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
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>
53 #include <net/net_debug.h>
60 struct ip_tunnel_parm;
61 struct macsec_context;
63 struct netdev_name_node;
68 /* 802.15.4 specific */
71 /* UDP Tunnel offloads */
72 struct udp_tunnel_info;
73 struct udp_tunnel_nic_info;
74 struct udp_tunnel_nic;
78 void synchronize_net(void);
79 void netdev_set_default_ethtool_ops(struct net_device *dev,
80 const struct ethtool_ops *ops);
81 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
83 /* Backlog congestion levels */
84 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
85 #define NET_RX_DROP 1 /* packet dropped */
87 #define MAX_NEST_DEV 8
90 * Transmit return codes: transmit return codes originate from three different
93 * - qdisc return codes
94 * - driver transmit return codes
97 * Drivers are allowed to return any one of those in their hard_start_xmit()
98 * function. Real network devices commonly used with qdiscs should only return
99 * the driver transmit return codes though - when qdiscs are used, the actual
100 * transmission happens asynchronously, so the value is not propagated to
101 * higher layers. Virtual network devices transmit synchronously; in this case
102 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
103 * others are propagated to higher layers.
106 /* qdisc ->enqueue() return codes. */
107 #define NET_XMIT_SUCCESS 0x00
108 #define NET_XMIT_DROP 0x01 /* skb dropped */
109 #define NET_XMIT_CN 0x02 /* congestion notification */
110 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
112 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
113 * indicates that the device will soon be dropping packets, or already drops
114 * some packets of the same priority; prompting us to send less aggressively. */
115 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
116 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
118 /* Driver transmit return codes */
119 #define NETDEV_TX_MASK 0xf0
122 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
123 NETDEV_TX_OK = 0x00, /* driver took care of packet */
124 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
126 typedef enum netdev_tx netdev_tx_t;
129 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
130 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
132 static inline bool dev_xmit_complete(int rc)
135 * Positive cases with an skb consumed by a driver:
136 * - successful transmission (rc == NETDEV_TX_OK)
137 * - error while transmitting (rc < 0)
138 * - error while queueing to a different device (rc & NET_XMIT_MASK)
140 if (likely(rc < NET_XMIT_MASK))
147 * Compute the worst-case header length according to the protocols
151 #if defined(CONFIG_HYPERV_NET)
152 # define LL_MAX_HEADER 128
153 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
154 # if defined(CONFIG_MAC80211_MESH)
155 # define LL_MAX_HEADER 128
157 # define LL_MAX_HEADER 96
160 # define LL_MAX_HEADER 32
163 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
164 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
165 #define MAX_HEADER LL_MAX_HEADER
167 #define MAX_HEADER (LL_MAX_HEADER + 48)
171 * Old network device statistics. Fields are native words
172 * (unsigned long) so they can be read and written atomically.
175 #define NET_DEV_STAT(FIELD) \
177 unsigned long FIELD; \
178 atomic_long_t __##FIELD; \
181 struct net_device_stats {
182 NET_DEV_STAT(rx_packets);
183 NET_DEV_STAT(tx_packets);
184 NET_DEV_STAT(rx_bytes);
185 NET_DEV_STAT(tx_bytes);
186 NET_DEV_STAT(rx_errors);
187 NET_DEV_STAT(tx_errors);
188 NET_DEV_STAT(rx_dropped);
189 NET_DEV_STAT(tx_dropped);
190 NET_DEV_STAT(multicast);
191 NET_DEV_STAT(collisions);
192 NET_DEV_STAT(rx_length_errors);
193 NET_DEV_STAT(rx_over_errors);
194 NET_DEV_STAT(rx_crc_errors);
195 NET_DEV_STAT(rx_frame_errors);
196 NET_DEV_STAT(rx_fifo_errors);
197 NET_DEV_STAT(rx_missed_errors);
198 NET_DEV_STAT(tx_aborted_errors);
199 NET_DEV_STAT(tx_carrier_errors);
200 NET_DEV_STAT(tx_fifo_errors);
201 NET_DEV_STAT(tx_heartbeat_errors);
202 NET_DEV_STAT(tx_window_errors);
203 NET_DEV_STAT(rx_compressed);
204 NET_DEV_STAT(tx_compressed);
208 /* per-cpu stats, allocated on demand.
209 * Try to fit them in a single cache line, for dev_get_stats() sake.
211 struct net_device_core_stats {
212 unsigned long rx_dropped;
213 unsigned long tx_dropped;
214 unsigned long rx_nohandler;
215 unsigned long rx_otherhost_dropped;
216 } __aligned(4 * sizeof(unsigned long));
218 #include <linux/cache.h>
219 #include <linux/skbuff.h>
222 #include <linux/static_key.h>
223 extern struct static_key_false rps_needed;
224 extern struct static_key_false rfs_needed;
231 struct netdev_hw_addr {
232 struct list_head list;
234 unsigned char addr[MAX_ADDR_LEN];
236 #define NETDEV_HW_ADDR_T_LAN 1
237 #define NETDEV_HW_ADDR_T_SAN 2
238 #define NETDEV_HW_ADDR_T_UNICAST 3
239 #define NETDEV_HW_ADDR_T_MULTICAST 4
244 struct rcu_head rcu_head;
247 struct netdev_hw_addr_list {
248 struct list_head list;
251 /* Auxiliary tree for faster lookup on addition and deletion */
255 #define netdev_hw_addr_list_count(l) ((l)->count)
256 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
257 #define netdev_hw_addr_list_for_each(ha, l) \
258 list_for_each_entry(ha, &(l)->list, list)
260 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
261 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
262 #define netdev_for_each_uc_addr(ha, dev) \
263 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
264 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
265 netdev_for_each_uc_addr((_ha), (_dev)) \
268 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
269 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
270 #define netdev_for_each_mc_addr(ha, dev) \
271 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
272 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
273 netdev_for_each_mc_addr((_ha), (_dev)) \
280 /* cached hardware header; allow for machine alignment needs. */
281 #define HH_DATA_MOD 16
282 #define HH_DATA_OFF(__len) \
283 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
284 #define HH_DATA_ALIGN(__len) \
285 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
286 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
289 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
291 * dev->hard_header_len ? (dev->hard_header_len +
292 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
294 * We could use other alignment values, but we must maintain the
295 * relationship HH alignment <= LL alignment.
297 #define LL_RESERVED_SPACE(dev) \
298 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
299 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
300 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
303 int (*create) (struct sk_buff *skb, struct net_device *dev,
304 unsigned short type, const void *daddr,
305 const void *saddr, unsigned int len);
306 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
307 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
308 void (*cache_update)(struct hh_cache *hh,
309 const struct net_device *dev,
310 const unsigned char *haddr);
311 bool (*validate)(const char *ll_header, unsigned int len);
312 __be16 (*parse_protocol)(const struct sk_buff *skb);
315 /* These flag bits are private to the generic network queueing
316 * layer; they may not be explicitly referenced by any other
320 enum netdev_state_t {
322 __LINK_STATE_PRESENT,
323 __LINK_STATE_NOCARRIER,
324 __LINK_STATE_LINKWATCH_PENDING,
325 __LINK_STATE_DORMANT,
326 __LINK_STATE_TESTING,
330 struct list_head list;
335 * size of gro hash buckets, must less than bit number of
336 * napi_struct::gro_bitmask
338 #define GRO_HASH_BUCKETS 8
341 * Structure for NAPI scheduling similar to tasklet but with weighting
344 /* The poll_list must only be managed by the entity which
345 * changes the state of the NAPI_STATE_SCHED bit. This means
346 * whoever atomically sets that bit can add this napi_struct
347 * to the per-CPU poll_list, and whoever clears that bit
348 * can remove from the list right before clearing the bit.
350 struct list_head poll_list;
354 int defer_hard_irqs_count;
355 unsigned long gro_bitmask;
356 int (*poll)(struct napi_struct *, int);
357 #ifdef CONFIG_NETPOLL
360 struct net_device *dev;
361 struct gro_list gro_hash[GRO_HASH_BUCKETS];
363 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
364 int rx_count; /* length of rx_list */
365 struct hrtimer timer;
366 struct list_head dev_list;
367 struct hlist_node napi_hash_node;
368 unsigned int napi_id;
369 struct task_struct *thread;
373 NAPI_STATE_SCHED, /* Poll is scheduled */
374 NAPI_STATE_MISSED, /* reschedule a napi */
375 NAPI_STATE_DISABLE, /* Disable pending */
376 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
377 NAPI_STATE_LISTED, /* NAPI added to system lists */
378 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
379 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
380 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
381 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
382 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
386 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
387 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
388 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
389 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
390 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
391 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
392 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
393 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
394 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
395 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
405 typedef enum gro_result gro_result_t;
408 * enum rx_handler_result - Possible return values for rx_handlers.
409 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
411 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
412 * case skb->dev was changed by rx_handler.
413 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
414 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
416 * rx_handlers are functions called from inside __netif_receive_skb(), to do
417 * special processing of the skb, prior to delivery to protocol handlers.
419 * Currently, a net_device can only have a single rx_handler registered. Trying
420 * to register a second rx_handler will return -EBUSY.
422 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
423 * To unregister a rx_handler on a net_device, use
424 * netdev_rx_handler_unregister().
426 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
429 * If the rx_handler consumed the skb in some way, it should return
430 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
431 * the skb to be delivered in some other way.
433 * If the rx_handler changed skb->dev, to divert the skb to another
434 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
435 * new device will be called if it exists.
437 * If the rx_handler decides the skb should be ignored, it should return
438 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
439 * are registered on exact device (ptype->dev == skb->dev).
441 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
442 * delivered, it should return RX_HANDLER_PASS.
444 * A device without a registered rx_handler will behave as if rx_handler
445 * returned RX_HANDLER_PASS.
448 enum rx_handler_result {
454 typedef enum rx_handler_result rx_handler_result_t;
455 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
457 void __napi_schedule(struct napi_struct *n);
458 void __napi_schedule_irqoff(struct napi_struct *n);
460 static inline bool napi_disable_pending(struct napi_struct *n)
462 return test_bit(NAPI_STATE_DISABLE, &n->state);
465 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
467 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
470 bool napi_schedule_prep(struct napi_struct *n);
473 * napi_schedule - schedule NAPI poll
476 * Schedule NAPI poll routine to be called if it is not already
479 static inline void napi_schedule(struct napi_struct *n)
481 if (napi_schedule_prep(n))
486 * napi_schedule_irqoff - schedule NAPI poll
489 * Variant of napi_schedule(), assuming hard irqs are masked.
491 static inline void napi_schedule_irqoff(struct napi_struct *n)
493 if (napi_schedule_prep(n))
494 __napi_schedule_irqoff(n);
497 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
498 static inline bool napi_reschedule(struct napi_struct *napi)
500 if (napi_schedule_prep(napi)) {
501 __napi_schedule(napi);
507 bool napi_complete_done(struct napi_struct *n, int work_done);
509 * napi_complete - NAPI processing complete
512 * Mark NAPI processing as complete.
513 * Consider using napi_complete_done() instead.
514 * Return false if device should avoid rearming interrupts.
516 static inline bool napi_complete(struct napi_struct *n)
518 return napi_complete_done(n, 0);
521 int dev_set_threaded(struct net_device *dev, bool threaded);
524 * napi_disable - prevent NAPI from scheduling
527 * Stop NAPI from being scheduled on this context.
528 * Waits till any outstanding processing completes.
530 void napi_disable(struct napi_struct *n);
532 void napi_enable(struct napi_struct *n);
535 * napi_synchronize - wait until NAPI is not running
538 * Wait until NAPI is done being scheduled on this context.
539 * Waits till any outstanding processing completes but
540 * does not disable future activations.
542 static inline void napi_synchronize(const struct napi_struct *n)
544 if (IS_ENABLED(CONFIG_SMP))
545 while (test_bit(NAPI_STATE_SCHED, &n->state))
552 * napi_if_scheduled_mark_missed - if napi is running, set the
556 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
559 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
561 unsigned long val, new;
563 val = READ_ONCE(n->state);
565 if (val & NAPIF_STATE_DISABLE)
568 if (!(val & NAPIF_STATE_SCHED))
571 new = val | NAPIF_STATE_MISSED;
572 } while (!try_cmpxchg(&n->state, &val, new));
577 enum netdev_queue_state_t {
578 __QUEUE_STATE_DRV_XOFF,
579 __QUEUE_STATE_STACK_XOFF,
580 __QUEUE_STATE_FROZEN,
583 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
584 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
585 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
587 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
588 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
590 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
594 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
595 * netif_tx_* functions below are used to manipulate this flag. The
596 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
597 * queue independently. The netif_xmit_*stopped functions below are called
598 * to check if the queue has been stopped by the driver or stack (either
599 * of the XOFF bits are set in the state). Drivers should not need to call
600 * netif_xmit*stopped functions, they should only be using netif_tx_*.
603 struct netdev_queue {
607 struct net_device *dev;
608 netdevice_tracker dev_tracker;
610 struct Qdisc __rcu *qdisc;
611 struct Qdisc *qdisc_sleeping;
615 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
618 unsigned long tx_maxrate;
620 * Number of TX timeouts for this queue
621 * (/sys/class/net/DEV/Q/trans_timeout)
623 atomic_long_t trans_timeout;
625 /* Subordinate device that the queue has been assigned to */
626 struct net_device *sb_dev;
627 #ifdef CONFIG_XDP_SOCKETS
628 struct xsk_buff_pool *pool;
633 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
636 * Time (in jiffies) of last Tx
638 unsigned long trans_start;
645 } ____cacheline_aligned_in_smp;
647 extern int sysctl_fb_tunnels_only_for_init_net;
648 extern int sysctl_devconf_inherit_init_net;
651 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
652 * == 1 : For initns only
655 static inline bool net_has_fallback_tunnels(const struct net *net)
657 #if IS_ENABLED(CONFIG_SYSCTL)
658 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
660 return !fb_tunnels_only_for_init_net ||
661 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
667 static inline int net_inherit_devconf(void)
669 #if IS_ENABLED(CONFIG_SYSCTL)
670 return READ_ONCE(sysctl_devconf_inherit_init_net);
676 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
678 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
685 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
687 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
694 * This structure holds an RPS map which can be of variable length. The
695 * map is an array of CPUs.
702 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
705 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
706 * tail pointer for that CPU's input queue at the time of last enqueue, and
707 * a hardware filter index.
709 struct rps_dev_flow {
712 unsigned int last_qtail;
714 #define RPS_NO_FILTER 0xffff
717 * The rps_dev_flow_table structure contains a table of flow mappings.
719 struct rps_dev_flow_table {
722 struct rps_dev_flow flows[];
724 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
725 ((_num) * sizeof(struct rps_dev_flow)))
728 * The rps_sock_flow_table contains mappings of flows to the last CPU
729 * on which they were processed by the application (set in recvmsg).
730 * Each entry is a 32bit value. Upper part is the high-order bits
731 * of flow hash, lower part is CPU number.
732 * rps_cpu_mask is used to partition the space, depending on number of
733 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
734 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
735 * meaning we use 32-6=26 bits for the hash.
737 struct rps_sock_flow_table {
740 u32 ents[] ____cacheline_aligned_in_smp;
742 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
744 #define RPS_NO_CPU 0xffff
746 extern u32 rps_cpu_mask;
747 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
749 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
753 unsigned int index = hash & table->mask;
754 u32 val = hash & ~rps_cpu_mask;
756 /* We only give a hint, preemption can change CPU under us */
757 val |= raw_smp_processor_id();
759 if (table->ents[index] != val)
760 table->ents[index] = val;
764 #ifdef CONFIG_RFS_ACCEL
765 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
768 #endif /* CONFIG_RPS */
770 /* This structure contains an instance of an RX queue. */
771 struct netdev_rx_queue {
772 struct xdp_rxq_info xdp_rxq;
774 struct rps_map __rcu *rps_map;
775 struct rps_dev_flow_table __rcu *rps_flow_table;
778 struct net_device *dev;
779 netdevice_tracker dev_tracker;
781 #ifdef CONFIG_XDP_SOCKETS
782 struct xsk_buff_pool *pool;
784 } ____cacheline_aligned_in_smp;
787 * RX queue sysfs structures and functions.
789 struct rx_queue_attribute {
790 struct attribute attr;
791 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
792 ssize_t (*store)(struct netdev_rx_queue *queue,
793 const char *buf, size_t len);
796 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
805 * This structure holds an XPS map which can be of variable length. The
806 * map is an array of queues.
810 unsigned int alloc_len;
814 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
815 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
816 - sizeof(struct xps_map)) / sizeof(u16))
819 * This structure holds all XPS maps for device. Maps are indexed by CPU.
821 * We keep track of the number of cpus/rxqs used when the struct is allocated,
822 * in nr_ids. This will help not accessing out-of-bound memory.
824 * We keep track of the number of traffic classes used when the struct is
825 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
826 * not crossing its upper bound, as the original dev->num_tc can be updated in
829 struct xps_dev_maps {
833 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
836 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
837 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
839 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
840 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
842 #endif /* CONFIG_XPS */
844 #define TC_MAX_QUEUE 16
845 #define TC_BITMASK 15
846 /* HW offloaded queuing disciplines txq count and offset maps */
847 struct netdev_tc_txq {
852 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
854 * This structure is to hold information about the device
855 * configured to run FCoE protocol stack.
857 struct netdev_fcoe_hbainfo {
858 char manufacturer[64];
859 char serial_number[64];
860 char hardware_version[64];
861 char driver_version[64];
862 char optionrom_version[64];
863 char firmware_version[64];
865 char model_description[256];
869 #define MAX_PHYS_ITEM_ID_LEN 32
871 /* This structure holds a unique identifier to identify some
872 * physical item (port for example) used by a netdevice.
874 struct netdev_phys_item_id {
875 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
876 unsigned char id_len;
879 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
880 struct netdev_phys_item_id *b)
882 return a->id_len == b->id_len &&
883 memcmp(a->id, b->id, a->id_len) == 0;
886 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
888 struct net_device *sb_dev);
890 enum net_device_path_type {
891 DEV_PATH_ETHERNET = 0,
899 struct net_device_path {
900 enum net_device_path_type type;
901 const struct net_device *dev;
910 DEV_PATH_BR_VLAN_KEEP,
911 DEV_PATH_BR_VLAN_TAG,
912 DEV_PATH_BR_VLAN_UNTAG,
913 DEV_PATH_BR_VLAN_UNTAG_HW,
931 #define NET_DEVICE_PATH_STACK_MAX 5
932 #define NET_DEVICE_PATH_VLAN_MAX 2
934 struct net_device_path_stack {
936 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
939 struct net_device_path_ctx {
940 const struct net_device *dev;
947 } vlan[NET_DEVICE_PATH_VLAN_MAX];
952 TC_SETUP_QDISC_MQPRIO,
955 TC_SETUP_CLSMATCHALL,
965 TC_SETUP_QDISC_TAPRIO,
974 /* These structures hold the attributes of bpf state that are being passed
975 * to the netdevice through the bpf op.
977 enum bpf_netdev_command {
978 /* Set or clear a bpf program used in the earliest stages of packet
979 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
980 * is responsible for calling bpf_prog_put on any old progs that are
981 * stored. In case of error, the callee need not release the new prog
982 * reference, but on success it takes ownership and must bpf_prog_put
983 * when it is no longer used.
987 /* BPF program for offload callbacks, invoked at program load time. */
988 BPF_OFFLOAD_MAP_ALLOC,
989 BPF_OFFLOAD_MAP_FREE,
993 struct bpf_prog_offload_ops;
994 struct netlink_ext_ack;
996 struct xdp_dev_bulk_queue;
1006 struct bpf_xdp_entity {
1007 struct bpf_prog *prog;
1008 struct bpf_xdp_link *link;
1012 enum bpf_netdev_command command;
1014 /* XDP_SETUP_PROG */
1017 struct bpf_prog *prog;
1018 struct netlink_ext_ack *extack;
1020 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1022 struct bpf_offloaded_map *offmap;
1024 /* XDP_SETUP_XSK_POOL */
1026 struct xsk_buff_pool *pool;
1032 /* Flags for ndo_xsk_wakeup. */
1033 #define XDP_WAKEUP_RX (1 << 0)
1034 #define XDP_WAKEUP_TX (1 << 1)
1036 #ifdef CONFIG_XFRM_OFFLOAD
1037 struct xfrmdev_ops {
1038 int (*xdo_dev_state_add) (struct xfrm_state *x);
1039 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1040 void (*xdo_dev_state_free) (struct xfrm_state *x);
1041 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1042 struct xfrm_state *x);
1043 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1044 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1045 int (*xdo_dev_policy_add) (struct xfrm_policy *x);
1046 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1047 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1051 struct dev_ifalias {
1052 struct rcu_head rcuhead;
1059 struct netdev_net_notifier {
1060 struct list_head list;
1061 struct notifier_block *nb;
1065 * This structure defines the management hooks for network devices.
1066 * The following hooks can be defined; unless noted otherwise, they are
1067 * optional and can be filled with a null pointer.
1069 * int (*ndo_init)(struct net_device *dev);
1070 * This function is called once when a network device is registered.
1071 * The network device can use this for any late stage initialization
1072 * or semantic validation. It can fail with an error code which will
1073 * be propagated back to register_netdev.
1075 * void (*ndo_uninit)(struct net_device *dev);
1076 * This function is called when device is unregistered or when registration
1077 * fails. It is not called if init fails.
1079 * int (*ndo_open)(struct net_device *dev);
1080 * This function is called when a network device transitions to the up
1083 * int (*ndo_stop)(struct net_device *dev);
1084 * This function is called when a network device transitions to the down
1087 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1088 * struct net_device *dev);
1089 * Called when a packet needs to be transmitted.
1090 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1091 * the queue before that can happen; it's for obsolete devices and weird
1092 * corner cases, but the stack really does a non-trivial amount
1093 * of useless work if you return NETDEV_TX_BUSY.
1094 * Required; cannot be NULL.
1096 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1097 * struct net_device *dev
1098 * netdev_features_t features);
1099 * Called by core transmit path to determine if device is capable of
1100 * performing offload operations on a given packet. This is to give
1101 * the device an opportunity to implement any restrictions that cannot
1102 * be otherwise expressed by feature flags. The check is called with
1103 * the set of features that the stack has calculated and it returns
1104 * those the driver believes to be appropriate.
1106 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1107 * struct net_device *sb_dev);
1108 * Called to decide which queue to use when device supports multiple
1111 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1112 * This function is called to allow device receiver to make
1113 * changes to configuration when multicast or promiscuous is enabled.
1115 * void (*ndo_set_rx_mode)(struct net_device *dev);
1116 * This function is called device changes address list filtering.
1117 * If driver handles unicast address filtering, it should set
1118 * IFF_UNICAST_FLT in its priv_flags.
1120 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1121 * This function is called when the Media Access Control address
1122 * needs to be changed. If this interface is not defined, the
1123 * MAC address can not be changed.
1125 * int (*ndo_validate_addr)(struct net_device *dev);
1126 * Test if Media Access Control address is valid for the device.
1128 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1129 * Old-style ioctl entry point. This is used internally by the
1130 * appletalk and ieee802154 subsystems but is no longer called by
1131 * the device ioctl handler.
1133 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1134 * Used by the bonding driver for its device specific ioctls:
1135 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1136 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1138 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1139 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1140 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1142 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1143 * Used to set network devices bus interface parameters. This interface
1144 * is retained for legacy reasons; new devices should use the bus
1145 * interface (PCI) for low level management.
1147 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1148 * Called when a user wants to change the Maximum Transfer Unit
1151 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1152 * Callback used when the transmitter has not made any progress
1153 * for dev->watchdog ticks.
1155 * void (*ndo_get_stats64)(struct net_device *dev,
1156 * struct rtnl_link_stats64 *storage);
1157 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1158 * Called when a user wants to get the network device usage
1159 * statistics. Drivers must do one of the following:
1160 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1161 * rtnl_link_stats64 structure passed by the caller.
1162 * 2. Define @ndo_get_stats to update a net_device_stats structure
1163 * (which should normally be dev->stats) and return a pointer to
1164 * it. The structure may be changed asynchronously only if each
1165 * field is written atomically.
1166 * 3. Update dev->stats asynchronously and atomically, and define
1167 * neither operation.
1169 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1170 * Return true if this device supports offload stats of this attr_id.
1172 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1174 * Get statistics for offload operations by attr_id. Write it into the
1175 * attr_data pointer.
1177 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1178 * If device supports VLAN filtering this function is called when a
1179 * VLAN id is registered.
1181 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1182 * If device supports VLAN filtering this function is called when a
1183 * VLAN id is unregistered.
1185 * void (*ndo_poll_controller)(struct net_device *dev);
1187 * SR-IOV management functions.
1188 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1189 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1190 * u8 qos, __be16 proto);
1191 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1193 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1194 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1195 * int (*ndo_get_vf_config)(struct net_device *dev,
1196 * int vf, struct ifla_vf_info *ivf);
1197 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1198 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1199 * struct nlattr *port[]);
1201 * Enable or disable the VF ability to query its RSS Redirection Table and
1202 * Hash Key. This is needed since on some devices VF share this information
1203 * with PF and querying it may introduce a theoretical security risk.
1204 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1205 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1206 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1208 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1209 * This is always called from the stack with the rtnl lock held and netif
1210 * tx queues stopped. This allows the netdevice to perform queue
1211 * management safely.
1213 * Fiber Channel over Ethernet (FCoE) offload functions.
1214 * int (*ndo_fcoe_enable)(struct net_device *dev);
1215 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1216 * so the underlying device can perform whatever needed configuration or
1217 * initialization to support acceleration of FCoE traffic.
1219 * int (*ndo_fcoe_disable)(struct net_device *dev);
1220 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1221 * so the underlying device can perform whatever needed clean-ups to
1222 * stop supporting acceleration of FCoE traffic.
1224 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1225 * struct scatterlist *sgl, unsigned int sgc);
1226 * Called when the FCoE Initiator wants to initialize an I/O that
1227 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1228 * perform necessary setup and returns 1 to indicate the device is set up
1229 * successfully to perform DDP on this I/O, otherwise this returns 0.
1231 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1232 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1233 * indicated by the FC exchange id 'xid', so the underlying device can
1234 * clean up and reuse resources for later DDP requests.
1236 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1237 * struct scatterlist *sgl, unsigned int sgc);
1238 * Called when the FCoE Target wants to initialize an I/O that
1239 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1240 * perform necessary setup and returns 1 to indicate the device is set up
1241 * successfully to perform DDP on this I/O, otherwise this returns 0.
1243 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1244 * struct netdev_fcoe_hbainfo *hbainfo);
1245 * Called when the FCoE Protocol stack wants information on the underlying
1246 * device. This information is utilized by the FCoE protocol stack to
1247 * register attributes with Fiber Channel management service as per the
1248 * FC-GS Fabric Device Management Information(FDMI) specification.
1250 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1251 * Called when the underlying device wants to override default World Wide
1252 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1253 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1254 * protocol stack to use.
1257 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1258 * u16 rxq_index, u32 flow_id);
1259 * Set hardware filter for RFS. rxq_index is the target queue index;
1260 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1261 * Return the filter ID on success, or a negative error code.
1263 * Slave management functions (for bridge, bonding, etc).
1264 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1265 * Called to make another netdev an underling.
1267 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1268 * Called to release previously enslaved netdev.
1270 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1271 * struct sk_buff *skb,
1273 * Get the xmit slave of master device. If all_slaves is true, function
1274 * assume all the slaves can transmit.
1276 * Feature/offload setting functions.
1277 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1278 * netdev_features_t features);
1279 * Adjusts the requested feature flags according to device-specific
1280 * constraints, and returns the resulting flags. Must not modify
1283 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1284 * Called to update device configuration to new features. Passed
1285 * feature set might be less than what was returned by ndo_fix_features()).
1286 * Must return >0 or -errno if it changed dev->features itself.
1288 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1289 * struct net_device *dev,
1290 * const unsigned char *addr, u16 vid, u16 flags,
1291 * struct netlink_ext_ack *extack);
1292 * Adds an FDB entry to dev for addr.
1293 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1294 * struct net_device *dev,
1295 * const unsigned char *addr, u16 vid)
1296 * Deletes the FDB entry from dev coresponding to addr.
1297 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1298 * struct net_device *dev,
1300 * struct netlink_ext_ack *extack);
1301 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1302 * struct net_device *dev, struct net_device *filter_dev,
1304 * Used to add FDB entries to dump requests. Implementers should add
1305 * entries to skb and update idx with the number of entries.
1307 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1308 * u16 flags, struct netlink_ext_ack *extack)
1309 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1310 * struct net_device *dev, u32 filter_mask,
1312 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1315 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1316 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1317 * which do not represent real hardware may define this to allow their
1318 * userspace components to manage their virtual carrier state. Devices
1319 * that determine carrier state from physical hardware properties (eg
1320 * network cables) or protocol-dependent mechanisms (eg
1321 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1323 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1324 * struct netdev_phys_item_id *ppid);
1325 * Called to get ID of physical port of this device. If driver does
1326 * not implement this, it is assumed that the hw is not able to have
1327 * multiple net devices on single physical port.
1329 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1330 * struct netdev_phys_item_id *ppid)
1331 * Called to get the parent ID of the physical port of this device.
1333 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1334 * struct net_device *dev)
1335 * Called by upper layer devices to accelerate switching or other
1336 * station functionality into hardware. 'pdev is the lowerdev
1337 * to use for the offload and 'dev' is the net device that will
1338 * back the offload. Returns a pointer to the private structure
1339 * the upper layer will maintain.
1340 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1341 * Called by upper layer device to delete the station created
1342 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1343 * the station and priv is the structure returned by the add
1345 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1346 * int queue_index, u32 maxrate);
1347 * Called when a user wants to set a max-rate limitation of specific
1349 * int (*ndo_get_iflink)(const struct net_device *dev);
1350 * Called to get the iflink value of this device.
1351 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1352 * This function is used to get egress tunnel information for given skb.
1353 * This is useful for retrieving outer tunnel header parameters while
1355 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1356 * This function is used to specify the headroom that the skb must
1357 * consider when allocation skb during packet reception. Setting
1358 * appropriate rx headroom value allows avoiding skb head copy on
1359 * forward. Setting a negative value resets the rx headroom to the
1361 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1362 * This function is used to set or query state related to XDP on the
1363 * netdevice and manage BPF offload. See definition of
1364 * enum bpf_netdev_command for details.
1365 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1367 * This function is used to submit @n XDP packets for transmit on a
1368 * netdevice. Returns number of frames successfully transmitted, frames
1369 * that got dropped are freed/returned via xdp_return_frame().
1370 * Returns negative number, means general error invoking ndo, meaning
1371 * no frames were xmit'ed and core-caller will free all frames.
1372 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1373 * struct xdp_buff *xdp);
1374 * Get the xmit slave of master device based on the xdp_buff.
1375 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1376 * This function is used to wake up the softirq, ksoftirqd or kthread
1377 * responsible for sending and/or receiving packets on a specific
1378 * queue id bound to an AF_XDP socket. The flags field specifies if
1379 * only RX, only Tx, or both should be woken up using the flags
1380 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1381 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1383 * Add, change, delete or get information on an IPv4 tunnel.
1384 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1385 * If a device is paired with a peer device, return the peer instance.
1386 * The caller must be under RCU read context.
1387 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1388 * Get the forwarding path to reach the real device from the HW destination address
1389 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1390 * const struct skb_shared_hwtstamps *hwtstamps,
1392 * Get hardware timestamp based on normal/adjustable time or free running
1393 * cycle counter. This function is required if physical clock supports a
1394 * free running cycle counter.
1396 struct net_device_ops {
1397 int (*ndo_init)(struct net_device *dev);
1398 void (*ndo_uninit)(struct net_device *dev);
1399 int (*ndo_open)(struct net_device *dev);
1400 int (*ndo_stop)(struct net_device *dev);
1401 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1402 struct net_device *dev);
1403 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1404 struct net_device *dev,
1405 netdev_features_t features);
1406 u16 (*ndo_select_queue)(struct net_device *dev,
1407 struct sk_buff *skb,
1408 struct net_device *sb_dev);
1409 void (*ndo_change_rx_flags)(struct net_device *dev,
1411 void (*ndo_set_rx_mode)(struct net_device *dev);
1412 int (*ndo_set_mac_address)(struct net_device *dev,
1414 int (*ndo_validate_addr)(struct net_device *dev);
1415 int (*ndo_do_ioctl)(struct net_device *dev,
1416 struct ifreq *ifr, int cmd);
1417 int (*ndo_eth_ioctl)(struct net_device *dev,
1418 struct ifreq *ifr, int cmd);
1419 int (*ndo_siocbond)(struct net_device *dev,
1420 struct ifreq *ifr, int cmd);
1421 int (*ndo_siocwandev)(struct net_device *dev,
1422 struct if_settings *ifs);
1423 int (*ndo_siocdevprivate)(struct net_device *dev,
1425 void __user *data, int cmd);
1426 int (*ndo_set_config)(struct net_device *dev,
1428 int (*ndo_change_mtu)(struct net_device *dev,
1430 int (*ndo_neigh_setup)(struct net_device *dev,
1431 struct neigh_parms *);
1432 void (*ndo_tx_timeout) (struct net_device *dev,
1433 unsigned int txqueue);
1435 void (*ndo_get_stats64)(struct net_device *dev,
1436 struct rtnl_link_stats64 *storage);
1437 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1438 int (*ndo_get_offload_stats)(int attr_id,
1439 const struct net_device *dev,
1441 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1443 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1444 __be16 proto, u16 vid);
1445 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1446 __be16 proto, u16 vid);
1447 #ifdef CONFIG_NET_POLL_CONTROLLER
1448 void (*ndo_poll_controller)(struct net_device *dev);
1449 int (*ndo_netpoll_setup)(struct net_device *dev,
1450 struct netpoll_info *info);
1451 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1453 int (*ndo_set_vf_mac)(struct net_device *dev,
1454 int queue, u8 *mac);
1455 int (*ndo_set_vf_vlan)(struct net_device *dev,
1456 int queue, u16 vlan,
1457 u8 qos, __be16 proto);
1458 int (*ndo_set_vf_rate)(struct net_device *dev,
1459 int vf, int min_tx_rate,
1461 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1462 int vf, bool setting);
1463 int (*ndo_set_vf_trust)(struct net_device *dev,
1464 int vf, bool setting);
1465 int (*ndo_get_vf_config)(struct net_device *dev,
1467 struct ifla_vf_info *ivf);
1468 int (*ndo_set_vf_link_state)(struct net_device *dev,
1469 int vf, int link_state);
1470 int (*ndo_get_vf_stats)(struct net_device *dev,
1472 struct ifla_vf_stats
1474 int (*ndo_set_vf_port)(struct net_device *dev,
1476 struct nlattr *port[]);
1477 int (*ndo_get_vf_port)(struct net_device *dev,
1478 int vf, struct sk_buff *skb);
1479 int (*ndo_get_vf_guid)(struct net_device *dev,
1481 struct ifla_vf_guid *node_guid,
1482 struct ifla_vf_guid *port_guid);
1483 int (*ndo_set_vf_guid)(struct net_device *dev,
1486 int (*ndo_set_vf_rss_query_en)(
1487 struct net_device *dev,
1488 int vf, bool setting);
1489 int (*ndo_setup_tc)(struct net_device *dev,
1490 enum tc_setup_type type,
1492 #if IS_ENABLED(CONFIG_FCOE)
1493 int (*ndo_fcoe_enable)(struct net_device *dev);
1494 int (*ndo_fcoe_disable)(struct net_device *dev);
1495 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1497 struct scatterlist *sgl,
1499 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1501 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1503 struct scatterlist *sgl,
1505 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1506 struct netdev_fcoe_hbainfo *hbainfo);
1509 #if IS_ENABLED(CONFIG_LIBFCOE)
1510 #define NETDEV_FCOE_WWNN 0
1511 #define NETDEV_FCOE_WWPN 1
1512 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1513 u64 *wwn, int type);
1516 #ifdef CONFIG_RFS_ACCEL
1517 int (*ndo_rx_flow_steer)(struct net_device *dev,
1518 const struct sk_buff *skb,
1522 int (*ndo_add_slave)(struct net_device *dev,
1523 struct net_device *slave_dev,
1524 struct netlink_ext_ack *extack);
1525 int (*ndo_del_slave)(struct net_device *dev,
1526 struct net_device *slave_dev);
1527 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1528 struct sk_buff *skb,
1530 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1532 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1533 netdev_features_t features);
1534 int (*ndo_set_features)(struct net_device *dev,
1535 netdev_features_t features);
1536 int (*ndo_neigh_construct)(struct net_device *dev,
1537 struct neighbour *n);
1538 void (*ndo_neigh_destroy)(struct net_device *dev,
1539 struct neighbour *n);
1541 int (*ndo_fdb_add)(struct ndmsg *ndm,
1542 struct nlattr *tb[],
1543 struct net_device *dev,
1544 const unsigned char *addr,
1547 struct netlink_ext_ack *extack);
1548 int (*ndo_fdb_del)(struct ndmsg *ndm,
1549 struct nlattr *tb[],
1550 struct net_device *dev,
1551 const unsigned char *addr,
1552 u16 vid, struct netlink_ext_ack *extack);
1553 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1554 struct nlattr *tb[],
1555 struct net_device *dev,
1557 struct netlink_ext_ack *extack);
1558 int (*ndo_fdb_dump)(struct sk_buff *skb,
1559 struct netlink_callback *cb,
1560 struct net_device *dev,
1561 struct net_device *filter_dev,
1563 int (*ndo_fdb_get)(struct sk_buff *skb,
1564 struct nlattr *tb[],
1565 struct net_device *dev,
1566 const unsigned char *addr,
1567 u16 vid, u32 portid, u32 seq,
1568 struct netlink_ext_ack *extack);
1569 int (*ndo_bridge_setlink)(struct net_device *dev,
1570 struct nlmsghdr *nlh,
1572 struct netlink_ext_ack *extack);
1573 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1575 struct net_device *dev,
1578 int (*ndo_bridge_dellink)(struct net_device *dev,
1579 struct nlmsghdr *nlh,
1581 int (*ndo_change_carrier)(struct net_device *dev,
1583 int (*ndo_get_phys_port_id)(struct net_device *dev,
1584 struct netdev_phys_item_id *ppid);
1585 int (*ndo_get_port_parent_id)(struct net_device *dev,
1586 struct netdev_phys_item_id *ppid);
1587 int (*ndo_get_phys_port_name)(struct net_device *dev,
1588 char *name, size_t len);
1589 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1590 struct net_device *dev);
1591 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1594 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1597 int (*ndo_get_iflink)(const struct net_device *dev);
1598 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1599 struct sk_buff *skb);
1600 void (*ndo_set_rx_headroom)(struct net_device *dev,
1601 int needed_headroom);
1602 int (*ndo_bpf)(struct net_device *dev,
1603 struct netdev_bpf *bpf);
1604 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1605 struct xdp_frame **xdp,
1607 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1608 struct xdp_buff *xdp);
1609 int (*ndo_xsk_wakeup)(struct net_device *dev,
1610 u32 queue_id, u32 flags);
1611 int (*ndo_tunnel_ctl)(struct net_device *dev,
1612 struct ip_tunnel_parm *p, int cmd);
1613 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1614 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1615 struct net_device_path *path);
1616 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1617 const struct skb_shared_hwtstamps *hwtstamps,
1622 * enum netdev_priv_flags - &struct net_device priv_flags
1624 * These are the &struct net_device, they are only set internally
1625 * by drivers and used in the kernel. These flags are invisible to
1626 * userspace; this means that the order of these flags can change
1627 * during any kernel release.
1629 * You should have a pretty good reason to be extending these flags.
1631 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1632 * @IFF_EBRIDGE: Ethernet bridging device
1633 * @IFF_BONDING: bonding master or slave
1634 * @IFF_ISATAP: ISATAP interface (RFC4214)
1635 * @IFF_WAN_HDLC: WAN HDLC device
1636 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1638 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1639 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1640 * @IFF_MACVLAN_PORT: device used as macvlan port
1641 * @IFF_BRIDGE_PORT: device used as bridge port
1642 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1643 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1644 * @IFF_UNICAST_FLT: Supports unicast filtering
1645 * @IFF_TEAM_PORT: device used as team port
1646 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1647 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1648 * change when it's running
1649 * @IFF_MACVLAN: Macvlan device
1650 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1651 * underlying stacked devices
1652 * @IFF_L3MDEV_MASTER: device is an L3 master device
1653 * @IFF_NO_QUEUE: device can run without qdisc attached
1654 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1655 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1656 * @IFF_TEAM: device is a team device
1657 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1658 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1659 * entity (i.e. the master device for bridged veth)
1660 * @IFF_MACSEC: device is a MACsec device
1661 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1662 * @IFF_FAILOVER: device is a failover master device
1663 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1664 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1665 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1666 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1667 * skb_headlen(skb) == 0 (data starts from frag0)
1668 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1670 enum netdev_priv_flags {
1671 IFF_802_1Q_VLAN = 1<<0,
1675 IFF_WAN_HDLC = 1<<4,
1676 IFF_XMIT_DST_RELEASE = 1<<5,
1677 IFF_DONT_BRIDGE = 1<<6,
1678 IFF_DISABLE_NETPOLL = 1<<7,
1679 IFF_MACVLAN_PORT = 1<<8,
1680 IFF_BRIDGE_PORT = 1<<9,
1681 IFF_OVS_DATAPATH = 1<<10,
1682 IFF_TX_SKB_SHARING = 1<<11,
1683 IFF_UNICAST_FLT = 1<<12,
1684 IFF_TEAM_PORT = 1<<13,
1685 IFF_SUPP_NOFCS = 1<<14,
1686 IFF_LIVE_ADDR_CHANGE = 1<<15,
1687 IFF_MACVLAN = 1<<16,
1688 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1689 IFF_L3MDEV_MASTER = 1<<18,
1690 IFF_NO_QUEUE = 1<<19,
1691 IFF_OPENVSWITCH = 1<<20,
1692 IFF_L3MDEV_SLAVE = 1<<21,
1694 IFF_RXFH_CONFIGURED = 1<<23,
1695 IFF_PHONY_HEADROOM = 1<<24,
1697 IFF_NO_RX_HANDLER = 1<<26,
1698 IFF_FAILOVER = 1<<27,
1699 IFF_FAILOVER_SLAVE = 1<<28,
1700 IFF_L3MDEV_RX_HANDLER = 1<<29,
1701 IFF_NO_ADDRCONF = BIT_ULL(30),
1702 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1703 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1706 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1707 #define IFF_EBRIDGE IFF_EBRIDGE
1708 #define IFF_BONDING IFF_BONDING
1709 #define IFF_ISATAP IFF_ISATAP
1710 #define IFF_WAN_HDLC IFF_WAN_HDLC
1711 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1712 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1713 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1714 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1715 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1716 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1717 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1718 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1719 #define IFF_TEAM_PORT IFF_TEAM_PORT
1720 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1721 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1722 #define IFF_MACVLAN IFF_MACVLAN
1723 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1724 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1725 #define IFF_NO_QUEUE IFF_NO_QUEUE
1726 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1727 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1728 #define IFF_TEAM IFF_TEAM
1729 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1730 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1731 #define IFF_MACSEC IFF_MACSEC
1732 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1733 #define IFF_FAILOVER IFF_FAILOVER
1734 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1735 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1736 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1738 /* Specifies the type of the struct net_device::ml_priv pointer */
1739 enum netdev_ml_priv_type {
1745 * struct net_device - The DEVICE structure.
1747 * Actually, this whole structure is a big mistake. It mixes I/O
1748 * data with strictly "high-level" data, and it has to know about
1749 * almost every data structure used in the INET module.
1751 * @name: This is the first field of the "visible" part of this structure
1752 * (i.e. as seen by users in the "Space.c" file). It is the name
1755 * @name_node: Name hashlist node
1756 * @ifalias: SNMP alias
1757 * @mem_end: Shared memory end
1758 * @mem_start: Shared memory start
1759 * @base_addr: Device I/O address
1760 * @irq: Device IRQ number
1762 * @state: Generic network queuing layer state, see netdev_state_t
1763 * @dev_list: The global list of network devices
1764 * @napi_list: List entry used for polling NAPI devices
1765 * @unreg_list: List entry when we are unregistering the
1766 * device; see the function unregister_netdev
1767 * @close_list: List entry used when we are closing the device
1768 * @ptype_all: Device-specific packet handlers for all protocols
1769 * @ptype_specific: Device-specific, protocol-specific packet handlers
1771 * @adj_list: Directly linked devices, like slaves for bonding
1772 * @features: Currently active device features
1773 * @hw_features: User-changeable features
1775 * @wanted_features: User-requested features
1776 * @vlan_features: Mask of features inheritable by VLAN devices
1778 * @hw_enc_features: Mask of features inherited by encapsulating devices
1779 * This field indicates what encapsulation
1780 * offloads the hardware is capable of doing,
1781 * and drivers will need to set them appropriately.
1783 * @mpls_features: Mask of features inheritable by MPLS
1784 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1786 * @ifindex: interface index
1787 * @group: The group the device belongs to
1789 * @stats: Statistics struct, which was left as a legacy, use
1790 * rtnl_link_stats64 instead
1792 * @core_stats: core networking counters,
1793 * do not use this in drivers
1794 * @carrier_up_count: Number of times the carrier has been up
1795 * @carrier_down_count: Number of times the carrier has been down
1797 * @wireless_handlers: List of functions to handle Wireless Extensions,
1799 * see <net/iw_handler.h> for details.
1800 * @wireless_data: Instance data managed by the core of wireless extensions
1802 * @netdev_ops: Includes several pointers to callbacks,
1803 * if one wants to override the ndo_*() functions
1804 * @ethtool_ops: Management operations
1805 * @l3mdev_ops: Layer 3 master device operations
1806 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1807 * discovery handling. Necessary for e.g. 6LoWPAN.
1808 * @xfrmdev_ops: Transformation offload operations
1809 * @tlsdev_ops: Transport Layer Security offload operations
1810 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1811 * of Layer 2 headers.
1813 * @flags: Interface flags (a la BSD)
1814 * @priv_flags: Like 'flags' but invisible to userspace,
1815 * see if.h for the definitions
1816 * @gflags: Global flags ( kept as legacy )
1817 * @padded: How much padding added by alloc_netdev()
1818 * @operstate: RFC2863 operstate
1819 * @link_mode: Mapping policy to operstate
1820 * @if_port: Selectable AUI, TP, ...
1822 * @mtu: Interface MTU value
1823 * @min_mtu: Interface Minimum MTU value
1824 * @max_mtu: Interface Maximum MTU value
1825 * @type: Interface hardware type
1826 * @hard_header_len: Maximum hardware header length.
1827 * @min_header_len: Minimum hardware header length
1829 * @needed_headroom: Extra headroom the hardware may need, but not in all
1830 * cases can this be guaranteed
1831 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1832 * cases can this be guaranteed. Some cases also use
1833 * LL_MAX_HEADER instead to allocate the skb
1835 * interface address info:
1837 * @perm_addr: Permanent hw address
1838 * @addr_assign_type: Hw address assignment type
1839 * @addr_len: Hardware address length
1840 * @upper_level: Maximum depth level of upper devices.
1841 * @lower_level: Maximum depth level of lower devices.
1842 * @neigh_priv_len: Used in neigh_alloc()
1843 * @dev_id: Used to differentiate devices that share
1844 * the same link layer address
1845 * @dev_port: Used to differentiate devices that share
1847 * @addr_list_lock: XXX: need comments on this one
1848 * @name_assign_type: network interface name assignment type
1849 * @uc_promisc: Counter that indicates promiscuous mode
1850 * has been enabled due to the need to listen to
1851 * additional unicast addresses in a device that
1852 * does not implement ndo_set_rx_mode()
1853 * @uc: unicast mac addresses
1854 * @mc: multicast mac addresses
1855 * @dev_addrs: list of device hw addresses
1856 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1857 * @promiscuity: Number of times the NIC is told to work in
1858 * promiscuous mode; if it becomes 0 the NIC will
1859 * exit promiscuous mode
1860 * @allmulti: Counter, enables or disables allmulticast mode
1862 * @vlan_info: VLAN info
1863 * @dsa_ptr: dsa specific data
1864 * @tipc_ptr: TIPC specific data
1865 * @atalk_ptr: AppleTalk link
1866 * @ip_ptr: IPv4 specific data
1867 * @ip6_ptr: IPv6 specific data
1868 * @ax25_ptr: AX.25 specific data
1869 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1870 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1872 * @mpls_ptr: mpls_dev struct pointer
1873 * @mctp_ptr: MCTP specific data
1875 * @dev_addr: Hw address (before bcast,
1876 * because most packets are unicast)
1878 * @_rx: Array of RX queues
1879 * @num_rx_queues: Number of RX queues
1880 * allocated at register_netdev() time
1881 * @real_num_rx_queues: Number of RX queues currently active in device
1882 * @xdp_prog: XDP sockets filter program pointer
1883 * @gro_flush_timeout: timeout for GRO layer in NAPI
1884 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1885 * allow to avoid NIC hard IRQ, on busy queues.
1887 * @rx_handler: handler for received packets
1888 * @rx_handler_data: XXX: need comments on this one
1889 * @miniq_ingress: ingress/clsact qdisc specific data for
1890 * ingress processing
1891 * @ingress_queue: XXX: need comments on this one
1892 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1893 * @broadcast: hw bcast address
1895 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1896 * indexed by RX queue number. Assigned by driver.
1897 * This must only be set if the ndo_rx_flow_steer
1898 * operation is defined
1899 * @index_hlist: Device index hash chain
1901 * @_tx: Array of TX queues
1902 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1903 * @real_num_tx_queues: Number of TX queues currently active in device
1904 * @qdisc: Root qdisc from userspace point of view
1905 * @tx_queue_len: Max frames per queue allowed
1906 * @tx_global_lock: XXX: need comments on this one
1907 * @xdp_bulkq: XDP device bulk queue
1908 * @xps_maps: all CPUs/RXQs maps for XPS device
1910 * @xps_maps: XXX: need comments on this one
1911 * @miniq_egress: clsact qdisc specific data for
1913 * @nf_hooks_egress: netfilter hooks executed for egress packets
1914 * @qdisc_hash: qdisc hash table
1915 * @watchdog_timeo: Represents the timeout that is used by
1916 * the watchdog (see dev_watchdog())
1917 * @watchdog_timer: List of timers
1919 * @proto_down_reason: reason a netdev interface is held down
1920 * @pcpu_refcnt: Number of references to this device
1921 * @dev_refcnt: Number of references to this device
1922 * @refcnt_tracker: Tracker directory for tracked references to this device
1923 * @todo_list: Delayed register/unregister
1924 * @link_watch_list: XXX: need comments on this one
1926 * @reg_state: Register/unregister state machine
1927 * @dismantle: Device is going to be freed
1928 * @rtnl_link_state: This enum represents the phases of creating
1931 * @needs_free_netdev: Should unregister perform free_netdev?
1932 * @priv_destructor: Called from unregister
1933 * @npinfo: XXX: need comments on this one
1934 * @nd_net: Network namespace this network device is inside
1936 * @ml_priv: Mid-layer private
1937 * @ml_priv_type: Mid-layer private type
1938 * @lstats: Loopback statistics
1939 * @tstats: Tunnel statistics
1940 * @dstats: Dummy statistics
1941 * @vstats: Virtual ethernet statistics
1946 * @dm_private: Drop monitor private
1948 * @dev: Class/net/name entry
1949 * @sysfs_groups: Space for optional device, statistics and wireless
1952 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1953 * @rtnl_link_ops: Rtnl_link_ops
1955 * @gso_max_size: Maximum size of generic segmentation offload
1956 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1957 * @gso_max_segs: Maximum number of segments that can be passed to the
1959 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1961 * @dcbnl_ops: Data Center Bridging netlink ops
1962 * @num_tc: Number of traffic classes in the net device
1963 * @tc_to_txq: XXX: need comments on this one
1964 * @prio_tc_map: XXX: need comments on this one
1966 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1968 * @priomap: XXX: need comments on this one
1969 * @phydev: Physical device may attach itself
1970 * for hardware timestamping
1971 * @sfp_bus: attached &struct sfp_bus structure.
1973 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1975 * @proto_down: protocol port state information can be sent to the
1976 * switch driver and used to set the phys state of the
1979 * @wol_enabled: Wake-on-LAN is enabled
1981 * @threaded: napi threaded mode is enabled
1983 * @net_notifier_list: List of per-net netdev notifier block
1984 * that follow this device when it is moved
1985 * to another network namespace.
1987 * @macsec_ops: MACsec offloading ops
1989 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1990 * offload capabilities of the device
1991 * @udp_tunnel_nic: UDP tunnel offload state
1992 * @xdp_state: stores info on attached XDP BPF programs
1994 * @nested_level: Used as a parameter of spin_lock_nested() of
1995 * dev->addr_list_lock.
1996 * @unlink_list: As netif_addr_lock() can be called recursively,
1997 * keep a list of interfaces to be deleted.
1998 * @gro_max_size: Maximum size of aggregated packet in generic
1999 * receive offload (GRO)
2001 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2002 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2003 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2004 * @dev_registered_tracker: tracker for reference held while
2006 * @offload_xstats_l3: L3 HW stats for this netdevice.
2008 * @devlink_port: Pointer to related devlink port structure.
2009 * Assigned by a driver before netdev registration using
2010 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2011 * during the time netdevice is registered.
2013 * FIXME: cleanup struct net_device such that network protocol info
2018 char name[IFNAMSIZ];
2019 struct netdev_name_node *name_node;
2020 struct dev_ifalias __rcu *ifalias;
2022 * I/O specific fields
2023 * FIXME: Merge these and struct ifmap into one
2025 unsigned long mem_end;
2026 unsigned long mem_start;
2027 unsigned long base_addr;
2030 * Some hardware also needs these fields (state,dev_list,
2031 * napi_list,unreg_list,close_list) but they are not
2032 * part of the usual set specified in Space.c.
2035 unsigned long state;
2037 struct list_head dev_list;
2038 struct list_head napi_list;
2039 struct list_head unreg_list;
2040 struct list_head close_list;
2041 struct list_head ptype_all;
2042 struct list_head ptype_specific;
2045 struct list_head upper;
2046 struct list_head lower;
2049 /* Read-mostly cache-line for fast-path access */
2051 unsigned long long priv_flags;
2052 const struct net_device_ops *netdev_ops;
2054 unsigned short gflags;
2055 unsigned short hard_header_len;
2057 /* Note : dev->mtu is often read without holding a lock.
2058 * Writers usually hold RTNL.
2059 * It is recommended to use READ_ONCE() to annotate the reads,
2060 * and to use WRITE_ONCE() to annotate the writes.
2063 unsigned short needed_headroom;
2064 unsigned short needed_tailroom;
2066 netdev_features_t features;
2067 netdev_features_t hw_features;
2068 netdev_features_t wanted_features;
2069 netdev_features_t vlan_features;
2070 netdev_features_t hw_enc_features;
2071 netdev_features_t mpls_features;
2072 netdev_features_t gso_partial_features;
2074 unsigned int min_mtu;
2075 unsigned int max_mtu;
2076 unsigned short type;
2077 unsigned char min_header_len;
2078 unsigned char name_assign_type;
2082 struct net_device_stats stats; /* not used by modern drivers */
2084 struct net_device_core_stats __percpu *core_stats;
2086 /* Stats to monitor link on/off, flapping */
2087 atomic_t carrier_up_count;
2088 atomic_t carrier_down_count;
2090 #ifdef CONFIG_WIRELESS_EXT
2091 const struct iw_handler_def *wireless_handlers;
2092 struct iw_public_data *wireless_data;
2094 const struct ethtool_ops *ethtool_ops;
2095 #ifdef CONFIG_NET_L3_MASTER_DEV
2096 const struct l3mdev_ops *l3mdev_ops;
2098 #if IS_ENABLED(CONFIG_IPV6)
2099 const struct ndisc_ops *ndisc_ops;
2102 #ifdef CONFIG_XFRM_OFFLOAD
2103 const struct xfrmdev_ops *xfrmdev_ops;
2106 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2107 const struct tlsdev_ops *tlsdev_ops;
2110 const struct header_ops *header_ops;
2112 unsigned char operstate;
2113 unsigned char link_mode;
2115 unsigned char if_port;
2118 /* Interface address info. */
2119 unsigned char perm_addr[MAX_ADDR_LEN];
2120 unsigned char addr_assign_type;
2121 unsigned char addr_len;
2122 unsigned char upper_level;
2123 unsigned char lower_level;
2125 unsigned short neigh_priv_len;
2126 unsigned short dev_id;
2127 unsigned short dev_port;
2128 unsigned short padded;
2130 spinlock_t addr_list_lock;
2133 struct netdev_hw_addr_list uc;
2134 struct netdev_hw_addr_list mc;
2135 struct netdev_hw_addr_list dev_addrs;
2138 struct kset *queues_kset;
2140 #ifdef CONFIG_LOCKDEP
2141 struct list_head unlink_list;
2143 unsigned int promiscuity;
2144 unsigned int allmulti;
2146 #ifdef CONFIG_LOCKDEP
2147 unsigned char nested_level;
2151 /* Protocol-specific pointers */
2153 struct in_device __rcu *ip_ptr;
2154 struct inet6_dev __rcu *ip6_ptr;
2155 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2156 struct vlan_info __rcu *vlan_info;
2158 #if IS_ENABLED(CONFIG_NET_DSA)
2159 struct dsa_port *dsa_ptr;
2161 #if IS_ENABLED(CONFIG_TIPC)
2162 struct tipc_bearer __rcu *tipc_ptr;
2164 #if IS_ENABLED(CONFIG_ATALK)
2167 #if IS_ENABLED(CONFIG_AX25)
2170 #if IS_ENABLED(CONFIG_CFG80211)
2171 struct wireless_dev *ieee80211_ptr;
2173 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2174 struct wpan_dev *ieee802154_ptr;
2176 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2177 struct mpls_dev __rcu *mpls_ptr;
2179 #if IS_ENABLED(CONFIG_MCTP)
2180 struct mctp_dev __rcu *mctp_ptr;
2184 * Cache lines mostly used on receive path (including eth_type_trans())
2186 /* Interface address info used in eth_type_trans() */
2187 const unsigned char *dev_addr;
2189 struct netdev_rx_queue *_rx;
2190 unsigned int num_rx_queues;
2191 unsigned int real_num_rx_queues;
2193 struct bpf_prog __rcu *xdp_prog;
2194 unsigned long gro_flush_timeout;
2195 int napi_defer_hard_irqs;
2196 #define GRO_LEGACY_MAX_SIZE 65536u
2197 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2198 * and shinfo->gso_segs is a 16bit field.
2200 #define GRO_MAX_SIZE (8 * 65535u)
2201 unsigned int gro_max_size;
2202 rx_handler_func_t __rcu *rx_handler;
2203 void __rcu *rx_handler_data;
2205 #ifdef CONFIG_NET_CLS_ACT
2206 struct mini_Qdisc __rcu *miniq_ingress;
2208 struct netdev_queue __rcu *ingress_queue;
2209 #ifdef CONFIG_NETFILTER_INGRESS
2210 struct nf_hook_entries __rcu *nf_hooks_ingress;
2213 unsigned char broadcast[MAX_ADDR_LEN];
2214 #ifdef CONFIG_RFS_ACCEL
2215 struct cpu_rmap *rx_cpu_rmap;
2217 struct hlist_node index_hlist;
2220 * Cache lines mostly used on transmit path
2222 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2223 unsigned int num_tx_queues;
2224 unsigned int real_num_tx_queues;
2225 struct Qdisc __rcu *qdisc;
2226 unsigned int tx_queue_len;
2227 spinlock_t tx_global_lock;
2229 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2232 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2234 #ifdef CONFIG_NET_CLS_ACT
2235 struct mini_Qdisc __rcu *miniq_egress;
2237 #ifdef CONFIG_NETFILTER_EGRESS
2238 struct nf_hook_entries __rcu *nf_hooks_egress;
2241 #ifdef CONFIG_NET_SCHED
2242 DECLARE_HASHTABLE (qdisc_hash, 4);
2244 /* These may be needed for future network-power-down code. */
2245 struct timer_list watchdog_timer;
2248 u32 proto_down_reason;
2250 struct list_head todo_list;
2252 #ifdef CONFIG_PCPU_DEV_REFCNT
2253 int __percpu *pcpu_refcnt;
2255 refcount_t dev_refcnt;
2257 struct ref_tracker_dir refcnt_tracker;
2259 struct list_head link_watch_list;
2261 enum { NETREG_UNINITIALIZED=0,
2262 NETREG_REGISTERED, /* completed register_netdevice */
2263 NETREG_UNREGISTERING, /* called unregister_netdevice */
2264 NETREG_UNREGISTERED, /* completed unregister todo */
2265 NETREG_RELEASED, /* called free_netdev */
2266 NETREG_DUMMY, /* dummy device for NAPI poll */
2272 RTNL_LINK_INITIALIZED,
2273 RTNL_LINK_INITIALIZING,
2274 } rtnl_link_state:16;
2276 bool needs_free_netdev;
2277 void (*priv_destructor)(struct net_device *dev);
2279 #ifdef CONFIG_NETPOLL
2280 struct netpoll_info __rcu *npinfo;
2283 possible_net_t nd_net;
2285 /* mid-layer private */
2287 enum netdev_ml_priv_type ml_priv_type;
2290 struct pcpu_lstats __percpu *lstats;
2291 struct pcpu_sw_netstats __percpu *tstats;
2292 struct pcpu_dstats __percpu *dstats;
2295 #if IS_ENABLED(CONFIG_GARP)
2296 struct garp_port __rcu *garp_port;
2298 #if IS_ENABLED(CONFIG_MRP)
2299 struct mrp_port __rcu *mrp_port;
2301 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2302 struct dm_hw_stat_delta __rcu *dm_private;
2305 const struct attribute_group *sysfs_groups[4];
2306 const struct attribute_group *sysfs_rx_queue_group;
2308 const struct rtnl_link_ops *rtnl_link_ops;
2310 /* for setting kernel sock attribute on TCP connection setup */
2311 #define GSO_MAX_SEGS 65535u
2312 #define GSO_LEGACY_MAX_SIZE 65536u
2313 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2314 * and shinfo->gso_segs is a 16bit field.
2316 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2318 unsigned int gso_max_size;
2319 #define TSO_LEGACY_MAX_SIZE 65536
2320 #define TSO_MAX_SIZE UINT_MAX
2321 unsigned int tso_max_size;
2323 #define TSO_MAX_SEGS U16_MAX
2327 const struct dcbnl_rtnl_ops *dcbnl_ops;
2330 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2331 u8 prio_tc_map[TC_BITMASK + 1];
2333 #if IS_ENABLED(CONFIG_FCOE)
2334 unsigned int fcoe_ddp_xid;
2336 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2337 struct netprio_map __rcu *priomap;
2339 struct phy_device *phydev;
2340 struct sfp_bus *sfp_bus;
2341 struct lock_class_key *qdisc_tx_busylock;
2343 unsigned wol_enabled:1;
2344 unsigned threaded:1;
2346 struct list_head net_notifier_list;
2348 #if IS_ENABLED(CONFIG_MACSEC)
2349 /* MACsec management functions */
2350 const struct macsec_ops *macsec_ops;
2352 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2353 struct udp_tunnel_nic *udp_tunnel_nic;
2355 /* protected by rtnl_lock */
2356 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2358 u8 dev_addr_shadow[MAX_ADDR_LEN];
2359 netdevice_tracker linkwatch_dev_tracker;
2360 netdevice_tracker watchdog_dev_tracker;
2361 netdevice_tracker dev_registered_tracker;
2362 struct rtnl_hw_stats64 *offload_xstats_l3;
2364 struct devlink_port *devlink_port;
2366 #define to_net_dev(d) container_of(d, struct net_device, dev)
2369 * Driver should use this to assign devlink port instance to a netdevice
2370 * before it registers the netdevice. Therefore devlink_port is static
2371 * during the netdev lifetime after it is registered.
2373 #define SET_NETDEV_DEVLINK_PORT(dev, port) \
2375 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2376 ((dev)->devlink_port = (port)); \
2379 static inline bool netif_elide_gro(const struct net_device *dev)
2381 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2386 #define NETDEV_ALIGN 32
2389 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2391 return dev->prio_tc_map[prio & TC_BITMASK];
2395 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2397 if (tc >= dev->num_tc)
2400 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2404 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2405 void netdev_reset_tc(struct net_device *dev);
2406 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2407 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2410 int netdev_get_num_tc(struct net_device *dev)
2415 static inline void net_prefetch(void *p)
2418 #if L1_CACHE_BYTES < 128
2419 prefetch((u8 *)p + L1_CACHE_BYTES);
2423 static inline void net_prefetchw(void *p)
2426 #if L1_CACHE_BYTES < 128
2427 prefetchw((u8 *)p + L1_CACHE_BYTES);
2431 void netdev_unbind_sb_channel(struct net_device *dev,
2432 struct net_device *sb_dev);
2433 int netdev_bind_sb_channel_queue(struct net_device *dev,
2434 struct net_device *sb_dev,
2435 u8 tc, u16 count, u16 offset);
2436 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2437 static inline int netdev_get_sb_channel(struct net_device *dev)
2439 return max_t(int, -dev->num_tc, 0);
2443 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2446 return &dev->_tx[index];
2449 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2450 const struct sk_buff *skb)
2452 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2455 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2456 void (*f)(struct net_device *,
2457 struct netdev_queue *,
2463 for (i = 0; i < dev->num_tx_queues; i++)
2464 f(dev, &dev->_tx[i], arg);
2467 #define netdev_lockdep_set_classes(dev) \
2469 static struct lock_class_key qdisc_tx_busylock_key; \
2470 static struct lock_class_key qdisc_xmit_lock_key; \
2471 static struct lock_class_key dev_addr_list_lock_key; \
2474 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2475 lockdep_set_class(&(dev)->addr_list_lock, \
2476 &dev_addr_list_lock_key); \
2477 for (i = 0; i < (dev)->num_tx_queues; i++) \
2478 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2479 &qdisc_xmit_lock_key); \
2482 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2483 struct net_device *sb_dev);
2484 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2485 struct sk_buff *skb,
2486 struct net_device *sb_dev);
2488 /* returns the headroom that the master device needs to take in account
2489 * when forwarding to this dev
2491 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2493 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2496 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2498 if (dev->netdev_ops->ndo_set_rx_headroom)
2499 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2502 /* set the device rx headroom to the dev's default */
2503 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2505 netdev_set_rx_headroom(dev, -1);
2508 static inline void *netdev_get_ml_priv(struct net_device *dev,
2509 enum netdev_ml_priv_type type)
2511 if (dev->ml_priv_type != type)
2514 return dev->ml_priv;
2517 static inline void netdev_set_ml_priv(struct net_device *dev,
2519 enum netdev_ml_priv_type type)
2521 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2522 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2523 dev->ml_priv_type, type);
2524 WARN(!dev->ml_priv_type && dev->ml_priv,
2525 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2527 dev->ml_priv = ml_priv;
2528 dev->ml_priv_type = type;
2532 * Net namespace inlines
2535 struct net *dev_net(const struct net_device *dev)
2537 return read_pnet(&dev->nd_net);
2541 void dev_net_set(struct net_device *dev, struct net *net)
2543 write_pnet(&dev->nd_net, net);
2547 * netdev_priv - access network device private data
2548 * @dev: network device
2550 * Get network device private data
2552 static inline void *netdev_priv(const struct net_device *dev)
2554 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2557 /* Set the sysfs physical device reference for the network logical device
2558 * if set prior to registration will cause a symlink during initialization.
2560 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2562 /* Set the sysfs device type for the network logical device to allow
2563 * fine-grained identification of different network device types. For
2564 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2566 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2568 /* Default NAPI poll() weight
2569 * Device drivers are strongly advised to not use bigger value
2571 #define NAPI_POLL_WEIGHT 64
2573 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2574 int (*poll)(struct napi_struct *, int), int weight);
2577 * netif_napi_add() - initialize a NAPI context
2578 * @dev: network device
2579 * @napi: NAPI context
2580 * @poll: polling function
2582 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2583 * *any* of the other NAPI-related functions.
2586 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2587 int (*poll)(struct napi_struct *, int))
2589 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2593 netif_napi_add_tx_weight(struct net_device *dev,
2594 struct napi_struct *napi,
2595 int (*poll)(struct napi_struct *, int),
2598 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2599 netif_napi_add_weight(dev, napi, poll, weight);
2603 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2604 * @dev: network device
2605 * @napi: NAPI context
2606 * @poll: polling function
2608 * This variant of netif_napi_add() should be used from drivers using NAPI
2609 * to exclusively poll a TX queue.
2610 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2612 static inline void netif_napi_add_tx(struct net_device *dev,
2613 struct napi_struct *napi,
2614 int (*poll)(struct napi_struct *, int))
2616 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2620 * __netif_napi_del - remove a NAPI context
2621 * @napi: NAPI context
2623 * Warning: caller must observe RCU grace period before freeing memory
2624 * containing @napi. Drivers might want to call this helper to combine
2625 * all the needed RCU grace periods into a single one.
2627 void __netif_napi_del(struct napi_struct *napi);
2630 * netif_napi_del - remove a NAPI context
2631 * @napi: NAPI context
2633 * netif_napi_del() removes a NAPI context from the network device NAPI list
2635 static inline void netif_napi_del(struct napi_struct *napi)
2637 __netif_napi_del(napi);
2641 struct packet_type {
2642 __be16 type; /* This is really htons(ether_type). */
2643 bool ignore_outgoing;
2644 struct net_device *dev; /* NULL is wildcarded here */
2645 netdevice_tracker dev_tracker;
2646 int (*func) (struct sk_buff *,
2647 struct net_device *,
2648 struct packet_type *,
2649 struct net_device *);
2650 void (*list_func) (struct list_head *,
2651 struct packet_type *,
2652 struct net_device *);
2653 bool (*id_match)(struct packet_type *ptype,
2655 struct net *af_packet_net;
2656 void *af_packet_priv;
2657 struct list_head list;
2660 struct offload_callbacks {
2661 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2662 netdev_features_t features);
2663 struct sk_buff *(*gro_receive)(struct list_head *head,
2664 struct sk_buff *skb);
2665 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2668 struct packet_offload {
2669 __be16 type; /* This is really htons(ether_type). */
2671 struct offload_callbacks callbacks;
2672 struct list_head list;
2675 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2676 struct pcpu_sw_netstats {
2677 u64_stats_t rx_packets;
2678 u64_stats_t rx_bytes;
2679 u64_stats_t tx_packets;
2680 u64_stats_t tx_bytes;
2681 struct u64_stats_sync syncp;
2682 } __aligned(4 * sizeof(u64));
2684 struct pcpu_lstats {
2685 u64_stats_t packets;
2687 struct u64_stats_sync syncp;
2688 } __aligned(2 * sizeof(u64));
2690 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2692 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2694 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2696 u64_stats_update_begin(&tstats->syncp);
2697 u64_stats_add(&tstats->rx_bytes, len);
2698 u64_stats_inc(&tstats->rx_packets);
2699 u64_stats_update_end(&tstats->syncp);
2702 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2703 unsigned int packets,
2706 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2708 u64_stats_update_begin(&tstats->syncp);
2709 u64_stats_add(&tstats->tx_bytes, len);
2710 u64_stats_add(&tstats->tx_packets, packets);
2711 u64_stats_update_end(&tstats->syncp);
2714 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2716 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2718 u64_stats_update_begin(&lstats->syncp);
2719 u64_stats_add(&lstats->bytes, len);
2720 u64_stats_inc(&lstats->packets);
2721 u64_stats_update_end(&lstats->syncp);
2724 #define __netdev_alloc_pcpu_stats(type, gfp) \
2726 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2729 for_each_possible_cpu(__cpu) { \
2730 typeof(type) *stat; \
2731 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2732 u64_stats_init(&stat->syncp); \
2738 #define netdev_alloc_pcpu_stats(type) \
2739 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2741 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2743 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2746 for_each_possible_cpu(__cpu) { \
2747 typeof(type) *stat; \
2748 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2749 u64_stats_init(&stat->syncp); \
2755 enum netdev_lag_tx_type {
2756 NETDEV_LAG_TX_TYPE_UNKNOWN,
2757 NETDEV_LAG_TX_TYPE_RANDOM,
2758 NETDEV_LAG_TX_TYPE_BROADCAST,
2759 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2760 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2761 NETDEV_LAG_TX_TYPE_HASH,
2764 enum netdev_lag_hash {
2765 NETDEV_LAG_HASH_NONE,
2767 NETDEV_LAG_HASH_L34,
2768 NETDEV_LAG_HASH_L23,
2769 NETDEV_LAG_HASH_E23,
2770 NETDEV_LAG_HASH_E34,
2771 NETDEV_LAG_HASH_VLAN_SRCMAC,
2772 NETDEV_LAG_HASH_UNKNOWN,
2775 struct netdev_lag_upper_info {
2776 enum netdev_lag_tx_type tx_type;
2777 enum netdev_lag_hash hash_type;
2780 struct netdev_lag_lower_state_info {
2785 #include <linux/notifier.h>
2787 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2788 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2792 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2794 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2795 detected a hardware crash and restarted
2796 - we can use this eg to kick tcp sessions
2798 NETDEV_CHANGE, /* Notify device state change */
2801 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2802 NETDEV_CHANGEADDR, /* notify after the address change */
2803 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2807 NETDEV_BONDING_FAILOVER,
2809 NETDEV_PRE_TYPE_CHANGE,
2810 NETDEV_POST_TYPE_CHANGE,
2814 NETDEV_NOTIFY_PEERS,
2818 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2819 NETDEV_CHANGEINFODATA,
2820 NETDEV_BONDING_INFO,
2821 NETDEV_PRECHANGEUPPER,
2822 NETDEV_CHANGELOWERSTATE,
2823 NETDEV_UDP_TUNNEL_PUSH_INFO,
2824 NETDEV_UDP_TUNNEL_DROP_INFO,
2825 NETDEV_CHANGE_TX_QUEUE_LEN,
2826 NETDEV_CVLAN_FILTER_PUSH_INFO,
2827 NETDEV_CVLAN_FILTER_DROP_INFO,
2828 NETDEV_SVLAN_FILTER_PUSH_INFO,
2829 NETDEV_SVLAN_FILTER_DROP_INFO,
2830 NETDEV_OFFLOAD_XSTATS_ENABLE,
2831 NETDEV_OFFLOAD_XSTATS_DISABLE,
2832 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2833 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2835 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2837 int register_netdevice_notifier(struct notifier_block *nb);
2838 int unregister_netdevice_notifier(struct notifier_block *nb);
2839 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2840 int unregister_netdevice_notifier_net(struct net *net,
2841 struct notifier_block *nb);
2842 void move_netdevice_notifier_net(struct net *src_net, struct net *dst_net,
2843 struct notifier_block *nb);
2844 int register_netdevice_notifier_dev_net(struct net_device *dev,
2845 struct notifier_block *nb,
2846 struct netdev_net_notifier *nn);
2847 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2848 struct notifier_block *nb,
2849 struct netdev_net_notifier *nn);
2851 struct netdev_notifier_info {
2852 struct net_device *dev;
2853 struct netlink_ext_ack *extack;
2856 struct netdev_notifier_info_ext {
2857 struct netdev_notifier_info info; /* must be first */
2863 struct netdev_notifier_change_info {
2864 struct netdev_notifier_info info; /* must be first */
2865 unsigned int flags_changed;
2868 struct netdev_notifier_changeupper_info {
2869 struct netdev_notifier_info info; /* must be first */
2870 struct net_device *upper_dev; /* new upper dev */
2871 bool master; /* is upper dev master */
2872 bool linking; /* is the notification for link or unlink */
2873 void *upper_info; /* upper dev info */
2876 struct netdev_notifier_changelowerstate_info {
2877 struct netdev_notifier_info info; /* must be first */
2878 void *lower_state_info; /* is lower dev state */
2881 struct netdev_notifier_pre_changeaddr_info {
2882 struct netdev_notifier_info info; /* must be first */
2883 const unsigned char *dev_addr;
2886 enum netdev_offload_xstats_type {
2887 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2890 struct netdev_notifier_offload_xstats_info {
2891 struct netdev_notifier_info info; /* must be first */
2892 enum netdev_offload_xstats_type type;
2895 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2896 struct netdev_notifier_offload_xstats_rd *report_delta;
2897 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2898 struct netdev_notifier_offload_xstats_ru *report_used;
2902 int netdev_offload_xstats_enable(struct net_device *dev,
2903 enum netdev_offload_xstats_type type,
2904 struct netlink_ext_ack *extack);
2905 int netdev_offload_xstats_disable(struct net_device *dev,
2906 enum netdev_offload_xstats_type type);
2907 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2908 enum netdev_offload_xstats_type type);
2909 int netdev_offload_xstats_get(struct net_device *dev,
2910 enum netdev_offload_xstats_type type,
2911 struct rtnl_hw_stats64 *stats, bool *used,
2912 struct netlink_ext_ack *extack);
2914 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2915 const struct rtnl_hw_stats64 *stats);
2917 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2918 void netdev_offload_xstats_push_delta(struct net_device *dev,
2919 enum netdev_offload_xstats_type type,
2920 const struct rtnl_hw_stats64 *stats);
2922 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2923 struct net_device *dev)
2926 info->extack = NULL;
2929 static inline struct net_device *
2930 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2935 static inline struct netlink_ext_ack *
2936 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2938 return info->extack;
2941 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2944 extern rwlock_t dev_base_lock; /* Device list lock */
2946 #define for_each_netdev(net, d) \
2947 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2948 #define for_each_netdev_reverse(net, d) \
2949 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2950 #define for_each_netdev_rcu(net, d) \
2951 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2952 #define for_each_netdev_safe(net, d, n) \
2953 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2954 #define for_each_netdev_continue(net, d) \
2955 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2956 #define for_each_netdev_continue_reverse(net, d) \
2957 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2959 #define for_each_netdev_continue_rcu(net, d) \
2960 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2961 #define for_each_netdev_in_bond_rcu(bond, slave) \
2962 for_each_netdev_rcu(&init_net, slave) \
2963 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2964 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2966 static inline struct net_device *next_net_device(struct net_device *dev)
2968 struct list_head *lh;
2972 lh = dev->dev_list.next;
2973 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2976 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2978 struct list_head *lh;
2982 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2983 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2986 static inline struct net_device *first_net_device(struct net *net)
2988 return list_empty(&net->dev_base_head) ? NULL :
2989 net_device_entry(net->dev_base_head.next);
2992 static inline struct net_device *first_net_device_rcu(struct net *net)
2994 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2996 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2999 int netdev_boot_setup_check(struct net_device *dev);
3000 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3001 const char *hwaddr);
3002 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3003 void dev_add_pack(struct packet_type *pt);
3004 void dev_remove_pack(struct packet_type *pt);
3005 void __dev_remove_pack(struct packet_type *pt);
3006 void dev_add_offload(struct packet_offload *po);
3007 void dev_remove_offload(struct packet_offload *po);
3009 int dev_get_iflink(const struct net_device *dev);
3010 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3011 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3012 struct net_device_path_stack *stack);
3013 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3014 unsigned short mask);
3015 struct net_device *dev_get_by_name(struct net *net, const char *name);
3016 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3017 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3018 bool netdev_name_in_use(struct net *net, const char *name);
3019 int dev_alloc_name(struct net_device *dev, const char *name);
3020 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3021 void dev_close(struct net_device *dev);
3022 void dev_close_many(struct list_head *head, bool unlink);
3023 void dev_disable_lro(struct net_device *dev);
3024 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3025 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3026 struct net_device *sb_dev);
3027 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3028 struct net_device *sb_dev);
3030 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3031 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3033 static inline int dev_queue_xmit(struct sk_buff *skb)
3035 return __dev_queue_xmit(skb, NULL);
3038 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3039 struct net_device *sb_dev)
3041 return __dev_queue_xmit(skb, sb_dev);
3044 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3048 ret = __dev_direct_xmit(skb, queue_id);
3049 if (!dev_xmit_complete(ret))
3054 int register_netdevice(struct net_device *dev);
3055 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3056 void unregister_netdevice_many(struct list_head *head);
3057 static inline void unregister_netdevice(struct net_device *dev)
3059 unregister_netdevice_queue(dev, NULL);
3062 int netdev_refcnt_read(const struct net_device *dev);
3063 void free_netdev(struct net_device *dev);
3064 void netdev_freemem(struct net_device *dev);
3065 int init_dummy_netdev(struct net_device *dev);
3067 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3068 struct sk_buff *skb,
3070 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3072 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3073 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3074 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3075 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3076 int dev_restart(struct net_device *dev);
3079 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3080 unsigned short type,
3081 const void *daddr, const void *saddr,
3084 if (!dev->header_ops || !dev->header_ops->create)
3087 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3090 static inline int dev_parse_header(const struct sk_buff *skb,
3091 unsigned char *haddr)
3093 const struct net_device *dev = skb->dev;
3095 if (!dev->header_ops || !dev->header_ops->parse)
3097 return dev->header_ops->parse(skb, haddr);
3100 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3102 const struct net_device *dev = skb->dev;
3104 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3106 return dev->header_ops->parse_protocol(skb);
3109 /* ll_header must have at least hard_header_len allocated */
3110 static inline bool dev_validate_header(const struct net_device *dev,
3111 char *ll_header, int len)
3113 if (likely(len >= dev->hard_header_len))
3115 if (len < dev->min_header_len)
3118 if (capable(CAP_SYS_RAWIO)) {
3119 memset(ll_header + len, 0, dev->hard_header_len - len);
3123 if (dev->header_ops && dev->header_ops->validate)
3124 return dev->header_ops->validate(ll_header, len);
3129 static inline bool dev_has_header(const struct net_device *dev)
3131 return dev->header_ops && dev->header_ops->create;
3135 * Incoming packets are placed on per-CPU queues
3137 struct softnet_data {
3138 struct list_head poll_list;
3139 struct sk_buff_head process_queue;
3142 unsigned int processed;
3143 unsigned int time_squeeze;
3145 struct softnet_data *rps_ipi_list;
3147 #ifdef CONFIG_NET_FLOW_LIMIT
3148 struct sd_flow_limit __rcu *flow_limit;
3150 struct Qdisc *output_queue;
3151 struct Qdisc **output_queue_tailp;
3152 struct sk_buff *completion_queue;
3153 #ifdef CONFIG_XFRM_OFFLOAD
3154 struct sk_buff_head xfrm_backlog;
3156 /* written and read only by owning cpu: */
3160 #ifdef CONFIG_NET_EGRESS
3165 /* input_queue_head should be written by cpu owning this struct,
3166 * and only read by other cpus. Worth using a cache line.
3168 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3170 /* Elements below can be accessed between CPUs for RPS/RFS */
3171 call_single_data_t csd ____cacheline_aligned_in_smp;
3172 struct softnet_data *rps_ipi_next;
3174 unsigned int input_queue_tail;
3176 unsigned int received_rps;
3177 unsigned int dropped;
3178 struct sk_buff_head input_pkt_queue;
3179 struct napi_struct backlog;
3181 /* Another possibly contended cache line */
3182 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3184 int defer_ipi_scheduled;
3185 struct sk_buff *defer_list;
3186 call_single_data_t defer_csd;
3189 static inline void input_queue_head_incr(struct softnet_data *sd)
3192 sd->input_queue_head++;
3196 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3197 unsigned int *qtail)
3200 *qtail = ++sd->input_queue_tail;
3204 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3206 static inline int dev_recursion_level(void)
3208 return this_cpu_read(softnet_data.xmit.recursion);
3211 #define XMIT_RECURSION_LIMIT 8
3212 static inline bool dev_xmit_recursion(void)
3214 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3215 XMIT_RECURSION_LIMIT);
3218 static inline void dev_xmit_recursion_inc(void)
3220 __this_cpu_inc(softnet_data.xmit.recursion);
3223 static inline void dev_xmit_recursion_dec(void)
3225 __this_cpu_dec(softnet_data.xmit.recursion);
3228 void __netif_schedule(struct Qdisc *q);
3229 void netif_schedule_queue(struct netdev_queue *txq);
3231 static inline void netif_tx_schedule_all(struct net_device *dev)
3235 for (i = 0; i < dev->num_tx_queues; i++)
3236 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3239 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3241 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3245 * netif_start_queue - allow transmit
3246 * @dev: network device
3248 * Allow upper layers to call the device hard_start_xmit routine.
3250 static inline void netif_start_queue(struct net_device *dev)
3252 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3255 static inline void netif_tx_start_all_queues(struct net_device *dev)
3259 for (i = 0; i < dev->num_tx_queues; i++) {
3260 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3261 netif_tx_start_queue(txq);
3265 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3268 * netif_wake_queue - restart transmit
3269 * @dev: network device
3271 * Allow upper layers to call the device hard_start_xmit routine.
3272 * Used for flow control when transmit resources are available.
3274 static inline void netif_wake_queue(struct net_device *dev)
3276 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3279 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3283 for (i = 0; i < dev->num_tx_queues; i++) {
3284 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3285 netif_tx_wake_queue(txq);
3289 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3291 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3295 * netif_stop_queue - stop transmitted packets
3296 * @dev: network device
3298 * Stop upper layers calling the device hard_start_xmit routine.
3299 * Used for flow control when transmit resources are unavailable.
3301 static inline void netif_stop_queue(struct net_device *dev)
3303 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3306 void netif_tx_stop_all_queues(struct net_device *dev);
3308 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3310 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3314 * netif_queue_stopped - test if transmit queue is flowblocked
3315 * @dev: network device
3317 * Test if transmit queue on device is currently unable to send.
3319 static inline bool netif_queue_stopped(const struct net_device *dev)
3321 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3324 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3326 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3330 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3332 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3336 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3338 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3342 * netdev_queue_set_dql_min_limit - set dql minimum limit
3343 * @dev_queue: pointer to transmit queue
3344 * @min_limit: dql minimum limit
3346 * Forces xmit_more() to return true until the minimum threshold
3347 * defined by @min_limit is reached (or until the tx queue is
3348 * empty). Warning: to be use with care, misuse will impact the
3351 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3352 unsigned int min_limit)
3355 dev_queue->dql.min_limit = min_limit;
3360 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3361 * @dev_queue: pointer to transmit queue
3363 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3364 * to give appropriate hint to the CPU.
3366 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3369 prefetchw(&dev_queue->dql.num_queued);
3374 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3375 * @dev_queue: pointer to transmit queue
3377 * BQL enabled drivers might use this helper in their TX completion path,
3378 * to give appropriate hint to the CPU.
3380 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3383 prefetchw(&dev_queue->dql.limit);
3388 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3389 * @dev_queue: network device queue
3390 * @bytes: number of bytes queued to the device queue
3392 * Report the number of bytes queued for sending/completion to the network
3393 * device hardware queue. @bytes should be a good approximation and should
3394 * exactly match netdev_completed_queue() @bytes.
3395 * This is typically called once per packet, from ndo_start_xmit().
3397 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3401 dql_queued(&dev_queue->dql, bytes);
3403 if (likely(dql_avail(&dev_queue->dql) >= 0))
3406 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3409 * The XOFF flag must be set before checking the dql_avail below,
3410 * because in netdev_tx_completed_queue we update the dql_completed
3411 * before checking the XOFF flag.
3415 /* check again in case another CPU has just made room avail */
3416 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3417 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3421 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3422 * that they should not test BQL status themselves.
3423 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3425 * Returns true if the doorbell must be used to kick the NIC.
3427 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3433 dql_queued(&dev_queue->dql, bytes);
3435 return netif_tx_queue_stopped(dev_queue);
3437 netdev_tx_sent_queue(dev_queue, bytes);
3442 * netdev_sent_queue - report the number of bytes queued to hardware
3443 * @dev: network device
3444 * @bytes: number of bytes queued to the hardware device queue
3446 * Report the number of bytes queued for sending/completion to the network
3447 * device hardware queue#0. @bytes should be a good approximation and should
3448 * exactly match netdev_completed_queue() @bytes.
3449 * This is typically called once per packet, from ndo_start_xmit().
3451 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3453 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3456 static inline bool __netdev_sent_queue(struct net_device *dev,
3460 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3465 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3466 * @dev_queue: network device queue
3467 * @pkts: number of packets (currently ignored)
3468 * @bytes: number of bytes dequeued from the device queue
3470 * Must be called at most once per TX completion round (and not per
3471 * individual packet), so that BQL can adjust its limits appropriately.
3473 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3474 unsigned int pkts, unsigned int bytes)
3477 if (unlikely(!bytes))
3480 dql_completed(&dev_queue->dql, bytes);
3483 * Without the memory barrier there is a small possiblity that
3484 * netdev_tx_sent_queue will miss the update and cause the queue to
3485 * be stopped forever
3489 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3492 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3493 netif_schedule_queue(dev_queue);
3498 * netdev_completed_queue - report bytes and packets completed by device
3499 * @dev: network device
3500 * @pkts: actual number of packets sent over the medium
3501 * @bytes: actual number of bytes sent over the medium
3503 * Report the number of bytes and packets transmitted by the network device
3504 * hardware queue over the physical medium, @bytes must exactly match the
3505 * @bytes amount passed to netdev_sent_queue()
3507 static inline void netdev_completed_queue(struct net_device *dev,
3508 unsigned int pkts, unsigned int bytes)
3510 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3513 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3516 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3522 * netdev_reset_queue - reset the packets and bytes count of a network device
3523 * @dev_queue: network device
3525 * Reset the bytes and packet count of a network device and clear the
3526 * software flow control OFF bit for this network device
3528 static inline void netdev_reset_queue(struct net_device *dev_queue)
3530 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3534 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3535 * @dev: network device
3536 * @queue_index: given tx queue index
3538 * Returns 0 if given tx queue index >= number of device tx queues,
3539 * otherwise returns the originally passed tx queue index.
3541 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3543 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3544 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3545 dev->name, queue_index,
3546 dev->real_num_tx_queues);
3554 * netif_running - test if up
3555 * @dev: network device
3557 * Test if the device has been brought up.
3559 static inline bool netif_running(const struct net_device *dev)
3561 return test_bit(__LINK_STATE_START, &dev->state);
3565 * Routines to manage the subqueues on a device. We only need start,
3566 * stop, and a check if it's stopped. All other device management is
3567 * done at the overall netdevice level.
3568 * Also test the device if we're multiqueue.
3572 * netif_start_subqueue - allow sending packets on subqueue
3573 * @dev: network device
3574 * @queue_index: sub queue index
3576 * Start individual transmit queue of a device with multiple transmit queues.
3578 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3580 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3582 netif_tx_start_queue(txq);
3586 * netif_stop_subqueue - stop sending packets on subqueue
3587 * @dev: network device
3588 * @queue_index: sub queue index
3590 * Stop individual transmit queue of a device with multiple transmit queues.
3592 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3594 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3595 netif_tx_stop_queue(txq);
3599 * __netif_subqueue_stopped - test status of subqueue
3600 * @dev: network device
3601 * @queue_index: sub queue index
3603 * Check individual transmit queue of a device with multiple transmit queues.
3605 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3608 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3610 return netif_tx_queue_stopped(txq);
3614 * netif_subqueue_stopped - test status of subqueue
3615 * @dev: network device
3616 * @skb: sub queue buffer pointer
3618 * Check individual transmit queue of a device with multiple transmit queues.
3620 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3621 struct sk_buff *skb)
3623 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3627 * netif_wake_subqueue - allow sending packets on subqueue
3628 * @dev: network device
3629 * @queue_index: sub queue index
3631 * Resume individual transmit queue of a device with multiple transmit queues.
3633 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3635 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3637 netif_tx_wake_queue(txq);
3641 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3643 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3644 u16 index, enum xps_map_type type);
3647 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3648 * @j: CPU/Rx queue index
3649 * @mask: bitmask of all cpus/rx queues
3650 * @nr_bits: number of bits in the bitmask
3652 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3654 static inline bool netif_attr_test_mask(unsigned long j,
3655 const unsigned long *mask,
3656 unsigned int nr_bits)
3658 cpu_max_bits_warn(j, nr_bits);
3659 return test_bit(j, mask);
3663 * netif_attr_test_online - Test for online CPU/Rx queue
3664 * @j: CPU/Rx queue index
3665 * @online_mask: bitmask for CPUs/Rx queues that are online
3666 * @nr_bits: number of bits in the bitmask
3668 * Returns true if a CPU/Rx queue is online.
3670 static inline bool netif_attr_test_online(unsigned long j,
3671 const unsigned long *online_mask,
3672 unsigned int nr_bits)
3674 cpu_max_bits_warn(j, nr_bits);
3677 return test_bit(j, online_mask);
3679 return (j < nr_bits);
3683 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3684 * @n: CPU/Rx queue index
3685 * @srcp: the cpumask/Rx queue mask pointer
3686 * @nr_bits: number of bits in the bitmask
3688 * Returns >= nr_bits if no further CPUs/Rx queues set.
3690 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3691 unsigned int nr_bits)
3693 /* -1 is a legal arg here. */
3695 cpu_max_bits_warn(n, nr_bits);
3698 return find_next_bit(srcp, nr_bits, n + 1);
3704 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3705 * @n: CPU/Rx queue index
3706 * @src1p: the first CPUs/Rx queues mask pointer
3707 * @src2p: the second CPUs/Rx queues mask pointer
3708 * @nr_bits: number of bits in the bitmask
3710 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3712 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3713 const unsigned long *src2p,
3714 unsigned int nr_bits)
3716 /* -1 is a legal arg here. */
3718 cpu_max_bits_warn(n, nr_bits);
3721 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3723 return find_next_bit(src1p, nr_bits, n + 1);
3725 return find_next_bit(src2p, nr_bits, n + 1);
3730 static inline int netif_set_xps_queue(struct net_device *dev,
3731 const struct cpumask *mask,
3737 static inline int __netif_set_xps_queue(struct net_device *dev,
3738 const unsigned long *mask,
3739 u16 index, enum xps_map_type type)
3746 * netif_is_multiqueue - test if device has multiple transmit queues
3747 * @dev: network device
3749 * Check if device has multiple transmit queues
3751 static inline bool netif_is_multiqueue(const struct net_device *dev)
3753 return dev->num_tx_queues > 1;
3756 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3759 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3761 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3764 dev->real_num_rx_queues = rxqs;
3768 int netif_set_real_num_queues(struct net_device *dev,
3769 unsigned int txq, unsigned int rxq);
3771 static inline struct netdev_rx_queue *
3772 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3774 return dev->_rx + rxq;
3778 static inline unsigned int get_netdev_rx_queue_index(
3779 struct netdev_rx_queue *queue)
3781 struct net_device *dev = queue->dev;
3782 int index = queue - dev->_rx;
3784 BUG_ON(index >= dev->num_rx_queues);
3789 int netif_get_num_default_rss_queues(void);
3791 enum skb_free_reason {
3792 SKB_REASON_CONSUMED,
3796 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3797 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3800 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3801 * interrupt context or with hardware interrupts being disabled.
3802 * (in_hardirq() || irqs_disabled())
3804 * We provide four helpers that can be used in following contexts :
3806 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3807 * replacing kfree_skb(skb)
3809 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3810 * Typically used in place of consume_skb(skb) in TX completion path
3812 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3813 * replacing kfree_skb(skb)
3815 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3816 * and consumed a packet. Used in place of consume_skb(skb)
3818 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3820 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3823 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3825 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3828 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3830 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3833 static inline void dev_consume_skb_any(struct sk_buff *skb)
3835 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3838 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3839 struct bpf_prog *xdp_prog);
3840 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3841 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3842 int netif_rx(struct sk_buff *skb);
3843 int __netif_rx(struct sk_buff *skb);
3845 int netif_receive_skb(struct sk_buff *skb);
3846 int netif_receive_skb_core(struct sk_buff *skb);
3847 void netif_receive_skb_list_internal(struct list_head *head);
3848 void netif_receive_skb_list(struct list_head *head);
3849 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3850 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3851 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3852 void napi_get_frags_check(struct napi_struct *napi);
3853 gro_result_t napi_gro_frags(struct napi_struct *napi);
3854 struct packet_offload *gro_find_receive_by_type(__be16 type);
3855 struct packet_offload *gro_find_complete_by_type(__be16 type);
3857 static inline void napi_free_frags(struct napi_struct *napi)
3859 kfree_skb(napi->skb);
3863 bool netdev_is_rx_handler_busy(struct net_device *dev);
3864 int netdev_rx_handler_register(struct net_device *dev,
3865 rx_handler_func_t *rx_handler,
3866 void *rx_handler_data);
3867 void netdev_rx_handler_unregister(struct net_device *dev);
3869 bool dev_valid_name(const char *name);
3870 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3872 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3874 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3875 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3876 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3877 void __user *data, bool *need_copyout);
3878 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3879 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3880 unsigned int dev_get_flags(const struct net_device *);
3881 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3882 struct netlink_ext_ack *extack);
3883 int dev_change_flags(struct net_device *dev, unsigned int flags,
3884 struct netlink_ext_ack *extack);
3885 int dev_set_alias(struct net_device *, const char *, size_t);
3886 int dev_get_alias(const struct net_device *, char *, size_t);
3887 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3888 const char *pat, int new_ifindex);
3890 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3893 return __dev_change_net_namespace(dev, net, pat, 0);
3895 int __dev_set_mtu(struct net_device *, int);
3896 int dev_set_mtu(struct net_device *, int);
3897 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3898 struct netlink_ext_ack *extack);
3899 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3900 struct netlink_ext_ack *extack);
3901 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3902 struct netlink_ext_ack *extack);
3903 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3904 int dev_get_port_parent_id(struct net_device *dev,
3905 struct netdev_phys_item_id *ppid, bool recurse);
3906 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3907 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3908 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3909 struct netdev_queue *txq, int *ret);
3911 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3912 u8 dev_xdp_prog_count(struct net_device *dev);
3913 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3915 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3916 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3917 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3918 bool is_skb_forwardable(const struct net_device *dev,
3919 const struct sk_buff *skb);
3921 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3922 const struct sk_buff *skb,
3923 const bool check_mtu)
3925 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3928 if (!(dev->flags & IFF_UP))
3934 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3935 if (skb->len <= len)
3938 /* if TSO is enabled, we don't care about the length as the packet
3939 * could be forwarded without being segmented before
3941 if (skb_is_gso(skb))
3947 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
3949 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
3951 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
3952 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
3957 return netdev_core_stats_alloc(dev);
3960 #define DEV_CORE_STATS_INC(FIELD) \
3961 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3963 struct net_device_core_stats __percpu *p; \
3965 p = dev_core_stats(dev); \
3967 this_cpu_inc(p->FIELD); \
3969 DEV_CORE_STATS_INC(rx_dropped)
3970 DEV_CORE_STATS_INC(tx_dropped)
3971 DEV_CORE_STATS_INC(rx_nohandler)
3972 DEV_CORE_STATS_INC(rx_otherhost_dropped)
3974 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3975 struct sk_buff *skb,
3976 const bool check_mtu)
3978 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3979 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3980 dev_core_stats_rx_dropped_inc(dev);
3985 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
3990 bool dev_nit_active(struct net_device *dev);
3991 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3993 static inline void __dev_put(struct net_device *dev)
3996 #ifdef CONFIG_PCPU_DEV_REFCNT
3997 this_cpu_dec(*dev->pcpu_refcnt);
3999 refcount_dec(&dev->dev_refcnt);
4004 static inline void __dev_hold(struct net_device *dev)
4007 #ifdef CONFIG_PCPU_DEV_REFCNT
4008 this_cpu_inc(*dev->pcpu_refcnt);
4010 refcount_inc(&dev->dev_refcnt);
4015 static inline void __netdev_tracker_alloc(struct net_device *dev,
4016 netdevice_tracker *tracker,
4019 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4020 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4024 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4025 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4027 static inline void netdev_tracker_alloc(struct net_device *dev,
4028 netdevice_tracker *tracker, gfp_t gfp)
4030 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4031 refcount_dec(&dev->refcnt_tracker.no_tracker);
4032 __netdev_tracker_alloc(dev, tracker, gfp);
4036 static inline void netdev_tracker_free(struct net_device *dev,
4037 netdevice_tracker *tracker)
4039 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4040 ref_tracker_free(&dev->refcnt_tracker, tracker);
4044 static inline void netdev_hold(struct net_device *dev,
4045 netdevice_tracker *tracker, gfp_t gfp)
4049 __netdev_tracker_alloc(dev, tracker, gfp);
4053 static inline void netdev_put(struct net_device *dev,
4054 netdevice_tracker *tracker)
4057 netdev_tracker_free(dev, tracker);
4063 * dev_hold - get reference to device
4064 * @dev: network device
4066 * Hold reference to device to keep it from being freed.
4067 * Try using netdev_hold() instead.
4069 static inline void dev_hold(struct net_device *dev)
4071 netdev_hold(dev, NULL, GFP_ATOMIC);
4075 * dev_put - release reference to device
4076 * @dev: network device
4078 * Release reference to device to allow it to be freed.
4079 * Try using netdev_put() instead.
4081 static inline void dev_put(struct net_device *dev)
4083 netdev_put(dev, NULL);
4086 static inline void netdev_ref_replace(struct net_device *odev,
4087 struct net_device *ndev,
4088 netdevice_tracker *tracker,
4092 netdev_tracker_free(odev, tracker);
4098 __netdev_tracker_alloc(ndev, tracker, gfp);
4101 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4102 * and _off may be called from IRQ context, but it is caller
4103 * who is responsible for serialization of these calls.
4105 * The name carrier is inappropriate, these functions should really be
4106 * called netif_lowerlayer_*() because they represent the state of any
4107 * kind of lower layer not just hardware media.
4109 void linkwatch_fire_event(struct net_device *dev);
4112 * netif_carrier_ok - test if carrier present
4113 * @dev: network device
4115 * Check if carrier is present on device
4117 static inline bool netif_carrier_ok(const struct net_device *dev)
4119 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4122 unsigned long dev_trans_start(struct net_device *dev);
4124 void __netdev_watchdog_up(struct net_device *dev);
4126 void netif_carrier_on(struct net_device *dev);
4127 void netif_carrier_off(struct net_device *dev);
4128 void netif_carrier_event(struct net_device *dev);
4131 * netif_dormant_on - mark device as dormant.
4132 * @dev: network device
4134 * Mark device as dormant (as per RFC2863).
4136 * The dormant state indicates that the relevant interface is not
4137 * actually in a condition to pass packets (i.e., it is not 'up') but is
4138 * in a "pending" state, waiting for some external event. For "on-
4139 * demand" interfaces, this new state identifies the situation where the
4140 * interface is waiting for events to place it in the up state.
4142 static inline void netif_dormant_on(struct net_device *dev)
4144 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4145 linkwatch_fire_event(dev);
4149 * netif_dormant_off - set device as not dormant.
4150 * @dev: network device
4152 * Device is not in dormant state.
4154 static inline void netif_dormant_off(struct net_device *dev)
4156 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4157 linkwatch_fire_event(dev);
4161 * netif_dormant - test if device is dormant
4162 * @dev: network device
4164 * Check if device is dormant.
4166 static inline bool netif_dormant(const struct net_device *dev)
4168 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4173 * netif_testing_on - mark device as under test.
4174 * @dev: network device
4176 * Mark device as under test (as per RFC2863).
4178 * The testing state indicates that some test(s) must be performed on
4179 * the interface. After completion, of the test, the interface state
4180 * will change to up, dormant, or down, as appropriate.
4182 static inline void netif_testing_on(struct net_device *dev)
4184 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4185 linkwatch_fire_event(dev);
4189 * netif_testing_off - set device as not under test.
4190 * @dev: network device
4192 * Device is not in testing state.
4194 static inline void netif_testing_off(struct net_device *dev)
4196 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4197 linkwatch_fire_event(dev);
4201 * netif_testing - test if device is under test
4202 * @dev: network device
4204 * Check if device is under test
4206 static inline bool netif_testing(const struct net_device *dev)
4208 return test_bit(__LINK_STATE_TESTING, &dev->state);
4213 * netif_oper_up - test if device is operational
4214 * @dev: network device
4216 * Check if carrier is operational
4218 static inline bool netif_oper_up(const struct net_device *dev)
4220 return (dev->operstate == IF_OPER_UP ||
4221 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4225 * netif_device_present - is device available or removed
4226 * @dev: network device
4228 * Check if device has not been removed from system.
4230 static inline bool netif_device_present(const struct net_device *dev)
4232 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4235 void netif_device_detach(struct net_device *dev);
4237 void netif_device_attach(struct net_device *dev);
4240 * Network interface message level settings
4245 NETIF_MSG_PROBE_BIT,
4247 NETIF_MSG_TIMER_BIT,
4248 NETIF_MSG_IFDOWN_BIT,
4250 NETIF_MSG_RX_ERR_BIT,
4251 NETIF_MSG_TX_ERR_BIT,
4252 NETIF_MSG_TX_QUEUED_BIT,
4254 NETIF_MSG_TX_DONE_BIT,
4255 NETIF_MSG_RX_STATUS_BIT,
4256 NETIF_MSG_PKTDATA_BIT,
4260 /* When you add a new bit above, update netif_msg_class_names array
4261 * in net/ethtool/common.c
4263 NETIF_MSG_CLASS_COUNT,
4265 /* Both ethtool_ops interface and internal driver implementation use u32 */
4266 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4268 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4269 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4271 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4272 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4273 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4274 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4275 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4276 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4277 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4278 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4279 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4280 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4281 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4282 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4283 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4284 #define NETIF_MSG_HW __NETIF_MSG(HW)
4285 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4287 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4288 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4289 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4290 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4291 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4292 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4293 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4294 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4295 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4296 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4297 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4298 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4299 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4300 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4301 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4303 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4306 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4307 return default_msg_enable_bits;
4308 if (debug_value == 0) /* no output */
4310 /* set low N bits */
4311 return (1U << debug_value) - 1;
4314 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4316 spin_lock(&txq->_xmit_lock);
4317 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4318 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4321 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4323 __acquire(&txq->_xmit_lock);
4327 static inline void __netif_tx_release(struct netdev_queue *txq)
4329 __release(&txq->_xmit_lock);
4332 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4334 spin_lock_bh(&txq->_xmit_lock);
4335 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4336 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4339 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4341 bool ok = spin_trylock(&txq->_xmit_lock);
4344 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4345 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4350 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4352 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4353 WRITE_ONCE(txq->xmit_lock_owner, -1);
4354 spin_unlock(&txq->_xmit_lock);
4357 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4359 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4360 WRITE_ONCE(txq->xmit_lock_owner, -1);
4361 spin_unlock_bh(&txq->_xmit_lock);
4365 * txq->trans_start can be read locklessly from dev_watchdog()
4367 static inline void txq_trans_update(struct netdev_queue *txq)
4369 if (txq->xmit_lock_owner != -1)
4370 WRITE_ONCE(txq->trans_start, jiffies);
4373 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4375 unsigned long now = jiffies;
4377 if (READ_ONCE(txq->trans_start) != now)
4378 WRITE_ONCE(txq->trans_start, now);
4381 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4382 static inline void netif_trans_update(struct net_device *dev)
4384 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4386 txq_trans_cond_update(txq);
4390 * netif_tx_lock - grab network device transmit lock
4391 * @dev: network device
4393 * Get network device transmit lock
4395 void netif_tx_lock(struct net_device *dev);
4397 static inline void netif_tx_lock_bh(struct net_device *dev)
4403 void netif_tx_unlock(struct net_device *dev);
4405 static inline void netif_tx_unlock_bh(struct net_device *dev)
4407 netif_tx_unlock(dev);
4411 #define HARD_TX_LOCK(dev, txq, cpu) { \
4412 if ((dev->features & NETIF_F_LLTX) == 0) { \
4413 __netif_tx_lock(txq, cpu); \
4415 __netif_tx_acquire(txq); \
4419 #define HARD_TX_TRYLOCK(dev, txq) \
4420 (((dev->features & NETIF_F_LLTX) == 0) ? \
4421 __netif_tx_trylock(txq) : \
4422 __netif_tx_acquire(txq))
4424 #define HARD_TX_UNLOCK(dev, txq) { \
4425 if ((dev->features & NETIF_F_LLTX) == 0) { \
4426 __netif_tx_unlock(txq); \
4428 __netif_tx_release(txq); \
4432 static inline void netif_tx_disable(struct net_device *dev)
4438 cpu = smp_processor_id();
4439 spin_lock(&dev->tx_global_lock);
4440 for (i = 0; i < dev->num_tx_queues; i++) {
4441 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4443 __netif_tx_lock(txq, cpu);
4444 netif_tx_stop_queue(txq);
4445 __netif_tx_unlock(txq);
4447 spin_unlock(&dev->tx_global_lock);
4451 static inline void netif_addr_lock(struct net_device *dev)
4453 unsigned char nest_level = 0;
4455 #ifdef CONFIG_LOCKDEP
4456 nest_level = dev->nested_level;
4458 spin_lock_nested(&dev->addr_list_lock, nest_level);
4461 static inline void netif_addr_lock_bh(struct net_device *dev)
4463 unsigned char nest_level = 0;
4465 #ifdef CONFIG_LOCKDEP
4466 nest_level = dev->nested_level;
4469 spin_lock_nested(&dev->addr_list_lock, nest_level);
4472 static inline void netif_addr_unlock(struct net_device *dev)
4474 spin_unlock(&dev->addr_list_lock);
4477 static inline void netif_addr_unlock_bh(struct net_device *dev)
4479 spin_unlock_bh(&dev->addr_list_lock);
4483 * dev_addrs walker. Should be used only for read access. Call with
4484 * rcu_read_lock held.
4486 #define for_each_dev_addr(dev, ha) \
4487 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4489 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4491 void ether_setup(struct net_device *dev);
4493 /* Support for loadable net-drivers */
4494 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4495 unsigned char name_assign_type,
4496 void (*setup)(struct net_device *),
4497 unsigned int txqs, unsigned int rxqs);
4498 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4499 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4501 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4502 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4505 int register_netdev(struct net_device *dev);
4506 void unregister_netdev(struct net_device *dev);
4508 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4510 /* General hardware address lists handling functions */
4511 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4512 struct netdev_hw_addr_list *from_list, int addr_len);
4513 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4514 struct netdev_hw_addr_list *from_list, int addr_len);
4515 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4516 struct net_device *dev,
4517 int (*sync)(struct net_device *, const unsigned char *),
4518 int (*unsync)(struct net_device *,
4519 const unsigned char *));
4520 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4521 struct net_device *dev,
4522 int (*sync)(struct net_device *,
4523 const unsigned char *, int),
4524 int (*unsync)(struct net_device *,
4525 const unsigned char *, int));
4526 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4527 struct net_device *dev,
4528 int (*unsync)(struct net_device *,
4529 const unsigned char *, int));
4530 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4531 struct net_device *dev,
4532 int (*unsync)(struct net_device *,
4533 const unsigned char *));
4534 void __hw_addr_init(struct netdev_hw_addr_list *list);
4536 /* Functions used for device addresses handling */
4537 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4538 const void *addr, size_t len);
4541 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4543 dev_addr_mod(dev, 0, addr, len);
4546 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4548 __dev_addr_set(dev, addr, dev->addr_len);
4551 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4552 unsigned char addr_type);
4553 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4554 unsigned char addr_type);
4556 /* Functions used for unicast addresses handling */
4557 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4558 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4559 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4560 int dev_uc_sync(struct net_device *to, struct net_device *from);
4561 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4562 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4563 void dev_uc_flush(struct net_device *dev);
4564 void dev_uc_init(struct net_device *dev);
4567 * __dev_uc_sync - Synchonize device's unicast list
4568 * @dev: device to sync
4569 * @sync: function to call if address should be added
4570 * @unsync: function to call if address should be removed
4572 * Add newly added addresses to the interface, and release
4573 * addresses that have been deleted.
4575 static inline int __dev_uc_sync(struct net_device *dev,
4576 int (*sync)(struct net_device *,
4577 const unsigned char *),
4578 int (*unsync)(struct net_device *,
4579 const unsigned char *))
4581 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4585 * __dev_uc_unsync - Remove synchronized addresses from device
4586 * @dev: device to sync
4587 * @unsync: function to call if address should be removed
4589 * Remove all addresses that were added to the device by dev_uc_sync().
4591 static inline void __dev_uc_unsync(struct net_device *dev,
4592 int (*unsync)(struct net_device *,
4593 const unsigned char *))
4595 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4598 /* Functions used for multicast addresses handling */
4599 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4600 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4601 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4602 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4603 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4604 int dev_mc_sync(struct net_device *to, struct net_device *from);
4605 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4606 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4607 void dev_mc_flush(struct net_device *dev);
4608 void dev_mc_init(struct net_device *dev);
4611 * __dev_mc_sync - Synchonize device's multicast list
4612 * @dev: device to sync
4613 * @sync: function to call if address should be added
4614 * @unsync: function to call if address should be removed
4616 * Add newly added addresses to the interface, and release
4617 * addresses that have been deleted.
4619 static inline int __dev_mc_sync(struct net_device *dev,
4620 int (*sync)(struct net_device *,
4621 const unsigned char *),
4622 int (*unsync)(struct net_device *,
4623 const unsigned char *))
4625 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4629 * __dev_mc_unsync - Remove synchronized addresses from device
4630 * @dev: device to sync
4631 * @unsync: function to call if address should be removed
4633 * Remove all addresses that were added to the device by dev_mc_sync().
4635 static inline void __dev_mc_unsync(struct net_device *dev,
4636 int (*unsync)(struct net_device *,
4637 const unsigned char *))
4639 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4642 /* Functions used for secondary unicast and multicast support */
4643 void dev_set_rx_mode(struct net_device *dev);
4644 int dev_set_promiscuity(struct net_device *dev, int inc);
4645 int dev_set_allmulti(struct net_device *dev, int inc);
4646 void netdev_state_change(struct net_device *dev);
4647 void __netdev_notify_peers(struct net_device *dev);
4648 void netdev_notify_peers(struct net_device *dev);
4649 void netdev_features_change(struct net_device *dev);
4650 /* Load a device via the kmod */
4651 void dev_load(struct net *net, const char *name);
4652 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4653 struct rtnl_link_stats64 *storage);
4654 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4655 const struct net_device_stats *netdev_stats);
4656 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4657 const struct pcpu_sw_netstats __percpu *netstats);
4658 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4660 extern int netdev_max_backlog;
4661 extern int dev_rx_weight;
4662 extern int dev_tx_weight;
4663 extern int gro_normal_batch;
4666 NESTED_SYNC_IMM_BIT,
4667 NESTED_SYNC_TODO_BIT,
4670 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4671 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4673 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4674 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4676 struct netdev_nested_priv {
4677 unsigned char flags;
4681 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4682 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4683 struct list_head **iter);
4685 /* iterate through upper list, must be called under RCU read lock */
4686 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4687 for (iter = &(dev)->adj_list.upper, \
4688 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4690 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4692 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4693 int (*fn)(struct net_device *upper_dev,
4694 struct netdev_nested_priv *priv),
4695 struct netdev_nested_priv *priv);
4697 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4698 struct net_device *upper_dev);
4700 bool netdev_has_any_upper_dev(struct net_device *dev);
4702 void *netdev_lower_get_next_private(struct net_device *dev,
4703 struct list_head **iter);
4704 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4705 struct list_head **iter);
4707 #define netdev_for_each_lower_private(dev, priv, iter) \
4708 for (iter = (dev)->adj_list.lower.next, \
4709 priv = netdev_lower_get_next_private(dev, &(iter)); \
4711 priv = netdev_lower_get_next_private(dev, &(iter)))
4713 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4714 for (iter = &(dev)->adj_list.lower, \
4715 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4717 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4719 void *netdev_lower_get_next(struct net_device *dev,
4720 struct list_head **iter);
4722 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4723 for (iter = (dev)->adj_list.lower.next, \
4724 ldev = netdev_lower_get_next(dev, &(iter)); \
4726 ldev = netdev_lower_get_next(dev, &(iter)))
4728 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4729 struct list_head **iter);
4730 int netdev_walk_all_lower_dev(struct net_device *dev,
4731 int (*fn)(struct net_device *lower_dev,
4732 struct netdev_nested_priv *priv),
4733 struct netdev_nested_priv *priv);
4734 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4735 int (*fn)(struct net_device *lower_dev,
4736 struct netdev_nested_priv *priv),
4737 struct netdev_nested_priv *priv);
4739 void *netdev_adjacent_get_private(struct list_head *adj_list);
4740 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4741 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4742 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4743 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4744 struct netlink_ext_ack *extack);
4745 int netdev_master_upper_dev_link(struct net_device *dev,
4746 struct net_device *upper_dev,
4747 void *upper_priv, void *upper_info,
4748 struct netlink_ext_ack *extack);
4749 void netdev_upper_dev_unlink(struct net_device *dev,
4750 struct net_device *upper_dev);
4751 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4752 struct net_device *new_dev,
4753 struct net_device *dev,
4754 struct netlink_ext_ack *extack);
4755 void netdev_adjacent_change_commit(struct net_device *old_dev,
4756 struct net_device *new_dev,
4757 struct net_device *dev);
4758 void netdev_adjacent_change_abort(struct net_device *old_dev,
4759 struct net_device *new_dev,
4760 struct net_device *dev);
4761 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4762 void *netdev_lower_dev_get_private(struct net_device *dev,
4763 struct net_device *lower_dev);
4764 void netdev_lower_state_changed(struct net_device *lower_dev,
4765 void *lower_state_info);
4767 /* RSS keys are 40 or 52 bytes long */
4768 #define NETDEV_RSS_KEY_LEN 52
4769 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4770 void netdev_rss_key_fill(void *buffer, size_t len);
4772 int skb_checksum_help(struct sk_buff *skb);
4773 int skb_crc32c_csum_help(struct sk_buff *skb);
4774 int skb_csum_hwoffload_help(struct sk_buff *skb,
4775 const netdev_features_t features);
4777 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4778 netdev_features_t features, bool tx_path);
4779 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb,
4780 netdev_features_t features, __be16 type);
4781 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4782 netdev_features_t features);
4784 struct netdev_bonding_info {
4789 struct netdev_notifier_bonding_info {
4790 struct netdev_notifier_info info; /* must be first */
4791 struct netdev_bonding_info bonding_info;
4794 void netdev_bonding_info_change(struct net_device *dev,
4795 struct netdev_bonding_info *bonding_info);
4797 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4798 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4800 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4807 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4809 return __skb_gso_segment(skb, features, true);
4811 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4813 static inline bool can_checksum_protocol(netdev_features_t features,
4816 if (protocol == htons(ETH_P_FCOE))
4817 return !!(features & NETIF_F_FCOE_CRC);
4819 /* Assume this is an IP checksum (not SCTP CRC) */
4821 if (features & NETIF_F_HW_CSUM) {
4822 /* Can checksum everything */
4827 case htons(ETH_P_IP):
4828 return !!(features & NETIF_F_IP_CSUM);
4829 case htons(ETH_P_IPV6):
4830 return !!(features & NETIF_F_IPV6_CSUM);
4837 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4839 static inline void netdev_rx_csum_fault(struct net_device *dev,
4840 struct sk_buff *skb)
4844 /* rx skb timestamps */
4845 void net_enable_timestamp(void);
4846 void net_disable_timestamp(void);
4848 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4849 const struct skb_shared_hwtstamps *hwtstamps,
4852 const struct net_device_ops *ops = dev->netdev_ops;
4854 if (ops->ndo_get_tstamp)
4855 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4857 return hwtstamps->hwtstamp;
4860 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4861 struct sk_buff *skb, struct net_device *dev,
4864 __this_cpu_write(softnet_data.xmit.more, more);
4865 return ops->ndo_start_xmit(skb, dev);
4868 static inline bool netdev_xmit_more(void)
4870 return __this_cpu_read(softnet_data.xmit.more);
4873 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4874 struct netdev_queue *txq, bool more)
4876 const struct net_device_ops *ops = dev->netdev_ops;
4879 rc = __netdev_start_xmit(ops, skb, dev, more);
4880 if (rc == NETDEV_TX_OK)
4881 txq_trans_update(txq);
4886 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4888 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4891 extern const struct kobj_ns_type_operations net_ns_type_operations;
4893 const char *netdev_drivername(const struct net_device *dev);
4895 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4896 netdev_features_t f2)
4898 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4899 if (f1 & NETIF_F_HW_CSUM)
4900 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4902 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4908 static inline netdev_features_t netdev_get_wanted_features(
4909 struct net_device *dev)
4911 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4913 netdev_features_t netdev_increment_features(netdev_features_t all,
4914 netdev_features_t one, netdev_features_t mask);
4916 /* Allow TSO being used on stacked device :
4917 * Performing the GSO segmentation before last device
4918 * is a performance improvement.
4920 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4921 netdev_features_t mask)
4923 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4926 int __netdev_update_features(struct net_device *dev);
4927 void netdev_update_features(struct net_device *dev);
4928 void netdev_change_features(struct net_device *dev);
4930 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4931 struct net_device *dev);
4933 netdev_features_t passthru_features_check(struct sk_buff *skb,
4934 struct net_device *dev,
4935 netdev_features_t features);
4936 netdev_features_t netif_skb_features(struct sk_buff *skb);
4938 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4940 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4942 /* check flags correspondence */
4943 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4944 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4945 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4946 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4947 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4948 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4949 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4950 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4951 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4952 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4953 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4954 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4955 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4956 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4957 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4958 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4959 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4960 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4961 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4963 return (features & feature) == feature;
4966 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4968 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4969 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4972 static inline bool netif_needs_gso(struct sk_buff *skb,
4973 netdev_features_t features)
4975 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4976 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4977 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4980 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
4981 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
4982 void netif_inherit_tso_max(struct net_device *to,
4983 const struct net_device *from);
4985 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4986 int pulled_hlen, u16 mac_offset,
4989 skb->protocol = protocol;
4990 skb->encapsulation = 1;
4991 skb_push(skb, pulled_hlen);
4992 skb_reset_transport_header(skb);
4993 skb->mac_header = mac_offset;
4994 skb->network_header = skb->mac_header + mac_len;
4995 skb->mac_len = mac_len;
4998 static inline bool netif_is_macsec(const struct net_device *dev)
5000 return dev->priv_flags & IFF_MACSEC;
5003 static inline bool netif_is_macvlan(const struct net_device *dev)
5005 return dev->priv_flags & IFF_MACVLAN;
5008 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5010 return dev->priv_flags & IFF_MACVLAN_PORT;
5013 static inline bool netif_is_bond_master(const struct net_device *dev)
5015 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5018 static inline bool netif_is_bond_slave(const struct net_device *dev)
5020 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5023 static inline bool netif_supports_nofcs(struct net_device *dev)
5025 return dev->priv_flags & IFF_SUPP_NOFCS;
5028 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5030 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5033 static inline bool netif_is_l3_master(const struct net_device *dev)
5035 return dev->priv_flags & IFF_L3MDEV_MASTER;
5038 static inline bool netif_is_l3_slave(const struct net_device *dev)
5040 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5043 static inline bool netif_is_bridge_master(const struct net_device *dev)
5045 return dev->priv_flags & IFF_EBRIDGE;
5048 static inline bool netif_is_bridge_port(const struct net_device *dev)
5050 return dev->priv_flags & IFF_BRIDGE_PORT;
5053 static inline bool netif_is_ovs_master(const struct net_device *dev)
5055 return dev->priv_flags & IFF_OPENVSWITCH;
5058 static inline bool netif_is_ovs_port(const struct net_device *dev)
5060 return dev->priv_flags & IFF_OVS_DATAPATH;
5063 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5065 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5068 static inline bool netif_is_team_master(const struct net_device *dev)
5070 return dev->priv_flags & IFF_TEAM;
5073 static inline bool netif_is_team_port(const struct net_device *dev)
5075 return dev->priv_flags & IFF_TEAM_PORT;
5078 static inline bool netif_is_lag_master(const struct net_device *dev)
5080 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5083 static inline bool netif_is_lag_port(const struct net_device *dev)
5085 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5088 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5090 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5093 static inline bool netif_is_failover(const struct net_device *dev)
5095 return dev->priv_flags & IFF_FAILOVER;
5098 static inline bool netif_is_failover_slave(const struct net_device *dev)
5100 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5103 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5104 static inline void netif_keep_dst(struct net_device *dev)
5106 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5109 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5110 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5112 /* TODO: reserve and use an additional IFF bit, if we get more users */
5113 return netif_is_macsec(dev);
5116 extern struct pernet_operations __net_initdata loopback_net_ops;
5118 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5120 /* netdev_printk helpers, similar to dev_printk */
5122 static inline const char *netdev_name(const struct net_device *dev)
5124 if (!dev->name[0] || strchr(dev->name, '%'))
5125 return "(unnamed net_device)";
5129 static inline const char *netdev_reg_state(const struct net_device *dev)
5131 switch (dev->reg_state) {
5132 case NETREG_UNINITIALIZED: return " (uninitialized)";
5133 case NETREG_REGISTERED: return "";
5134 case NETREG_UNREGISTERING: return " (unregistering)";
5135 case NETREG_UNREGISTERED: return " (unregistered)";
5136 case NETREG_RELEASED: return " (released)";
5137 case NETREG_DUMMY: return " (dummy)";
5140 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5141 return " (unknown)";
5144 #define MODULE_ALIAS_NETDEV(device) \
5145 MODULE_ALIAS("netdev-" device)
5148 * netdev_WARN() acts like dev_printk(), but with the key difference
5149 * of using a WARN/WARN_ON to get the message out, including the
5150 * file/line information and a backtrace.
5152 #define netdev_WARN(dev, format, args...) \
5153 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5154 netdev_reg_state(dev), ##args)
5156 #define netdev_WARN_ONCE(dev, format, args...) \
5157 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5158 netdev_reg_state(dev), ##args)
5161 * The list of packet types we will receive (as opposed to discard)
5162 * and the routines to invoke.
5164 * Why 16. Because with 16 the only overlap we get on a hash of the
5165 * low nibble of the protocol value is RARP/SNAP/X.25.
5179 #define PTYPE_HASH_SIZE (16)
5180 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5182 extern struct list_head ptype_all __read_mostly;
5183 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5185 extern struct net_device *blackhole_netdev;
5187 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5188 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5189 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5190 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5192 #endif /* _LINUX_NETDEVICE_H */