NAPI_STATE_DISABLE, &n->state);
494 * napi_enable - enable NAPI scheduling
495 * @n: napi context
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
500 static inline void napi_enable(struct napi_struct *n)
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
503 smp_mb__before_clear_bit();
504 clear_bit(NAPI_STATE_SCHED, &n->state);
507 #ifdef CONFIG_SMP
509 * napi_synchronize - wait until NAPI is not running
510 * @n: napi context
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
516 static inline void napi_synchronize(const struct napi_struct *n)
518 while (test_bit(NAPI_STATE_SCHED, &n->state))
519 msleep(1);
521 #else
522 # define napi_synchronize(n) barrier()
523 #endif
525 enum netdev_queue_state_t {
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
528 __QUEUE_STATE_FROZEN,
529 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
530 (1 << __QUEUE_STATE_STACK_XOFF))
531 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
532 (1 << __QUEUE_STATE_FROZEN))
535 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
536 * netif_tx_* functions below are used to manipulate this flag. The
537 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
538 * queue independently. The netif_xmit_*stopped functions below are called
539 * to check if the queue has been stopped by the driver or stack (either
540 * of the XOFF bits are set in the state). Drivers should not need to call
541 * netif_xmit*stopped functions, they should only be using netif_tx_*.
544 struct netdev_queue {
546 * read mostly part
548 struct net_device *dev;
549 struct Qdisc *qdisc;
550 struct Qdisc *qdisc_sleeping;
551 #ifdef CONFIG_SYSFS
552 struct kobject kobj;
553 #endif
554 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
555 int numa_node;
556 #endif
558 * write mostly part
560 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
561 int xmit_lock_owner;
563 * please use this field instead of dev->trans_start
565 unsigned long trans_start;
568 * Number of TX timeouts for this queue
569 * (/sys/class/net/DEV/Q/trans_timeout)
571 unsigned long trans_timeout;
573 unsigned long state;
575 #ifdef CONFIG_BQL
576 struct dql dql;
577 #endif
578 } ____cacheline_aligned_in_smp;
580 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
582 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
583 return q->numa_node;
584 #else
585 return NUMA_NO_NODE;
586 #endif
589 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
591 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
592 q->numa_node = node;
593 #endif
596 #ifdef CONFIG_RPS
598 * This structure holds an RPS map which can be of variable length. The
599 * map is an array of CPUs.
601 struct rps_map {
602 unsigned int len;
603 struct rcu_head rcu;
604 u16 cpus[0];
606 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
609 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
610 * tail pointer for that CPU's input queue at the time of last enqueue, and
611 * a hardware filter index.
613 struct rps_dev_flow {
614 u16 cpu;
615 u16 filter;
616 unsigned int last_qtail;
618 #define RPS_NO_FILTER 0xffff
621 * The rps_dev_flow_table structure contains a table of flow mappings.
623 struct rps_dev_flow_table {
624 unsigned int mask;
625 struct rcu_head rcu;
626 struct rps_dev_flow flows[0];
628 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
629 ((_num) * sizeof(struct rps_dev_flow)))
632 * The rps_sock_flow_table contains mappings of flows to the last CPU
633 * on which they were processed by the application (set in recvmsg).
635 struct rps_sock_flow_table {
636 unsigned int mask;
637 u16 ents[0];
639 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
640 ((_num) * sizeof(u16)))
642 #define RPS_NO_CPU 0xffff
644 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
645 u32 hash)
647 if (table && hash) {
648 unsigned int cpu, index = hash & table->mask;
650 /* We only give a hint, preemption can change cpu under us */
651 cpu = raw_smp_processor_id();
653 if (table->ents[index] != cpu)
654 table->ents[index] = cpu;
658 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
659 u32 hash)
661 if (table && hash)
662 table->ents[hash & table->mask] = RPS_NO_CPU;
665 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
667 #ifdef CONFIG_RFS_ACCEL
668 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
669 u16 filter_id);
670 #endif
672 /* This structure contains an instance of an RX queue. */
673 struct netdev_rx_queue {
674 struct rps_map __rcu *rps_map;
675 struct rps_dev_flow_table __rcu *rps_flow_table;
676 struct kobject kobj;
677 struct net_device *dev;
678 } ____cacheline_aligned_in_smp;
679 #endif /* CONFIG_RPS */
681 #ifdef CONFIG_XPS
683 * This structure holds an XPS map which can be of variable length. The
684 * map is an array of queues.
686 struct xps_map {
687 unsigned int len;
688 unsigned int alloc_len;
689 struct rcu_head rcu;
690 u16 queues[0];
692 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
693 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
694 / sizeof(u16))
697 * This structure holds all XPS maps for device. Maps are indexed by CPU.
699 struct xps_dev_maps {
700 struct rcu_head rcu;
701 struct xps_map __rcu *cpu_map[0];
703 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
704 (nr_cpu_ids * sizeof(struct xps_map *)))
705 #endif /* CONFIG_XPS */
707 #define TC_MAX_QUEUE 16
708 #define TC_BITMASK 15
709 /* HW offloaded queuing disciplines txq count and offset maps */
710 struct netdev_tc_txq {
711 u16 count;
712 u16 offset;
715 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
717 * This structure is to hold information about the device
718 * configured to run FCoE protocol stack.
720 struct netdev_fcoe_hbainfo {
721 char manufacturer[64];
722 char serial_number[64];
723 char hardware_version[64];
724 char driver_version[64];
725 char optionrom_version[64];
726 char firmware_version[64];
727 char model[256];
728 char model_description[256];
730 #endif
732 #define MAX_PHYS_PORT_ID_LEN 32
734 /* This structure holds a unique identifier to identify the
735 * physical port used by a netdevice.
737 struct netdev_phys_port_id {
738 unsigned char id[MAX_PHYS_PORT_ID_LEN];
739 unsigned char id_len;
743 * This structure defines the management hooks for network devices.
744 * The following hooks can be defined; unless noted otherwise, they are
745 * optional and can be filled with a null pointer.
747 * int (*ndo_init)(struct net_device *dev);
748 * This function is called once when network device is registered.
749 * The network device can use this to any late stage initializaton
750 * or semantic validattion. It can fail with an error code which will
751 * be propogated back to register_netdev
753 * void (*ndo_uninit)(struct net_device *dev);
754 * This function is called when device is unregistered or when registration
755 * fails. It is not called if init fails.
757 * int (*ndo_open)(struct net_device *dev);
758 * This function is called when network device transistions to the up
759 * state.
761 * int (*ndo_stop)(struct net_device *dev);
762 * This function is called when network device transistions to the down
763 * state.
765 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
766 * struct net_device *dev);
767 * Called when a packet needs to be transmitted.
768 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
769 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
770 * Required can not be NULL.
772 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
773 * Called to decide which queue to when device supports multiple
774 * transmit queues.
776 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
777 * This function is called to allow device receiver to make
778 * changes to configuration when multicast or promiscious is enabled.
780 * void (*ndo_set_rx_mode)(struct net_device *dev);
781 * This function is called device changes address list filtering.
782 * If driver handles unicast address filtering, it should set
783 * IFF_UNICAST_FLT to its priv_flags.
785 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
786 * This function is called when the Media Access Control address
787 * needs to be changed. If this interface is not defined, the
788 * mac address can not be changed.
790 * int (*ndo_validate_addr)(struct net_device *dev);
791 * Test if Media Access Control address is valid for the device.
793 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
794 * Called when a user request an ioctl which can't be handled by
795 * the generic interface code. If not defined ioctl's return
796 * not supported error code.
798 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
799 * Used to set network devices bus interface parameters. This interface
800 * is retained for legacy reason, new devices should use the bus
801 * interface (PCI) for low level management.
803 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
804 * Called when a user wants to change the Maximum Transfer Unit
805 * of a device. If not defined, any request to change MTU will
806 * will return an error.
808 * void (*ndo_tx_timeout)(struct net_device *dev);
809 * Callback uses when the transmitter has not made any progress
810 * for dev->watchdog ticks.
812 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
813 * struct rtnl_link_stats64 *storage);
814 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
815 * Called when a user wants to get the network device usage
816 * statistics. Drivers must do one of the following:
817 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
818 * rtnl_link_stats64 structure passed by the caller.
819 * 2. Define @ndo_get_stats to update a net_device_stats structure
820 * (which should normally be dev->stats) and return a pointer to
821 * it. The structure may be changed asynchronously only if each
822 * field is written atomically.
823 * 3. Update dev->stats asynchronously and atomically, and define
824 * neither operation.
826 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
827 * If device support VLAN filtering this function is called when a
828 * VLAN id is registered.
830 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
831 * If device support VLAN filtering this function is called when a
832 * VLAN id is unregistered.
834 * void (*ndo_poll_controller)(struct net_device *dev);
836 * SR-IOV management functions.
837 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
838 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
839 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
840 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
841 * int (*ndo_get_vf_config)(struct net_device *dev,
842 * int vf, struct ifla_vf_info *ivf);
843 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
844 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
845 * struct nlattr *port[]);
846 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
847 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
848 * Called to setup 'tc' number of traffic classes in the net device. This
849 * is always called from the stack with the rtnl lock held and netif tx
850 * queues stopped. This allows the netdevice to perform queue management
851 * safely.
853 * Fiber Channel over Ethernet (FCoE) offload functions.
854 * int (*ndo_fcoe_enable)(struct net_device *dev);
855 * Called when the FCoE protocol stack wants to start using LLD for FCoE
856 * so the underlying device can perform whatever needed configuration or
857 * initialization to support acceleration of FCoE traffic.
859 * int (*ndo_fcoe_disable)(struct net_device *dev);
860 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
861 * so the underlying device can perform whatever needed clean-ups to
862 * stop supporting acceleration of FCoE traffic.
864 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
865 * struct scatterlist *sgl, unsigned int sgc);
866 * Called when the FCoE Initiator wants to initialize an I/O that
867 * is a possible candidate for Direct Data Placement (DDP). The LLD can
868 * perform necessary setup and returns 1 to indicate the device is set up
869 * successfully to perform DDP on this I/O, otherwise this returns 0.
871 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
872 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
873 * indicated by the FC exchange id 'xid', so the underlying device can
874 * clean up and reuse resources for later DDP requests.
876 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
877 * struct scatterlist *sgl, unsigned int sgc);
878 * Called when the FCoE Target wants to initialize an I/O that
879 * is a possible candidate for Direct Data Placement (DDP). The LLD can
880 * perform necessary setup and returns 1 to indicate the device is set up
881 * successfully to perform DDP on this I/O, otherwise this returns 0.
883 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
884 * struct netdev_fcoe_hbainfo *hbainfo);
885 * Called when the FCoE Protocol stack wants information on the underlying
886 * device. This information is utilized by the FCoE protocol stack to
887 * register attributes with Fiber Channel management service as per the
888 * FC-GS Fabric Device Management Information(FDMI) specification.
890 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
891 * Called when the underlying device wants to override default World Wide
892 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
893 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
894 * protocol stack to use.
896 * RFS acceleration.
897 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
898 * u16 rxq_index, u32 flow_id);
899 * Set hardware filter for RFS. rxq_index is the target queue index;
900 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
901 * Return the filter ID on success, or a negative error code.
903 * Slave management functions (for bridge, bonding, etc).
904 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
905 * Called to make another netdev an underling.
907 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
908 * Called to release previously enslaved netdev.
910 * Feature/offload setting functions.
911 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
912 * netdev_features_t features);
913 * Adjusts the requested feature flags according to device-specific
914 * constraints, and returns the resulting flags. Must not modify
915 * the device state.
917 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
918 * Called to update device configuration to new features. Passed
919 * feature set might be less than what was returned by ndo_fix_features()).
920 * Must return >0 or -errno if it changed dev->features itself.
922 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
923 * struct net_device *dev,
924 * const unsigned char *addr, u16 flags)
925 * Adds an FDB entry to dev for addr.
926 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
927 * struct net_device *dev,
928 * const unsigned char *addr)
929 * Deletes the FDB entry from dev coresponding to addr.
930 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
931 * struct net_device *dev, int idx)
932 * Used to add FDB entries to dump requests. Implementers should add
933 * entries to skb and update idx with the number of entries.
935 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
936 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
937 * struct net_device *dev, u32 filter_mask)
939 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
940 * Called to change device carrier. Soft-devices (like dummy, team, etc)
941 * which do not represent real hardware may define this to allow their
942 * userspace components to manage their virtual carrier state. Devices
943 * that determine carrier state from physical hardware properties (eg
944 * network cables) or protocol-dependent mechanisms (eg
945 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
947 * int (*ndo_get_phys_port_id)(struct net_device *dev,
948 * struct netdev_phys_port_id *ppid);
949 * Called to get ID of physical port of this device. If driver does
950 * not implement this, it is assumed that the hw is not able to have
951 * multiple net devices on single physical port.
953 * void (*ndo_add_vxlan_port)(struct net_device *dev,
954 * sa_family_t sa_family, __be16 port);
955 * Called by vxlan to notiy a driver about the UDP port and socket
956 * address family that vxlan is listnening to. It is called only when
957 * a new port starts listening. The operation is protected by the
958 * vxlan_net->sock_lock.
960 * void (*ndo_del_vxlan_port)(struct net_device *dev,
961 * sa_family_t sa_family, __be16 port);
962 * Called by vxlan to notify the driver about a UDP port and socket
963 * address family that vxlan is not listening to anymore. The operation
964 * is protected by the vxlan_net->sock_lock.
966 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
967 * struct net_device *dev)
968 * Called by upper layer devices to accelerate switching or other
969 * station functionality into hardware. 'pdev is the lowerdev
970 * to use for the offload and 'dev' is the net device that will
971 * back the offload. Returns a pointer to the private structure
972 * the upper layer will maintain.
973 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
974 * Called by upper layer device to delete the station created
975 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
976 * the station and priv is the structure returned by the add
977 * operation.
978 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
979 * struct net_device *dev,
980 * void *priv);
981 * Callback to use for xmit over the accelerated station. This
982 * is used in place of ndo_start_xmit on accelerated net
983 * devices.
985 struct net_device_ops {
986 int (*ndo_init)(struct net_device *dev);
987 void (*ndo_uninit)(struct net_device *dev);
988 int (*ndo_open)(struct net_device *dev);
989 int (*ndo_stop)(struct net_device *dev);
990 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
991 struct net_device *dev);
992 u16 (*ndo_select_queue)(struct net_device *dev,
993 struct sk_buff *skb);
994 void (*ndo_change_rx_flags)(struct net_device *dev,
995 int flags);
996 void (*ndo_set_rx_mode)(struct net_device *dev);
997 int (*ndo_set_mac_address)(struct net_device *dev,
998 void *addr);
999 int (*ndo_validate_addr)(struct net_device *dev);
1000 int (*ndo_do_ioctl)(struct net_device *dev,
1001 struct ifreq *ifr, int cmd);
1002 int (*ndo_set_config)(struct net_device *dev,
1003 struct ifmap *map);
1004 int (*ndo_change_mtu)(struct net_device *dev,
1005 int new_mtu);
1006 int (*ndo_neigh_setup)(struct net_device *dev,
1007 struct neigh_parms *);
1008 void (*ndo_tx_timeout) (struct net_device *dev);
1010 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1011 struct rtnl_link_stats64 *storage);
1012 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1014 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1015 __be16 proto, u16 vid);
1016 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1017 __be16 proto, u16 vid);
1018 #ifdef CONFIG_NET_POLL_CONTROLLER
1019 void (*ndo_poll_controller)(struct net_device *dev);
1020 int (*ndo_netpoll_setup)(struct net_device *dev,
1021 struct netpoll_info *info,
1022 gfp_t gfp);
1023 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1024 #endif
1025 #ifdef CONFIG_NET_RX_BUSY_POLL
1026 int (*ndo_busy_poll)(struct napi_struct *dev);
1027 #endif
1028 int (*ndo_set_vf_mac)(struct net_device *dev,
1029 int queue, u8 *mac);
1030 int (*ndo_set_vf_vlan)(struct net_device *dev,
1031 int queue, u16 vlan, u8 qos);
1032 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
1033 int vf, int rate);
1034 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1035 int vf, bool setting);
1036 int (*ndo_get_vf_config)(struct net_device *dev,
1037 int vf,
1038 struct ifla_vf_info *ivf);
1039 int (*ndo_set_vf_link_state)(struct net_device *dev,
1040 int vf, int link_state);
1041 int (*ndo_set_vf_port)(struct net_device *dev,
1042 int vf,
1043 struct nlattr *port[]);
1044 int (*ndo_get_vf_port)(struct net_device *dev,
1045 int vf, struct sk_buff *skb);
1046 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1047 #if IS_ENABLED(CONFIG_FCOE)
1048 int (*ndo_fcoe_enable)(struct net_device *dev);
1049 int (*ndo_fcoe_disable)(struct net_device *dev);
1050 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1051 u16 xid,
1052 struct scatterlist *sgl,
1053 unsigned int sgc);
1054 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1055 u16 xid);
1056 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1057 u16 xid,
1058 struct scatterlist *sgl,
1059 unsigned int sgc);
1060 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1061 struct netdev_fcoe_hbainfo *hbainfo);
1062 #endif
1064 #if IS_ENABLED(CONFIG_LIBFCOE)
1065 #define NETDEV_FCOE_WWNN 0
1066 #define NETDEV_FCOE_WWPN 1
1067 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1068 u64 *wwn, int type);
1069 #endif
1071 #ifdef CONFIG_RFS_ACCEL
1072 int (*ndo_rx_flow_steer)(struct net_device *dev,
1073 const struct sk_buff *skb,
1074 u16 rxq_index,
1075 u32 flow_id);
1076 #endif
1077 int (*ndo_add_slave)(struct net_device *dev,
1078 struct net_device *slave_dev);
1079 int (*ndo_del_slave)(struct net_device *dev,
1080 struct net_device *slave_dev);
1081 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1082 netdev_features_t features);
1083 int (*ndo_set_features)(struct net_device *dev,
1084 netdev_features_t features);
1085 int (*ndo_neigh_construct)(struct neighbour *n);
1086 void (*ndo_neigh_destroy)(struct neighbour *n);
1088 int (*ndo_fdb_add)(struct ndmsg *ndm,
1089 struct nlattr *tb[],
1090 struct net_device *dev,
1091 const unsigned char *addr,
1092 u16 flags);
1093 int (*ndo_fdb_del)(struct ndmsg *ndm,
1094 struct nlattr *tb[],
1095 struct net_device *dev,
1096 const unsigned char *addr);
1097 int (*ndo_fdb_dump)(struct sk_buff *skb,
1098 struct netlink_callback *cb,
1099 struct net_device *dev,
1100 int idx);
1102 int (*ndo_bridge_setlink)(struct net_device *dev,
1103 struct nlmsghdr *nlh);
1104 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1105 u32 pid, u32 seq,
1106 struct net_device *dev,
1107 u32 filter_mask);
1108 int (*ndo_bridge_dellink)(struct net_device *dev,
1109 struct nlmsghdr *nlh);
1110 int (*ndo_change_carrier)(struct net_device *dev,
1111 bool new_carrier);
1112 int (*ndo_get_phys_port_id)(struct net_device *dev,
1113 struct netdev_phys_port_id *ppid);
1114 void (*ndo_add_vxlan_port)(struct net_device *dev,
1115 sa_family_t sa_family,
1116 __be16 port);
1117 void (*ndo_del_vxlan_port)(struct net_device *dev,
1118 sa_family_t sa_family,
1119 __be16 port);
1121 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1122 struct net_device *dev);
1123 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1124 void *priv);
1126 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1127 struct net_device *dev,
1128 void *priv);
1132 * The DEVICE structure.
1133 * Actually, this whole structure is a big mistake. It mixes I/O
1134 * data with strictly "high-level" data, and it has to know about
1135 * almost every data structure used in the INET module.
1137 * FIXME: cleanup struct net_device such that network protocol info
1138 * moves out.
1141 struct net_device {
1144 * This is the first field of the "visible" part of this structure
1145 * (i.e. as seen by users in the "Space.c" file). It is the name
1146 * of the interface.
1148 char name[IFNAMSIZ];
1150 /* device name hash chain, please keep it close to name[] */
1151 struct hlist_node name_hlist;
1153 /* snmp alias */
1154 char *ifalias;
1157 * I/O specific fields
1158 * FIXME: Merge these and struct ifmap into one
1160 unsigned long mem_end; /* shared mem end */
1161 unsigned long mem_start; /* shared mem start */
1162 unsigned long base_addr; /* device I/O address */
1163 int irq; /* device IRQ number */
1166 * Some hardware also needs these fields, but they are not
1167 * part of the usual set specified in Space.c.
1170 unsigned long state;
1172 struct list_head dev_list;
1173 struct list_head napi_list;
1174 struct list_head unreg_list;
1175 struct list_head close_list;
1177 /* directly linked devices, like slaves for bonding */
1178 struct {
1179 struct list_head upper;
1180 struct list_head lower;
1181 } adj_list;
1183 /* all linked devices, *including* neighbours */
1184 struct {
1185 struct list_head upper;
1186 struct list_head lower;
1187 } all_adj_list;
1190 /* currently active device features */
1191 netdev_features_t features;
1192 /* user-changeable features */
1193 netdev_features_t hw_features;
1194 /* user-requested features */
1195 netdev_features_t wanted_features;
1196 /* mask of features inheritable by VLAN devices */
1197 netdev_features_t vlan_features;
1198 /* mask of features inherited by encapsulating devices
1199 * This field indicates what encapsulation offloads
1200 * the hardware is capable of doing, and drivers will
1201 * need to set them appropriately.
1203 netdev_features_t hw_enc_features;
1204 /* mask of fetures inheritable by MPLS */
1205 netdev_features_t mpls_features;
1207 /* Interface index. Unique device identifier */
1208 int ifindex;
1209 int iflink;
1211 struct net_device_stats stats;
1212 atomic_long_t rx_dropped; /* dropped packets by core network
1213 * Do not use this in drivers.
1216 #ifdef CONFIG_WIRELESS_EXT
1217 /* List of functions to handle Wireless Extensions (instead of ioctl).
1218 * See <net/iw_handler.h> for details. Jean II */
1219 const struct iw_handler_def * wireless_handlers;
1220 /* Instance data managed by the core of Wireless Extensions. */
1221 struct iw_public_data * wireless_data;
1222 #endif
1223 /* Management operations */
1224 const struct net_device_ops *netdev_ops;
1225 const struct ethtool_ops *ethtool_ops;
1226 const struct forwarding_accel_ops *fwd_ops;
1228 /* Hardware header description */
1229 const struct header_ops *header_ops;
1231 unsigned int flags; /* interface flags (a la BSD) */
1232 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1233 * See if.h for definitions. */
1234 unsigned short gflags;
1235 unsigned short padded; /* How much padding added by alloc_netdev() */
1237 unsigned char operstate; /* RFC2863 operstate */
1238 unsigned char link_mode; /* mapping policy to operstate */
1240 unsigned char if_port; /* Selectable AUI, TP,..*/
1241 unsigned char dma; /* DMA channel */
1243 unsigned int mtu; /* interface MTU value */
1244 unsigned short type; /* interface hardware type */
1245 unsigned short hard_header_len; /* hardware hdr length */
1247 /* extra head- and tailroom the hardware may need, but not in all cases
1248 * can this be guaranteed, especially tailroom. Some cases also use
1249 * LL_MAX_HEADER instead to allocate the skb.
1251 unsigned short needed_headroom;
1252 unsigned short needed_tailroom;
1254 /* Interface address info. */
1255 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1256 unsigned char addr_assign_type; /* hw address assignment type */
1257 unsigned char addr_len; /* hardware address length */
1258 unsigned char neigh_priv_len;
1259 unsigned short dev_id; /* Used to differentiate devices
1260 * that share the same link
1261 * layer address
1263 spinlock_t addr_list_lock;
1264 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1265 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1266 struct netdev_hw_addr_list dev_addrs; /* list of device
1267 * hw addresses
1269 #ifdef CONFIG_SYSFS
1270 struct kset *queues_kset;
1271 #endif
1273 bool uc_promisc;
1274 unsigned int promiscuity;
1275 unsigned int allmulti;
1278 /* Protocol specific pointers */
1280 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1281 struct vlan_info __rcu *vlan_info; /* VLAN info */
1282 #endif
1283 #if IS_ENABLED(CONFIG_NET_DSA)
1284 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1285 #endif
1286 void *atalk_ptr; /* AppleTalk link */
1287 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1288 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1289 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1290 void *ax25_ptr; /* AX.25 specific data */
1291 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1292 assign before registering */
1295 * Cache lines mostly used on receive path (including eth_type_trans())
1297 unsigned long last_rx; /* Time of last Rx
1298 * This should not be set in
1299 * drivers, unless really needed,
1300 * because network stack (bonding)
1301 * use it if/when necessary, to
1302 * avoid dirtying this cache line.
1305 /* Interface address info used in eth_type_trans() */
1306 unsigned char *dev_addr; /* hw address, (before bcast
1307 because most packets are
1308 unicast) */
1311 #ifdef CONFIG_RPS
1312 struct netdev_rx_queue *_rx;
1314 /* Number of RX queues allocated at register_netdev() time */
1315 unsigned int num_rx_queues;
1317 /* Number of RX queues currently active in device */
1318 unsigned int real_num_rx_queues;
1320 #endif
1322 rx_handler_func_t __rcu *rx_handler;
1323 void __rcu *rx_handler_data;
1325 struct netdev_queue __rcu *ingress_queue;
1326 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1330 * Cache lines mostly used on transmit path
1332 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1334 /* Number of TX queues allocated at alloc_netdev_mq() time */
1335 unsigned int num_tx_queues;
1337 /* Number of TX queues currently active in device */
1338 unsigned int real_num_tx_queues;
1340 /* root qdisc from userspace point of view */
1341 struct Qdisc *qdisc;
1343 unsigned long tx_queue_len; /* Max frames per queue allowed */
1344 spinlock_t tx_global_lock;
1346 #ifdef CONFIG_XPS
1347 struct xps_dev_maps __rcu *xps_maps;
1348 #endif
1349 #ifdef CONFIG_RFS_ACCEL
1350 /* CPU reverse-mapping for RX completion interrupts, indexed
1351 * by RX queue number. Assigned by driver. This must only be
1352 * set if the ndo_rx_flow_steer operation is defined. */
1353 struct cpu_rmap *rx_cpu_rmap;
1354 #endif
1356 /* These may be needed for future network-power-down code. */
1359 * trans_start here is expensive for high speed devices on SMP,
1360 * please use netdev_queue->trans_start instead.
1362 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1364 int watchdog_timeo; /* used by dev_watchdog() */
1365 struct timer_list watchdog_timer;
1367 /* Number of references to this device */
1368 int __percpu *pcpu_refcnt;
1370 /* delayed register/unregister */
1371 struct list_head todo_list;
1372 /* device index hash chain */
1373 struct hlist_node index_hlist;
1375 struct list_head link_watch_list;
1377 /* register/unregister state machine */
1378 enum { NETREG_UNINITIALIZED=0,
1379 NETREG_REGISTERED, /* completed register_netdevice */
1380 NETREG_UNREGISTERING, /* called unregister_netdevice */
1381 NETREG_UNREGISTERED, /* completed unregister todo */
1382 NETREG_RELEASED, /* called free_netdev */
1383 NETREG_DUMMY, /* dummy device for NAPI poll */
1384 } reg_state:8;
1386 bool dismantle; /* device is going do be freed */
1388 enum {
1389 RTNL_LINK_INITIALIZED,
1390 RTNL_LINK_INITIALIZING,
1391 } rtnl_link_state:16;
1393 /* Called from unregister, can be used to call free_netdev */
1394 void (*destructor)(struct net_device *dev);
1396 #ifdef CONFIG_NETPOLL
1397 struct netpoll_info __rcu *npinfo;
1398 #endif
1400 #ifdef CONFIG_NET_NS
1401 /* Network namespace this network device is inside */
1402 struct net *nd_net;
1403 #endif
1405 /* mid-layer private */
1406 union {
1407 void *ml_priv;
1408 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1409 struct pcpu_tstats __percpu *tstats; /* tunnel stats */
1410 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1411 struct pcpu_vstats __percpu *vstats; /* veth stats */
1413 /* GARP */
1414 struct garp_port __rcu *garp_port;
1415 /* MRP */
1416 struct mrp_port __rcu *mrp_port;
1418 /* class/net/name entry */
1419 struct device dev;
1420 /* space for optional device, statistics, and wireless sysfs groups */
1421 const struct attribute_group *sysfs_groups[4];
1423 /* rtnetlink link ops */
1424 const struct rtnl_link_ops *rtnl_link_ops;
1426 /* for setting kernel sock attribute on TCP connection setup */
1427 #define GSO_MAX_SIZE 65536
1428 unsigned int gso_max_size;
1429 #define GSO_MAX_SEGS 65535
1430 u16 gso_max_segs;
1432 #ifdef CONFIG_DCB
1433 /* Data Center Bridging netlink ops */
1434 const struct dcbnl_rtnl_ops *dcbnl_ops;
1435 #endif
1436 u8 num_tc;
1437 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1438 u8 prio_tc_map[TC_BITMASK + 1];
1440 #if IS_ENABLED(CONFIG_FCOE)
1441 /* max exchange id for FCoE LRO by ddp */
1442 unsigned int fcoe_ddp_xid;
1443 #endif
1444 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
1445 struct netprio_map __rcu *priomap;
1446 #endif
1447 /* phy device may attach itself for hardware timestamping */
1448 struct phy_device *phydev;
1450 struct lock_class_key *qdisc_tx_busylock;
1452 /* group the device belongs to */
1453 int group;
1455 struct pm_qos_request pm_qos_req;
1457 #define to_net_dev(d) container_of(d, struct net_device, dev)
1459 #define NETDEV_ALIGN 32
1461 static inline
1462 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1464 return dev->prio_tc_map[prio & TC_BITMASK];
1467 static inline
1468 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1470 if (tc >= dev->num_tc)
1471 return -EINVAL;
1473 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1474 return 0;
1477 static inline
1478 void netdev_reset_tc(struct net_device *dev)
1480 dev->num_tc = 0;
1481 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1482 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1485 static inline
1486 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1488 if (tc >= dev->num_tc)
1489 return -EINVAL;
1491 dev->tc_to_txq[tc].count = count;
1492 dev->tc_to_txq[tc].offset = offset;
1493 return 0;
1496 static inline
1497 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1499 if (num_tc > TC_MAX_QUEUE)
1500 return -EINVAL;
1502 dev->num_tc = num_tc;
1503 return 0;
1506 static inline
1507 int netdev_get_num_tc(struct net_device *dev)
1509 return dev->num_tc;
1512 static inline
1513 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1514 unsigned int index)
1516 return &dev->_tx[index];
1519 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1520 void (*f)(struct net_device *,
1521 struct netdev_queue *,
1522 void *),
1523 void *arg)
1525 unsigned int i;
1527 for (i = 0; i < dev->num_tx_queues; i++)
1528 f(dev, &dev->_tx[i], arg);
1531 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1532 struct sk_buff *skb);
1533 u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb);
1536 * Net namespace inlines
1538 static inline
1539 struct net *dev_net(const struct net_device *dev)
1541 return read_pnet(&dev->nd_net);
1544 static inline
1545 void dev_net_set(struct net_device *dev, struct net *net)
1547 #ifdef CONFIG_NET_NS
1548 release_net(dev->nd_net);
1549 dev->nd_net = hold_net(net);
1550 #endif
1553 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1555 #ifdef CONFIG_NET_DSA_TAG_DSA
1556 if (dev->dsa_ptr != NULL)
1557 return dsa_uses_dsa_tags(dev->dsa_ptr);
1558 #endif
1560 return 0;
1563 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1565 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1566 if (dev->dsa_ptr != NULL)
1567 return dsa_uses_trailer_tags(dev->dsa_ptr);
1568 #endif
1570 return 0;
1574 * netdev_priv - access network device private data
1575 * @dev: network device
1577 * Get network device private data
1579 static inline void *netdev_priv(const struct net_device *dev)
1581 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1584 /* Set the sysfs physical device reference for the network logical device
1585 * if set prior to registration will cause a symlink during initialization.
1587 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1589 /* Set the sysfs device type for the network logical device to allow
1590 * fine-grained identification of different network device types. For
1591 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1593 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1595 /* Default NAPI poll() weight
1596 * Device drivers are strongly advised to not use bigger value
1598 #define NAPI_POLL_WEIGHT 64
1601 * netif_napi_add - initialize a napi context
1602 * @dev: network device
1603 * @napi: napi context
1604 * @poll: polling function
1605 * @weight: default weight
1607 * netif_napi_add() must be used to initialize a napi context prior to calling
1608 * *any* of the other napi related functions.
1610 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1611 int (*poll)(struct napi_struct *, int), int weight);
1614 * netif_napi_del - remove a napi context
1615 * @napi: napi context
1617 * netif_napi_del() removes a napi context from the network device napi list
1619 void netif_napi_del(struct napi_struct *napi);
1621 struct napi_gro_cb {
1622 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1623 void *frag0;
1625 /* Length of frag0. */
1626 unsigned int frag0_len;
1628 /* This indicates where we are processing relative to skb->data. */
1629 int data_offset;
1631 /* This is non-zero if the packet cannot be merged with the new skb. */
1632 int flush;
1634 /* Number of segments aggregated. */
1635 u16 count;
1637 /* This is non-zero if the packet may be of the same flow. */
1638 u8 same_flow;
1640 /* Free the skb? */
1641 u8 free;
1642 #define NAPI_GRO_FREE 1
1643 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1645 /* jiffies when first packet was created/queued */
1646 unsigned long age;
1648 /* Used in ipv6_gro_receive() */
1649 int proto;
1651 /* used in skb_gro_receive() slow path */
1652 struct sk_buff *last;
1655 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1657 struct packet_type {
1658 __be16 type; /* This is really htons(ether_type). */
1659 struct net_device *dev; /* NULL is wildcarded here */
1660 int (*func) (struct sk_buff *,
1661 struct net_device *,
1662 struct packet_type *,
1663 struct net_device *);
1664 bool (*id_match)(struct packet_type *ptype,
1665 struct sock *sk);
1666 void *af_packet_priv;
1667 struct list_head list;
1670 struct offload_callbacks {
1671 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1672 netdev_features_t features);
1673 int (*gso_send_check)(struct sk_buff *skb);
1674 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1675 struct sk_buff *skb);
1676 int (*gro_complete)(struct sk_buff *skb);
1679 struct packet_offload {
1680 __be16 type; /* This is really htons(ether_type). */
1681 struct offload_callbacks callbacks;
1682 struct list_head list;
1685 #include <linux/notifier.h>
1687 /* netdevice notifier chain. Please remember to update the rtnetlink
1688 * notification exclusion list in rtnetlink_event() when adding new
1689 * types.
1691 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1692 #define NETDEV_DOWN 0x0002
1693 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1694 detected a hardware crash and restarted
1695 - we can use this eg to kick tcp sessions
1696 once done */
1697 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1698 #define NETDEV_REGISTER 0x0005
1699 #define NETDEV_UNREGISTER 0x0006
1700 #define NETDEV_CHANGEMTU 0x0007
1701 #define NETDEV_CHANGEADDR 0x0008
1702 #define NETDEV_GOING_DOWN 0x0009
1703 #define NETDEV_CHANGENAME 0x000A
1704 #define NETDEV_FEAT_CHANGE 0x000B
1705 #define NETDEV_BONDING_FAILOVER 0x000C
1706 #define NETDEV_PRE_UP 0x000D
1707 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1708 #define NETDEV_POST_TYPE_CHANGE 0x000F
1709 #define NETDEV_POST_INIT 0x0010
1710 #define NETDEV_UNREGISTER_FINAL 0x0011
1711 #define NETDEV_RELEASE 0x0012
1712 #define NETDEV_NOTIFY_PEERS 0x0013
1713 #define NETDEV_JOIN 0x0014
1714 #define NETDEV_CHANGEUPPER 0x0015
1715 #define NETDEV_RESEND_IGMP 0x0016
1717 int register_netdevice_notifier(struct notifier_block *nb);
1718 int unregister_netdevice_notifier(struct notifier_block *nb);
1720 struct netdev_notifier_info {
1721 struct net_device *dev;
1724 struct netdev_notifier_change_info {
1725 struct netdev_notifier_info info; /* must be first */
1726 unsigned int flags_changed;
1729 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1730 struct net_device *dev)
1732 info->dev = dev;
1735 static inline struct net_device *
1736 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1738 return info->dev;
1741 int call_netdevice_notifiers_info(unsigned long val, struct net_device *dev,
1742 struct netdev_notifier_info *info);
1743 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1746 extern rwlock_t dev_base_lock; /* Device list lock */
1748 #define for_each_netdev(net, d) \
1749 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1750 #define for_each_netdev_reverse(net, d) \
1751 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1752 #define for_each_netdev_rcu(net, d) \
1753 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1754 #define for_each_netdev_safe(net, d, n) \
1755 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1756 #define for_each_netdev_continue(net, d) \
1757 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1758 #define for_each_netdev_continue_rcu(net, d) \
1759 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1760 #define for_each_netdev_in_bond_rcu(bond, slave) \
1761 for_each_netdev_rcu(&init_net, slave) \
1762 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1763 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1765 static inline struct net_device *next_net_device(struct net_device *dev)
1767 struct list_head *lh;
1768 struct net *net;
1770 net = dev_net(dev);
1771 lh = dev->dev_list.next;
1772 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1775 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1777 struct list_head *lh;
1778 struct net *net;
1780 net = dev_net(dev);
1781 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1782 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1785 static inline struct net_device *first_net_device(struct net *net)
1787 return list_empty(&net->dev_base_head) ? NULL :
1788 net_device_entry(net->dev_base_head.next);
1791 static inline struct net_device *first_net_device_rcu(struct net *net)
1793 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1795 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1798 int netdev_boot_setup_check(struct net_device *dev);
1799 unsigned long netdev_boot_base(const char *prefix, int unit);
1800 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1801 const char *hwaddr);
1802 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1803 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1804 void dev_add_pack(struct packet_type *pt);
1805 void dev_remove_pack(struct packet_type *pt);
1806 void __dev_remove_pack(struct packet_type *pt);
1807 void dev_add_offload(struct packet_offload *po);
1808 void dev_remove_offload(struct packet_offload *po);
1809 void __dev_remove_offload(struct packet_offload *po);
1811 struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1812 unsigned short mask);
1813 struct net_device *dev_get_by_name(struct net *net, const char *name);
1814 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1815 struct net_device *__dev_get_by_name(struct net *net, const char *name);
1816 int dev_alloc_name(struct net_device *dev, const char *name);
1817 int dev_open(struct net_device *dev);
1818 int dev_close(struct net_device *dev);
1819 void dev_disable_lro(struct net_device *dev);
1820 int dev_loopback_xmit(struct sk_buff *newskb);
1821 int dev_queue_xmit(struct sk_buff *skb);
1822 int register_netdevice(struct net_device *dev);
1823 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
1824 void unregister_netdevice_many(struct list_head *head);
1825 static inline void unregister_netdevice(struct net_device *dev)
1827 unregister_netdevice_queue(dev, NULL);
1830 int netdev_refcnt_read(const struct net_device *dev);
1831 void free_netdev(struct net_device *dev);
1832 void netdev_freemem(struct net_device *dev);
1833 void synchronize_net(void);
1834 int init_dummy_netdev(struct net_device *dev);
1836 struct net_device *dev_get_by_index(struct net *net, int ifindex);
1837 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1838 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1839 int netdev_get_name(struct net *net, char *name, int ifindex);
1840 int dev_restart(struct net_device *dev);
1841 #ifdef CONFIG_NETPOLL_TRAP
1842 int netpoll_trap(void);
1843 #endif
1844 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
1846 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1848 return NAPI_GRO_CB(skb)->data_offset;
1851 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1853 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1856 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1858 NAPI_GRO_CB(skb)->data_offset += len;
1861 static inline void *skb_gro_header_fast(struct sk_buff *skb,
1862 unsigned int offset)
1864 return NAPI_GRO_CB(skb)->frag0 + offset;
1867 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1869 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1872 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1873 unsigned int offset)
1875 if (!pskb_may_pull(skb, hlen))
1876 return NULL;
1878 NAPI_GRO_CB(skb)->frag0 = NULL;
1879 NAPI_GRO_CB(skb)->frag0_len = 0;
1880 return skb->data + offset;
1883 static inline void *skb_gro_mac_header(struct sk_buff *skb)
1885 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
1888 static inline void *skb_gro_network_header(struct sk_buff *skb)
1890 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1891 skb_network_offset(skb);
1894 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1895 unsigned short type,
1896 const void *daddr, const void *saddr,
1897 unsigned int len)
1899 if (!dev->header_ops || !dev->header_ops->create)
1900 return 0;
1902 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
1905 static inline int dev_parse_header(const struct sk_buff *skb,
1906 unsigned char *haddr)
1908 const struct net_device *dev = skb->dev;
1910 if (!dev->header_ops || !dev->header_ops->parse)
1911 return 0;
1912 return dev->header_ops->parse(skb, haddr);
1915 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
1916 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1917 static inline int unregister_gifconf(unsigned int family)
1919 return register_gifconf(family, NULL);
1922 #ifdef CONFIG_NET_FLOW_LIMIT
1923 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
1924 struct sd_flow_limit {
1925 u64 count;
1926 unsigned int num_buckets;
1927 unsigned int history_head;
1928 u16 history[FLOW_LIMIT_HISTORY];
1929 u8 buckets[];
1932 extern int netdev_flow_limit_table_len;
1933 #endif /* CONFIG_NET_FLOW_LIMIT */
1936 * Incoming packets are placed on per-cpu queues
1938 struct softnet_data {
1939 struct Qdisc *output_queue;
1940 struct Qdisc **output_queue_tailp;
1941 struct list_head poll_list;
1942 struct sk_buff *completion_queue;
1943 struct sk_buff_head process_queue;
1945 /* stats */
1946 unsigned int processed;
1947 unsigned int time_squeeze;
1948 unsigned int cpu_collision;
1949 unsigned int received_rps;
1951 #ifdef CONFIG_RPS
1952 struct softnet_data *rps_ipi_list;
1954 /* Elements below can be accessed between CPUs for RPS */
1955 struct call_single_data csd ____cacheline_aligned_in_smp;
1956 struct softnet_data *rps_ipi_next;
1957 unsigned int cpu;
1958 unsigned int input_queue_head;
1959 unsigned int input_queue_tail;
1960 #endif
1961 unsigned int dropped;
1962 struct sk_buff_head input_pkt_queue;
1963 struct napi_struct backlog;
1965 #ifdef CONFIG_NET_FLOW_LIMIT
1966 struct sd_flow_limit __rcu *flow_limit;
1967 #endif
1970 static inline void input_queue_head_incr(struct softnet_data *sd)
1972 #ifdef CONFIG_RPS
1973 sd->input_queue_head++;
1974 #endif
1977 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1978 unsigned int *qtail)
1980 #ifdef CONFIG_RPS
1981 *qtail = ++sd->input_queue_tail;
1982 #endif
1985 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1987 void __netif_schedule(struct Qdisc *q);
1989 static inline void netif_schedule_queue(struct netdev_queue *txq)
1991 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
1992 __netif_schedule(txq->qdisc);
1995 static inline void netif_tx_schedule_all(struct net_device *dev)
1997 unsigned int i;
1999 for (i = 0; i < dev->num_tx_queues; i++)
2000 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2003 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2005 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2009 * netif_start_queue - allow transmit
2010 * @dev: network device
2012 * Allow upper layers to call the device hard_start_xmit routine.
2014 static inline void netif_start_queue(struct net_device *dev)
2016 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2019 static inline void netif_tx_start_all_queues(struct net_device *dev)
2021 unsigned int i;
2023 for (i = 0; i < dev->num_tx_queues; i++) {
2024 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2025 netif_tx_start_queue(txq);
2029 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
2031 #ifdef CONFIG_NETPOLL_TRAP
2032 if (netpoll_trap()) {
2033 netif_tx_start_queue(dev_queue);
2034 return;
2036 #endif
2037 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
2038 __netif_schedule(dev_queue->qdisc);
2042 * netif_wake_queue - restart transmit
2043 * @dev: network device
2045 * Allow upper layers to call the device hard_start_xmit routine.
2046 * Used for flow control when transmit resources are available.
2048 static inline void netif_wake_queue(struct net_device *dev)
2050 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2053 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2055 unsigned int i;
2057 for (i = 0; i < dev->num_tx_queues; i++) {
2058 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2059 netif_tx_wake_queue(txq);
2063 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2065 if (WARN_ON(!dev_queue)) {
2066 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2067 return;
2069 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2073 * netif_stop_queue - stop transmitted packets
2074 * @dev: network device
2076 * Stop upper layers calling the device hard_start_xmit routine.
2077 * Used for flow control when transmit resources are unavailable.
2079 static inline void netif_stop_queue(struct net_device *dev)
2081 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2084 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2086 unsigned int i;
2088 for (i = 0; i < dev->num_tx_queues; i++) {
2089 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2090 netif_tx_stop_queue(txq);
2094 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2096 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2100 * netif_queue_stopped - test if transmit queue is flowblocked
2101 * @dev: network device
2103 * Test if transmit queue on device is currently unable to send.
2105 static inline bool netif_queue_stopped(const struct net_device *dev)
2107 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2110 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2112 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2115 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2117 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2120 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2121 unsigned int bytes)
2123 #ifdef CONFIG_BQL
2124 dql_queued(&dev_queue->dql, bytes);
2126 if (likely(dql_avail(&dev_queue->dql) >= 0))
2127 return;
2129 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2132 * The XOFF flag must be set before checking the dql_avail below,
2133 * because in netdev_tx_completed_queue we update the dql_completed
2134 * before checking the XOFF flag.
2136 smp_mb();
2138 /* check again in case another CPU has just made room avail */
2139 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2140 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2141 #endif
2145 * netdev_sent_queue - report the number of bytes queued to hardware
2146 * @dev: network device
2147 * @bytes: number of bytes queued to the hardware device queue
2149 * Report the number of bytes queued for sending/completion to the network
2150 * device hardware queue. @bytes should be a good approximation and should
2151 * exactly match netdev_completed_queue() @bytes
2153 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2155 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2158 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2159 unsigned int pkts, unsigned int bytes)
2161 #ifdef CONFIG_BQL
2162 if (unlikely(!bytes))
2163 return;
2165 dql_completed(&dev_queue->dql, bytes);
2168 * Without the memory barrier there is a small possiblity that
2169 * netdev_tx_sent_queue will miss the update and cause the queue to
2170 * be stopped forever
2172 smp_mb();
2174 if (dql_avail(&dev_queue->dql) < 0)
2175 return;
2177 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2178 netif_schedule_queue(dev_queue);
2179 #endif
2183 * netdev_completed_queue - report bytes and packets completed by device
2184 * @dev: network device
2185 * @pkts: actual number of packets sent over the medium
2186 * @bytes: actual number of bytes sent over the medium
2188 * Report the number of bytes and packets transmitted by the network device
2189 * hardware queue over the physical medium, @bytes must exactly match the
2190 * @bytes amount passed to netdev_sent_queue()
2192 static inline void netdev_completed_queue(struct net_device *dev,
2193 unsigned int pkts, unsigned int bytes)
2195 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2198 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2200 #ifdef CONFIG_BQL
2201 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2202 dql_reset(&q->dql);
2203 #endif
2207 * netdev_reset_queue - reset the packets and bytes count of a network device
2208 * @dev_queue: network device
2210 * Reset the bytes and packet count of a network device and clear the
2211 * software flow control OFF bit for this network device
2213 static inline void netdev_reset_queue(struct net_device *dev_queue)
2215 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2219 * netif_running - test if up
2220 * @dev: network device
2222 * Test if the device has been brought up.
2224 static inline bool netif_running(const struct net_device *dev)
2226 return test_bit(__LINK_STATE_START, &dev->state);
2230 * Routines to manage the subqueues on a device. We only need start
2231 * stop, and a check if it's stopped. All other device management is
2232 * done at the overall netdevice level.
2233 * Also test the device if we're multiqueue.
2237 * netif_start_subqueue - allow sending packets on subqueue
2238 * @dev: network device
2239 * @queue_index: sub queue index
2241 * Start individual transmit queue of a device with multiple transmit queues.
2243 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2245 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2247 netif_tx_start_queue(txq);
2251 * netif_stop_subqueue - stop sending packets on subqueue
2252 * @dev: network device
2253 * @queue_index: sub queue index
2255 * Stop individual transmit queue of a device with multiple transmit queues.
2257 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2259 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2260 #ifdef CONFIG_NETPOLL_TRAP
2261 if (netpoll_trap())
2262 return;
2263 #endif
2264 netif_tx_stop_queue(txq);
2268 * netif_subqueue_stopped - test status of subqueue
2269 * @dev: network device
2270 * @queue_index: sub queue index
2272 * Check individual transmit queue of a device with multiple transmit queues.
2274 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2275 u16 queue_index)
2277 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2279 return netif_tx_queue_stopped(txq);
2282 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2283 struct sk_buff *skb)
2285 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2289 * netif_wake_subqueue - allow sending packets on subqueue
2290 * @dev: network device
2291 * @queue_index: sub queue index
2293 * Resume individual transmit queue of a device with multiple transmit queues.
2295 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2297 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2298 #ifdef CONFIG_NETPOLL_TRAP
2299 if (netpoll_trap())
2300 return;
2301 #endif
2302 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2303 __netif_schedule(txq->qdisc);
2306 #ifdef CONFIG_XPS
2307 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2308 u16 index);
2309 #else
2310 static inline int netif_set_xps_queue(struct net_device *dev,
2311 const struct cpumask *mask,
2312 u16 index)
2314 return 0;
2316 #endif
2319 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2320 * as a distribution range limit for the returned value.
2322 static inline u16 skb_tx_hash(const struct net_device *dev,
2323 const struct sk_buff *skb)
2325 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2329 * netif_is_multiqueue - test if device has multiple transmit queues
2330 * @dev: network device
2332 * Check if device has multiple transmit queues
2334 static inline bool netif_is_multiqueue(const struct net_device *dev)
2336 return dev->num_tx_queues > 1;
2339 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2341 #ifdef CONFIG_RPS
2342 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2343 #else
2344 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2345 unsigned int rxq)
2347 return 0;
2349 #endif
2351 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2352 const struct net_device *from_dev)
2354 int err;
2356 err = netif_set_real_num_tx_queues(to_dev,
2357 from_dev->real_num_tx_queues);
2358 if (err)
2359 return err;
2360 #ifdef CONFIG_RPS
2361 return netif_set_real_num_rx_queues(to_dev,
2362 from_dev->real_num_rx_queues);
2363 #else
2364 return 0;
2365 #endif
2368 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2369 int netif_get_num_default_rss_queues(void);
2371 /* Use this variant when it is known for sure that it
2372 * is executing from hardware interrupt context or with hardware interrupts
2373 * disabled.
2375 void dev_kfree_skb_irq(struct sk_buff *skb);
2377 /* Use this variant in places where it could be invoked
2378 * from either hardware interrupt or other context, with hardware interrupts
2379 * either disabled or enabled.
2381 void dev_kfree_skb_any(struct sk_buff *skb);
2383 int netif_rx(struct sk_buff *skb);
2384 int netif_rx_ni(struct sk_buff *skb);
2385 int netif_receive_skb(struct sk_buff *skb);
2386 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2387 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2388 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2389 gro_result_t napi_gro_frags(struct napi_struct *napi);
2391 static inline void napi_free_frags(struct napi_struct *napi)
2393 kfree_skb(napi->skb);
2394 napi->skb = NULL;
2397 int netdev_rx_handler_register(struct net_device *dev,
2398 rx_handler_func_t *rx_handler,
2399 void *rx_handler_data);
2400 void netdev_rx_handler_unregister(struct net_device *dev);
2402 bool dev_valid_name(const char *name);
2403 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2404 int dev_ethtool(struct net *net, struct ifreq *);
2405 unsigned int dev_get_flags(const struct net_device *);
2406 int __dev_change_flags(struct net_device *, unsigned int flags);
2407 int dev_change_flags(struct net_device *, unsigned int);
2408 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2409 unsigned int gchanges);
2410 int dev_change_name(struct net_device *, const char *);
2411 int dev_set_alias(struct net_device *, const char *, size_t);
2412 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2413 int dev_set_mtu(struct net_device *, int);
2414 void dev_set_group(struct net_device *, int);
2415 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2416 int dev_change_carrier(struct net_device *, bool new_carrier);
2417 int dev_get_phys_port_id(struct net_device *dev,
2418 struct netdev_phys_port_id *ppid);
2419 int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2420 struct netdev_queue *txq, void *accel_priv);
2421 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2423 extern int netdev_budget;
2425 /* Called by rtnetlink.c:rtnl_unlock() */
2426 void netdev_run_todo(void);
2429 * dev_put - release reference to device
2430 * @dev: network device
2432 * Release reference to device to allow it to be freed.
2434 static inline void dev_put(struct net_device *dev)
2436 this_cpu_dec(*dev->pcpu_refcnt);
2440 * dev_hold - get reference to device
2441 * @dev: network device
2443 * Hold reference to device to keep it from being freed.
2445 static inline void dev_hold(struct net_device *dev)
2447 this_cpu_inc(*dev->pcpu_refcnt);
2450 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2451 * and _off may be called from IRQ context, but it is caller
2452 * who is responsible for serialization of these calls.
2454 * The name carrier is inappropriate, these functions should really be
2455 * called netif_lowerlayer_*() because they represent the state of any
2456 * kind of lower layer not just hardware media.
2459 void linkwatch_init_dev(struct net_device *dev);
2460 void linkwatch_fire_event(struct net_device *dev);
2461 void linkwatch_forget_dev(struct net_device *dev);
2464 * netif_carrier_ok - test if carrier present
2465 * @dev: network device
2467 * Check if carrier is present on device
2469 static inline bool netif_carrier_ok(const struct net_device *dev)
2471 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2474 unsigned long dev_trans_start(struct net_device *dev);
2476 void __netdev_watchdog_up(struct net_device *dev);
2478 void netif_carrier_on(struct net_device *dev);
2480 void netif_carrier_off(struct net_device *dev);
2483 * netif_dormant_on - mark device as dormant.
2484 * @dev: network device
2486 * Mark device as dormant (as per RFC2863).
2488 * The dormant state indicates that the relevant interface is not
2489 * actually in a condition to pass packets (i.e., it is not 'up') but is
2490 * in a "pending" state, waiting for some external event. For "on-
2491 * demand" interfaces, this new state identifies the situation where the
2492 * interface is waiting for events to place it in the up state.
2495 static inline void netif_dormant_on(struct net_device *dev)
2497 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2498 linkwatch_fire_event(dev);
2502 * netif_dormant_off - set device as not dormant.
2503 * @dev: network device
2505 * Device is not in dormant state.
2507 static inline void netif_dormant_off(struct net_device *dev)
2509 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2510 linkwatch_fire_event(dev);
2514 * netif_dormant - test if carrier present
2515 * @dev: network device
2517 * Check if carrier is present on device
2519 static inline bool netif_dormant(const struct net_device *dev)
2521 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2526 * netif_oper_up - test if device is operational
2527 * @dev: network device
2529 * Check if carrier is operational
2531 static inline bool netif_oper_up(const struct net_device *dev)
2533 return (dev->operstate == IF_OPER_UP ||
2534 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2538 * netif_device_present - is device available or removed
2539 * @dev: network device
2541 * Check if device has not been removed from system.
2543 static inline bool netif_device_present(struct net_device *dev)
2545 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2548 void netif_device_detach(struct net_device *dev);
2550 void netif_device_attach(struct net_device *dev);
2553 * Network interface message level settings
2556 enum {
2557 NETIF_MSG_DRV = 0x0001,
2558 NETIF_MSG_PROBE = 0x0002,
2559 NETIF_MSG_LINK = 0x0004,
2560 NETIF_MSG_TIMER = 0x0008,
2561 NETIF_MSG_IFDOWN = 0x0010,
2562 NETIF_MSG_IFUP = 0x0020,
2563 NETIF_MSG_RX_ERR = 0x0040,
2564 NETIF_MSG_TX_ERR = 0x0080,
2565 NETIF_MSG_TX_QUEUED = 0x0100,
2566 NETIF_MSG_INTR = 0x0200,
2567 NETIF_MSG_TX_DONE = 0x0400,
2568 NETIF_MSG_RX_STATUS = 0x0800,
2569 NETIF_MSG_PKTDATA = 0x1000,
2570 NETIF_MSG_HW = 0x2000,
2571 NETIF_MSG_WOL = 0x4000,
2574 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2575 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2576 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2577 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2578 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2579 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2580 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2581 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2582 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2583 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2584 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2585 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2586 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2587 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2588 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2590 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2592 /* use default */
2593 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2594 return default_msg_enable_bits;
2595 if (debug_value == 0) /* no output */
2596 return 0;
2597 /* set low N bits */
2598 return (1 << debug_value) - 1;
2601 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2603 spin_lock(&txq->_xmit_lock);
2604 txq->xmit_lock_owner = cpu;
2607 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2609 spin_lock_bh(&txq->_xmit_lock);
2610 txq->xmit_lock_owner = smp_processor_id();
2613 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2615 bool ok = spin_trylock(&txq->_xmit_lock);
2616 if (likely(ok))
2617 txq->xmit_lock_owner = smp_processor_id();
2618 return ok;
2621 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2623 txq->xmit_lock_owner = -1;
2624 spin_unlock(&txq->_xmit_lock);
2627 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2629 txq->xmit_lock_owner = -1;
2630 spin_unlock_bh(&txq->_xmit_lock);
2633 static inline void txq_trans_update(struct netdev_queue *txq)
2635 if (txq->xmit_lock_owner != -1)
2636 txq->trans_start = jiffies;
2640 * netif_tx_lock - grab network device transmit lock
2641 * @dev: network device
2643 * Get network device transmit lock
2645 static inline void netif_tx_lock(struct net_device *dev)
2647 unsigned int i;
2648 int cpu;
2650 spin_lock(&dev->tx_global_lock);
2651 cpu = smp_processor_id();
2652 for (i = 0; i < dev->num_tx_queues; i++) {
2653 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2655 /* We are the only thread of execution doing a
2656 * freeze, but we have to grab the _xmit_lock in
2657 * order to synchronize with threads which are in
2658 * the ->hard_start_xmit() handler and already
2659 * checked the frozen bit.
2661 __netif_tx_lock(txq, cpu);
2662 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2663 __netif_tx_unlock(txq);
2667 static inline void netif_tx_lock_bh(struct net_device *dev)
2669 local_bh_disable();
2670 netif_tx_lock(dev);
2673 static inline void netif_tx_unlock(struct net_device *dev)
2675 unsigned int i;
2677 for (i = 0; i < dev->num_tx_queues; i++) {
2678 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2680 /* No need to grab the _xmit_lock here. If the
2681 * queue is not stopped for another reason, we
2682 * force a schedule.
2684 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2685 netif_schedule_queue(txq);
2687 spin_unlock(&dev->tx_global_lock);
2690 static inline void netif_tx_unlock_bh(struct net_device *dev)
2692 netif_tx_unlock(dev);
2693 local_bh_enable();
2696 #define HARD_TX_LOCK(dev, txq, cpu) { \
2697 if ((dev->features & NETIF_F_LLTX) == 0) { \
2698 __netif_tx_lock(txq, cpu); \
2702 #define HARD_TX_UNLOCK(dev, txq) { \
2703 if ((dev->features & NETIF_F_LLTX) == 0) { \
2704 __netif_tx_unlock(txq); \
2708 static inline void netif_tx_disable(struct net_device *dev)
2710 unsigned int i;
2711 int cpu;
2713 local_bh_disable();
2714 cpu = smp_processor_id();
2715 for (i = 0; i < dev->num_tx_queues; i++) {
2716 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2718 __netif_tx_lock(txq, cpu);
2719 netif_tx_stop_queue(txq);
2720 __netif_tx_unlock(txq);
2722 local_bh_enable();
2725 static inline void netif_addr_lock(struct net_device *dev)
2727 spin_lock(&dev->addr_list_lock);
2730 static inline void netif_addr_lock_nested(struct net_device *dev)
2732 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2735 static inline void netif_addr_lock_bh(struct net_device *dev)
2737 spin_lock_bh(&dev->addr_list_lock);
2740 static inline void netif_addr_unlock(struct net_device *dev)
2742 spin_unlock(&dev->addr_list_lock);
2745 static inline void netif_addr_unlock_bh(struct net_device *dev)
2747 spin_unlock_bh(&dev->addr_list_lock);
2751 * dev_addrs walker. Should be used only for read access. Call with
2752 * rcu_read_lock held.
2754 #define for_each_dev_addr(dev, ha) \
2755 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2757 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2759 void ether_setup(struct net_device *dev);
2761 /* Support for loadable net-drivers */
2762 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2763 void (*setup)(struct net_device *),
2764 unsigned int txqs, unsigned int rxqs);
2765 #define alloc_netdev(sizeof_priv, name, setup) \
2766 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2768 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2769 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2771 int register_netdev(struct net_device *dev);
2772 void unregister_netdev(struct net_device *dev);
2774 /* General hardware address lists handling functions */
2775 int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2776 struct netdev_hw_addr_list *from_list,
2777 int addr_len, unsigned char addr_type);
2778 void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2779 struct netdev_hw_addr_list *from_list,
2780 int addr_len, unsigned char addr_type);
2781 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2782 struct netdev_hw_addr_list *from_list, int addr_len);
2783 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2784 struct netdev_hw_addr_list *from_list, int addr_len);
2785 void __hw_addr_flush(struct netdev_hw_addr_list *list);
2786 void __hw_addr_init(struct netdev_hw_addr_list *list);
2788 /* Functions used for device addresses handling */
2789 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2790 unsigned char addr_type);
2791 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2792 unsigned char addr_type);
2793 int dev_addr_add_multiple(struct net_device *to_dev,
2794 struct net_device *from_dev, unsigned char addr_type);
2795 int dev_addr_del_multiple(struct net_device *to_dev,
2796 struct net_device *from_dev, unsigned char addr_type);
2797 void dev_addr_flush(struct net_device *dev);
2798 int dev_addr_init(struct net_device *dev);
2800 /* Functions used for unicast addresses handling */
2801 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2802 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2803 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
2804 int dev_uc_sync(struct net_device *to, struct net_device *from);
2805 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
2806 void dev_uc_unsync(struct net_device *to, struct net_device *from);
2807 void dev_uc_flush(struct net_device *dev);
2808 void dev_uc_init(struct net_device *dev);
2810 /* Functions used for multicast addresses handling */
2811 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
2812 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
2813 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
2814 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
2815 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
2816 int dev_mc_sync(struct net_device *to, struct net_device *from);
2817 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
2818 void dev_mc_unsync(struct net_device *to, struct net_device *from);
2819 void dev_mc_flush(struct net_device *dev);
2820 void dev_mc_init(struct net_device *dev);
2822 /* Functions used for secondary unicast and multicast support */
2823 void dev_set_rx_mode(struct net_device *dev);
2824 void __dev_set_rx_mode(struct net_device *dev);
2825 int dev_set_promiscuity(struct net_device *dev, int inc);
2826 int dev_set_allmulti(struct net_device *dev, int inc);
2827 void netdev_state_change(struct net_device *dev);
2828 void netdev_notify_peers(struct net_device *dev);
2829 void netdev_features_change(struct net_device *dev);
2830 /* Load a device via the kmod */
2831 void dev_load(struct net *net, const char *name);
2832 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2833 struct rtnl_link_stats64 *storage);
2834 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
2835 const struct net_device_stats *netdev_stats);
2837 extern int netdev_max_backlog;
2838 extern int netdev_tstamp_prequeue;
2839 extern int weight_p;
2840 extern int bpf_jit_enable;
2842 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
2843 bool netdev_has_any_upper_dev(struct net_device *dev);
2844 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
2845 struct list_head **iter);
2847 /* iterate through upper list, must be called under RCU read lock */
2848 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
2849 for (iter = &(dev)->all_adj_list.upper, \
2850 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
2851 updev; \
2852 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
2854 void *netdev_lower_get_next_private(struct net_device *dev,
2855 struct list_head **iter);
2856 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
2857 struct list_head **iter);
2859 #define netdev_for_each_lower_private(dev, priv, iter) \
2860 for (iter = (dev)->adj_list.lower.next, \
2861 priv = netdev_lower_get_next_private(dev, &(iter)); \
2862 priv; \
2863 priv = netdev_lower_get_next_private(dev, &(iter)))
2865 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
2866 for (iter = &(dev)->adj_list.lower, \
2867 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
2868 priv; \
2869 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
2871 void *netdev_adjacent_get_private(struct list_head *adj_list);
2872 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
2873 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
2874 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
2875 int netdev_master_upper_dev_link(struct net_device *dev,
2876 struct net_device *upper_dev);
2877 int netdev_master_upper_dev_link_private(struct net_device *dev,
2878 struct net_device *upper_dev,
2879 void *private);
2880 void netdev_upper_dev_unlink(struct net_device *dev,
2881 struct net_device *upper_dev);
2882 void *netdev_lower_dev_get_private_rcu(struct net_device *dev,
2883 struct net_device *lower_dev);
2884 void *netdev_lower_dev_get_private(struct net_device *dev,
2885 struct net_device *lower_dev);
2886 int skb_checksum_help(struct sk_buff *skb);
2887 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
2888 netdev_features_t features, bool tx_path);
2889 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
2890 netdev_features_t features);
2892 static inline
2893 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
2895 return __skb_gso_segment(skb, features, true);
2897 __be16 skb_network_protocol(struct sk_buff *skb);
2899 static inline bool can_checksum_protocol(netdev_features_t features,
2900 __be16 protocol)
2902 return ((features & NETIF_F_GEN_CSUM) ||
2903 ((features & NETIF_F_V4_CSUM) &&
2904 protocol == htons(ETH_P_IP)) ||
2905 ((features & NETIF_F_V6_CSUM) &&
2906 protocol == htons(ETH_P_IPV6)) ||
2907 ((features & NETIF_F_FCOE_CRC) &&
2908 protocol == htons(ETH_P_FCOE)));
2911 #ifdef CONFIG_BUG
2912 void netdev_rx_csum_fault(struct net_device *dev);
2913 #else
2914 static inline void netdev_rx_csum_fault(struct net_device *dev)
2917 #endif
2918 /* rx skb timestamps */
2919 void net_enable_timestamp(void);
2920 void net_disable_timestamp(void);
2922 #ifdef CONFIG_PROC_FS
2923 int __init dev_proc_init(void);
2924 #else
2925 #define dev_proc_init() 0
2926 #endif
2928 int netdev_class_create_file_ns(struct class_attribute *class_attr,
2929 const void *ns);
2930 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
2931 const void *ns);
2933 static inline int netdev_class_create_file(struct class_attribute *class_attr)
2935 return netdev_class_create_file_ns(class_attr, NULL);
2938 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
2940 netdev_class_remove_file_ns(class_attr, NULL);
2943 extern struct kobj_ns_type_operations net_ns_type_operations;
2945 const char *netdev_drivername(const struct net_device *dev);
2947 void linkwatch_run_queue(void);
2949 static inline netdev_features_t netdev_get_wanted_features(
2950 struct net_device *dev)
2952 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2954 netdev_features_t netdev_increment_features(netdev_features_t all,
2955 netdev_features_t one, netdev_features_t mask);
2957 /* Allow TSO being used on stacked device :
2958 * Performing the GSO segmentation before last device
2959 * is a performance improvement.
2961 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
2962 netdev_features_t mask)
2964 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
2967 int __netdev_update_features(struct net_device *dev);
2968 void netdev_update_features(struct net_device *dev);
2969 void netdev_change_features(struct net_device *dev);
2971 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
2972 struct net_device *dev);
2974 netdev_features_t netif_skb_features(struct sk_buff *skb);
2976 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
2978 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
2980 /* check flags correspondence */
2981 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
2982 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
2983 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
2984 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
2985 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
2986 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
2988 return (features & feature) == feature;
2991 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
2993 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
2994 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
2997 static inline bool netif_needs_gso(struct sk_buff *skb,
2998 netdev_features_t features)
3000 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3001 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3002 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3005 static inline void netif_set_gso_max_size(struct net_device *dev,
3006 unsigned int size)
3008 dev->gso_max_size = size;
3011 static inline bool netif_is_macvlan(struct net_device *dev)
3013 return dev->priv_flags & IFF_MACVLAN;
3016 static inline bool netif_is_bond_master(struct net_device *dev)
3018 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3021 static inline bool netif_is_bond_slave(struct net_device *dev)
3023 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3026 static inline bool netif_supports_nofcs(struct net_device *dev)
3028 return dev->priv_flags & IFF_SUPP_NOFCS;
3031 extern struct pernet_operations __net_initdata loopback_net_ops;
3033 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3035 /* netdev_printk helpers, similar to dev_printk */
3037 static inline const char *netdev_name(const struct net_device *dev)
3039 if (dev->reg_state != NETREG_REGISTERED)
3040 return "(unregistered net_device)";
3041 return dev->name;
3044 __printf(3, 4)
3045 int netdev_printk(const char *level, const struct net_device *dev,
3046 const char *format, ...);
3047 __printf(2, 3)
3048 int netdev_emerg(const struct net_device *dev, const char *format, ...);
3049 __printf(2, 3)
3050 int netdev_alert(const struct net_device *dev, const char *format, ...);
3051 __printf(2, 3)
3052 int netdev_crit(const struct net_device *dev, const char *format, ...);
3053 __printf(2, 3)
3054 int netdev_err(const struct net_device *dev, const char *format, ...);
3055 __printf(2, 3)
3056 int netdev_warn(const struct net_device *dev, const char *format, ...);
3057 __printf(2, 3)
3058 int netdev_notice(const struct net_device *dev, const char *format, ...);
3059 __printf(2, 3)
3060 int netdev_info(const struct net_device *dev, const char *format, ...);
3062 #define MODULE_ALIAS_NETDEV(device) \
3063 MODULE_ALIAS("netdev-" device)
3065 #if defined(CONFIG_DYNAMIC_DEBUG)
3066 #define netdev_dbg(__dev, format, args...) \
3067 do { \
3068 dynamic_netdev_dbg(__dev, format, ##args); \
3069 } while (0)
3070 #elif defined(DEBUG)
3071 #define netdev_dbg(__dev, format, args...) \
3072 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3073 #else
3074 #define netdev_dbg(__dev, format, args...) \
3075 ({ \
3076 if (0) \
3077 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3078 0; \
3080 #endif
3082 #if defined(VERBOSE_DEBUG)
3083 #define netdev_vdbg netdev_dbg
3084 #else
3086 #define netdev_vdbg(dev, format, args...) \
3087 ({ \
3088 if (0) \
3089 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3090 0; \
3092 #endif
3095 * netdev_WARN() acts like dev_printk(), but with the key difference
3096 * of using a WARN/WARN_ON to get the message out, including the
3097 * file/line information and a backtrace.
3099 #define netdev_WARN(dev, format, args...) \
3100 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
3102 /* netif printk helpers, similar to netdev_printk */
3104 #define netif_printk(priv, type, level, dev, fmt, args...) \
3105 do { \
3106 if (netif_msg_##type(priv)) \
3107 netdev_printk(level, (dev), fmt, ##args); \
3108 } while (0)
3110 #define netif_level(level, priv, type, dev, fmt, args...) \
3111 do { \
3112 if (netif_msg_##type(priv)) \
3113 netdev_##level(dev, fmt, ##args); \
3114 } while (0)
3116 #define netif_emerg(priv, type, dev, fmt, args...) \
3117 netif_level(emerg, priv, type, dev, fmt, ##args)
3118 #define netif_alert(priv, type, dev, fmt, args...) \
3119 netif_level(alert, priv, type, dev, fmt, ##args)
3120 #define netif_crit(priv, type, dev, fmt, args...) \
3121 netif_level(crit, priv, type, dev, fmt, ##args)
3122 #define netif_err(priv, type, dev, fmt, args...) \
3123 netif_level(err, priv, type, dev, fmt, ##args)
3124 #define netif_warn(priv, type, dev, fmt, args...) \
3125 netif_level(warn, priv, type, dev, fmt, ##args)
3126 #define netif_notice(priv, type, dev, fmt, args...) \
3127 netif_level(notice, priv, type, dev, fmt, ##args)
3128 #define netif_info(priv, type, dev, fmt, args...) \
3129 netif_level(info, priv, type, dev, fmt, ##args)
3131 #if defined(CONFIG_DYNAMIC_DEBUG)
3132 #define netif_dbg(priv, type, netdev, format, args...) \
3133 do { \
3134 if (netif_msg_##type(priv)) \
3135 dynamic_netdev_dbg(netdev, format, ##args); \
3136 } while (0)
3137 #elif defined(DEBUG)
3138 #define netif_dbg(priv, type, dev, format, args...) \
3139 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3140 #else
3141 #define netif_dbg(priv, type, dev, format, args...) \
3142 ({ \
3143 if (0) \
3144 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3145 0; \
3147 #endif
3149 #if defined(VERBOSE_DEBUG)
3150 #define netif_vdbg netif_dbg
3151 #else
3152 #define netif_vdbg(priv, type, dev, format, args...) \
3153 ({ \
3154 if (0) \
3155 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3156 0; \
3158 #endif
3161 * The list of packet types we will receive (as opposed to discard)
3162 * and the routines to invoke.
3164 * Why 16. Because with 16 the only overlap we get on a hash of the
3165 * low nibble of the protocol value is RARP/SNAP/X.25.
3167 * NOTE: That is no longer true with the addition of VLAN tags. Not
3168 * sure which should go first, but I bet it won't make much
3169 * difference if we are running VLANs. The good news is that
3170 * this protocol won't be in the list unless compiled in, so
3171 * the average user (w/out VLANs) will not be adversely affected.
3172 * --BLG
3174 * 0800 IP
3175 * 8100 802.1Q VLAN
3176 * 0001 802.3
3177 * 0002 AX.25
3178 * 0004 802.2
3179 * 8035 RARP
3180 * 0005 SNAP
3181 * 0805 X.25
3182 * 0806 ARP
3183 * 8137 IPX
3184 * 0009 Localtalk
3185 * 86DD IPv6
3187 #define PTYPE_HASH_SIZE (16)
3188 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3190 #endif /* _LINUX_NETDEVICE_H */