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
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
60 /* 802.15.4 specific */
64 void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
67 /* Backlog congestion levels */
68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69 #define NET_RX_DROP 1 /* packet dropped */
72 * Transmit return codes: transmit return codes originate from three different
75 * - qdisc return codes
76 * - driver transmit return codes
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
88 /* qdisc ->enqueue() return codes. */
89 #define NET_XMIT_SUCCESS 0x00
90 #define NET_XMIT_DROP 0x01 /* skb dropped */
91 #define NET_XMIT_CN 0x02 /* congestion notification */
92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101 /* Driver transmit return codes */
102 #define NETDEV_TX_MASK 0xf0
105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
110 typedef enum netdev_tx netdev_tx_t;
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 static inline bool dev_xmit_complete(int rc)
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 if (likely(rc < NET_XMIT_MASK))
131 * Compute the worst case header length according to the protocols
135 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
136 # if defined(CONFIG_MAC80211_MESH)
137 # define LL_MAX_HEADER 128
139 # define LL_MAX_HEADER 96
142 # define LL_MAX_HEADER 32
145 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
147 #define MAX_HEADER LL_MAX_HEADER
149 #define MAX_HEADER (LL_MAX_HEADER + 48)
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
157 struct net_device_stats {
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
167 unsigned long collisions;
168 unsigned long rx_length_errors;
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
184 #include <linux/cache.h>
185 #include <linux/skbuff.h>
188 #include <linux/static_key.h>
189 extern struct static_key rps_needed;
196 struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
200 #define NETDEV_HW_ADDR_T_LAN 1
201 #define NETDEV_HW_ADDR_T_SAN 2
202 #define NETDEV_HW_ADDR_T_SLAVE 3
203 #define NETDEV_HW_ADDR_T_UNICAST 4
204 #define NETDEV_HW_ADDR_T_MULTICAST 5
209 struct rcu_head rcu_head;
212 struct netdev_hw_addr_list {
213 struct list_head list;
217 #define netdev_hw_addr_list_count(l) ((l)->count)
218 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219 #define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
222 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224 #define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
227 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
229 #define netdev_for_each_mc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
237 /* cached hardware header; allow for machine alignment needs. */
238 #define HH_DATA_MOD 16
239 #define HH_DATA_OFF(__len) \
240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
241 #define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
246 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
254 #define LL_RESERVED_SPACE(dev) \
255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
256 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
262 const void *saddr, unsigned int len);
263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268 bool (*validate)(const char *ll_header, unsigned int len);
271 /* These flag bits are private to the generic network queueing
272 * layer, they may not be explicitly referenced by any other
276 enum netdev_state_t {
278 __LINK_STATE_PRESENT,
279 __LINK_STATE_NOCARRIER,
280 __LINK_STATE_LINKWATCH_PENDING,
281 __LINK_STATE_DORMANT,
286 * This structure holds at boot time configured netdevice settings. They
287 * are then used in the device probing.
289 struct netdev_boot_setup {
293 #define NETDEV_BOOT_SETUP_MAX 8
295 int __init netdev_boot_setup(char *str);
298 * Structure for NAPI scheduling similar to tasklet but with weighting
301 /* The poll_list must only be managed by the entity which
302 * changes the state of the NAPI_STATE_SCHED bit. This means
303 * whoever atomically sets that bit can add this napi_struct
304 * to the per-cpu poll_list, and whoever clears that bit
305 * can remove from the list right before clearing the bit.
307 struct list_head poll_list;
311 unsigned int gro_count;
312 int (*poll)(struct napi_struct *, int);
313 #ifdef CONFIG_NETPOLL
314 spinlock_t poll_lock;
317 struct net_device *dev;
318 struct sk_buff *gro_list;
320 struct hrtimer timer;
321 struct list_head dev_list;
322 struct hlist_node napi_hash_node;
323 unsigned int napi_id;
327 NAPI_STATE_SCHED, /* Poll is scheduled */
328 NAPI_STATE_DISABLE, /* Disable pending */
329 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
330 NAPI_STATE_HASHED, /* In NAPI hash */
340 typedef enum gro_result gro_result_t;
343 * enum rx_handler_result - Possible return values for rx_handlers.
344 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
346 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
347 * case skb->dev was changed by rx_handler.
348 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
349 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
351 * rx_handlers are functions called from inside __netif_receive_skb(), to do
352 * special processing of the skb, prior to delivery to protocol handlers.
354 * Currently, a net_device can only have a single rx_handler registered. Trying
355 * to register a second rx_handler will return -EBUSY.
357 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
358 * To unregister a rx_handler on a net_device, use
359 * netdev_rx_handler_unregister().
361 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
364 * If the rx_handler consumed to skb in some way, it should return
365 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
366 * the skb to be delivered in some other ways.
368 * If the rx_handler changed skb->dev, to divert the skb to another
369 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
370 * new device will be called if it exists.
372 * If the rx_handler consider the skb should be ignored, it should return
373 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
374 * are registered on exact device (ptype->dev == skb->dev).
376 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
377 * delivered, it should return RX_HANDLER_PASS.
379 * A device without a registered rx_handler will behave as if rx_handler
380 * returned RX_HANDLER_PASS.
383 enum rx_handler_result {
389 typedef enum rx_handler_result rx_handler_result_t;
390 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
392 void __napi_schedule(struct napi_struct *n);
395 * When PREEMPT_RT_FULL is defined, all device interrupt handlers
396 * run as threads, and they can also be preempted (without PREEMPT_RT
397 * interrupt threads can not be preempted). Which means that calling
398 * __napi_schedule_irqoff() from an interrupt handler can be preempted
399 * and can corrupt the napi->poll_list.
401 #ifdef CONFIG_PREEMPT_RT_FULL
402 #define __napi_schedule_irqoff(n) __napi_schedule(n)
404 void __napi_schedule_irqoff(struct napi_struct *n);
407 static inline bool napi_disable_pending(struct napi_struct *n)
409 return test_bit(NAPI_STATE_DISABLE, &n->state);
413 * napi_schedule_prep - check if napi can be scheduled
416 * Test if NAPI routine is already running, and if not mark
417 * it as running. This is used as a condition variable
418 * insure only one NAPI poll instance runs. We also make
419 * sure there is no pending NAPI disable.
421 static inline bool napi_schedule_prep(struct napi_struct *n)
423 return !napi_disable_pending(n) &&
424 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
428 * napi_schedule - schedule NAPI poll
431 * Schedule NAPI poll routine to be called if it is not already
434 static inline void napi_schedule(struct napi_struct *n)
436 if (napi_schedule_prep(n))
441 * napi_schedule_irqoff - schedule NAPI poll
444 * Variant of napi_schedule(), assuming hard irqs are masked.
446 static inline void napi_schedule_irqoff(struct napi_struct *n)
448 if (napi_schedule_prep(n))
449 __napi_schedule_irqoff(n);
452 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
453 static inline bool napi_reschedule(struct napi_struct *napi)
455 if (napi_schedule_prep(napi)) {
456 __napi_schedule(napi);
462 void __napi_complete(struct napi_struct *n);
463 void napi_complete_done(struct napi_struct *n, int work_done);
465 * napi_complete - NAPI processing complete
468 * Mark NAPI processing as complete.
469 * Consider using napi_complete_done() instead.
471 static inline void napi_complete(struct napi_struct *n)
473 return napi_complete_done(n, 0);
477 * napi_by_id - lookup a NAPI by napi_id
478 * @napi_id: hashed napi_id
480 * lookup @napi_id in napi_hash table
481 * must be called under rcu_read_lock()
483 struct napi_struct *napi_by_id(unsigned int napi_id);
486 * napi_hash_add - add a NAPI to global hashtable
487 * @napi: napi context
489 * generate a new napi_id and store a @napi under it in napi_hash
491 void napi_hash_add(struct napi_struct *napi);
494 * napi_hash_del - remove a NAPI from global table
495 * @napi: napi context
497 * Warning: caller must observe rcu grace period
498 * before freeing memory containing @napi
500 void napi_hash_del(struct napi_struct *napi);
503 * napi_disable - prevent NAPI from scheduling
506 * Stop NAPI from being scheduled on this context.
507 * Waits till any outstanding processing completes.
509 void napi_disable(struct napi_struct *n);
512 * napi_enable - enable NAPI scheduling
515 * Resume NAPI from being scheduled on this context.
516 * Must be paired with napi_disable.
518 static inline void napi_enable(struct napi_struct *n)
520 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
521 smp_mb__before_atomic();
522 clear_bit(NAPI_STATE_SCHED, &n->state);
523 clear_bit(NAPI_STATE_NPSVC, &n->state);
527 * napi_synchronize - wait until NAPI is not running
530 * Wait until NAPI is done being scheduled on this context.
531 * Waits till any outstanding processing completes but
532 * does not disable future activations.
534 static inline void napi_synchronize(const struct napi_struct *n)
536 if (IS_ENABLED(CONFIG_SMP))
537 while (test_bit(NAPI_STATE_SCHED, &n->state))
543 enum netdev_queue_state_t {
544 __QUEUE_STATE_DRV_XOFF,
545 __QUEUE_STATE_STACK_XOFF,
546 __QUEUE_STATE_FROZEN,
549 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
550 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
551 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
553 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
554 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
556 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
560 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
561 * netif_tx_* functions below are used to manipulate this flag. The
562 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
563 * queue independently. The netif_xmit_*stopped functions below are called
564 * to check if the queue has been stopped by the driver or stack (either
565 * of the XOFF bits are set in the state). Drivers should not need to call
566 * netif_xmit*stopped functions, they should only be using netif_tx_*.
569 struct netdev_queue {
573 struct net_device *dev;
574 struct Qdisc __rcu *qdisc;
575 struct Qdisc *qdisc_sleeping;
579 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
585 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
588 * please use this field instead of dev->trans_start
590 unsigned long trans_start;
593 * Number of TX timeouts for this queue
594 * (/sys/class/net/DEV/Q/trans_timeout)
596 unsigned long trans_timeout;
603 unsigned long tx_maxrate;
604 } ____cacheline_aligned_in_smp;
606 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
608 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
615 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
617 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
624 * This structure holds an RPS map which can be of variable length. The
625 * map is an array of CPUs.
632 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
635 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
636 * tail pointer for that CPU's input queue at the time of last enqueue, and
637 * a hardware filter index.
639 struct rps_dev_flow {
642 unsigned int last_qtail;
644 #define RPS_NO_FILTER 0xffff
647 * The rps_dev_flow_table structure contains a table of flow mappings.
649 struct rps_dev_flow_table {
652 struct rps_dev_flow flows[0];
654 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
655 ((_num) * sizeof(struct rps_dev_flow)))
658 * The rps_sock_flow_table contains mappings of flows to the last CPU
659 * on which they were processed by the application (set in recvmsg).
660 * Each entry is a 32bit value. Upper part is the high order bits
661 * of flow hash, lower part is cpu number.
662 * rps_cpu_mask is used to partition the space, depending on number of
663 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
664 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
665 * meaning we use 32-6=26 bits for the hash.
667 struct rps_sock_flow_table {
670 u32 ents[0] ____cacheline_aligned_in_smp;
672 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
674 #define RPS_NO_CPU 0xffff
676 extern u32 rps_cpu_mask;
677 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
679 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
683 unsigned int index = hash & table->mask;
684 u32 val = hash & ~rps_cpu_mask;
686 /* We only give a hint, preemption can change cpu under us */
687 val |= raw_smp_processor_id();
689 if (table->ents[index] != val)
690 table->ents[index] = val;
694 #ifdef CONFIG_RFS_ACCEL
695 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
698 #endif /* CONFIG_RPS */
700 /* This structure contains an instance of an RX queue. */
701 struct netdev_rx_queue {
703 struct rps_map __rcu *rps_map;
704 struct rps_dev_flow_table __rcu *rps_flow_table;
707 struct net_device *dev;
708 } ____cacheline_aligned_in_smp;
711 * RX queue sysfs structures and functions.
713 struct rx_queue_attribute {
714 struct attribute attr;
715 ssize_t (*show)(struct netdev_rx_queue *queue,
716 struct rx_queue_attribute *attr, char *buf);
717 ssize_t (*store)(struct netdev_rx_queue *queue,
718 struct rx_queue_attribute *attr, const char *buf, size_t len);
723 * This structure holds an XPS map which can be of variable length. The
724 * map is an array of queues.
728 unsigned int alloc_len;
732 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
733 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
734 - sizeof(struct xps_map)) / sizeof(u16))
737 * This structure holds all XPS maps for device. Maps are indexed by CPU.
739 struct xps_dev_maps {
741 struct xps_map __rcu *cpu_map[0];
743 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
744 (nr_cpu_ids * sizeof(struct xps_map *)))
745 #endif /* CONFIG_XPS */
747 #define TC_MAX_QUEUE 16
748 #define TC_BITMASK 15
749 /* HW offloaded queuing disciplines txq count and offset maps */
750 struct netdev_tc_txq {
755 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
757 * This structure is to hold information about the device
758 * configured to run FCoE protocol stack.
760 struct netdev_fcoe_hbainfo {
761 char manufacturer[64];
762 char serial_number[64];
763 char hardware_version[64];
764 char driver_version[64];
765 char optionrom_version[64];
766 char firmware_version[64];
768 char model_description[256];
772 #define MAX_PHYS_ITEM_ID_LEN 32
774 /* This structure holds a unique identifier to identify some
775 * physical item (port for example) used by a netdevice.
777 struct netdev_phys_item_id {
778 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
779 unsigned char id_len;
782 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
783 struct netdev_phys_item_id *b)
785 return a->id_len == b->id_len &&
786 memcmp(a->id, b->id, a->id_len) == 0;
789 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
790 struct sk_buff *skb);
793 * This structure defines the management hooks for network devices.
794 * The following hooks can be defined; unless noted otherwise, they are
795 * optional and can be filled with a null pointer.
797 * int (*ndo_init)(struct net_device *dev);
798 * This function is called once when network device is registered.
799 * The network device can use this to any late stage initializaton
800 * or semantic validattion. It can fail with an error code which will
801 * be propogated back to register_netdev
803 * void (*ndo_uninit)(struct net_device *dev);
804 * This function is called when device is unregistered or when registration
805 * fails. It is not called if init fails.
807 * int (*ndo_open)(struct net_device *dev);
808 * This function is called when network device transistions to the up
811 * int (*ndo_stop)(struct net_device *dev);
812 * This function is called when network device transistions to the down
815 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
816 * struct net_device *dev);
817 * Called when a packet needs to be transmitted.
818 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
819 * the queue before that can happen; it's for obsolete devices and weird
820 * corner cases, but the stack really does a non-trivial amount
821 * of useless work if you return NETDEV_TX_BUSY.
822 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
823 * Required can not be NULL.
825 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
826 * void *accel_priv, select_queue_fallback_t fallback);
827 * Called to decide which queue to when device supports multiple
830 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
831 * This function is called to allow device receiver to make
832 * changes to configuration when multicast or promiscious is enabled.
834 * void (*ndo_set_rx_mode)(struct net_device *dev);
835 * This function is called device changes address list filtering.
836 * If driver handles unicast address filtering, it should set
837 * IFF_UNICAST_FLT to its priv_flags.
839 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
840 * This function is called when the Media Access Control address
841 * needs to be changed. If this interface is not defined, the
842 * mac address can not be changed.
844 * int (*ndo_validate_addr)(struct net_device *dev);
845 * Test if Media Access Control address is valid for the device.
847 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
848 * Called when a user request an ioctl which can't be handled by
849 * the generic interface code. If not defined ioctl's return
850 * not supported error code.
852 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
853 * Used to set network devices bus interface parameters. This interface
854 * is retained for legacy reason, new devices should use the bus
855 * interface (PCI) for low level management.
857 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
858 * Called when a user wants to change the Maximum Transfer Unit
859 * of a device. If not defined, any request to change MTU will
860 * will return an error.
862 * void (*ndo_tx_timeout)(struct net_device *dev);
863 * Callback uses when the transmitter has not made any progress
864 * for dev->watchdog ticks.
866 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
867 * struct rtnl_link_stats64 *storage);
868 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
869 * Called when a user wants to get the network device usage
870 * statistics. Drivers must do one of the following:
871 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
872 * rtnl_link_stats64 structure passed by the caller.
873 * 2. Define @ndo_get_stats to update a net_device_stats structure
874 * (which should normally be dev->stats) and return a pointer to
875 * it. The structure may be changed asynchronously only if each
876 * field is written atomically.
877 * 3. Update dev->stats asynchronously and atomically, and define
880 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
881 * If device support VLAN filtering this function is called when a
882 * VLAN id is registered.
884 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
885 * If device support VLAN filtering this function is called when a
886 * VLAN id is unregistered.
888 * void (*ndo_poll_controller)(struct net_device *dev);
890 * SR-IOV management functions.
891 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
892 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
893 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
895 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
896 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
897 * int (*ndo_get_vf_config)(struct net_device *dev,
898 * int vf, struct ifla_vf_info *ivf);
899 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
900 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
901 * struct nlattr *port[]);
903 * Enable or disable the VF ability to query its RSS Redirection Table and
904 * Hash Key. This is needed since on some devices VF share this information
905 * with PF and querying it may adduce a theoretical security risk.
906 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
907 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
908 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
909 * Called to setup 'tc' number of traffic classes in the net device. This
910 * is always called from the stack with the rtnl lock held and netif tx
911 * queues stopped. This allows the netdevice to perform queue management
914 * Fiber Channel over Ethernet (FCoE) offload functions.
915 * int (*ndo_fcoe_enable)(struct net_device *dev);
916 * Called when the FCoE protocol stack wants to start using LLD for FCoE
917 * so the underlying device can perform whatever needed configuration or
918 * initialization to support acceleration of FCoE traffic.
920 * int (*ndo_fcoe_disable)(struct net_device *dev);
921 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
922 * so the underlying device can perform whatever needed clean-ups to
923 * stop supporting acceleration of FCoE traffic.
925 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
926 * struct scatterlist *sgl, unsigned int sgc);
927 * Called when the FCoE Initiator wants to initialize an I/O that
928 * is a possible candidate for Direct Data Placement (DDP). The LLD can
929 * perform necessary setup and returns 1 to indicate the device is set up
930 * successfully to perform DDP on this I/O, otherwise this returns 0.
932 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
933 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
934 * indicated by the FC exchange id 'xid', so the underlying device can
935 * clean up and reuse resources for later DDP requests.
937 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
938 * struct scatterlist *sgl, unsigned int sgc);
939 * Called when the FCoE Target wants to initialize an I/O that
940 * is a possible candidate for Direct Data Placement (DDP). The LLD can
941 * perform necessary setup and returns 1 to indicate the device is set up
942 * successfully to perform DDP on this I/O, otherwise this returns 0.
944 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
945 * struct netdev_fcoe_hbainfo *hbainfo);
946 * Called when the FCoE Protocol stack wants information on the underlying
947 * device. This information is utilized by the FCoE protocol stack to
948 * register attributes with Fiber Channel management service as per the
949 * FC-GS Fabric Device Management Information(FDMI) specification.
951 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
952 * Called when the underlying device wants to override default World Wide
953 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
954 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
955 * protocol stack to use.
958 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
959 * u16 rxq_index, u32 flow_id);
960 * Set hardware filter for RFS. rxq_index is the target queue index;
961 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
962 * Return the filter ID on success, or a negative error code.
964 * Slave management functions (for bridge, bonding, etc).
965 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
966 * Called to make another netdev an underling.
968 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
969 * Called to release previously enslaved netdev.
971 * Feature/offload setting functions.
972 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
973 * netdev_features_t features);
974 * Adjusts the requested feature flags according to device-specific
975 * constraints, and returns the resulting flags. Must not modify
978 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
979 * Called to update device configuration to new features. Passed
980 * feature set might be less than what was returned by ndo_fix_features()).
981 * Must return >0 or -errno if it changed dev->features itself.
983 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
984 * struct net_device *dev,
985 * const unsigned char *addr, u16 vid, u16 flags)
986 * Adds an FDB entry to dev for addr.
987 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
988 * struct net_device *dev,
989 * const unsigned char *addr, u16 vid)
990 * Deletes the FDB entry from dev coresponding to addr.
991 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
992 * struct net_device *dev, struct net_device *filter_dev,
994 * Used to add FDB entries to dump requests. Implementers should add
995 * entries to skb and update idx with the number of entries.
997 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
999 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1000 * struct net_device *dev, u32 filter_mask,
1002 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1005 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1006 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1007 * which do not represent real hardware may define this to allow their
1008 * userspace components to manage their virtual carrier state. Devices
1009 * that determine carrier state from physical hardware properties (eg
1010 * network cables) or protocol-dependent mechanisms (eg
1011 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1013 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1014 * struct netdev_phys_item_id *ppid);
1015 * Called to get ID of physical port of this device. If driver does
1016 * not implement this, it is assumed that the hw is not able to have
1017 * multiple net devices on single physical port.
1019 * void (*ndo_add_vxlan_port)(struct net_device *dev,
1020 * sa_family_t sa_family, __be16 port);
1021 * Called by vxlan to notiy a driver about the UDP port and socket
1022 * address family that vxlan is listnening to. It is called only when
1023 * a new port starts listening. The operation is protected by the
1024 * vxlan_net->sock_lock.
1026 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1027 * sa_family_t sa_family, __be16 port);
1028 * Called by vxlan to notify the driver about a UDP port and socket
1029 * address family that vxlan is not listening to anymore. The operation
1030 * is protected by the vxlan_net->sock_lock.
1032 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1033 * struct net_device *dev)
1034 * Called by upper layer devices to accelerate switching or other
1035 * station functionality into hardware. 'pdev is the lowerdev
1036 * to use for the offload and 'dev' is the net device that will
1037 * back the offload. Returns a pointer to the private structure
1038 * the upper layer will maintain.
1039 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1040 * Called by upper layer device to delete the station created
1041 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1042 * the station and priv is the structure returned by the add
1044 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1045 * struct net_device *dev,
1047 * Callback to use for xmit over the accelerated station. This
1048 * is used in place of ndo_start_xmit on accelerated net
1050 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1051 * struct net_device *dev
1052 * netdev_features_t features);
1053 * Called by core transmit path to determine if device is capable of
1054 * performing offload operations on a given packet. This is to give
1055 * the device an opportunity to implement any restrictions that cannot
1056 * be otherwise expressed by feature flags. The check is called with
1057 * the set of features that the stack has calculated and it returns
1058 * those the driver believes to be appropriate.
1059 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1060 * int queue_index, u32 maxrate);
1061 * Called when a user wants to set a max-rate limitation of specific
1063 * int (*ndo_get_iflink)(const struct net_device *dev);
1064 * Called to get the iflink value of this device.
1065 * void (*ndo_change_proto_down)(struct net_device *dev,
1067 * This function is used to pass protocol port error state information
1068 * to the switch driver. The switch driver can react to the proto_down
1069 * by doing a phys down on the associated switch port.
1070 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1071 * This function is used to get egress tunnel information for given skb.
1072 * This is useful for retrieving outer tunnel header parameters while
1076 struct net_device_ops {
1077 int (*ndo_init)(struct net_device *dev);
1078 void (*ndo_uninit)(struct net_device *dev);
1079 int (*ndo_open)(struct net_device *dev);
1080 int (*ndo_stop)(struct net_device *dev);
1081 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1082 struct net_device *dev);
1083 u16 (*ndo_select_queue)(struct net_device *dev,
1084 struct sk_buff *skb,
1086 select_queue_fallback_t fallback);
1087 void (*ndo_change_rx_flags)(struct net_device *dev,
1089 void (*ndo_set_rx_mode)(struct net_device *dev);
1090 int (*ndo_set_mac_address)(struct net_device *dev,
1092 int (*ndo_validate_addr)(struct net_device *dev);
1093 int (*ndo_do_ioctl)(struct net_device *dev,
1094 struct ifreq *ifr, int cmd);
1095 int (*ndo_set_config)(struct net_device *dev,
1097 int (*ndo_change_mtu)(struct net_device *dev,
1099 int (*ndo_neigh_setup)(struct net_device *dev,
1100 struct neigh_parms *);
1101 void (*ndo_tx_timeout) (struct net_device *dev);
1103 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1104 struct rtnl_link_stats64 *storage);
1105 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1107 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1108 __be16 proto, u16 vid);
1109 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1110 __be16 proto, u16 vid);
1111 #ifdef CONFIG_NET_POLL_CONTROLLER
1112 void (*ndo_poll_controller)(struct net_device *dev);
1113 int (*ndo_netpoll_setup)(struct net_device *dev,
1114 struct netpoll_info *info);
1115 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1117 #ifdef CONFIG_NET_RX_BUSY_POLL
1118 int (*ndo_busy_poll)(struct napi_struct *dev);
1120 int (*ndo_set_vf_mac)(struct net_device *dev,
1121 int queue, u8 *mac);
1122 int (*ndo_set_vf_vlan)(struct net_device *dev,
1123 int queue, u16 vlan, u8 qos);
1124 int (*ndo_set_vf_rate)(struct net_device *dev,
1125 int vf, int min_tx_rate,
1127 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1128 int vf, bool setting);
1129 int (*ndo_set_vf_trust)(struct net_device *dev,
1130 int vf, bool setting);
1131 int (*ndo_get_vf_config)(struct net_device *dev,
1133 struct ifla_vf_info *ivf);
1134 int (*ndo_set_vf_link_state)(struct net_device *dev,
1135 int vf, int link_state);
1136 int (*ndo_get_vf_stats)(struct net_device *dev,
1138 struct ifla_vf_stats
1140 int (*ndo_set_vf_port)(struct net_device *dev,
1142 struct nlattr *port[]);
1143 int (*ndo_get_vf_port)(struct net_device *dev,
1144 int vf, struct sk_buff *skb);
1145 int (*ndo_set_vf_rss_query_en)(
1146 struct net_device *dev,
1147 int vf, bool setting);
1148 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1149 #if IS_ENABLED(CONFIG_FCOE)
1150 int (*ndo_fcoe_enable)(struct net_device *dev);
1151 int (*ndo_fcoe_disable)(struct net_device *dev);
1152 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1154 struct scatterlist *sgl,
1156 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1158 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1160 struct scatterlist *sgl,
1162 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1163 struct netdev_fcoe_hbainfo *hbainfo);
1166 #if IS_ENABLED(CONFIG_LIBFCOE)
1167 #define NETDEV_FCOE_WWNN 0
1168 #define NETDEV_FCOE_WWPN 1
1169 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1170 u64 *wwn, int type);
1173 #ifdef CONFIG_RFS_ACCEL
1174 int (*ndo_rx_flow_steer)(struct net_device *dev,
1175 const struct sk_buff *skb,
1179 int (*ndo_add_slave)(struct net_device *dev,
1180 struct net_device *slave_dev);
1181 int (*ndo_del_slave)(struct net_device *dev,
1182 struct net_device *slave_dev);
1183 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1184 netdev_features_t features);
1185 int (*ndo_set_features)(struct net_device *dev,
1186 netdev_features_t features);
1187 int (*ndo_neigh_construct)(struct neighbour *n);
1188 void (*ndo_neigh_destroy)(struct neighbour *n);
1190 int (*ndo_fdb_add)(struct ndmsg *ndm,
1191 struct nlattr *tb[],
1192 struct net_device *dev,
1193 const unsigned char *addr,
1196 int (*ndo_fdb_del)(struct ndmsg *ndm,
1197 struct nlattr *tb[],
1198 struct net_device *dev,
1199 const unsigned char *addr,
1201 int (*ndo_fdb_dump)(struct sk_buff *skb,
1202 struct netlink_callback *cb,
1203 struct net_device *dev,
1204 struct net_device *filter_dev,
1207 int (*ndo_bridge_setlink)(struct net_device *dev,
1208 struct nlmsghdr *nlh,
1210 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1212 struct net_device *dev,
1215 int (*ndo_bridge_dellink)(struct net_device *dev,
1216 struct nlmsghdr *nlh,
1218 int (*ndo_change_carrier)(struct net_device *dev,
1220 int (*ndo_get_phys_port_id)(struct net_device *dev,
1221 struct netdev_phys_item_id *ppid);
1222 int (*ndo_get_phys_port_name)(struct net_device *dev,
1223 char *name, size_t len);
1224 void (*ndo_add_vxlan_port)(struct net_device *dev,
1225 sa_family_t sa_family,
1227 void (*ndo_del_vxlan_port)(struct net_device *dev,
1228 sa_family_t sa_family,
1231 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1232 struct net_device *dev);
1233 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1236 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1237 struct net_device *dev,
1239 int (*ndo_get_lock_subclass)(struct net_device *dev);
1240 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1241 struct net_device *dev,
1242 netdev_features_t features);
1243 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1246 int (*ndo_get_iflink)(const struct net_device *dev);
1247 int (*ndo_change_proto_down)(struct net_device *dev,
1249 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1250 struct sk_buff *skb);
1254 * enum net_device_priv_flags - &struct net_device priv_flags
1256 * These are the &struct net_device, they are only set internally
1257 * by drivers and used in the kernel. These flags are invisible to
1258 * userspace, this means that the order of these flags can change
1259 * during any kernel release.
1261 * You should have a pretty good reason to be extending these flags.
1263 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1264 * @IFF_EBRIDGE: Ethernet bridging device
1265 * @IFF_BONDING: bonding master or slave
1266 * @IFF_ISATAP: ISATAP interface (RFC4214)
1267 * @IFF_WAN_HDLC: WAN HDLC device
1268 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1270 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1271 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1272 * @IFF_MACVLAN_PORT: device used as macvlan port
1273 * @IFF_BRIDGE_PORT: device used as bridge port
1274 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1275 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1276 * @IFF_UNICAST_FLT: Supports unicast filtering
1277 * @IFF_TEAM_PORT: device used as team port
1278 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1279 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1280 * change when it's running
1281 * @IFF_MACVLAN: Macvlan device
1282 * @IFF_L3MDEV_MASTER: device is an L3 master device
1283 * @IFF_NO_QUEUE: device can run without qdisc attached
1284 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1285 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1287 enum netdev_priv_flags {
1288 IFF_802_1Q_VLAN = 1<<0,
1292 IFF_WAN_HDLC = 1<<4,
1293 IFF_XMIT_DST_RELEASE = 1<<5,
1294 IFF_DONT_BRIDGE = 1<<6,
1295 IFF_DISABLE_NETPOLL = 1<<7,
1296 IFF_MACVLAN_PORT = 1<<8,
1297 IFF_BRIDGE_PORT = 1<<9,
1298 IFF_OVS_DATAPATH = 1<<10,
1299 IFF_TX_SKB_SHARING = 1<<11,
1300 IFF_UNICAST_FLT = 1<<12,
1301 IFF_TEAM_PORT = 1<<13,
1302 IFF_SUPP_NOFCS = 1<<14,
1303 IFF_LIVE_ADDR_CHANGE = 1<<15,
1304 IFF_MACVLAN = 1<<16,
1305 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1306 IFF_IPVLAN_MASTER = 1<<18,
1307 IFF_IPVLAN_SLAVE = 1<<19,
1308 IFF_L3MDEV_MASTER = 1<<20,
1309 IFF_NO_QUEUE = 1<<21,
1310 IFF_OPENVSWITCH = 1<<22,
1311 IFF_L3MDEV_SLAVE = 1<<23,
1314 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1315 #define IFF_EBRIDGE IFF_EBRIDGE
1316 #define IFF_BONDING IFF_BONDING
1317 #define IFF_ISATAP IFF_ISATAP
1318 #define IFF_WAN_HDLC IFF_WAN_HDLC
1319 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1320 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1321 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1322 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1323 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1324 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1325 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1326 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1327 #define IFF_TEAM_PORT IFF_TEAM_PORT
1328 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1329 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1330 #define IFF_MACVLAN IFF_MACVLAN
1331 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1332 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1333 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1334 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1335 #define IFF_NO_QUEUE IFF_NO_QUEUE
1336 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1337 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1340 * struct net_device - The DEVICE structure.
1341 * Actually, this whole structure is a big mistake. It mixes I/O
1342 * data with strictly "high-level" data, and it has to know about
1343 * almost every data structure used in the INET module.
1345 * @name: This is the first field of the "visible" part of this structure
1346 * (i.e. as seen by users in the "Space.c" file). It is the name
1349 * @name_hlist: Device name hash chain, please keep it close to name[]
1350 * @ifalias: SNMP alias
1351 * @mem_end: Shared memory end
1352 * @mem_start: Shared memory start
1353 * @base_addr: Device I/O address
1354 * @irq: Device IRQ number
1356 * @carrier_changes: Stats to monitor carrier on<->off transitions
1358 * @state: Generic network queuing layer state, see netdev_state_t
1359 * @dev_list: The global list of network devices
1360 * @napi_list: List entry, that is used for polling napi devices
1361 * @unreg_list: List entry, that is used, when we are unregistering the
1362 * device, see the function unregister_netdev
1363 * @close_list: List entry, that is used, when we are closing the device
1365 * @adj_list: Directly linked devices, like slaves for bonding
1366 * @all_adj_list: All linked devices, *including* neighbours
1367 * @features: Currently active device features
1368 * @hw_features: User-changeable features
1370 * @wanted_features: User-requested features
1371 * @vlan_features: Mask of features inheritable by VLAN devices
1373 * @hw_enc_features: Mask of features inherited by encapsulating devices
1374 * This field indicates what encapsulation
1375 * offloads the hardware is capable of doing,
1376 * and drivers will need to set them appropriately.
1378 * @mpls_features: Mask of features inheritable by MPLS
1380 * @ifindex: interface index
1381 * @group: The group, that the device belongs to
1383 * @stats: Statistics struct, which was left as a legacy, use
1384 * rtnl_link_stats64 instead
1386 * @rx_dropped: Dropped packets by core network,
1387 * do not use this in drivers
1388 * @tx_dropped: Dropped packets by core network,
1389 * do not use this in drivers
1391 * @wireless_handlers: List of functions to handle Wireless Extensions,
1393 * see <net/iw_handler.h> for details.
1394 * @wireless_data: Instance data managed by the core of wireless extensions
1396 * @netdev_ops: Includes several pointers to callbacks,
1397 * if one wants to override the ndo_*() functions
1398 * @ethtool_ops: Management operations
1399 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1400 * of Layer 2 headers.
1402 * @flags: Interface flags (a la BSD)
1403 * @priv_flags: Like 'flags' but invisible to userspace,
1404 * see if.h for the definitions
1405 * @gflags: Global flags ( kept as legacy )
1406 * @padded: How much padding added by alloc_netdev()
1407 * @operstate: RFC2863 operstate
1408 * @link_mode: Mapping policy to operstate
1409 * @if_port: Selectable AUI, TP, ...
1411 * @mtu: Interface MTU value
1412 * @type: Interface hardware type
1413 * @hard_header_len: Maximum hardware header length.
1414 * @min_header_len: Minimum hardware header length
1416 * @needed_headroom: Extra headroom the hardware may need, but not in all
1417 * cases can this be guaranteed
1418 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1419 * cases can this be guaranteed. Some cases also use
1420 * LL_MAX_HEADER instead to allocate the skb
1422 * interface address info:
1424 * @perm_addr: Permanent hw address
1425 * @addr_assign_type: Hw address assignment type
1426 * @addr_len: Hardware address length
1427 * @neigh_priv_len; Used in neigh_alloc(),
1428 * initialized only in atm/clip.c
1429 * @dev_id: Used to differentiate devices that share
1430 * the same link layer address
1431 * @dev_port: Used to differentiate devices that share
1433 * @addr_list_lock: XXX: need comments on this one
1434 * @uc_promisc: Counter, that indicates, that promiscuous mode
1435 * has been enabled due to the need to listen to
1436 * additional unicast addresses in a device that
1437 * does not implement ndo_set_rx_mode()
1438 * @uc: unicast mac addresses
1439 * @mc: multicast mac addresses
1440 * @dev_addrs: list of device hw addresses
1441 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1442 * @promiscuity: Number of times, the NIC is told to work in
1443 * Promiscuous mode, if it becomes 0 the NIC will
1444 * exit from working in Promiscuous mode
1445 * @allmulti: Counter, enables or disables allmulticast mode
1447 * @vlan_info: VLAN info
1448 * @dsa_ptr: dsa specific data
1449 * @tipc_ptr: TIPC specific data
1450 * @atalk_ptr: AppleTalk link
1451 * @ip_ptr: IPv4 specific data
1452 * @dn_ptr: DECnet specific data
1453 * @ip6_ptr: IPv6 specific data
1454 * @ax25_ptr: AX.25 specific data
1455 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1457 * @last_rx: Time of last Rx
1458 * @dev_addr: Hw address (before bcast,
1459 * because most packets are unicast)
1461 * @_rx: Array of RX queues
1462 * @num_rx_queues: Number of RX queues
1463 * allocated at register_netdev() time
1464 * @real_num_rx_queues: Number of RX queues currently active in device
1466 * @rx_handler: handler for received packets
1467 * @rx_handler_data: XXX: need comments on this one
1468 * @ingress_queue: XXX: need comments on this one
1469 * @broadcast: hw bcast address
1471 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1472 * indexed by RX queue number. Assigned by driver.
1473 * This must only be set if the ndo_rx_flow_steer
1474 * operation is defined
1475 * @index_hlist: Device index hash chain
1477 * @_tx: Array of TX queues
1478 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1479 * @real_num_tx_queues: Number of TX queues currently active in device
1480 * @qdisc: Root qdisc from userspace point of view
1481 * @tx_queue_len: Max frames per queue allowed
1482 * @tx_global_lock: XXX: need comments on this one
1484 * @xps_maps: XXX: need comments on this one
1486 * @offload_fwd_mark: Offload device fwding mark
1488 * @trans_start: Time (in jiffies) of last Tx
1489 * @watchdog_timeo: Represents the timeout that is used by
1490 * the watchdog ( see dev_watchdog() )
1491 * @watchdog_timer: List of timers
1493 * @pcpu_refcnt: Number of references to this device
1494 * @todo_list: Delayed register/unregister
1495 * @link_watch_list: XXX: need comments on this one
1497 * @reg_state: Register/unregister state machine
1498 * @dismantle: Device is going to be freed
1499 * @rtnl_link_state: This enum represents the phases of creating
1502 * @destructor: Called from unregister,
1503 * can be used to call free_netdev
1504 * @npinfo: XXX: need comments on this one
1505 * @nd_net: Network namespace this network device is inside
1507 * @ml_priv: Mid-layer private
1508 * @lstats: Loopback statistics
1509 * @tstats: Tunnel statistics
1510 * @dstats: Dummy statistics
1511 * @vstats: Virtual ethernet statistics
1516 * @dev: Class/net/name entry
1517 * @sysfs_groups: Space for optional device, statistics and wireless
1520 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1521 * @rtnl_link_ops: Rtnl_link_ops
1523 * @gso_max_size: Maximum size of generic segmentation offload
1524 * @gso_max_segs: Maximum number of segments that can be passed to the
1526 * @gso_min_segs: Minimum number of segments that can be passed to the
1529 * @dcbnl_ops: Data Center Bridging netlink ops
1530 * @num_tc: Number of traffic classes in the net device
1531 * @tc_to_txq: XXX: need comments on this one
1532 * @prio_tc_map XXX: need comments on this one
1534 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1536 * @priomap: XXX: need comments on this one
1537 * @phydev: Physical device may attach itself
1538 * for hardware timestamping
1540 * @qdisc_tx_busylock: XXX: need comments on this one
1542 * @proto_down: protocol port state information can be sent to the
1543 * switch driver and used to set the phys state of the
1546 * FIXME: cleanup struct net_device such that network protocol info
1551 char name[IFNAMSIZ];
1552 struct hlist_node name_hlist;
1555 * I/O specific fields
1556 * FIXME: Merge these and struct ifmap into one
1558 unsigned long mem_end;
1559 unsigned long mem_start;
1560 unsigned long base_addr;
1563 atomic_t carrier_changes;
1566 * Some hardware also needs these fields (state,dev_list,
1567 * napi_list,unreg_list,close_list) but they are not
1568 * part of the usual set specified in Space.c.
1571 unsigned long state;
1573 struct list_head dev_list;
1574 struct list_head napi_list;
1575 struct list_head unreg_list;
1576 struct list_head close_list;
1577 struct list_head ptype_all;
1578 struct list_head ptype_specific;
1581 struct list_head upper;
1582 struct list_head lower;
1586 struct list_head upper;
1587 struct list_head lower;
1590 netdev_features_t features;
1591 netdev_features_t hw_features;
1592 netdev_features_t wanted_features;
1593 netdev_features_t vlan_features;
1594 netdev_features_t hw_enc_features;
1595 netdev_features_t mpls_features;
1600 struct net_device_stats stats;
1602 atomic_long_t rx_dropped;
1603 atomic_long_t tx_dropped;
1605 #ifdef CONFIG_WIRELESS_EXT
1606 const struct iw_handler_def * wireless_handlers;
1607 struct iw_public_data * wireless_data;
1609 const struct net_device_ops *netdev_ops;
1610 const struct ethtool_ops *ethtool_ops;
1611 #ifdef CONFIG_NET_SWITCHDEV
1612 const struct switchdev_ops *switchdev_ops;
1614 #ifdef CONFIG_NET_L3_MASTER_DEV
1615 const struct l3mdev_ops *l3mdev_ops;
1618 const struct header_ops *header_ops;
1621 unsigned int priv_flags;
1623 unsigned short gflags;
1624 unsigned short padded;
1626 unsigned char operstate;
1627 unsigned char link_mode;
1629 unsigned char if_port;
1633 unsigned short type;
1634 unsigned short hard_header_len;
1635 unsigned short min_header_len;
1637 unsigned short needed_headroom;
1638 unsigned short needed_tailroom;
1640 /* Interface address info. */
1641 unsigned char perm_addr[MAX_ADDR_LEN];
1642 unsigned char addr_assign_type;
1643 unsigned char addr_len;
1644 unsigned short neigh_priv_len;
1645 unsigned short dev_id;
1646 unsigned short dev_port;
1647 spinlock_t addr_list_lock;
1648 unsigned char name_assign_type;
1650 struct netdev_hw_addr_list uc;
1651 struct netdev_hw_addr_list mc;
1652 struct netdev_hw_addr_list dev_addrs;
1655 struct kset *queues_kset;
1657 unsigned int promiscuity;
1658 unsigned int allmulti;
1661 /* Protocol specific pointers */
1663 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1664 struct vlan_info __rcu *vlan_info;
1666 #if IS_ENABLED(CONFIG_NET_DSA)
1667 struct dsa_switch_tree *dsa_ptr;
1669 #if IS_ENABLED(CONFIG_TIPC)
1670 struct tipc_bearer __rcu *tipc_ptr;
1673 struct in_device __rcu *ip_ptr;
1674 struct dn_dev __rcu *dn_ptr;
1675 struct inet6_dev __rcu *ip6_ptr;
1677 struct wireless_dev *ieee80211_ptr;
1678 struct wpan_dev *ieee802154_ptr;
1679 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1680 struct mpls_dev __rcu *mpls_ptr;
1684 * Cache lines mostly used on receive path (including eth_type_trans())
1686 unsigned long last_rx;
1688 /* Interface address info used in eth_type_trans() */
1689 unsigned char *dev_addr;
1693 struct netdev_rx_queue *_rx;
1695 unsigned int num_rx_queues;
1696 unsigned int real_num_rx_queues;
1700 unsigned long gro_flush_timeout;
1701 rx_handler_func_t __rcu *rx_handler;
1702 void __rcu *rx_handler_data;
1704 #ifdef CONFIG_NET_CLS_ACT
1705 struct tcf_proto __rcu *ingress_cl_list;
1707 struct netdev_queue __rcu *ingress_queue;
1708 #ifdef CONFIG_NETFILTER_INGRESS
1709 struct list_head nf_hooks_ingress;
1712 unsigned char broadcast[MAX_ADDR_LEN];
1713 #ifdef CONFIG_RFS_ACCEL
1714 struct cpu_rmap *rx_cpu_rmap;
1716 struct hlist_node index_hlist;
1719 * Cache lines mostly used on transmit path
1721 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1722 unsigned int num_tx_queues;
1723 unsigned int real_num_tx_queues;
1724 struct Qdisc *qdisc;
1725 unsigned long tx_queue_len;
1726 spinlock_t tx_global_lock;
1730 struct xps_dev_maps __rcu *xps_maps;
1733 #ifdef CONFIG_NET_SWITCHDEV
1734 u32 offload_fwd_mark;
1737 /* These may be needed for future network-power-down code. */
1740 * trans_start here is expensive for high speed devices on SMP,
1741 * please use netdev_queue->trans_start instead.
1743 unsigned long trans_start;
1745 struct timer_list watchdog_timer;
1747 int __percpu *pcpu_refcnt;
1748 struct list_head todo_list;
1750 struct list_head link_watch_list;
1752 enum { NETREG_UNINITIALIZED=0,
1753 NETREG_REGISTERED, /* completed register_netdevice */
1754 NETREG_UNREGISTERING, /* called unregister_netdevice */
1755 NETREG_UNREGISTERED, /* completed unregister todo */
1756 NETREG_RELEASED, /* called free_netdev */
1757 NETREG_DUMMY, /* dummy device for NAPI poll */
1763 RTNL_LINK_INITIALIZED,
1764 RTNL_LINK_INITIALIZING,
1765 } rtnl_link_state:16;
1767 void (*destructor)(struct net_device *dev);
1769 #ifdef CONFIG_NETPOLL
1770 struct netpoll_info __rcu *npinfo;
1773 possible_net_t nd_net;
1775 /* mid-layer private */
1778 struct pcpu_lstats __percpu *lstats;
1779 struct pcpu_sw_netstats __percpu *tstats;
1780 struct pcpu_dstats __percpu *dstats;
1781 struct pcpu_vstats __percpu *vstats;
1784 struct garp_port __rcu *garp_port;
1785 struct mrp_port __rcu *mrp_port;
1788 const struct attribute_group *sysfs_groups[4];
1789 const struct attribute_group *sysfs_rx_queue_group;
1791 const struct rtnl_link_ops *rtnl_link_ops;
1793 /* for setting kernel sock attribute on TCP connection setup */
1794 #define GSO_MAX_SIZE 65536
1795 unsigned int gso_max_size;
1796 #define GSO_MAX_SEGS 65535
1800 const struct dcbnl_rtnl_ops *dcbnl_ops;
1803 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1804 u8 prio_tc_map[TC_BITMASK + 1];
1806 #if IS_ENABLED(CONFIG_FCOE)
1807 unsigned int fcoe_ddp_xid;
1809 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1810 struct netprio_map __rcu *priomap;
1812 struct phy_device *phydev;
1813 struct lock_class_key *qdisc_tx_busylock;
1816 #define to_net_dev(d) container_of(d, struct net_device, dev)
1818 #define NETDEV_ALIGN 32
1821 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1823 return dev->prio_tc_map[prio & TC_BITMASK];
1827 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1829 if (tc >= dev->num_tc)
1832 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1837 void netdev_reset_tc(struct net_device *dev)
1840 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1841 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1845 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1847 if (tc >= dev->num_tc)
1850 dev->tc_to_txq[tc].count = count;
1851 dev->tc_to_txq[tc].offset = offset;
1856 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1858 if (num_tc > TC_MAX_QUEUE)
1861 dev->num_tc = num_tc;
1866 int netdev_get_num_tc(struct net_device *dev)
1872 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1875 return &dev->_tx[index];
1878 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1879 const struct sk_buff *skb)
1881 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1884 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1885 void (*f)(struct net_device *,
1886 struct netdev_queue *,
1892 for (i = 0; i < dev->num_tx_queues; i++)
1893 f(dev, &dev->_tx[i], arg);
1896 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1897 struct sk_buff *skb,
1901 * Net namespace inlines
1904 struct net *dev_net(const struct net_device *dev)
1906 return read_pnet(&dev->nd_net);
1910 void dev_net_set(struct net_device *dev, struct net *net)
1912 write_pnet(&dev->nd_net, net);
1915 static inline bool netdev_uses_dsa(struct net_device *dev)
1917 #if IS_ENABLED(CONFIG_NET_DSA)
1918 if (dev->dsa_ptr != NULL)
1919 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1925 * netdev_priv - access network device private data
1926 * @dev: network device
1928 * Get network device private data
1930 static inline void *netdev_priv(const struct net_device *dev)
1932 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1935 /* Set the sysfs physical device reference for the network logical device
1936 * if set prior to registration will cause a symlink during initialization.
1938 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1940 /* Set the sysfs device type for the network logical device to allow
1941 * fine-grained identification of different network device types. For
1942 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1944 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1946 /* Default NAPI poll() weight
1947 * Device drivers are strongly advised to not use bigger value
1949 #define NAPI_POLL_WEIGHT 64
1952 * netif_napi_add - initialize a napi context
1953 * @dev: network device
1954 * @napi: napi context
1955 * @poll: polling function
1956 * @weight: default weight
1958 * netif_napi_add() must be used to initialize a napi context prior to calling
1959 * *any* of the other napi related functions.
1961 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1962 int (*poll)(struct napi_struct *, int), int weight);
1965 * netif_napi_del - remove a napi context
1966 * @napi: napi context
1968 * netif_napi_del() removes a napi context from the network device napi list
1970 void netif_napi_del(struct napi_struct *napi);
1972 struct napi_gro_cb {
1973 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1976 /* Length of frag0. */
1977 unsigned int frag0_len;
1979 /* This indicates where we are processing relative to skb->data. */
1982 /* This is non-zero if the packet cannot be merged with the new skb. */
1985 /* Save the IP ID here and check when we get to the transport layer */
1988 /* Number of segments aggregated. */
1991 /* Start offset for remote checksum offload */
1992 u16 gro_remcsum_start;
1994 /* jiffies when first packet was created/queued */
1997 /* Used in ipv6_gro_receive() and foo-over-udp */
2000 /* This is non-zero if the packet may be of the same flow. */
2003 /* Used in tunnel GRO receive */
2006 /* GRO checksum is valid */
2009 /* Number of checksums via CHECKSUM_UNNECESSARY */
2014 #define NAPI_GRO_FREE 1
2015 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2017 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2020 /* Number of gro_receive callbacks this packet already went through */
2021 u8 recursion_counter:4;
2025 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2028 /* used in skb_gro_receive() slow path */
2029 struct sk_buff *last;
2032 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2034 #define GRO_RECURSION_LIMIT 15
2035 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2037 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2040 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2041 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2042 struct sk_buff **head,
2043 struct sk_buff *skb)
2045 if (unlikely(gro_recursion_inc_test(skb))) {
2046 NAPI_GRO_CB(skb)->flush |= 1;
2050 return cb(head, skb);
2053 struct packet_type {
2054 __be16 type; /* This is really htons(ether_type). */
2055 struct net_device *dev; /* NULL is wildcarded here */
2056 int (*func) (struct sk_buff *,
2057 struct net_device *,
2058 struct packet_type *,
2059 struct net_device *);
2060 bool (*id_match)(struct packet_type *ptype,
2062 void *af_packet_priv;
2063 struct list_head list;
2066 struct offload_callbacks {
2067 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2068 netdev_features_t features);
2069 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2070 struct sk_buff *skb);
2071 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2074 struct packet_offload {
2075 __be16 type; /* This is really htons(ether_type). */
2077 struct offload_callbacks callbacks;
2078 struct list_head list;
2083 struct udp_offload_callbacks {
2084 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2085 struct sk_buff *skb,
2086 struct udp_offload *uoff);
2087 int (*gro_complete)(struct sk_buff *skb,
2089 struct udp_offload *uoff);
2092 struct udp_offload {
2095 struct udp_offload_callbacks callbacks;
2098 typedef struct sk_buff **(*gro_receive_udp_t)(struct sk_buff **,
2100 struct udp_offload *);
2101 static inline struct sk_buff **call_gro_receive_udp(gro_receive_udp_t cb,
2102 struct sk_buff **head,
2103 struct sk_buff *skb,
2104 struct udp_offload *uoff)
2106 if (unlikely(gro_recursion_inc_test(skb))) {
2107 NAPI_GRO_CB(skb)->flush |= 1;
2111 return cb(head, skb, uoff);
2114 /* often modified stats are per cpu, other are shared (netdev->stats) */
2115 struct pcpu_sw_netstats {
2120 struct u64_stats_sync syncp;
2123 #define __netdev_alloc_pcpu_stats(type, gfp) \
2125 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2128 for_each_possible_cpu(__cpu) { \
2129 typeof(type) *stat; \
2130 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2131 u64_stats_init(&stat->syncp); \
2137 #define netdev_alloc_pcpu_stats(type) \
2138 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2140 #include <linux/notifier.h>
2142 /* netdevice notifier chain. Please remember to update the rtnetlink
2143 * notification exclusion list in rtnetlink_event() when adding new
2146 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2147 #define NETDEV_DOWN 0x0002
2148 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2149 detected a hardware crash and restarted
2150 - we can use this eg to kick tcp sessions
2152 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2153 #define NETDEV_REGISTER 0x0005
2154 #define NETDEV_UNREGISTER 0x0006
2155 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2156 #define NETDEV_CHANGEADDR 0x0008
2157 #define NETDEV_GOING_DOWN 0x0009
2158 #define NETDEV_CHANGENAME 0x000A
2159 #define NETDEV_FEAT_CHANGE 0x000B
2160 #define NETDEV_BONDING_FAILOVER 0x000C
2161 #define NETDEV_PRE_UP 0x000D
2162 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2163 #define NETDEV_POST_TYPE_CHANGE 0x000F
2164 #define NETDEV_POST_INIT 0x0010
2165 #define NETDEV_UNREGISTER_FINAL 0x0011
2166 #define NETDEV_RELEASE 0x0012
2167 #define NETDEV_NOTIFY_PEERS 0x0013
2168 #define NETDEV_JOIN 0x0014
2169 #define NETDEV_CHANGEUPPER 0x0015
2170 #define NETDEV_RESEND_IGMP 0x0016
2171 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2172 #define NETDEV_CHANGEINFODATA 0x0018
2173 #define NETDEV_BONDING_INFO 0x0019
2174 #define NETDEV_PRECHANGEUPPER 0x001A
2176 int register_netdevice_notifier(struct notifier_block *nb);
2177 int unregister_netdevice_notifier(struct notifier_block *nb);
2179 struct netdev_notifier_info {
2180 struct net_device *dev;
2183 struct netdev_notifier_change_info {
2184 struct netdev_notifier_info info; /* must be first */
2185 unsigned int flags_changed;
2188 struct netdev_notifier_changeupper_info {
2189 struct netdev_notifier_info info; /* must be first */
2190 struct net_device *upper_dev; /* new upper dev */
2191 bool master; /* is upper dev master */
2192 bool linking; /* is the nofication for link or unlink */
2195 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2196 struct net_device *dev)
2201 static inline struct net_device *
2202 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2207 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2210 extern rwlock_t dev_base_lock; /* Device list lock */
2212 #define for_each_netdev(net, d) \
2213 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2214 #define for_each_netdev_reverse(net, d) \
2215 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2216 #define for_each_netdev_rcu(net, d) \
2217 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2218 #define for_each_netdev_safe(net, d, n) \
2219 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2220 #define for_each_netdev_continue(net, d) \
2221 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2222 #define for_each_netdev_continue_rcu(net, d) \
2223 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2224 #define for_each_netdev_in_bond_rcu(bond, slave) \
2225 for_each_netdev_rcu(&init_net, slave) \
2226 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2227 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2229 static inline struct net_device *next_net_device(struct net_device *dev)
2231 struct list_head *lh;
2235 lh = dev->dev_list.next;
2236 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2239 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2241 struct list_head *lh;
2245 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2246 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2249 static inline struct net_device *first_net_device(struct net *net)
2251 return list_empty(&net->dev_base_head) ? NULL :
2252 net_device_entry(net->dev_base_head.next);
2255 static inline struct net_device *first_net_device_rcu(struct net *net)
2257 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2259 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2262 int netdev_boot_setup_check(struct net_device *dev);
2263 unsigned long netdev_boot_base(const char *prefix, int unit);
2264 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2265 const char *hwaddr);
2266 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2267 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2268 void dev_add_pack(struct packet_type *pt);
2269 void dev_remove_pack(struct packet_type *pt);
2270 void __dev_remove_pack(struct packet_type *pt);
2271 void dev_add_offload(struct packet_offload *po);
2272 void dev_remove_offload(struct packet_offload *po);
2274 int dev_get_iflink(const struct net_device *dev);
2275 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2276 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2277 unsigned short mask);
2278 struct net_device *dev_get_by_name(struct net *net, const char *name);
2279 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2280 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2281 int dev_alloc_name(struct net_device *dev, const char *name);
2282 int dev_open(struct net_device *dev);
2283 int dev_close(struct net_device *dev);
2284 int dev_close_many(struct list_head *head, bool unlink);
2285 void dev_disable_lro(struct net_device *dev);
2286 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2287 int dev_queue_xmit(struct sk_buff *skb);
2288 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2289 int register_netdevice(struct net_device *dev);
2290 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2291 void unregister_netdevice_many(struct list_head *head);
2292 static inline void unregister_netdevice(struct net_device *dev)
2294 unregister_netdevice_queue(dev, NULL);
2297 int netdev_refcnt_read(const struct net_device *dev);
2298 void free_netdev(struct net_device *dev);
2299 void netdev_freemem(struct net_device *dev);
2300 void synchronize_net(void);
2301 int init_dummy_netdev(struct net_device *dev);
2303 #ifdef CONFIG_PREEMPT_RT_FULL
2304 static inline int dev_recursion_level(void)
2306 return current->xmit_recursion;
2311 DECLARE_PER_CPU(int, xmit_recursion);
2312 static inline int dev_recursion_level(void)
2314 return this_cpu_read(xmit_recursion);
2318 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2319 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2320 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2321 int netdev_get_name(struct net *net, char *name, int ifindex);
2322 int dev_restart(struct net_device *dev);
2323 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2325 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2327 return NAPI_GRO_CB(skb)->data_offset;
2330 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2332 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2335 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2337 NAPI_GRO_CB(skb)->data_offset += len;
2340 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2341 unsigned int offset)
2343 return NAPI_GRO_CB(skb)->frag0 + offset;
2346 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2348 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2351 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2353 NAPI_GRO_CB(skb)->frag0 = NULL;
2354 NAPI_GRO_CB(skb)->frag0_len = 0;
2357 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2358 unsigned int offset)
2360 if (!pskb_may_pull(skb, hlen))
2363 skb_gro_frag0_invalidate(skb);
2364 return skb->data + offset;
2367 static inline void *skb_gro_network_header(struct sk_buff *skb)
2369 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2370 skb_network_offset(skb);
2373 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2374 const void *start, unsigned int len)
2376 if (NAPI_GRO_CB(skb)->csum_valid)
2377 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2378 csum_partial(start, len, 0));
2381 /* GRO checksum functions. These are logical equivalents of the normal
2382 * checksum functions (in skbuff.h) except that they operate on the GRO
2383 * offsets and fields in sk_buff.
2386 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2388 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2390 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2393 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2397 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2398 skb_checksum_start_offset(skb) <
2399 skb_gro_offset(skb)) &&
2400 !skb_at_gro_remcsum_start(skb) &&
2401 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2402 (!zero_okay || check));
2405 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2408 if (NAPI_GRO_CB(skb)->csum_valid &&
2409 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2412 NAPI_GRO_CB(skb)->csum = psum;
2414 return __skb_gro_checksum_complete(skb);
2417 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2419 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2420 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2421 NAPI_GRO_CB(skb)->csum_cnt--;
2423 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2424 * verified a new top level checksum or an encapsulated one
2425 * during GRO. This saves work if we fallback to normal path.
2427 __skb_incr_checksum_unnecessary(skb);
2431 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2434 __sum16 __ret = 0; \
2435 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2436 __ret = __skb_gro_checksum_validate_complete(skb, \
2437 compute_pseudo(skb, proto)); \
2439 __skb_mark_checksum_bad(skb); \
2441 skb_gro_incr_csum_unnecessary(skb); \
2445 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2446 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2448 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2450 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2452 #define skb_gro_checksum_simple_validate(skb) \
2453 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2455 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2457 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2458 !NAPI_GRO_CB(skb)->csum_valid);
2461 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2462 __sum16 check, __wsum pseudo)
2464 NAPI_GRO_CB(skb)->csum = ~pseudo;
2465 NAPI_GRO_CB(skb)->csum_valid = 1;
2468 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2470 if (__skb_gro_checksum_convert_check(skb)) \
2471 __skb_gro_checksum_convert(skb, check, \
2472 compute_pseudo(skb, proto)); \
2475 struct gro_remcsum {
2480 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2486 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2487 unsigned int off, size_t hdrlen,
2488 int start, int offset,
2489 struct gro_remcsum *grc,
2493 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2495 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2498 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2502 ptr = skb_gro_header_fast(skb, off);
2503 if (skb_gro_header_hard(skb, off + plen)) {
2504 ptr = skb_gro_header_slow(skb, off + plen, off);
2509 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2512 /* Adjust skb->csum since we changed the packet */
2513 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2515 grc->offset = off + hdrlen + offset;
2521 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2522 struct gro_remcsum *grc)
2525 size_t plen = grc->offset + sizeof(u16);
2530 ptr = skb_gro_header_fast(skb, grc->offset);
2531 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2532 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2537 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2540 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2541 unsigned short type,
2542 const void *daddr, const void *saddr,
2545 if (!dev->header_ops || !dev->header_ops->create)
2548 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2551 static inline int dev_parse_header(const struct sk_buff *skb,
2552 unsigned char *haddr)
2554 const struct net_device *dev = skb->dev;
2556 if (!dev->header_ops || !dev->header_ops->parse)
2558 return dev->header_ops->parse(skb, haddr);
2561 /* ll_header must have at least hard_header_len allocated */
2562 static inline bool dev_validate_header(const struct net_device *dev,
2563 char *ll_header, int len)
2565 if (likely(len >= dev->hard_header_len))
2567 if (len < dev->min_header_len)
2570 if (capable(CAP_SYS_RAWIO)) {
2571 memset(ll_header + len, 0, dev->hard_header_len - len);
2575 if (dev->header_ops && dev->header_ops->validate)
2576 return dev->header_ops->validate(ll_header, len);
2581 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2582 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2583 static inline int unregister_gifconf(unsigned int family)
2585 return register_gifconf(family, NULL);
2588 #ifdef CONFIG_NET_FLOW_LIMIT
2589 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2590 struct sd_flow_limit {
2592 unsigned int num_buckets;
2593 unsigned int history_head;
2594 u16 history[FLOW_LIMIT_HISTORY];
2598 extern int netdev_flow_limit_table_len;
2599 #endif /* CONFIG_NET_FLOW_LIMIT */
2602 * Incoming packets are placed on per-cpu queues
2604 struct softnet_data {
2605 struct list_head poll_list;
2606 struct sk_buff_head process_queue;
2609 unsigned int processed;
2610 unsigned int time_squeeze;
2611 unsigned int cpu_collision;
2612 unsigned int received_rps;
2614 struct softnet_data *rps_ipi_list;
2616 #ifdef CONFIG_NET_FLOW_LIMIT
2617 struct sd_flow_limit __rcu *flow_limit;
2619 struct Qdisc *output_queue;
2620 struct Qdisc **output_queue_tailp;
2621 struct sk_buff *completion_queue;
2624 /* Elements below can be accessed between CPUs for RPS */
2625 struct call_single_data csd ____cacheline_aligned_in_smp;
2626 struct softnet_data *rps_ipi_next;
2628 unsigned int input_queue_head;
2629 unsigned int input_queue_tail;
2631 unsigned int dropped;
2632 struct sk_buff_head input_pkt_queue;
2633 struct napi_struct backlog;
2634 struct sk_buff_head tofree_queue;
2638 static inline void input_queue_head_incr(struct softnet_data *sd)
2641 sd->input_queue_head++;
2645 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2646 unsigned int *qtail)
2649 *qtail = ++sd->input_queue_tail;
2653 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2655 void __netif_schedule(struct Qdisc *q);
2656 void netif_schedule_queue(struct netdev_queue *txq);
2658 static inline void netif_tx_schedule_all(struct net_device *dev)
2662 for (i = 0; i < dev->num_tx_queues; i++)
2663 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2666 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2668 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2672 * netif_start_queue - allow transmit
2673 * @dev: network device
2675 * Allow upper layers to call the device hard_start_xmit routine.
2677 static inline void netif_start_queue(struct net_device *dev)
2679 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2682 static inline void netif_tx_start_all_queues(struct net_device *dev)
2686 for (i = 0; i < dev->num_tx_queues; i++) {
2687 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2688 netif_tx_start_queue(txq);
2692 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2695 * netif_wake_queue - restart transmit
2696 * @dev: network device
2698 * Allow upper layers to call the device hard_start_xmit routine.
2699 * Used for flow control when transmit resources are available.
2701 static inline void netif_wake_queue(struct net_device *dev)
2703 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2706 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2710 for (i = 0; i < dev->num_tx_queues; i++) {
2711 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2712 netif_tx_wake_queue(txq);
2716 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2718 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2722 * netif_stop_queue - stop transmitted packets
2723 * @dev: network device
2725 * Stop upper layers calling the device hard_start_xmit routine.
2726 * Used for flow control when transmit resources are unavailable.
2728 static inline void netif_stop_queue(struct net_device *dev)
2730 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2733 void netif_tx_stop_all_queues(struct net_device *dev);
2735 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2737 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2741 * netif_queue_stopped - test if transmit queue is flowblocked
2742 * @dev: network device
2744 * Test if transmit queue on device is currently unable to send.
2746 static inline bool netif_queue_stopped(const struct net_device *dev)
2748 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2751 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2753 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2757 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2759 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2763 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2765 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2769 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2770 * @dev_queue: pointer to transmit queue
2772 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2773 * to give appropriate hint to the cpu.
2775 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2778 prefetchw(&dev_queue->dql.num_queued);
2783 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2784 * @dev_queue: pointer to transmit queue
2786 * BQL enabled drivers might use this helper in their TX completion path,
2787 * to give appropriate hint to the cpu.
2789 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2792 prefetchw(&dev_queue->dql.limit);
2796 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2800 dql_queued(&dev_queue->dql, bytes);
2802 if (likely(dql_avail(&dev_queue->dql) >= 0))
2805 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2808 * The XOFF flag must be set before checking the dql_avail below,
2809 * because in netdev_tx_completed_queue we update the dql_completed
2810 * before checking the XOFF flag.
2814 /* check again in case another CPU has just made room avail */
2815 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2816 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2821 * netdev_sent_queue - report the number of bytes queued to hardware
2822 * @dev: network device
2823 * @bytes: number of bytes queued to the hardware device queue
2825 * Report the number of bytes queued for sending/completion to the network
2826 * device hardware queue. @bytes should be a good approximation and should
2827 * exactly match netdev_completed_queue() @bytes
2829 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2831 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2834 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2835 unsigned int pkts, unsigned int bytes)
2838 if (unlikely(!bytes))
2841 dql_completed(&dev_queue->dql, bytes);
2844 * Without the memory barrier there is a small possiblity that
2845 * netdev_tx_sent_queue will miss the update and cause the queue to
2846 * be stopped forever
2850 if (dql_avail(&dev_queue->dql) < 0)
2853 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2854 netif_schedule_queue(dev_queue);
2859 * netdev_completed_queue - report bytes and packets completed by device
2860 * @dev: network device
2861 * @pkts: actual number of packets sent over the medium
2862 * @bytes: actual number of bytes sent over the medium
2864 * Report the number of bytes and packets transmitted by the network device
2865 * hardware queue over the physical medium, @bytes must exactly match the
2866 * @bytes amount passed to netdev_sent_queue()
2868 static inline void netdev_completed_queue(struct net_device *dev,
2869 unsigned int pkts, unsigned int bytes)
2871 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2874 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2877 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2883 * netdev_reset_queue - reset the packets and bytes count of a network device
2884 * @dev_queue: network device
2886 * Reset the bytes and packet count of a network device and clear the
2887 * software flow control OFF bit for this network device
2889 static inline void netdev_reset_queue(struct net_device *dev_queue)
2891 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2895 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2896 * @dev: network device
2897 * @queue_index: given tx queue index
2899 * Returns 0 if given tx queue index >= number of device tx queues,
2900 * otherwise returns the originally passed tx queue index.
2902 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2904 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2905 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2906 dev->name, queue_index,
2907 dev->real_num_tx_queues);
2915 * netif_running - test if up
2916 * @dev: network device
2918 * Test if the device has been brought up.
2920 static inline bool netif_running(const struct net_device *dev)
2922 return test_bit(__LINK_STATE_START, &dev->state);
2926 * Routines to manage the subqueues on a device. We only need start
2927 * stop, and a check if it's stopped. All other device management is
2928 * done at the overall netdevice level.
2929 * Also test the device if we're multiqueue.
2933 * netif_start_subqueue - allow sending packets on subqueue
2934 * @dev: network device
2935 * @queue_index: sub queue index
2937 * Start individual transmit queue of a device with multiple transmit queues.
2939 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2941 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2943 netif_tx_start_queue(txq);
2947 * netif_stop_subqueue - stop sending packets on subqueue
2948 * @dev: network device
2949 * @queue_index: sub queue index
2951 * Stop individual transmit queue of a device with multiple transmit queues.
2953 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2955 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2956 netif_tx_stop_queue(txq);
2960 * netif_subqueue_stopped - test status of subqueue
2961 * @dev: network device
2962 * @queue_index: sub queue index
2964 * Check individual transmit queue of a device with multiple transmit queues.
2966 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2969 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2971 return netif_tx_queue_stopped(txq);
2974 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2975 struct sk_buff *skb)
2977 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2980 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2983 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2986 static inline int netif_set_xps_queue(struct net_device *dev,
2987 const struct cpumask *mask,
2994 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2995 unsigned int num_tx_queues);
2998 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2999 * as a distribution range limit for the returned value.
3001 static inline u16 skb_tx_hash(const struct net_device *dev,
3002 struct sk_buff *skb)
3004 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3008 * netif_is_multiqueue - test if device has multiple transmit queues
3009 * @dev: network device
3011 * Check if device has multiple transmit queues
3013 static inline bool netif_is_multiqueue(const struct net_device *dev)
3015 return dev->num_tx_queues > 1;
3018 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3021 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3023 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3031 static inline unsigned int get_netdev_rx_queue_index(
3032 struct netdev_rx_queue *queue)
3034 struct net_device *dev = queue->dev;
3035 int index = queue - dev->_rx;
3037 BUG_ON(index >= dev->num_rx_queues);
3042 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3043 int netif_get_num_default_rss_queues(void);
3045 enum skb_free_reason {
3046 SKB_REASON_CONSUMED,
3050 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3051 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3054 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3055 * interrupt context or with hardware interrupts being disabled.
3056 * (in_irq() || irqs_disabled())
3058 * We provide four helpers that can be used in following contexts :
3060 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3061 * replacing kfree_skb(skb)
3063 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3064 * Typically used in place of consume_skb(skb) in TX completion path
3066 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3067 * replacing kfree_skb(skb)
3069 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3070 * and consumed a packet. Used in place of consume_skb(skb)
3072 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3074 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3077 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3079 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3082 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3084 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3087 static inline void dev_consume_skb_any(struct sk_buff *skb)
3089 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3092 int netif_rx(struct sk_buff *skb);
3093 int netif_rx_ni(struct sk_buff *skb);
3094 int netif_receive_skb(struct sk_buff *skb);
3095 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3096 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3097 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3098 gro_result_t napi_gro_frags(struct napi_struct *napi);
3099 struct packet_offload *gro_find_receive_by_type(__be16 type);
3100 struct packet_offload *gro_find_complete_by_type(__be16 type);
3102 static inline void napi_free_frags(struct napi_struct *napi)
3104 kfree_skb(napi->skb);
3108 bool netdev_is_rx_handler_busy(struct net_device *dev);
3109 int netdev_rx_handler_register(struct net_device *dev,
3110 rx_handler_func_t *rx_handler,
3111 void *rx_handler_data);
3112 void netdev_rx_handler_unregister(struct net_device *dev);
3114 bool dev_valid_name(const char *name);
3115 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3116 int dev_ethtool(struct net *net, struct ifreq *);
3117 unsigned int dev_get_flags(const struct net_device *);
3118 int __dev_change_flags(struct net_device *, unsigned int flags);
3119 int dev_change_flags(struct net_device *, unsigned int);
3120 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3121 unsigned int gchanges);
3122 int dev_change_name(struct net_device *, const char *);
3123 int dev_set_alias(struct net_device *, const char *, size_t);
3124 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3125 int dev_set_mtu(struct net_device *, int);
3126 void dev_set_group(struct net_device *, int);
3127 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3128 int dev_change_carrier(struct net_device *, bool new_carrier);
3129 int dev_get_phys_port_id(struct net_device *dev,
3130 struct netdev_phys_item_id *ppid);
3131 int dev_get_phys_port_name(struct net_device *dev,
3132 char *name, size_t len);
3133 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3134 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3135 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3136 struct netdev_queue *txq, int *ret);
3137 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3138 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3139 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
3141 extern int netdev_budget;
3143 /* Called by rtnetlink.c:rtnl_unlock() */
3144 void netdev_run_todo(void);
3147 * dev_put - release reference to device
3148 * @dev: network device
3150 * Release reference to device to allow it to be freed.
3152 static inline void dev_put(struct net_device *dev)
3154 this_cpu_dec(*dev->pcpu_refcnt);
3158 * dev_hold - get reference to device
3159 * @dev: network device
3161 * Hold reference to device to keep it from being freed.
3163 static inline void dev_hold(struct net_device *dev)
3165 this_cpu_inc(*dev->pcpu_refcnt);
3168 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3169 * and _off may be called from IRQ context, but it is caller
3170 * who is responsible for serialization of these calls.
3172 * The name carrier is inappropriate, these functions should really be
3173 * called netif_lowerlayer_*() because they represent the state of any
3174 * kind of lower layer not just hardware media.
3177 void linkwatch_init_dev(struct net_device *dev);
3178 void linkwatch_fire_event(struct net_device *dev);
3179 void linkwatch_forget_dev(struct net_device *dev);
3182 * netif_carrier_ok - test if carrier present
3183 * @dev: network device
3185 * Check if carrier is present on device
3187 static inline bool netif_carrier_ok(const struct net_device *dev)
3189 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3192 unsigned long dev_trans_start(struct net_device *dev);
3194 void __netdev_watchdog_up(struct net_device *dev);
3196 void netif_carrier_on(struct net_device *dev);
3198 void netif_carrier_off(struct net_device *dev);
3201 * netif_dormant_on - mark device as dormant.
3202 * @dev: network device
3204 * Mark device as dormant (as per RFC2863).
3206 * The dormant state indicates that the relevant interface is not
3207 * actually in a condition to pass packets (i.e., it is not 'up') but is
3208 * in a "pending" state, waiting for some external event. For "on-
3209 * demand" interfaces, this new state identifies the situation where the
3210 * interface is waiting for events to place it in the up state.
3213 static inline void netif_dormant_on(struct net_device *dev)
3215 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3216 linkwatch_fire_event(dev);
3220 * netif_dormant_off - set device as not dormant.
3221 * @dev: network device
3223 * Device is not in dormant state.
3225 static inline void netif_dormant_off(struct net_device *dev)
3227 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3228 linkwatch_fire_event(dev);
3232 * netif_dormant - test if carrier present
3233 * @dev: network device
3235 * Check if carrier is present on device
3237 static inline bool netif_dormant(const struct net_device *dev)
3239 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3244 * netif_oper_up - test if device is operational
3245 * @dev: network device
3247 * Check if carrier is operational
3249 static inline bool netif_oper_up(const struct net_device *dev)
3251 return (dev->operstate == IF_OPER_UP ||
3252 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3256 * netif_device_present - is device available or removed
3257 * @dev: network device
3259 * Check if device has not been removed from system.
3261 static inline bool netif_device_present(struct net_device *dev)
3263 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3266 void netif_device_detach(struct net_device *dev);
3268 void netif_device_attach(struct net_device *dev);
3271 * Network interface message level settings
3275 NETIF_MSG_DRV = 0x0001,
3276 NETIF_MSG_PROBE = 0x0002,
3277 NETIF_MSG_LINK = 0x0004,
3278 NETIF_MSG_TIMER = 0x0008,
3279 NETIF_MSG_IFDOWN = 0x0010,
3280 NETIF_MSG_IFUP = 0x0020,
3281 NETIF_MSG_RX_ERR = 0x0040,
3282 NETIF_MSG_TX_ERR = 0x0080,
3283 NETIF_MSG_TX_QUEUED = 0x0100,
3284 NETIF_MSG_INTR = 0x0200,
3285 NETIF_MSG_TX_DONE = 0x0400,
3286 NETIF_MSG_RX_STATUS = 0x0800,
3287 NETIF_MSG_PKTDATA = 0x1000,
3288 NETIF_MSG_HW = 0x2000,
3289 NETIF_MSG_WOL = 0x4000,
3292 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3293 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3294 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3295 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3296 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3297 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3298 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3299 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3300 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3301 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3302 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3303 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3304 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3305 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3306 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3308 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3311 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3312 return default_msg_enable_bits;
3313 if (debug_value == 0) /* no output */
3315 /* set low N bits */
3316 return (1 << debug_value) - 1;
3319 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3321 spin_lock(&txq->_xmit_lock);
3322 txq->xmit_lock_owner = cpu;
3325 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3327 spin_lock_bh(&txq->_xmit_lock);
3328 txq->xmit_lock_owner = smp_processor_id();
3331 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3333 bool ok = spin_trylock(&txq->_xmit_lock);
3335 txq->xmit_lock_owner = smp_processor_id();
3339 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3341 txq->xmit_lock_owner = -1;
3342 spin_unlock(&txq->_xmit_lock);
3345 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3347 txq->xmit_lock_owner = -1;
3348 spin_unlock_bh(&txq->_xmit_lock);
3351 static inline void txq_trans_update(struct netdev_queue *txq)
3353 if (txq->xmit_lock_owner != -1)
3354 txq->trans_start = jiffies;
3358 * netif_tx_lock - grab network device transmit lock
3359 * @dev: network device
3361 * Get network device transmit lock
3363 static inline void netif_tx_lock(struct net_device *dev)
3368 spin_lock(&dev->tx_global_lock);
3369 cpu = smp_processor_id();
3370 for (i = 0; i < dev->num_tx_queues; i++) {
3371 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3373 /* We are the only thread of execution doing a
3374 * freeze, but we have to grab the _xmit_lock in
3375 * order to synchronize with threads which are in
3376 * the ->hard_start_xmit() handler and already
3377 * checked the frozen bit.
3379 __netif_tx_lock(txq, cpu);
3380 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3381 __netif_tx_unlock(txq);
3385 static inline void netif_tx_lock_bh(struct net_device *dev)
3391 static inline void netif_tx_unlock(struct net_device *dev)
3395 for (i = 0; i < dev->num_tx_queues; i++) {
3396 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3398 /* No need to grab the _xmit_lock here. If the
3399 * queue is not stopped for another reason, we
3402 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3403 netif_schedule_queue(txq);
3405 spin_unlock(&dev->tx_global_lock);
3408 static inline void netif_tx_unlock_bh(struct net_device *dev)
3410 netif_tx_unlock(dev);
3414 #define HARD_TX_LOCK(dev, txq, cpu) { \
3415 if ((dev->features & NETIF_F_LLTX) == 0) { \
3416 __netif_tx_lock(txq, cpu); \
3420 #define HARD_TX_TRYLOCK(dev, txq) \
3421 (((dev->features & NETIF_F_LLTX) == 0) ? \
3422 __netif_tx_trylock(txq) : \
3425 #define HARD_TX_UNLOCK(dev, txq) { \
3426 if ((dev->features & NETIF_F_LLTX) == 0) { \
3427 __netif_tx_unlock(txq); \
3431 static inline void netif_tx_disable(struct net_device *dev)
3437 cpu = smp_processor_id();
3438 for (i = 0; i < dev->num_tx_queues; i++) {
3439 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3441 __netif_tx_lock(txq, cpu);
3442 netif_tx_stop_queue(txq);
3443 __netif_tx_unlock(txq);
3448 static inline void netif_addr_lock(struct net_device *dev)
3450 spin_lock(&dev->addr_list_lock);
3453 static inline void netif_addr_lock_nested(struct net_device *dev)
3455 int subclass = SINGLE_DEPTH_NESTING;
3457 if (dev->netdev_ops->ndo_get_lock_subclass)
3458 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3460 spin_lock_nested(&dev->addr_list_lock, subclass);
3463 static inline void netif_addr_lock_bh(struct net_device *dev)
3465 spin_lock_bh(&dev->addr_list_lock);
3468 static inline void netif_addr_unlock(struct net_device *dev)
3470 spin_unlock(&dev->addr_list_lock);
3473 static inline void netif_addr_unlock_bh(struct net_device *dev)
3475 spin_unlock_bh(&dev->addr_list_lock);
3479 * dev_addrs walker. Should be used only for read access. Call with
3480 * rcu_read_lock held.
3482 #define for_each_dev_addr(dev, ha) \
3483 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3485 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3487 void ether_setup(struct net_device *dev);
3489 /* Support for loadable net-drivers */
3490 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3491 unsigned char name_assign_type,
3492 void (*setup)(struct net_device *),
3493 unsigned int txqs, unsigned int rxqs);
3494 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3495 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3497 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3498 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3501 int register_netdev(struct net_device *dev);
3502 void unregister_netdev(struct net_device *dev);
3504 /* General hardware address lists handling functions */
3505 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3506 struct netdev_hw_addr_list *from_list, int addr_len);
3507 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3508 struct netdev_hw_addr_list *from_list, int addr_len);
3509 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3510 struct net_device *dev,
3511 int (*sync)(struct net_device *, const unsigned char *),
3512 int (*unsync)(struct net_device *,
3513 const unsigned char *));
3514 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3515 struct net_device *dev,
3516 int (*unsync)(struct net_device *,
3517 const unsigned char *));
3518 void __hw_addr_init(struct netdev_hw_addr_list *list);
3520 /* Functions used for device addresses handling */
3521 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3522 unsigned char addr_type);
3523 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3524 unsigned char addr_type);
3525 void dev_addr_flush(struct net_device *dev);
3526 int dev_addr_init(struct net_device *dev);
3528 /* Functions used for unicast addresses handling */
3529 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3530 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3531 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3532 int dev_uc_sync(struct net_device *to, struct net_device *from);
3533 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3534 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3535 void dev_uc_flush(struct net_device *dev);
3536 void dev_uc_init(struct net_device *dev);
3539 * __dev_uc_sync - Synchonize device's unicast list
3540 * @dev: device to sync
3541 * @sync: function to call if address should be added
3542 * @unsync: function to call if address should be removed
3544 * Add newly added addresses to the interface, and release
3545 * addresses that have been deleted.
3547 static inline int __dev_uc_sync(struct net_device *dev,
3548 int (*sync)(struct net_device *,
3549 const unsigned char *),
3550 int (*unsync)(struct net_device *,
3551 const unsigned char *))
3553 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3557 * __dev_uc_unsync - Remove synchronized addresses from device
3558 * @dev: device to sync
3559 * @unsync: function to call if address should be removed
3561 * Remove all addresses that were added to the device by dev_uc_sync().
3563 static inline void __dev_uc_unsync(struct net_device *dev,
3564 int (*unsync)(struct net_device *,
3565 const unsigned char *))
3567 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3570 /* Functions used for multicast addresses handling */
3571 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3572 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3573 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3574 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3575 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3576 int dev_mc_sync(struct net_device *to, struct net_device *from);
3577 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3578 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3579 void dev_mc_flush(struct net_device *dev);
3580 void dev_mc_init(struct net_device *dev);
3583 * __dev_mc_sync - Synchonize device's multicast list
3584 * @dev: device to sync
3585 * @sync: function to call if address should be added
3586 * @unsync: function to call if address should be removed
3588 * Add newly added addresses to the interface, and release
3589 * addresses that have been deleted.
3591 static inline int __dev_mc_sync(struct net_device *dev,
3592 int (*sync)(struct net_device *,
3593 const unsigned char *),
3594 int (*unsync)(struct net_device *,
3595 const unsigned char *))
3597 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3601 * __dev_mc_unsync - Remove synchronized addresses from device
3602 * @dev: device to sync
3603 * @unsync: function to call if address should be removed
3605 * Remove all addresses that were added to the device by dev_mc_sync().
3607 static inline void __dev_mc_unsync(struct net_device *dev,
3608 int (*unsync)(struct net_device *,
3609 const unsigned char *))
3611 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3614 /* Functions used for secondary unicast and multicast support */
3615 void dev_set_rx_mode(struct net_device *dev);
3616 void __dev_set_rx_mode(struct net_device *dev);
3617 int dev_set_promiscuity(struct net_device *dev, int inc);
3618 int dev_set_allmulti(struct net_device *dev, int inc);
3619 void netdev_state_change(struct net_device *dev);
3620 void netdev_notify_peers(struct net_device *dev);
3621 void netdev_features_change(struct net_device *dev);
3622 /* Load a device via the kmod */
3623 void dev_load(struct net *net, const char *name);
3624 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3625 struct rtnl_link_stats64 *storage);
3626 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3627 const struct net_device_stats *netdev_stats);
3629 extern int netdev_max_backlog;
3630 extern int netdev_tstamp_prequeue;
3631 extern int weight_p;
3632 extern int bpf_jit_enable;
3634 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3635 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3636 struct list_head **iter);
3637 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3638 struct list_head **iter);
3640 /* iterate through upper list, must be called under RCU read lock */
3641 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3642 for (iter = &(dev)->adj_list.upper, \
3643 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3645 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3647 /* iterate through upper list, must be called under RCU read lock */
3648 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3649 for (iter = &(dev)->all_adj_list.upper, \
3650 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3652 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3654 void *netdev_lower_get_next_private(struct net_device *dev,
3655 struct list_head **iter);
3656 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3657 struct list_head **iter);
3659 #define netdev_for_each_lower_private(dev, priv, iter) \
3660 for (iter = (dev)->adj_list.lower.next, \
3661 priv = netdev_lower_get_next_private(dev, &(iter)); \
3663 priv = netdev_lower_get_next_private(dev, &(iter)))
3665 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3666 for (iter = &(dev)->adj_list.lower, \
3667 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3669 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3671 void *netdev_lower_get_next(struct net_device *dev,
3672 struct list_head **iter);
3673 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3674 for (iter = &(dev)->adj_list.lower, \
3675 ldev = netdev_lower_get_next(dev, &(iter)); \
3677 ldev = netdev_lower_get_next(dev, &(iter)))
3679 void *netdev_adjacent_get_private(struct list_head *adj_list);
3680 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3681 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3682 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3683 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3684 int netdev_master_upper_dev_link(struct net_device *dev,
3685 struct net_device *upper_dev);
3686 int netdev_master_upper_dev_link_private(struct net_device *dev,
3687 struct net_device *upper_dev,
3689 void netdev_upper_dev_unlink(struct net_device *dev,
3690 struct net_device *upper_dev);
3691 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3692 void *netdev_lower_dev_get_private(struct net_device *dev,
3693 struct net_device *lower_dev);
3695 /* RSS keys are 40 or 52 bytes long */
3696 #define NETDEV_RSS_KEY_LEN 52
3697 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3698 void netdev_rss_key_fill(void *buffer, size_t len);
3700 int dev_get_nest_level(struct net_device *dev,
3701 bool (*type_check)(struct net_device *dev));
3702 int skb_checksum_help(struct sk_buff *skb);
3703 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3704 netdev_features_t features, bool tx_path);
3705 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3706 netdev_features_t features);
3708 struct netdev_bonding_info {
3713 struct netdev_notifier_bonding_info {
3714 struct netdev_notifier_info info; /* must be first */
3715 struct netdev_bonding_info bonding_info;
3718 void netdev_bonding_info_change(struct net_device *dev,
3719 struct netdev_bonding_info *bonding_info);
3722 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3724 return __skb_gso_segment(skb, features, true);
3726 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3728 static inline bool can_checksum_protocol(netdev_features_t features,
3731 return ((features & NETIF_F_GEN_CSUM) ||
3732 ((features & NETIF_F_V4_CSUM) &&
3733 protocol == htons(ETH_P_IP)) ||
3734 ((features & NETIF_F_V6_CSUM) &&
3735 protocol == htons(ETH_P_IPV6)) ||
3736 ((features & NETIF_F_FCOE_CRC) &&
3737 protocol == htons(ETH_P_FCOE)));
3741 void netdev_rx_csum_fault(struct net_device *dev);
3743 static inline void netdev_rx_csum_fault(struct net_device *dev)
3747 /* rx skb timestamps */
3748 void net_enable_timestamp(void);
3749 void net_disable_timestamp(void);
3751 #ifdef CONFIG_PROC_FS
3752 int __init dev_proc_init(void);
3754 #define dev_proc_init() 0
3757 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3758 struct sk_buff *skb, struct net_device *dev,
3761 skb->xmit_more = more ? 1 : 0;
3762 return ops->ndo_start_xmit(skb, dev);
3765 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3766 struct netdev_queue *txq, bool more)
3768 const struct net_device_ops *ops = dev->netdev_ops;
3771 rc = __netdev_start_xmit(ops, skb, dev, more);
3772 if (rc == NETDEV_TX_OK)
3773 txq_trans_update(txq);
3778 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3780 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3783 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3785 return netdev_class_create_file_ns(class_attr, NULL);
3788 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3790 netdev_class_remove_file_ns(class_attr, NULL);
3793 extern struct kobj_ns_type_operations net_ns_type_operations;
3795 const char *netdev_drivername(const struct net_device *dev);
3797 void linkwatch_run_queue(void);
3799 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3800 netdev_features_t f2)
3802 if (f1 & NETIF_F_GEN_CSUM)
3803 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3804 if (f2 & NETIF_F_GEN_CSUM)
3805 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3807 if (f1 & NETIF_F_GEN_CSUM)
3808 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3813 static inline netdev_features_t netdev_get_wanted_features(
3814 struct net_device *dev)
3816 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3818 netdev_features_t netdev_increment_features(netdev_features_t all,
3819 netdev_features_t one, netdev_features_t mask);
3821 /* Allow TSO being used on stacked device :
3822 * Performing the GSO segmentation before last device
3823 * is a performance improvement.
3825 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3826 netdev_features_t mask)
3828 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3831 int __netdev_update_features(struct net_device *dev);
3832 void netdev_update_features(struct net_device *dev);
3833 void netdev_change_features(struct net_device *dev);
3835 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3836 struct net_device *dev);
3838 netdev_features_t passthru_features_check(struct sk_buff *skb,
3839 struct net_device *dev,
3840 netdev_features_t features);
3841 netdev_features_t netif_skb_features(struct sk_buff *skb);
3843 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3845 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3847 /* check flags correspondence */
3848 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3849 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3850 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3851 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3852 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3853 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3854 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3855 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3856 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3857 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3858 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3859 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3860 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3862 return (features & feature) == feature;
3865 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3867 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3868 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3871 static inline bool netif_needs_gso(struct sk_buff *skb,
3872 netdev_features_t features)
3874 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3875 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3876 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3879 static inline void netif_set_gso_max_size(struct net_device *dev,
3882 dev->gso_max_size = size;
3885 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3886 int pulled_hlen, u16 mac_offset,
3889 skb->protocol = protocol;
3890 skb->encapsulation = 1;
3891 skb_push(skb, pulled_hlen);
3892 skb_reset_transport_header(skb);
3893 skb->mac_header = mac_offset;
3894 skb->network_header = skb->mac_header + mac_len;
3895 skb->mac_len = mac_len;
3898 static inline bool netif_is_macvlan(struct net_device *dev)
3900 return dev->priv_flags & IFF_MACVLAN;
3903 static inline bool netif_is_macvlan_port(struct net_device *dev)
3905 return dev->priv_flags & IFF_MACVLAN_PORT;
3908 static inline bool netif_is_ipvlan(struct net_device *dev)
3910 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3913 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3915 return dev->priv_flags & IFF_IPVLAN_MASTER;
3918 static inline bool netif_is_bond_master(struct net_device *dev)
3920 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3923 static inline bool netif_is_bond_slave(struct net_device *dev)
3925 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3928 static inline bool netif_supports_nofcs(struct net_device *dev)
3930 return dev->priv_flags & IFF_SUPP_NOFCS;
3933 static inline bool netif_is_l3_master(const struct net_device *dev)
3935 return dev->priv_flags & IFF_L3MDEV_MASTER;
3938 static inline bool netif_is_l3_slave(const struct net_device *dev)
3940 return dev->priv_flags & IFF_L3MDEV_SLAVE;
3943 static inline bool netif_is_bridge_master(const struct net_device *dev)
3945 return dev->priv_flags & IFF_EBRIDGE;
3948 static inline bool netif_is_bridge_port(const struct net_device *dev)
3950 return dev->priv_flags & IFF_BRIDGE_PORT;
3953 static inline bool netif_is_ovs_master(const struct net_device *dev)
3955 return dev->priv_flags & IFF_OPENVSWITCH;
3958 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3959 static inline void netif_keep_dst(struct net_device *dev)
3961 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3964 extern struct pernet_operations __net_initdata loopback_net_ops;
3966 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3968 /* netdev_printk helpers, similar to dev_printk */
3970 static inline const char *netdev_name(const struct net_device *dev)
3972 if (!dev->name[0] || strchr(dev->name, '%'))
3973 return "(unnamed net_device)";
3977 static inline const char *netdev_reg_state(const struct net_device *dev)
3979 switch (dev->reg_state) {
3980 case NETREG_UNINITIALIZED: return " (uninitialized)";
3981 case NETREG_REGISTERED: return "";
3982 case NETREG_UNREGISTERING: return " (unregistering)";
3983 case NETREG_UNREGISTERED: return " (unregistered)";
3984 case NETREG_RELEASED: return " (released)";
3985 case NETREG_DUMMY: return " (dummy)";
3988 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3989 return " (unknown)";
3993 void netdev_printk(const char *level, const struct net_device *dev,
3994 const char *format, ...);
3996 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3998 void netdev_alert(const struct net_device *dev, const char *format, ...);
4000 void netdev_crit(const struct net_device *dev, const char *format, ...);
4002 void netdev_err(const struct net_device *dev, const char *format, ...);
4004 void netdev_warn(const struct net_device *dev, const char *format, ...);
4006 void netdev_notice(const struct net_device *dev, const char *format, ...);
4008 void netdev_info(const struct net_device *dev, const char *format, ...);
4010 #define MODULE_ALIAS_NETDEV(device) \
4011 MODULE_ALIAS("netdev-" device)
4013 #if defined(CONFIG_DYNAMIC_DEBUG)
4014 #define netdev_dbg(__dev, format, args...) \
4016 dynamic_netdev_dbg(__dev, format, ##args); \
4018 #elif defined(DEBUG)
4019 #define netdev_dbg(__dev, format, args...) \
4020 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4022 #define netdev_dbg(__dev, format, args...) \
4025 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4029 #if defined(VERBOSE_DEBUG)
4030 #define netdev_vdbg netdev_dbg
4033 #define netdev_vdbg(dev, format, args...) \
4036 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4042 * netdev_WARN() acts like dev_printk(), but with the key difference
4043 * of using a WARN/WARN_ON to get the message out, including the
4044 * file/line information and a backtrace.
4046 #define netdev_WARN(dev, format, args...) \
4047 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4048 netdev_reg_state(dev), ##args)
4050 /* netif printk helpers, similar to netdev_printk */
4052 #define netif_printk(priv, type, level, dev, fmt, args...) \
4054 if (netif_msg_##type(priv)) \
4055 netdev_printk(level, (dev), fmt, ##args); \
4058 #define netif_level(level, priv, type, dev, fmt, args...) \
4060 if (netif_msg_##type(priv)) \
4061 netdev_##level(dev, fmt, ##args); \
4064 #define netif_emerg(priv, type, dev, fmt, args...) \
4065 netif_level(emerg, priv, type, dev, fmt, ##args)
4066 #define netif_alert(priv, type, dev, fmt, args...) \
4067 netif_level(alert, priv, type, dev, fmt, ##args)
4068 #define netif_crit(priv, type, dev, fmt, args...) \
4069 netif_level(crit, priv, type, dev, fmt, ##args)
4070 #define netif_err(priv, type, dev, fmt, args...) \
4071 netif_level(err, priv, type, dev, fmt, ##args)
4072 #define netif_warn(priv, type, dev, fmt, args...) \
4073 netif_level(warn, priv, type, dev, fmt, ##args)
4074 #define netif_notice(priv, type, dev, fmt, args...) \
4075 netif_level(notice, priv, type, dev, fmt, ##args)
4076 #define netif_info(priv, type, dev, fmt, args...) \
4077 netif_level(info, priv, type, dev, fmt, ##args)
4079 #if defined(CONFIG_DYNAMIC_DEBUG)
4080 #define netif_dbg(priv, type, netdev, format, args...) \
4082 if (netif_msg_##type(priv)) \
4083 dynamic_netdev_dbg(netdev, format, ##args); \
4085 #elif defined(DEBUG)
4086 #define netif_dbg(priv, type, dev, format, args...) \
4087 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4089 #define netif_dbg(priv, type, dev, format, args...) \
4092 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4097 #if defined(VERBOSE_DEBUG)
4098 #define netif_vdbg netif_dbg
4100 #define netif_vdbg(priv, type, dev, format, args...) \
4103 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4109 * The list of packet types we will receive (as opposed to discard)
4110 * and the routines to invoke.
4112 * Why 16. Because with 16 the only overlap we get on a hash of the
4113 * low nibble of the protocol value is RARP/SNAP/X.25.
4115 * NOTE: That is no longer true with the addition of VLAN tags. Not
4116 * sure which should go first, but I bet it won't make much
4117 * difference if we are running VLANs. The good news is that
4118 * this protocol won't be in the list unless compiled in, so
4119 * the average user (w/out VLANs) will not be adversely affected.
4135 #define PTYPE_HASH_SIZE (16)
4136 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4138 #endif /* _LINUX_NETDEVICE_H */