1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Definitions for the Interfaces handler.
9 * Version: @(#)dev.h 1.0.10 08/12/93
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
19 * Moved to /usr/include/linux for NET3
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
32 #include <linux/percpu.h>
33 #include <linux/rculist.h>
34 #include <linux/workqueue.h>
35 #include <linux/dynamic_queue_limits.h>
37 #include <linux/ethtool.h>
38 #include <net/net_namespace.h>
40 #include <net/dcbnl.h>
42 #include <net/netprio_cgroup.h>
45 #include <linux/netdev_features.h>
46 #include <linux/neighbour.h>
47 #include <uapi/linux/netdevice.h>
48 #include <uapi/linux/if_bonding.h>
49 #include <uapi/linux/pkt_cls.h>
50 #include <linux/hashtable.h>
56 struct macsec_context;
62 /* 802.15.4 specific */
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
70 void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
77 #define MAX_NEST_DEV 8
80 * Transmit return codes: transmit return codes originate from three different
83 * - qdisc return codes
84 * - driver transmit return codes
87 * Drivers are allowed to return any one of those in their hard_start_xmit()
88 * function. Real network devices commonly used with qdiscs should only return
89 * the driver transmit return codes though - when qdiscs are used, the actual
90 * transmission happens asynchronously, so the value is not propagated to
91 * higher layers. Virtual network devices transmit synchronously; in this case
92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
93 * others are propagated to higher layers.
96 /* qdisc ->enqueue() return codes. */
97 #define NET_XMIT_SUCCESS 0x00
98 #define NET_XMIT_DROP 0x01 /* skb dropped */
99 #define NET_XMIT_CN 0x02 /* congestion notification */
100 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
102 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
103 * indicates that the device will soon be dropping packets, or already drops
104 * some packets of the same priority; prompting us to send less aggressively. */
105 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
106 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
108 /* Driver transmit return codes */
109 #define NETDEV_TX_MASK 0xf0
112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
113 NETDEV_TX_OK = 0x00, /* driver took care of packet */
114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
116 typedef enum netdev_tx netdev_tx_t;
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
122 static inline bool dev_xmit_complete(int rc)
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
130 if (likely(rc < NET_XMIT_MASK))
137 * Compute the worst-case header length according to the protocols
141 #if defined(CONFIG_HYPERV_NET)
142 # define LL_MAX_HEADER 128
143 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
144 # if defined(CONFIG_MAC80211_MESH)
145 # define LL_MAX_HEADER 128
147 # define LL_MAX_HEADER 96
150 # define LL_MAX_HEADER 32
153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
155 #define MAX_HEADER LL_MAX_HEADER
157 #define MAX_HEADER (LL_MAX_HEADER + 48)
161 * Old network device statistics. Fields are native words
162 * (unsigned long) so they can be read and written atomically.
165 struct net_device_stats {
166 unsigned long rx_packets;
167 unsigned long tx_packets;
168 unsigned long rx_bytes;
169 unsigned long tx_bytes;
170 unsigned long rx_errors;
171 unsigned long tx_errors;
172 unsigned long rx_dropped;
173 unsigned long tx_dropped;
174 unsigned long multicast;
175 unsigned long collisions;
176 unsigned long rx_length_errors;
177 unsigned long rx_over_errors;
178 unsigned long rx_crc_errors;
179 unsigned long rx_frame_errors;
180 unsigned long rx_fifo_errors;
181 unsigned long rx_missed_errors;
182 unsigned long tx_aborted_errors;
183 unsigned long tx_carrier_errors;
184 unsigned long tx_fifo_errors;
185 unsigned long tx_heartbeat_errors;
186 unsigned long tx_window_errors;
187 unsigned long rx_compressed;
188 unsigned long tx_compressed;
192 #include <linux/cache.h>
193 #include <linux/skbuff.h>
196 #include <linux/static_key.h>
197 extern struct static_key_false rps_needed;
198 extern struct static_key_false rfs_needed;
205 struct netdev_hw_addr {
206 struct list_head list;
207 unsigned char addr[MAX_ADDR_LEN];
209 #define NETDEV_HW_ADDR_T_LAN 1
210 #define NETDEV_HW_ADDR_T_SAN 2
211 #define NETDEV_HW_ADDR_T_SLAVE 3
212 #define NETDEV_HW_ADDR_T_UNICAST 4
213 #define NETDEV_HW_ADDR_T_MULTICAST 5
218 struct rcu_head rcu_head;
221 struct netdev_hw_addr_list {
222 struct list_head list;
226 #define netdev_hw_addr_list_count(l) ((l)->count)
227 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
228 #define netdev_hw_addr_list_for_each(ha, l) \
229 list_for_each_entry(ha, &(l)->list, list)
231 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
232 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
233 #define netdev_for_each_uc_addr(ha, dev) \
234 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
236 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
237 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
238 #define netdev_for_each_mc_addr(ha, dev) \
239 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
245 /* cached hardware header; allow for machine alignment needs. */
246 #define HH_DATA_MOD 16
247 #define HH_DATA_OFF(__len) \
248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
249 #define HH_DATA_ALIGN(__len) \
250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
254 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
256 * dev->hard_header_len ? (dev->hard_header_len +
257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
259 * We could use other alignment values, but we must maintain the
260 * relationship HH alignment <= LL alignment.
262 #define LL_RESERVED_SPACE(dev) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
268 int (*create) (struct sk_buff *skb, struct net_device *dev,
269 unsigned short type, const void *daddr,
270 const void *saddr, unsigned int len);
271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
273 void (*cache_update)(struct hh_cache *hh,
274 const struct net_device *dev,
275 const unsigned char *haddr);
276 bool (*validate)(const char *ll_header, unsigned int len);
277 __be16 (*parse_protocol)(const struct sk_buff *skb);
280 /* These flag bits are private to the generic network queueing
281 * layer; they may not be explicitly referenced by any other
285 enum netdev_state_t {
287 __LINK_STATE_PRESENT,
288 __LINK_STATE_NOCARRIER,
289 __LINK_STATE_LINKWATCH_PENDING,
290 __LINK_STATE_DORMANT,
291 __LINK_STATE_TESTING,
296 * This structure holds boot-time configured netdevice settings. They
297 * are then used in the device probing.
299 struct netdev_boot_setup {
303 #define NETDEV_BOOT_SETUP_MAX 8
305 int __init netdev_boot_setup(char *str);
308 struct list_head list;
313 * size of gro hash buckets, must less than bit number of
314 * napi_struct::gro_bitmask
316 #define GRO_HASH_BUCKETS 8
319 * Structure for NAPI scheduling similar to tasklet but with weighting
322 /* The poll_list must only be managed by the entity which
323 * changes the state of the NAPI_STATE_SCHED bit. This means
324 * whoever atomically sets that bit can add this napi_struct
325 * to the per-CPU poll_list, and whoever clears that bit
326 * can remove from the list right before clearing the bit.
328 struct list_head poll_list;
332 int defer_hard_irqs_count;
333 unsigned long gro_bitmask;
334 int (*poll)(struct napi_struct *, int);
335 #ifdef CONFIG_NETPOLL
338 struct net_device *dev;
339 struct gro_list gro_hash[GRO_HASH_BUCKETS];
341 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
342 int rx_count; /* length of rx_list */
343 struct hrtimer timer;
344 struct list_head dev_list;
345 struct hlist_node napi_hash_node;
346 unsigned int napi_id;
350 NAPI_STATE_SCHED, /* Poll is scheduled */
351 NAPI_STATE_MISSED, /* reschedule a napi */
352 NAPI_STATE_DISABLE, /* Disable pending */
353 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
354 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
355 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
356 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
360 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
361 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
362 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
363 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
364 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
365 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
366 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
377 typedef enum gro_result gro_result_t;
380 * enum rx_handler_result - Possible return values for rx_handlers.
381 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
383 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
384 * case skb->dev was changed by rx_handler.
385 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
386 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
388 * rx_handlers are functions called from inside __netif_receive_skb(), to do
389 * special processing of the skb, prior to delivery to protocol handlers.
391 * Currently, a net_device can only have a single rx_handler registered. Trying
392 * to register a second rx_handler will return -EBUSY.
394 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
395 * To unregister a rx_handler on a net_device, use
396 * netdev_rx_handler_unregister().
398 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
401 * If the rx_handler consumed the skb in some way, it should return
402 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
403 * the skb to be delivered in some other way.
405 * If the rx_handler changed skb->dev, to divert the skb to another
406 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
407 * new device will be called if it exists.
409 * If the rx_handler decides the skb should be ignored, it should return
410 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
411 * are registered on exact device (ptype->dev == skb->dev).
413 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
414 * delivered, it should return RX_HANDLER_PASS.
416 * A device without a registered rx_handler will behave as if rx_handler
417 * returned RX_HANDLER_PASS.
420 enum rx_handler_result {
426 typedef enum rx_handler_result rx_handler_result_t;
427 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
429 void __napi_schedule(struct napi_struct *n);
430 void __napi_schedule_irqoff(struct napi_struct *n);
432 static inline bool napi_disable_pending(struct napi_struct *n)
434 return test_bit(NAPI_STATE_DISABLE, &n->state);
437 bool napi_schedule_prep(struct napi_struct *n);
440 * napi_schedule - schedule NAPI poll
443 * Schedule NAPI poll routine to be called if it is not already
446 static inline void napi_schedule(struct napi_struct *n)
448 if (napi_schedule_prep(n))
453 * napi_schedule_irqoff - schedule NAPI poll
456 * Variant of napi_schedule(), assuming hard irqs are masked.
458 static inline void napi_schedule_irqoff(struct napi_struct *n)
460 if (napi_schedule_prep(n))
461 __napi_schedule_irqoff(n);
464 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
465 static inline bool napi_reschedule(struct napi_struct *napi)
467 if (napi_schedule_prep(napi)) {
468 __napi_schedule(napi);
474 bool napi_complete_done(struct napi_struct *n, int work_done);
476 * napi_complete - NAPI processing complete
479 * Mark NAPI processing as complete.
480 * Consider using napi_complete_done() instead.
481 * Return false if device should avoid rearming interrupts.
483 static inline bool napi_complete(struct napi_struct *n)
485 return napi_complete_done(n, 0);
489 * napi_hash_del - remove a NAPI from global table
490 * @napi: NAPI context
492 * Warning: caller must observe RCU grace period
493 * before freeing memory containing @napi, if
494 * this function returns true.
495 * Note: core networking stack automatically calls it
496 * from netif_napi_del().
497 * Drivers might want to call this helper to combine all
498 * the needed RCU grace periods into a single one.
500 bool 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))
544 * napi_if_scheduled_mark_missed - if napi is running, set the
548 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
551 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
553 unsigned long val, new;
556 val = READ_ONCE(n->state);
557 if (val & NAPIF_STATE_DISABLE)
560 if (!(val & NAPIF_STATE_SCHED))
563 new = val | NAPIF_STATE_MISSED;
564 } while (cmpxchg(&n->state, val, new) != val);
569 enum netdev_queue_state_t {
570 __QUEUE_STATE_DRV_XOFF,
571 __QUEUE_STATE_STACK_XOFF,
572 __QUEUE_STATE_FROZEN,
575 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
576 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
577 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
579 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
580 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
582 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
586 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
587 * netif_tx_* functions below are used to manipulate this flag. The
588 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
589 * queue independently. The netif_xmit_*stopped functions below are called
590 * to check if the queue has been stopped by the driver or stack (either
591 * of the XOFF bits are set in the state). Drivers should not need to call
592 * netif_xmit*stopped functions, they should only be using netif_tx_*.
595 struct netdev_queue {
599 struct net_device *dev;
600 struct Qdisc __rcu *qdisc;
601 struct Qdisc *qdisc_sleeping;
605 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 unsigned long tx_maxrate;
610 * Number of TX timeouts for this queue
611 * (/sys/class/net/DEV/Q/trans_timeout)
613 unsigned long trans_timeout;
615 /* Subordinate device that the queue has been assigned to */
616 struct net_device *sb_dev;
617 #ifdef CONFIG_XDP_SOCKETS
618 struct xdp_umem *umem;
623 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
626 * Time (in jiffies) of last Tx
628 unsigned long trans_start;
635 } ____cacheline_aligned_in_smp;
637 extern int sysctl_fb_tunnels_only_for_init_net;
638 extern int sysctl_devconf_inherit_init_net;
640 static inline bool net_has_fallback_tunnels(const struct net *net)
642 return net == &init_net ||
643 !IS_ENABLED(CONFIG_SYSCTL) ||
644 !sysctl_fb_tunnels_only_for_init_net;
647 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
649 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
656 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
658 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
665 * This structure holds an RPS map which can be of variable length. The
666 * map is an array of CPUs.
673 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
676 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
677 * tail pointer for that CPU's input queue at the time of last enqueue, and
678 * a hardware filter index.
680 struct rps_dev_flow {
683 unsigned int last_qtail;
685 #define RPS_NO_FILTER 0xffff
688 * The rps_dev_flow_table structure contains a table of flow mappings.
690 struct rps_dev_flow_table {
693 struct rps_dev_flow flows[];
695 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
696 ((_num) * sizeof(struct rps_dev_flow)))
699 * The rps_sock_flow_table contains mappings of flows to the last CPU
700 * on which they were processed by the application (set in recvmsg).
701 * Each entry is a 32bit value. Upper part is the high-order bits
702 * of flow hash, lower part is CPU number.
703 * rps_cpu_mask is used to partition the space, depending on number of
704 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
705 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
706 * meaning we use 32-6=26 bits for the hash.
708 struct rps_sock_flow_table {
711 u32 ents[] ____cacheline_aligned_in_smp;
713 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
715 #define RPS_NO_CPU 0xffff
717 extern u32 rps_cpu_mask;
718 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
720 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
724 unsigned int index = hash & table->mask;
725 u32 val = hash & ~rps_cpu_mask;
727 /* We only give a hint, preemption can change CPU under us */
728 val |= raw_smp_processor_id();
730 if (table->ents[index] != val)
731 table->ents[index] = val;
735 #ifdef CONFIG_RFS_ACCEL
736 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
739 #endif /* CONFIG_RPS */
741 /* This structure contains an instance of an RX queue. */
742 struct netdev_rx_queue {
744 struct rps_map __rcu *rps_map;
745 struct rps_dev_flow_table __rcu *rps_flow_table;
748 struct net_device *dev;
749 struct xdp_rxq_info xdp_rxq;
750 #ifdef CONFIG_XDP_SOCKETS
751 struct xdp_umem *umem;
753 } ____cacheline_aligned_in_smp;
756 * RX queue sysfs structures and functions.
758 struct rx_queue_attribute {
759 struct attribute attr;
760 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
761 ssize_t (*store)(struct netdev_rx_queue *queue,
762 const char *buf, size_t len);
767 * This structure holds an XPS map which can be of variable length. The
768 * map is an array of queues.
772 unsigned int alloc_len;
776 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
777 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
778 - sizeof(struct xps_map)) / sizeof(u16))
781 * This structure holds all XPS maps for device. Maps are indexed by CPU.
783 struct xps_dev_maps {
785 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
788 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
789 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
791 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
792 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
794 #endif /* CONFIG_XPS */
796 #define TC_MAX_QUEUE 16
797 #define TC_BITMASK 15
798 /* HW offloaded queuing disciplines txq count and offset maps */
799 struct netdev_tc_txq {
804 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
806 * This structure is to hold information about the device
807 * configured to run FCoE protocol stack.
809 struct netdev_fcoe_hbainfo {
810 char manufacturer[64];
811 char serial_number[64];
812 char hardware_version[64];
813 char driver_version[64];
814 char optionrom_version[64];
815 char firmware_version[64];
817 char model_description[256];
821 #define MAX_PHYS_ITEM_ID_LEN 32
823 /* This structure holds a unique identifier to identify some
824 * physical item (port for example) used by a netdevice.
826 struct netdev_phys_item_id {
827 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
828 unsigned char id_len;
831 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
832 struct netdev_phys_item_id *b)
834 return a->id_len == b->id_len &&
835 memcmp(a->id, b->id, a->id_len) == 0;
838 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
840 struct net_device *sb_dev);
843 TC_SETUP_QDISC_MQPRIO,
846 TC_SETUP_CLSMATCHALL,
856 TC_SETUP_QDISC_TAPRIO,
863 /* These structures hold the attributes of bpf state that are being passed
864 * to the netdevice through the bpf op.
866 enum bpf_netdev_command {
867 /* Set or clear a bpf program used in the earliest stages of packet
868 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
869 * is responsible for calling bpf_prog_put on any old progs that are
870 * stored. In case of error, the callee need not release the new prog
871 * reference, but on success it takes ownership and must bpf_prog_put
872 * when it is no longer used.
878 /* BPF program for offload callbacks, invoked at program load time. */
879 BPF_OFFLOAD_MAP_ALLOC,
880 BPF_OFFLOAD_MAP_FREE,
884 struct bpf_prog_offload_ops;
885 struct netlink_ext_ack;
887 struct xdp_dev_bulk_queue;
890 enum bpf_netdev_command command;
895 struct bpf_prog *prog;
896 struct netlink_ext_ack *extack;
898 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
901 /* flags with which program was installed */
904 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
906 struct bpf_offloaded_map *offmap;
908 /* XDP_SETUP_XSK_UMEM */
910 struct xdp_umem *umem;
916 /* Flags for ndo_xsk_wakeup. */
917 #define XDP_WAKEUP_RX (1 << 0)
918 #define XDP_WAKEUP_TX (1 << 1)
920 #ifdef CONFIG_XFRM_OFFLOAD
922 int (*xdo_dev_state_add) (struct xfrm_state *x);
923 void (*xdo_dev_state_delete) (struct xfrm_state *x);
924 void (*xdo_dev_state_free) (struct xfrm_state *x);
925 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
926 struct xfrm_state *x);
927 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
932 struct rcu_head rcuhead;
939 struct netdev_name_node {
940 struct hlist_node hlist;
941 struct list_head list;
942 struct net_device *dev;
946 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
947 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
949 struct netdev_net_notifier {
950 struct list_head list;
951 struct notifier_block *nb;
955 * This structure defines the management hooks for network devices.
956 * The following hooks can be defined; unless noted otherwise, they are
957 * optional and can be filled with a null pointer.
959 * int (*ndo_init)(struct net_device *dev);
960 * This function is called once when a network device is registered.
961 * The network device can use this for any late stage initialization
962 * or semantic validation. It can fail with an error code which will
963 * be propagated back to register_netdev.
965 * void (*ndo_uninit)(struct net_device *dev);
966 * This function is called when device is unregistered or when registration
967 * fails. It is not called if init fails.
969 * int (*ndo_open)(struct net_device *dev);
970 * This function is called when a network device transitions to the up
973 * int (*ndo_stop)(struct net_device *dev);
974 * This function is called when a network device transitions to the down
977 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
978 * struct net_device *dev);
979 * Called when a packet needs to be transmitted.
980 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
981 * the queue before that can happen; it's for obsolete devices and weird
982 * corner cases, but the stack really does a non-trivial amount
983 * of useless work if you return NETDEV_TX_BUSY.
984 * Required; cannot be NULL.
986 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
987 * struct net_device *dev
988 * netdev_features_t features);
989 * Called by core transmit path to determine if device is capable of
990 * performing offload operations on a given packet. This is to give
991 * the device an opportunity to implement any restrictions that cannot
992 * be otherwise expressed by feature flags. The check is called with
993 * the set of features that the stack has calculated and it returns
994 * those the driver believes to be appropriate.
996 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
997 * struct net_device *sb_dev);
998 * Called to decide which queue to use when device supports multiple
1001 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1002 * This function is called to allow device receiver to make
1003 * changes to configuration when multicast or promiscuous is enabled.
1005 * void (*ndo_set_rx_mode)(struct net_device *dev);
1006 * This function is called device changes address list filtering.
1007 * If driver handles unicast address filtering, it should set
1008 * IFF_UNICAST_FLT in its priv_flags.
1010 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1011 * This function is called when the Media Access Control address
1012 * needs to be changed. If this interface is not defined, the
1013 * MAC address can not be changed.
1015 * int (*ndo_validate_addr)(struct net_device *dev);
1016 * Test if Media Access Control address is valid for the device.
1018 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1019 * Called when a user requests an ioctl which can't be handled by
1020 * the generic interface code. If not defined ioctls return
1021 * not supported error code.
1023 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1024 * Used to set network devices bus interface parameters. This interface
1025 * is retained for legacy reasons; new devices should use the bus
1026 * interface (PCI) for low level management.
1028 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1029 * Called when a user wants to change the Maximum Transfer Unit
1032 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1033 * Callback used when the transmitter has not made any progress
1034 * for dev->watchdog ticks.
1036 * void (*ndo_get_stats64)(struct net_device *dev,
1037 * struct rtnl_link_stats64 *storage);
1038 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1039 * Called when a user wants to get the network device usage
1040 * statistics. Drivers must do one of the following:
1041 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1042 * rtnl_link_stats64 structure passed by the caller.
1043 * 2. Define @ndo_get_stats to update a net_device_stats structure
1044 * (which should normally be dev->stats) and return a pointer to
1045 * it. The structure may be changed asynchronously only if each
1046 * field is written atomically.
1047 * 3. Update dev->stats asynchronously and atomically, and define
1048 * neither operation.
1050 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1051 * Return true if this device supports offload stats of this attr_id.
1053 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1055 * Get statistics for offload operations by attr_id. Write it into the
1056 * attr_data pointer.
1058 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1059 * If device supports VLAN filtering this function is called when a
1060 * VLAN id is registered.
1062 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1063 * If device supports VLAN filtering this function is called when a
1064 * VLAN id is unregistered.
1066 * void (*ndo_poll_controller)(struct net_device *dev);
1068 * SR-IOV management functions.
1069 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1070 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1071 * u8 qos, __be16 proto);
1072 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1074 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1075 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1076 * int (*ndo_get_vf_config)(struct net_device *dev,
1077 * int vf, struct ifla_vf_info *ivf);
1078 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1079 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1080 * struct nlattr *port[]);
1082 * Enable or disable the VF ability to query its RSS Redirection Table and
1083 * Hash Key. This is needed since on some devices VF share this information
1084 * with PF and querying it may introduce a theoretical security risk.
1085 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1086 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1087 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1089 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1090 * This is always called from the stack with the rtnl lock held and netif
1091 * tx queues stopped. This allows the netdevice to perform queue
1092 * management safely.
1094 * Fiber Channel over Ethernet (FCoE) offload functions.
1095 * int (*ndo_fcoe_enable)(struct net_device *dev);
1096 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1097 * so the underlying device can perform whatever needed configuration or
1098 * initialization to support acceleration of FCoE traffic.
1100 * int (*ndo_fcoe_disable)(struct net_device *dev);
1101 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1102 * so the underlying device can perform whatever needed clean-ups to
1103 * stop supporting acceleration of FCoE traffic.
1105 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1106 * struct scatterlist *sgl, unsigned int sgc);
1107 * Called when the FCoE Initiator wants to initialize an I/O that
1108 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1109 * perform necessary setup and returns 1 to indicate the device is set up
1110 * successfully to perform DDP on this I/O, otherwise this returns 0.
1112 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1113 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1114 * indicated by the FC exchange id 'xid', so the underlying device can
1115 * clean up and reuse resources for later DDP requests.
1117 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1118 * struct scatterlist *sgl, unsigned int sgc);
1119 * Called when the FCoE Target wants to initialize an I/O that
1120 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1121 * perform necessary setup and returns 1 to indicate the device is set up
1122 * successfully to perform DDP on this I/O, otherwise this returns 0.
1124 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1125 * struct netdev_fcoe_hbainfo *hbainfo);
1126 * Called when the FCoE Protocol stack wants information on the underlying
1127 * device. This information is utilized by the FCoE protocol stack to
1128 * register attributes with Fiber Channel management service as per the
1129 * FC-GS Fabric Device Management Information(FDMI) specification.
1131 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1132 * Called when the underlying device wants to override default World Wide
1133 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1134 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1135 * protocol stack to use.
1138 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1139 * u16 rxq_index, u32 flow_id);
1140 * Set hardware filter for RFS. rxq_index is the target queue index;
1141 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1142 * Return the filter ID on success, or a negative error code.
1144 * Slave management functions (for bridge, bonding, etc).
1145 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1146 * Called to make another netdev an underling.
1148 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1149 * Called to release previously enslaved netdev.
1151 * Feature/offload setting functions.
1152 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1153 * netdev_features_t features);
1154 * Adjusts the requested feature flags according to device-specific
1155 * constraints, and returns the resulting flags. Must not modify
1158 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1159 * Called to update device configuration to new features. Passed
1160 * feature set might be less than what was returned by ndo_fix_features()).
1161 * Must return >0 or -errno if it changed dev->features itself.
1163 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1164 * struct net_device *dev,
1165 * const unsigned char *addr, u16 vid, u16 flags,
1166 * struct netlink_ext_ack *extack);
1167 * Adds an FDB entry to dev for addr.
1168 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1169 * struct net_device *dev,
1170 * const unsigned char *addr, u16 vid)
1171 * Deletes the FDB entry from dev coresponding to addr.
1172 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1173 * struct net_device *dev, struct net_device *filter_dev,
1175 * Used to add FDB entries to dump requests. Implementers should add
1176 * entries to skb and update idx with the number of entries.
1178 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1179 * u16 flags, struct netlink_ext_ack *extack)
1180 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1181 * struct net_device *dev, u32 filter_mask,
1183 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1186 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1187 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1188 * which do not represent real hardware may define this to allow their
1189 * userspace components to manage their virtual carrier state. Devices
1190 * that determine carrier state from physical hardware properties (eg
1191 * network cables) or protocol-dependent mechanisms (eg
1192 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1194 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1195 * struct netdev_phys_item_id *ppid);
1196 * Called to get ID of physical port of this device. If driver does
1197 * not implement this, it is assumed that the hw is not able to have
1198 * multiple net devices on single physical port.
1200 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1201 * struct netdev_phys_item_id *ppid)
1202 * Called to get the parent ID of the physical port of this device.
1204 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1205 * struct udp_tunnel_info *ti);
1206 * Called by UDP tunnel to notify a driver about the UDP port and socket
1207 * address family that a UDP tunnel is listnening to. It is called only
1208 * when a new port starts listening. The operation is protected by the
1211 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1212 * struct udp_tunnel_info *ti);
1213 * Called by UDP tunnel to notify the driver about a UDP port and socket
1214 * address family that the UDP tunnel is not listening to anymore. The
1215 * operation is protected by the RTNL.
1217 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1218 * struct net_device *dev)
1219 * Called by upper layer devices to accelerate switching or other
1220 * station functionality into hardware. 'pdev is the lowerdev
1221 * to use for the offload and 'dev' is the net device that will
1222 * back the offload. Returns a pointer to the private structure
1223 * the upper layer will maintain.
1224 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1225 * Called by upper layer device to delete the station created
1226 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1227 * the station and priv is the structure returned by the add
1229 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1230 * int queue_index, u32 maxrate);
1231 * Called when a user wants to set a max-rate limitation of specific
1233 * int (*ndo_get_iflink)(const struct net_device *dev);
1234 * Called to get the iflink value of this device.
1235 * void (*ndo_change_proto_down)(struct net_device *dev,
1237 * This function is used to pass protocol port error state information
1238 * to the switch driver. The switch driver can react to the proto_down
1239 * by doing a phys down on the associated switch port.
1240 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1241 * This function is used to get egress tunnel information for given skb.
1242 * This is useful for retrieving outer tunnel header parameters while
1244 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1245 * This function is used to specify the headroom that the skb must
1246 * consider when allocation skb during packet reception. Setting
1247 * appropriate rx headroom value allows avoiding skb head copy on
1248 * forward. Setting a negative value resets the rx headroom to the
1250 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1251 * This function is used to set or query state related to XDP on the
1252 * netdevice and manage BPF offload. See definition of
1253 * enum bpf_netdev_command for details.
1254 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1256 * This function is used to submit @n XDP packets for transmit on a
1257 * netdevice. Returns number of frames successfully transmitted, frames
1258 * that got dropped are freed/returned via xdp_return_frame().
1259 * Returns negative number, means general error invoking ndo, meaning
1260 * no frames were xmit'ed and core-caller will free all frames.
1261 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1262 * This function is used to wake up the softirq, ksoftirqd or kthread
1263 * responsible for sending and/or receiving packets on a specific
1264 * queue id bound to an AF_XDP socket. The flags field specifies if
1265 * only RX, only Tx, or both should be woken up using the flags
1266 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1267 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1268 * Get devlink port instance associated with a given netdev.
1269 * Called with a reference on the netdevice and devlink locks only,
1270 * rtnl_lock is not held.
1272 struct net_device_ops {
1273 int (*ndo_init)(struct net_device *dev);
1274 void (*ndo_uninit)(struct net_device *dev);
1275 int (*ndo_open)(struct net_device *dev);
1276 int (*ndo_stop)(struct net_device *dev);
1277 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1278 struct net_device *dev);
1279 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1280 struct net_device *dev,
1281 netdev_features_t features);
1282 u16 (*ndo_select_queue)(struct net_device *dev,
1283 struct sk_buff *skb,
1284 struct net_device *sb_dev);
1285 void (*ndo_change_rx_flags)(struct net_device *dev,
1287 void (*ndo_set_rx_mode)(struct net_device *dev);
1288 int (*ndo_set_mac_address)(struct net_device *dev,
1290 int (*ndo_validate_addr)(struct net_device *dev);
1291 int (*ndo_do_ioctl)(struct net_device *dev,
1292 struct ifreq *ifr, int cmd);
1293 int (*ndo_set_config)(struct net_device *dev,
1295 int (*ndo_change_mtu)(struct net_device *dev,
1297 int (*ndo_neigh_setup)(struct net_device *dev,
1298 struct neigh_parms *);
1299 void (*ndo_tx_timeout) (struct net_device *dev,
1300 unsigned int txqueue);
1302 void (*ndo_get_stats64)(struct net_device *dev,
1303 struct rtnl_link_stats64 *storage);
1304 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1305 int (*ndo_get_offload_stats)(int attr_id,
1306 const struct net_device *dev,
1308 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1310 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1311 __be16 proto, u16 vid);
1312 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1313 __be16 proto, u16 vid);
1314 #ifdef CONFIG_NET_POLL_CONTROLLER
1315 void (*ndo_poll_controller)(struct net_device *dev);
1316 int (*ndo_netpoll_setup)(struct net_device *dev,
1317 struct netpoll_info *info);
1318 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1320 int (*ndo_set_vf_mac)(struct net_device *dev,
1321 int queue, u8 *mac);
1322 int (*ndo_set_vf_vlan)(struct net_device *dev,
1323 int queue, u16 vlan,
1324 u8 qos, __be16 proto);
1325 int (*ndo_set_vf_rate)(struct net_device *dev,
1326 int vf, int min_tx_rate,
1328 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1329 int vf, bool setting);
1330 int (*ndo_set_vf_trust)(struct net_device *dev,
1331 int vf, bool setting);
1332 int (*ndo_get_vf_config)(struct net_device *dev,
1334 struct ifla_vf_info *ivf);
1335 int (*ndo_set_vf_link_state)(struct net_device *dev,
1336 int vf, int link_state);
1337 int (*ndo_get_vf_stats)(struct net_device *dev,
1339 struct ifla_vf_stats
1341 int (*ndo_set_vf_port)(struct net_device *dev,
1343 struct nlattr *port[]);
1344 int (*ndo_get_vf_port)(struct net_device *dev,
1345 int vf, struct sk_buff *skb);
1346 int (*ndo_get_vf_guid)(struct net_device *dev,
1348 struct ifla_vf_guid *node_guid,
1349 struct ifla_vf_guid *port_guid);
1350 int (*ndo_set_vf_guid)(struct net_device *dev,
1353 int (*ndo_set_vf_rss_query_en)(
1354 struct net_device *dev,
1355 int vf, bool setting);
1356 int (*ndo_setup_tc)(struct net_device *dev,
1357 enum tc_setup_type type,
1359 #if IS_ENABLED(CONFIG_FCOE)
1360 int (*ndo_fcoe_enable)(struct net_device *dev);
1361 int (*ndo_fcoe_disable)(struct net_device *dev);
1362 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1364 struct scatterlist *sgl,
1366 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1368 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1370 struct scatterlist *sgl,
1372 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1373 struct netdev_fcoe_hbainfo *hbainfo);
1376 #if IS_ENABLED(CONFIG_LIBFCOE)
1377 #define NETDEV_FCOE_WWNN 0
1378 #define NETDEV_FCOE_WWPN 1
1379 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1380 u64 *wwn, int type);
1383 #ifdef CONFIG_RFS_ACCEL
1384 int (*ndo_rx_flow_steer)(struct net_device *dev,
1385 const struct sk_buff *skb,
1389 int (*ndo_add_slave)(struct net_device *dev,
1390 struct net_device *slave_dev,
1391 struct netlink_ext_ack *extack);
1392 int (*ndo_del_slave)(struct net_device *dev,
1393 struct net_device *slave_dev);
1394 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1395 netdev_features_t features);
1396 int (*ndo_set_features)(struct net_device *dev,
1397 netdev_features_t features);
1398 int (*ndo_neigh_construct)(struct net_device *dev,
1399 struct neighbour *n);
1400 void (*ndo_neigh_destroy)(struct net_device *dev,
1401 struct neighbour *n);
1403 int (*ndo_fdb_add)(struct ndmsg *ndm,
1404 struct nlattr *tb[],
1405 struct net_device *dev,
1406 const unsigned char *addr,
1409 struct netlink_ext_ack *extack);
1410 int (*ndo_fdb_del)(struct ndmsg *ndm,
1411 struct nlattr *tb[],
1412 struct net_device *dev,
1413 const unsigned char *addr,
1415 int (*ndo_fdb_dump)(struct sk_buff *skb,
1416 struct netlink_callback *cb,
1417 struct net_device *dev,
1418 struct net_device *filter_dev,
1420 int (*ndo_fdb_get)(struct sk_buff *skb,
1421 struct nlattr *tb[],
1422 struct net_device *dev,
1423 const unsigned char *addr,
1424 u16 vid, u32 portid, u32 seq,
1425 struct netlink_ext_ack *extack);
1426 int (*ndo_bridge_setlink)(struct net_device *dev,
1427 struct nlmsghdr *nlh,
1429 struct netlink_ext_ack *extack);
1430 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1432 struct net_device *dev,
1435 int (*ndo_bridge_dellink)(struct net_device *dev,
1436 struct nlmsghdr *nlh,
1438 int (*ndo_change_carrier)(struct net_device *dev,
1440 int (*ndo_get_phys_port_id)(struct net_device *dev,
1441 struct netdev_phys_item_id *ppid);
1442 int (*ndo_get_port_parent_id)(struct net_device *dev,
1443 struct netdev_phys_item_id *ppid);
1444 int (*ndo_get_phys_port_name)(struct net_device *dev,
1445 char *name, size_t len);
1446 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1447 struct udp_tunnel_info *ti);
1448 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1449 struct udp_tunnel_info *ti);
1450 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1451 struct net_device *dev);
1452 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1455 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1458 int (*ndo_get_iflink)(const struct net_device *dev);
1459 int (*ndo_change_proto_down)(struct net_device *dev,
1461 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1462 struct sk_buff *skb);
1463 void (*ndo_set_rx_headroom)(struct net_device *dev,
1464 int needed_headroom);
1465 int (*ndo_bpf)(struct net_device *dev,
1466 struct netdev_bpf *bpf);
1467 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1468 struct xdp_frame **xdp,
1470 int (*ndo_xsk_wakeup)(struct net_device *dev,
1471 u32 queue_id, u32 flags);
1472 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1476 * enum net_device_priv_flags - &struct net_device priv_flags
1478 * These are the &struct net_device, they are only set internally
1479 * by drivers and used in the kernel. These flags are invisible to
1480 * userspace; this means that the order of these flags can change
1481 * during any kernel release.
1483 * You should have a pretty good reason to be extending these flags.
1485 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1486 * @IFF_EBRIDGE: Ethernet bridging device
1487 * @IFF_BONDING: bonding master or slave
1488 * @IFF_ISATAP: ISATAP interface (RFC4214)
1489 * @IFF_WAN_HDLC: WAN HDLC device
1490 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1492 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1493 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1494 * @IFF_MACVLAN_PORT: device used as macvlan port
1495 * @IFF_BRIDGE_PORT: device used as bridge port
1496 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1497 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1498 * @IFF_UNICAST_FLT: Supports unicast filtering
1499 * @IFF_TEAM_PORT: device used as team port
1500 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1501 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1502 * change when it's running
1503 * @IFF_MACVLAN: Macvlan device
1504 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1505 * underlying stacked devices
1506 * @IFF_L3MDEV_MASTER: device is an L3 master device
1507 * @IFF_NO_QUEUE: device can run without qdisc attached
1508 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1509 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1510 * @IFF_TEAM: device is a team device
1511 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1512 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1513 * entity (i.e. the master device for bridged veth)
1514 * @IFF_MACSEC: device is a MACsec device
1515 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1516 * @IFF_FAILOVER: device is a failover master device
1517 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1518 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1519 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1521 enum netdev_priv_flags {
1522 IFF_802_1Q_VLAN = 1<<0,
1526 IFF_WAN_HDLC = 1<<4,
1527 IFF_XMIT_DST_RELEASE = 1<<5,
1528 IFF_DONT_BRIDGE = 1<<6,
1529 IFF_DISABLE_NETPOLL = 1<<7,
1530 IFF_MACVLAN_PORT = 1<<8,
1531 IFF_BRIDGE_PORT = 1<<9,
1532 IFF_OVS_DATAPATH = 1<<10,
1533 IFF_TX_SKB_SHARING = 1<<11,
1534 IFF_UNICAST_FLT = 1<<12,
1535 IFF_TEAM_PORT = 1<<13,
1536 IFF_SUPP_NOFCS = 1<<14,
1537 IFF_LIVE_ADDR_CHANGE = 1<<15,
1538 IFF_MACVLAN = 1<<16,
1539 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1540 IFF_L3MDEV_MASTER = 1<<18,
1541 IFF_NO_QUEUE = 1<<19,
1542 IFF_OPENVSWITCH = 1<<20,
1543 IFF_L3MDEV_SLAVE = 1<<21,
1545 IFF_RXFH_CONFIGURED = 1<<23,
1546 IFF_PHONY_HEADROOM = 1<<24,
1548 IFF_NO_RX_HANDLER = 1<<26,
1549 IFF_FAILOVER = 1<<27,
1550 IFF_FAILOVER_SLAVE = 1<<28,
1551 IFF_L3MDEV_RX_HANDLER = 1<<29,
1552 IFF_LIVE_RENAME_OK = 1<<30,
1555 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1556 #define IFF_EBRIDGE IFF_EBRIDGE
1557 #define IFF_BONDING IFF_BONDING
1558 #define IFF_ISATAP IFF_ISATAP
1559 #define IFF_WAN_HDLC IFF_WAN_HDLC
1560 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1561 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1562 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1563 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1564 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1565 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1566 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1567 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1568 #define IFF_TEAM_PORT IFF_TEAM_PORT
1569 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1570 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1571 #define IFF_MACVLAN IFF_MACVLAN
1572 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1573 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1574 #define IFF_NO_QUEUE IFF_NO_QUEUE
1575 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1576 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1577 #define IFF_TEAM IFF_TEAM
1578 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1579 #define IFF_MACSEC IFF_MACSEC
1580 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1581 #define IFF_FAILOVER IFF_FAILOVER
1582 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1583 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1584 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1587 * struct net_device - The DEVICE structure.
1589 * Actually, this whole structure is a big mistake. It mixes I/O
1590 * data with strictly "high-level" data, and it has to know about
1591 * almost every data structure used in the INET module.
1593 * @name: This is the first field of the "visible" part of this structure
1594 * (i.e. as seen by users in the "Space.c" file). It is the name
1597 * @name_node: Name hashlist node
1598 * @ifalias: SNMP alias
1599 * @mem_end: Shared memory end
1600 * @mem_start: Shared memory start
1601 * @base_addr: Device I/O address
1602 * @irq: Device IRQ number
1604 * @state: Generic network queuing layer state, see netdev_state_t
1605 * @dev_list: The global list of network devices
1606 * @napi_list: List entry used for polling NAPI devices
1607 * @unreg_list: List entry when we are unregistering the
1608 * device; see the function unregister_netdev
1609 * @close_list: List entry used when we are closing the device
1610 * @ptype_all: Device-specific packet handlers for all protocols
1611 * @ptype_specific: Device-specific, protocol-specific packet handlers
1613 * @adj_list: Directly linked devices, like slaves for bonding
1614 * @features: Currently active device features
1615 * @hw_features: User-changeable features
1617 * @wanted_features: User-requested features
1618 * @vlan_features: Mask of features inheritable by VLAN devices
1620 * @hw_enc_features: Mask of features inherited by encapsulating devices
1621 * This field indicates what encapsulation
1622 * offloads the hardware is capable of doing,
1623 * and drivers will need to set them appropriately.
1625 * @mpls_features: Mask of features inheritable by MPLS
1626 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1628 * @ifindex: interface index
1629 * @group: The group the device belongs to
1631 * @stats: Statistics struct, which was left as a legacy, use
1632 * rtnl_link_stats64 instead
1634 * @rx_dropped: Dropped packets by core network,
1635 * do not use this in drivers
1636 * @tx_dropped: Dropped packets by core network,
1637 * do not use this in drivers
1638 * @rx_nohandler: nohandler dropped packets by core network on
1639 * inactive devices, do not use this in drivers
1640 * @carrier_up_count: Number of times the carrier has been up
1641 * @carrier_down_count: Number of times the carrier has been down
1643 * @wireless_handlers: List of functions to handle Wireless Extensions,
1645 * see <net/iw_handler.h> for details.
1646 * @wireless_data: Instance data managed by the core of wireless extensions
1648 * @netdev_ops: Includes several pointers to callbacks,
1649 * if one wants to override the ndo_*() functions
1650 * @ethtool_ops: Management operations
1651 * @l3mdev_ops: Layer 3 master device operations
1652 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1653 * discovery handling. Necessary for e.g. 6LoWPAN.
1654 * @xfrmdev_ops: Transformation offload operations
1655 * @tlsdev_ops: Transport Layer Security offload operations
1656 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1657 * of Layer 2 headers.
1659 * @flags: Interface flags (a la BSD)
1660 * @priv_flags: Like 'flags' but invisible to userspace,
1661 * see if.h for the definitions
1662 * @gflags: Global flags ( kept as legacy )
1663 * @padded: How much padding added by alloc_netdev()
1664 * @operstate: RFC2863 operstate
1665 * @link_mode: Mapping policy to operstate
1666 * @if_port: Selectable AUI, TP, ...
1668 * @mtu: Interface MTU value
1669 * @min_mtu: Interface Minimum MTU value
1670 * @max_mtu: Interface Maximum MTU value
1671 * @type: Interface hardware type
1672 * @hard_header_len: Maximum hardware header length.
1673 * @min_header_len: Minimum hardware header length
1675 * @needed_headroom: Extra headroom the hardware may need, but not in all
1676 * cases can this be guaranteed
1677 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1678 * cases can this be guaranteed. Some cases also use
1679 * LL_MAX_HEADER instead to allocate the skb
1681 * interface address info:
1683 * @perm_addr: Permanent hw address
1684 * @addr_assign_type: Hw address assignment type
1685 * @addr_len: Hardware address length
1686 * @upper_level: Maximum depth level of upper devices.
1687 * @lower_level: Maximum depth level of lower devices.
1688 * @neigh_priv_len: Used in neigh_alloc()
1689 * @dev_id: Used to differentiate devices that share
1690 * the same link layer address
1691 * @dev_port: Used to differentiate devices that share
1693 * @addr_list_lock: XXX: need comments on this one
1694 * @name_assign_type: network interface name assignment type
1695 * @uc_promisc: Counter that indicates promiscuous mode
1696 * has been enabled due to the need to listen to
1697 * additional unicast addresses in a device that
1698 * does not implement ndo_set_rx_mode()
1699 * @uc: unicast mac addresses
1700 * @mc: multicast mac addresses
1701 * @dev_addrs: list of device hw addresses
1702 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1703 * @promiscuity: Number of times the NIC is told to work in
1704 * promiscuous mode; if it becomes 0 the NIC will
1705 * exit promiscuous mode
1706 * @allmulti: Counter, enables or disables allmulticast mode
1708 * @vlan_info: VLAN info
1709 * @dsa_ptr: dsa specific data
1710 * @tipc_ptr: TIPC specific data
1711 * @atalk_ptr: AppleTalk link
1712 * @ip_ptr: IPv4 specific data
1713 * @dn_ptr: DECnet specific data
1714 * @ip6_ptr: IPv6 specific data
1715 * @ax25_ptr: AX.25 specific data
1716 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1717 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1719 * @mpls_ptr: mpls_dev struct pointer
1721 * @dev_addr: Hw address (before bcast,
1722 * because most packets are unicast)
1724 * @_rx: Array of RX queues
1725 * @num_rx_queues: Number of RX queues
1726 * allocated at register_netdev() time
1727 * @real_num_rx_queues: Number of RX queues currently active in device
1728 * @xdp_prog: XDP sockets filter program pointer
1729 * @gro_flush_timeout: timeout for GRO layer in NAPI
1731 * @rx_handler: handler for received packets
1732 * @rx_handler_data: XXX: need comments on this one
1733 * @miniq_ingress: ingress/clsact qdisc specific data for
1734 * ingress processing
1735 * @ingress_queue: XXX: need comments on this one
1736 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1737 * @broadcast: hw bcast address
1739 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1740 * indexed by RX queue number. Assigned by driver.
1741 * This must only be set if the ndo_rx_flow_steer
1742 * operation is defined
1743 * @index_hlist: Device index hash chain
1745 * @_tx: Array of TX queues
1746 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1747 * @real_num_tx_queues: Number of TX queues currently active in device
1748 * @qdisc: Root qdisc from userspace point of view
1749 * @tx_queue_len: Max frames per queue allowed
1750 * @tx_global_lock: XXX: need comments on this one
1751 * @xdp_bulkq: XDP device bulk queue
1752 * @xps_cpus_map: all CPUs map for XPS device
1753 * @xps_rxqs_map: all RXQs map for XPS device
1755 * @xps_maps: XXX: need comments on this one
1756 * @miniq_egress: clsact qdisc specific data for
1758 * @qdisc_hash: qdisc hash table
1759 * @watchdog_timeo: Represents the timeout that is used by
1760 * the watchdog (see dev_watchdog())
1761 * @watchdog_timer: List of timers
1763 * @pcpu_refcnt: Number of references to this device
1764 * @todo_list: Delayed register/unregister
1765 * @link_watch_list: XXX: need comments on this one
1767 * @reg_state: Register/unregister state machine
1768 * @dismantle: Device is going to be freed
1769 * @rtnl_link_state: This enum represents the phases of creating
1772 * @needs_free_netdev: Should unregister perform free_netdev?
1773 * @priv_destructor: Called from unregister
1774 * @npinfo: XXX: need comments on this one
1775 * @nd_net: Network namespace this network device is inside
1777 * @ml_priv: Mid-layer private
1778 * @lstats: Loopback statistics
1779 * @tstats: Tunnel statistics
1780 * @dstats: Dummy statistics
1781 * @vstats: Virtual ethernet statistics
1786 * @dev: Class/net/name entry
1787 * @sysfs_groups: Space for optional device, statistics and wireless
1790 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1791 * @rtnl_link_ops: Rtnl_link_ops
1793 * @gso_max_size: Maximum size of generic segmentation offload
1794 * @gso_max_segs: Maximum number of segments that can be passed to the
1797 * @dcbnl_ops: Data Center Bridging netlink ops
1798 * @num_tc: Number of traffic classes in the net device
1799 * @tc_to_txq: XXX: need comments on this one
1800 * @prio_tc_map: XXX: need comments on this one
1802 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1804 * @priomap: XXX: need comments on this one
1805 * @phydev: Physical device may attach itself
1806 * for hardware timestamping
1807 * @sfp_bus: attached &struct sfp_bus structure.
1808 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock
1810 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1811 * @qdisc_xmit_lock_key: lockdep class annotating
1812 * netdev_queue->_xmit_lock spinlock
1813 * @addr_list_lock_key: lockdep class annotating
1814 * net_device->addr_list_lock spinlock
1816 * @proto_down: protocol port state information can be sent to the
1817 * switch driver and used to set the phys state of the
1820 * @wol_enabled: Wake-on-LAN is enabled
1822 * @net_notifier_list: List of per-net netdev notifier block
1823 * that follow this device when it is moved
1824 * to another network namespace.
1826 * @macsec_ops: MACsec offloading ops
1828 * FIXME: cleanup struct net_device such that network protocol info
1833 char name[IFNAMSIZ];
1834 struct netdev_name_node *name_node;
1835 struct dev_ifalias __rcu *ifalias;
1837 * I/O specific fields
1838 * FIXME: Merge these and struct ifmap into one
1840 unsigned long mem_end;
1841 unsigned long mem_start;
1842 unsigned long base_addr;
1846 * Some hardware also needs these fields (state,dev_list,
1847 * napi_list,unreg_list,close_list) but they are not
1848 * part of the usual set specified in Space.c.
1851 unsigned long state;
1853 struct list_head dev_list;
1854 struct list_head napi_list;
1855 struct list_head unreg_list;
1856 struct list_head close_list;
1857 struct list_head ptype_all;
1858 struct list_head ptype_specific;
1861 struct list_head upper;
1862 struct list_head lower;
1865 netdev_features_t features;
1866 netdev_features_t hw_features;
1867 netdev_features_t wanted_features;
1868 netdev_features_t vlan_features;
1869 netdev_features_t hw_enc_features;
1870 netdev_features_t mpls_features;
1871 netdev_features_t gso_partial_features;
1876 struct net_device_stats stats;
1878 atomic_long_t rx_dropped;
1879 atomic_long_t tx_dropped;
1880 atomic_long_t rx_nohandler;
1882 /* Stats to monitor link on/off, flapping */
1883 atomic_t carrier_up_count;
1884 atomic_t carrier_down_count;
1886 #ifdef CONFIG_WIRELESS_EXT
1887 const struct iw_handler_def *wireless_handlers;
1888 struct iw_public_data *wireless_data;
1890 const struct net_device_ops *netdev_ops;
1891 const struct ethtool_ops *ethtool_ops;
1892 #ifdef CONFIG_NET_L3_MASTER_DEV
1893 const struct l3mdev_ops *l3mdev_ops;
1895 #if IS_ENABLED(CONFIG_IPV6)
1896 const struct ndisc_ops *ndisc_ops;
1899 #ifdef CONFIG_XFRM_OFFLOAD
1900 const struct xfrmdev_ops *xfrmdev_ops;
1903 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1904 const struct tlsdev_ops *tlsdev_ops;
1907 const struct header_ops *header_ops;
1910 unsigned int priv_flags;
1912 unsigned short gflags;
1913 unsigned short padded;
1915 unsigned char operstate;
1916 unsigned char link_mode;
1918 unsigned char if_port;
1921 /* Note : dev->mtu is often read without holding a lock.
1922 * Writers usually hold RTNL.
1923 * It is recommended to use READ_ONCE() to annotate the reads,
1924 * and to use WRITE_ONCE() to annotate the writes.
1927 unsigned int min_mtu;
1928 unsigned int max_mtu;
1929 unsigned short type;
1930 unsigned short hard_header_len;
1931 unsigned char min_header_len;
1933 unsigned short needed_headroom;
1934 unsigned short needed_tailroom;
1936 /* Interface address info. */
1937 unsigned char perm_addr[MAX_ADDR_LEN];
1938 unsigned char addr_assign_type;
1939 unsigned char addr_len;
1940 unsigned char upper_level;
1941 unsigned char lower_level;
1942 unsigned short neigh_priv_len;
1943 unsigned short dev_id;
1944 unsigned short dev_port;
1945 spinlock_t addr_list_lock;
1946 unsigned char name_assign_type;
1948 struct netdev_hw_addr_list uc;
1949 struct netdev_hw_addr_list mc;
1950 struct netdev_hw_addr_list dev_addrs;
1953 struct kset *queues_kset;
1955 unsigned int promiscuity;
1956 unsigned int allmulti;
1959 /* Protocol-specific pointers */
1961 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1962 struct vlan_info __rcu *vlan_info;
1964 #if IS_ENABLED(CONFIG_NET_DSA)
1965 struct dsa_port *dsa_ptr;
1967 #if IS_ENABLED(CONFIG_TIPC)
1968 struct tipc_bearer __rcu *tipc_ptr;
1970 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1973 struct in_device __rcu *ip_ptr;
1974 #if IS_ENABLED(CONFIG_DECNET)
1975 struct dn_dev __rcu *dn_ptr;
1977 struct inet6_dev __rcu *ip6_ptr;
1978 #if IS_ENABLED(CONFIG_AX25)
1981 struct wireless_dev *ieee80211_ptr;
1982 struct wpan_dev *ieee802154_ptr;
1983 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1984 struct mpls_dev __rcu *mpls_ptr;
1988 * Cache lines mostly used on receive path (including eth_type_trans())
1990 /* Interface address info used in eth_type_trans() */
1991 unsigned char *dev_addr;
1993 struct netdev_rx_queue *_rx;
1994 unsigned int num_rx_queues;
1995 unsigned int real_num_rx_queues;
1997 struct bpf_prog __rcu *xdp_prog;
1998 unsigned long gro_flush_timeout;
1999 int napi_defer_hard_irqs;
2000 rx_handler_func_t __rcu *rx_handler;
2001 void __rcu *rx_handler_data;
2003 #ifdef CONFIG_NET_CLS_ACT
2004 struct mini_Qdisc __rcu *miniq_ingress;
2006 struct netdev_queue __rcu *ingress_queue;
2007 #ifdef CONFIG_NETFILTER_INGRESS
2008 struct nf_hook_entries __rcu *nf_hooks_ingress;
2011 unsigned char broadcast[MAX_ADDR_LEN];
2012 #ifdef CONFIG_RFS_ACCEL
2013 struct cpu_rmap *rx_cpu_rmap;
2015 struct hlist_node index_hlist;
2018 * Cache lines mostly used on transmit path
2020 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2021 unsigned int num_tx_queues;
2022 unsigned int real_num_tx_queues;
2023 struct Qdisc *qdisc;
2024 unsigned int tx_queue_len;
2025 spinlock_t tx_global_lock;
2027 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2030 struct xps_dev_maps __rcu *xps_cpus_map;
2031 struct xps_dev_maps __rcu *xps_rxqs_map;
2033 #ifdef CONFIG_NET_CLS_ACT
2034 struct mini_Qdisc __rcu *miniq_egress;
2037 #ifdef CONFIG_NET_SCHED
2038 DECLARE_HASHTABLE (qdisc_hash, 4);
2040 /* These may be needed for future network-power-down code. */
2041 struct timer_list watchdog_timer;
2044 struct list_head todo_list;
2045 int __percpu *pcpu_refcnt;
2047 struct list_head link_watch_list;
2049 enum { NETREG_UNINITIALIZED=0,
2050 NETREG_REGISTERED, /* completed register_netdevice */
2051 NETREG_UNREGISTERING, /* called unregister_netdevice */
2052 NETREG_UNREGISTERED, /* completed unregister todo */
2053 NETREG_RELEASED, /* called free_netdev */
2054 NETREG_DUMMY, /* dummy device for NAPI poll */
2060 RTNL_LINK_INITIALIZED,
2061 RTNL_LINK_INITIALIZING,
2062 } rtnl_link_state:16;
2064 bool needs_free_netdev;
2065 void (*priv_destructor)(struct net_device *dev);
2067 #ifdef CONFIG_NETPOLL
2068 struct netpoll_info __rcu *npinfo;
2071 possible_net_t nd_net;
2073 /* mid-layer private */
2076 struct pcpu_lstats __percpu *lstats;
2077 struct pcpu_sw_netstats __percpu *tstats;
2078 struct pcpu_dstats __percpu *dstats;
2081 #if IS_ENABLED(CONFIG_GARP)
2082 struct garp_port __rcu *garp_port;
2084 #if IS_ENABLED(CONFIG_MRP)
2085 struct mrp_port __rcu *mrp_port;
2089 const struct attribute_group *sysfs_groups[4];
2090 const struct attribute_group *sysfs_rx_queue_group;
2092 const struct rtnl_link_ops *rtnl_link_ops;
2094 /* for setting kernel sock attribute on TCP connection setup */
2095 #define GSO_MAX_SIZE 65536
2096 unsigned int gso_max_size;
2097 #define GSO_MAX_SEGS 65535
2101 const struct dcbnl_rtnl_ops *dcbnl_ops;
2104 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2105 u8 prio_tc_map[TC_BITMASK + 1];
2107 #if IS_ENABLED(CONFIG_FCOE)
2108 unsigned int fcoe_ddp_xid;
2110 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2111 struct netprio_map __rcu *priomap;
2113 struct phy_device *phydev;
2114 struct sfp_bus *sfp_bus;
2115 struct lock_class_key qdisc_tx_busylock_key;
2116 struct lock_class_key qdisc_running_key;
2117 struct lock_class_key qdisc_xmit_lock_key;
2118 struct lock_class_key addr_list_lock_key;
2120 unsigned wol_enabled:1;
2122 struct list_head net_notifier_list;
2124 #if IS_ENABLED(CONFIG_MACSEC)
2125 /* MACsec management functions */
2126 const struct macsec_ops *macsec_ops;
2129 #define to_net_dev(d) container_of(d, struct net_device, dev)
2131 static inline bool netif_elide_gro(const struct net_device *dev)
2133 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2138 #define NETDEV_ALIGN 32
2141 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2143 return dev->prio_tc_map[prio & TC_BITMASK];
2147 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2149 if (tc >= dev->num_tc)
2152 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2156 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2157 void netdev_reset_tc(struct net_device *dev);
2158 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2159 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2162 int netdev_get_num_tc(struct net_device *dev)
2167 void netdev_unbind_sb_channel(struct net_device *dev,
2168 struct net_device *sb_dev);
2169 int netdev_bind_sb_channel_queue(struct net_device *dev,
2170 struct net_device *sb_dev,
2171 u8 tc, u16 count, u16 offset);
2172 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2173 static inline int netdev_get_sb_channel(struct net_device *dev)
2175 return max_t(int, -dev->num_tc, 0);
2179 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2182 return &dev->_tx[index];
2185 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2186 const struct sk_buff *skb)
2188 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2191 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2192 void (*f)(struct net_device *,
2193 struct netdev_queue *,
2199 for (i = 0; i < dev->num_tx_queues; i++)
2200 f(dev, &dev->_tx[i], arg);
2203 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2204 struct net_device *sb_dev);
2205 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2206 struct sk_buff *skb,
2207 struct net_device *sb_dev);
2209 /* returns the headroom that the master device needs to take in account
2210 * when forwarding to this dev
2212 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2214 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2217 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2219 if (dev->netdev_ops->ndo_set_rx_headroom)
2220 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2223 /* set the device rx headroom to the dev's default */
2224 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2226 netdev_set_rx_headroom(dev, -1);
2230 * Net namespace inlines
2233 struct net *dev_net(const struct net_device *dev)
2235 return read_pnet(&dev->nd_net);
2239 void dev_net_set(struct net_device *dev, struct net *net)
2241 write_pnet(&dev->nd_net, net);
2245 * netdev_priv - access network device private data
2246 * @dev: network device
2248 * Get network device private data
2250 static inline void *netdev_priv(const struct net_device *dev)
2252 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2255 /* Set the sysfs physical device reference for the network logical device
2256 * if set prior to registration will cause a symlink during initialization.
2258 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2260 /* Set the sysfs device type for the network logical device to allow
2261 * fine-grained identification of different network device types. For
2262 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2264 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2266 /* Default NAPI poll() weight
2267 * Device drivers are strongly advised to not use bigger value
2269 #define NAPI_POLL_WEIGHT 64
2272 * netif_napi_add - initialize a NAPI context
2273 * @dev: network device
2274 * @napi: NAPI context
2275 * @poll: polling function
2276 * @weight: default weight
2278 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2279 * *any* of the other NAPI-related functions.
2281 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2282 int (*poll)(struct napi_struct *, int), int weight);
2285 * netif_tx_napi_add - initialize a NAPI context
2286 * @dev: network device
2287 * @napi: NAPI context
2288 * @poll: polling function
2289 * @weight: default weight
2291 * This variant of netif_napi_add() should be used from drivers using NAPI
2292 * to exclusively poll a TX queue.
2293 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2295 static inline void netif_tx_napi_add(struct net_device *dev,
2296 struct napi_struct *napi,
2297 int (*poll)(struct napi_struct *, int),
2300 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2301 netif_napi_add(dev, napi, poll, weight);
2305 * netif_napi_del - remove a NAPI context
2306 * @napi: NAPI context
2308 * netif_napi_del() removes a NAPI context from the network device NAPI list
2310 void netif_napi_del(struct napi_struct *napi);
2312 struct napi_gro_cb {
2313 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2316 /* Length of frag0. */
2317 unsigned int frag0_len;
2319 /* This indicates where we are processing relative to skb->data. */
2322 /* This is non-zero if the packet cannot be merged with the new skb. */
2325 /* Save the IP ID here and check when we get to the transport layer */
2328 /* Number of segments aggregated. */
2331 /* Start offset for remote checksum offload */
2332 u16 gro_remcsum_start;
2334 /* jiffies when first packet was created/queued */
2337 /* Used in ipv6_gro_receive() and foo-over-udp */
2340 /* This is non-zero if the packet may be of the same flow. */
2343 /* Used in tunnel GRO receive */
2346 /* GRO checksum is valid */
2349 /* Number of checksums via CHECKSUM_UNNECESSARY */
2354 #define NAPI_GRO_FREE 1
2355 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2357 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2360 /* Used in GRE, set in fou/gue_gro_receive */
2363 /* Used to determine if flush_id can be ignored */
2366 /* Number of gro_receive callbacks this packet already went through */
2367 u8 recursion_counter:4;
2369 /* GRO is done by frag_list pointer chaining. */
2372 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2375 /* used in skb_gro_receive() slow path */
2376 struct sk_buff *last;
2379 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2381 #define GRO_RECURSION_LIMIT 15
2382 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2384 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2387 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2388 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2389 struct list_head *head,
2390 struct sk_buff *skb)
2392 if (unlikely(gro_recursion_inc_test(skb))) {
2393 NAPI_GRO_CB(skb)->flush |= 1;
2397 return cb(head, skb);
2400 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2402 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2404 struct list_head *head,
2405 struct sk_buff *skb)
2407 if (unlikely(gro_recursion_inc_test(skb))) {
2408 NAPI_GRO_CB(skb)->flush |= 1;
2412 return cb(sk, head, skb);
2415 struct packet_type {
2416 __be16 type; /* This is really htons(ether_type). */
2417 bool ignore_outgoing;
2418 struct net_device *dev; /* NULL is wildcarded here */
2419 int (*func) (struct sk_buff *,
2420 struct net_device *,
2421 struct packet_type *,
2422 struct net_device *);
2423 void (*list_func) (struct list_head *,
2424 struct packet_type *,
2425 struct net_device *);
2426 bool (*id_match)(struct packet_type *ptype,
2428 void *af_packet_priv;
2429 struct list_head list;
2432 struct offload_callbacks {
2433 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2434 netdev_features_t features);
2435 struct sk_buff *(*gro_receive)(struct list_head *head,
2436 struct sk_buff *skb);
2437 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2440 struct packet_offload {
2441 __be16 type; /* This is really htons(ether_type). */
2443 struct offload_callbacks callbacks;
2444 struct list_head list;
2447 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2448 struct pcpu_sw_netstats {
2453 struct u64_stats_sync syncp;
2454 } __aligned(4 * sizeof(u64));
2456 struct pcpu_lstats {
2457 u64_stats_t packets;
2459 struct u64_stats_sync syncp;
2460 } __aligned(2 * sizeof(u64));
2462 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2464 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2466 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2468 u64_stats_update_begin(&lstats->syncp);
2469 u64_stats_add(&lstats->bytes, len);
2470 u64_stats_inc(&lstats->packets);
2471 u64_stats_update_end(&lstats->syncp);
2474 #define __netdev_alloc_pcpu_stats(type, gfp) \
2476 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2479 for_each_possible_cpu(__cpu) { \
2480 typeof(type) *stat; \
2481 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2482 u64_stats_init(&stat->syncp); \
2488 #define netdev_alloc_pcpu_stats(type) \
2489 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2491 enum netdev_lag_tx_type {
2492 NETDEV_LAG_TX_TYPE_UNKNOWN,
2493 NETDEV_LAG_TX_TYPE_RANDOM,
2494 NETDEV_LAG_TX_TYPE_BROADCAST,
2495 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2496 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2497 NETDEV_LAG_TX_TYPE_HASH,
2500 enum netdev_lag_hash {
2501 NETDEV_LAG_HASH_NONE,
2503 NETDEV_LAG_HASH_L34,
2504 NETDEV_LAG_HASH_L23,
2505 NETDEV_LAG_HASH_E23,
2506 NETDEV_LAG_HASH_E34,
2507 NETDEV_LAG_HASH_UNKNOWN,
2510 struct netdev_lag_upper_info {
2511 enum netdev_lag_tx_type tx_type;
2512 enum netdev_lag_hash hash_type;
2515 struct netdev_lag_lower_state_info {
2520 #include <linux/notifier.h>
2522 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2523 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2527 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2529 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2530 detected a hardware crash and restarted
2531 - we can use this eg to kick tcp sessions
2533 NETDEV_CHANGE, /* Notify device state change */
2536 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2537 NETDEV_CHANGEADDR, /* notify after the address change */
2538 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2542 NETDEV_BONDING_FAILOVER,
2544 NETDEV_PRE_TYPE_CHANGE,
2545 NETDEV_POST_TYPE_CHANGE,
2548 NETDEV_NOTIFY_PEERS,
2552 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2553 NETDEV_CHANGEINFODATA,
2554 NETDEV_BONDING_INFO,
2555 NETDEV_PRECHANGEUPPER,
2556 NETDEV_CHANGELOWERSTATE,
2557 NETDEV_UDP_TUNNEL_PUSH_INFO,
2558 NETDEV_UDP_TUNNEL_DROP_INFO,
2559 NETDEV_CHANGE_TX_QUEUE_LEN,
2560 NETDEV_CVLAN_FILTER_PUSH_INFO,
2561 NETDEV_CVLAN_FILTER_DROP_INFO,
2562 NETDEV_SVLAN_FILTER_PUSH_INFO,
2563 NETDEV_SVLAN_FILTER_DROP_INFO,
2565 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2567 int register_netdevice_notifier(struct notifier_block *nb);
2568 int unregister_netdevice_notifier(struct notifier_block *nb);
2569 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2570 int unregister_netdevice_notifier_net(struct net *net,
2571 struct notifier_block *nb);
2572 int register_netdevice_notifier_dev_net(struct net_device *dev,
2573 struct notifier_block *nb,
2574 struct netdev_net_notifier *nn);
2575 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2576 struct notifier_block *nb,
2577 struct netdev_net_notifier *nn);
2579 struct netdev_notifier_info {
2580 struct net_device *dev;
2581 struct netlink_ext_ack *extack;
2584 struct netdev_notifier_info_ext {
2585 struct netdev_notifier_info info; /* must be first */
2591 struct netdev_notifier_change_info {
2592 struct netdev_notifier_info info; /* must be first */
2593 unsigned int flags_changed;
2596 struct netdev_notifier_changeupper_info {
2597 struct netdev_notifier_info info; /* must be first */
2598 struct net_device *upper_dev; /* new upper dev */
2599 bool master; /* is upper dev master */
2600 bool linking; /* is the notification for link or unlink */
2601 void *upper_info; /* upper dev info */
2604 struct netdev_notifier_changelowerstate_info {
2605 struct netdev_notifier_info info; /* must be first */
2606 void *lower_state_info; /* is lower dev state */
2609 struct netdev_notifier_pre_changeaddr_info {
2610 struct netdev_notifier_info info; /* must be first */
2611 const unsigned char *dev_addr;
2614 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2615 struct net_device *dev)
2618 info->extack = NULL;
2621 static inline struct net_device *
2622 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2627 static inline struct netlink_ext_ack *
2628 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2630 return info->extack;
2633 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2636 extern rwlock_t dev_base_lock; /* Device list lock */
2638 #define for_each_netdev(net, d) \
2639 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2640 #define for_each_netdev_reverse(net, d) \
2641 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2642 #define for_each_netdev_rcu(net, d) \
2643 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2644 #define for_each_netdev_safe(net, d, n) \
2645 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2646 #define for_each_netdev_continue(net, d) \
2647 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2648 #define for_each_netdev_continue_reverse(net, d) \
2649 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2651 #define for_each_netdev_continue_rcu(net, d) \
2652 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2653 #define for_each_netdev_in_bond_rcu(bond, slave) \
2654 for_each_netdev_rcu(&init_net, slave) \
2655 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2656 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2658 static inline struct net_device *next_net_device(struct net_device *dev)
2660 struct list_head *lh;
2664 lh = dev->dev_list.next;
2665 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2668 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2670 struct list_head *lh;
2674 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2675 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2678 static inline struct net_device *first_net_device(struct net *net)
2680 return list_empty(&net->dev_base_head) ? NULL :
2681 net_device_entry(net->dev_base_head.next);
2684 static inline struct net_device *first_net_device_rcu(struct net *net)
2686 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2688 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2691 int netdev_boot_setup_check(struct net_device *dev);
2692 unsigned long netdev_boot_base(const char *prefix, int unit);
2693 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2694 const char *hwaddr);
2695 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2696 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2697 void dev_add_pack(struct packet_type *pt);
2698 void dev_remove_pack(struct packet_type *pt);
2699 void __dev_remove_pack(struct packet_type *pt);
2700 void dev_add_offload(struct packet_offload *po);
2701 void dev_remove_offload(struct packet_offload *po);
2703 int dev_get_iflink(const struct net_device *dev);
2704 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2705 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2706 unsigned short mask);
2707 struct net_device *dev_get_by_name(struct net *net, const char *name);
2708 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2709 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2710 int dev_alloc_name(struct net_device *dev, const char *name);
2711 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2712 void dev_close(struct net_device *dev);
2713 void dev_close_many(struct list_head *head, bool unlink);
2714 void dev_disable_lro(struct net_device *dev);
2715 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2716 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2717 struct net_device *sb_dev);
2718 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2719 struct net_device *sb_dev);
2720 int dev_queue_xmit(struct sk_buff *skb);
2721 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2722 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2723 int register_netdevice(struct net_device *dev);
2724 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2725 void unregister_netdevice_many(struct list_head *head);
2726 static inline void unregister_netdevice(struct net_device *dev)
2728 unregister_netdevice_queue(dev, NULL);
2731 int netdev_refcnt_read(const struct net_device *dev);
2732 void free_netdev(struct net_device *dev);
2733 void netdev_freemem(struct net_device *dev);
2734 void synchronize_net(void);
2735 int init_dummy_netdev(struct net_device *dev);
2737 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2738 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2739 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2740 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2741 int netdev_get_name(struct net *net, char *name, int ifindex);
2742 int dev_restart(struct net_device *dev);
2743 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2744 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2746 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2748 return NAPI_GRO_CB(skb)->data_offset;
2751 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2753 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2756 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2758 NAPI_GRO_CB(skb)->data_offset += len;
2761 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2762 unsigned int offset)
2764 return NAPI_GRO_CB(skb)->frag0 + offset;
2767 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2769 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2772 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2774 NAPI_GRO_CB(skb)->frag0 = NULL;
2775 NAPI_GRO_CB(skb)->frag0_len = 0;
2778 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2779 unsigned int offset)
2781 if (!pskb_may_pull(skb, hlen))
2784 skb_gro_frag0_invalidate(skb);
2785 return skb->data + offset;
2788 static inline void *skb_gro_network_header(struct sk_buff *skb)
2790 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2791 skb_network_offset(skb);
2794 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2795 const void *start, unsigned int len)
2797 if (NAPI_GRO_CB(skb)->csum_valid)
2798 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2799 csum_partial(start, len, 0));
2802 /* GRO checksum functions. These are logical equivalents of the normal
2803 * checksum functions (in skbuff.h) except that they operate on the GRO
2804 * offsets and fields in sk_buff.
2807 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2809 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2811 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2814 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2818 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2819 skb_checksum_start_offset(skb) <
2820 skb_gro_offset(skb)) &&
2821 !skb_at_gro_remcsum_start(skb) &&
2822 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2823 (!zero_okay || check));
2826 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2829 if (NAPI_GRO_CB(skb)->csum_valid &&
2830 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2833 NAPI_GRO_CB(skb)->csum = psum;
2835 return __skb_gro_checksum_complete(skb);
2838 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2840 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2841 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2842 NAPI_GRO_CB(skb)->csum_cnt--;
2844 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2845 * verified a new top level checksum or an encapsulated one
2846 * during GRO. This saves work if we fallback to normal path.
2848 __skb_incr_checksum_unnecessary(skb);
2852 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2855 __sum16 __ret = 0; \
2856 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2857 __ret = __skb_gro_checksum_validate_complete(skb, \
2858 compute_pseudo(skb, proto)); \
2860 skb_gro_incr_csum_unnecessary(skb); \
2864 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2865 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2867 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2869 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2871 #define skb_gro_checksum_simple_validate(skb) \
2872 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2874 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2876 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2877 !NAPI_GRO_CB(skb)->csum_valid);
2880 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2883 NAPI_GRO_CB(skb)->csum = ~pseudo;
2884 NAPI_GRO_CB(skb)->csum_valid = 1;
2887 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
2889 if (__skb_gro_checksum_convert_check(skb)) \
2890 __skb_gro_checksum_convert(skb, \
2891 compute_pseudo(skb, proto)); \
2894 struct gro_remcsum {
2899 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2905 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2906 unsigned int off, size_t hdrlen,
2907 int start, int offset,
2908 struct gro_remcsum *grc,
2912 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2914 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2917 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2921 ptr = skb_gro_header_fast(skb, off);
2922 if (skb_gro_header_hard(skb, off + plen)) {
2923 ptr = skb_gro_header_slow(skb, off + plen, off);
2928 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2931 /* Adjust skb->csum since we changed the packet */
2932 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2934 grc->offset = off + hdrlen + offset;
2940 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2941 struct gro_remcsum *grc)
2944 size_t plen = grc->offset + sizeof(u16);
2949 ptr = skb_gro_header_fast(skb, grc->offset);
2950 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2951 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2956 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2959 #ifdef CONFIG_XFRM_OFFLOAD
2960 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2962 if (PTR_ERR(pp) != -EINPROGRESS)
2963 NAPI_GRO_CB(skb)->flush |= flush;
2965 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2968 struct gro_remcsum *grc)
2970 if (PTR_ERR(pp) != -EINPROGRESS) {
2971 NAPI_GRO_CB(skb)->flush |= flush;
2972 skb_gro_remcsum_cleanup(skb, grc);
2973 skb->remcsum_offload = 0;
2977 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2979 NAPI_GRO_CB(skb)->flush |= flush;
2981 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2984 struct gro_remcsum *grc)
2986 NAPI_GRO_CB(skb)->flush |= flush;
2987 skb_gro_remcsum_cleanup(skb, grc);
2988 skb->remcsum_offload = 0;
2992 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2993 unsigned short type,
2994 const void *daddr, const void *saddr,
2997 if (!dev->header_ops || !dev->header_ops->create)
3000 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3003 static inline int dev_parse_header(const struct sk_buff *skb,
3004 unsigned char *haddr)
3006 const struct net_device *dev = skb->dev;
3008 if (!dev->header_ops || !dev->header_ops->parse)
3010 return dev->header_ops->parse(skb, haddr);
3013 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3015 const struct net_device *dev = skb->dev;
3017 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3019 return dev->header_ops->parse_protocol(skb);
3022 /* ll_header must have at least hard_header_len allocated */
3023 static inline bool dev_validate_header(const struct net_device *dev,
3024 char *ll_header, int len)
3026 if (likely(len >= dev->hard_header_len))
3028 if (len < dev->min_header_len)
3031 if (capable(CAP_SYS_RAWIO)) {
3032 memset(ll_header + len, 0, dev->hard_header_len - len);
3036 if (dev->header_ops && dev->header_ops->validate)
3037 return dev->header_ops->validate(ll_header, len);
3042 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3044 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3045 static inline int unregister_gifconf(unsigned int family)
3047 return register_gifconf(family, NULL);
3050 #ifdef CONFIG_NET_FLOW_LIMIT
3051 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3052 struct sd_flow_limit {
3054 unsigned int num_buckets;
3055 unsigned int history_head;
3056 u16 history[FLOW_LIMIT_HISTORY];
3060 extern int netdev_flow_limit_table_len;
3061 #endif /* CONFIG_NET_FLOW_LIMIT */
3064 * Incoming packets are placed on per-CPU queues
3066 struct softnet_data {
3067 struct list_head poll_list;
3068 struct sk_buff_head process_queue;
3071 unsigned int processed;
3072 unsigned int time_squeeze;
3073 unsigned int received_rps;
3075 struct softnet_data *rps_ipi_list;
3077 #ifdef CONFIG_NET_FLOW_LIMIT
3078 struct sd_flow_limit __rcu *flow_limit;
3080 struct Qdisc *output_queue;
3081 struct Qdisc **output_queue_tailp;
3082 struct sk_buff *completion_queue;
3083 #ifdef CONFIG_XFRM_OFFLOAD
3084 struct sk_buff_head xfrm_backlog;
3086 /* written and read only by owning cpu: */
3092 /* input_queue_head should be written by cpu owning this struct,
3093 * and only read by other cpus. Worth using a cache line.
3095 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3097 /* Elements below can be accessed between CPUs for RPS/RFS */
3098 call_single_data_t csd ____cacheline_aligned_in_smp;
3099 struct softnet_data *rps_ipi_next;
3101 unsigned int input_queue_tail;
3103 unsigned int dropped;
3104 struct sk_buff_head input_pkt_queue;
3105 struct napi_struct backlog;
3109 static inline void input_queue_head_incr(struct softnet_data *sd)
3112 sd->input_queue_head++;
3116 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3117 unsigned int *qtail)
3120 *qtail = ++sd->input_queue_tail;
3124 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3126 static inline int dev_recursion_level(void)
3128 return this_cpu_read(softnet_data.xmit.recursion);
3131 #define XMIT_RECURSION_LIMIT 10
3132 static inline bool dev_xmit_recursion(void)
3134 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3135 XMIT_RECURSION_LIMIT);
3138 static inline void dev_xmit_recursion_inc(void)
3140 __this_cpu_inc(softnet_data.xmit.recursion);
3143 static inline void dev_xmit_recursion_dec(void)
3145 __this_cpu_dec(softnet_data.xmit.recursion);
3148 void __netif_schedule(struct Qdisc *q);
3149 void netif_schedule_queue(struct netdev_queue *txq);
3151 static inline void netif_tx_schedule_all(struct net_device *dev)
3155 for (i = 0; i < dev->num_tx_queues; i++)
3156 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3159 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3161 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3165 * netif_start_queue - allow transmit
3166 * @dev: network device
3168 * Allow upper layers to call the device hard_start_xmit routine.
3170 static inline void netif_start_queue(struct net_device *dev)
3172 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3175 static inline void netif_tx_start_all_queues(struct net_device *dev)
3179 for (i = 0; i < dev->num_tx_queues; i++) {
3180 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3181 netif_tx_start_queue(txq);
3185 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3188 * netif_wake_queue - restart transmit
3189 * @dev: network device
3191 * Allow upper layers to call the device hard_start_xmit routine.
3192 * Used for flow control when transmit resources are available.
3194 static inline void netif_wake_queue(struct net_device *dev)
3196 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3199 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3203 for (i = 0; i < dev->num_tx_queues; i++) {
3204 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3205 netif_tx_wake_queue(txq);
3209 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3211 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3215 * netif_stop_queue - stop transmitted packets
3216 * @dev: network device
3218 * Stop upper layers calling the device hard_start_xmit routine.
3219 * Used for flow control when transmit resources are unavailable.
3221 static inline void netif_stop_queue(struct net_device *dev)
3223 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3226 void netif_tx_stop_all_queues(struct net_device *dev);
3227 void netdev_update_lockdep_key(struct net_device *dev);
3229 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3231 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3235 * netif_queue_stopped - test if transmit queue is flowblocked
3236 * @dev: network device
3238 * Test if transmit queue on device is currently unable to send.
3240 static inline bool netif_queue_stopped(const struct net_device *dev)
3242 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3245 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3247 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3251 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3253 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3257 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3259 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3263 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3264 * @dev_queue: pointer to transmit queue
3266 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3267 * to give appropriate hint to the CPU.
3269 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3272 prefetchw(&dev_queue->dql.num_queued);
3277 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3278 * @dev_queue: pointer to transmit queue
3280 * BQL enabled drivers might use this helper in their TX completion path,
3281 * to give appropriate hint to the CPU.
3283 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3286 prefetchw(&dev_queue->dql.limit);
3290 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3294 dql_queued(&dev_queue->dql, bytes);
3296 if (likely(dql_avail(&dev_queue->dql) >= 0))
3299 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3302 * The XOFF flag must be set before checking the dql_avail below,
3303 * because in netdev_tx_completed_queue we update the dql_completed
3304 * before checking the XOFF flag.
3308 /* check again in case another CPU has just made room avail */
3309 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3310 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3314 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3315 * that they should not test BQL status themselves.
3316 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3318 * Returns true if the doorbell must be used to kick the NIC.
3320 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3326 dql_queued(&dev_queue->dql, bytes);
3328 return netif_tx_queue_stopped(dev_queue);
3330 netdev_tx_sent_queue(dev_queue, bytes);
3335 * netdev_sent_queue - report the number of bytes queued to hardware
3336 * @dev: network device
3337 * @bytes: number of bytes queued to the hardware device queue
3339 * Report the number of bytes queued for sending/completion to the network
3340 * device hardware queue. @bytes should be a good approximation and should
3341 * exactly match netdev_completed_queue() @bytes
3343 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3345 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3348 static inline bool __netdev_sent_queue(struct net_device *dev,
3352 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3356 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3357 unsigned int pkts, unsigned int bytes)
3360 if (unlikely(!bytes))
3363 dql_completed(&dev_queue->dql, bytes);
3366 * Without the memory barrier there is a small possiblity that
3367 * netdev_tx_sent_queue will miss the update and cause the queue to
3368 * be stopped forever
3372 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3375 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3376 netif_schedule_queue(dev_queue);
3381 * netdev_completed_queue - report bytes and packets completed by device
3382 * @dev: network device
3383 * @pkts: actual number of packets sent over the medium
3384 * @bytes: actual number of bytes sent over the medium
3386 * Report the number of bytes and packets transmitted by the network device
3387 * hardware queue over the physical medium, @bytes must exactly match the
3388 * @bytes amount passed to netdev_sent_queue()
3390 static inline void netdev_completed_queue(struct net_device *dev,
3391 unsigned int pkts, unsigned int bytes)
3393 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3396 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3399 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3405 * netdev_reset_queue - reset the packets and bytes count of a network device
3406 * @dev_queue: network device
3408 * Reset the bytes and packet count of a network device and clear the
3409 * software flow control OFF bit for this network device
3411 static inline void netdev_reset_queue(struct net_device *dev_queue)
3413 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3417 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3418 * @dev: network device
3419 * @queue_index: given tx queue index
3421 * Returns 0 if given tx queue index >= number of device tx queues,
3422 * otherwise returns the originally passed tx queue index.
3424 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3426 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3427 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3428 dev->name, queue_index,
3429 dev->real_num_tx_queues);
3437 * netif_running - test if up
3438 * @dev: network device
3440 * Test if the device has been brought up.
3442 static inline bool netif_running(const struct net_device *dev)
3444 return test_bit(__LINK_STATE_START, &dev->state);
3448 * Routines to manage the subqueues on a device. We only need start,
3449 * stop, and a check if it's stopped. All other device management is
3450 * done at the overall netdevice level.
3451 * Also test the device if we're multiqueue.
3455 * netif_start_subqueue - allow sending packets on subqueue
3456 * @dev: network device
3457 * @queue_index: sub queue index
3459 * Start individual transmit queue of a device with multiple transmit queues.
3461 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3463 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3465 netif_tx_start_queue(txq);
3469 * netif_stop_subqueue - stop sending packets on subqueue
3470 * @dev: network device
3471 * @queue_index: sub queue index
3473 * Stop individual transmit queue of a device with multiple transmit queues.
3475 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3477 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3478 netif_tx_stop_queue(txq);
3482 * netif_subqueue_stopped - test status of subqueue
3483 * @dev: network device
3484 * @queue_index: sub queue index
3486 * Check individual transmit queue of a device with multiple transmit queues.
3488 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3491 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3493 return netif_tx_queue_stopped(txq);
3496 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3497 struct sk_buff *skb)
3499 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3503 * netif_wake_subqueue - allow sending packets on subqueue
3504 * @dev: network device
3505 * @queue_index: sub queue index
3507 * Resume individual transmit queue of a device with multiple transmit queues.
3509 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3511 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3513 netif_tx_wake_queue(txq);
3517 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3519 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3520 u16 index, bool is_rxqs_map);
3523 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3524 * @j: CPU/Rx queue index
3525 * @mask: bitmask of all cpus/rx queues
3526 * @nr_bits: number of bits in the bitmask
3528 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3530 static inline bool netif_attr_test_mask(unsigned long j,
3531 const unsigned long *mask,
3532 unsigned int nr_bits)
3534 cpu_max_bits_warn(j, nr_bits);
3535 return test_bit(j, mask);
3539 * netif_attr_test_online - Test for online CPU/Rx queue
3540 * @j: CPU/Rx queue index
3541 * @online_mask: bitmask for CPUs/Rx queues that are online
3542 * @nr_bits: number of bits in the bitmask
3544 * Returns true if a CPU/Rx queue is online.
3546 static inline bool netif_attr_test_online(unsigned long j,
3547 const unsigned long *online_mask,
3548 unsigned int nr_bits)
3550 cpu_max_bits_warn(j, nr_bits);
3553 return test_bit(j, online_mask);
3555 return (j < nr_bits);
3559 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3560 * @n: CPU/Rx queue index
3561 * @srcp: the cpumask/Rx queue mask pointer
3562 * @nr_bits: number of bits in the bitmask
3564 * Returns >= nr_bits if no further CPUs/Rx queues set.
3566 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3567 unsigned int nr_bits)
3569 /* -1 is a legal arg here. */
3571 cpu_max_bits_warn(n, nr_bits);
3574 return find_next_bit(srcp, nr_bits, n + 1);
3580 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3581 * @n: CPU/Rx queue index
3582 * @src1p: the first CPUs/Rx queues mask pointer
3583 * @src2p: the second CPUs/Rx queues mask pointer
3584 * @nr_bits: number of bits in the bitmask
3586 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3588 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3589 const unsigned long *src2p,
3590 unsigned int nr_bits)
3592 /* -1 is a legal arg here. */
3594 cpu_max_bits_warn(n, nr_bits);
3597 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3599 return find_next_bit(src1p, nr_bits, n + 1);
3601 return find_next_bit(src2p, nr_bits, n + 1);
3606 static inline int netif_set_xps_queue(struct net_device *dev,
3607 const struct cpumask *mask,
3613 static inline int __netif_set_xps_queue(struct net_device *dev,
3614 const unsigned long *mask,
3615 u16 index, bool is_rxqs_map)
3622 * netif_is_multiqueue - test if device has multiple transmit queues
3623 * @dev: network device
3625 * Check if device has multiple transmit queues
3627 static inline bool netif_is_multiqueue(const struct net_device *dev)
3629 return dev->num_tx_queues > 1;
3632 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3635 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3637 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3640 dev->real_num_rx_queues = rxqs;
3645 static inline struct netdev_rx_queue *
3646 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3648 return dev->_rx + rxq;
3652 static inline unsigned int get_netdev_rx_queue_index(
3653 struct netdev_rx_queue *queue)
3655 struct net_device *dev = queue->dev;
3656 int index = queue - dev->_rx;
3658 BUG_ON(index >= dev->num_rx_queues);
3663 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3664 int netif_get_num_default_rss_queues(void);
3666 enum skb_free_reason {
3667 SKB_REASON_CONSUMED,
3671 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3672 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3675 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3676 * interrupt context or with hardware interrupts being disabled.
3677 * (in_irq() || irqs_disabled())
3679 * We provide four helpers that can be used in following contexts :
3681 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3682 * replacing kfree_skb(skb)
3684 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3685 * Typically used in place of consume_skb(skb) in TX completion path
3687 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3688 * replacing kfree_skb(skb)
3690 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3691 * and consumed a packet. Used in place of consume_skb(skb)
3693 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3695 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3698 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3700 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3703 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3705 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3708 static inline void dev_consume_skb_any(struct sk_buff *skb)
3710 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3713 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3714 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3715 int netif_rx(struct sk_buff *skb);
3716 int netif_rx_ni(struct sk_buff *skb);
3717 int netif_receive_skb(struct sk_buff *skb);
3718 int netif_receive_skb_core(struct sk_buff *skb);
3719 void netif_receive_skb_list(struct list_head *head);
3720 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3721 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3722 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3723 gro_result_t napi_gro_frags(struct napi_struct *napi);
3724 struct packet_offload *gro_find_receive_by_type(__be16 type);
3725 struct packet_offload *gro_find_complete_by_type(__be16 type);
3727 static inline void napi_free_frags(struct napi_struct *napi)
3729 kfree_skb(napi->skb);
3733 bool netdev_is_rx_handler_busy(struct net_device *dev);
3734 int netdev_rx_handler_register(struct net_device *dev,
3735 rx_handler_func_t *rx_handler,
3736 void *rx_handler_data);
3737 void netdev_rx_handler_unregister(struct net_device *dev);
3739 bool dev_valid_name(const char *name);
3740 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3741 bool *need_copyout);
3742 int dev_ifconf(struct net *net, struct ifconf *, int);
3743 int dev_ethtool(struct net *net, struct ifreq *);
3744 unsigned int dev_get_flags(const struct net_device *);
3745 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3746 struct netlink_ext_ack *extack);
3747 int dev_change_flags(struct net_device *dev, unsigned int flags,
3748 struct netlink_ext_ack *extack);
3749 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3750 unsigned int gchanges);
3751 int dev_change_name(struct net_device *, const char *);
3752 int dev_set_alias(struct net_device *, const char *, size_t);
3753 int dev_get_alias(const struct net_device *, char *, size_t);
3754 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3755 int __dev_set_mtu(struct net_device *, int);
3756 int dev_validate_mtu(struct net_device *dev, int mtu,
3757 struct netlink_ext_ack *extack);
3758 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3759 struct netlink_ext_ack *extack);
3760 int dev_set_mtu(struct net_device *, int);
3761 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3762 void dev_set_group(struct net_device *, int);
3763 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3764 struct netlink_ext_ack *extack);
3765 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3766 struct netlink_ext_ack *extack);
3767 int dev_change_carrier(struct net_device *, bool new_carrier);
3768 int dev_get_phys_port_id(struct net_device *dev,
3769 struct netdev_phys_item_id *ppid);
3770 int dev_get_phys_port_name(struct net_device *dev,
3771 char *name, size_t len);
3772 int dev_get_port_parent_id(struct net_device *dev,
3773 struct netdev_phys_item_id *ppid, bool recurse);
3774 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3775 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3776 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3777 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3778 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3779 struct netdev_queue *txq, int *ret);
3781 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3782 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3783 int fd, int expected_fd, u32 flags);
3784 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3785 enum bpf_netdev_command cmd);
3786 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3788 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3789 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3790 bool is_skb_forwardable(const struct net_device *dev,
3791 const struct sk_buff *skb);
3793 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3794 struct sk_buff *skb)
3796 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3797 unlikely(!is_skb_forwardable(dev, skb))) {
3798 atomic_long_inc(&dev->rx_dropped);
3803 skb_scrub_packet(skb, true);
3808 bool dev_nit_active(struct net_device *dev);
3809 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3811 extern int netdev_budget;
3812 extern unsigned int netdev_budget_usecs;
3814 /* Called by rtnetlink.c:rtnl_unlock() */
3815 void netdev_run_todo(void);
3818 * dev_put - release reference to device
3819 * @dev: network device
3821 * Release reference to device to allow it to be freed.
3823 static inline void dev_put(struct net_device *dev)
3825 this_cpu_dec(*dev->pcpu_refcnt);
3829 * dev_hold - get reference to device
3830 * @dev: network device
3832 * Hold reference to device to keep it from being freed.
3834 static inline void dev_hold(struct net_device *dev)
3836 this_cpu_inc(*dev->pcpu_refcnt);
3839 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3840 * and _off may be called from IRQ context, but it is caller
3841 * who is responsible for serialization of these calls.
3843 * The name carrier is inappropriate, these functions should really be
3844 * called netif_lowerlayer_*() because they represent the state of any
3845 * kind of lower layer not just hardware media.
3848 void linkwatch_init_dev(struct net_device *dev);
3849 void linkwatch_fire_event(struct net_device *dev);
3850 void linkwatch_forget_dev(struct net_device *dev);
3853 * netif_carrier_ok - test if carrier present
3854 * @dev: network device
3856 * Check if carrier is present on device
3858 static inline bool netif_carrier_ok(const struct net_device *dev)
3860 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3863 unsigned long dev_trans_start(struct net_device *dev);
3865 void __netdev_watchdog_up(struct net_device *dev);
3867 void netif_carrier_on(struct net_device *dev);
3869 void netif_carrier_off(struct net_device *dev);
3872 * netif_dormant_on - mark device as dormant.
3873 * @dev: network device
3875 * Mark device as dormant (as per RFC2863).
3877 * The dormant state indicates that the relevant interface is not
3878 * actually in a condition to pass packets (i.e., it is not 'up') but is
3879 * in a "pending" state, waiting for some external event. For "on-
3880 * demand" interfaces, this new state identifies the situation where the
3881 * interface is waiting for events to place it in the up state.
3883 static inline void netif_dormant_on(struct net_device *dev)
3885 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3886 linkwatch_fire_event(dev);
3890 * netif_dormant_off - set device as not dormant.
3891 * @dev: network device
3893 * Device is not in dormant state.
3895 static inline void netif_dormant_off(struct net_device *dev)
3897 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3898 linkwatch_fire_event(dev);
3902 * netif_dormant - test if device is dormant
3903 * @dev: network device
3905 * Check if device is dormant.
3907 static inline bool netif_dormant(const struct net_device *dev)
3909 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3914 * netif_testing_on - mark device as under test.
3915 * @dev: network device
3917 * Mark device as under test (as per RFC2863).
3919 * The testing state indicates that some test(s) must be performed on
3920 * the interface. After completion, of the test, the interface state
3921 * will change to up, dormant, or down, as appropriate.
3923 static inline void netif_testing_on(struct net_device *dev)
3925 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
3926 linkwatch_fire_event(dev);
3930 * netif_testing_off - set device as not under test.
3931 * @dev: network device
3933 * Device is not in testing state.
3935 static inline void netif_testing_off(struct net_device *dev)
3937 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
3938 linkwatch_fire_event(dev);
3942 * netif_testing - test if device is under test
3943 * @dev: network device
3945 * Check if device is under test
3947 static inline bool netif_testing(const struct net_device *dev)
3949 return test_bit(__LINK_STATE_TESTING, &dev->state);
3954 * netif_oper_up - test if device is operational
3955 * @dev: network device
3957 * Check if carrier is operational
3959 static inline bool netif_oper_up(const struct net_device *dev)
3961 return (dev->operstate == IF_OPER_UP ||
3962 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3966 * netif_device_present - is device available or removed
3967 * @dev: network device
3969 * Check if device has not been removed from system.
3971 static inline bool netif_device_present(struct net_device *dev)
3973 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3976 void netif_device_detach(struct net_device *dev);
3978 void netif_device_attach(struct net_device *dev);
3981 * Network interface message level settings
3986 NETIF_MSG_PROBE_BIT,
3988 NETIF_MSG_TIMER_BIT,
3989 NETIF_MSG_IFDOWN_BIT,
3991 NETIF_MSG_RX_ERR_BIT,
3992 NETIF_MSG_TX_ERR_BIT,
3993 NETIF_MSG_TX_QUEUED_BIT,
3995 NETIF_MSG_TX_DONE_BIT,
3996 NETIF_MSG_RX_STATUS_BIT,
3997 NETIF_MSG_PKTDATA_BIT,
4001 /* When you add a new bit above, update netif_msg_class_names array
4002 * in net/ethtool/common.c
4004 NETIF_MSG_CLASS_COUNT,
4006 /* Both ethtool_ops interface and internal driver implementation use u32 */
4007 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4009 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4010 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4012 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4013 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4014 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4015 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4016 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4017 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4018 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4019 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4020 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4021 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4022 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4023 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4024 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4025 #define NETIF_MSG_HW __NETIF_MSG(HW)
4026 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4028 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4029 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4030 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4031 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4032 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4033 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4034 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4035 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4036 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4037 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4038 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4039 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4040 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4041 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4042 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4044 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4047 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4048 return default_msg_enable_bits;
4049 if (debug_value == 0) /* no output */
4051 /* set low N bits */
4052 return (1U << debug_value) - 1;
4055 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4057 spin_lock(&txq->_xmit_lock);
4058 txq->xmit_lock_owner = cpu;
4061 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4063 __acquire(&txq->_xmit_lock);
4067 static inline void __netif_tx_release(struct netdev_queue *txq)
4069 __release(&txq->_xmit_lock);
4072 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4074 spin_lock_bh(&txq->_xmit_lock);
4075 txq->xmit_lock_owner = smp_processor_id();
4078 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4080 bool ok = spin_trylock(&txq->_xmit_lock);
4082 txq->xmit_lock_owner = smp_processor_id();
4086 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4088 txq->xmit_lock_owner = -1;
4089 spin_unlock(&txq->_xmit_lock);
4092 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4094 txq->xmit_lock_owner = -1;
4095 spin_unlock_bh(&txq->_xmit_lock);
4098 static inline void txq_trans_update(struct netdev_queue *txq)
4100 if (txq->xmit_lock_owner != -1)
4101 txq->trans_start = jiffies;
4104 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4105 static inline void netif_trans_update(struct net_device *dev)
4107 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4109 if (txq->trans_start != jiffies)
4110 txq->trans_start = jiffies;
4114 * netif_tx_lock - grab network device transmit lock
4115 * @dev: network device
4117 * Get network device transmit lock
4119 static inline void netif_tx_lock(struct net_device *dev)
4124 spin_lock(&dev->tx_global_lock);
4125 cpu = smp_processor_id();
4126 for (i = 0; i < dev->num_tx_queues; i++) {
4127 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4129 /* We are the only thread of execution doing a
4130 * freeze, but we have to grab the _xmit_lock in
4131 * order to synchronize with threads which are in
4132 * the ->hard_start_xmit() handler and already
4133 * checked the frozen bit.
4135 __netif_tx_lock(txq, cpu);
4136 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4137 __netif_tx_unlock(txq);
4141 static inline void netif_tx_lock_bh(struct net_device *dev)
4147 static inline void netif_tx_unlock(struct net_device *dev)
4151 for (i = 0; i < dev->num_tx_queues; i++) {
4152 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4154 /* No need to grab the _xmit_lock here. If the
4155 * queue is not stopped for another reason, we
4158 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4159 netif_schedule_queue(txq);
4161 spin_unlock(&dev->tx_global_lock);
4164 static inline void netif_tx_unlock_bh(struct net_device *dev)
4166 netif_tx_unlock(dev);
4170 #define HARD_TX_LOCK(dev, txq, cpu) { \
4171 if ((dev->features & NETIF_F_LLTX) == 0) { \
4172 __netif_tx_lock(txq, cpu); \
4174 __netif_tx_acquire(txq); \
4178 #define HARD_TX_TRYLOCK(dev, txq) \
4179 (((dev->features & NETIF_F_LLTX) == 0) ? \
4180 __netif_tx_trylock(txq) : \
4181 __netif_tx_acquire(txq))
4183 #define HARD_TX_UNLOCK(dev, txq) { \
4184 if ((dev->features & NETIF_F_LLTX) == 0) { \
4185 __netif_tx_unlock(txq); \
4187 __netif_tx_release(txq); \
4191 static inline void netif_tx_disable(struct net_device *dev)
4197 cpu = smp_processor_id();
4198 for (i = 0; i < dev->num_tx_queues; i++) {
4199 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4201 __netif_tx_lock(txq, cpu);
4202 netif_tx_stop_queue(txq);
4203 __netif_tx_unlock(txq);
4208 static inline void netif_addr_lock(struct net_device *dev)
4210 spin_lock(&dev->addr_list_lock);
4213 static inline void netif_addr_lock_bh(struct net_device *dev)
4215 spin_lock_bh(&dev->addr_list_lock);
4218 static inline void netif_addr_unlock(struct net_device *dev)
4220 spin_unlock(&dev->addr_list_lock);
4223 static inline void netif_addr_unlock_bh(struct net_device *dev)
4225 spin_unlock_bh(&dev->addr_list_lock);
4229 * dev_addrs walker. Should be used only for read access. Call with
4230 * rcu_read_lock held.
4232 #define for_each_dev_addr(dev, ha) \
4233 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4235 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4237 void ether_setup(struct net_device *dev);
4239 /* Support for loadable net-drivers */
4240 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4241 unsigned char name_assign_type,
4242 void (*setup)(struct net_device *),
4243 unsigned int txqs, unsigned int rxqs);
4244 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4245 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4247 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4248 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4251 int register_netdev(struct net_device *dev);
4252 void unregister_netdev(struct net_device *dev);
4254 /* General hardware address lists handling functions */
4255 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4256 struct netdev_hw_addr_list *from_list, int addr_len);
4257 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4258 struct netdev_hw_addr_list *from_list, int addr_len);
4259 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4260 struct net_device *dev,
4261 int (*sync)(struct net_device *, const unsigned char *),
4262 int (*unsync)(struct net_device *,
4263 const unsigned char *));
4264 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4265 struct net_device *dev,
4266 int (*sync)(struct net_device *,
4267 const unsigned char *, int),
4268 int (*unsync)(struct net_device *,
4269 const unsigned char *, int));
4270 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4271 struct net_device *dev,
4272 int (*unsync)(struct net_device *,
4273 const unsigned char *, int));
4274 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4275 struct net_device *dev,
4276 int (*unsync)(struct net_device *,
4277 const unsigned char *));
4278 void __hw_addr_init(struct netdev_hw_addr_list *list);
4280 /* Functions used for device addresses handling */
4281 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4282 unsigned char addr_type);
4283 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4284 unsigned char addr_type);
4285 void dev_addr_flush(struct net_device *dev);
4286 int dev_addr_init(struct net_device *dev);
4288 /* Functions used for unicast addresses handling */
4289 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4290 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4291 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4292 int dev_uc_sync(struct net_device *to, struct net_device *from);
4293 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4294 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4295 void dev_uc_flush(struct net_device *dev);
4296 void dev_uc_init(struct net_device *dev);
4299 * __dev_uc_sync - Synchonize device's unicast list
4300 * @dev: device to sync
4301 * @sync: function to call if address should be added
4302 * @unsync: function to call if address should be removed
4304 * Add newly added addresses to the interface, and release
4305 * addresses that have been deleted.
4307 static inline int __dev_uc_sync(struct net_device *dev,
4308 int (*sync)(struct net_device *,
4309 const unsigned char *),
4310 int (*unsync)(struct net_device *,
4311 const unsigned char *))
4313 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4317 * __dev_uc_unsync - Remove synchronized addresses from device
4318 * @dev: device to sync
4319 * @unsync: function to call if address should be removed
4321 * Remove all addresses that were added to the device by dev_uc_sync().
4323 static inline void __dev_uc_unsync(struct net_device *dev,
4324 int (*unsync)(struct net_device *,
4325 const unsigned char *))
4327 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4330 /* Functions used for multicast addresses handling */
4331 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4332 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4333 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4334 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4335 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4336 int dev_mc_sync(struct net_device *to, struct net_device *from);
4337 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4338 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4339 void dev_mc_flush(struct net_device *dev);
4340 void dev_mc_init(struct net_device *dev);
4343 * __dev_mc_sync - Synchonize device's multicast list
4344 * @dev: device to sync
4345 * @sync: function to call if address should be added
4346 * @unsync: function to call if address should be removed
4348 * Add newly added addresses to the interface, and release
4349 * addresses that have been deleted.
4351 static inline int __dev_mc_sync(struct net_device *dev,
4352 int (*sync)(struct net_device *,
4353 const unsigned char *),
4354 int (*unsync)(struct net_device *,
4355 const unsigned char *))
4357 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4361 * __dev_mc_unsync - Remove synchronized addresses from device
4362 * @dev: device to sync
4363 * @unsync: function to call if address should be removed
4365 * Remove all addresses that were added to the device by dev_mc_sync().
4367 static inline void __dev_mc_unsync(struct net_device *dev,
4368 int (*unsync)(struct net_device *,
4369 const unsigned char *))
4371 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4374 /* Functions used for secondary unicast and multicast support */
4375 void dev_set_rx_mode(struct net_device *dev);
4376 void __dev_set_rx_mode(struct net_device *dev);
4377 int dev_set_promiscuity(struct net_device *dev, int inc);
4378 int dev_set_allmulti(struct net_device *dev, int inc);
4379 void netdev_state_change(struct net_device *dev);
4380 void netdev_notify_peers(struct net_device *dev);
4381 void netdev_features_change(struct net_device *dev);
4382 /* Load a device via the kmod */
4383 void dev_load(struct net *net, const char *name);
4384 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4385 struct rtnl_link_stats64 *storage);
4386 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4387 const struct net_device_stats *netdev_stats);
4389 extern int netdev_max_backlog;
4390 extern int netdev_tstamp_prequeue;
4391 extern int weight_p;
4392 extern int dev_weight_rx_bias;
4393 extern int dev_weight_tx_bias;
4394 extern int dev_rx_weight;
4395 extern int dev_tx_weight;
4396 extern int gro_normal_batch;
4398 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4399 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4400 struct list_head **iter);
4401 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4402 struct list_head **iter);
4404 /* iterate through upper list, must be called under RCU read lock */
4405 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4406 for (iter = &(dev)->adj_list.upper, \
4407 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4409 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4411 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4412 int (*fn)(struct net_device *upper_dev,
4416 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4417 struct net_device *upper_dev);
4419 bool netdev_has_any_upper_dev(struct net_device *dev);
4421 void *netdev_lower_get_next_private(struct net_device *dev,
4422 struct list_head **iter);
4423 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4424 struct list_head **iter);
4426 #define netdev_for_each_lower_private(dev, priv, iter) \
4427 for (iter = (dev)->adj_list.lower.next, \
4428 priv = netdev_lower_get_next_private(dev, &(iter)); \
4430 priv = netdev_lower_get_next_private(dev, &(iter)))
4432 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4433 for (iter = &(dev)->adj_list.lower, \
4434 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4436 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4438 void *netdev_lower_get_next(struct net_device *dev,
4439 struct list_head **iter);
4441 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4442 for (iter = (dev)->adj_list.lower.next, \
4443 ldev = netdev_lower_get_next(dev, &(iter)); \
4445 ldev = netdev_lower_get_next(dev, &(iter)))
4447 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4448 struct list_head **iter);
4449 int netdev_walk_all_lower_dev(struct net_device *dev,
4450 int (*fn)(struct net_device *lower_dev,
4453 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4454 int (*fn)(struct net_device *lower_dev,
4458 void *netdev_adjacent_get_private(struct list_head *adj_list);
4459 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4460 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4461 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4462 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4463 struct netlink_ext_ack *extack);
4464 int netdev_master_upper_dev_link(struct net_device *dev,
4465 struct net_device *upper_dev,
4466 void *upper_priv, void *upper_info,
4467 struct netlink_ext_ack *extack);
4468 void netdev_upper_dev_unlink(struct net_device *dev,
4469 struct net_device *upper_dev);
4470 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4471 struct net_device *new_dev,
4472 struct net_device *dev,
4473 struct netlink_ext_ack *extack);
4474 void netdev_adjacent_change_commit(struct net_device *old_dev,
4475 struct net_device *new_dev,
4476 struct net_device *dev);
4477 void netdev_adjacent_change_abort(struct net_device *old_dev,
4478 struct net_device *new_dev,
4479 struct net_device *dev);
4480 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4481 void *netdev_lower_dev_get_private(struct net_device *dev,
4482 struct net_device *lower_dev);
4483 void netdev_lower_state_changed(struct net_device *lower_dev,
4484 void *lower_state_info);
4486 /* RSS keys are 40 or 52 bytes long */
4487 #define NETDEV_RSS_KEY_LEN 52
4488 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4489 void netdev_rss_key_fill(void *buffer, size_t len);
4491 int skb_checksum_help(struct sk_buff *skb);
4492 int skb_crc32c_csum_help(struct sk_buff *skb);
4493 int skb_csum_hwoffload_help(struct sk_buff *skb,
4494 const netdev_features_t features);
4496 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4497 netdev_features_t features, bool tx_path);
4498 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4499 netdev_features_t features);
4501 struct netdev_bonding_info {
4506 struct netdev_notifier_bonding_info {
4507 struct netdev_notifier_info info; /* must be first */
4508 struct netdev_bonding_info bonding_info;
4511 void netdev_bonding_info_change(struct net_device *dev,
4512 struct netdev_bonding_info *bonding_info);
4514 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4515 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4517 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4524 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4526 return __skb_gso_segment(skb, features, true);
4528 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4530 static inline bool can_checksum_protocol(netdev_features_t features,
4533 if (protocol == htons(ETH_P_FCOE))
4534 return !!(features & NETIF_F_FCOE_CRC);
4536 /* Assume this is an IP checksum (not SCTP CRC) */
4538 if (features & NETIF_F_HW_CSUM) {
4539 /* Can checksum everything */
4544 case htons(ETH_P_IP):
4545 return !!(features & NETIF_F_IP_CSUM);
4546 case htons(ETH_P_IPV6):
4547 return !!(features & NETIF_F_IPV6_CSUM);
4554 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4556 static inline void netdev_rx_csum_fault(struct net_device *dev,
4557 struct sk_buff *skb)
4561 /* rx skb timestamps */
4562 void net_enable_timestamp(void);
4563 void net_disable_timestamp(void);
4565 #ifdef CONFIG_PROC_FS
4566 int __init dev_proc_init(void);
4568 #define dev_proc_init() 0
4571 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4572 struct sk_buff *skb, struct net_device *dev,
4575 __this_cpu_write(softnet_data.xmit.more, more);
4576 return ops->ndo_start_xmit(skb, dev);
4579 static inline bool netdev_xmit_more(void)
4581 return __this_cpu_read(softnet_data.xmit.more);
4584 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4585 struct netdev_queue *txq, bool more)
4587 const struct net_device_ops *ops = dev->netdev_ops;
4590 rc = __netdev_start_xmit(ops, skb, dev, more);
4591 if (rc == NETDEV_TX_OK)
4592 txq_trans_update(txq);
4597 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4599 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4602 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4604 return netdev_class_create_file_ns(class_attr, NULL);
4607 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4609 netdev_class_remove_file_ns(class_attr, NULL);
4612 extern const struct kobj_ns_type_operations net_ns_type_operations;
4614 const char *netdev_drivername(const struct net_device *dev);
4616 void linkwatch_run_queue(void);
4618 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4619 netdev_features_t f2)
4621 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4622 if (f1 & NETIF_F_HW_CSUM)
4623 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4625 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4631 static inline netdev_features_t netdev_get_wanted_features(
4632 struct net_device *dev)
4634 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4636 netdev_features_t netdev_increment_features(netdev_features_t all,
4637 netdev_features_t one, netdev_features_t mask);
4639 /* Allow TSO being used on stacked device :
4640 * Performing the GSO segmentation before last device
4641 * is a performance improvement.
4643 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4644 netdev_features_t mask)
4646 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4649 int __netdev_update_features(struct net_device *dev);
4650 void netdev_update_features(struct net_device *dev);
4651 void netdev_change_features(struct net_device *dev);
4653 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4654 struct net_device *dev);
4656 netdev_features_t passthru_features_check(struct sk_buff *skb,
4657 struct net_device *dev,
4658 netdev_features_t features);
4659 netdev_features_t netif_skb_features(struct sk_buff *skb);
4661 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4663 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4665 /* check flags correspondence */
4666 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4667 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4668 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4669 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4670 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4671 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4672 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4673 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4674 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4675 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4676 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4677 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4678 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4679 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4680 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4681 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4682 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4683 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4684 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4686 return (features & feature) == feature;
4689 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4691 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4692 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4695 static inline bool netif_needs_gso(struct sk_buff *skb,
4696 netdev_features_t features)
4698 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4699 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4700 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4703 static inline void netif_set_gso_max_size(struct net_device *dev,
4706 dev->gso_max_size = size;
4709 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4710 int pulled_hlen, u16 mac_offset,
4713 skb->protocol = protocol;
4714 skb->encapsulation = 1;
4715 skb_push(skb, pulled_hlen);
4716 skb_reset_transport_header(skb);
4717 skb->mac_header = mac_offset;
4718 skb->network_header = skb->mac_header + mac_len;
4719 skb->mac_len = mac_len;
4722 static inline bool netif_is_macsec(const struct net_device *dev)
4724 return dev->priv_flags & IFF_MACSEC;
4727 static inline bool netif_is_macvlan(const struct net_device *dev)
4729 return dev->priv_flags & IFF_MACVLAN;
4732 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4734 return dev->priv_flags & IFF_MACVLAN_PORT;
4737 static inline bool netif_is_bond_master(const struct net_device *dev)
4739 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4742 static inline bool netif_is_bond_slave(const struct net_device *dev)
4744 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4747 static inline bool netif_supports_nofcs(struct net_device *dev)
4749 return dev->priv_flags & IFF_SUPP_NOFCS;
4752 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4754 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4757 static inline bool netif_is_l3_master(const struct net_device *dev)
4759 return dev->priv_flags & IFF_L3MDEV_MASTER;
4762 static inline bool netif_is_l3_slave(const struct net_device *dev)
4764 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4767 static inline bool netif_is_bridge_master(const struct net_device *dev)
4769 return dev->priv_flags & IFF_EBRIDGE;
4772 static inline bool netif_is_bridge_port(const struct net_device *dev)
4774 return dev->priv_flags & IFF_BRIDGE_PORT;
4777 static inline bool netif_is_ovs_master(const struct net_device *dev)
4779 return dev->priv_flags & IFF_OPENVSWITCH;
4782 static inline bool netif_is_ovs_port(const struct net_device *dev)
4784 return dev->priv_flags & IFF_OVS_DATAPATH;
4787 static inline bool netif_is_team_master(const struct net_device *dev)
4789 return dev->priv_flags & IFF_TEAM;
4792 static inline bool netif_is_team_port(const struct net_device *dev)
4794 return dev->priv_flags & IFF_TEAM_PORT;
4797 static inline bool netif_is_lag_master(const struct net_device *dev)
4799 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4802 static inline bool netif_is_lag_port(const struct net_device *dev)
4804 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4807 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4809 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4812 static inline bool netif_is_failover(const struct net_device *dev)
4814 return dev->priv_flags & IFF_FAILOVER;
4817 static inline bool netif_is_failover_slave(const struct net_device *dev)
4819 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4822 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4823 static inline void netif_keep_dst(struct net_device *dev)
4825 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4828 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4829 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4831 /* TODO: reserve and use an additional IFF bit, if we get more users */
4832 return dev->priv_flags & IFF_MACSEC;
4835 extern struct pernet_operations __net_initdata loopback_net_ops;
4837 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4839 /* netdev_printk helpers, similar to dev_printk */
4841 static inline const char *netdev_name(const struct net_device *dev)
4843 if (!dev->name[0] || strchr(dev->name, '%'))
4844 return "(unnamed net_device)";
4848 static inline bool netdev_unregistering(const struct net_device *dev)
4850 return dev->reg_state == NETREG_UNREGISTERING;
4853 static inline const char *netdev_reg_state(const struct net_device *dev)
4855 switch (dev->reg_state) {
4856 case NETREG_UNINITIALIZED: return " (uninitialized)";
4857 case NETREG_REGISTERED: return "";
4858 case NETREG_UNREGISTERING: return " (unregistering)";
4859 case NETREG_UNREGISTERED: return " (unregistered)";
4860 case NETREG_RELEASED: return " (released)";
4861 case NETREG_DUMMY: return " (dummy)";
4864 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4865 return " (unknown)";
4868 __printf(3, 4) __cold
4869 void netdev_printk(const char *level, const struct net_device *dev,
4870 const char *format, ...);
4871 __printf(2, 3) __cold
4872 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4873 __printf(2, 3) __cold
4874 void netdev_alert(const struct net_device *dev, const char *format, ...);
4875 __printf(2, 3) __cold
4876 void netdev_crit(const struct net_device *dev, const char *format, ...);
4877 __printf(2, 3) __cold
4878 void netdev_err(const struct net_device *dev, const char *format, ...);
4879 __printf(2, 3) __cold
4880 void netdev_warn(const struct net_device *dev, const char *format, ...);
4881 __printf(2, 3) __cold
4882 void netdev_notice(const struct net_device *dev, const char *format, ...);
4883 __printf(2, 3) __cold
4884 void netdev_info(const struct net_device *dev, const char *format, ...);
4886 #define netdev_level_once(level, dev, fmt, ...) \
4888 static bool __print_once __read_mostly; \
4890 if (!__print_once) { \
4891 __print_once = true; \
4892 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4896 #define netdev_emerg_once(dev, fmt, ...) \
4897 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4898 #define netdev_alert_once(dev, fmt, ...) \
4899 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4900 #define netdev_crit_once(dev, fmt, ...) \
4901 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4902 #define netdev_err_once(dev, fmt, ...) \
4903 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4904 #define netdev_warn_once(dev, fmt, ...) \
4905 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4906 #define netdev_notice_once(dev, fmt, ...) \
4907 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4908 #define netdev_info_once(dev, fmt, ...) \
4909 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4911 #define MODULE_ALIAS_NETDEV(device) \
4912 MODULE_ALIAS("netdev-" device)
4914 #if defined(CONFIG_DYNAMIC_DEBUG)
4915 #define netdev_dbg(__dev, format, args...) \
4917 dynamic_netdev_dbg(__dev, format, ##args); \
4919 #elif defined(DEBUG)
4920 #define netdev_dbg(__dev, format, args...) \
4921 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4923 #define netdev_dbg(__dev, format, args...) \
4926 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4930 #if defined(VERBOSE_DEBUG)
4931 #define netdev_vdbg netdev_dbg
4934 #define netdev_vdbg(dev, format, args...) \
4937 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4943 * netdev_WARN() acts like dev_printk(), but with the key difference
4944 * of using a WARN/WARN_ON to get the message out, including the
4945 * file/line information and a backtrace.
4947 #define netdev_WARN(dev, format, args...) \
4948 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4949 netdev_reg_state(dev), ##args)
4951 #define netdev_WARN_ONCE(dev, format, args...) \
4952 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4953 netdev_reg_state(dev), ##args)
4955 /* netif printk helpers, similar to netdev_printk */
4957 #define netif_printk(priv, type, level, dev, fmt, args...) \
4959 if (netif_msg_##type(priv)) \
4960 netdev_printk(level, (dev), fmt, ##args); \
4963 #define netif_level(level, priv, type, dev, fmt, args...) \
4965 if (netif_msg_##type(priv)) \
4966 netdev_##level(dev, fmt, ##args); \
4969 #define netif_emerg(priv, type, dev, fmt, args...) \
4970 netif_level(emerg, priv, type, dev, fmt, ##args)
4971 #define netif_alert(priv, type, dev, fmt, args...) \
4972 netif_level(alert, priv, type, dev, fmt, ##args)
4973 #define netif_crit(priv, type, dev, fmt, args...) \
4974 netif_level(crit, priv, type, dev, fmt, ##args)
4975 #define netif_err(priv, type, dev, fmt, args...) \
4976 netif_level(err, priv, type, dev, fmt, ##args)
4977 #define netif_warn(priv, type, dev, fmt, args...) \
4978 netif_level(warn, priv, type, dev, fmt, ##args)
4979 #define netif_notice(priv, type, dev, fmt, args...) \
4980 netif_level(notice, priv, type, dev, fmt, ##args)
4981 #define netif_info(priv, type, dev, fmt, args...) \
4982 netif_level(info, priv, type, dev, fmt, ##args)
4984 #if defined(CONFIG_DYNAMIC_DEBUG)
4985 #define netif_dbg(priv, type, netdev, format, args...) \
4987 if (netif_msg_##type(priv)) \
4988 dynamic_netdev_dbg(netdev, format, ##args); \
4990 #elif defined(DEBUG)
4991 #define netif_dbg(priv, type, dev, format, args...) \
4992 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4994 #define netif_dbg(priv, type, dev, format, args...) \
4997 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5002 /* if @cond then downgrade to debug, else print at @level */
5003 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5006 netif_dbg(priv, type, netdev, fmt, ##args); \
5008 netif_ ## level(priv, type, netdev, fmt, ##args); \
5011 #if defined(VERBOSE_DEBUG)
5012 #define netif_vdbg netif_dbg
5014 #define netif_vdbg(priv, type, dev, format, args...) \
5017 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5023 * The list of packet types we will receive (as opposed to discard)
5024 * and the routines to invoke.
5026 * Why 16. Because with 16 the only overlap we get on a hash of the
5027 * low nibble of the protocol value is RARP/SNAP/X.25.
5041 #define PTYPE_HASH_SIZE (16)
5042 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5044 extern struct net_device *blackhole_netdev;
5046 #endif /* _LINUX_NETDEVICE_H */