netdevice: Support DSA tagging when DSA is built as a module
[linux-2.6/btrfs-unstable.git] / include / linux / netdevice.h
blobf9e81d10a3b9704b409756fb98ce0e0a268dc1f9
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;
60 void netdev_set_default_ethtool_ops(struct net_device *dev,
61 const struct ethtool_ops *ops);
63 /* Backlog congestion levels */
64 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
65 #define NET_RX_DROP 1 /* packet dropped */
68 * Transmit return codes: transmit return codes originate from three different
69 * namespaces:
71 * - qdisc return codes
72 * - driver transmit return codes
73 * - errno values
75 * Drivers are allowed to return any one of those in their hard_start_xmit()
76 * function. Real network devices commonly used with qdiscs should only return
77 * the driver transmit return codes though - when qdiscs are used, the actual
78 * transmission happens asynchronously, so the value is not propagated to
79 * higher layers. Virtual network devices transmit synchronously, in this case
80 * the driver transmit return codes are consumed by dev_queue_xmit(), all
81 * others are propagated to higher layers.
84 /* qdisc ->enqueue() return codes. */
85 #define NET_XMIT_SUCCESS 0x00
86 #define NET_XMIT_DROP 0x01 /* skb dropped */
87 #define NET_XMIT_CN 0x02 /* congestion notification */
88 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
89 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
91 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
92 * indicates that the device will soon be dropping packets, or already drops
93 * some packets of the same priority; prompting us to send less aggressively. */
94 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
95 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
97 /* Driver transmit return codes */
98 #define NETDEV_TX_MASK 0xf0
100 enum netdev_tx {
101 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
102 NETDEV_TX_OK = 0x00, /* driver took care of packet */
103 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
104 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
106 typedef enum netdev_tx netdev_tx_t;
109 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
110 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
112 static inline bool dev_xmit_complete(int rc)
115 * Positive cases with an skb consumed by a driver:
116 * - successful transmission (rc == NETDEV_TX_OK)
117 * - error while transmitting (rc < 0)
118 * - error while queueing to a different device (rc & NET_XMIT_MASK)
120 if (likely(rc < NET_XMIT_MASK))
121 return true;
123 return false;
127 * Compute the worst case header length according to the protocols
128 * used.
131 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
132 # if defined(CONFIG_MAC80211_MESH)
133 # define LL_MAX_HEADER 128
134 # else
135 # define LL_MAX_HEADER 96
136 # endif
137 #else
138 # define LL_MAX_HEADER 32
139 #endif
141 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
142 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
143 #define MAX_HEADER LL_MAX_HEADER
144 #else
145 #define MAX_HEADER (LL_MAX_HEADER + 48)
146 #endif
149 * Old network device statistics. Fields are native words
150 * (unsigned long) so they can be read and written atomically.
153 struct net_device_stats {
154 unsigned long rx_packets;
155 unsigned long tx_packets;
156 unsigned long rx_bytes;
157 unsigned long tx_bytes;
158 unsigned long rx_errors;
159 unsigned long tx_errors;
160 unsigned long rx_dropped;
161 unsigned long tx_dropped;
162 unsigned long multicast;
163 unsigned long collisions;
164 unsigned long rx_length_errors;
165 unsigned long rx_over_errors;
166 unsigned long rx_crc_errors;
167 unsigned long rx_frame_errors;
168 unsigned long rx_fifo_errors;
169 unsigned long rx_missed_errors;
170 unsigned long tx_aborted_errors;
171 unsigned long tx_carrier_errors;
172 unsigned long tx_fifo_errors;
173 unsigned long tx_heartbeat_errors;
174 unsigned long tx_window_errors;
175 unsigned long rx_compressed;
176 unsigned long tx_compressed;
180 #include <linux/cache.h>
181 #include <linux/skbuff.h>
183 #ifdef CONFIG_RPS
184 #include <linux/static_key.h>
185 extern struct static_key rps_needed;
186 #endif
188 struct neighbour;
189 struct neigh_parms;
190 struct sk_buff;
192 struct netdev_hw_addr {
193 struct list_head list;
194 unsigned char addr[MAX_ADDR_LEN];
195 unsigned char type;
196 #define NETDEV_HW_ADDR_T_LAN 1
197 #define NETDEV_HW_ADDR_T_SAN 2
198 #define NETDEV_HW_ADDR_T_SLAVE 3
199 #define NETDEV_HW_ADDR_T_UNICAST 4
200 #define NETDEV_HW_ADDR_T_MULTICAST 5
201 bool global_use;
202 int sync_cnt;
203 int refcount;
204 int synced;
205 struct rcu_head rcu_head;
208 struct netdev_hw_addr_list {
209 struct list_head list;
210 int count;
213 #define netdev_hw_addr_list_count(l) ((l)->count)
214 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
215 #define netdev_hw_addr_list_for_each(ha, l) \
216 list_for_each_entry(ha, &(l)->list, list)
218 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
219 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
220 #define netdev_for_each_uc_addr(ha, dev) \
221 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
223 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
224 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
225 #define netdev_for_each_mc_addr(ha, dev) \
226 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
228 struct hh_cache {
229 u16 hh_len;
230 u16 __pad;
231 seqlock_t hh_lock;
233 /* cached hardware header; allow for machine alignment needs. */
234 #define HH_DATA_MOD 16
235 #define HH_DATA_OFF(__len) \
236 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
237 #define HH_DATA_ALIGN(__len) \
238 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
239 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
242 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
243 * Alternative is:
244 * dev->hard_header_len ? (dev->hard_header_len +
245 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
247 * We could use other alignment values, but we must maintain the
248 * relationship HH alignment <= LL alignment.
250 #define LL_RESERVED_SPACE(dev) \
251 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
252 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
253 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
255 struct header_ops {
256 int (*create) (struct sk_buff *skb, struct net_device *dev,
257 unsigned short type, const void *daddr,
258 const void *saddr, unsigned int len);
259 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
260 int (*rebuild)(struct sk_buff *skb);
261 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
262 void (*cache_update)(struct hh_cache *hh,
263 const struct net_device *dev,
264 const unsigned char *haddr);
267 /* These flag bits are private to the generic network queueing
268 * layer, they may not be explicitly referenced by any other
269 * code.
272 enum netdev_state_t {
273 __LINK_STATE_START,
274 __LINK_STATE_PRESENT,
275 __LINK_STATE_NOCARRIER,
276 __LINK_STATE_LINKWATCH_PENDING,
277 __LINK_STATE_DORMANT,
282 * This structure holds at boot time configured netdevice settings. They
283 * are then used in the device probing.
285 struct netdev_boot_setup {
286 char name[IFNAMSIZ];
287 struct ifmap map;
289 #define NETDEV_BOOT_SETUP_MAX 8
291 int __init netdev_boot_setup(char *str);
294 * Structure for NAPI scheduling similar to tasklet but with weighting
296 struct napi_struct {
297 /* The poll_list must only be managed by the entity which
298 * changes the state of the NAPI_STATE_SCHED bit. This means
299 * whoever atomically sets that bit can add this napi_struct
300 * to the per-cpu poll_list, and whoever clears that bit
301 * can remove from the list right before clearing the bit.
303 struct list_head poll_list;
305 unsigned long state;
306 int weight;
307 unsigned int gro_count;
308 int (*poll)(struct napi_struct *, int);
309 #ifdef CONFIG_NETPOLL
310 spinlock_t poll_lock;
311 int poll_owner;
312 #endif
313 struct net_device *dev;
314 struct sk_buff *gro_list;
315 struct sk_buff *skb;
316 struct list_head dev_list;
317 struct hlist_node napi_hash_node;
318 unsigned int napi_id;
321 enum {
322 NAPI_STATE_SCHED, /* Poll is scheduled */
323 NAPI_STATE_DISABLE, /* Disable pending */
324 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
325 NAPI_STATE_HASHED, /* In NAPI hash */
328 enum gro_result {
329 GRO_MERGED,
330 GRO_MERGED_FREE,
331 GRO_HELD,
332 GRO_NORMAL,
333 GRO_DROP,
335 typedef enum gro_result gro_result_t;
338 * enum rx_handler_result - Possible return values for rx_handlers.
339 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
340 * further.
341 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
342 * case skb->dev was changed by rx_handler.
343 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
344 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
346 * rx_handlers are functions called from inside __netif_receive_skb(), to do
347 * special processing of the skb, prior to delivery to protocol handlers.
349 * Currently, a net_device can only have a single rx_handler registered. Trying
350 * to register a second rx_handler will return -EBUSY.
352 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
353 * To unregister a rx_handler on a net_device, use
354 * netdev_rx_handler_unregister().
356 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
357 * do with the skb.
359 * If the rx_handler consumed to skb in some way, it should return
360 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
361 * the skb to be delivered in some other ways.
363 * If the rx_handler changed skb->dev, to divert the skb to another
364 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
365 * new device will be called if it exists.
367 * If the rx_handler consider the skb should be ignored, it should return
368 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
369 * are registered on exact device (ptype->dev == skb->dev).
371 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
372 * delivered, it should return RX_HANDLER_PASS.
374 * A device without a registered rx_handler will behave as if rx_handler
375 * returned RX_HANDLER_PASS.
378 enum rx_handler_result {
379 RX_HANDLER_CONSUMED,
380 RX_HANDLER_ANOTHER,
381 RX_HANDLER_EXACT,
382 RX_HANDLER_PASS,
384 typedef enum rx_handler_result rx_handler_result_t;
385 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
387 void __napi_schedule(struct napi_struct *n);
389 static inline bool napi_disable_pending(struct napi_struct *n)
391 return test_bit(NAPI_STATE_DISABLE, &n->state);
395 * napi_schedule_prep - check if napi can be scheduled
396 * @n: napi context
398 * Test if NAPI routine is already running, and if not mark
399 * it as running. This is used as a condition variable
400 * insure only one NAPI poll instance runs. We also make
401 * sure there is no pending NAPI disable.
403 static inline bool napi_schedule_prep(struct napi_struct *n)
405 return !napi_disable_pending(n) &&
406 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
410 * napi_schedule - schedule NAPI poll
411 * @n: napi context
413 * Schedule NAPI poll routine to be called if it is not already
414 * running.
416 static inline void napi_schedule(struct napi_struct *n)
418 if (napi_schedule_prep(n))
419 __napi_schedule(n);
422 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
423 static inline bool napi_reschedule(struct napi_struct *napi)
425 if (napi_schedule_prep(napi)) {
426 __napi_schedule(napi);
427 return true;
429 return false;
433 * napi_complete - NAPI processing complete
434 * @n: napi context
436 * Mark NAPI processing as complete.
438 void __napi_complete(struct napi_struct *n);
439 void napi_complete(struct napi_struct *n);
442 * napi_by_id - lookup a NAPI by napi_id
443 * @napi_id: hashed napi_id
445 * lookup @napi_id in napi_hash table
446 * must be called under rcu_read_lock()
448 struct napi_struct *napi_by_id(unsigned int napi_id);
451 * napi_hash_add - add a NAPI to global hashtable
452 * @napi: napi context
454 * generate a new napi_id and store a @napi under it in napi_hash
456 void napi_hash_add(struct napi_struct *napi);
459 * napi_hash_del - remove a NAPI from global table
460 * @napi: napi context
462 * Warning: caller must observe rcu grace period
463 * before freeing memory containing @napi
465 void napi_hash_del(struct napi_struct *napi);
468 * napi_disable - prevent NAPI from scheduling
469 * @n: napi context
471 * Stop NAPI from being scheduled on this context.
472 * Waits till any outstanding processing completes.
474 static inline void napi_disable(struct napi_struct *n)
476 might_sleep();
477 set_bit(NAPI_STATE_DISABLE, &n->state);
478 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
479 msleep(1);
480 clear_bit(NAPI_STATE_DISABLE, &n->state);
484 * napi_enable - enable NAPI scheduling
485 * @n: napi context
487 * Resume NAPI from being scheduled on this context.
488 * Must be paired with napi_disable.
490 static inline void napi_enable(struct napi_struct *n)
492 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
493 smp_mb__before_atomic();
494 clear_bit(NAPI_STATE_SCHED, &n->state);
497 #ifdef CONFIG_SMP
499 * napi_synchronize - wait until NAPI is not running
500 * @n: napi context
502 * Wait until NAPI is done being scheduled on this context.
503 * Waits till any outstanding processing completes but
504 * does not disable future activations.
506 static inline void napi_synchronize(const struct napi_struct *n)
508 while (test_bit(NAPI_STATE_SCHED, &n->state))
509 msleep(1);
511 #else
512 # define napi_synchronize(n) barrier()
513 #endif
515 enum netdev_queue_state_t {
516 __QUEUE_STATE_DRV_XOFF,
517 __QUEUE_STATE_STACK_XOFF,
518 __QUEUE_STATE_FROZEN,
521 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
522 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
523 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
525 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
526 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
527 QUEUE_STATE_FROZEN)
528 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
529 QUEUE_STATE_FROZEN)
532 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
533 * netif_tx_* functions below are used to manipulate this flag. The
534 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
535 * queue independently. The netif_xmit_*stopped functions below are called
536 * to check if the queue has been stopped by the driver or stack (either
537 * of the XOFF bits are set in the state). Drivers should not need to call
538 * netif_xmit*stopped functions, they should only be using netif_tx_*.
541 struct netdev_queue {
543 * read mostly part
545 struct net_device *dev;
546 struct Qdisc __rcu *qdisc;
547 struct Qdisc *qdisc_sleeping;
548 #ifdef CONFIG_SYSFS
549 struct kobject kobj;
550 #endif
551 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
552 int numa_node;
553 #endif
555 * write mostly part
557 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
558 int xmit_lock_owner;
560 * please use this field instead of dev->trans_start
562 unsigned long trans_start;
565 * Number of TX timeouts for this queue
566 * (/sys/class/net/DEV/Q/trans_timeout)
568 unsigned long trans_timeout;
570 unsigned long state;
572 #ifdef CONFIG_BQL
573 struct dql dql;
574 #endif
575 } ____cacheline_aligned_in_smp;
577 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
579 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
580 return q->numa_node;
581 #else
582 return NUMA_NO_NODE;
583 #endif
586 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
588 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
589 q->numa_node = node;
590 #endif
593 #ifdef CONFIG_RPS
595 * This structure holds an RPS map which can be of variable length. The
596 * map is an array of CPUs.
598 struct rps_map {
599 unsigned int len;
600 struct rcu_head rcu;
601 u16 cpus[0];
603 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
606 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
607 * tail pointer for that CPU's input queue at the time of last enqueue, and
608 * a hardware filter index.
610 struct rps_dev_flow {
611 u16 cpu;
612 u16 filter;
613 unsigned int last_qtail;
615 #define RPS_NO_FILTER 0xffff
618 * The rps_dev_flow_table structure contains a table of flow mappings.
620 struct rps_dev_flow_table {
621 unsigned int mask;
622 struct rcu_head rcu;
623 struct rps_dev_flow flows[0];
625 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
626 ((_num) * sizeof(struct rps_dev_flow)))
629 * The rps_sock_flow_table contains mappings of flows to the last CPU
630 * on which they were processed by the application (set in recvmsg).
632 struct rps_sock_flow_table {
633 unsigned int mask;
634 u16 ents[0];
636 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
637 ((_num) * sizeof(u16)))
639 #define RPS_NO_CPU 0xffff
641 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
642 u32 hash)
644 if (table && hash) {
645 unsigned int cpu, index = hash & table->mask;
647 /* We only give a hint, preemption can change cpu under us */
648 cpu = raw_smp_processor_id();
650 if (table->ents[index] != cpu)
651 table->ents[index] = cpu;
655 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
656 u32 hash)
658 if (table && hash)
659 table->ents[hash & table->mask] = RPS_NO_CPU;
662 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
664 #ifdef CONFIG_RFS_ACCEL
665 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
666 u16 filter_id);
667 #endif
668 #endif /* CONFIG_RPS */
670 /* This structure contains an instance of an RX queue. */
671 struct netdev_rx_queue {
672 #ifdef CONFIG_RPS
673 struct rps_map __rcu *rps_map;
674 struct rps_dev_flow_table __rcu *rps_flow_table;
675 #endif
676 struct kobject kobj;
677 struct net_device *dev;
678 } ____cacheline_aligned_in_smp;
681 * RX queue sysfs structures and functions.
683 struct rx_queue_attribute {
684 struct attribute attr;
685 ssize_t (*show)(struct netdev_rx_queue *queue,
686 struct rx_queue_attribute *attr, char *buf);
687 ssize_t (*store)(struct netdev_rx_queue *queue,
688 struct rx_queue_attribute *attr, const char *buf, size_t len);
691 #ifdef CONFIG_XPS
693 * This structure holds an XPS map which can be of variable length. The
694 * map is an array of queues.
696 struct xps_map {
697 unsigned int len;
698 unsigned int alloc_len;
699 struct rcu_head rcu;
700 u16 queues[0];
702 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
703 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
704 / sizeof(u16))
707 * This structure holds all XPS maps for device. Maps are indexed by CPU.
709 struct xps_dev_maps {
710 struct rcu_head rcu;
711 struct xps_map __rcu *cpu_map[0];
713 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
714 (nr_cpu_ids * sizeof(struct xps_map *)))
715 #endif /* CONFIG_XPS */
717 #define TC_MAX_QUEUE 16
718 #define TC_BITMASK 15
719 /* HW offloaded queuing disciplines txq count and offset maps */
720 struct netdev_tc_txq {
721 u16 count;
722 u16 offset;
725 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
727 * This structure is to hold information about the device
728 * configured to run FCoE protocol stack.
730 struct netdev_fcoe_hbainfo {
731 char manufacturer[64];
732 char serial_number[64];
733 char hardware_version[64];
734 char driver_version[64];
735 char optionrom_version[64];
736 char firmware_version[64];
737 char model[256];
738 char model_description[256];
740 #endif
742 #define MAX_PHYS_PORT_ID_LEN 32
744 /* This structure holds a unique identifier to identify the
745 * physical port used by a netdevice.
747 struct netdev_phys_port_id {
748 unsigned char id[MAX_PHYS_PORT_ID_LEN];
749 unsigned char id_len;
752 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
753 struct sk_buff *skb);
756 * This structure defines the management hooks for network devices.
757 * The following hooks can be defined; unless noted otherwise, they are
758 * optional and can be filled with a null pointer.
760 * int (*ndo_init)(struct net_device *dev);
761 * This function is called once when network device is registered.
762 * The network device can use this to any late stage initializaton
763 * or semantic validattion. It can fail with an error code which will
764 * be propogated back to register_netdev
766 * void (*ndo_uninit)(struct net_device *dev);
767 * This function is called when device is unregistered or when registration
768 * fails. It is not called if init fails.
770 * int (*ndo_open)(struct net_device *dev);
771 * This function is called when network device transistions to the up
772 * state.
774 * int (*ndo_stop)(struct net_device *dev);
775 * This function is called when network device transistions to the down
776 * state.
778 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
779 * struct net_device *dev);
780 * Called when a packet needs to be transmitted.
781 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
782 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
783 * Required can not be NULL.
785 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
786 * void *accel_priv, select_queue_fallback_t fallback);
787 * Called to decide which queue to when device supports multiple
788 * transmit queues.
790 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
791 * This function is called to allow device receiver to make
792 * changes to configuration when multicast or promiscious is enabled.
794 * void (*ndo_set_rx_mode)(struct net_device *dev);
795 * This function is called device changes address list filtering.
796 * If driver handles unicast address filtering, it should set
797 * IFF_UNICAST_FLT to its priv_flags.
799 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
800 * This function is called when the Media Access Control address
801 * needs to be changed. If this interface is not defined, the
802 * mac address can not be changed.
804 * int (*ndo_validate_addr)(struct net_device *dev);
805 * Test if Media Access Control address is valid for the device.
807 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
808 * Called when a user request an ioctl which can't be handled by
809 * the generic interface code. If not defined ioctl's return
810 * not supported error code.
812 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
813 * Used to set network devices bus interface parameters. This interface
814 * is retained for legacy reason, new devices should use the bus
815 * interface (PCI) for low level management.
817 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
818 * Called when a user wants to change the Maximum Transfer Unit
819 * of a device. If not defined, any request to change MTU will
820 * will return an error.
822 * void (*ndo_tx_timeout)(struct net_device *dev);
823 * Callback uses when the transmitter has not made any progress
824 * for dev->watchdog ticks.
826 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
827 * struct rtnl_link_stats64 *storage);
828 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
829 * Called when a user wants to get the network device usage
830 * statistics. Drivers must do one of the following:
831 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
832 * rtnl_link_stats64 structure passed by the caller.
833 * 2. Define @ndo_get_stats to update a net_device_stats structure
834 * (which should normally be dev->stats) and return a pointer to
835 * it. The structure may be changed asynchronously only if each
836 * field is written atomically.
837 * 3. Update dev->stats asynchronously and atomically, and define
838 * neither operation.
840 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
841 * If device support VLAN filtering this function is called when a
842 * VLAN id is registered.
844 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
845 * If device support VLAN filtering this function is called when a
846 * VLAN id is unregistered.
848 * void (*ndo_poll_controller)(struct net_device *dev);
850 * SR-IOV management functions.
851 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
852 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
853 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
854 * int max_tx_rate);
855 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
856 * int (*ndo_get_vf_config)(struct net_device *dev,
857 * int vf, struct ifla_vf_info *ivf);
858 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
859 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
860 * struct nlattr *port[]);
861 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
862 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
863 * Called to setup 'tc' number of traffic classes in the net device. This
864 * is always called from the stack with the rtnl lock held and netif tx
865 * queues stopped. This allows the netdevice to perform queue management
866 * safely.
868 * Fiber Channel over Ethernet (FCoE) offload functions.
869 * int (*ndo_fcoe_enable)(struct net_device *dev);
870 * Called when the FCoE protocol stack wants to start using LLD for FCoE
871 * so the underlying device can perform whatever needed configuration or
872 * initialization to support acceleration of FCoE traffic.
874 * int (*ndo_fcoe_disable)(struct net_device *dev);
875 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
876 * so the underlying device can perform whatever needed clean-ups to
877 * stop supporting acceleration of FCoE traffic.
879 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
880 * struct scatterlist *sgl, unsigned int sgc);
881 * Called when the FCoE Initiator wants to initialize an I/O that
882 * is a possible candidate for Direct Data Placement (DDP). The LLD can
883 * perform necessary setup and returns 1 to indicate the device is set up
884 * successfully to perform DDP on this I/O, otherwise this returns 0.
886 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
887 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
888 * indicated by the FC exchange id 'xid', so the underlying device can
889 * clean up and reuse resources for later DDP requests.
891 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
892 * struct scatterlist *sgl, unsigned int sgc);
893 * Called when the FCoE Target wants to initialize an I/O that
894 * is a possible candidate for Direct Data Placement (DDP). The LLD can
895 * perform necessary setup and returns 1 to indicate the device is set up
896 * successfully to perform DDP on this I/O, otherwise this returns 0.
898 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
899 * struct netdev_fcoe_hbainfo *hbainfo);
900 * Called when the FCoE Protocol stack wants information on the underlying
901 * device. This information is utilized by the FCoE protocol stack to
902 * register attributes with Fiber Channel management service as per the
903 * FC-GS Fabric Device Management Information(FDMI) specification.
905 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
906 * Called when the underlying device wants to override default World Wide
907 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
908 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
909 * protocol stack to use.
911 * RFS acceleration.
912 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
913 * u16 rxq_index, u32 flow_id);
914 * Set hardware filter for RFS. rxq_index is the target queue index;
915 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
916 * Return the filter ID on success, or a negative error code.
918 * Slave management functions (for bridge, bonding, etc).
919 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
920 * Called to make another netdev an underling.
922 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
923 * Called to release previously enslaved netdev.
925 * Feature/offload setting functions.
926 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
927 * netdev_features_t features);
928 * Adjusts the requested feature flags according to device-specific
929 * constraints, and returns the resulting flags. Must not modify
930 * the device state.
932 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
933 * Called to update device configuration to new features. Passed
934 * feature set might be less than what was returned by ndo_fix_features()).
935 * Must return >0 or -errno if it changed dev->features itself.
937 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
938 * struct net_device *dev,
939 * const unsigned char *addr, u16 flags)
940 * Adds an FDB entry to dev for addr.
941 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
942 * struct net_device *dev,
943 * const unsigned char *addr)
944 * Deletes the FDB entry from dev coresponding to addr.
945 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
946 * struct net_device *dev, struct net_device *filter_dev,
947 * int idx)
948 * Used to add FDB entries to dump requests. Implementers should add
949 * entries to skb and update idx with the number of entries.
951 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
952 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
953 * struct net_device *dev, u32 filter_mask)
955 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
956 * Called to change device carrier. Soft-devices (like dummy, team, etc)
957 * which do not represent real hardware may define this to allow their
958 * userspace components to manage their virtual carrier state. Devices
959 * that determine carrier state from physical hardware properties (eg
960 * network cables) or protocol-dependent mechanisms (eg
961 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
963 * int (*ndo_get_phys_port_id)(struct net_device *dev,
964 * struct netdev_phys_port_id *ppid);
965 * Called to get ID of physical port of this device. If driver does
966 * not implement this, it is assumed that the hw is not able to have
967 * multiple net devices on single physical port.
969 * void (*ndo_add_vxlan_port)(struct net_device *dev,
970 * sa_family_t sa_family, __be16 port);
971 * Called by vxlan to notiy a driver about the UDP port and socket
972 * address family that vxlan is listnening to. It is called only when
973 * a new port starts listening. The operation is protected by the
974 * vxlan_net->sock_lock.
976 * void (*ndo_del_vxlan_port)(struct net_device *dev,
977 * sa_family_t sa_family, __be16 port);
978 * Called by vxlan to notify the driver about a UDP port and socket
979 * address family that vxlan is not listening to anymore. The operation
980 * is protected by the vxlan_net->sock_lock.
982 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
983 * struct net_device *dev)
984 * Called by upper layer devices to accelerate switching or other
985 * station functionality into hardware. 'pdev is the lowerdev
986 * to use for the offload and 'dev' is the net device that will
987 * back the offload. Returns a pointer to the private structure
988 * the upper layer will maintain.
989 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
990 * Called by upper layer device to delete the station created
991 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
992 * the station and priv is the structure returned by the add
993 * operation.
994 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
995 * struct net_device *dev,
996 * void *priv);
997 * Callback to use for xmit over the accelerated station. This
998 * is used in place of ndo_start_xmit on accelerated net
999 * devices.
1001 struct net_device_ops {
1002 int (*ndo_init)(struct net_device *dev);
1003 void (*ndo_uninit)(struct net_device *dev);
1004 int (*ndo_open)(struct net_device *dev);
1005 int (*ndo_stop)(struct net_device *dev);
1006 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1007 struct net_device *dev);
1008 u16 (*ndo_select_queue)(struct net_device *dev,
1009 struct sk_buff *skb,
1010 void *accel_priv,
1011 select_queue_fallback_t fallback);
1012 void (*ndo_change_rx_flags)(struct net_device *dev,
1013 int flags);
1014 void (*ndo_set_rx_mode)(struct net_device *dev);
1015 int (*ndo_set_mac_address)(struct net_device *dev,
1016 void *addr);
1017 int (*ndo_validate_addr)(struct net_device *dev);
1018 int (*ndo_do_ioctl)(struct net_device *dev,
1019 struct ifreq *ifr, int cmd);
1020 int (*ndo_set_config)(struct net_device *dev,
1021 struct ifmap *map);
1022 int (*ndo_change_mtu)(struct net_device *dev,
1023 int new_mtu);
1024 int (*ndo_neigh_setup)(struct net_device *dev,
1025 struct neigh_parms *);
1026 void (*ndo_tx_timeout) (struct net_device *dev);
1028 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1029 struct rtnl_link_stats64 *storage);
1030 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1032 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1033 __be16 proto, u16 vid);
1034 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1035 __be16 proto, u16 vid);
1036 #ifdef CONFIG_NET_POLL_CONTROLLER
1037 void (*ndo_poll_controller)(struct net_device *dev);
1038 int (*ndo_netpoll_setup)(struct net_device *dev,
1039 struct netpoll_info *info);
1040 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1041 #endif
1042 #ifdef CONFIG_NET_RX_BUSY_POLL
1043 int (*ndo_busy_poll)(struct napi_struct *dev);
1044 #endif
1045 int (*ndo_set_vf_mac)(struct net_device *dev,
1046 int queue, u8 *mac);
1047 int (*ndo_set_vf_vlan)(struct net_device *dev,
1048 int queue, u16 vlan, u8 qos);
1049 int (*ndo_set_vf_rate)(struct net_device *dev,
1050 int vf, int min_tx_rate,
1051 int max_tx_rate);
1052 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1053 int vf, bool setting);
1054 int (*ndo_get_vf_config)(struct net_device *dev,
1055 int vf,
1056 struct ifla_vf_info *ivf);
1057 int (*ndo_set_vf_link_state)(struct net_device *dev,
1058 int vf, int link_state);
1059 int (*ndo_set_vf_port)(struct net_device *dev,
1060 int vf,
1061 struct nlattr *port[]);
1062 int (*ndo_get_vf_port)(struct net_device *dev,
1063 int vf, struct sk_buff *skb);
1064 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1065 #if IS_ENABLED(CONFIG_FCOE)
1066 int (*ndo_fcoe_enable)(struct net_device *dev);
1067 int (*ndo_fcoe_disable)(struct net_device *dev);
1068 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1069 u16 xid,
1070 struct scatterlist *sgl,
1071 unsigned int sgc);
1072 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1073 u16 xid);
1074 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1075 u16 xid,
1076 struct scatterlist *sgl,
1077 unsigned int sgc);
1078 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1079 struct netdev_fcoe_hbainfo *hbainfo);
1080 #endif
1082 #if IS_ENABLED(CONFIG_LIBFCOE)
1083 #define NETDEV_FCOE_WWNN 0
1084 #define NETDEV_FCOE_WWPN 1
1085 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1086 u64 *wwn, int type);
1087 #endif
1089 #ifdef CONFIG_RFS_ACCEL
1090 int (*ndo_rx_flow_steer)(struct net_device *dev,
1091 const struct sk_buff *skb,
1092 u16 rxq_index,
1093 u32 flow_id);
1094 #endif
1095 int (*ndo_add_slave)(struct net_device *dev,
1096 struct net_device *slave_dev);
1097 int (*ndo_del_slave)(struct net_device *dev,
1098 struct net_device *slave_dev);
1099 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1100 netdev_features_t features);
1101 int (*ndo_set_features)(struct net_device *dev,
1102 netdev_features_t features);
1103 int (*ndo_neigh_construct)(struct neighbour *n);
1104 void (*ndo_neigh_destroy)(struct neighbour *n);
1106 int (*ndo_fdb_add)(struct ndmsg *ndm,
1107 struct nlattr *tb[],
1108 struct net_device *dev,
1109 const unsigned char *addr,
1110 u16 flags);
1111 int (*ndo_fdb_del)(struct ndmsg *ndm,
1112 struct nlattr *tb[],
1113 struct net_device *dev,
1114 const unsigned char *addr);
1115 int (*ndo_fdb_dump)(struct sk_buff *skb,
1116 struct netlink_callback *cb,
1117 struct net_device *dev,
1118 struct net_device *filter_dev,
1119 int idx);
1121 int (*ndo_bridge_setlink)(struct net_device *dev,
1122 struct nlmsghdr *nlh);
1123 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1124 u32 pid, u32 seq,
1125 struct net_device *dev,
1126 u32 filter_mask);
1127 int (*ndo_bridge_dellink)(struct net_device *dev,
1128 struct nlmsghdr *nlh);
1129 int (*ndo_change_carrier)(struct net_device *dev,
1130 bool new_carrier);
1131 int (*ndo_get_phys_port_id)(struct net_device *dev,
1132 struct netdev_phys_port_id *ppid);
1133 void (*ndo_add_vxlan_port)(struct net_device *dev,
1134 sa_family_t sa_family,
1135 __be16 port);
1136 void (*ndo_del_vxlan_port)(struct net_device *dev,
1137 sa_family_t sa_family,
1138 __be16 port);
1140 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1141 struct net_device *dev);
1142 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1143 void *priv);
1145 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1146 struct net_device *dev,
1147 void *priv);
1148 int (*ndo_get_lock_subclass)(struct net_device *dev);
1152 * enum net_device_priv_flags - &struct net_device priv_flags
1154 * These are the &struct net_device, they are only set internally
1155 * by drivers and used in the kernel. These flags are invisible to
1156 * userspace, this means that the order of these flags can change
1157 * during any kernel release.
1159 * You should have a pretty good reason to be extending these flags.
1161 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1162 * @IFF_EBRIDGE: Ethernet bridging device
1163 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1164 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1165 * @IFF_MASTER_ALB: bonding master, balance-alb
1166 * @IFF_BONDING: bonding master or slave
1167 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1168 * @IFF_ISATAP: ISATAP interface (RFC4214)
1169 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1170 * @IFF_WAN_HDLC: WAN HDLC device
1171 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1172 * release skb->dst
1173 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1174 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1175 * @IFF_MACVLAN_PORT: device used as macvlan port
1176 * @IFF_BRIDGE_PORT: device used as bridge port
1177 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1178 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1179 * @IFF_UNICAST_FLT: Supports unicast filtering
1180 * @IFF_TEAM_PORT: device used as team port
1181 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1182 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1183 * change when it's running
1184 * @IFF_MACVLAN: Macvlan device
1186 enum netdev_priv_flags {
1187 IFF_802_1Q_VLAN = 1<<0,
1188 IFF_EBRIDGE = 1<<1,
1189 IFF_SLAVE_INACTIVE = 1<<2,
1190 IFF_MASTER_8023AD = 1<<3,
1191 IFF_MASTER_ALB = 1<<4,
1192 IFF_BONDING = 1<<5,
1193 IFF_SLAVE_NEEDARP = 1<<6,
1194 IFF_ISATAP = 1<<7,
1195 IFF_MASTER_ARPMON = 1<<8,
1196 IFF_WAN_HDLC = 1<<9,
1197 IFF_XMIT_DST_RELEASE = 1<<10,
1198 IFF_DONT_BRIDGE = 1<<11,
1199 IFF_DISABLE_NETPOLL = 1<<12,
1200 IFF_MACVLAN_PORT = 1<<13,
1201 IFF_BRIDGE_PORT = 1<<14,
1202 IFF_OVS_DATAPATH = 1<<15,
1203 IFF_TX_SKB_SHARING = 1<<16,
1204 IFF_UNICAST_FLT = 1<<17,
1205 IFF_TEAM_PORT = 1<<18,
1206 IFF_SUPP_NOFCS = 1<<19,
1207 IFF_LIVE_ADDR_CHANGE = 1<<20,
1208 IFF_MACVLAN = 1<<21,
1211 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1212 #define IFF_EBRIDGE IFF_EBRIDGE
1213 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1214 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1215 #define IFF_MASTER_ALB IFF_MASTER_ALB
1216 #define IFF_BONDING IFF_BONDING
1217 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1218 #define IFF_ISATAP IFF_ISATAP
1219 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1220 #define IFF_WAN_HDLC IFF_WAN_HDLC
1221 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1222 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1223 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1224 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1225 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1226 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1227 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1228 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1229 #define IFF_TEAM_PORT IFF_TEAM_PORT
1230 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1231 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1232 #define IFF_MACVLAN IFF_MACVLAN
1235 * struct net_device - The DEVICE structure.
1236 * Actually, this whole structure is a big mistake. It mixes I/O
1237 * data with strictly "high-level" data, and it has to know about
1238 * almost every data structure used in the INET module.
1240 * @name: This is the first field of the "visible" part of this structure
1241 * (i.e. as seen by users in the "Space.c" file). It is the name
1242 * of the interface.
1244 * @name_hlist: Device name hash chain, please keep it close to name[]
1245 * @ifalias: SNMP alias
1246 * @mem_end: Shared memory end
1247 * @mem_start: Shared memory start
1248 * @base_addr: Device I/O address
1249 * @irq: Device IRQ number
1251 * @state: Generic network queuing layer state, see netdev_state_t
1252 * @dev_list: The global list of network devices
1253 * @napi_list: List entry, that is used for polling napi devices
1254 * @unreg_list: List entry, that is used, when we are unregistering the
1255 * device, see the function unregister_netdev
1256 * @close_list: List entry, that is used, when we are closing the device
1258 * @adj_list: Directly linked devices, like slaves for bonding
1259 * @all_adj_list: All linked devices, *including* neighbours
1260 * @features: Currently active device features
1261 * @hw_features: User-changeable features
1263 * @wanted_features: User-requested features
1264 * @vlan_features: Mask of features inheritable by VLAN devices
1266 * @hw_enc_features: Mask of features inherited by encapsulating devices
1267 * This field indicates what encapsulation
1268 * offloads the hardware is capable of doing,
1269 * and drivers will need to set them appropriately.
1271 * @mpls_features: Mask of features inheritable by MPLS
1273 * @ifindex: interface index
1274 * @iflink: unique device identifier
1276 * @stats: Statistics struct, which was left as a legacy, use
1277 * rtnl_link_stats64 instead
1279 * @rx_dropped: Dropped packets by core network,
1280 * do not use this in drivers
1281 * @tx_dropped: Dropped packets by core network,
1282 * do not use this in drivers
1284 * @carrier_changes: Stats to monitor carrier on<->off transitions
1286 * @wireless_handlers: List of functions to handle Wireless Extensions,
1287 * instead of ioctl,
1288 * see <net/iw_handler.h> for details.
1289 * @wireless_data: Instance data managed by the core of wireless extensions
1291 * @netdev_ops: Includes several pointers to callbacks,
1292 * if one wants to override the ndo_*() functions
1293 * @ethtool_ops: Management operations
1294 * @fwd_ops: Management operations
1295 * @header_ops: Includes callbacks for creating,parsing,rebuilding,etc
1296 * of Layer 2 headers.
1298 * @flags: Interface flags (a la BSD)
1299 * @priv_flags: Like 'flags' but invisible to userspace,
1300 * see if.h for the definitions
1301 * @gflags: Global flags ( kept as legacy )
1302 * @padded: How much padding added by alloc_netdev()
1303 * @operstate: RFC2863 operstate
1304 * @link_mode: Mapping policy to operstate
1305 * @if_port: Selectable AUI, TP, ...
1306 * @dma: DMA channel
1307 * @mtu: Interface MTU value
1308 * @type: Interface hardware type
1309 * @hard_header_len: Hardware header length
1311 * @needed_headroom: Extra headroom the hardware may need, but not in all
1312 * cases can this be guaranteed
1313 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1314 * cases can this be guaranteed. Some cases also use
1315 * LL_MAX_HEADER instead to allocate the skb
1317 * interface address info:
1319 * @perm_addr: Permanent hw address
1320 * @addr_assign_type: Hw address assignment type
1321 * @addr_len: Hardware address length
1322 * @neigh_priv_len; Used in neigh_alloc(),
1323 * initialized only in atm/clip.c
1324 * @dev_id: Used to differentiate devices that share
1325 * the same link layer address
1326 * @dev_port: Used to differentiate devices that share
1327 * the same function
1328 * @addr_list_lock: XXX: need comments on this one
1329 * @uc: unicast mac addresses
1330 * @mc: multicast mac addresses
1331 * @dev_addrs: list of device hw addresses
1332 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1333 * @uc_promisc: Counter, that indicates, that promiscuous mode
1334 * has been enabled due to the need to listen to
1335 * additional unicast addresses in a device that
1336 * does not implement ndo_set_rx_mode()
1337 * @promiscuity: Number of times, the NIC is told to work in
1338 * Promiscuous mode, if it becomes 0 the NIC will
1339 * exit from working in Promiscuous mode
1340 * @allmulti: Counter, enables or disables allmulticast mode
1342 * @vlan_info: VLAN info
1343 * @dsa_ptr: dsa specific data
1344 * @tipc_ptr: TIPC specific data
1345 * @atalk_ptr: AppleTalk link
1346 * @ip_ptr: IPv4 specific data
1347 * @dn_ptr: DECnet specific data
1348 * @ip6_ptr: IPv6 specific data
1349 * @ax25_ptr: AX.25 specific data
1350 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1352 * @last_rx: Time of last Rx
1353 * @dev_addr: Hw address (before bcast,
1354 * because most packets are unicast)
1356 * @_rx: Array of RX queues
1357 * @num_rx_queues: Number of RX queues
1358 * allocated at register_netdev() time
1359 * @real_num_rx_queues: Number of RX queues currently active in device
1361 * @rx_handler: handler for received packets
1362 * @rx_handler_data: XXX: need comments on this one
1363 * @ingress_queue: XXX: need comments on this one
1364 * @broadcast: hw bcast address
1366 * @_tx: Array of TX queues
1367 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1368 * @real_num_tx_queues: Number of TX queues currently active in device
1369 * @qdisc: Root qdisc from userspace point of view
1370 * @tx_queue_len: Max frames per queue allowed
1371 * @tx_global_lock: XXX: need comments on this one
1373 * @xps_maps: XXX: need comments on this one
1375 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1376 * indexed by RX queue number. Assigned by driver.
1377 * This must only be set if the ndo_rx_flow_steer
1378 * operation is defined
1380 * @trans_start: Time (in jiffies) of last Tx
1381 * @watchdog_timeo: Represents the timeout that is used by
1382 * the watchdog ( see dev_watchdog() )
1383 * @watchdog_timer: List of timers
1385 * @pcpu_refcnt: Number of references to this device
1386 * @todo_list: Delayed register/unregister
1387 * @index_hlist: Device index hash chain
1388 * @link_watch_list: XXX: need comments on this one
1390 * @reg_state: Register/unregister state machine
1391 * @dismantle: Device is going to be freed
1392 * @rtnl_link_state: This enum represents the phases of creating
1393 * a new link
1395 * @destructor: Called from unregister,
1396 * can be used to call free_netdev
1397 * @npinfo: XXX: need comments on this one
1398 * @nd_net: Network namespace this network device is inside
1400 * @ml_priv: Mid-layer private
1401 * @lstats: Loopback statistics
1402 * @tstats: Tunnel statistics
1403 * @dstats: Dummy statistics
1404 * @vstats: Virtual ethernet statistics
1406 * @garp_port: GARP
1407 * @mrp_port: MRP
1409 * @dev: Class/net/name entry
1410 * @sysfs_groups: Space for optional device, statistics and wireless
1411 * sysfs groups
1413 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1414 * @rtnl_link_ops: Rtnl_link_ops
1416 * @gso_max_size: Maximum size of generic segmentation offload
1417 * @gso_max_segs: Maximum number of segments that can be passed to the
1418 * NIC for GSO
1420 * @dcbnl_ops: Data Center Bridging netlink ops
1421 * @num_tc: Number of traffic classes in the net device
1422 * @tc_to_txq: XXX: need comments on this one
1423 * @prio_tc_map XXX: need comments on this one
1425 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1427 * @priomap: XXX: need comments on this one
1428 * @phydev: Physical device may attach itself
1429 * for hardware timestamping
1431 * @qdisc_tx_busylock: XXX: need comments on this one
1433 * @group: The group, that the device belongs to
1434 * @pm_qos_req: Power Management QoS object
1436 * FIXME: cleanup struct net_device such that network protocol info
1437 * moves out.
1440 struct net_device {
1441 char name[IFNAMSIZ];
1442 struct hlist_node name_hlist;
1443 char *ifalias;
1445 * I/O specific fields
1446 * FIXME: Merge these and struct ifmap into one
1448 unsigned long mem_end;
1449 unsigned long mem_start;
1450 unsigned long base_addr;
1451 int irq;
1454 * Some hardware also needs these fields (state,dev_list,
1455 * napi_list,unreg_list,close_list) but they are not
1456 * part of the usual set specified in Space.c.
1459 unsigned long state;
1461 struct list_head dev_list;
1462 struct list_head napi_list;
1463 struct list_head unreg_list;
1464 struct list_head close_list;
1466 struct {
1467 struct list_head upper;
1468 struct list_head lower;
1469 } adj_list;
1471 struct {
1472 struct list_head upper;
1473 struct list_head lower;
1474 } all_adj_list;
1476 netdev_features_t features;
1477 netdev_features_t hw_features;
1478 netdev_features_t wanted_features;
1479 netdev_features_t vlan_features;
1480 netdev_features_t hw_enc_features;
1481 netdev_features_t mpls_features;
1483 int ifindex;
1484 int iflink;
1486 struct net_device_stats stats;
1488 atomic_long_t rx_dropped;
1489 atomic_long_t tx_dropped;
1491 atomic_t carrier_changes;
1493 #ifdef CONFIG_WIRELESS_EXT
1494 const struct iw_handler_def * wireless_handlers;
1495 struct iw_public_data * wireless_data;
1496 #endif
1497 const struct net_device_ops *netdev_ops;
1498 const struct ethtool_ops *ethtool_ops;
1499 const struct forwarding_accel_ops *fwd_ops;
1501 const struct header_ops *header_ops;
1503 unsigned int flags;
1504 unsigned int priv_flags;
1506 unsigned short gflags;
1507 unsigned short padded;
1509 unsigned char operstate;
1510 unsigned char link_mode;
1512 unsigned char if_port;
1513 unsigned char dma;
1515 unsigned int mtu;
1516 unsigned short type;
1517 unsigned short hard_header_len;
1519 unsigned short needed_headroom;
1520 unsigned short needed_tailroom;
1522 /* Interface address info. */
1523 unsigned char perm_addr[MAX_ADDR_LEN];
1524 unsigned char addr_assign_type;
1525 unsigned char addr_len;
1526 unsigned short neigh_priv_len;
1527 unsigned short dev_id;
1528 unsigned short dev_port;
1529 spinlock_t addr_list_lock;
1530 struct netdev_hw_addr_list uc;
1531 struct netdev_hw_addr_list mc;
1532 struct netdev_hw_addr_list dev_addrs;
1534 #ifdef CONFIG_SYSFS
1535 struct kset *queues_kset;
1536 #endif
1538 unsigned char name_assign_type;
1540 bool uc_promisc;
1541 unsigned int promiscuity;
1542 unsigned int allmulti;
1545 /* Protocol specific pointers */
1547 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1548 struct vlan_info __rcu *vlan_info;
1549 #endif
1550 #if IS_ENABLED(CONFIG_NET_DSA)
1551 struct dsa_switch_tree *dsa_ptr;
1552 #endif
1553 #if IS_ENABLED(CONFIG_TIPC)
1554 struct tipc_bearer __rcu *tipc_ptr;
1555 #endif
1556 void *atalk_ptr;
1557 struct in_device __rcu *ip_ptr;
1558 struct dn_dev __rcu *dn_ptr;
1559 struct inet6_dev __rcu *ip6_ptr;
1560 void *ax25_ptr;
1561 struct wireless_dev *ieee80211_ptr;
1564 * Cache lines mostly used on receive path (including eth_type_trans())
1566 unsigned long last_rx;
1568 /* Interface address info used in eth_type_trans() */
1569 unsigned char *dev_addr;
1572 #ifdef CONFIG_SYSFS
1573 struct netdev_rx_queue *_rx;
1575 unsigned int num_rx_queues;
1576 unsigned int real_num_rx_queues;
1578 #endif
1580 rx_handler_func_t __rcu *rx_handler;
1581 void __rcu *rx_handler_data;
1583 struct netdev_queue __rcu *ingress_queue;
1584 unsigned char broadcast[MAX_ADDR_LEN];
1588 * Cache lines mostly used on transmit path
1590 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1591 unsigned int num_tx_queues;
1592 unsigned int real_num_tx_queues;
1593 struct Qdisc *qdisc;
1594 unsigned long tx_queue_len;
1595 spinlock_t tx_global_lock;
1597 #ifdef CONFIG_XPS
1598 struct xps_dev_maps __rcu *xps_maps;
1599 #endif
1600 #ifdef CONFIG_RFS_ACCEL
1601 struct cpu_rmap *rx_cpu_rmap;
1602 #endif
1604 /* These may be needed for future network-power-down code. */
1607 * trans_start here is expensive for high speed devices on SMP,
1608 * please use netdev_queue->trans_start instead.
1610 unsigned long trans_start;
1612 int watchdog_timeo;
1613 struct timer_list watchdog_timer;
1615 int __percpu *pcpu_refcnt;
1616 struct list_head todo_list;
1618 struct hlist_node index_hlist;
1619 struct list_head link_watch_list;
1621 enum { NETREG_UNINITIALIZED=0,
1622 NETREG_REGISTERED, /* completed register_netdevice */
1623 NETREG_UNREGISTERING, /* called unregister_netdevice */
1624 NETREG_UNREGISTERED, /* completed unregister todo */
1625 NETREG_RELEASED, /* called free_netdev */
1626 NETREG_DUMMY, /* dummy device for NAPI poll */
1627 } reg_state:8;
1629 bool dismantle;
1631 enum {
1632 RTNL_LINK_INITIALIZED,
1633 RTNL_LINK_INITIALIZING,
1634 } rtnl_link_state:16;
1636 void (*destructor)(struct net_device *dev);
1638 #ifdef CONFIG_NETPOLL
1639 struct netpoll_info __rcu *npinfo;
1640 #endif
1642 #ifdef CONFIG_NET_NS
1643 struct net *nd_net;
1644 #endif
1646 /* mid-layer private */
1647 union {
1648 void *ml_priv;
1649 struct pcpu_lstats __percpu *lstats;
1650 struct pcpu_sw_netstats __percpu *tstats;
1651 struct pcpu_dstats __percpu *dstats;
1652 struct pcpu_vstats __percpu *vstats;
1655 struct garp_port __rcu *garp_port;
1656 struct mrp_port __rcu *mrp_port;
1658 struct device dev;
1659 const struct attribute_group *sysfs_groups[4];
1660 const struct attribute_group *sysfs_rx_queue_group;
1662 const struct rtnl_link_ops *rtnl_link_ops;
1664 /* for setting kernel sock attribute on TCP connection setup */
1665 #define GSO_MAX_SIZE 65536
1666 unsigned int gso_max_size;
1667 #define GSO_MAX_SEGS 65535
1668 u16 gso_max_segs;
1670 #ifdef CONFIG_DCB
1671 const struct dcbnl_rtnl_ops *dcbnl_ops;
1672 #endif
1673 u8 num_tc;
1674 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1675 u8 prio_tc_map[TC_BITMASK + 1];
1677 #if IS_ENABLED(CONFIG_FCOE)
1678 unsigned int fcoe_ddp_xid;
1679 #endif
1680 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1681 struct netprio_map __rcu *priomap;
1682 #endif
1683 struct phy_device *phydev;
1684 struct lock_class_key *qdisc_tx_busylock;
1685 int group;
1686 struct pm_qos_request pm_qos_req;
1688 #define to_net_dev(d) container_of(d, struct net_device, dev)
1690 #define NETDEV_ALIGN 32
1692 static inline
1693 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1695 return dev->prio_tc_map[prio & TC_BITMASK];
1698 static inline
1699 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1701 if (tc >= dev->num_tc)
1702 return -EINVAL;
1704 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1705 return 0;
1708 static inline
1709 void netdev_reset_tc(struct net_device *dev)
1711 dev->num_tc = 0;
1712 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1713 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1716 static inline
1717 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1719 if (tc >= dev->num_tc)
1720 return -EINVAL;
1722 dev->tc_to_txq[tc].count = count;
1723 dev->tc_to_txq[tc].offset = offset;
1724 return 0;
1727 static inline
1728 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1730 if (num_tc > TC_MAX_QUEUE)
1731 return -EINVAL;
1733 dev->num_tc = num_tc;
1734 return 0;
1737 static inline
1738 int netdev_get_num_tc(struct net_device *dev)
1740 return dev->num_tc;
1743 static inline
1744 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1745 unsigned int index)
1747 return &dev->_tx[index];
1750 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1751 const struct sk_buff *skb)
1753 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1756 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1757 void (*f)(struct net_device *,
1758 struct netdev_queue *,
1759 void *),
1760 void *arg)
1762 unsigned int i;
1764 for (i = 0; i < dev->num_tx_queues; i++)
1765 f(dev, &dev->_tx[i], arg);
1768 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1769 struct sk_buff *skb,
1770 void *accel_priv);
1773 * Net namespace inlines
1775 static inline
1776 struct net *dev_net(const struct net_device *dev)
1778 return read_pnet(&dev->nd_net);
1781 static inline
1782 void dev_net_set(struct net_device *dev, struct net *net)
1784 #ifdef CONFIG_NET_NS
1785 release_net(dev->nd_net);
1786 dev->nd_net = hold_net(net);
1787 #endif
1790 static inline bool netdev_uses_dsa(struct net_device *dev)
1792 #if IS_ENABLED(CONFIG_NET_DSA)
1793 if (dev->dsa_ptr != NULL)
1794 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1795 #endif
1796 return false;
1800 * netdev_priv - access network device private data
1801 * @dev: network device
1803 * Get network device private data
1805 static inline void *netdev_priv(const struct net_device *dev)
1807 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1810 /* Set the sysfs physical device reference for the network logical device
1811 * if set prior to registration will cause a symlink during initialization.
1813 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1815 /* Set the sysfs device type for the network logical device to allow
1816 * fine-grained identification of different network device types. For
1817 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1819 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1821 /* Default NAPI poll() weight
1822 * Device drivers are strongly advised to not use bigger value
1824 #define NAPI_POLL_WEIGHT 64
1827 * netif_napi_add - initialize a napi context
1828 * @dev: network device
1829 * @napi: napi context
1830 * @poll: polling function
1831 * @weight: default weight
1833 * netif_napi_add() must be used to initialize a napi context prior to calling
1834 * *any* of the other napi related functions.
1836 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1837 int (*poll)(struct napi_struct *, int), int weight);
1840 * netif_napi_del - remove a napi context
1841 * @napi: napi context
1843 * netif_napi_del() removes a napi context from the network device napi list
1845 void netif_napi_del(struct napi_struct *napi);
1847 struct napi_gro_cb {
1848 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1849 void *frag0;
1851 /* Length of frag0. */
1852 unsigned int frag0_len;
1854 /* This indicates where we are processing relative to skb->data. */
1855 int data_offset;
1857 /* This is non-zero if the packet cannot be merged with the new skb. */
1858 u16 flush;
1860 /* Save the IP ID here and check when we get to the transport layer */
1861 u16 flush_id;
1863 /* Number of segments aggregated. */
1864 u16 count;
1866 /* This is non-zero if the packet may be of the same flow. */
1867 u8 same_flow;
1869 /* Free the skb? */
1870 u8 free;
1871 #define NAPI_GRO_FREE 1
1872 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1874 /* jiffies when first packet was created/queued */
1875 unsigned long age;
1877 /* Used in ipv6_gro_receive() */
1878 u16 proto;
1880 /* Used in udp_gro_receive */
1881 u8 udp_mark:1;
1883 /* GRO checksum is valid */
1884 u8 csum_valid:1;
1886 /* Number of checksums via CHECKSUM_UNNECESSARY */
1887 u8 csum_cnt:3;
1889 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1890 __wsum csum;
1892 /* used in skb_gro_receive() slow path */
1893 struct sk_buff *last;
1896 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1898 struct packet_type {
1899 __be16 type; /* This is really htons(ether_type). */
1900 struct net_device *dev; /* NULL is wildcarded here */
1901 int (*func) (struct sk_buff *,
1902 struct net_device *,
1903 struct packet_type *,
1904 struct net_device *);
1905 bool (*id_match)(struct packet_type *ptype,
1906 struct sock *sk);
1907 void *af_packet_priv;
1908 struct list_head list;
1911 struct offload_callbacks {
1912 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1913 netdev_features_t features);
1914 int (*gso_send_check)(struct sk_buff *skb);
1915 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1916 struct sk_buff *skb);
1917 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1920 struct packet_offload {
1921 __be16 type; /* This is really htons(ether_type). */
1922 struct offload_callbacks callbacks;
1923 struct list_head list;
1926 struct udp_offload {
1927 __be16 port;
1928 struct offload_callbacks callbacks;
1931 struct dsa_device_ops {
1932 netdev_tx_t (*xmit)(struct sk_buff *skb, struct net_device *dev);
1933 int (*rcv)(struct sk_buff *skb, struct net_device *dev,
1934 struct packet_type *pt, struct net_device *orig_dev);
1938 /* often modified stats are per cpu, other are shared (netdev->stats) */
1939 struct pcpu_sw_netstats {
1940 u64 rx_packets;
1941 u64 rx_bytes;
1942 u64 tx_packets;
1943 u64 tx_bytes;
1944 struct u64_stats_sync syncp;
1947 #define netdev_alloc_pcpu_stats(type) \
1948 ({ \
1949 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1950 if (pcpu_stats) { \
1951 int i; \
1952 for_each_possible_cpu(i) { \
1953 typeof(type) *stat; \
1954 stat = per_cpu_ptr(pcpu_stats, i); \
1955 u64_stats_init(&stat->syncp); \
1958 pcpu_stats; \
1961 #include <linux/notifier.h>
1963 /* netdevice notifier chain. Please remember to update the rtnetlink
1964 * notification exclusion list in rtnetlink_event() when adding new
1965 * types.
1967 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1968 #define NETDEV_DOWN 0x0002
1969 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1970 detected a hardware crash and restarted
1971 - we can use this eg to kick tcp sessions
1972 once done */
1973 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1974 #define NETDEV_REGISTER 0x0005
1975 #define NETDEV_UNREGISTER 0x0006
1976 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
1977 #define NETDEV_CHANGEADDR 0x0008
1978 #define NETDEV_GOING_DOWN 0x0009
1979 #define NETDEV_CHANGENAME 0x000A
1980 #define NETDEV_FEAT_CHANGE 0x000B
1981 #define NETDEV_BONDING_FAILOVER 0x000C
1982 #define NETDEV_PRE_UP 0x000D
1983 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1984 #define NETDEV_POST_TYPE_CHANGE 0x000F
1985 #define NETDEV_POST_INIT 0x0010
1986 #define NETDEV_UNREGISTER_FINAL 0x0011
1987 #define NETDEV_RELEASE 0x0012
1988 #define NETDEV_NOTIFY_PEERS 0x0013
1989 #define NETDEV_JOIN 0x0014
1990 #define NETDEV_CHANGEUPPER 0x0015
1991 #define NETDEV_RESEND_IGMP 0x0016
1992 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
1993 #define NETDEV_CHANGEINFODATA 0x0018
1995 int register_netdevice_notifier(struct notifier_block *nb);
1996 int unregister_netdevice_notifier(struct notifier_block *nb);
1998 struct netdev_notifier_info {
1999 struct net_device *dev;
2002 struct netdev_notifier_change_info {
2003 struct netdev_notifier_info info; /* must be first */
2004 unsigned int flags_changed;
2007 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2008 struct net_device *dev)
2010 info->dev = dev;
2013 static inline struct net_device *
2014 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2016 return info->dev;
2019 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2022 extern rwlock_t dev_base_lock; /* Device list lock */
2024 #define for_each_netdev(net, d) \
2025 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2026 #define for_each_netdev_reverse(net, d) \
2027 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2028 #define for_each_netdev_rcu(net, d) \
2029 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2030 #define for_each_netdev_safe(net, d, n) \
2031 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2032 #define for_each_netdev_continue(net, d) \
2033 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2034 #define for_each_netdev_continue_rcu(net, d) \
2035 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2036 #define for_each_netdev_in_bond_rcu(bond, slave) \
2037 for_each_netdev_rcu(&init_net, slave) \
2038 if (netdev_master_upper_dev_get_rcu(slave) == bond)
2039 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2041 static inline struct net_device *next_net_device(struct net_device *dev)
2043 struct list_head *lh;
2044 struct net *net;
2046 net = dev_net(dev);
2047 lh = dev->dev_list.next;
2048 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2051 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2053 struct list_head *lh;
2054 struct net *net;
2056 net = dev_net(dev);
2057 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2058 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2061 static inline struct net_device *first_net_device(struct net *net)
2063 return list_empty(&net->dev_base_head) ? NULL :
2064 net_device_entry(net->dev_base_head.next);
2067 static inline struct net_device *first_net_device_rcu(struct net *net)
2069 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2071 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2074 int netdev_boot_setup_check(struct net_device *dev);
2075 unsigned long netdev_boot_base(const char *prefix, int unit);
2076 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2077 const char *hwaddr);
2078 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2079 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2080 void dev_add_pack(struct packet_type *pt);
2081 void dev_remove_pack(struct packet_type *pt);
2082 void __dev_remove_pack(struct packet_type *pt);
2083 void dev_add_offload(struct packet_offload *po);
2084 void dev_remove_offload(struct packet_offload *po);
2086 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2087 unsigned short mask);
2088 struct net_device *dev_get_by_name(struct net *net, const char *name);
2089 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2090 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2091 int dev_alloc_name(struct net_device *dev, const char *name);
2092 int dev_open(struct net_device *dev);
2093 int dev_close(struct net_device *dev);
2094 void dev_disable_lro(struct net_device *dev);
2095 int dev_loopback_xmit(struct sk_buff *newskb);
2096 int dev_queue_xmit(struct sk_buff *skb);
2097 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2098 int register_netdevice(struct net_device *dev);
2099 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2100 void unregister_netdevice_many(struct list_head *head);
2101 static inline void unregister_netdevice(struct net_device *dev)
2103 unregister_netdevice_queue(dev, NULL);
2106 int netdev_refcnt_read(const struct net_device *dev);
2107 void free_netdev(struct net_device *dev);
2108 void netdev_freemem(struct net_device *dev);
2109 void synchronize_net(void);
2110 int init_dummy_netdev(struct net_device *dev);
2112 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2113 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2114 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2115 int netdev_get_name(struct net *net, char *name, int ifindex);
2116 int dev_restart(struct net_device *dev);
2117 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2119 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2121 return NAPI_GRO_CB(skb)->data_offset;
2124 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2126 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2129 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2131 NAPI_GRO_CB(skb)->data_offset += len;
2134 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2135 unsigned int offset)
2137 return NAPI_GRO_CB(skb)->frag0 + offset;
2140 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2142 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2145 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2146 unsigned int offset)
2148 if (!pskb_may_pull(skb, hlen))
2149 return NULL;
2151 NAPI_GRO_CB(skb)->frag0 = NULL;
2152 NAPI_GRO_CB(skb)->frag0_len = 0;
2153 return skb->data + offset;
2156 static inline void *skb_gro_network_header(struct sk_buff *skb)
2158 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2159 skb_network_offset(skb);
2162 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2163 const void *start, unsigned int len)
2165 if (NAPI_GRO_CB(skb)->csum_valid)
2166 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2167 csum_partial(start, len, 0));
2170 /* GRO checksum functions. These are logical equivalents of the normal
2171 * checksum functions (in skbuff.h) except that they operate on the GRO
2172 * offsets and fields in sk_buff.
2175 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2177 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2178 bool zero_okay,
2179 __sum16 check)
2181 return (skb->ip_summed != CHECKSUM_PARTIAL &&
2182 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2183 (!zero_okay || check));
2186 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2187 __wsum psum)
2189 if (NAPI_GRO_CB(skb)->csum_valid &&
2190 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2191 return 0;
2193 NAPI_GRO_CB(skb)->csum = psum;
2195 return __skb_gro_checksum_complete(skb);
2198 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2200 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2201 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2202 NAPI_GRO_CB(skb)->csum_cnt--;
2203 } else {
2204 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2205 * verified a new top level checksum or an encapsulated one
2206 * during GRO. This saves work if we fallback to normal path.
2208 __skb_incr_checksum_unnecessary(skb);
2212 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2213 compute_pseudo) \
2214 ({ \
2215 __sum16 __ret = 0; \
2216 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2217 __ret = __skb_gro_checksum_validate_complete(skb, \
2218 compute_pseudo(skb, proto)); \
2219 if (__ret) \
2220 __skb_mark_checksum_bad(skb); \
2221 else \
2222 skb_gro_incr_csum_unnecessary(skb); \
2223 __ret; \
2226 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2227 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2229 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2230 compute_pseudo) \
2231 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2233 #define skb_gro_checksum_simple_validate(skb) \
2234 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2236 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2238 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2239 !NAPI_GRO_CB(skb)->csum_valid);
2242 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2243 __sum16 check, __wsum pseudo)
2245 NAPI_GRO_CB(skb)->csum = ~pseudo;
2246 NAPI_GRO_CB(skb)->csum_valid = 1;
2249 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2250 do { \
2251 if (__skb_gro_checksum_convert_check(skb)) \
2252 __skb_gro_checksum_convert(skb, check, \
2253 compute_pseudo(skb, proto)); \
2254 } while (0)
2256 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2257 unsigned short type,
2258 const void *daddr, const void *saddr,
2259 unsigned int len)
2261 if (!dev->header_ops || !dev->header_ops->create)
2262 return 0;
2264 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2267 static inline int dev_parse_header(const struct sk_buff *skb,
2268 unsigned char *haddr)
2270 const struct net_device *dev = skb->dev;
2272 if (!dev->header_ops || !dev->header_ops->parse)
2273 return 0;
2274 return dev->header_ops->parse(skb, haddr);
2277 static inline int dev_rebuild_header(struct sk_buff *skb)
2279 const struct net_device *dev = skb->dev;
2281 if (!dev->header_ops || !dev->header_ops->rebuild)
2282 return 0;
2283 return dev->header_ops->rebuild(skb);
2286 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2287 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2288 static inline int unregister_gifconf(unsigned int family)
2290 return register_gifconf(family, NULL);
2293 #ifdef CONFIG_NET_FLOW_LIMIT
2294 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2295 struct sd_flow_limit {
2296 u64 count;
2297 unsigned int num_buckets;
2298 unsigned int history_head;
2299 u16 history[FLOW_LIMIT_HISTORY];
2300 u8 buckets[];
2303 extern int netdev_flow_limit_table_len;
2304 #endif /* CONFIG_NET_FLOW_LIMIT */
2307 * Incoming packets are placed on per-cpu queues
2309 struct softnet_data {
2310 struct Qdisc *output_queue;
2311 struct Qdisc **output_queue_tailp;
2312 struct list_head poll_list;
2313 struct sk_buff *completion_queue;
2314 struct sk_buff_head process_queue;
2316 /* stats */
2317 unsigned int processed;
2318 unsigned int time_squeeze;
2319 unsigned int cpu_collision;
2320 unsigned int received_rps;
2322 #ifdef CONFIG_RPS
2323 struct softnet_data *rps_ipi_list;
2325 /* Elements below can be accessed between CPUs for RPS */
2326 struct call_single_data csd ____cacheline_aligned_in_smp;
2327 struct softnet_data *rps_ipi_next;
2328 unsigned int cpu;
2329 unsigned int input_queue_head;
2330 unsigned int input_queue_tail;
2331 #endif
2332 unsigned int dropped;
2333 struct sk_buff_head input_pkt_queue;
2334 struct napi_struct backlog;
2336 #ifdef CONFIG_NET_FLOW_LIMIT
2337 struct sd_flow_limit __rcu *flow_limit;
2338 #endif
2341 static inline void input_queue_head_incr(struct softnet_data *sd)
2343 #ifdef CONFIG_RPS
2344 sd->input_queue_head++;
2345 #endif
2348 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2349 unsigned int *qtail)
2351 #ifdef CONFIG_RPS
2352 *qtail = ++sd->input_queue_tail;
2353 #endif
2356 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2358 void __netif_schedule(struct Qdisc *q);
2359 void netif_schedule_queue(struct netdev_queue *txq);
2361 static inline void netif_tx_schedule_all(struct net_device *dev)
2363 unsigned int i;
2365 for (i = 0; i < dev->num_tx_queues; i++)
2366 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2369 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2371 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2375 * netif_start_queue - allow transmit
2376 * @dev: network device
2378 * Allow upper layers to call the device hard_start_xmit routine.
2380 static inline void netif_start_queue(struct net_device *dev)
2382 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2385 static inline void netif_tx_start_all_queues(struct net_device *dev)
2387 unsigned int i;
2389 for (i = 0; i < dev->num_tx_queues; i++) {
2390 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2391 netif_tx_start_queue(txq);
2395 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2398 * netif_wake_queue - restart transmit
2399 * @dev: network device
2401 * Allow upper layers to call the device hard_start_xmit routine.
2402 * Used for flow control when transmit resources are available.
2404 static inline void netif_wake_queue(struct net_device *dev)
2406 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2409 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2411 unsigned int i;
2413 for (i = 0; i < dev->num_tx_queues; i++) {
2414 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2415 netif_tx_wake_queue(txq);
2419 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2421 if (WARN_ON(!dev_queue)) {
2422 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2423 return;
2425 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2429 * netif_stop_queue - stop transmitted packets
2430 * @dev: network device
2432 * Stop upper layers calling the device hard_start_xmit routine.
2433 * Used for flow control when transmit resources are unavailable.
2435 static inline void netif_stop_queue(struct net_device *dev)
2437 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2440 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2442 unsigned int i;
2444 for (i = 0; i < dev->num_tx_queues; i++) {
2445 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2446 netif_tx_stop_queue(txq);
2450 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2452 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2456 * netif_queue_stopped - test if transmit queue is flowblocked
2457 * @dev: network device
2459 * Test if transmit queue on device is currently unable to send.
2461 static inline bool netif_queue_stopped(const struct net_device *dev)
2463 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2466 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2468 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2471 static inline bool
2472 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2474 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2477 static inline bool
2478 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2480 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2483 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2484 unsigned int bytes)
2486 #ifdef CONFIG_BQL
2487 dql_queued(&dev_queue->dql, bytes);
2489 if (likely(dql_avail(&dev_queue->dql) >= 0))
2490 return;
2492 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2495 * The XOFF flag must be set before checking the dql_avail below,
2496 * because in netdev_tx_completed_queue we update the dql_completed
2497 * before checking the XOFF flag.
2499 smp_mb();
2501 /* check again in case another CPU has just made room avail */
2502 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2503 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2504 #endif
2508 * netdev_sent_queue - report the number of bytes queued to hardware
2509 * @dev: network device
2510 * @bytes: number of bytes queued to the hardware device queue
2512 * Report the number of bytes queued for sending/completion to the network
2513 * device hardware queue. @bytes should be a good approximation and should
2514 * exactly match netdev_completed_queue() @bytes
2516 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2518 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2521 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2522 unsigned int pkts, unsigned int bytes)
2524 #ifdef CONFIG_BQL
2525 if (unlikely(!bytes))
2526 return;
2528 dql_completed(&dev_queue->dql, bytes);
2531 * Without the memory barrier there is a small possiblity that
2532 * netdev_tx_sent_queue will miss the update and cause the queue to
2533 * be stopped forever
2535 smp_mb();
2537 if (dql_avail(&dev_queue->dql) < 0)
2538 return;
2540 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2541 netif_schedule_queue(dev_queue);
2542 #endif
2546 * netdev_completed_queue - report bytes and packets completed by device
2547 * @dev: network device
2548 * @pkts: actual number of packets sent over the medium
2549 * @bytes: actual number of bytes sent over the medium
2551 * Report the number of bytes and packets transmitted by the network device
2552 * hardware queue over the physical medium, @bytes must exactly match the
2553 * @bytes amount passed to netdev_sent_queue()
2555 static inline void netdev_completed_queue(struct net_device *dev,
2556 unsigned int pkts, unsigned int bytes)
2558 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2561 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2563 #ifdef CONFIG_BQL
2564 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2565 dql_reset(&q->dql);
2566 #endif
2570 * netdev_reset_queue - reset the packets and bytes count of a network device
2571 * @dev_queue: network device
2573 * Reset the bytes and packet count of a network device and clear the
2574 * software flow control OFF bit for this network device
2576 static inline void netdev_reset_queue(struct net_device *dev_queue)
2578 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2582 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2583 * @dev: network device
2584 * @queue_index: given tx queue index
2586 * Returns 0 if given tx queue index >= number of device tx queues,
2587 * otherwise returns the originally passed tx queue index.
2589 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2591 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2592 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2593 dev->name, queue_index,
2594 dev->real_num_tx_queues);
2595 return 0;
2598 return queue_index;
2602 * netif_running - test if up
2603 * @dev: network device
2605 * Test if the device has been brought up.
2607 static inline bool netif_running(const struct net_device *dev)
2609 return test_bit(__LINK_STATE_START, &dev->state);
2613 * Routines to manage the subqueues on a device. We only need start
2614 * stop, and a check if it's stopped. All other device management is
2615 * done at the overall netdevice level.
2616 * Also test the device if we're multiqueue.
2620 * netif_start_subqueue - allow sending packets on subqueue
2621 * @dev: network device
2622 * @queue_index: sub queue index
2624 * Start individual transmit queue of a device with multiple transmit queues.
2626 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2628 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2630 netif_tx_start_queue(txq);
2634 * netif_stop_subqueue - stop sending packets on subqueue
2635 * @dev: network device
2636 * @queue_index: sub queue index
2638 * Stop individual transmit queue of a device with multiple transmit queues.
2640 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2642 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2643 netif_tx_stop_queue(txq);
2647 * netif_subqueue_stopped - test status of subqueue
2648 * @dev: network device
2649 * @queue_index: sub queue index
2651 * Check individual transmit queue of a device with multiple transmit queues.
2653 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2654 u16 queue_index)
2656 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2658 return netif_tx_queue_stopped(txq);
2661 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2662 struct sk_buff *skb)
2664 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2667 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2669 #ifdef CONFIG_XPS
2670 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2671 u16 index);
2672 #else
2673 static inline int netif_set_xps_queue(struct net_device *dev,
2674 const struct cpumask *mask,
2675 u16 index)
2677 return 0;
2679 #endif
2682 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2683 * as a distribution range limit for the returned value.
2685 static inline u16 skb_tx_hash(const struct net_device *dev,
2686 struct sk_buff *skb)
2688 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2692 * netif_is_multiqueue - test if device has multiple transmit queues
2693 * @dev: network device
2695 * Check if device has multiple transmit queues
2697 static inline bool netif_is_multiqueue(const struct net_device *dev)
2699 return dev->num_tx_queues > 1;
2702 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2704 #ifdef CONFIG_SYSFS
2705 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2706 #else
2707 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2708 unsigned int rxq)
2710 return 0;
2712 #endif
2714 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2715 const struct net_device *from_dev)
2717 int err;
2719 err = netif_set_real_num_tx_queues(to_dev,
2720 from_dev->real_num_tx_queues);
2721 if (err)
2722 return err;
2723 #ifdef CONFIG_SYSFS
2724 return netif_set_real_num_rx_queues(to_dev,
2725 from_dev->real_num_rx_queues);
2726 #else
2727 return 0;
2728 #endif
2731 #ifdef CONFIG_SYSFS
2732 static inline unsigned int get_netdev_rx_queue_index(
2733 struct netdev_rx_queue *queue)
2735 struct net_device *dev = queue->dev;
2736 int index = queue - dev->_rx;
2738 BUG_ON(index >= dev->num_rx_queues);
2739 return index;
2741 #endif
2743 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2744 int netif_get_num_default_rss_queues(void);
2746 enum skb_free_reason {
2747 SKB_REASON_CONSUMED,
2748 SKB_REASON_DROPPED,
2751 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2752 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2755 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2756 * interrupt context or with hardware interrupts being disabled.
2757 * (in_irq() || irqs_disabled())
2759 * We provide four helpers that can be used in following contexts :
2761 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2762 * replacing kfree_skb(skb)
2764 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2765 * Typically used in place of consume_skb(skb) in TX completion path
2767 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2768 * replacing kfree_skb(skb)
2770 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2771 * and consumed a packet. Used in place of consume_skb(skb)
2773 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2775 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2778 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2780 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2783 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2785 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2788 static inline void dev_consume_skb_any(struct sk_buff *skb)
2790 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2793 int netif_rx(struct sk_buff *skb);
2794 int netif_rx_ni(struct sk_buff *skb);
2795 int netif_receive_skb(struct sk_buff *skb);
2796 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2797 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2798 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2799 gro_result_t napi_gro_frags(struct napi_struct *napi);
2800 struct packet_offload *gro_find_receive_by_type(__be16 type);
2801 struct packet_offload *gro_find_complete_by_type(__be16 type);
2803 static inline void napi_free_frags(struct napi_struct *napi)
2805 kfree_skb(napi->skb);
2806 napi->skb = NULL;
2809 int netdev_rx_handler_register(struct net_device *dev,
2810 rx_handler_func_t *rx_handler,
2811 void *rx_handler_data);
2812 void netdev_rx_handler_unregister(struct net_device *dev);
2814 bool dev_valid_name(const char *name);
2815 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2816 int dev_ethtool(struct net *net, struct ifreq *);
2817 unsigned int dev_get_flags(const struct net_device *);
2818 int __dev_change_flags(struct net_device *, unsigned int flags);
2819 int dev_change_flags(struct net_device *, unsigned int);
2820 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2821 unsigned int gchanges);
2822 int dev_change_name(struct net_device *, const char *);
2823 int dev_set_alias(struct net_device *, const char *, size_t);
2824 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2825 int dev_set_mtu(struct net_device *, int);
2826 void dev_set_group(struct net_device *, int);
2827 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2828 int dev_change_carrier(struct net_device *, bool new_carrier);
2829 int dev_get_phys_port_id(struct net_device *dev,
2830 struct netdev_phys_port_id *ppid);
2831 struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev);
2832 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2833 struct netdev_queue *txq, int *ret);
2834 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2835 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2836 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2838 extern int netdev_budget;
2840 /* Called by rtnetlink.c:rtnl_unlock() */
2841 void netdev_run_todo(void);
2844 * dev_put - release reference to device
2845 * @dev: network device
2847 * Release reference to device to allow it to be freed.
2849 static inline void dev_put(struct net_device *dev)
2851 this_cpu_dec(*dev->pcpu_refcnt);
2855 * dev_hold - get reference to device
2856 * @dev: network device
2858 * Hold reference to device to keep it from being freed.
2860 static inline void dev_hold(struct net_device *dev)
2862 this_cpu_inc(*dev->pcpu_refcnt);
2865 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2866 * and _off may be called from IRQ context, but it is caller
2867 * who is responsible for serialization of these calls.
2869 * The name carrier is inappropriate, these functions should really be
2870 * called netif_lowerlayer_*() because they represent the state of any
2871 * kind of lower layer not just hardware media.
2874 void linkwatch_init_dev(struct net_device *dev);
2875 void linkwatch_fire_event(struct net_device *dev);
2876 void linkwatch_forget_dev(struct net_device *dev);
2879 * netif_carrier_ok - test if carrier present
2880 * @dev: network device
2882 * Check if carrier is present on device
2884 static inline bool netif_carrier_ok(const struct net_device *dev)
2886 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2889 unsigned long dev_trans_start(struct net_device *dev);
2891 void __netdev_watchdog_up(struct net_device *dev);
2893 void netif_carrier_on(struct net_device *dev);
2895 void netif_carrier_off(struct net_device *dev);
2898 * netif_dormant_on - mark device as dormant.
2899 * @dev: network device
2901 * Mark device as dormant (as per RFC2863).
2903 * The dormant state indicates that the relevant interface is not
2904 * actually in a condition to pass packets (i.e., it is not 'up') but is
2905 * in a "pending" state, waiting for some external event. For "on-
2906 * demand" interfaces, this new state identifies the situation where the
2907 * interface is waiting for events to place it in the up state.
2910 static inline void netif_dormant_on(struct net_device *dev)
2912 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2913 linkwatch_fire_event(dev);
2917 * netif_dormant_off - set device as not dormant.
2918 * @dev: network device
2920 * Device is not in dormant state.
2922 static inline void netif_dormant_off(struct net_device *dev)
2924 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2925 linkwatch_fire_event(dev);
2929 * netif_dormant - test if carrier present
2930 * @dev: network device
2932 * Check if carrier is present on device
2934 static inline bool netif_dormant(const struct net_device *dev)
2936 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2941 * netif_oper_up - test if device is operational
2942 * @dev: network device
2944 * Check if carrier is operational
2946 static inline bool netif_oper_up(const struct net_device *dev)
2948 return (dev->operstate == IF_OPER_UP ||
2949 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2953 * netif_device_present - is device available or removed
2954 * @dev: network device
2956 * Check if device has not been removed from system.
2958 static inline bool netif_device_present(struct net_device *dev)
2960 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2963 void netif_device_detach(struct net_device *dev);
2965 void netif_device_attach(struct net_device *dev);
2968 * Network interface message level settings
2971 enum {
2972 NETIF_MSG_DRV = 0x0001,
2973 NETIF_MSG_PROBE = 0x0002,
2974 NETIF_MSG_LINK = 0x0004,
2975 NETIF_MSG_TIMER = 0x0008,
2976 NETIF_MSG_IFDOWN = 0x0010,
2977 NETIF_MSG_IFUP = 0x0020,
2978 NETIF_MSG_RX_ERR = 0x0040,
2979 NETIF_MSG_TX_ERR = 0x0080,
2980 NETIF_MSG_TX_QUEUED = 0x0100,
2981 NETIF_MSG_INTR = 0x0200,
2982 NETIF_MSG_TX_DONE = 0x0400,
2983 NETIF_MSG_RX_STATUS = 0x0800,
2984 NETIF_MSG_PKTDATA = 0x1000,
2985 NETIF_MSG_HW = 0x2000,
2986 NETIF_MSG_WOL = 0x4000,
2989 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2990 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2991 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2992 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2993 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2994 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2995 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2996 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2997 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2998 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2999 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3000 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3001 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3002 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3003 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3005 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3007 /* use default */
3008 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3009 return default_msg_enable_bits;
3010 if (debug_value == 0) /* no output */
3011 return 0;
3012 /* set low N bits */
3013 return (1 << debug_value) - 1;
3016 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3018 spin_lock(&txq->_xmit_lock);
3019 txq->xmit_lock_owner = cpu;
3022 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3024 spin_lock_bh(&txq->_xmit_lock);
3025 txq->xmit_lock_owner = smp_processor_id();
3028 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3030 bool ok = spin_trylock(&txq->_xmit_lock);
3031 if (likely(ok))
3032 txq->xmit_lock_owner = smp_processor_id();
3033 return ok;
3036 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3038 txq->xmit_lock_owner = -1;
3039 spin_unlock(&txq->_xmit_lock);
3042 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3044 txq->xmit_lock_owner = -1;
3045 spin_unlock_bh(&txq->_xmit_lock);
3048 static inline void txq_trans_update(struct netdev_queue *txq)
3050 if (txq->xmit_lock_owner != -1)
3051 txq->trans_start = jiffies;
3055 * netif_tx_lock - grab network device transmit lock
3056 * @dev: network device
3058 * Get network device transmit lock
3060 static inline void netif_tx_lock(struct net_device *dev)
3062 unsigned int i;
3063 int cpu;
3065 spin_lock(&dev->tx_global_lock);
3066 cpu = smp_processor_id();
3067 for (i = 0; i < dev->num_tx_queues; i++) {
3068 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3070 /* We are the only thread of execution doing a
3071 * freeze, but we have to grab the _xmit_lock in
3072 * order to synchronize with threads which are in
3073 * the ->hard_start_xmit() handler and already
3074 * checked the frozen bit.
3076 __netif_tx_lock(txq, cpu);
3077 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3078 __netif_tx_unlock(txq);
3082 static inline void netif_tx_lock_bh(struct net_device *dev)
3084 local_bh_disable();
3085 netif_tx_lock(dev);
3088 static inline void netif_tx_unlock(struct net_device *dev)
3090 unsigned int i;
3092 for (i = 0; i < dev->num_tx_queues; i++) {
3093 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3095 /* No need to grab the _xmit_lock here. If the
3096 * queue is not stopped for another reason, we
3097 * force a schedule.
3099 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3100 netif_schedule_queue(txq);
3102 spin_unlock(&dev->tx_global_lock);
3105 static inline void netif_tx_unlock_bh(struct net_device *dev)
3107 netif_tx_unlock(dev);
3108 local_bh_enable();
3111 #define HARD_TX_LOCK(dev, txq, cpu) { \
3112 if ((dev->features & NETIF_F_LLTX) == 0) { \
3113 __netif_tx_lock(txq, cpu); \
3117 #define HARD_TX_TRYLOCK(dev, txq) \
3118 (((dev->features & NETIF_F_LLTX) == 0) ? \
3119 __netif_tx_trylock(txq) : \
3120 true )
3122 #define HARD_TX_UNLOCK(dev, txq) { \
3123 if ((dev->features & NETIF_F_LLTX) == 0) { \
3124 __netif_tx_unlock(txq); \
3128 static inline void netif_tx_disable(struct net_device *dev)
3130 unsigned int i;
3131 int cpu;
3133 local_bh_disable();
3134 cpu = smp_processor_id();
3135 for (i = 0; i < dev->num_tx_queues; i++) {
3136 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3138 __netif_tx_lock(txq, cpu);
3139 netif_tx_stop_queue(txq);
3140 __netif_tx_unlock(txq);
3142 local_bh_enable();
3145 static inline void netif_addr_lock(struct net_device *dev)
3147 spin_lock(&dev->addr_list_lock);
3150 static inline void netif_addr_lock_nested(struct net_device *dev)
3152 int subclass = SINGLE_DEPTH_NESTING;
3154 if (dev->netdev_ops->ndo_get_lock_subclass)
3155 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3157 spin_lock_nested(&dev->addr_list_lock, subclass);
3160 static inline void netif_addr_lock_bh(struct net_device *dev)
3162 spin_lock_bh(&dev->addr_list_lock);
3165 static inline void netif_addr_unlock(struct net_device *dev)
3167 spin_unlock(&dev->addr_list_lock);
3170 static inline void netif_addr_unlock_bh(struct net_device *dev)
3172 spin_unlock_bh(&dev->addr_list_lock);
3176 * dev_addrs walker. Should be used only for read access. Call with
3177 * rcu_read_lock held.
3179 #define for_each_dev_addr(dev, ha) \
3180 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3182 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3184 void ether_setup(struct net_device *dev);
3186 /* Support for loadable net-drivers */
3187 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3188 unsigned char name_assign_type,
3189 void (*setup)(struct net_device *),
3190 unsigned int txqs, unsigned int rxqs);
3191 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3192 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3194 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3195 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3196 count)
3198 int register_netdev(struct net_device *dev);
3199 void unregister_netdev(struct net_device *dev);
3201 /* General hardware address lists handling functions */
3202 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3203 struct netdev_hw_addr_list *from_list, int addr_len);
3204 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3205 struct netdev_hw_addr_list *from_list, int addr_len);
3206 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3207 struct net_device *dev,
3208 int (*sync)(struct net_device *, const unsigned char *),
3209 int (*unsync)(struct net_device *,
3210 const unsigned char *));
3211 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3212 struct net_device *dev,
3213 int (*unsync)(struct net_device *,
3214 const unsigned char *));
3215 void __hw_addr_init(struct netdev_hw_addr_list *list);
3217 /* Functions used for device addresses handling */
3218 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3219 unsigned char addr_type);
3220 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3221 unsigned char addr_type);
3222 void dev_addr_flush(struct net_device *dev);
3223 int dev_addr_init(struct net_device *dev);
3225 /* Functions used for unicast addresses handling */
3226 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3227 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3228 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3229 int dev_uc_sync(struct net_device *to, struct net_device *from);
3230 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3231 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3232 void dev_uc_flush(struct net_device *dev);
3233 void dev_uc_init(struct net_device *dev);
3236 * __dev_uc_sync - Synchonize device's unicast list
3237 * @dev: device to sync
3238 * @sync: function to call if address should be added
3239 * @unsync: function to call if address should be removed
3241 * Add newly added addresses to the interface, and release
3242 * addresses that have been deleted.
3244 static inline int __dev_uc_sync(struct net_device *dev,
3245 int (*sync)(struct net_device *,
3246 const unsigned char *),
3247 int (*unsync)(struct net_device *,
3248 const unsigned char *))
3250 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3254 * __dev_uc_unsync - Remove synchronized addresses from device
3255 * @dev: device to sync
3256 * @unsync: function to call if address should be removed
3258 * Remove all addresses that were added to the device by dev_uc_sync().
3260 static inline void __dev_uc_unsync(struct net_device *dev,
3261 int (*unsync)(struct net_device *,
3262 const unsigned char *))
3264 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3267 /* Functions used for multicast addresses handling */
3268 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3269 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3270 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3271 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3272 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3273 int dev_mc_sync(struct net_device *to, struct net_device *from);
3274 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3275 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3276 void dev_mc_flush(struct net_device *dev);
3277 void dev_mc_init(struct net_device *dev);
3280 * __dev_mc_sync - Synchonize device's multicast list
3281 * @dev: device to sync
3282 * @sync: function to call if address should be added
3283 * @unsync: function to call if address should be removed
3285 * Add newly added addresses to the interface, and release
3286 * addresses that have been deleted.
3288 static inline int __dev_mc_sync(struct net_device *dev,
3289 int (*sync)(struct net_device *,
3290 const unsigned char *),
3291 int (*unsync)(struct net_device *,
3292 const unsigned char *))
3294 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3298 * __dev_mc_unsync - Remove synchronized addresses from device
3299 * @dev: device to sync
3300 * @unsync: function to call if address should be removed
3302 * Remove all addresses that were added to the device by dev_mc_sync().
3304 static inline void __dev_mc_unsync(struct net_device *dev,
3305 int (*unsync)(struct net_device *,
3306 const unsigned char *))
3308 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3311 /* Functions used for secondary unicast and multicast support */
3312 void dev_set_rx_mode(struct net_device *dev);
3313 void __dev_set_rx_mode(struct net_device *dev);
3314 int dev_set_promiscuity(struct net_device *dev, int inc);
3315 int dev_set_allmulti(struct net_device *dev, int inc);
3316 void netdev_state_change(struct net_device *dev);
3317 void netdev_notify_peers(struct net_device *dev);
3318 void netdev_features_change(struct net_device *dev);
3319 /* Load a device via the kmod */
3320 void dev_load(struct net *net, const char *name);
3321 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3322 struct rtnl_link_stats64 *storage);
3323 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3324 const struct net_device_stats *netdev_stats);
3326 extern int netdev_max_backlog;
3327 extern int netdev_tstamp_prequeue;
3328 extern int weight_p;
3329 extern int bpf_jit_enable;
3331 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3332 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3333 struct list_head **iter);
3334 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3335 struct list_head **iter);
3337 /* iterate through upper list, must be called under RCU read lock */
3338 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3339 for (iter = &(dev)->adj_list.upper, \
3340 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3341 updev; \
3342 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3344 /* iterate through upper list, must be called under RCU read lock */
3345 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3346 for (iter = &(dev)->all_adj_list.upper, \
3347 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3348 updev; \
3349 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3351 void *netdev_lower_get_next_private(struct net_device *dev,
3352 struct list_head **iter);
3353 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3354 struct list_head **iter);
3356 #define netdev_for_each_lower_private(dev, priv, iter) \
3357 for (iter = (dev)->adj_list.lower.next, \
3358 priv = netdev_lower_get_next_private(dev, &(iter)); \
3359 priv; \
3360 priv = netdev_lower_get_next_private(dev, &(iter)))
3362 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3363 for (iter = &(dev)->adj_list.lower, \
3364 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3365 priv; \
3366 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3368 void *netdev_lower_get_next(struct net_device *dev,
3369 struct list_head **iter);
3370 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3371 for (iter = &(dev)->adj_list.lower, \
3372 ldev = netdev_lower_get_next(dev, &(iter)); \
3373 ldev; \
3374 ldev = netdev_lower_get_next(dev, &(iter)))
3376 void *netdev_adjacent_get_private(struct list_head *adj_list);
3377 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3378 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3379 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3380 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3381 int netdev_master_upper_dev_link(struct net_device *dev,
3382 struct net_device *upper_dev);
3383 int netdev_master_upper_dev_link_private(struct net_device *dev,
3384 struct net_device *upper_dev,
3385 void *private);
3386 void netdev_upper_dev_unlink(struct net_device *dev,
3387 struct net_device *upper_dev);
3388 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3389 void *netdev_lower_dev_get_private(struct net_device *dev,
3390 struct net_device *lower_dev);
3391 int dev_get_nest_level(struct net_device *dev,
3392 bool (*type_check)(struct net_device *dev));
3393 int skb_checksum_help(struct sk_buff *skb);
3394 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3395 netdev_features_t features, bool tx_path);
3396 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3397 netdev_features_t features);
3399 static inline
3400 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3402 return __skb_gso_segment(skb, features, true);
3404 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3406 static inline bool can_checksum_protocol(netdev_features_t features,
3407 __be16 protocol)
3409 return ((features & NETIF_F_GEN_CSUM) ||
3410 ((features & NETIF_F_V4_CSUM) &&
3411 protocol == htons(ETH_P_IP)) ||
3412 ((features & NETIF_F_V6_CSUM) &&
3413 protocol == htons(ETH_P_IPV6)) ||
3414 ((features & NETIF_F_FCOE_CRC) &&
3415 protocol == htons(ETH_P_FCOE)));
3418 #ifdef CONFIG_BUG
3419 void netdev_rx_csum_fault(struct net_device *dev);
3420 #else
3421 static inline void netdev_rx_csum_fault(struct net_device *dev)
3424 #endif
3425 /* rx skb timestamps */
3426 void net_enable_timestamp(void);
3427 void net_disable_timestamp(void);
3429 #ifdef CONFIG_PROC_FS
3430 int __init dev_proc_init(void);
3431 #else
3432 #define dev_proc_init() 0
3433 #endif
3435 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3436 struct sk_buff *skb, struct net_device *dev,
3437 bool more)
3439 skb->xmit_more = more ? 1 : 0;
3440 return ops->ndo_start_xmit(skb, dev);
3443 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3444 struct netdev_queue *txq, bool more)
3446 const struct net_device_ops *ops = dev->netdev_ops;
3447 int rc;
3449 rc = __netdev_start_xmit(ops, skb, dev, more);
3450 if (rc == NETDEV_TX_OK)
3451 txq_trans_update(txq);
3453 return rc;
3456 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3457 const void *ns);
3458 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3459 const void *ns);
3461 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3463 return netdev_class_create_file_ns(class_attr, NULL);
3466 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3468 netdev_class_remove_file_ns(class_attr, NULL);
3471 extern struct kobj_ns_type_operations net_ns_type_operations;
3473 const char *netdev_drivername(const struct net_device *dev);
3475 void linkwatch_run_queue(void);
3477 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3478 netdev_features_t f2)
3480 if (f1 & NETIF_F_GEN_CSUM)
3481 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3482 if (f2 & NETIF_F_GEN_CSUM)
3483 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3484 f1 &= f2;
3485 if (f1 & NETIF_F_GEN_CSUM)
3486 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3488 return f1;
3491 static inline netdev_features_t netdev_get_wanted_features(
3492 struct net_device *dev)
3494 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3496 netdev_features_t netdev_increment_features(netdev_features_t all,
3497 netdev_features_t one, netdev_features_t mask);
3499 /* Allow TSO being used on stacked device :
3500 * Performing the GSO segmentation before last device
3501 * is a performance improvement.
3503 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3504 netdev_features_t mask)
3506 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3509 int __netdev_update_features(struct net_device *dev);
3510 void netdev_update_features(struct net_device *dev);
3511 void netdev_change_features(struct net_device *dev);
3513 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3514 struct net_device *dev);
3516 netdev_features_t netif_skb_features(struct sk_buff *skb);
3518 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3520 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3522 /* check flags correspondence */
3523 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3524 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3525 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3526 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3527 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3528 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3529 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3530 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3531 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3532 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3533 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3534 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3535 BUILD_BUG_ON(SKB_GSO_MPLS != (NETIF_F_GSO_MPLS >> NETIF_F_GSO_SHIFT));
3537 return (features & feature) == feature;
3540 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3542 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3543 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3546 static inline bool netif_needs_gso(struct sk_buff *skb,
3547 netdev_features_t features)
3549 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3550 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3551 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3554 static inline void netif_set_gso_max_size(struct net_device *dev,
3555 unsigned int size)
3557 dev->gso_max_size = size;
3560 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3561 int pulled_hlen, u16 mac_offset,
3562 int mac_len)
3564 skb->protocol = protocol;
3565 skb->encapsulation = 1;
3566 skb_push(skb, pulled_hlen);
3567 skb_reset_transport_header(skb);
3568 skb->mac_header = mac_offset;
3569 skb->network_header = skb->mac_header + mac_len;
3570 skb->mac_len = mac_len;
3573 static inline bool netif_is_macvlan(struct net_device *dev)
3575 return dev->priv_flags & IFF_MACVLAN;
3578 static inline bool netif_is_bond_master(struct net_device *dev)
3580 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3583 static inline bool netif_is_bond_slave(struct net_device *dev)
3585 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3588 static inline bool netif_supports_nofcs(struct net_device *dev)
3590 return dev->priv_flags & IFF_SUPP_NOFCS;
3593 extern struct pernet_operations __net_initdata loopback_net_ops;
3595 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3597 /* netdev_printk helpers, similar to dev_printk */
3599 static inline const char *netdev_name(const struct net_device *dev)
3601 if (!dev->name[0] || strchr(dev->name, '%'))
3602 return "(unnamed net_device)";
3603 return dev->name;
3606 static inline const char *netdev_reg_state(const struct net_device *dev)
3608 switch (dev->reg_state) {
3609 case NETREG_UNINITIALIZED: return " (uninitialized)";
3610 case NETREG_REGISTERED: return "";
3611 case NETREG_UNREGISTERING: return " (unregistering)";
3612 case NETREG_UNREGISTERED: return " (unregistered)";
3613 case NETREG_RELEASED: return " (released)";
3614 case NETREG_DUMMY: return " (dummy)";
3617 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3618 return " (unknown)";
3621 __printf(3, 4)
3622 int netdev_printk(const char *level, const struct net_device *dev,
3623 const char *format, ...);
3624 __printf(2, 3)
3625 int netdev_emerg(const struct net_device *dev, const char *format, ...);
3626 __printf(2, 3)
3627 int netdev_alert(const struct net_device *dev, const char *format, ...);
3628 __printf(2, 3)
3629 int netdev_crit(const struct net_device *dev, const char *format, ...);
3630 __printf(2, 3)
3631 int netdev_err(const struct net_device *dev, const char *format, ...);
3632 __printf(2, 3)
3633 int netdev_warn(const struct net_device *dev, const char *format, ...);
3634 __printf(2, 3)
3635 int netdev_notice(const struct net_device *dev, const char *format, ...);
3636 __printf(2, 3)
3637 int netdev_info(const struct net_device *dev, const char *format, ...);
3639 #define MODULE_ALIAS_NETDEV(device) \
3640 MODULE_ALIAS("netdev-" device)
3642 #if defined(CONFIG_DYNAMIC_DEBUG)
3643 #define netdev_dbg(__dev, format, args...) \
3644 do { \
3645 dynamic_netdev_dbg(__dev, format, ##args); \
3646 } while (0)
3647 #elif defined(DEBUG)
3648 #define netdev_dbg(__dev, format, args...) \
3649 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3650 #else
3651 #define netdev_dbg(__dev, format, args...) \
3652 ({ \
3653 if (0) \
3654 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3655 0; \
3657 #endif
3659 #if defined(VERBOSE_DEBUG)
3660 #define netdev_vdbg netdev_dbg
3661 #else
3663 #define netdev_vdbg(dev, format, args...) \
3664 ({ \
3665 if (0) \
3666 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3667 0; \
3669 #endif
3672 * netdev_WARN() acts like dev_printk(), but with the key difference
3673 * of using a WARN/WARN_ON to get the message out, including the
3674 * file/line information and a backtrace.
3676 #define netdev_WARN(dev, format, args...) \
3677 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3678 netdev_reg_state(dev), ##args)
3680 /* netif printk helpers, similar to netdev_printk */
3682 #define netif_printk(priv, type, level, dev, fmt, args...) \
3683 do { \
3684 if (netif_msg_##type(priv)) \
3685 netdev_printk(level, (dev), fmt, ##args); \
3686 } while (0)
3688 #define netif_level(level, priv, type, dev, fmt, args...) \
3689 do { \
3690 if (netif_msg_##type(priv)) \
3691 netdev_##level(dev, fmt, ##args); \
3692 } while (0)
3694 #define netif_emerg(priv, type, dev, fmt, args...) \
3695 netif_level(emerg, priv, type, dev, fmt, ##args)
3696 #define netif_alert(priv, type, dev, fmt, args...) \
3697 netif_level(alert, priv, type, dev, fmt, ##args)
3698 #define netif_crit(priv, type, dev, fmt, args...) \
3699 netif_level(crit, priv, type, dev, fmt, ##args)
3700 #define netif_err(priv, type, dev, fmt, args...) \
3701 netif_level(err, priv, type, dev, fmt, ##args)
3702 #define netif_warn(priv, type, dev, fmt, args...) \
3703 netif_level(warn, priv, type, dev, fmt, ##args)
3704 #define netif_notice(priv, type, dev, fmt, args...) \
3705 netif_level(notice, priv, type, dev, fmt, ##args)
3706 #define netif_info(priv, type, dev, fmt, args...) \
3707 netif_level(info, priv, type, dev, fmt, ##args)
3709 #if defined(CONFIG_DYNAMIC_DEBUG)
3710 #define netif_dbg(priv, type, netdev, format, args...) \
3711 do { \
3712 if (netif_msg_##type(priv)) \
3713 dynamic_netdev_dbg(netdev, format, ##args); \
3714 } while (0)
3715 #elif defined(DEBUG)
3716 #define netif_dbg(priv, type, dev, format, args...) \
3717 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3718 #else
3719 #define netif_dbg(priv, type, dev, format, args...) \
3720 ({ \
3721 if (0) \
3722 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3723 0; \
3725 #endif
3727 #if defined(VERBOSE_DEBUG)
3728 #define netif_vdbg netif_dbg
3729 #else
3730 #define netif_vdbg(priv, type, dev, format, args...) \
3731 ({ \
3732 if (0) \
3733 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3734 0; \
3736 #endif
3739 * The list of packet types we will receive (as opposed to discard)
3740 * and the routines to invoke.
3742 * Why 16. Because with 16 the only overlap we get on a hash of the
3743 * low nibble of the protocol value is RARP/SNAP/X.25.
3745 * NOTE: That is no longer true with the addition of VLAN tags. Not
3746 * sure which should go first, but I bet it won't make much
3747 * difference if we are running VLANs. The good news is that
3748 * this protocol won't be in the list unless compiled in, so
3749 * the average user (w/out VLANs) will not be adversely affected.
3750 * --BLG
3752 * 0800 IP
3753 * 8100 802.1Q VLAN
3754 * 0001 802.3
3755 * 0002 AX.25
3756 * 0004 802.2
3757 * 8035 RARP
3758 * 0005 SNAP
3759 * 0805 X.25
3760 * 0806 ARP
3761 * 8137 IPX
3762 * 0009 Localtalk
3763 * 86DD IPv6
3765 #define PTYPE_HASH_SIZE (16)
3766 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3768 #endif /* _LINUX_NETDEVICE_H */