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>
31 #include <asm/local.h>
33 #include <linux/percpu.h>
34 #include <linux/rculist.h>
35 #include <linux/workqueue.h>
36 #include <linux/dynamic_queue_limits.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>
51 #include <linux/rbtree.h>
52 #include <net/net_trackers.h>
59 struct ip_tunnel_parm;
60 struct macsec_context;
62 struct netdev_name_node;
67 /* 802.15.4 specific */
70 /* UDP Tunnel offloads */
71 struct udp_tunnel_info;
72 struct udp_tunnel_nic_info;
73 struct udp_tunnel_nic;
77 void synchronize_net(void);
78 void netdev_set_default_ethtool_ops(struct net_device *dev,
79 const struct ethtool_ops *ops);
81 /* Backlog congestion levels */
82 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
83 #define NET_RX_DROP 1 /* packet dropped */
85 #define MAX_NEST_DEV 8
88 * Transmit return codes: transmit return codes originate from three different
91 * - qdisc return codes
92 * - driver transmit return codes
95 * Drivers are allowed to return any one of those in their hard_start_xmit()
96 * function. Real network devices commonly used with qdiscs should only return
97 * the driver transmit return codes though - when qdiscs are used, the actual
98 * transmission happens asynchronously, so the value is not propagated to
99 * higher layers. Virtual network devices transmit synchronously; in this case
100 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
101 * others are propagated to higher layers.
104 /* qdisc ->enqueue() return codes. */
105 #define NET_XMIT_SUCCESS 0x00
106 #define NET_XMIT_DROP 0x01 /* skb dropped */
107 #define NET_XMIT_CN 0x02 /* congestion notification */
108 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
110 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
111 * indicates that the device will soon be dropping packets, or already drops
112 * some packets of the same priority; prompting us to send less aggressively. */
113 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
114 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
116 /* Driver transmit return codes */
117 #define NETDEV_TX_MASK 0xf0
120 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
121 NETDEV_TX_OK = 0x00, /* driver took care of packet */
122 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
124 typedef enum netdev_tx netdev_tx_t;
127 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
128 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
130 static inline bool dev_xmit_complete(int rc)
133 * Positive cases with an skb consumed by a driver:
134 * - successful transmission (rc == NETDEV_TX_OK)
135 * - error while transmitting (rc < 0)
136 * - error while queueing to a different device (rc & NET_XMIT_MASK)
138 if (likely(rc < NET_XMIT_MASK))
145 * Compute the worst-case header length according to the protocols
149 #if defined(CONFIG_HYPERV_NET)
150 # define LL_MAX_HEADER 128
151 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
152 # if defined(CONFIG_MAC80211_MESH)
153 # define LL_MAX_HEADER 128
155 # define LL_MAX_HEADER 96
158 # define LL_MAX_HEADER 32
161 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
162 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
163 #define MAX_HEADER LL_MAX_HEADER
165 #define MAX_HEADER (LL_MAX_HEADER + 48)
169 * Old network device statistics. Fields are native words
170 * (unsigned long) so they can be read and written atomically.
173 struct net_device_stats {
174 unsigned long rx_packets;
175 unsigned long tx_packets;
176 unsigned long rx_bytes;
177 unsigned long tx_bytes;
178 unsigned long rx_errors;
179 unsigned long tx_errors;
180 unsigned long rx_dropped;
181 unsigned long tx_dropped;
182 unsigned long multicast;
183 unsigned long collisions;
184 unsigned long rx_length_errors;
185 unsigned long rx_over_errors;
186 unsigned long rx_crc_errors;
187 unsigned long rx_frame_errors;
188 unsigned long rx_fifo_errors;
189 unsigned long rx_missed_errors;
190 unsigned long tx_aborted_errors;
191 unsigned long tx_carrier_errors;
192 unsigned long tx_fifo_errors;
193 unsigned long tx_heartbeat_errors;
194 unsigned long tx_window_errors;
195 unsigned long rx_compressed;
196 unsigned long tx_compressed;
199 /* per-cpu stats, allocated on demand.
200 * Try to fit them in a single cache line, for dev_get_stats() sake.
202 struct net_device_core_stats {
205 local_t rx_nohandler;
206 local_t rx_otherhost_dropped;
207 } __aligned(4 * sizeof(local_t));
209 #include <linux/cache.h>
210 #include <linux/skbuff.h>
213 #include <linux/static_key.h>
214 extern struct static_key_false rps_needed;
215 extern struct static_key_false rfs_needed;
222 struct netdev_hw_addr {
223 struct list_head list;
225 unsigned char addr[MAX_ADDR_LEN];
227 #define NETDEV_HW_ADDR_T_LAN 1
228 #define NETDEV_HW_ADDR_T_SAN 2
229 #define NETDEV_HW_ADDR_T_UNICAST 3
230 #define NETDEV_HW_ADDR_T_MULTICAST 4
235 struct rcu_head rcu_head;
238 struct netdev_hw_addr_list {
239 struct list_head list;
242 /* Auxiliary tree for faster lookup on addition and deletion */
246 #define netdev_hw_addr_list_count(l) ((l)->count)
247 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
248 #define netdev_hw_addr_list_for_each(ha, l) \
249 list_for_each_entry(ha, &(l)->list, list)
251 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
252 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
253 #define netdev_for_each_uc_addr(ha, dev) \
254 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
256 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
257 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
258 #define netdev_for_each_mc_addr(ha, dev) \
259 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
265 /* cached hardware header; allow for machine alignment needs. */
266 #define HH_DATA_MOD 16
267 #define HH_DATA_OFF(__len) \
268 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
269 #define HH_DATA_ALIGN(__len) \
270 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
271 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
274 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
276 * dev->hard_header_len ? (dev->hard_header_len +
277 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
279 * We could use other alignment values, but we must maintain the
280 * relationship HH alignment <= LL alignment.
282 #define LL_RESERVED_SPACE(dev) \
283 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
284 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
285 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
288 int (*create) (struct sk_buff *skb, struct net_device *dev,
289 unsigned short type, const void *daddr,
290 const void *saddr, unsigned int len);
291 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
292 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
293 void (*cache_update)(struct hh_cache *hh,
294 const struct net_device *dev,
295 const unsigned char *haddr);
296 bool (*validate)(const char *ll_header, unsigned int len);
297 __be16 (*parse_protocol)(const struct sk_buff *skb);
300 /* These flag bits are private to the generic network queueing
301 * layer; they may not be explicitly referenced by any other
305 enum netdev_state_t {
307 __LINK_STATE_PRESENT,
308 __LINK_STATE_NOCARRIER,
309 __LINK_STATE_LINKWATCH_PENDING,
310 __LINK_STATE_DORMANT,
311 __LINK_STATE_TESTING,
315 struct list_head list;
320 * size of gro hash buckets, must less than bit number of
321 * napi_struct::gro_bitmask
323 #define GRO_HASH_BUCKETS 8
326 * Structure for NAPI scheduling similar to tasklet but with weighting
329 /* The poll_list must only be managed by the entity which
330 * changes the state of the NAPI_STATE_SCHED bit. This means
331 * whoever atomically sets that bit can add this napi_struct
332 * to the per-CPU poll_list, and whoever clears that bit
333 * can remove from the list right before clearing the bit.
335 struct list_head poll_list;
339 int defer_hard_irqs_count;
340 unsigned long gro_bitmask;
341 int (*poll)(struct napi_struct *, int);
342 #ifdef CONFIG_NETPOLL
345 struct net_device *dev;
346 struct gro_list gro_hash[GRO_HASH_BUCKETS];
348 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
349 int rx_count; /* length of rx_list */
350 struct hrtimer timer;
351 struct list_head dev_list;
352 struct hlist_node napi_hash_node;
353 unsigned int napi_id;
354 struct task_struct *thread;
358 NAPI_STATE_SCHED, /* Poll is scheduled */
359 NAPI_STATE_MISSED, /* reschedule a napi */
360 NAPI_STATE_DISABLE, /* Disable pending */
361 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
362 NAPI_STATE_LISTED, /* NAPI added to system lists */
363 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
364 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
365 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
366 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
367 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
371 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
372 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
373 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
374 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
375 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
376 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
377 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
378 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
379 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
380 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
390 typedef enum gro_result gro_result_t;
393 * enum rx_handler_result - Possible return values for rx_handlers.
394 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
396 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
397 * case skb->dev was changed by rx_handler.
398 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
399 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
401 * rx_handlers are functions called from inside __netif_receive_skb(), to do
402 * special processing of the skb, prior to delivery to protocol handlers.
404 * Currently, a net_device can only have a single rx_handler registered. Trying
405 * to register a second rx_handler will return -EBUSY.
407 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
408 * To unregister a rx_handler on a net_device, use
409 * netdev_rx_handler_unregister().
411 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
414 * If the rx_handler consumed the skb in some way, it should return
415 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
416 * the skb to be delivered in some other way.
418 * If the rx_handler changed skb->dev, to divert the skb to another
419 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
420 * new device will be called if it exists.
422 * If the rx_handler decides the skb should be ignored, it should return
423 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
424 * are registered on exact device (ptype->dev == skb->dev).
426 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
427 * delivered, it should return RX_HANDLER_PASS.
429 * A device without a registered rx_handler will behave as if rx_handler
430 * returned RX_HANDLER_PASS.
433 enum rx_handler_result {
439 typedef enum rx_handler_result rx_handler_result_t;
440 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
442 void __napi_schedule(struct napi_struct *n);
443 void __napi_schedule_irqoff(struct napi_struct *n);
445 static inline bool napi_disable_pending(struct napi_struct *n)
447 return test_bit(NAPI_STATE_DISABLE, &n->state);
450 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
452 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
455 bool napi_schedule_prep(struct napi_struct *n);
458 * napi_schedule - schedule NAPI poll
461 * Schedule NAPI poll routine to be called if it is not already
464 static inline void napi_schedule(struct napi_struct *n)
466 if (napi_schedule_prep(n))
471 * napi_schedule_irqoff - schedule NAPI poll
474 * Variant of napi_schedule(), assuming hard irqs are masked.
476 static inline void napi_schedule_irqoff(struct napi_struct *n)
478 if (napi_schedule_prep(n))
479 __napi_schedule_irqoff(n);
482 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
483 static inline bool napi_reschedule(struct napi_struct *napi)
485 if (napi_schedule_prep(napi)) {
486 __napi_schedule(napi);
492 bool napi_complete_done(struct napi_struct *n, int work_done);
494 * napi_complete - NAPI processing complete
497 * Mark NAPI processing as complete.
498 * Consider using napi_complete_done() instead.
499 * Return false if device should avoid rearming interrupts.
501 static inline bool napi_complete(struct napi_struct *n)
503 return napi_complete_done(n, 0);
506 int dev_set_threaded(struct net_device *dev, bool threaded);
509 * napi_disable - prevent NAPI from scheduling
512 * Stop NAPI from being scheduled on this context.
513 * Waits till any outstanding processing completes.
515 void napi_disable(struct napi_struct *n);
517 void napi_enable(struct napi_struct *n);
520 * napi_synchronize - wait until NAPI is not running
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
527 static inline void napi_synchronize(const struct napi_struct *n)
529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
537 * napi_if_scheduled_mark_missed - if napi is running, set the
541 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
544 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
546 unsigned long val, new;
549 val = READ_ONCE(n->state);
550 if (val & NAPIF_STATE_DISABLE)
553 if (!(val & NAPIF_STATE_SCHED))
556 new = val | NAPIF_STATE_MISSED;
557 } while (cmpxchg(&n->state, val, new) != val);
562 enum netdev_queue_state_t {
563 __QUEUE_STATE_DRV_XOFF,
564 __QUEUE_STATE_STACK_XOFF,
565 __QUEUE_STATE_FROZEN,
568 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
569 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
570 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
572 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
573 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
575 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
579 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
580 * netif_tx_* functions below are used to manipulate this flag. The
581 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
582 * queue independently. The netif_xmit_*stopped functions below are called
583 * to check if the queue has been stopped by the driver or stack (either
584 * of the XOFF bits are set in the state). Drivers should not need to call
585 * netif_xmit*stopped functions, they should only be using netif_tx_*.
588 struct netdev_queue {
592 struct net_device *dev;
593 netdevice_tracker dev_tracker;
595 struct Qdisc __rcu *qdisc;
596 struct Qdisc *qdisc_sleeping;
600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
603 unsigned long tx_maxrate;
605 * Number of TX timeouts for this queue
606 * (/sys/class/net/DEV/Q/trans_timeout)
608 atomic_long_t trans_timeout;
610 /* Subordinate device that the queue has been assigned to */
611 struct net_device *sb_dev;
612 #ifdef CONFIG_XDP_SOCKETS
613 struct xsk_buff_pool *pool;
618 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
621 * Time (in jiffies) of last Tx
623 unsigned long trans_start;
630 } ____cacheline_aligned_in_smp;
632 extern int sysctl_fb_tunnels_only_for_init_net;
633 extern int sysctl_devconf_inherit_init_net;
636 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
637 * == 1 : For initns only
640 static inline bool net_has_fallback_tunnels(const struct net *net)
642 return !IS_ENABLED(CONFIG_SYSCTL) ||
643 !sysctl_fb_tunnels_only_for_init_net ||
644 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
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 {
743 struct xdp_rxq_info xdp_rxq;
745 struct rps_map __rcu *rps_map;
746 struct rps_dev_flow_table __rcu *rps_flow_table;
749 struct net_device *dev;
750 netdevice_tracker dev_tracker;
752 #ifdef CONFIG_XDP_SOCKETS
753 struct xsk_buff_pool *pool;
755 } ____cacheline_aligned_in_smp;
758 * RX queue sysfs structures and functions.
760 struct rx_queue_attribute {
761 struct attribute attr;
762 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
763 ssize_t (*store)(struct netdev_rx_queue *queue,
764 const char *buf, size_t len);
767 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
776 * This structure holds an XPS map which can be of variable length. The
777 * map is an array of queues.
781 unsigned int alloc_len;
785 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
786 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
787 - sizeof(struct xps_map)) / sizeof(u16))
790 * This structure holds all XPS maps for device. Maps are indexed by CPU.
792 * We keep track of the number of cpus/rxqs used when the struct is allocated,
793 * in nr_ids. This will help not accessing out-of-bound memory.
795 * We keep track of the number of traffic classes used when the struct is
796 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
797 * not crossing its upper bound, as the original dev->num_tc can be updated in
800 struct xps_dev_maps {
804 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
807 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
808 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
810 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
811 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
813 #endif /* CONFIG_XPS */
815 #define TC_MAX_QUEUE 16
816 #define TC_BITMASK 15
817 /* HW offloaded queuing disciplines txq count and offset maps */
818 struct netdev_tc_txq {
823 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
825 * This structure is to hold information about the device
826 * configured to run FCoE protocol stack.
828 struct netdev_fcoe_hbainfo {
829 char manufacturer[64];
830 char serial_number[64];
831 char hardware_version[64];
832 char driver_version[64];
833 char optionrom_version[64];
834 char firmware_version[64];
836 char model_description[256];
840 #define MAX_PHYS_ITEM_ID_LEN 32
842 /* This structure holds a unique identifier to identify some
843 * physical item (port for example) used by a netdevice.
845 struct netdev_phys_item_id {
846 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
847 unsigned char id_len;
850 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
851 struct netdev_phys_item_id *b)
853 return a->id_len == b->id_len &&
854 memcmp(a->id, b->id, a->id_len) == 0;
857 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
859 struct net_device *sb_dev);
861 enum net_device_path_type {
862 DEV_PATH_ETHERNET = 0,
870 struct net_device_path {
871 enum net_device_path_type type;
872 const struct net_device *dev;
881 DEV_PATH_BR_VLAN_KEEP,
882 DEV_PATH_BR_VLAN_TAG,
883 DEV_PATH_BR_VLAN_UNTAG,
884 DEV_PATH_BR_VLAN_UNTAG_HW,
902 #define NET_DEVICE_PATH_STACK_MAX 5
903 #define NET_DEVICE_PATH_VLAN_MAX 2
905 struct net_device_path_stack {
907 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
910 struct net_device_path_ctx {
911 const struct net_device *dev;
918 } vlan[NET_DEVICE_PATH_VLAN_MAX];
922 TC_SETUP_QDISC_MQPRIO,
925 TC_SETUP_CLSMATCHALL,
935 TC_SETUP_QDISC_TAPRIO,
944 /* These structures hold the attributes of bpf state that are being passed
945 * to the netdevice through the bpf op.
947 enum bpf_netdev_command {
948 /* Set or clear a bpf program used in the earliest stages of packet
949 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
950 * is responsible for calling bpf_prog_put on any old progs that are
951 * stored. In case of error, the callee need not release the new prog
952 * reference, but on success it takes ownership and must bpf_prog_put
953 * when it is no longer used.
957 /* BPF program for offload callbacks, invoked at program load time. */
958 BPF_OFFLOAD_MAP_ALLOC,
959 BPF_OFFLOAD_MAP_FREE,
963 struct bpf_prog_offload_ops;
964 struct netlink_ext_ack;
966 struct xdp_dev_bulk_queue;
976 struct bpf_xdp_entity {
977 struct bpf_prog *prog;
978 struct bpf_xdp_link *link;
982 enum bpf_netdev_command command;
987 struct bpf_prog *prog;
988 struct netlink_ext_ack *extack;
990 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
992 struct bpf_offloaded_map *offmap;
994 /* XDP_SETUP_XSK_POOL */
996 struct xsk_buff_pool *pool;
1002 /* Flags for ndo_xsk_wakeup. */
1003 #define XDP_WAKEUP_RX (1 << 0)
1004 #define XDP_WAKEUP_TX (1 << 1)
1006 #ifdef CONFIG_XFRM_OFFLOAD
1007 struct xfrmdev_ops {
1008 int (*xdo_dev_state_add) (struct xfrm_state *x);
1009 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1010 void (*xdo_dev_state_free) (struct xfrm_state *x);
1011 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1012 struct xfrm_state *x);
1013 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1017 struct dev_ifalias {
1018 struct rcu_head rcuhead;
1025 struct netdev_net_notifier {
1026 struct list_head list;
1027 struct notifier_block *nb;
1031 * This structure defines the management hooks for network devices.
1032 * The following hooks can be defined; unless noted otherwise, they are
1033 * optional and can be filled with a null pointer.
1035 * int (*ndo_init)(struct net_device *dev);
1036 * This function is called once when a network device is registered.
1037 * The network device can use this for any late stage initialization
1038 * or semantic validation. It can fail with an error code which will
1039 * be propagated back to register_netdev.
1041 * void (*ndo_uninit)(struct net_device *dev);
1042 * This function is called when device is unregistered or when registration
1043 * fails. It is not called if init fails.
1045 * int (*ndo_open)(struct net_device *dev);
1046 * This function is called when a network device transitions to the up
1049 * int (*ndo_stop)(struct net_device *dev);
1050 * This function is called when a network device transitions to the down
1053 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1054 * struct net_device *dev);
1055 * Called when a packet needs to be transmitted.
1056 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1057 * the queue before that can happen; it's for obsolete devices and weird
1058 * corner cases, but the stack really does a non-trivial amount
1059 * of useless work if you return NETDEV_TX_BUSY.
1060 * Required; cannot be NULL.
1062 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1063 * struct net_device *dev
1064 * netdev_features_t features);
1065 * Called by core transmit path to determine if device is capable of
1066 * performing offload operations on a given packet. This is to give
1067 * the device an opportunity to implement any restrictions that cannot
1068 * be otherwise expressed by feature flags. The check is called with
1069 * the set of features that the stack has calculated and it returns
1070 * those the driver believes to be appropriate.
1072 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1073 * struct net_device *sb_dev);
1074 * Called to decide which queue to use when device supports multiple
1077 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1078 * This function is called to allow device receiver to make
1079 * changes to configuration when multicast or promiscuous is enabled.
1081 * void (*ndo_set_rx_mode)(struct net_device *dev);
1082 * This function is called device changes address list filtering.
1083 * If driver handles unicast address filtering, it should set
1084 * IFF_UNICAST_FLT in its priv_flags.
1086 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1087 * This function is called when the Media Access Control address
1088 * needs to be changed. If this interface is not defined, the
1089 * MAC address can not be changed.
1091 * int (*ndo_validate_addr)(struct net_device *dev);
1092 * Test if Media Access Control address is valid for the device.
1094 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1095 * Old-style ioctl entry point. This is used internally by the
1096 * appletalk and ieee802154 subsystems but is no longer called by
1097 * the device ioctl handler.
1099 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1100 * Used by the bonding driver for its device specific ioctls:
1101 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1102 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1104 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1105 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1106 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1108 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1109 * Used to set network devices bus interface parameters. This interface
1110 * is retained for legacy reasons; new devices should use the bus
1111 * interface (PCI) for low level management.
1113 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1114 * Called when a user wants to change the Maximum Transfer Unit
1117 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1118 * Callback used when the transmitter has not made any progress
1119 * for dev->watchdog ticks.
1121 * void (*ndo_get_stats64)(struct net_device *dev,
1122 * struct rtnl_link_stats64 *storage);
1123 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1124 * Called when a user wants to get the network device usage
1125 * statistics. Drivers must do one of the following:
1126 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1127 * rtnl_link_stats64 structure passed by the caller.
1128 * 2. Define @ndo_get_stats to update a net_device_stats structure
1129 * (which should normally be dev->stats) and return a pointer to
1130 * it. The structure may be changed asynchronously only if each
1131 * field is written atomically.
1132 * 3. Update dev->stats asynchronously and atomically, and define
1133 * neither operation.
1135 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1136 * Return true if this device supports offload stats of this attr_id.
1138 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1140 * Get statistics for offload operations by attr_id. Write it into the
1141 * attr_data pointer.
1143 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1144 * If device supports VLAN filtering this function is called when a
1145 * VLAN id is registered.
1147 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1148 * If device supports VLAN filtering this function is called when a
1149 * VLAN id is unregistered.
1151 * void (*ndo_poll_controller)(struct net_device *dev);
1153 * SR-IOV management functions.
1154 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1155 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1156 * u8 qos, __be16 proto);
1157 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1159 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1160 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1161 * int (*ndo_get_vf_config)(struct net_device *dev,
1162 * int vf, struct ifla_vf_info *ivf);
1163 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1164 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1165 * struct nlattr *port[]);
1167 * Enable or disable the VF ability to query its RSS Redirection Table and
1168 * Hash Key. This is needed since on some devices VF share this information
1169 * with PF and querying it may introduce a theoretical security risk.
1170 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1171 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1172 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1174 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1175 * This is always called from the stack with the rtnl lock held and netif
1176 * tx queues stopped. This allows the netdevice to perform queue
1177 * management safely.
1179 * Fiber Channel over Ethernet (FCoE) offload functions.
1180 * int (*ndo_fcoe_enable)(struct net_device *dev);
1181 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1182 * so the underlying device can perform whatever needed configuration or
1183 * initialization to support acceleration of FCoE traffic.
1185 * int (*ndo_fcoe_disable)(struct net_device *dev);
1186 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1187 * so the underlying device can perform whatever needed clean-ups to
1188 * stop supporting acceleration of FCoE traffic.
1190 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1191 * struct scatterlist *sgl, unsigned int sgc);
1192 * Called when the FCoE Initiator wants to initialize an I/O that
1193 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1194 * perform necessary setup and returns 1 to indicate the device is set up
1195 * successfully to perform DDP on this I/O, otherwise this returns 0.
1197 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1198 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1199 * indicated by the FC exchange id 'xid', so the underlying device can
1200 * clean up and reuse resources for later DDP requests.
1202 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1203 * struct scatterlist *sgl, unsigned int sgc);
1204 * Called when the FCoE Target wants to initialize an I/O that
1205 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1206 * perform necessary setup and returns 1 to indicate the device is set up
1207 * successfully to perform DDP on this I/O, otherwise this returns 0.
1209 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1210 * struct netdev_fcoe_hbainfo *hbainfo);
1211 * Called when the FCoE Protocol stack wants information on the underlying
1212 * device. This information is utilized by the FCoE protocol stack to
1213 * register attributes with Fiber Channel management service as per the
1214 * FC-GS Fabric Device Management Information(FDMI) specification.
1216 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1217 * Called when the underlying device wants to override default World Wide
1218 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1219 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1220 * protocol stack to use.
1223 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1224 * u16 rxq_index, u32 flow_id);
1225 * Set hardware filter for RFS. rxq_index is the target queue index;
1226 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1227 * Return the filter ID on success, or a negative error code.
1229 * Slave management functions (for bridge, bonding, etc).
1230 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1231 * Called to make another netdev an underling.
1233 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1234 * Called to release previously enslaved netdev.
1236 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1237 * struct sk_buff *skb,
1239 * Get the xmit slave of master device. If all_slaves is true, function
1240 * assume all the slaves can transmit.
1242 * Feature/offload setting functions.
1243 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1244 * netdev_features_t features);
1245 * Adjusts the requested feature flags according to device-specific
1246 * constraints, and returns the resulting flags. Must not modify
1249 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1250 * Called to update device configuration to new features. Passed
1251 * feature set might be less than what was returned by ndo_fix_features()).
1252 * Must return >0 or -errno if it changed dev->features itself.
1254 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1255 * struct net_device *dev,
1256 * const unsigned char *addr, u16 vid, u16 flags,
1257 * struct netlink_ext_ack *extack);
1258 * Adds an FDB entry to dev for addr.
1259 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1260 * struct net_device *dev,
1261 * const unsigned char *addr, u16 vid)
1262 * Deletes the FDB entry from dev coresponding to addr.
1263 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1264 * struct net_device *dev, struct net_device *filter_dev,
1266 * Used to add FDB entries to dump requests. Implementers should add
1267 * entries to skb and update idx with the number of entries.
1269 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1270 * u16 flags, struct netlink_ext_ack *extack)
1271 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1272 * struct net_device *dev, u32 filter_mask,
1274 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1277 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1278 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1279 * which do not represent real hardware may define this to allow their
1280 * userspace components to manage their virtual carrier state. Devices
1281 * that determine carrier state from physical hardware properties (eg
1282 * network cables) or protocol-dependent mechanisms (eg
1283 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1285 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1286 * struct netdev_phys_item_id *ppid);
1287 * Called to get ID of physical port of this device. If driver does
1288 * not implement this, it is assumed that the hw is not able to have
1289 * multiple net devices on single physical port.
1291 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1292 * struct netdev_phys_item_id *ppid)
1293 * Called to get the parent ID of the physical port of this device.
1295 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1296 * struct net_device *dev)
1297 * Called by upper layer devices to accelerate switching or other
1298 * station functionality into hardware. 'pdev is the lowerdev
1299 * to use for the offload and 'dev' is the net device that will
1300 * back the offload. Returns a pointer to the private structure
1301 * the upper layer will maintain.
1302 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1303 * Called by upper layer device to delete the station created
1304 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1305 * the station and priv is the structure returned by the add
1307 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1308 * int queue_index, u32 maxrate);
1309 * Called when a user wants to set a max-rate limitation of specific
1311 * int (*ndo_get_iflink)(const struct net_device *dev);
1312 * Called to get the iflink value of this device.
1313 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1314 * This function is used to get egress tunnel information for given skb.
1315 * This is useful for retrieving outer tunnel header parameters while
1317 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1318 * This function is used to specify the headroom that the skb must
1319 * consider when allocation skb during packet reception. Setting
1320 * appropriate rx headroom value allows avoiding skb head copy on
1321 * forward. Setting a negative value resets the rx headroom to the
1323 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1324 * This function is used to set or query state related to XDP on the
1325 * netdevice and manage BPF offload. See definition of
1326 * enum bpf_netdev_command for details.
1327 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1329 * This function is used to submit @n XDP packets for transmit on a
1330 * netdevice. Returns number of frames successfully transmitted, frames
1331 * that got dropped are freed/returned via xdp_return_frame().
1332 * Returns negative number, means general error invoking ndo, meaning
1333 * no frames were xmit'ed and core-caller will free all frames.
1334 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1335 * struct xdp_buff *xdp);
1336 * Get the xmit slave of master device based on the xdp_buff.
1337 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1338 * This function is used to wake up the softirq, ksoftirqd or kthread
1339 * responsible for sending and/or receiving packets on a specific
1340 * queue id bound to an AF_XDP socket. The flags field specifies if
1341 * only RX, only Tx, or both should be woken up using the flags
1342 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1343 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1344 * Get devlink port instance associated with a given netdev.
1345 * Called with a reference on the netdevice and devlink locks only,
1346 * rtnl_lock is not held.
1347 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1349 * Add, change, delete or get information on an IPv4 tunnel.
1350 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1351 * If a device is paired with a peer device, return the peer instance.
1352 * The caller must be under RCU read context.
1353 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1354 * Get the forwarding path to reach the real device from the HW destination address
1356 struct net_device_ops {
1357 int (*ndo_init)(struct net_device *dev);
1358 void (*ndo_uninit)(struct net_device *dev);
1359 int (*ndo_open)(struct net_device *dev);
1360 int (*ndo_stop)(struct net_device *dev);
1361 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1362 struct net_device *dev);
1363 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1364 struct net_device *dev,
1365 netdev_features_t features);
1366 u16 (*ndo_select_queue)(struct net_device *dev,
1367 struct sk_buff *skb,
1368 struct net_device *sb_dev);
1369 void (*ndo_change_rx_flags)(struct net_device *dev,
1371 void (*ndo_set_rx_mode)(struct net_device *dev);
1372 int (*ndo_set_mac_address)(struct net_device *dev,
1374 int (*ndo_validate_addr)(struct net_device *dev);
1375 int (*ndo_do_ioctl)(struct net_device *dev,
1376 struct ifreq *ifr, int cmd);
1377 int (*ndo_eth_ioctl)(struct net_device *dev,
1378 struct ifreq *ifr, int cmd);
1379 int (*ndo_siocbond)(struct net_device *dev,
1380 struct ifreq *ifr, int cmd);
1381 int (*ndo_siocwandev)(struct net_device *dev,
1382 struct if_settings *ifs);
1383 int (*ndo_siocdevprivate)(struct net_device *dev,
1385 void __user *data, int cmd);
1386 int (*ndo_set_config)(struct net_device *dev,
1388 int (*ndo_change_mtu)(struct net_device *dev,
1390 int (*ndo_neigh_setup)(struct net_device *dev,
1391 struct neigh_parms *);
1392 void (*ndo_tx_timeout) (struct net_device *dev,
1393 unsigned int txqueue);
1395 void (*ndo_get_stats64)(struct net_device *dev,
1396 struct rtnl_link_stats64 *storage);
1397 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1398 int (*ndo_get_offload_stats)(int attr_id,
1399 const struct net_device *dev,
1401 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1403 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1404 __be16 proto, u16 vid);
1405 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1406 __be16 proto, u16 vid);
1407 #ifdef CONFIG_NET_POLL_CONTROLLER
1408 void (*ndo_poll_controller)(struct net_device *dev);
1409 int (*ndo_netpoll_setup)(struct net_device *dev,
1410 struct netpoll_info *info);
1411 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1413 int (*ndo_set_vf_mac)(struct net_device *dev,
1414 int queue, u8 *mac);
1415 int (*ndo_set_vf_vlan)(struct net_device *dev,
1416 int queue, u16 vlan,
1417 u8 qos, __be16 proto);
1418 int (*ndo_set_vf_rate)(struct net_device *dev,
1419 int vf, int min_tx_rate,
1421 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1422 int vf, bool setting);
1423 int (*ndo_set_vf_trust)(struct net_device *dev,
1424 int vf, bool setting);
1425 int (*ndo_get_vf_config)(struct net_device *dev,
1427 struct ifla_vf_info *ivf);
1428 int (*ndo_set_vf_link_state)(struct net_device *dev,
1429 int vf, int link_state);
1430 int (*ndo_get_vf_stats)(struct net_device *dev,
1432 struct ifla_vf_stats
1434 int (*ndo_set_vf_port)(struct net_device *dev,
1436 struct nlattr *port[]);
1437 int (*ndo_get_vf_port)(struct net_device *dev,
1438 int vf, struct sk_buff *skb);
1439 int (*ndo_get_vf_guid)(struct net_device *dev,
1441 struct ifla_vf_guid *node_guid,
1442 struct ifla_vf_guid *port_guid);
1443 int (*ndo_set_vf_guid)(struct net_device *dev,
1446 int (*ndo_set_vf_rss_query_en)(
1447 struct net_device *dev,
1448 int vf, bool setting);
1449 int (*ndo_setup_tc)(struct net_device *dev,
1450 enum tc_setup_type type,
1452 #if IS_ENABLED(CONFIG_FCOE)
1453 int (*ndo_fcoe_enable)(struct net_device *dev);
1454 int (*ndo_fcoe_disable)(struct net_device *dev);
1455 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1457 struct scatterlist *sgl,
1459 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1461 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1463 struct scatterlist *sgl,
1465 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1466 struct netdev_fcoe_hbainfo *hbainfo);
1469 #if IS_ENABLED(CONFIG_LIBFCOE)
1470 #define NETDEV_FCOE_WWNN 0
1471 #define NETDEV_FCOE_WWPN 1
1472 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1473 u64 *wwn, int type);
1476 #ifdef CONFIG_RFS_ACCEL
1477 int (*ndo_rx_flow_steer)(struct net_device *dev,
1478 const struct sk_buff *skb,
1482 int (*ndo_add_slave)(struct net_device *dev,
1483 struct net_device *slave_dev,
1484 struct netlink_ext_ack *extack);
1485 int (*ndo_del_slave)(struct net_device *dev,
1486 struct net_device *slave_dev);
1487 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1488 struct sk_buff *skb,
1490 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1492 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1493 netdev_features_t features);
1494 int (*ndo_set_features)(struct net_device *dev,
1495 netdev_features_t features);
1496 int (*ndo_neigh_construct)(struct net_device *dev,
1497 struct neighbour *n);
1498 void (*ndo_neigh_destroy)(struct net_device *dev,
1499 struct neighbour *n);
1501 int (*ndo_fdb_add)(struct ndmsg *ndm,
1502 struct nlattr *tb[],
1503 struct net_device *dev,
1504 const unsigned char *addr,
1507 struct netlink_ext_ack *extack);
1508 int (*ndo_fdb_del)(struct ndmsg *ndm,
1509 struct nlattr *tb[],
1510 struct net_device *dev,
1511 const unsigned char *addr,
1513 int (*ndo_fdb_dump)(struct sk_buff *skb,
1514 struct netlink_callback *cb,
1515 struct net_device *dev,
1516 struct net_device *filter_dev,
1518 int (*ndo_fdb_get)(struct sk_buff *skb,
1519 struct nlattr *tb[],
1520 struct net_device *dev,
1521 const unsigned char *addr,
1522 u16 vid, u32 portid, u32 seq,
1523 struct netlink_ext_ack *extack);
1524 int (*ndo_bridge_setlink)(struct net_device *dev,
1525 struct nlmsghdr *nlh,
1527 struct netlink_ext_ack *extack);
1528 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1530 struct net_device *dev,
1533 int (*ndo_bridge_dellink)(struct net_device *dev,
1534 struct nlmsghdr *nlh,
1536 int (*ndo_change_carrier)(struct net_device *dev,
1538 int (*ndo_get_phys_port_id)(struct net_device *dev,
1539 struct netdev_phys_item_id *ppid);
1540 int (*ndo_get_port_parent_id)(struct net_device *dev,
1541 struct netdev_phys_item_id *ppid);
1542 int (*ndo_get_phys_port_name)(struct net_device *dev,
1543 char *name, size_t len);
1544 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1545 struct net_device *dev);
1546 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1549 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1552 int (*ndo_get_iflink)(const struct net_device *dev);
1553 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1554 struct sk_buff *skb);
1555 void (*ndo_set_rx_headroom)(struct net_device *dev,
1556 int needed_headroom);
1557 int (*ndo_bpf)(struct net_device *dev,
1558 struct netdev_bpf *bpf);
1559 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1560 struct xdp_frame **xdp,
1562 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1563 struct xdp_buff *xdp);
1564 int (*ndo_xsk_wakeup)(struct net_device *dev,
1565 u32 queue_id, u32 flags);
1566 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1567 int (*ndo_tunnel_ctl)(struct net_device *dev,
1568 struct ip_tunnel_parm *p, int cmd);
1569 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1570 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1571 struct net_device_path *path);
1575 * enum netdev_priv_flags - &struct net_device priv_flags
1577 * These are the &struct net_device, they are only set internally
1578 * by drivers and used in the kernel. These flags are invisible to
1579 * userspace; this means that the order of these flags can change
1580 * during any kernel release.
1582 * You should have a pretty good reason to be extending these flags.
1584 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1585 * @IFF_EBRIDGE: Ethernet bridging device
1586 * @IFF_BONDING: bonding master or slave
1587 * @IFF_ISATAP: ISATAP interface (RFC4214)
1588 * @IFF_WAN_HDLC: WAN HDLC device
1589 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1591 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1592 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1593 * @IFF_MACVLAN_PORT: device used as macvlan port
1594 * @IFF_BRIDGE_PORT: device used as bridge port
1595 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1596 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1597 * @IFF_UNICAST_FLT: Supports unicast filtering
1598 * @IFF_TEAM_PORT: device used as team port
1599 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1600 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1601 * change when it's running
1602 * @IFF_MACVLAN: Macvlan device
1603 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1604 * underlying stacked devices
1605 * @IFF_L3MDEV_MASTER: device is an L3 master device
1606 * @IFF_NO_QUEUE: device can run without qdisc attached
1607 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1608 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1609 * @IFF_TEAM: device is a team device
1610 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1611 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1612 * entity (i.e. the master device for bridged veth)
1613 * @IFF_MACSEC: device is a MACsec device
1614 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1615 * @IFF_FAILOVER: device is a failover master device
1616 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1617 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1618 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1619 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1620 * skb_headlen(skb) == 0 (data starts from frag0)
1621 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1623 enum netdev_priv_flags {
1624 IFF_802_1Q_VLAN = 1<<0,
1628 IFF_WAN_HDLC = 1<<4,
1629 IFF_XMIT_DST_RELEASE = 1<<5,
1630 IFF_DONT_BRIDGE = 1<<6,
1631 IFF_DISABLE_NETPOLL = 1<<7,
1632 IFF_MACVLAN_PORT = 1<<8,
1633 IFF_BRIDGE_PORT = 1<<9,
1634 IFF_OVS_DATAPATH = 1<<10,
1635 IFF_TX_SKB_SHARING = 1<<11,
1636 IFF_UNICAST_FLT = 1<<12,
1637 IFF_TEAM_PORT = 1<<13,
1638 IFF_SUPP_NOFCS = 1<<14,
1639 IFF_LIVE_ADDR_CHANGE = 1<<15,
1640 IFF_MACVLAN = 1<<16,
1641 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1642 IFF_L3MDEV_MASTER = 1<<18,
1643 IFF_NO_QUEUE = 1<<19,
1644 IFF_OPENVSWITCH = 1<<20,
1645 IFF_L3MDEV_SLAVE = 1<<21,
1647 IFF_RXFH_CONFIGURED = 1<<23,
1648 IFF_PHONY_HEADROOM = 1<<24,
1650 IFF_NO_RX_HANDLER = 1<<26,
1651 IFF_FAILOVER = 1<<27,
1652 IFF_FAILOVER_SLAVE = 1<<28,
1653 IFF_L3MDEV_RX_HANDLER = 1<<29,
1654 IFF_LIVE_RENAME_OK = 1<<30,
1655 IFF_TX_SKB_NO_LINEAR = 1<<31,
1656 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1659 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1660 #define IFF_EBRIDGE IFF_EBRIDGE
1661 #define IFF_BONDING IFF_BONDING
1662 #define IFF_ISATAP IFF_ISATAP
1663 #define IFF_WAN_HDLC IFF_WAN_HDLC
1664 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1665 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1666 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1667 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1668 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1669 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1670 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1671 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1672 #define IFF_TEAM_PORT IFF_TEAM_PORT
1673 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1674 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1675 #define IFF_MACVLAN IFF_MACVLAN
1676 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1677 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1678 #define IFF_NO_QUEUE IFF_NO_QUEUE
1679 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1680 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1681 #define IFF_TEAM IFF_TEAM
1682 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1683 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1684 #define IFF_MACSEC IFF_MACSEC
1685 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1686 #define IFF_FAILOVER IFF_FAILOVER
1687 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1688 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1689 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1690 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1692 /* Specifies the type of the struct net_device::ml_priv pointer */
1693 enum netdev_ml_priv_type {
1699 * struct net_device - The DEVICE structure.
1701 * Actually, this whole structure is a big mistake. It mixes I/O
1702 * data with strictly "high-level" data, and it has to know about
1703 * almost every data structure used in the INET module.
1705 * @name: This is the first field of the "visible" part of this structure
1706 * (i.e. as seen by users in the "Space.c" file). It is the name
1709 * @name_node: Name hashlist node
1710 * @ifalias: SNMP alias
1711 * @mem_end: Shared memory end
1712 * @mem_start: Shared memory start
1713 * @base_addr: Device I/O address
1714 * @irq: Device IRQ number
1716 * @state: Generic network queuing layer state, see netdev_state_t
1717 * @dev_list: The global list of network devices
1718 * @napi_list: List entry used for polling NAPI devices
1719 * @unreg_list: List entry when we are unregistering the
1720 * device; see the function unregister_netdev
1721 * @close_list: List entry used when we are closing the device
1722 * @ptype_all: Device-specific packet handlers for all protocols
1723 * @ptype_specific: Device-specific, protocol-specific packet handlers
1725 * @adj_list: Directly linked devices, like slaves for bonding
1726 * @features: Currently active device features
1727 * @hw_features: User-changeable features
1729 * @wanted_features: User-requested features
1730 * @vlan_features: Mask of features inheritable by VLAN devices
1732 * @hw_enc_features: Mask of features inherited by encapsulating devices
1733 * This field indicates what encapsulation
1734 * offloads the hardware is capable of doing,
1735 * and drivers will need to set them appropriately.
1737 * @mpls_features: Mask of features inheritable by MPLS
1738 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1740 * @ifindex: interface index
1741 * @group: The group the device belongs to
1743 * @stats: Statistics struct, which was left as a legacy, use
1744 * rtnl_link_stats64 instead
1746 * @core_stats: core networking counters,
1747 * do not use this in drivers
1748 * @carrier_up_count: Number of times the carrier has been up
1749 * @carrier_down_count: Number of times the carrier has been down
1751 * @wireless_handlers: List of functions to handle Wireless Extensions,
1753 * see <net/iw_handler.h> for details.
1754 * @wireless_data: Instance data managed by the core of wireless extensions
1756 * @netdev_ops: Includes several pointers to callbacks,
1757 * if one wants to override the ndo_*() functions
1758 * @ethtool_ops: Management operations
1759 * @l3mdev_ops: Layer 3 master device operations
1760 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1761 * discovery handling. Necessary for e.g. 6LoWPAN.
1762 * @xfrmdev_ops: Transformation offload operations
1763 * @tlsdev_ops: Transport Layer Security offload operations
1764 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1765 * of Layer 2 headers.
1767 * @flags: Interface flags (a la BSD)
1768 * @priv_flags: Like 'flags' but invisible to userspace,
1769 * see if.h for the definitions
1770 * @gflags: Global flags ( kept as legacy )
1771 * @padded: How much padding added by alloc_netdev()
1772 * @operstate: RFC2863 operstate
1773 * @link_mode: Mapping policy to operstate
1774 * @if_port: Selectable AUI, TP, ...
1776 * @mtu: Interface MTU value
1777 * @min_mtu: Interface Minimum MTU value
1778 * @max_mtu: Interface Maximum MTU value
1779 * @type: Interface hardware type
1780 * @hard_header_len: Maximum hardware header length.
1781 * @min_header_len: Minimum hardware header length
1783 * @needed_headroom: Extra headroom the hardware may need, but not in all
1784 * cases can this be guaranteed
1785 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1786 * cases can this be guaranteed. Some cases also use
1787 * LL_MAX_HEADER instead to allocate the skb
1789 * interface address info:
1791 * @perm_addr: Permanent hw address
1792 * @addr_assign_type: Hw address assignment type
1793 * @addr_len: Hardware address length
1794 * @upper_level: Maximum depth level of upper devices.
1795 * @lower_level: Maximum depth level of lower devices.
1796 * @neigh_priv_len: Used in neigh_alloc()
1797 * @dev_id: Used to differentiate devices that share
1798 * the same link layer address
1799 * @dev_port: Used to differentiate devices that share
1801 * @addr_list_lock: XXX: need comments on this one
1802 * @name_assign_type: network interface name assignment type
1803 * @uc_promisc: Counter that indicates promiscuous mode
1804 * has been enabled due to the need to listen to
1805 * additional unicast addresses in a device that
1806 * does not implement ndo_set_rx_mode()
1807 * @uc: unicast mac addresses
1808 * @mc: multicast mac addresses
1809 * @dev_addrs: list of device hw addresses
1810 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1811 * @promiscuity: Number of times the NIC is told to work in
1812 * promiscuous mode; if it becomes 0 the NIC will
1813 * exit promiscuous mode
1814 * @allmulti: Counter, enables or disables allmulticast mode
1816 * @vlan_info: VLAN info
1817 * @dsa_ptr: dsa specific data
1818 * @tipc_ptr: TIPC specific data
1819 * @atalk_ptr: AppleTalk link
1820 * @ip_ptr: IPv4 specific data
1821 * @dn_ptr: DECnet specific data
1822 * @ip6_ptr: IPv6 specific data
1823 * @ax25_ptr: AX.25 specific data
1824 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1825 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1827 * @mpls_ptr: mpls_dev struct pointer
1828 * @mctp_ptr: MCTP specific data
1830 * @dev_addr: Hw address (before bcast,
1831 * because most packets are unicast)
1833 * @_rx: Array of RX queues
1834 * @num_rx_queues: Number of RX queues
1835 * allocated at register_netdev() time
1836 * @real_num_rx_queues: Number of RX queues currently active in device
1837 * @xdp_prog: XDP sockets filter program pointer
1838 * @gro_flush_timeout: timeout for GRO layer in NAPI
1839 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1840 * allow to avoid NIC hard IRQ, on busy queues.
1842 * @rx_handler: handler for received packets
1843 * @rx_handler_data: XXX: need comments on this one
1844 * @miniq_ingress: ingress/clsact qdisc specific data for
1845 * ingress processing
1846 * @ingress_queue: XXX: need comments on this one
1847 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1848 * @broadcast: hw bcast address
1850 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1851 * indexed by RX queue number. Assigned by driver.
1852 * This must only be set if the ndo_rx_flow_steer
1853 * operation is defined
1854 * @index_hlist: Device index hash chain
1856 * @_tx: Array of TX queues
1857 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1858 * @real_num_tx_queues: Number of TX queues currently active in device
1859 * @qdisc: Root qdisc from userspace point of view
1860 * @tx_queue_len: Max frames per queue allowed
1861 * @tx_global_lock: XXX: need comments on this one
1862 * @xdp_bulkq: XDP device bulk queue
1863 * @xps_maps: all CPUs/RXQs maps for XPS device
1865 * @xps_maps: XXX: need comments on this one
1866 * @miniq_egress: clsact qdisc specific data for
1868 * @nf_hooks_egress: netfilter hooks executed for egress packets
1869 * @qdisc_hash: qdisc hash table
1870 * @watchdog_timeo: Represents the timeout that is used by
1871 * the watchdog (see dev_watchdog())
1872 * @watchdog_timer: List of timers
1874 * @proto_down_reason: reason a netdev interface is held down
1875 * @pcpu_refcnt: Number of references to this device
1876 * @dev_refcnt: Number of references to this device
1877 * @refcnt_tracker: Tracker directory for tracked references to this device
1878 * @todo_list: Delayed register/unregister
1879 * @link_watch_list: XXX: need comments on this one
1881 * @reg_state: Register/unregister state machine
1882 * @dismantle: Device is going to be freed
1883 * @rtnl_link_state: This enum represents the phases of creating
1886 * @needs_free_netdev: Should unregister perform free_netdev?
1887 * @priv_destructor: Called from unregister
1888 * @npinfo: XXX: need comments on this one
1889 * @nd_net: Network namespace this network device is inside
1891 * @ml_priv: Mid-layer private
1892 * @ml_priv_type: Mid-layer private type
1893 * @lstats: Loopback statistics
1894 * @tstats: Tunnel statistics
1895 * @dstats: Dummy statistics
1896 * @vstats: Virtual ethernet statistics
1901 * @dm_private: Drop monitor private
1903 * @dev: Class/net/name entry
1904 * @sysfs_groups: Space for optional device, statistics and wireless
1907 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1908 * @rtnl_link_ops: Rtnl_link_ops
1910 * @gso_max_size: Maximum size of generic segmentation offload
1911 * @gso_max_segs: Maximum number of segments that can be passed to the
1914 * @dcbnl_ops: Data Center Bridging netlink ops
1915 * @num_tc: Number of traffic classes in the net device
1916 * @tc_to_txq: XXX: need comments on this one
1917 * @prio_tc_map: XXX: need comments on this one
1919 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1921 * @priomap: XXX: need comments on this one
1922 * @phydev: Physical device may attach itself
1923 * for hardware timestamping
1924 * @sfp_bus: attached &struct sfp_bus structure.
1926 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1928 * @proto_down: protocol port state information can be sent to the
1929 * switch driver and used to set the phys state of the
1932 * @wol_enabled: Wake-on-LAN is enabled
1934 * @threaded: napi threaded mode is enabled
1936 * @net_notifier_list: List of per-net netdev notifier block
1937 * that follow this device when it is moved
1938 * to another network namespace.
1940 * @macsec_ops: MACsec offloading ops
1942 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1943 * offload capabilities of the device
1944 * @udp_tunnel_nic: UDP tunnel offload state
1945 * @xdp_state: stores info on attached XDP BPF programs
1947 * @nested_level: Used as a parameter of spin_lock_nested() of
1948 * dev->addr_list_lock.
1949 * @unlink_list: As netif_addr_lock() can be called recursively,
1950 * keep a list of interfaces to be deleted.
1951 * @gro_max_size: Maximum size of aggregated packet in generic
1952 * receive offload (GRO)
1954 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
1955 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
1956 * @watchdog_dev_tracker: refcount tracker used by watchdog.
1957 * @dev_registered_tracker: tracker for reference held while
1959 * @offload_xstats_l3: L3 HW stats for this netdevice.
1961 * FIXME: cleanup struct net_device such that network protocol info
1966 char name[IFNAMSIZ];
1967 struct netdev_name_node *name_node;
1968 struct dev_ifalias __rcu *ifalias;
1970 * I/O specific fields
1971 * FIXME: Merge these and struct ifmap into one
1973 unsigned long mem_end;
1974 unsigned long mem_start;
1975 unsigned long base_addr;
1978 * Some hardware also needs these fields (state,dev_list,
1979 * napi_list,unreg_list,close_list) but they are not
1980 * part of the usual set specified in Space.c.
1983 unsigned long state;
1985 struct list_head dev_list;
1986 struct list_head napi_list;
1987 struct list_head unreg_list;
1988 struct list_head close_list;
1989 struct list_head ptype_all;
1990 struct list_head ptype_specific;
1993 struct list_head upper;
1994 struct list_head lower;
1997 /* Read-mostly cache-line for fast-path access */
1999 unsigned long long priv_flags;
2000 const struct net_device_ops *netdev_ops;
2002 unsigned short gflags;
2003 unsigned short hard_header_len;
2005 /* Note : dev->mtu is often read without holding a lock.
2006 * Writers usually hold RTNL.
2007 * It is recommended to use READ_ONCE() to annotate the reads,
2008 * and to use WRITE_ONCE() to annotate the writes.
2011 unsigned short needed_headroom;
2012 unsigned short needed_tailroom;
2014 netdev_features_t features;
2015 netdev_features_t hw_features;
2016 netdev_features_t wanted_features;
2017 netdev_features_t vlan_features;
2018 netdev_features_t hw_enc_features;
2019 netdev_features_t mpls_features;
2020 netdev_features_t gso_partial_features;
2022 unsigned int min_mtu;
2023 unsigned int max_mtu;
2024 unsigned short type;
2025 unsigned char min_header_len;
2026 unsigned char name_assign_type;
2030 struct net_device_stats stats; /* not used by modern drivers */
2032 struct net_device_core_stats __percpu *core_stats;
2034 /* Stats to monitor link on/off, flapping */
2035 atomic_t carrier_up_count;
2036 atomic_t carrier_down_count;
2038 #ifdef CONFIG_WIRELESS_EXT
2039 const struct iw_handler_def *wireless_handlers;
2040 struct iw_public_data *wireless_data;
2042 const struct ethtool_ops *ethtool_ops;
2043 #ifdef CONFIG_NET_L3_MASTER_DEV
2044 const struct l3mdev_ops *l3mdev_ops;
2046 #if IS_ENABLED(CONFIG_IPV6)
2047 const struct ndisc_ops *ndisc_ops;
2050 #ifdef CONFIG_XFRM_OFFLOAD
2051 const struct xfrmdev_ops *xfrmdev_ops;
2054 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2055 const struct tlsdev_ops *tlsdev_ops;
2058 const struct header_ops *header_ops;
2060 unsigned char operstate;
2061 unsigned char link_mode;
2063 unsigned char if_port;
2066 /* Interface address info. */
2067 unsigned char perm_addr[MAX_ADDR_LEN];
2068 unsigned char addr_assign_type;
2069 unsigned char addr_len;
2070 unsigned char upper_level;
2071 unsigned char lower_level;
2073 unsigned short neigh_priv_len;
2074 unsigned short dev_id;
2075 unsigned short dev_port;
2076 unsigned short padded;
2078 spinlock_t addr_list_lock;
2081 struct netdev_hw_addr_list uc;
2082 struct netdev_hw_addr_list mc;
2083 struct netdev_hw_addr_list dev_addrs;
2086 struct kset *queues_kset;
2088 #ifdef CONFIG_LOCKDEP
2089 struct list_head unlink_list;
2091 unsigned int promiscuity;
2092 unsigned int allmulti;
2094 #ifdef CONFIG_LOCKDEP
2095 unsigned char nested_level;
2099 /* Protocol-specific pointers */
2101 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2102 struct vlan_info __rcu *vlan_info;
2104 #if IS_ENABLED(CONFIG_NET_DSA)
2105 struct dsa_port *dsa_ptr;
2107 #if IS_ENABLED(CONFIG_TIPC)
2108 struct tipc_bearer __rcu *tipc_ptr;
2110 #if IS_ENABLED(CONFIG_ATALK)
2113 struct in_device __rcu *ip_ptr;
2114 #if IS_ENABLED(CONFIG_DECNET)
2115 struct dn_dev __rcu *dn_ptr;
2117 struct inet6_dev __rcu *ip6_ptr;
2118 #if IS_ENABLED(CONFIG_AX25)
2121 struct wireless_dev *ieee80211_ptr;
2122 struct wpan_dev *ieee802154_ptr;
2123 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2124 struct mpls_dev __rcu *mpls_ptr;
2126 #if IS_ENABLED(CONFIG_MCTP)
2127 struct mctp_dev __rcu *mctp_ptr;
2131 * Cache lines mostly used on receive path (including eth_type_trans())
2133 /* Interface address info used in eth_type_trans() */
2134 const unsigned char *dev_addr;
2136 struct netdev_rx_queue *_rx;
2137 unsigned int num_rx_queues;
2138 unsigned int real_num_rx_queues;
2140 struct bpf_prog __rcu *xdp_prog;
2141 unsigned long gro_flush_timeout;
2142 int napi_defer_hard_irqs;
2143 #define GRO_MAX_SIZE 65536
2144 unsigned int gro_max_size;
2145 rx_handler_func_t __rcu *rx_handler;
2146 void __rcu *rx_handler_data;
2148 #ifdef CONFIG_NET_CLS_ACT
2149 struct mini_Qdisc __rcu *miniq_ingress;
2151 struct netdev_queue __rcu *ingress_queue;
2152 #ifdef CONFIG_NETFILTER_INGRESS
2153 struct nf_hook_entries __rcu *nf_hooks_ingress;
2156 unsigned char broadcast[MAX_ADDR_LEN];
2157 #ifdef CONFIG_RFS_ACCEL
2158 struct cpu_rmap *rx_cpu_rmap;
2160 struct hlist_node index_hlist;
2163 * Cache lines mostly used on transmit path
2165 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2166 unsigned int num_tx_queues;
2167 unsigned int real_num_tx_queues;
2168 struct Qdisc __rcu *qdisc;
2169 unsigned int tx_queue_len;
2170 spinlock_t tx_global_lock;
2172 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2175 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2177 #ifdef CONFIG_NET_CLS_ACT
2178 struct mini_Qdisc __rcu *miniq_egress;
2180 #ifdef CONFIG_NETFILTER_EGRESS
2181 struct nf_hook_entries __rcu *nf_hooks_egress;
2184 #ifdef CONFIG_NET_SCHED
2185 DECLARE_HASHTABLE (qdisc_hash, 4);
2187 /* These may be needed for future network-power-down code. */
2188 struct timer_list watchdog_timer;
2191 u32 proto_down_reason;
2193 struct list_head todo_list;
2195 #ifdef CONFIG_PCPU_DEV_REFCNT
2196 int __percpu *pcpu_refcnt;
2198 refcount_t dev_refcnt;
2200 struct ref_tracker_dir refcnt_tracker;
2202 struct list_head link_watch_list;
2204 enum { NETREG_UNINITIALIZED=0,
2205 NETREG_REGISTERED, /* completed register_netdevice */
2206 NETREG_UNREGISTERING, /* called unregister_netdevice */
2207 NETREG_UNREGISTERED, /* completed unregister todo */
2208 NETREG_RELEASED, /* called free_netdev */
2209 NETREG_DUMMY, /* dummy device for NAPI poll */
2215 RTNL_LINK_INITIALIZED,
2216 RTNL_LINK_INITIALIZING,
2217 } rtnl_link_state:16;
2219 bool needs_free_netdev;
2220 void (*priv_destructor)(struct net_device *dev);
2222 #ifdef CONFIG_NETPOLL
2223 struct netpoll_info __rcu *npinfo;
2226 possible_net_t nd_net;
2228 /* mid-layer private */
2230 enum netdev_ml_priv_type ml_priv_type;
2233 struct pcpu_lstats __percpu *lstats;
2234 struct pcpu_sw_netstats __percpu *tstats;
2235 struct pcpu_dstats __percpu *dstats;
2238 #if IS_ENABLED(CONFIG_GARP)
2239 struct garp_port __rcu *garp_port;
2241 #if IS_ENABLED(CONFIG_MRP)
2242 struct mrp_port __rcu *mrp_port;
2244 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2245 struct dm_hw_stat_delta __rcu *dm_private;
2248 const struct attribute_group *sysfs_groups[4];
2249 const struct attribute_group *sysfs_rx_queue_group;
2251 const struct rtnl_link_ops *rtnl_link_ops;
2253 /* for setting kernel sock attribute on TCP connection setup */
2254 #define GSO_MAX_SIZE 65536
2255 unsigned int gso_max_size;
2256 #define GSO_MAX_SEGS 65535
2260 const struct dcbnl_rtnl_ops *dcbnl_ops;
2263 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2264 u8 prio_tc_map[TC_BITMASK + 1];
2266 #if IS_ENABLED(CONFIG_FCOE)
2267 unsigned int fcoe_ddp_xid;
2269 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2270 struct netprio_map __rcu *priomap;
2272 struct phy_device *phydev;
2273 struct sfp_bus *sfp_bus;
2274 struct lock_class_key *qdisc_tx_busylock;
2276 unsigned wol_enabled:1;
2277 unsigned threaded:1;
2279 struct list_head net_notifier_list;
2281 #if IS_ENABLED(CONFIG_MACSEC)
2282 /* MACsec management functions */
2283 const struct macsec_ops *macsec_ops;
2285 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2286 struct udp_tunnel_nic *udp_tunnel_nic;
2288 /* protected by rtnl_lock */
2289 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2291 u8 dev_addr_shadow[MAX_ADDR_LEN];
2292 netdevice_tracker linkwatch_dev_tracker;
2293 netdevice_tracker watchdog_dev_tracker;
2294 netdevice_tracker dev_registered_tracker;
2295 struct rtnl_hw_stats64 *offload_xstats_l3;
2297 #define to_net_dev(d) container_of(d, struct net_device, dev)
2299 static inline bool netif_elide_gro(const struct net_device *dev)
2301 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2306 #define NETDEV_ALIGN 32
2309 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2311 return dev->prio_tc_map[prio & TC_BITMASK];
2315 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2317 if (tc >= dev->num_tc)
2320 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2324 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2325 void netdev_reset_tc(struct net_device *dev);
2326 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2327 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2330 int netdev_get_num_tc(struct net_device *dev)
2335 static inline void net_prefetch(void *p)
2338 #if L1_CACHE_BYTES < 128
2339 prefetch((u8 *)p + L1_CACHE_BYTES);
2343 static inline void net_prefetchw(void *p)
2346 #if L1_CACHE_BYTES < 128
2347 prefetchw((u8 *)p + L1_CACHE_BYTES);
2351 void netdev_unbind_sb_channel(struct net_device *dev,
2352 struct net_device *sb_dev);
2353 int netdev_bind_sb_channel_queue(struct net_device *dev,
2354 struct net_device *sb_dev,
2355 u8 tc, u16 count, u16 offset);
2356 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2357 static inline int netdev_get_sb_channel(struct net_device *dev)
2359 return max_t(int, -dev->num_tc, 0);
2363 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2366 return &dev->_tx[index];
2369 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2370 const struct sk_buff *skb)
2372 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2375 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2376 void (*f)(struct net_device *,
2377 struct netdev_queue *,
2383 for (i = 0; i < dev->num_tx_queues; i++)
2384 f(dev, &dev->_tx[i], arg);
2387 #define netdev_lockdep_set_classes(dev) \
2389 static struct lock_class_key qdisc_tx_busylock_key; \
2390 static struct lock_class_key qdisc_xmit_lock_key; \
2391 static struct lock_class_key dev_addr_list_lock_key; \
2394 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2395 lockdep_set_class(&(dev)->addr_list_lock, \
2396 &dev_addr_list_lock_key); \
2397 for (i = 0; i < (dev)->num_tx_queues; i++) \
2398 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2399 &qdisc_xmit_lock_key); \
2402 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2403 struct net_device *sb_dev);
2404 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2405 struct sk_buff *skb,
2406 struct net_device *sb_dev);
2408 /* returns the headroom that the master device needs to take in account
2409 * when forwarding to this dev
2411 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2413 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2416 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2418 if (dev->netdev_ops->ndo_set_rx_headroom)
2419 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2422 /* set the device rx headroom to the dev's default */
2423 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2425 netdev_set_rx_headroom(dev, -1);
2428 static inline void *netdev_get_ml_priv(struct net_device *dev,
2429 enum netdev_ml_priv_type type)
2431 if (dev->ml_priv_type != type)
2434 return dev->ml_priv;
2437 static inline void netdev_set_ml_priv(struct net_device *dev,
2439 enum netdev_ml_priv_type type)
2441 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2442 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2443 dev->ml_priv_type, type);
2444 WARN(!dev->ml_priv_type && dev->ml_priv,
2445 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2447 dev->ml_priv = ml_priv;
2448 dev->ml_priv_type = type;
2452 * Net namespace inlines
2455 struct net *dev_net(const struct net_device *dev)
2457 return read_pnet(&dev->nd_net);
2461 void dev_net_set(struct net_device *dev, struct net *net)
2463 write_pnet(&dev->nd_net, net);
2467 * netdev_priv - access network device private data
2468 * @dev: network device
2470 * Get network device private data
2472 static inline void *netdev_priv(const struct net_device *dev)
2474 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2477 /* Set the sysfs physical device reference for the network logical device
2478 * if set prior to registration will cause a symlink during initialization.
2480 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2482 /* Set the sysfs device type for the network logical device to allow
2483 * fine-grained identification of different network device types. For
2484 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2486 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2488 /* Default NAPI poll() weight
2489 * Device drivers are strongly advised to not use bigger value
2491 #define NAPI_POLL_WEIGHT 64
2494 * netif_napi_add - initialize a NAPI context
2495 * @dev: network device
2496 * @napi: NAPI context
2497 * @poll: polling function
2498 * @weight: default weight
2500 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2501 * *any* of the other NAPI-related functions.
2503 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2504 int (*poll)(struct napi_struct *, int), int weight);
2507 * netif_tx_napi_add - initialize a NAPI context
2508 * @dev: network device
2509 * @napi: NAPI context
2510 * @poll: polling function
2511 * @weight: default weight
2513 * This variant of netif_napi_add() should be used from drivers using NAPI
2514 * to exclusively poll a TX queue.
2515 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2517 static inline void netif_tx_napi_add(struct net_device *dev,
2518 struct napi_struct *napi,
2519 int (*poll)(struct napi_struct *, int),
2522 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2523 netif_napi_add(dev, napi, poll, weight);
2527 * __netif_napi_del - remove a NAPI context
2528 * @napi: NAPI context
2530 * Warning: caller must observe RCU grace period before freeing memory
2531 * containing @napi. Drivers might want to call this helper to combine
2532 * all the needed RCU grace periods into a single one.
2534 void __netif_napi_del(struct napi_struct *napi);
2537 * netif_napi_del - remove a NAPI context
2538 * @napi: NAPI context
2540 * netif_napi_del() removes a NAPI context from the network device NAPI list
2542 static inline void netif_napi_del(struct napi_struct *napi)
2544 __netif_napi_del(napi);
2548 struct packet_type {
2549 __be16 type; /* This is really htons(ether_type). */
2550 bool ignore_outgoing;
2551 struct net_device *dev; /* NULL is wildcarded here */
2552 netdevice_tracker dev_tracker;
2553 int (*func) (struct sk_buff *,
2554 struct net_device *,
2555 struct packet_type *,
2556 struct net_device *);
2557 void (*list_func) (struct list_head *,
2558 struct packet_type *,
2559 struct net_device *);
2560 bool (*id_match)(struct packet_type *ptype,
2562 struct net *af_packet_net;
2563 void *af_packet_priv;
2564 struct list_head list;
2567 struct offload_callbacks {
2568 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2569 netdev_features_t features);
2570 struct sk_buff *(*gro_receive)(struct list_head *head,
2571 struct sk_buff *skb);
2572 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2575 struct packet_offload {
2576 __be16 type; /* This is really htons(ether_type). */
2578 struct offload_callbacks callbacks;
2579 struct list_head list;
2582 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2583 struct pcpu_sw_netstats {
2588 struct u64_stats_sync syncp;
2589 } __aligned(4 * sizeof(u64));
2591 struct pcpu_lstats {
2592 u64_stats_t packets;
2594 struct u64_stats_sync syncp;
2595 } __aligned(2 * sizeof(u64));
2597 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2599 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2601 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2603 u64_stats_update_begin(&tstats->syncp);
2604 tstats->rx_bytes += len;
2605 tstats->rx_packets++;
2606 u64_stats_update_end(&tstats->syncp);
2609 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2610 unsigned int packets,
2613 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2615 u64_stats_update_begin(&tstats->syncp);
2616 tstats->tx_bytes += len;
2617 tstats->tx_packets += packets;
2618 u64_stats_update_end(&tstats->syncp);
2621 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2623 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2625 u64_stats_update_begin(&lstats->syncp);
2626 u64_stats_add(&lstats->bytes, len);
2627 u64_stats_inc(&lstats->packets);
2628 u64_stats_update_end(&lstats->syncp);
2631 #define __netdev_alloc_pcpu_stats(type, gfp) \
2633 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2636 for_each_possible_cpu(__cpu) { \
2637 typeof(type) *stat; \
2638 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2639 u64_stats_init(&stat->syncp); \
2645 #define netdev_alloc_pcpu_stats(type) \
2646 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2648 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2650 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2653 for_each_possible_cpu(__cpu) { \
2654 typeof(type) *stat; \
2655 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2656 u64_stats_init(&stat->syncp); \
2662 enum netdev_lag_tx_type {
2663 NETDEV_LAG_TX_TYPE_UNKNOWN,
2664 NETDEV_LAG_TX_TYPE_RANDOM,
2665 NETDEV_LAG_TX_TYPE_BROADCAST,
2666 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2667 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2668 NETDEV_LAG_TX_TYPE_HASH,
2671 enum netdev_lag_hash {
2672 NETDEV_LAG_HASH_NONE,
2674 NETDEV_LAG_HASH_L34,
2675 NETDEV_LAG_HASH_L23,
2676 NETDEV_LAG_HASH_E23,
2677 NETDEV_LAG_HASH_E34,
2678 NETDEV_LAG_HASH_VLAN_SRCMAC,
2679 NETDEV_LAG_HASH_UNKNOWN,
2682 struct netdev_lag_upper_info {
2683 enum netdev_lag_tx_type tx_type;
2684 enum netdev_lag_hash hash_type;
2687 struct netdev_lag_lower_state_info {
2692 #include <linux/notifier.h>
2694 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2695 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2699 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2701 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2702 detected a hardware crash and restarted
2703 - we can use this eg to kick tcp sessions
2705 NETDEV_CHANGE, /* Notify device state change */
2708 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2709 NETDEV_CHANGEADDR, /* notify after the address change */
2710 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2714 NETDEV_BONDING_FAILOVER,
2716 NETDEV_PRE_TYPE_CHANGE,
2717 NETDEV_POST_TYPE_CHANGE,
2720 NETDEV_NOTIFY_PEERS,
2724 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2725 NETDEV_CHANGEINFODATA,
2726 NETDEV_BONDING_INFO,
2727 NETDEV_PRECHANGEUPPER,
2728 NETDEV_CHANGELOWERSTATE,
2729 NETDEV_UDP_TUNNEL_PUSH_INFO,
2730 NETDEV_UDP_TUNNEL_DROP_INFO,
2731 NETDEV_CHANGE_TX_QUEUE_LEN,
2732 NETDEV_CVLAN_FILTER_PUSH_INFO,
2733 NETDEV_CVLAN_FILTER_DROP_INFO,
2734 NETDEV_SVLAN_FILTER_PUSH_INFO,
2735 NETDEV_SVLAN_FILTER_DROP_INFO,
2736 NETDEV_OFFLOAD_XSTATS_ENABLE,
2737 NETDEV_OFFLOAD_XSTATS_DISABLE,
2738 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2739 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2741 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2743 int register_netdevice_notifier(struct notifier_block *nb);
2744 int unregister_netdevice_notifier(struct notifier_block *nb);
2745 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2746 int unregister_netdevice_notifier_net(struct net *net,
2747 struct notifier_block *nb);
2748 int register_netdevice_notifier_dev_net(struct net_device *dev,
2749 struct notifier_block *nb,
2750 struct netdev_net_notifier *nn);
2751 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2752 struct notifier_block *nb,
2753 struct netdev_net_notifier *nn);
2755 struct netdev_notifier_info {
2756 struct net_device *dev;
2757 struct netlink_ext_ack *extack;
2760 struct netdev_notifier_info_ext {
2761 struct netdev_notifier_info info; /* must be first */
2767 struct netdev_notifier_change_info {
2768 struct netdev_notifier_info info; /* must be first */
2769 unsigned int flags_changed;
2772 struct netdev_notifier_changeupper_info {
2773 struct netdev_notifier_info info; /* must be first */
2774 struct net_device *upper_dev; /* new upper dev */
2775 bool master; /* is upper dev master */
2776 bool linking; /* is the notification for link or unlink */
2777 void *upper_info; /* upper dev info */
2780 struct netdev_notifier_changelowerstate_info {
2781 struct netdev_notifier_info info; /* must be first */
2782 void *lower_state_info; /* is lower dev state */
2785 struct netdev_notifier_pre_changeaddr_info {
2786 struct netdev_notifier_info info; /* must be first */
2787 const unsigned char *dev_addr;
2790 enum netdev_offload_xstats_type {
2791 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2794 struct netdev_notifier_offload_xstats_info {
2795 struct netdev_notifier_info info; /* must be first */
2796 enum netdev_offload_xstats_type type;
2799 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2800 struct netdev_notifier_offload_xstats_rd *report_delta;
2801 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2802 struct netdev_notifier_offload_xstats_ru *report_used;
2806 int netdev_offload_xstats_enable(struct net_device *dev,
2807 enum netdev_offload_xstats_type type,
2808 struct netlink_ext_ack *extack);
2809 int netdev_offload_xstats_disable(struct net_device *dev,
2810 enum netdev_offload_xstats_type type);
2811 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2812 enum netdev_offload_xstats_type type);
2813 int netdev_offload_xstats_get(struct net_device *dev,
2814 enum netdev_offload_xstats_type type,
2815 struct rtnl_hw_stats64 *stats, bool *used,
2816 struct netlink_ext_ack *extack);
2818 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2819 const struct rtnl_hw_stats64 *stats);
2821 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2822 void netdev_offload_xstats_push_delta(struct net_device *dev,
2823 enum netdev_offload_xstats_type type,
2824 const struct rtnl_hw_stats64 *stats);
2826 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2827 struct net_device *dev)
2830 info->extack = NULL;
2833 static inline struct net_device *
2834 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2839 static inline struct netlink_ext_ack *
2840 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2842 return info->extack;
2845 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2848 extern rwlock_t dev_base_lock; /* Device list lock */
2850 #define for_each_netdev(net, d) \
2851 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2852 #define for_each_netdev_reverse(net, d) \
2853 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2854 #define for_each_netdev_rcu(net, d) \
2855 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2856 #define for_each_netdev_safe(net, d, n) \
2857 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2858 #define for_each_netdev_continue(net, d) \
2859 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2860 #define for_each_netdev_continue_reverse(net, d) \
2861 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2863 #define for_each_netdev_continue_rcu(net, d) \
2864 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2865 #define for_each_netdev_in_bond_rcu(bond, slave) \
2866 for_each_netdev_rcu(&init_net, slave) \
2867 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2868 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2870 static inline struct net_device *next_net_device(struct net_device *dev)
2872 struct list_head *lh;
2876 lh = dev->dev_list.next;
2877 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2880 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2882 struct list_head *lh;
2886 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2887 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2890 static inline struct net_device *first_net_device(struct net *net)
2892 return list_empty(&net->dev_base_head) ? NULL :
2893 net_device_entry(net->dev_base_head.next);
2896 static inline struct net_device *first_net_device_rcu(struct net *net)
2898 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2900 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2903 int netdev_boot_setup_check(struct net_device *dev);
2904 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2905 const char *hwaddr);
2906 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2907 void dev_add_pack(struct packet_type *pt);
2908 void dev_remove_pack(struct packet_type *pt);
2909 void __dev_remove_pack(struct packet_type *pt);
2910 void dev_add_offload(struct packet_offload *po);
2911 void dev_remove_offload(struct packet_offload *po);
2913 int dev_get_iflink(const struct net_device *dev);
2914 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2915 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2916 struct net_device_path_stack *stack);
2917 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2918 unsigned short mask);
2919 struct net_device *dev_get_by_name(struct net *net, const char *name);
2920 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2921 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2922 bool netdev_name_in_use(struct net *net, const char *name);
2923 int dev_alloc_name(struct net_device *dev, const char *name);
2924 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2925 void dev_close(struct net_device *dev);
2926 void dev_close_many(struct list_head *head, bool unlink);
2927 void dev_disable_lro(struct net_device *dev);
2928 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2929 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2930 struct net_device *sb_dev);
2931 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2932 struct net_device *sb_dev);
2934 int dev_queue_xmit(struct sk_buff *skb);
2935 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2936 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2938 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2942 ret = __dev_direct_xmit(skb, queue_id);
2943 if (!dev_xmit_complete(ret))
2948 int register_netdevice(struct net_device *dev);
2949 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2950 void unregister_netdevice_many(struct list_head *head);
2951 static inline void unregister_netdevice(struct net_device *dev)
2953 unregister_netdevice_queue(dev, NULL);
2956 int netdev_refcnt_read(const struct net_device *dev);
2957 void free_netdev(struct net_device *dev);
2958 void netdev_freemem(struct net_device *dev);
2959 int init_dummy_netdev(struct net_device *dev);
2961 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2962 struct sk_buff *skb,
2964 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2966 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2967 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2968 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2969 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2970 int dev_restart(struct net_device *dev);
2973 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2974 unsigned short type,
2975 const void *daddr, const void *saddr,
2978 if (!dev->header_ops || !dev->header_ops->create)
2981 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2984 static inline int dev_parse_header(const struct sk_buff *skb,
2985 unsigned char *haddr)
2987 const struct net_device *dev = skb->dev;
2989 if (!dev->header_ops || !dev->header_ops->parse)
2991 return dev->header_ops->parse(skb, haddr);
2994 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2996 const struct net_device *dev = skb->dev;
2998 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3000 return dev->header_ops->parse_protocol(skb);
3003 /* ll_header must have at least hard_header_len allocated */
3004 static inline bool dev_validate_header(const struct net_device *dev,
3005 char *ll_header, int len)
3007 if (likely(len >= dev->hard_header_len))
3009 if (len < dev->min_header_len)
3012 if (capable(CAP_SYS_RAWIO)) {
3013 memset(ll_header + len, 0, dev->hard_header_len - len);
3017 if (dev->header_ops && dev->header_ops->validate)
3018 return dev->header_ops->validate(ll_header, len);
3023 static inline bool dev_has_header(const struct net_device *dev)
3025 return dev->header_ops && dev->header_ops->create;
3029 * Incoming packets are placed on per-CPU queues
3031 struct softnet_data {
3032 struct list_head poll_list;
3033 struct sk_buff_head process_queue;
3036 unsigned int processed;
3037 unsigned int time_squeeze;
3038 unsigned int received_rps;
3040 struct softnet_data *rps_ipi_list;
3042 #ifdef CONFIG_NET_FLOW_LIMIT
3043 struct sd_flow_limit __rcu *flow_limit;
3045 struct Qdisc *output_queue;
3046 struct Qdisc **output_queue_tailp;
3047 struct sk_buff *completion_queue;
3048 #ifdef CONFIG_XFRM_OFFLOAD
3049 struct sk_buff_head xfrm_backlog;
3051 /* written and read only by owning cpu: */
3057 /* input_queue_head should be written by cpu owning this struct,
3058 * and only read by other cpus. Worth using a cache line.
3060 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3062 /* Elements below can be accessed between CPUs for RPS/RFS */
3063 call_single_data_t csd ____cacheline_aligned_in_smp;
3064 struct softnet_data *rps_ipi_next;
3066 unsigned int input_queue_tail;
3068 unsigned int dropped;
3069 struct sk_buff_head input_pkt_queue;
3070 struct napi_struct backlog;
3074 static inline void input_queue_head_incr(struct softnet_data *sd)
3077 sd->input_queue_head++;
3081 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3082 unsigned int *qtail)
3085 *qtail = ++sd->input_queue_tail;
3089 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3091 static inline int dev_recursion_level(void)
3093 return this_cpu_read(softnet_data.xmit.recursion);
3096 #define XMIT_RECURSION_LIMIT 8
3097 static inline bool dev_xmit_recursion(void)
3099 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3100 XMIT_RECURSION_LIMIT);
3103 static inline void dev_xmit_recursion_inc(void)
3105 __this_cpu_inc(softnet_data.xmit.recursion);
3108 static inline void dev_xmit_recursion_dec(void)
3110 __this_cpu_dec(softnet_data.xmit.recursion);
3113 void __netif_schedule(struct Qdisc *q);
3114 void netif_schedule_queue(struct netdev_queue *txq);
3116 static inline void netif_tx_schedule_all(struct net_device *dev)
3120 for (i = 0; i < dev->num_tx_queues; i++)
3121 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3124 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3126 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3130 * netif_start_queue - allow transmit
3131 * @dev: network device
3133 * Allow upper layers to call the device hard_start_xmit routine.
3135 static inline void netif_start_queue(struct net_device *dev)
3137 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3140 static inline void netif_tx_start_all_queues(struct net_device *dev)
3144 for (i = 0; i < dev->num_tx_queues; i++) {
3145 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3146 netif_tx_start_queue(txq);
3150 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3153 * netif_wake_queue - restart transmit
3154 * @dev: network device
3156 * Allow upper layers to call the device hard_start_xmit routine.
3157 * Used for flow control when transmit resources are available.
3159 static inline void netif_wake_queue(struct net_device *dev)
3161 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3164 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3168 for (i = 0; i < dev->num_tx_queues; i++) {
3169 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3170 netif_tx_wake_queue(txq);
3174 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3176 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3180 * netif_stop_queue - stop transmitted packets
3181 * @dev: network device
3183 * Stop upper layers calling the device hard_start_xmit routine.
3184 * Used for flow control when transmit resources are unavailable.
3186 static inline void netif_stop_queue(struct net_device *dev)
3188 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3191 void netif_tx_stop_all_queues(struct net_device *dev);
3193 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3195 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3199 * netif_queue_stopped - test if transmit queue is flowblocked
3200 * @dev: network device
3202 * Test if transmit queue on device is currently unable to send.
3204 static inline bool netif_queue_stopped(const struct net_device *dev)
3206 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3209 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3211 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3215 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3217 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3221 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3223 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3227 * netdev_queue_set_dql_min_limit - set dql minimum limit
3228 * @dev_queue: pointer to transmit queue
3229 * @min_limit: dql minimum limit
3231 * Forces xmit_more() to return true until the minimum threshold
3232 * defined by @min_limit is reached (or until the tx queue is
3233 * empty). Warning: to be use with care, misuse will impact the
3236 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3237 unsigned int min_limit)
3240 dev_queue->dql.min_limit = min_limit;
3245 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3246 * @dev_queue: pointer to transmit queue
3248 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3249 * to give appropriate hint to the CPU.
3251 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3254 prefetchw(&dev_queue->dql.num_queued);
3259 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3260 * @dev_queue: pointer to transmit queue
3262 * BQL enabled drivers might use this helper in their TX completion path,
3263 * to give appropriate hint to the CPU.
3265 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3268 prefetchw(&dev_queue->dql.limit);
3272 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3276 dql_queued(&dev_queue->dql, bytes);
3278 if (likely(dql_avail(&dev_queue->dql) >= 0))
3281 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3284 * The XOFF flag must be set before checking the dql_avail below,
3285 * because in netdev_tx_completed_queue we update the dql_completed
3286 * before checking the XOFF flag.
3290 /* check again in case another CPU has just made room avail */
3291 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3292 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3296 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3297 * that they should not test BQL status themselves.
3298 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3300 * Returns true if the doorbell must be used to kick the NIC.
3302 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3308 dql_queued(&dev_queue->dql, bytes);
3310 return netif_tx_queue_stopped(dev_queue);
3312 netdev_tx_sent_queue(dev_queue, bytes);
3317 * netdev_sent_queue - report the number of bytes queued to hardware
3318 * @dev: network device
3319 * @bytes: number of bytes queued to the hardware device queue
3321 * Report the number of bytes queued for sending/completion to the network
3322 * device hardware queue. @bytes should be a good approximation and should
3323 * exactly match netdev_completed_queue() @bytes
3325 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3327 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3330 static inline bool __netdev_sent_queue(struct net_device *dev,
3334 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3338 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3339 unsigned int pkts, unsigned int bytes)
3342 if (unlikely(!bytes))
3345 dql_completed(&dev_queue->dql, bytes);
3348 * Without the memory barrier there is a small possiblity that
3349 * netdev_tx_sent_queue will miss the update and cause the queue to
3350 * be stopped forever
3354 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3357 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3358 netif_schedule_queue(dev_queue);
3363 * netdev_completed_queue - report bytes and packets completed by device
3364 * @dev: network device
3365 * @pkts: actual number of packets sent over the medium
3366 * @bytes: actual number of bytes sent over the medium
3368 * Report the number of bytes and packets transmitted by the network device
3369 * hardware queue over the physical medium, @bytes must exactly match the
3370 * @bytes amount passed to netdev_sent_queue()
3372 static inline void netdev_completed_queue(struct net_device *dev,
3373 unsigned int pkts, unsigned int bytes)
3375 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3378 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3381 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3387 * netdev_reset_queue - reset the packets and bytes count of a network device
3388 * @dev_queue: network device
3390 * Reset the bytes and packet count of a network device and clear the
3391 * software flow control OFF bit for this network device
3393 static inline void netdev_reset_queue(struct net_device *dev_queue)
3395 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3399 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3400 * @dev: network device
3401 * @queue_index: given tx queue index
3403 * Returns 0 if given tx queue index >= number of device tx queues,
3404 * otherwise returns the originally passed tx queue index.
3406 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3408 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3409 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3410 dev->name, queue_index,
3411 dev->real_num_tx_queues);
3419 * netif_running - test if up
3420 * @dev: network device
3422 * Test if the device has been brought up.
3424 static inline bool netif_running(const struct net_device *dev)
3426 return test_bit(__LINK_STATE_START, &dev->state);
3430 * Routines to manage the subqueues on a device. We only need start,
3431 * stop, and a check if it's stopped. All other device management is
3432 * done at the overall netdevice level.
3433 * Also test the device if we're multiqueue.
3437 * netif_start_subqueue - allow sending packets on subqueue
3438 * @dev: network device
3439 * @queue_index: sub queue index
3441 * Start individual transmit queue of a device with multiple transmit queues.
3443 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3445 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3447 netif_tx_start_queue(txq);
3451 * netif_stop_subqueue - stop sending packets on subqueue
3452 * @dev: network device
3453 * @queue_index: sub queue index
3455 * Stop individual transmit queue of a device with multiple transmit queues.
3457 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3459 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3460 netif_tx_stop_queue(txq);
3464 * __netif_subqueue_stopped - test status of subqueue
3465 * @dev: network device
3466 * @queue_index: sub queue index
3468 * Check individual transmit queue of a device with multiple transmit queues.
3470 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3473 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3475 return netif_tx_queue_stopped(txq);
3479 * netif_subqueue_stopped - test status of subqueue
3480 * @dev: network device
3481 * @skb: sub queue buffer pointer
3483 * Check individual transmit queue of a device with multiple transmit queues.
3485 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3486 struct sk_buff *skb)
3488 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3492 * netif_wake_subqueue - allow sending packets on subqueue
3493 * @dev: network device
3494 * @queue_index: sub queue index
3496 * Resume individual transmit queue of a device with multiple transmit queues.
3498 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3500 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3502 netif_tx_wake_queue(txq);
3506 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3508 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3509 u16 index, enum xps_map_type type);
3512 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3513 * @j: CPU/Rx queue index
3514 * @mask: bitmask of all cpus/rx queues
3515 * @nr_bits: number of bits in the bitmask
3517 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3519 static inline bool netif_attr_test_mask(unsigned long j,
3520 const unsigned long *mask,
3521 unsigned int nr_bits)
3523 cpu_max_bits_warn(j, nr_bits);
3524 return test_bit(j, mask);
3528 * netif_attr_test_online - Test for online CPU/Rx queue
3529 * @j: CPU/Rx queue index
3530 * @online_mask: bitmask for CPUs/Rx queues that are online
3531 * @nr_bits: number of bits in the bitmask
3533 * Returns true if a CPU/Rx queue is online.
3535 static inline bool netif_attr_test_online(unsigned long j,
3536 const unsigned long *online_mask,
3537 unsigned int nr_bits)
3539 cpu_max_bits_warn(j, nr_bits);
3542 return test_bit(j, online_mask);
3544 return (j < nr_bits);
3548 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3549 * @n: CPU/Rx queue index
3550 * @srcp: the cpumask/Rx queue mask pointer
3551 * @nr_bits: number of bits in the bitmask
3553 * Returns >= nr_bits if no further CPUs/Rx queues set.
3555 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3556 unsigned int nr_bits)
3558 /* -1 is a legal arg here. */
3560 cpu_max_bits_warn(n, nr_bits);
3563 return find_next_bit(srcp, nr_bits, n + 1);
3569 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3570 * @n: CPU/Rx queue index
3571 * @src1p: the first CPUs/Rx queues mask pointer
3572 * @src2p: the second CPUs/Rx queues mask pointer
3573 * @nr_bits: number of bits in the bitmask
3575 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3577 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3578 const unsigned long *src2p,
3579 unsigned int nr_bits)
3581 /* -1 is a legal arg here. */
3583 cpu_max_bits_warn(n, nr_bits);
3586 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3588 return find_next_bit(src1p, nr_bits, n + 1);
3590 return find_next_bit(src2p, nr_bits, n + 1);
3595 static inline int netif_set_xps_queue(struct net_device *dev,
3596 const struct cpumask *mask,
3602 static inline int __netif_set_xps_queue(struct net_device *dev,
3603 const unsigned long *mask,
3604 u16 index, enum xps_map_type type)
3611 * netif_is_multiqueue - test if device has multiple transmit queues
3612 * @dev: network device
3614 * Check if device has multiple transmit queues
3616 static inline bool netif_is_multiqueue(const struct net_device *dev)
3618 return dev->num_tx_queues > 1;
3621 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3624 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3626 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3629 dev->real_num_rx_queues = rxqs;
3633 int netif_set_real_num_queues(struct net_device *dev,
3634 unsigned int txq, unsigned int rxq);
3636 static inline struct netdev_rx_queue *
3637 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3639 return dev->_rx + rxq;
3643 static inline unsigned int get_netdev_rx_queue_index(
3644 struct netdev_rx_queue *queue)
3646 struct net_device *dev = queue->dev;
3647 int index = queue - dev->_rx;
3649 BUG_ON(index >= dev->num_rx_queues);
3654 int netif_get_num_default_rss_queues(void);
3656 enum skb_free_reason {
3657 SKB_REASON_CONSUMED,
3661 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3662 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3665 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3666 * interrupt context or with hardware interrupts being disabled.
3667 * (in_hardirq() || irqs_disabled())
3669 * We provide four helpers that can be used in following contexts :
3671 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3672 * replacing kfree_skb(skb)
3674 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3675 * Typically used in place of consume_skb(skb) in TX completion path
3677 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3678 * replacing kfree_skb(skb)
3680 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3681 * and consumed a packet. Used in place of consume_skb(skb)
3683 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3685 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3688 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3690 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3693 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3695 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3698 static inline void dev_consume_skb_any(struct sk_buff *skb)
3700 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3703 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3704 struct bpf_prog *xdp_prog);
3705 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3706 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3707 int netif_rx(struct sk_buff *skb);
3708 int __netif_rx(struct sk_buff *skb);
3710 int netif_receive_skb(struct sk_buff *skb);
3711 int netif_receive_skb_core(struct sk_buff *skb);
3712 void netif_receive_skb_list_internal(struct list_head *head);
3713 void netif_receive_skb_list(struct list_head *head);
3714 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3715 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3716 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3717 gro_result_t napi_gro_frags(struct napi_struct *napi);
3718 struct packet_offload *gro_find_receive_by_type(__be16 type);
3719 struct packet_offload *gro_find_complete_by_type(__be16 type);
3721 static inline void napi_free_frags(struct napi_struct *napi)
3723 kfree_skb(napi->skb);
3727 bool netdev_is_rx_handler_busy(struct net_device *dev);
3728 int netdev_rx_handler_register(struct net_device *dev,
3729 rx_handler_func_t *rx_handler,
3730 void *rx_handler_data);
3731 void netdev_rx_handler_unregister(struct net_device *dev);
3733 bool dev_valid_name(const char *name);
3734 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3736 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3738 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3739 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3740 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3741 void __user *data, bool *need_copyout);
3742 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3743 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
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_set_alias(struct net_device *, const char *, size_t);
3752 int dev_get_alias(const struct net_device *, char *, size_t);
3753 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3754 const char *pat, int new_ifindex);
3756 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3759 return __dev_change_net_namespace(dev, net, pat, 0);
3761 int __dev_set_mtu(struct net_device *, int);
3762 int dev_set_mtu(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_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3768 struct netlink_ext_ack *extack);
3769 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3770 int dev_get_port_parent_id(struct net_device *dev,
3771 struct netdev_phys_item_id *ppid, bool recurse);
3772 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3773 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3774 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3775 struct netdev_queue *txq, int *ret);
3777 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3778 u8 dev_xdp_prog_count(struct net_device *dev);
3779 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3781 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3782 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3783 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3784 bool is_skb_forwardable(const struct net_device *dev,
3785 const struct sk_buff *skb);
3787 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3788 const struct sk_buff *skb,
3789 const bool check_mtu)
3791 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3794 if (!(dev->flags & IFF_UP))
3800 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3801 if (skb->len <= len)
3804 /* if TSO is enabled, we don't care about the length as the packet
3805 * could be forwarded without being segmented before
3807 if (skb_is_gso(skb))
3813 struct net_device_core_stats *netdev_core_stats_alloc(struct net_device *dev);
3815 static inline struct net_device_core_stats *dev_core_stats(struct net_device *dev)
3817 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
3818 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
3821 return this_cpu_ptr(p);
3823 return netdev_core_stats_alloc(dev);
3826 #define DEV_CORE_STATS_INC(FIELD) \
3827 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3829 struct net_device_core_stats *p; \
3831 preempt_disable(); \
3832 p = dev_core_stats(dev); \
3835 local_inc(&p->FIELD); \
3838 DEV_CORE_STATS_INC(rx_dropped)
3839 DEV_CORE_STATS_INC(tx_dropped)
3840 DEV_CORE_STATS_INC(rx_nohandler)
3841 DEV_CORE_STATS_INC(rx_otherhost_dropped)
3843 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3844 struct sk_buff *skb,
3845 const bool check_mtu)
3847 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3848 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3849 dev_core_stats_rx_dropped_inc(dev);
3854 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
3859 bool dev_nit_active(struct net_device *dev);
3860 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3862 static inline void __dev_put(struct net_device *dev)
3865 #ifdef CONFIG_PCPU_DEV_REFCNT
3866 this_cpu_dec(*dev->pcpu_refcnt);
3868 refcount_dec(&dev->dev_refcnt);
3873 static inline void __dev_hold(struct net_device *dev)
3876 #ifdef CONFIG_PCPU_DEV_REFCNT
3877 this_cpu_inc(*dev->pcpu_refcnt);
3879 refcount_inc(&dev->dev_refcnt);
3884 static inline void __netdev_tracker_alloc(struct net_device *dev,
3885 netdevice_tracker *tracker,
3888 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3889 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
3893 /* netdev_tracker_alloc() can upgrade a prior untracked reference
3894 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
3896 static inline void netdev_tracker_alloc(struct net_device *dev,
3897 netdevice_tracker *tracker, gfp_t gfp)
3899 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3900 refcount_dec(&dev->refcnt_tracker.no_tracker);
3901 __netdev_tracker_alloc(dev, tracker, gfp);
3905 static inline void netdev_tracker_free(struct net_device *dev,
3906 netdevice_tracker *tracker)
3908 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3909 ref_tracker_free(&dev->refcnt_tracker, tracker);
3913 static inline void dev_hold_track(struct net_device *dev,
3914 netdevice_tracker *tracker, gfp_t gfp)
3918 __netdev_tracker_alloc(dev, tracker, gfp);
3922 static inline void dev_put_track(struct net_device *dev,
3923 netdevice_tracker *tracker)
3926 netdev_tracker_free(dev, tracker);
3932 * dev_hold - get reference to device
3933 * @dev: network device
3935 * Hold reference to device to keep it from being freed.
3936 * Try using dev_hold_track() instead.
3938 static inline void dev_hold(struct net_device *dev)
3940 dev_hold_track(dev, NULL, GFP_ATOMIC);
3944 * dev_put - release reference to device
3945 * @dev: network device
3947 * Release reference to device to allow it to be freed.
3948 * Try using dev_put_track() instead.
3950 static inline void dev_put(struct net_device *dev)
3952 dev_put_track(dev, NULL);
3955 static inline void dev_replace_track(struct net_device *odev,
3956 struct net_device *ndev,
3957 netdevice_tracker *tracker,
3961 netdev_tracker_free(odev, tracker);
3967 __netdev_tracker_alloc(ndev, tracker, gfp);
3970 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3971 * and _off may be called from IRQ context, but it is caller
3972 * who is responsible for serialization of these calls.
3974 * The name carrier is inappropriate, these functions should really be
3975 * called netif_lowerlayer_*() because they represent the state of any
3976 * kind of lower layer not just hardware media.
3978 void linkwatch_fire_event(struct net_device *dev);
3981 * netif_carrier_ok - test if carrier present
3982 * @dev: network device
3984 * Check if carrier is present on device
3986 static inline bool netif_carrier_ok(const struct net_device *dev)
3988 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3991 unsigned long dev_trans_start(struct net_device *dev);
3993 void __netdev_watchdog_up(struct net_device *dev);
3995 void netif_carrier_on(struct net_device *dev);
3996 void netif_carrier_off(struct net_device *dev);
3997 void netif_carrier_event(struct net_device *dev);
4000 * netif_dormant_on - mark device as dormant.
4001 * @dev: network device
4003 * Mark device as dormant (as per RFC2863).
4005 * The dormant state indicates that the relevant interface is not
4006 * actually in a condition to pass packets (i.e., it is not 'up') but is
4007 * in a "pending" state, waiting for some external event. For "on-
4008 * demand" interfaces, this new state identifies the situation where the
4009 * interface is waiting for events to place it in the up state.
4011 static inline void netif_dormant_on(struct net_device *dev)
4013 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4014 linkwatch_fire_event(dev);
4018 * netif_dormant_off - set device as not dormant.
4019 * @dev: network device
4021 * Device is not in dormant state.
4023 static inline void netif_dormant_off(struct net_device *dev)
4025 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4026 linkwatch_fire_event(dev);
4030 * netif_dormant - test if device is dormant
4031 * @dev: network device
4033 * Check if device is dormant.
4035 static inline bool netif_dormant(const struct net_device *dev)
4037 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4042 * netif_testing_on - mark device as under test.
4043 * @dev: network device
4045 * Mark device as under test (as per RFC2863).
4047 * The testing state indicates that some test(s) must be performed on
4048 * the interface. After completion, of the test, the interface state
4049 * will change to up, dormant, or down, as appropriate.
4051 static inline void netif_testing_on(struct net_device *dev)
4053 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4054 linkwatch_fire_event(dev);
4058 * netif_testing_off - set device as not under test.
4059 * @dev: network device
4061 * Device is not in testing state.
4063 static inline void netif_testing_off(struct net_device *dev)
4065 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4066 linkwatch_fire_event(dev);
4070 * netif_testing - test if device is under test
4071 * @dev: network device
4073 * Check if device is under test
4075 static inline bool netif_testing(const struct net_device *dev)
4077 return test_bit(__LINK_STATE_TESTING, &dev->state);
4082 * netif_oper_up - test if device is operational
4083 * @dev: network device
4085 * Check if carrier is operational
4087 static inline bool netif_oper_up(const struct net_device *dev)
4089 return (dev->operstate == IF_OPER_UP ||
4090 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4094 * netif_device_present - is device available or removed
4095 * @dev: network device
4097 * Check if device has not been removed from system.
4099 static inline bool netif_device_present(const struct net_device *dev)
4101 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4104 void netif_device_detach(struct net_device *dev);
4106 void netif_device_attach(struct net_device *dev);
4109 * Network interface message level settings
4114 NETIF_MSG_PROBE_BIT,
4116 NETIF_MSG_TIMER_BIT,
4117 NETIF_MSG_IFDOWN_BIT,
4119 NETIF_MSG_RX_ERR_BIT,
4120 NETIF_MSG_TX_ERR_BIT,
4121 NETIF_MSG_TX_QUEUED_BIT,
4123 NETIF_MSG_TX_DONE_BIT,
4124 NETIF_MSG_RX_STATUS_BIT,
4125 NETIF_MSG_PKTDATA_BIT,
4129 /* When you add a new bit above, update netif_msg_class_names array
4130 * in net/ethtool/common.c
4132 NETIF_MSG_CLASS_COUNT,
4134 /* Both ethtool_ops interface and internal driver implementation use u32 */
4135 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4137 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4138 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4140 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4141 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4142 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4143 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4144 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4145 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4146 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4147 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4148 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4149 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4150 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4151 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4152 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4153 #define NETIF_MSG_HW __NETIF_MSG(HW)
4154 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4156 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4157 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4158 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4159 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4160 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4161 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4162 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4163 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4164 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4165 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4166 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4167 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4168 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4169 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4170 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4172 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4175 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4176 return default_msg_enable_bits;
4177 if (debug_value == 0) /* no output */
4179 /* set low N bits */
4180 return (1U << debug_value) - 1;
4183 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4185 spin_lock(&txq->_xmit_lock);
4186 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4187 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4190 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4192 __acquire(&txq->_xmit_lock);
4196 static inline void __netif_tx_release(struct netdev_queue *txq)
4198 __release(&txq->_xmit_lock);
4201 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4203 spin_lock_bh(&txq->_xmit_lock);
4204 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4205 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4208 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4210 bool ok = spin_trylock(&txq->_xmit_lock);
4213 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4214 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4219 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4221 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4222 WRITE_ONCE(txq->xmit_lock_owner, -1);
4223 spin_unlock(&txq->_xmit_lock);
4226 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4228 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4229 WRITE_ONCE(txq->xmit_lock_owner, -1);
4230 spin_unlock_bh(&txq->_xmit_lock);
4234 * txq->trans_start can be read locklessly from dev_watchdog()
4236 static inline void txq_trans_update(struct netdev_queue *txq)
4238 if (txq->xmit_lock_owner != -1)
4239 WRITE_ONCE(txq->trans_start, jiffies);
4242 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4244 unsigned long now = jiffies;
4246 if (READ_ONCE(txq->trans_start) != now)
4247 WRITE_ONCE(txq->trans_start, now);
4250 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4251 static inline void netif_trans_update(struct net_device *dev)
4253 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4255 txq_trans_cond_update(txq);
4259 * netif_tx_lock - grab network device transmit lock
4260 * @dev: network device
4262 * Get network device transmit lock
4264 void netif_tx_lock(struct net_device *dev);
4266 static inline void netif_tx_lock_bh(struct net_device *dev)
4272 void netif_tx_unlock(struct net_device *dev);
4274 static inline void netif_tx_unlock_bh(struct net_device *dev)
4276 netif_tx_unlock(dev);
4280 #define HARD_TX_LOCK(dev, txq, cpu) { \
4281 if ((dev->features & NETIF_F_LLTX) == 0) { \
4282 __netif_tx_lock(txq, cpu); \
4284 __netif_tx_acquire(txq); \
4288 #define HARD_TX_TRYLOCK(dev, txq) \
4289 (((dev->features & NETIF_F_LLTX) == 0) ? \
4290 __netif_tx_trylock(txq) : \
4291 __netif_tx_acquire(txq))
4293 #define HARD_TX_UNLOCK(dev, txq) { \
4294 if ((dev->features & NETIF_F_LLTX) == 0) { \
4295 __netif_tx_unlock(txq); \
4297 __netif_tx_release(txq); \
4301 static inline void netif_tx_disable(struct net_device *dev)
4307 cpu = smp_processor_id();
4308 spin_lock(&dev->tx_global_lock);
4309 for (i = 0; i < dev->num_tx_queues; i++) {
4310 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4312 __netif_tx_lock(txq, cpu);
4313 netif_tx_stop_queue(txq);
4314 __netif_tx_unlock(txq);
4316 spin_unlock(&dev->tx_global_lock);
4320 static inline void netif_addr_lock(struct net_device *dev)
4322 unsigned char nest_level = 0;
4324 #ifdef CONFIG_LOCKDEP
4325 nest_level = dev->nested_level;
4327 spin_lock_nested(&dev->addr_list_lock, nest_level);
4330 static inline void netif_addr_lock_bh(struct net_device *dev)
4332 unsigned char nest_level = 0;
4334 #ifdef CONFIG_LOCKDEP
4335 nest_level = dev->nested_level;
4338 spin_lock_nested(&dev->addr_list_lock, nest_level);
4341 static inline void netif_addr_unlock(struct net_device *dev)
4343 spin_unlock(&dev->addr_list_lock);
4346 static inline void netif_addr_unlock_bh(struct net_device *dev)
4348 spin_unlock_bh(&dev->addr_list_lock);
4352 * dev_addrs walker. Should be used only for read access. Call with
4353 * rcu_read_lock held.
4355 #define for_each_dev_addr(dev, ha) \
4356 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4358 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4360 void ether_setup(struct net_device *dev);
4362 /* Support for loadable net-drivers */
4363 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4364 unsigned char name_assign_type,
4365 void (*setup)(struct net_device *),
4366 unsigned int txqs, unsigned int rxqs);
4367 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4368 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4370 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4371 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4374 int register_netdev(struct net_device *dev);
4375 void unregister_netdev(struct net_device *dev);
4377 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4379 /* General hardware address lists handling functions */
4380 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4381 struct netdev_hw_addr_list *from_list, int addr_len);
4382 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4383 struct netdev_hw_addr_list *from_list, int addr_len);
4384 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4385 struct net_device *dev,
4386 int (*sync)(struct net_device *, const unsigned char *),
4387 int (*unsync)(struct net_device *,
4388 const unsigned char *));
4389 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4390 struct net_device *dev,
4391 int (*sync)(struct net_device *,
4392 const unsigned char *, int),
4393 int (*unsync)(struct net_device *,
4394 const unsigned char *, int));
4395 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4396 struct net_device *dev,
4397 int (*unsync)(struct net_device *,
4398 const unsigned char *, int));
4399 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4400 struct net_device *dev,
4401 int (*unsync)(struct net_device *,
4402 const unsigned char *));
4403 void __hw_addr_init(struct netdev_hw_addr_list *list);
4405 /* Functions used for device addresses handling */
4406 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4407 const void *addr, size_t len);
4410 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4412 dev_addr_mod(dev, 0, addr, len);
4415 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4417 __dev_addr_set(dev, addr, dev->addr_len);
4420 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4421 unsigned char addr_type);
4422 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4423 unsigned char addr_type);
4425 /* Functions used for unicast addresses handling */
4426 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4427 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4428 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4429 int dev_uc_sync(struct net_device *to, struct net_device *from);
4430 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4431 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4432 void dev_uc_flush(struct net_device *dev);
4433 void dev_uc_init(struct net_device *dev);
4436 * __dev_uc_sync - Synchonize device's unicast list
4437 * @dev: device to sync
4438 * @sync: function to call if address should be added
4439 * @unsync: function to call if address should be removed
4441 * Add newly added addresses to the interface, and release
4442 * addresses that have been deleted.
4444 static inline int __dev_uc_sync(struct net_device *dev,
4445 int (*sync)(struct net_device *,
4446 const unsigned char *),
4447 int (*unsync)(struct net_device *,
4448 const unsigned char *))
4450 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4454 * __dev_uc_unsync - Remove synchronized addresses from device
4455 * @dev: device to sync
4456 * @unsync: function to call if address should be removed
4458 * Remove all addresses that were added to the device by dev_uc_sync().
4460 static inline void __dev_uc_unsync(struct net_device *dev,
4461 int (*unsync)(struct net_device *,
4462 const unsigned char *))
4464 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4467 /* Functions used for multicast addresses handling */
4468 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4469 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4470 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4471 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4472 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4473 int dev_mc_sync(struct net_device *to, struct net_device *from);
4474 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4475 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4476 void dev_mc_flush(struct net_device *dev);
4477 void dev_mc_init(struct net_device *dev);
4480 * __dev_mc_sync - Synchonize device's multicast list
4481 * @dev: device to sync
4482 * @sync: function to call if address should be added
4483 * @unsync: function to call if address should be removed
4485 * Add newly added addresses to the interface, and release
4486 * addresses that have been deleted.
4488 static inline int __dev_mc_sync(struct net_device *dev,
4489 int (*sync)(struct net_device *,
4490 const unsigned char *),
4491 int (*unsync)(struct net_device *,
4492 const unsigned char *))
4494 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4498 * __dev_mc_unsync - Remove synchronized addresses from device
4499 * @dev: device to sync
4500 * @unsync: function to call if address should be removed
4502 * Remove all addresses that were added to the device by dev_mc_sync().
4504 static inline void __dev_mc_unsync(struct net_device *dev,
4505 int (*unsync)(struct net_device *,
4506 const unsigned char *))
4508 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4511 /* Functions used for secondary unicast and multicast support */
4512 void dev_set_rx_mode(struct net_device *dev);
4513 int dev_set_promiscuity(struct net_device *dev, int inc);
4514 int dev_set_allmulti(struct net_device *dev, int inc);
4515 void netdev_state_change(struct net_device *dev);
4516 void __netdev_notify_peers(struct net_device *dev);
4517 void netdev_notify_peers(struct net_device *dev);
4518 void netdev_features_change(struct net_device *dev);
4519 /* Load a device via the kmod */
4520 void dev_load(struct net *net, const char *name);
4521 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4522 struct rtnl_link_stats64 *storage);
4523 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4524 const struct net_device_stats *netdev_stats);
4525 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4526 const struct pcpu_sw_netstats __percpu *netstats);
4527 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4529 extern int netdev_max_backlog;
4530 extern int dev_rx_weight;
4531 extern int dev_tx_weight;
4532 extern int gro_normal_batch;
4535 NESTED_SYNC_IMM_BIT,
4536 NESTED_SYNC_TODO_BIT,
4539 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4540 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4542 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4543 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4545 struct netdev_nested_priv {
4546 unsigned char flags;
4550 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4551 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4552 struct list_head **iter);
4554 /* iterate through upper list, must be called under RCU read lock */
4555 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4556 for (iter = &(dev)->adj_list.upper, \
4557 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4559 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4561 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4562 int (*fn)(struct net_device *upper_dev,
4563 struct netdev_nested_priv *priv),
4564 struct netdev_nested_priv *priv);
4566 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4567 struct net_device *upper_dev);
4569 bool netdev_has_any_upper_dev(struct net_device *dev);
4571 void *netdev_lower_get_next_private(struct net_device *dev,
4572 struct list_head **iter);
4573 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4574 struct list_head **iter);
4576 #define netdev_for_each_lower_private(dev, priv, iter) \
4577 for (iter = (dev)->adj_list.lower.next, \
4578 priv = netdev_lower_get_next_private(dev, &(iter)); \
4580 priv = netdev_lower_get_next_private(dev, &(iter)))
4582 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4583 for (iter = &(dev)->adj_list.lower, \
4584 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4586 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4588 void *netdev_lower_get_next(struct net_device *dev,
4589 struct list_head **iter);
4591 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4592 for (iter = (dev)->adj_list.lower.next, \
4593 ldev = netdev_lower_get_next(dev, &(iter)); \
4595 ldev = netdev_lower_get_next(dev, &(iter)))
4597 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4598 struct list_head **iter);
4599 int netdev_walk_all_lower_dev(struct net_device *dev,
4600 int (*fn)(struct net_device *lower_dev,
4601 struct netdev_nested_priv *priv),
4602 struct netdev_nested_priv *priv);
4603 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4604 int (*fn)(struct net_device *lower_dev,
4605 struct netdev_nested_priv *priv),
4606 struct netdev_nested_priv *priv);
4608 void *netdev_adjacent_get_private(struct list_head *adj_list);
4609 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4610 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4611 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4612 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4613 struct netlink_ext_ack *extack);
4614 int netdev_master_upper_dev_link(struct net_device *dev,
4615 struct net_device *upper_dev,
4616 void *upper_priv, void *upper_info,
4617 struct netlink_ext_ack *extack);
4618 void netdev_upper_dev_unlink(struct net_device *dev,
4619 struct net_device *upper_dev);
4620 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4621 struct net_device *new_dev,
4622 struct net_device *dev,
4623 struct netlink_ext_ack *extack);
4624 void netdev_adjacent_change_commit(struct net_device *old_dev,
4625 struct net_device *new_dev,
4626 struct net_device *dev);
4627 void netdev_adjacent_change_abort(struct net_device *old_dev,
4628 struct net_device *new_dev,
4629 struct net_device *dev);
4630 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4631 void *netdev_lower_dev_get_private(struct net_device *dev,
4632 struct net_device *lower_dev);
4633 void netdev_lower_state_changed(struct net_device *lower_dev,
4634 void *lower_state_info);
4636 /* RSS keys are 40 or 52 bytes long */
4637 #define NETDEV_RSS_KEY_LEN 52
4638 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4639 void netdev_rss_key_fill(void *buffer, size_t len);
4641 int skb_checksum_help(struct sk_buff *skb);
4642 int skb_crc32c_csum_help(struct sk_buff *skb);
4643 int skb_csum_hwoffload_help(struct sk_buff *skb,
4644 const netdev_features_t features);
4646 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4647 netdev_features_t features, bool tx_path);
4648 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb,
4649 netdev_features_t features, __be16 type);
4650 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4651 netdev_features_t features);
4653 struct netdev_bonding_info {
4658 struct netdev_notifier_bonding_info {
4659 struct netdev_notifier_info info; /* must be first */
4660 struct netdev_bonding_info bonding_info;
4663 void netdev_bonding_info_change(struct net_device *dev,
4664 struct netdev_bonding_info *bonding_info);
4666 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4667 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4669 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4676 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4678 return __skb_gso_segment(skb, features, true);
4680 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4682 static inline bool can_checksum_protocol(netdev_features_t features,
4685 if (protocol == htons(ETH_P_FCOE))
4686 return !!(features & NETIF_F_FCOE_CRC);
4688 /* Assume this is an IP checksum (not SCTP CRC) */
4690 if (features & NETIF_F_HW_CSUM) {
4691 /* Can checksum everything */
4696 case htons(ETH_P_IP):
4697 return !!(features & NETIF_F_IP_CSUM);
4698 case htons(ETH_P_IPV6):
4699 return !!(features & NETIF_F_IPV6_CSUM);
4706 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4708 static inline void netdev_rx_csum_fault(struct net_device *dev,
4709 struct sk_buff *skb)
4713 /* rx skb timestamps */
4714 void net_enable_timestamp(void);
4715 void net_disable_timestamp(void);
4717 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4718 struct sk_buff *skb, struct net_device *dev,
4721 __this_cpu_write(softnet_data.xmit.more, more);
4722 return ops->ndo_start_xmit(skb, dev);
4725 static inline bool netdev_xmit_more(void)
4727 return __this_cpu_read(softnet_data.xmit.more);
4730 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4731 struct netdev_queue *txq, bool more)
4733 const struct net_device_ops *ops = dev->netdev_ops;
4736 rc = __netdev_start_xmit(ops, skb, dev, more);
4737 if (rc == NETDEV_TX_OK)
4738 txq_trans_update(txq);
4743 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4745 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4748 extern const struct kobj_ns_type_operations net_ns_type_operations;
4750 const char *netdev_drivername(const struct net_device *dev);
4752 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4753 netdev_features_t f2)
4755 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4756 if (f1 & NETIF_F_HW_CSUM)
4757 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4759 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4765 static inline netdev_features_t netdev_get_wanted_features(
4766 struct net_device *dev)
4768 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4770 netdev_features_t netdev_increment_features(netdev_features_t all,
4771 netdev_features_t one, netdev_features_t mask);
4773 /* Allow TSO being used on stacked device :
4774 * Performing the GSO segmentation before last device
4775 * is a performance improvement.
4777 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4778 netdev_features_t mask)
4780 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4783 int __netdev_update_features(struct net_device *dev);
4784 void netdev_update_features(struct net_device *dev);
4785 void netdev_change_features(struct net_device *dev);
4787 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4788 struct net_device *dev);
4790 netdev_features_t passthru_features_check(struct sk_buff *skb,
4791 struct net_device *dev,
4792 netdev_features_t features);
4793 netdev_features_t netif_skb_features(struct sk_buff *skb);
4795 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4797 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4799 /* check flags correspondence */
4800 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4801 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4802 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4803 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4804 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4805 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4806 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4807 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4808 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4809 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4810 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4811 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4812 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4813 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4814 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4815 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4816 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4817 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4818 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4820 return (features & feature) == feature;
4823 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4825 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4826 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4829 static inline bool netif_needs_gso(struct sk_buff *skb,
4830 netdev_features_t features)
4832 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4833 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4834 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4837 static inline void netif_set_gso_max_size(struct net_device *dev,
4840 /* dev->gso_max_size is read locklessly from sk_setup_caps() */
4841 WRITE_ONCE(dev->gso_max_size, size);
4844 static inline void netif_set_gso_max_segs(struct net_device *dev,
4847 /* dev->gso_max_segs is read locklessly from sk_setup_caps() */
4848 WRITE_ONCE(dev->gso_max_segs, segs);
4851 static inline void netif_set_gro_max_size(struct net_device *dev,
4854 /* This pairs with the READ_ONCE() in skb_gro_receive() */
4855 WRITE_ONCE(dev->gro_max_size, size);
4858 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4859 int pulled_hlen, u16 mac_offset,
4862 skb->protocol = protocol;
4863 skb->encapsulation = 1;
4864 skb_push(skb, pulled_hlen);
4865 skb_reset_transport_header(skb);
4866 skb->mac_header = mac_offset;
4867 skb->network_header = skb->mac_header + mac_len;
4868 skb->mac_len = mac_len;
4871 static inline bool netif_is_macsec(const struct net_device *dev)
4873 return dev->priv_flags & IFF_MACSEC;
4876 static inline bool netif_is_macvlan(const struct net_device *dev)
4878 return dev->priv_flags & IFF_MACVLAN;
4881 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4883 return dev->priv_flags & IFF_MACVLAN_PORT;
4886 static inline bool netif_is_bond_master(const struct net_device *dev)
4888 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4891 static inline bool netif_is_bond_slave(const struct net_device *dev)
4893 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4896 static inline bool netif_supports_nofcs(struct net_device *dev)
4898 return dev->priv_flags & IFF_SUPP_NOFCS;
4901 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4903 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4906 static inline bool netif_is_l3_master(const struct net_device *dev)
4908 return dev->priv_flags & IFF_L3MDEV_MASTER;
4911 static inline bool netif_is_l3_slave(const struct net_device *dev)
4913 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4916 static inline bool netif_is_bridge_master(const struct net_device *dev)
4918 return dev->priv_flags & IFF_EBRIDGE;
4921 static inline bool netif_is_bridge_port(const struct net_device *dev)
4923 return dev->priv_flags & IFF_BRIDGE_PORT;
4926 static inline bool netif_is_ovs_master(const struct net_device *dev)
4928 return dev->priv_flags & IFF_OPENVSWITCH;
4931 static inline bool netif_is_ovs_port(const struct net_device *dev)
4933 return dev->priv_flags & IFF_OVS_DATAPATH;
4936 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
4938 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
4941 static inline bool netif_is_team_master(const struct net_device *dev)
4943 return dev->priv_flags & IFF_TEAM;
4946 static inline bool netif_is_team_port(const struct net_device *dev)
4948 return dev->priv_flags & IFF_TEAM_PORT;
4951 static inline bool netif_is_lag_master(const struct net_device *dev)
4953 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4956 static inline bool netif_is_lag_port(const struct net_device *dev)
4958 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4961 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4963 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4966 static inline bool netif_is_failover(const struct net_device *dev)
4968 return dev->priv_flags & IFF_FAILOVER;
4971 static inline bool netif_is_failover_slave(const struct net_device *dev)
4973 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4976 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4977 static inline void netif_keep_dst(struct net_device *dev)
4979 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4982 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4983 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4985 /* TODO: reserve and use an additional IFF bit, if we get more users */
4986 return netif_is_macsec(dev);
4989 extern struct pernet_operations __net_initdata loopback_net_ops;
4991 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4993 /* netdev_printk helpers, similar to dev_printk */
4995 static inline const char *netdev_name(const struct net_device *dev)
4997 if (!dev->name[0] || strchr(dev->name, '%'))
4998 return "(unnamed net_device)";
5002 static inline bool netdev_unregistering(const struct net_device *dev)
5004 return dev->reg_state == NETREG_UNREGISTERING;
5007 static inline const char *netdev_reg_state(const struct net_device *dev)
5009 switch (dev->reg_state) {
5010 case NETREG_UNINITIALIZED: return " (uninitialized)";
5011 case NETREG_REGISTERED: return "";
5012 case NETREG_UNREGISTERING: return " (unregistering)";
5013 case NETREG_UNREGISTERED: return " (unregistered)";
5014 case NETREG_RELEASED: return " (released)";
5015 case NETREG_DUMMY: return " (dummy)";
5018 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5019 return " (unknown)";
5022 __printf(3, 4) __cold
5023 void netdev_printk(const char *level, const struct net_device *dev,
5024 const char *format, ...);
5025 __printf(2, 3) __cold
5026 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5027 __printf(2, 3) __cold
5028 void netdev_alert(const struct net_device *dev, const char *format, ...);
5029 __printf(2, 3) __cold
5030 void netdev_crit(const struct net_device *dev, const char *format, ...);
5031 __printf(2, 3) __cold
5032 void netdev_err(const struct net_device *dev, const char *format, ...);
5033 __printf(2, 3) __cold
5034 void netdev_warn(const struct net_device *dev, const char *format, ...);
5035 __printf(2, 3) __cold
5036 void netdev_notice(const struct net_device *dev, const char *format, ...);
5037 __printf(2, 3) __cold
5038 void netdev_info(const struct net_device *dev, const char *format, ...);
5040 #define netdev_level_once(level, dev, fmt, ...) \
5042 static bool __section(".data.once") __print_once; \
5044 if (!__print_once) { \
5045 __print_once = true; \
5046 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5050 #define netdev_emerg_once(dev, fmt, ...) \
5051 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5052 #define netdev_alert_once(dev, fmt, ...) \
5053 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5054 #define netdev_crit_once(dev, fmt, ...) \
5055 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5056 #define netdev_err_once(dev, fmt, ...) \
5057 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5058 #define netdev_warn_once(dev, fmt, ...) \
5059 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5060 #define netdev_notice_once(dev, fmt, ...) \
5061 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5062 #define netdev_info_once(dev, fmt, ...) \
5063 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5065 #define MODULE_ALIAS_NETDEV(device) \
5066 MODULE_ALIAS("netdev-" device)
5068 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5069 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5070 #define netdev_dbg(__dev, format, args...) \
5072 dynamic_netdev_dbg(__dev, format, ##args); \
5074 #elif defined(DEBUG)
5075 #define netdev_dbg(__dev, format, args...) \
5076 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5078 #define netdev_dbg(__dev, format, args...) \
5081 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5085 #if defined(VERBOSE_DEBUG)
5086 #define netdev_vdbg netdev_dbg
5089 #define netdev_vdbg(dev, format, args...) \
5092 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5098 * netdev_WARN() acts like dev_printk(), but with the key difference
5099 * of using a WARN/WARN_ON to get the message out, including the
5100 * file/line information and a backtrace.
5102 #define netdev_WARN(dev, format, args...) \
5103 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5104 netdev_reg_state(dev), ##args)
5106 #define netdev_WARN_ONCE(dev, format, args...) \
5107 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5108 netdev_reg_state(dev), ##args)
5110 /* netif printk helpers, similar to netdev_printk */
5112 #define netif_printk(priv, type, level, dev, fmt, args...) \
5114 if (netif_msg_##type(priv)) \
5115 netdev_printk(level, (dev), fmt, ##args); \
5118 #define netif_level(level, priv, type, dev, fmt, args...) \
5120 if (netif_msg_##type(priv)) \
5121 netdev_##level(dev, fmt, ##args); \
5124 #define netif_emerg(priv, type, dev, fmt, args...) \
5125 netif_level(emerg, priv, type, dev, fmt, ##args)
5126 #define netif_alert(priv, type, dev, fmt, args...) \
5127 netif_level(alert, priv, type, dev, fmt, ##args)
5128 #define netif_crit(priv, type, dev, fmt, args...) \
5129 netif_level(crit, priv, type, dev, fmt, ##args)
5130 #define netif_err(priv, type, dev, fmt, args...) \
5131 netif_level(err, priv, type, dev, fmt, ##args)
5132 #define netif_warn(priv, type, dev, fmt, args...) \
5133 netif_level(warn, priv, type, dev, fmt, ##args)
5134 #define netif_notice(priv, type, dev, fmt, args...) \
5135 netif_level(notice, priv, type, dev, fmt, ##args)
5136 #define netif_info(priv, type, dev, fmt, args...) \
5137 netif_level(info, priv, type, dev, fmt, ##args)
5139 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5140 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5141 #define netif_dbg(priv, type, netdev, format, args...) \
5143 if (netif_msg_##type(priv)) \
5144 dynamic_netdev_dbg(netdev, format, ##args); \
5146 #elif defined(DEBUG)
5147 #define netif_dbg(priv, type, dev, format, args...) \
5148 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5150 #define netif_dbg(priv, type, dev, format, args...) \
5153 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5158 /* if @cond then downgrade to debug, else print at @level */
5159 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5162 netif_dbg(priv, type, netdev, fmt, ##args); \
5164 netif_ ## level(priv, type, netdev, fmt, ##args); \
5167 #if defined(VERBOSE_DEBUG)
5168 #define netif_vdbg netif_dbg
5170 #define netif_vdbg(priv, type, dev, format, args...) \
5173 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5179 * The list of packet types we will receive (as opposed to discard)
5180 * and the routines to invoke.
5182 * Why 16. Because with 16 the only overlap we get on a hash of the
5183 * low nibble of the protocol value is RARP/SNAP/X.25.
5197 #define PTYPE_HASH_SIZE (16)
5198 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5200 extern struct list_head ptype_all __read_mostly;
5201 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5203 extern struct net_device *blackhole_netdev;
5205 #endif /* _LINUX_NETDEVICE_H */