1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Definitions for the Interfaces handler.
9 * Version: @(#)dev.h 1.0.10 08/12/93
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
19 * Moved to /usr/include/linux for NET3
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
31 #include <asm/local.h>
33 #include <linux/percpu.h>
34 #include <linux/rculist.h>
35 #include <linux/workqueue.h>
36 #include <linux/dynamic_queue_limits.h>
38 #include <net/net_namespace.h>
40 #include <net/dcbnl.h>
42 #include <net/netprio_cgroup.h>
44 #include <linux/netdev_features.h>
45 #include <linux/neighbour.h>
46 #include <uapi/linux/netdevice.h>
47 #include <uapi/linux/if_bonding.h>
48 #include <uapi/linux/pkt_cls.h>
49 #include <uapi/linux/netdev.h>
50 #include <uapi/linux/net_tstamp.h>
51 #include <linux/hashtable.h>
52 #include <linux/rbtree.h>
53 #include <net/net_trackers.h>
54 #include <net/net_debug.h>
55 #include <net/dropreason-core.h>
60 struct kernel_hwtstamp_config;
63 struct ip_tunnel_parm;
64 struct macsec_context;
66 struct netdev_name_node;
71 /* 802.15.4 specific */
74 /* UDP Tunnel offloads */
75 struct udp_tunnel_info;
76 struct udp_tunnel_nic_info;
77 struct udp_tunnel_nic;
81 struct xdp_metadata_ops;
86 typedef u32 xdp_features_t;
88 void synchronize_net(void);
89 void netdev_set_default_ethtool_ops(struct net_device *dev,
90 const struct ethtool_ops *ops);
91 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
93 /* Backlog congestion levels */
94 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
95 #define NET_RX_DROP 1 /* packet dropped */
97 #define MAX_NEST_DEV 8
100 * Transmit return codes: transmit return codes originate from three different
103 * - qdisc return codes
104 * - driver transmit return codes
107 * Drivers are allowed to return any one of those in their hard_start_xmit()
108 * function. Real network devices commonly used with qdiscs should only return
109 * the driver transmit return codes though - when qdiscs are used, the actual
110 * transmission happens asynchronously, so the value is not propagated to
111 * higher layers. Virtual network devices transmit synchronously; in this case
112 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
113 * others are propagated to higher layers.
116 /* qdisc ->enqueue() return codes. */
117 #define NET_XMIT_SUCCESS 0x00
118 #define NET_XMIT_DROP 0x01 /* skb dropped */
119 #define NET_XMIT_CN 0x02 /* congestion notification */
120 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
122 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
123 * indicates that the device will soon be dropping packets, or already drops
124 * some packets of the same priority; prompting us to send less aggressively. */
125 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
126 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
128 /* Driver transmit return codes */
129 #define NETDEV_TX_MASK 0xf0
132 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
133 NETDEV_TX_OK = 0x00, /* driver took care of packet */
134 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
136 typedef enum netdev_tx netdev_tx_t;
139 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
140 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
142 static inline bool dev_xmit_complete(int rc)
145 * Positive cases with an skb consumed by a driver:
146 * - successful transmission (rc == NETDEV_TX_OK)
147 * - error while transmitting (rc < 0)
148 * - error while queueing to a different device (rc & NET_XMIT_MASK)
150 if (likely(rc < NET_XMIT_MASK))
157 * Compute the worst-case header length according to the protocols
161 #if defined(CONFIG_HYPERV_NET)
162 # define LL_MAX_HEADER 128
163 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
164 # if defined(CONFIG_MAC80211_MESH)
165 # define LL_MAX_HEADER 128
167 # define LL_MAX_HEADER 96
170 # define LL_MAX_HEADER 32
173 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
174 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
175 #define MAX_HEADER LL_MAX_HEADER
177 #define MAX_HEADER (LL_MAX_HEADER + 48)
181 * Old network device statistics. Fields are native words
182 * (unsigned long) so they can be read and written atomically.
185 #define NET_DEV_STAT(FIELD) \
187 unsigned long FIELD; \
188 atomic_long_t __##FIELD; \
191 struct net_device_stats {
192 NET_DEV_STAT(rx_packets);
193 NET_DEV_STAT(tx_packets);
194 NET_DEV_STAT(rx_bytes);
195 NET_DEV_STAT(tx_bytes);
196 NET_DEV_STAT(rx_errors);
197 NET_DEV_STAT(tx_errors);
198 NET_DEV_STAT(rx_dropped);
199 NET_DEV_STAT(tx_dropped);
200 NET_DEV_STAT(multicast);
201 NET_DEV_STAT(collisions);
202 NET_DEV_STAT(rx_length_errors);
203 NET_DEV_STAT(rx_over_errors);
204 NET_DEV_STAT(rx_crc_errors);
205 NET_DEV_STAT(rx_frame_errors);
206 NET_DEV_STAT(rx_fifo_errors);
207 NET_DEV_STAT(rx_missed_errors);
208 NET_DEV_STAT(tx_aborted_errors);
209 NET_DEV_STAT(tx_carrier_errors);
210 NET_DEV_STAT(tx_fifo_errors);
211 NET_DEV_STAT(tx_heartbeat_errors);
212 NET_DEV_STAT(tx_window_errors);
213 NET_DEV_STAT(rx_compressed);
214 NET_DEV_STAT(tx_compressed);
218 /* per-cpu stats, allocated on demand.
219 * Try to fit them in a single cache line, for dev_get_stats() sake.
221 struct net_device_core_stats {
222 unsigned long rx_dropped;
223 unsigned long tx_dropped;
224 unsigned long rx_nohandler;
225 unsigned long rx_otherhost_dropped;
226 } __aligned(4 * sizeof(unsigned long));
228 #include <linux/cache.h>
229 #include <linux/skbuff.h>
232 #include <linux/static_key.h>
233 extern struct static_key_false rps_needed;
234 extern struct static_key_false rfs_needed;
241 struct netdev_hw_addr {
242 struct list_head list;
244 unsigned char addr[MAX_ADDR_LEN];
246 #define NETDEV_HW_ADDR_T_LAN 1
247 #define NETDEV_HW_ADDR_T_SAN 2
248 #define NETDEV_HW_ADDR_T_UNICAST 3
249 #define NETDEV_HW_ADDR_T_MULTICAST 4
254 struct rcu_head rcu_head;
257 struct netdev_hw_addr_list {
258 struct list_head list;
261 /* Auxiliary tree for faster lookup on addition and deletion */
265 #define netdev_hw_addr_list_count(l) ((l)->count)
266 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
267 #define netdev_hw_addr_list_for_each(ha, l) \
268 list_for_each_entry(ha, &(l)->list, list)
270 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
271 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
272 #define netdev_for_each_uc_addr(ha, dev) \
273 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
274 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
275 netdev_for_each_uc_addr((_ha), (_dev)) \
278 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
279 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
280 #define netdev_for_each_mc_addr(ha, dev) \
281 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
282 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
283 netdev_for_each_mc_addr((_ha), (_dev)) \
290 /* cached hardware header; allow for machine alignment needs. */
291 #define HH_DATA_MOD 16
292 #define HH_DATA_OFF(__len) \
293 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
294 #define HH_DATA_ALIGN(__len) \
295 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
296 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
299 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
301 * dev->hard_header_len ? (dev->hard_header_len +
302 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
304 * We could use other alignment values, but we must maintain the
305 * relationship HH alignment <= LL alignment.
307 #define LL_RESERVED_SPACE(dev) \
308 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
309 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
310 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
311 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
312 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
315 int (*create) (struct sk_buff *skb, struct net_device *dev,
316 unsigned short type, const void *daddr,
317 const void *saddr, unsigned int len);
318 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
319 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
320 void (*cache_update)(struct hh_cache *hh,
321 const struct net_device *dev,
322 const unsigned char *haddr);
323 bool (*validate)(const char *ll_header, unsigned int len);
324 __be16 (*parse_protocol)(const struct sk_buff *skb);
327 /* These flag bits are private to the generic network queueing
328 * layer; they may not be explicitly referenced by any other
332 enum netdev_state_t {
334 __LINK_STATE_PRESENT,
335 __LINK_STATE_NOCARRIER,
336 __LINK_STATE_LINKWATCH_PENDING,
337 __LINK_STATE_DORMANT,
338 __LINK_STATE_TESTING,
342 struct list_head list;
347 * size of gro hash buckets, must less than bit number of
348 * napi_struct::gro_bitmask
350 #define GRO_HASH_BUCKETS 8
353 * Structure for NAPI scheduling similar to tasklet but with weighting
356 /* The poll_list must only be managed by the entity which
357 * changes the state of the NAPI_STATE_SCHED bit. This means
358 * whoever atomically sets that bit can add this napi_struct
359 * to the per-CPU poll_list, and whoever clears that bit
360 * can remove from the list right before clearing the bit.
362 struct list_head poll_list;
366 int defer_hard_irqs_count;
367 unsigned long gro_bitmask;
368 int (*poll)(struct napi_struct *, int);
369 #ifdef CONFIG_NETPOLL
370 /* CPU actively polling if netpoll is configured */
373 /* CPU on which NAPI has been scheduled for processing */
375 struct net_device *dev;
376 struct gro_list gro_hash[GRO_HASH_BUCKETS];
378 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
379 int rx_count; /* length of rx_list */
380 unsigned int napi_id;
381 struct hrtimer timer;
382 struct task_struct *thread;
383 /* control-path-only fields follow */
384 struct list_head dev_list;
385 struct hlist_node napi_hash_node;
389 NAPI_STATE_SCHED, /* Poll is scheduled */
390 NAPI_STATE_MISSED, /* reschedule a napi */
391 NAPI_STATE_DISABLE, /* Disable pending */
392 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
393 NAPI_STATE_LISTED, /* NAPI added to system lists */
394 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
395 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
396 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
397 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
398 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
402 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
403 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
404 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
405 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
406 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
407 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
408 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
409 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
410 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
411 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
421 typedef enum gro_result gro_result_t;
424 * enum rx_handler_result - Possible return values for rx_handlers.
425 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
427 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
428 * case skb->dev was changed by rx_handler.
429 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
430 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
432 * rx_handlers are functions called from inside __netif_receive_skb(), to do
433 * special processing of the skb, prior to delivery to protocol handlers.
435 * Currently, a net_device can only have a single rx_handler registered. Trying
436 * to register a second rx_handler will return -EBUSY.
438 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
439 * To unregister a rx_handler on a net_device, use
440 * netdev_rx_handler_unregister().
442 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
445 * If the rx_handler consumed the skb in some way, it should return
446 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
447 * the skb to be delivered in some other way.
449 * If the rx_handler changed skb->dev, to divert the skb to another
450 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
451 * new device will be called if it exists.
453 * If the rx_handler decides the skb should be ignored, it should return
454 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
455 * are registered on exact device (ptype->dev == skb->dev).
457 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
458 * delivered, it should return RX_HANDLER_PASS.
460 * A device without a registered rx_handler will behave as if rx_handler
461 * returned RX_HANDLER_PASS.
464 enum rx_handler_result {
470 typedef enum rx_handler_result rx_handler_result_t;
471 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
473 void __napi_schedule(struct napi_struct *n);
474 void __napi_schedule_irqoff(struct napi_struct *n);
476 static inline bool napi_disable_pending(struct napi_struct *n)
478 return test_bit(NAPI_STATE_DISABLE, &n->state);
481 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
483 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
487 * napi_is_scheduled - test if NAPI is scheduled
490 * This check is "best-effort". With no locking implemented,
491 * a NAPI can be scheduled or terminate right after this check
492 * and produce not precise results.
494 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
495 * should not be used normally and napi_schedule should be
498 * Use only if the driver really needs to check if a NAPI
499 * is scheduled for example in the context of delayed timer
500 * that can be skipped if a NAPI is already scheduled.
502 * Return True if NAPI is scheduled, False otherwise.
504 static inline bool napi_is_scheduled(struct napi_struct *n)
506 return test_bit(NAPI_STATE_SCHED, &n->state);
509 bool napi_schedule_prep(struct napi_struct *n);
512 * napi_schedule - schedule NAPI poll
515 * Schedule NAPI poll routine to be called if it is not already
517 * Return true if we schedule a NAPI or false if not.
518 * Refer to napi_schedule_prep() for additional reason on why
519 * a NAPI might not be scheduled.
521 static inline bool napi_schedule(struct napi_struct *n)
523 if (napi_schedule_prep(n)) {
532 * napi_schedule_irqoff - schedule NAPI poll
535 * Variant of napi_schedule(), assuming hard irqs are masked.
537 static inline void napi_schedule_irqoff(struct napi_struct *n)
539 if (napi_schedule_prep(n))
540 __napi_schedule_irqoff(n);
544 * napi_complete_done - NAPI processing complete
546 * @work_done: number of packets processed
548 * Mark NAPI processing as complete. Should only be called if poll budget
549 * has not been completely consumed.
550 * Prefer over napi_complete().
551 * Return false if device should avoid rearming interrupts.
553 bool napi_complete_done(struct napi_struct *n, int work_done);
555 static inline bool napi_complete(struct napi_struct *n)
557 return napi_complete_done(n, 0);
560 int dev_set_threaded(struct net_device *dev, bool threaded);
563 * napi_disable - prevent NAPI from scheduling
566 * Stop NAPI from being scheduled on this context.
567 * Waits till any outstanding processing completes.
569 void napi_disable(struct napi_struct *n);
571 void napi_enable(struct napi_struct *n);
574 * napi_synchronize - wait until NAPI is not running
577 * Wait until NAPI is done being scheduled on this context.
578 * Waits till any outstanding processing completes but
579 * does not disable future activations.
581 static inline void napi_synchronize(const struct napi_struct *n)
583 if (IS_ENABLED(CONFIG_SMP))
584 while (test_bit(NAPI_STATE_SCHED, &n->state))
591 * napi_if_scheduled_mark_missed - if napi is running, set the
595 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
598 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
600 unsigned long val, new;
602 val = READ_ONCE(n->state);
604 if (val & NAPIF_STATE_DISABLE)
607 if (!(val & NAPIF_STATE_SCHED))
610 new = val | NAPIF_STATE_MISSED;
611 } while (!try_cmpxchg(&n->state, &val, new));
616 enum netdev_queue_state_t {
617 __QUEUE_STATE_DRV_XOFF,
618 __QUEUE_STATE_STACK_XOFF,
619 __QUEUE_STATE_FROZEN,
622 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
623 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
624 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
626 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
627 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
629 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
633 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
634 * netif_tx_* functions below are used to manipulate this flag. The
635 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
636 * queue independently. The netif_xmit_*stopped functions below are called
637 * to check if the queue has been stopped by the driver or stack (either
638 * of the XOFF bits are set in the state). Drivers should not need to call
639 * netif_xmit*stopped functions, they should only be using netif_tx_*.
642 struct netdev_queue {
646 struct net_device *dev;
647 netdevice_tracker dev_tracker;
649 struct Qdisc __rcu *qdisc;
650 struct Qdisc __rcu *qdisc_sleeping;
654 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
657 unsigned long tx_maxrate;
659 * Number of TX timeouts for this queue
660 * (/sys/class/net/DEV/Q/trans_timeout)
662 atomic_long_t trans_timeout;
664 /* Subordinate device that the queue has been assigned to */
665 struct net_device *sb_dev;
666 #ifdef CONFIG_XDP_SOCKETS
667 struct xsk_buff_pool *pool;
672 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
675 * Time (in jiffies) of last Tx
677 unsigned long trans_start;
684 } ____cacheline_aligned_in_smp;
686 extern int sysctl_fb_tunnels_only_for_init_net;
687 extern int sysctl_devconf_inherit_init_net;
690 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
691 * == 1 : For initns only
694 static inline bool net_has_fallback_tunnels(const struct net *net)
696 #if IS_ENABLED(CONFIG_SYSCTL)
697 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
699 return !fb_tunnels_only_for_init_net ||
700 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
706 static inline int net_inherit_devconf(void)
708 #if IS_ENABLED(CONFIG_SYSCTL)
709 return READ_ONCE(sysctl_devconf_inherit_init_net);
715 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
717 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
724 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
726 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
733 * This structure holds an RPS map which can be of variable length. The
734 * map is an array of CPUs.
741 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
744 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
745 * tail pointer for that CPU's input queue at the time of last enqueue, and
746 * a hardware filter index.
748 struct rps_dev_flow {
751 unsigned int last_qtail;
753 #define RPS_NO_FILTER 0xffff
756 * The rps_dev_flow_table structure contains a table of flow mappings.
758 struct rps_dev_flow_table {
761 struct rps_dev_flow flows[];
763 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
764 ((_num) * sizeof(struct rps_dev_flow)))
767 * The rps_sock_flow_table contains mappings of flows to the last CPU
768 * on which they were processed by the application (set in recvmsg).
769 * Each entry is a 32bit value. Upper part is the high-order bits
770 * of flow hash, lower part is CPU number.
771 * rps_cpu_mask is used to partition the space, depending on number of
772 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
773 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
774 * meaning we use 32-6=26 bits for the hash.
776 struct rps_sock_flow_table {
779 u32 ents[] ____cacheline_aligned_in_smp;
781 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
783 #define RPS_NO_CPU 0xffff
785 extern u32 rps_cpu_mask;
786 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
788 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
792 unsigned int index = hash & table->mask;
793 u32 val = hash & ~rps_cpu_mask;
795 /* We only give a hint, preemption can change CPU under us */
796 val |= raw_smp_processor_id();
798 /* The following WRITE_ONCE() is paired with the READ_ONCE()
799 * here, and another one in get_rps_cpu().
801 if (READ_ONCE(table->ents[index]) != val)
802 WRITE_ONCE(table->ents[index], val);
806 #ifdef CONFIG_RFS_ACCEL
807 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
810 #endif /* CONFIG_RPS */
812 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
821 * This structure holds an XPS map which can be of variable length. The
822 * map is an array of queues.
826 unsigned int alloc_len;
830 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
831 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
832 - sizeof(struct xps_map)) / sizeof(u16))
835 * This structure holds all XPS maps for device. Maps are indexed by CPU.
837 * We keep track of the number of cpus/rxqs used when the struct is allocated,
838 * in nr_ids. This will help not accessing out-of-bound memory.
840 * We keep track of the number of traffic classes used when the struct is
841 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
842 * not crossing its upper bound, as the original dev->num_tc can be updated in
845 struct xps_dev_maps {
849 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
852 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
853 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
855 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
856 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
858 #endif /* CONFIG_XPS */
860 #define TC_MAX_QUEUE 16
861 #define TC_BITMASK 15
862 /* HW offloaded queuing disciplines txq count and offset maps */
863 struct netdev_tc_txq {
868 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
870 * This structure is to hold information about the device
871 * configured to run FCoE protocol stack.
873 struct netdev_fcoe_hbainfo {
874 char manufacturer[64];
875 char serial_number[64];
876 char hardware_version[64];
877 char driver_version[64];
878 char optionrom_version[64];
879 char firmware_version[64];
881 char model_description[256];
885 #define MAX_PHYS_ITEM_ID_LEN 32
887 /* This structure holds a unique identifier to identify some
888 * physical item (port for example) used by a netdevice.
890 struct netdev_phys_item_id {
891 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
892 unsigned char id_len;
895 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
896 struct netdev_phys_item_id *b)
898 return a->id_len == b->id_len &&
899 memcmp(a->id, b->id, a->id_len) == 0;
902 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
904 struct net_device *sb_dev);
906 enum net_device_path_type {
907 DEV_PATH_ETHERNET = 0,
915 struct net_device_path {
916 enum net_device_path_type type;
917 const struct net_device *dev;
926 DEV_PATH_BR_VLAN_KEEP,
927 DEV_PATH_BR_VLAN_TAG,
928 DEV_PATH_BR_VLAN_UNTAG,
929 DEV_PATH_BR_VLAN_UNTAG_HW,
948 #define NET_DEVICE_PATH_STACK_MAX 5
949 #define NET_DEVICE_PATH_VLAN_MAX 2
951 struct net_device_path_stack {
953 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
956 struct net_device_path_ctx {
957 const struct net_device *dev;
964 } vlan[NET_DEVICE_PATH_VLAN_MAX];
969 TC_SETUP_QDISC_MQPRIO,
972 TC_SETUP_CLSMATCHALL,
982 TC_SETUP_QDISC_TAPRIO,
991 /* These structures hold the attributes of bpf state that are being passed
992 * to the netdevice through the bpf op.
994 enum bpf_netdev_command {
995 /* Set or clear a bpf program used in the earliest stages of packet
996 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
997 * is responsible for calling bpf_prog_put on any old progs that are
998 * stored. In case of error, the callee need not release the new prog
999 * reference, but on success it takes ownership and must bpf_prog_put
1000 * when it is no longer used.
1004 /* BPF program for offload callbacks, invoked at program load time. */
1005 BPF_OFFLOAD_MAP_ALLOC,
1006 BPF_OFFLOAD_MAP_FREE,
1010 struct bpf_prog_offload_ops;
1011 struct netlink_ext_ack;
1013 struct xdp_dev_bulk_queue;
1014 struct bpf_xdp_link;
1023 struct bpf_xdp_entity {
1024 struct bpf_prog *prog;
1025 struct bpf_xdp_link *link;
1029 enum bpf_netdev_command command;
1031 /* XDP_SETUP_PROG */
1034 struct bpf_prog *prog;
1035 struct netlink_ext_ack *extack;
1037 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1039 struct bpf_offloaded_map *offmap;
1041 /* XDP_SETUP_XSK_POOL */
1043 struct xsk_buff_pool *pool;
1049 /* Flags for ndo_xsk_wakeup. */
1050 #define XDP_WAKEUP_RX (1 << 0)
1051 #define XDP_WAKEUP_TX (1 << 1)
1053 #ifdef CONFIG_XFRM_OFFLOAD
1054 struct xfrmdev_ops {
1055 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1056 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1057 void (*xdo_dev_state_free) (struct xfrm_state *x);
1058 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1059 struct xfrm_state *x);
1060 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1061 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1062 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1063 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1064 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1068 struct dev_ifalias {
1069 struct rcu_head rcuhead;
1076 struct netdev_net_notifier {
1077 struct list_head list;
1078 struct notifier_block *nb;
1082 * This structure defines the management hooks for network devices.
1083 * The following hooks can be defined; unless noted otherwise, they are
1084 * optional and can be filled with a null pointer.
1086 * int (*ndo_init)(struct net_device *dev);
1087 * This function is called once when a network device is registered.
1088 * The network device can use this for any late stage initialization
1089 * or semantic validation. It can fail with an error code which will
1090 * be propagated back to register_netdev.
1092 * void (*ndo_uninit)(struct net_device *dev);
1093 * This function is called when device is unregistered or when registration
1094 * fails. It is not called if init fails.
1096 * int (*ndo_open)(struct net_device *dev);
1097 * This function is called when a network device transitions to the up
1100 * int (*ndo_stop)(struct net_device *dev);
1101 * This function is called when a network device transitions to the down
1104 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1105 * struct net_device *dev);
1106 * Called when a packet needs to be transmitted.
1107 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1108 * the queue before that can happen; it's for obsolete devices and weird
1109 * corner cases, but the stack really does a non-trivial amount
1110 * of useless work if you return NETDEV_TX_BUSY.
1111 * Required; cannot be NULL.
1113 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1114 * struct net_device *dev
1115 * netdev_features_t features);
1116 * Called by core transmit path to determine if device is capable of
1117 * performing offload operations on a given packet. This is to give
1118 * the device an opportunity to implement any restrictions that cannot
1119 * be otherwise expressed by feature flags. The check is called with
1120 * the set of features that the stack has calculated and it returns
1121 * those the driver believes to be appropriate.
1123 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1124 * struct net_device *sb_dev);
1125 * Called to decide which queue to use when device supports multiple
1128 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1129 * This function is called to allow device receiver to make
1130 * changes to configuration when multicast or promiscuous is enabled.
1132 * void (*ndo_set_rx_mode)(struct net_device *dev);
1133 * This function is called device changes address list filtering.
1134 * If driver handles unicast address filtering, it should set
1135 * IFF_UNICAST_FLT in its priv_flags.
1137 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1138 * This function is called when the Media Access Control address
1139 * needs to be changed. If this interface is not defined, the
1140 * MAC address can not be changed.
1142 * int (*ndo_validate_addr)(struct net_device *dev);
1143 * Test if Media Access Control address is valid for the device.
1145 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1146 * Old-style ioctl entry point. This is used internally by the
1147 * appletalk and ieee802154 subsystems but is no longer called by
1148 * the device ioctl handler.
1150 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1151 * Used by the bonding driver for its device specific ioctls:
1152 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1153 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1155 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1156 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1157 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1159 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1160 * Used to set network devices bus interface parameters. This interface
1161 * is retained for legacy reasons; new devices should use the bus
1162 * interface (PCI) for low level management.
1164 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1165 * Called when a user wants to change the Maximum Transfer Unit
1168 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1169 * Callback used when the transmitter has not made any progress
1170 * for dev->watchdog ticks.
1172 * void (*ndo_get_stats64)(struct net_device *dev,
1173 * struct rtnl_link_stats64 *storage);
1174 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1175 * Called when a user wants to get the network device usage
1176 * statistics. Drivers must do one of the following:
1177 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1178 * rtnl_link_stats64 structure passed by the caller.
1179 * 2. Define @ndo_get_stats to update a net_device_stats structure
1180 * (which should normally be dev->stats) and return a pointer to
1181 * it. The structure may be changed asynchronously only if each
1182 * field is written atomically.
1183 * 3. Update dev->stats asynchronously and atomically, and define
1184 * neither operation.
1186 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1187 * Return true if this device supports offload stats of this attr_id.
1189 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1191 * Get statistics for offload operations by attr_id. Write it into the
1192 * attr_data pointer.
1194 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1195 * If device supports VLAN filtering this function is called when a
1196 * VLAN id is registered.
1198 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1199 * If device supports VLAN filtering this function is called when a
1200 * VLAN id is unregistered.
1202 * void (*ndo_poll_controller)(struct net_device *dev);
1204 * SR-IOV management functions.
1205 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1206 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1207 * u8 qos, __be16 proto);
1208 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1210 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1211 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1212 * int (*ndo_get_vf_config)(struct net_device *dev,
1213 * int vf, struct ifla_vf_info *ivf);
1214 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1215 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1216 * struct nlattr *port[]);
1218 * Enable or disable the VF ability to query its RSS Redirection Table and
1219 * Hash Key. This is needed since on some devices VF share this information
1220 * with PF and querying it may introduce a theoretical security risk.
1221 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1222 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1223 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1225 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1226 * This is always called from the stack with the rtnl lock held and netif
1227 * tx queues stopped. This allows the netdevice to perform queue
1228 * management safely.
1230 * Fiber Channel over Ethernet (FCoE) offload functions.
1231 * int (*ndo_fcoe_enable)(struct net_device *dev);
1232 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1233 * so the underlying device can perform whatever needed configuration or
1234 * initialization to support acceleration of FCoE traffic.
1236 * int (*ndo_fcoe_disable)(struct net_device *dev);
1237 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1238 * so the underlying device can perform whatever needed clean-ups to
1239 * stop supporting acceleration of FCoE traffic.
1241 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1242 * struct scatterlist *sgl, unsigned int sgc);
1243 * Called when the FCoE Initiator wants to initialize an I/O that
1244 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1245 * perform necessary setup and returns 1 to indicate the device is set up
1246 * successfully to perform DDP on this I/O, otherwise this returns 0.
1248 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1249 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1250 * indicated by the FC exchange id 'xid', so the underlying device can
1251 * clean up and reuse resources for later DDP requests.
1253 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1254 * struct scatterlist *sgl, unsigned int sgc);
1255 * Called when the FCoE Target wants to initialize an I/O that
1256 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1257 * perform necessary setup and returns 1 to indicate the device is set up
1258 * successfully to perform DDP on this I/O, otherwise this returns 0.
1260 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1261 * struct netdev_fcoe_hbainfo *hbainfo);
1262 * Called when the FCoE Protocol stack wants information on the underlying
1263 * device. This information is utilized by the FCoE protocol stack to
1264 * register attributes with Fiber Channel management service as per the
1265 * FC-GS Fabric Device Management Information(FDMI) specification.
1267 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1268 * Called when the underlying device wants to override default World Wide
1269 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1270 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1271 * protocol stack to use.
1274 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1275 * u16 rxq_index, u32 flow_id);
1276 * Set hardware filter for RFS. rxq_index is the target queue index;
1277 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1278 * Return the filter ID on success, or a negative error code.
1280 * Slave management functions (for bridge, bonding, etc).
1281 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1282 * Called to make another netdev an underling.
1284 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1285 * Called to release previously enslaved netdev.
1287 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1288 * struct sk_buff *skb,
1290 * Get the xmit slave of master device. If all_slaves is true, function
1291 * assume all the slaves can transmit.
1293 * Feature/offload setting functions.
1294 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1295 * netdev_features_t features);
1296 * Adjusts the requested feature flags according to device-specific
1297 * constraints, and returns the resulting flags. Must not modify
1300 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1301 * Called to update device configuration to new features. Passed
1302 * feature set might be less than what was returned by ndo_fix_features()).
1303 * Must return >0 or -errno if it changed dev->features itself.
1305 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1306 * struct net_device *dev,
1307 * const unsigned char *addr, u16 vid, u16 flags,
1308 * struct netlink_ext_ack *extack);
1309 * Adds an FDB entry to dev for addr.
1310 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1311 * struct net_device *dev,
1312 * const unsigned char *addr, u16 vid)
1313 * Deletes the FDB entry from dev coresponding to addr.
1314 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1315 * struct netlink_ext_ack *extack);
1316 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1317 * struct net_device *dev, struct net_device *filter_dev,
1319 * Used to add FDB entries to dump requests. Implementers should add
1320 * entries to skb and update idx with the number of entries.
1322 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1323 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1324 * Adds an MDB entry to dev.
1325 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1326 * struct netlink_ext_ack *extack);
1327 * Deletes the MDB entry from dev.
1328 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1329 * struct netlink_callback *cb);
1330 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1331 * callback is used by core rtnetlink code.
1333 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1334 * u16 flags, struct netlink_ext_ack *extack)
1335 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1336 * struct net_device *dev, u32 filter_mask,
1338 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1341 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1342 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1343 * which do not represent real hardware may define this to allow their
1344 * userspace components to manage their virtual carrier state. Devices
1345 * that determine carrier state from physical hardware properties (eg
1346 * network cables) or protocol-dependent mechanisms (eg
1347 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1349 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1350 * struct netdev_phys_item_id *ppid);
1351 * Called to get ID of physical port of this device. If driver does
1352 * not implement this, it is assumed that the hw is not able to have
1353 * multiple net devices on single physical port.
1355 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1356 * struct netdev_phys_item_id *ppid)
1357 * Called to get the parent ID of the physical port of this device.
1359 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1360 * struct net_device *dev)
1361 * Called by upper layer devices to accelerate switching or other
1362 * station functionality into hardware. 'pdev is the lowerdev
1363 * to use for the offload and 'dev' is the net device that will
1364 * back the offload. Returns a pointer to the private structure
1365 * the upper layer will maintain.
1366 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1367 * Called by upper layer device to delete the station created
1368 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1369 * the station and priv is the structure returned by the add
1371 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1372 * int queue_index, u32 maxrate);
1373 * Called when a user wants to set a max-rate limitation of specific
1375 * int (*ndo_get_iflink)(const struct net_device *dev);
1376 * Called to get the iflink value of this device.
1377 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1378 * This function is used to get egress tunnel information for given skb.
1379 * This is useful for retrieving outer tunnel header parameters while
1381 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1382 * This function is used to specify the headroom that the skb must
1383 * consider when allocation skb during packet reception. Setting
1384 * appropriate rx headroom value allows avoiding skb head copy on
1385 * forward. Setting a negative value resets the rx headroom to the
1387 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1388 * This function is used to set or query state related to XDP on the
1389 * netdevice and manage BPF offload. See definition of
1390 * enum bpf_netdev_command for details.
1391 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1393 * This function is used to submit @n XDP packets for transmit on a
1394 * netdevice. Returns number of frames successfully transmitted, frames
1395 * that got dropped are freed/returned via xdp_return_frame().
1396 * Returns negative number, means general error invoking ndo, meaning
1397 * no frames were xmit'ed and core-caller will free all frames.
1398 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1399 * struct xdp_buff *xdp);
1400 * Get the xmit slave of master device based on the xdp_buff.
1401 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1402 * This function is used to wake up the softirq, ksoftirqd or kthread
1403 * responsible for sending and/or receiving packets on a specific
1404 * queue id bound to an AF_XDP socket. The flags field specifies if
1405 * only RX, only Tx, or both should be woken up using the flags
1406 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1407 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1409 * Add, change, delete or get information on an IPv4 tunnel.
1410 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1411 * If a device is paired with a peer device, return the peer instance.
1412 * The caller must be under RCU read context.
1413 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1414 * Get the forwarding path to reach the real device from the HW destination address
1415 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1416 * const struct skb_shared_hwtstamps *hwtstamps,
1418 * Get hardware timestamp based on normal/adjustable time or free running
1419 * cycle counter. This function is required if physical clock supports a
1420 * free running cycle counter.
1422 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1423 * struct kernel_hwtstamp_config *kernel_config);
1424 * Get the currently configured hardware timestamping parameters for the
1427 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1428 * struct kernel_hwtstamp_config *kernel_config,
1429 * struct netlink_ext_ack *extack);
1430 * Change the hardware timestamping parameters for NIC device.
1432 struct net_device_ops {
1433 int (*ndo_init)(struct net_device *dev);
1434 void (*ndo_uninit)(struct net_device *dev);
1435 int (*ndo_open)(struct net_device *dev);
1436 int (*ndo_stop)(struct net_device *dev);
1437 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1438 struct net_device *dev);
1439 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1440 struct net_device *dev,
1441 netdev_features_t features);
1442 u16 (*ndo_select_queue)(struct net_device *dev,
1443 struct sk_buff *skb,
1444 struct net_device *sb_dev);
1445 void (*ndo_change_rx_flags)(struct net_device *dev,
1447 void (*ndo_set_rx_mode)(struct net_device *dev);
1448 int (*ndo_set_mac_address)(struct net_device *dev,
1450 int (*ndo_validate_addr)(struct net_device *dev);
1451 int (*ndo_do_ioctl)(struct net_device *dev,
1452 struct ifreq *ifr, int cmd);
1453 int (*ndo_eth_ioctl)(struct net_device *dev,
1454 struct ifreq *ifr, int cmd);
1455 int (*ndo_siocbond)(struct net_device *dev,
1456 struct ifreq *ifr, int cmd);
1457 int (*ndo_siocwandev)(struct net_device *dev,
1458 struct if_settings *ifs);
1459 int (*ndo_siocdevprivate)(struct net_device *dev,
1461 void __user *data, int cmd);
1462 int (*ndo_set_config)(struct net_device *dev,
1464 int (*ndo_change_mtu)(struct net_device *dev,
1466 int (*ndo_neigh_setup)(struct net_device *dev,
1467 struct neigh_parms *);
1468 void (*ndo_tx_timeout) (struct net_device *dev,
1469 unsigned int txqueue);
1471 void (*ndo_get_stats64)(struct net_device *dev,
1472 struct rtnl_link_stats64 *storage);
1473 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1474 int (*ndo_get_offload_stats)(int attr_id,
1475 const struct net_device *dev,
1477 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1479 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1480 __be16 proto, u16 vid);
1481 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1482 __be16 proto, u16 vid);
1483 #ifdef CONFIG_NET_POLL_CONTROLLER
1484 void (*ndo_poll_controller)(struct net_device *dev);
1485 int (*ndo_netpoll_setup)(struct net_device *dev,
1486 struct netpoll_info *info);
1487 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1489 int (*ndo_set_vf_mac)(struct net_device *dev,
1490 int queue, u8 *mac);
1491 int (*ndo_set_vf_vlan)(struct net_device *dev,
1492 int queue, u16 vlan,
1493 u8 qos, __be16 proto);
1494 int (*ndo_set_vf_rate)(struct net_device *dev,
1495 int vf, int min_tx_rate,
1497 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1498 int vf, bool setting);
1499 int (*ndo_set_vf_trust)(struct net_device *dev,
1500 int vf, bool setting);
1501 int (*ndo_get_vf_config)(struct net_device *dev,
1503 struct ifla_vf_info *ivf);
1504 int (*ndo_set_vf_link_state)(struct net_device *dev,
1505 int vf, int link_state);
1506 int (*ndo_get_vf_stats)(struct net_device *dev,
1508 struct ifla_vf_stats
1510 int (*ndo_set_vf_port)(struct net_device *dev,
1512 struct nlattr *port[]);
1513 int (*ndo_get_vf_port)(struct net_device *dev,
1514 int vf, struct sk_buff *skb);
1515 int (*ndo_get_vf_guid)(struct net_device *dev,
1517 struct ifla_vf_guid *node_guid,
1518 struct ifla_vf_guid *port_guid);
1519 int (*ndo_set_vf_guid)(struct net_device *dev,
1522 int (*ndo_set_vf_rss_query_en)(
1523 struct net_device *dev,
1524 int vf, bool setting);
1525 int (*ndo_setup_tc)(struct net_device *dev,
1526 enum tc_setup_type type,
1528 #if IS_ENABLED(CONFIG_FCOE)
1529 int (*ndo_fcoe_enable)(struct net_device *dev);
1530 int (*ndo_fcoe_disable)(struct net_device *dev);
1531 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1533 struct scatterlist *sgl,
1535 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1537 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1539 struct scatterlist *sgl,
1541 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1542 struct netdev_fcoe_hbainfo *hbainfo);
1545 #if IS_ENABLED(CONFIG_LIBFCOE)
1546 #define NETDEV_FCOE_WWNN 0
1547 #define NETDEV_FCOE_WWPN 1
1548 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1549 u64 *wwn, int type);
1552 #ifdef CONFIG_RFS_ACCEL
1553 int (*ndo_rx_flow_steer)(struct net_device *dev,
1554 const struct sk_buff *skb,
1558 int (*ndo_add_slave)(struct net_device *dev,
1559 struct net_device *slave_dev,
1560 struct netlink_ext_ack *extack);
1561 int (*ndo_del_slave)(struct net_device *dev,
1562 struct net_device *slave_dev);
1563 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1564 struct sk_buff *skb,
1566 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1568 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1569 netdev_features_t features);
1570 int (*ndo_set_features)(struct net_device *dev,
1571 netdev_features_t features);
1572 int (*ndo_neigh_construct)(struct net_device *dev,
1573 struct neighbour *n);
1574 void (*ndo_neigh_destroy)(struct net_device *dev,
1575 struct neighbour *n);
1577 int (*ndo_fdb_add)(struct ndmsg *ndm,
1578 struct nlattr *tb[],
1579 struct net_device *dev,
1580 const unsigned char *addr,
1583 struct netlink_ext_ack *extack);
1584 int (*ndo_fdb_del)(struct ndmsg *ndm,
1585 struct nlattr *tb[],
1586 struct net_device *dev,
1587 const unsigned char *addr,
1588 u16 vid, struct netlink_ext_ack *extack);
1589 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1590 struct net_device *dev,
1591 struct netlink_ext_ack *extack);
1592 int (*ndo_fdb_dump)(struct sk_buff *skb,
1593 struct netlink_callback *cb,
1594 struct net_device *dev,
1595 struct net_device *filter_dev,
1597 int (*ndo_fdb_get)(struct sk_buff *skb,
1598 struct nlattr *tb[],
1599 struct net_device *dev,
1600 const unsigned char *addr,
1601 u16 vid, u32 portid, u32 seq,
1602 struct netlink_ext_ack *extack);
1603 int (*ndo_mdb_add)(struct net_device *dev,
1604 struct nlattr *tb[],
1606 struct netlink_ext_ack *extack);
1607 int (*ndo_mdb_del)(struct net_device *dev,
1608 struct nlattr *tb[],
1609 struct netlink_ext_ack *extack);
1610 int (*ndo_mdb_dump)(struct net_device *dev,
1611 struct sk_buff *skb,
1612 struct netlink_callback *cb);
1613 int (*ndo_mdb_get)(struct net_device *dev,
1614 struct nlattr *tb[], u32 portid,
1616 struct netlink_ext_ack *extack);
1617 int (*ndo_bridge_setlink)(struct net_device *dev,
1618 struct nlmsghdr *nlh,
1620 struct netlink_ext_ack *extack);
1621 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1623 struct net_device *dev,
1626 int (*ndo_bridge_dellink)(struct net_device *dev,
1627 struct nlmsghdr *nlh,
1629 int (*ndo_change_carrier)(struct net_device *dev,
1631 int (*ndo_get_phys_port_id)(struct net_device *dev,
1632 struct netdev_phys_item_id *ppid);
1633 int (*ndo_get_port_parent_id)(struct net_device *dev,
1634 struct netdev_phys_item_id *ppid);
1635 int (*ndo_get_phys_port_name)(struct net_device *dev,
1636 char *name, size_t len);
1637 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1638 struct net_device *dev);
1639 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1642 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1645 int (*ndo_get_iflink)(const struct net_device *dev);
1646 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1647 struct sk_buff *skb);
1648 void (*ndo_set_rx_headroom)(struct net_device *dev,
1649 int needed_headroom);
1650 int (*ndo_bpf)(struct net_device *dev,
1651 struct netdev_bpf *bpf);
1652 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1653 struct xdp_frame **xdp,
1655 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1656 struct xdp_buff *xdp);
1657 int (*ndo_xsk_wakeup)(struct net_device *dev,
1658 u32 queue_id, u32 flags);
1659 int (*ndo_tunnel_ctl)(struct net_device *dev,
1660 struct ip_tunnel_parm *p, int cmd);
1661 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1662 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1663 struct net_device_path *path);
1664 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1665 const struct skb_shared_hwtstamps *hwtstamps,
1667 int (*ndo_hwtstamp_get)(struct net_device *dev,
1668 struct kernel_hwtstamp_config *kernel_config);
1669 int (*ndo_hwtstamp_set)(struct net_device *dev,
1670 struct kernel_hwtstamp_config *kernel_config,
1671 struct netlink_ext_ack *extack);
1675 * enum netdev_priv_flags - &struct net_device priv_flags
1677 * These are the &struct net_device, they are only set internally
1678 * by drivers and used in the kernel. These flags are invisible to
1679 * userspace; this means that the order of these flags can change
1680 * during any kernel release.
1682 * You should have a pretty good reason to be extending these flags.
1684 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1685 * @IFF_EBRIDGE: Ethernet bridging device
1686 * @IFF_BONDING: bonding master or slave
1687 * @IFF_ISATAP: ISATAP interface (RFC4214)
1688 * @IFF_WAN_HDLC: WAN HDLC device
1689 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1691 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1692 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1693 * @IFF_MACVLAN_PORT: device used as macvlan port
1694 * @IFF_BRIDGE_PORT: device used as bridge port
1695 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1696 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1697 * @IFF_UNICAST_FLT: Supports unicast filtering
1698 * @IFF_TEAM_PORT: device used as team port
1699 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1700 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1701 * change when it's running
1702 * @IFF_MACVLAN: Macvlan device
1703 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1704 * underlying stacked devices
1705 * @IFF_L3MDEV_MASTER: device is an L3 master device
1706 * @IFF_NO_QUEUE: device can run without qdisc attached
1707 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1708 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1709 * @IFF_TEAM: device is a team device
1710 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1711 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1712 * entity (i.e. the master device for bridged veth)
1713 * @IFF_MACSEC: device is a MACsec device
1714 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1715 * @IFF_FAILOVER: device is a failover master device
1716 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1717 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1718 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1719 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1720 * skb_headlen(skb) == 0 (data starts from frag0)
1721 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1722 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1723 * ndo_hwtstamp_set() for all timestamp requests regardless of source,
1724 * even if those aren't HWTSTAMP_SOURCE_NETDEV.
1726 enum netdev_priv_flags {
1727 IFF_802_1Q_VLAN = 1<<0,
1731 IFF_WAN_HDLC = 1<<4,
1732 IFF_XMIT_DST_RELEASE = 1<<5,
1733 IFF_DONT_BRIDGE = 1<<6,
1734 IFF_DISABLE_NETPOLL = 1<<7,
1735 IFF_MACVLAN_PORT = 1<<8,
1736 IFF_BRIDGE_PORT = 1<<9,
1737 IFF_OVS_DATAPATH = 1<<10,
1738 IFF_TX_SKB_SHARING = 1<<11,
1739 IFF_UNICAST_FLT = 1<<12,
1740 IFF_TEAM_PORT = 1<<13,
1741 IFF_SUPP_NOFCS = 1<<14,
1742 IFF_LIVE_ADDR_CHANGE = 1<<15,
1743 IFF_MACVLAN = 1<<16,
1744 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1745 IFF_L3MDEV_MASTER = 1<<18,
1746 IFF_NO_QUEUE = 1<<19,
1747 IFF_OPENVSWITCH = 1<<20,
1748 IFF_L3MDEV_SLAVE = 1<<21,
1750 IFF_RXFH_CONFIGURED = 1<<23,
1751 IFF_PHONY_HEADROOM = 1<<24,
1753 IFF_NO_RX_HANDLER = 1<<26,
1754 IFF_FAILOVER = 1<<27,
1755 IFF_FAILOVER_SLAVE = 1<<28,
1756 IFF_L3MDEV_RX_HANDLER = 1<<29,
1757 IFF_NO_ADDRCONF = BIT_ULL(30),
1758 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1759 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1760 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33),
1763 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1764 #define IFF_EBRIDGE IFF_EBRIDGE
1765 #define IFF_BONDING IFF_BONDING
1766 #define IFF_ISATAP IFF_ISATAP
1767 #define IFF_WAN_HDLC IFF_WAN_HDLC
1768 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1769 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1770 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1771 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1772 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1773 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1774 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1775 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1776 #define IFF_TEAM_PORT IFF_TEAM_PORT
1777 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1778 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1779 #define IFF_MACVLAN IFF_MACVLAN
1780 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1781 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1782 #define IFF_NO_QUEUE IFF_NO_QUEUE
1783 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1784 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1785 #define IFF_TEAM IFF_TEAM
1786 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1787 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1788 #define IFF_MACSEC IFF_MACSEC
1789 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1790 #define IFF_FAILOVER IFF_FAILOVER
1791 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1792 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1793 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1795 /* Specifies the type of the struct net_device::ml_priv pointer */
1796 enum netdev_ml_priv_type {
1802 * struct net_device - The DEVICE structure.
1804 * Actually, this whole structure is a big mistake. It mixes I/O
1805 * data with strictly "high-level" data, and it has to know about
1806 * almost every data structure used in the INET module.
1808 * @name: This is the first field of the "visible" part of this structure
1809 * (i.e. as seen by users in the "Space.c" file). It is the name
1812 * @name_node: Name hashlist node
1813 * @ifalias: SNMP alias
1814 * @mem_end: Shared memory end
1815 * @mem_start: Shared memory start
1816 * @base_addr: Device I/O address
1817 * @irq: Device IRQ number
1819 * @state: Generic network queuing layer state, see netdev_state_t
1820 * @dev_list: The global list of network devices
1821 * @napi_list: List entry used for polling NAPI devices
1822 * @unreg_list: List entry when we are unregistering the
1823 * device; see the function unregister_netdev
1824 * @close_list: List entry used when we are closing the device
1825 * @ptype_all: Device-specific packet handlers for all protocols
1826 * @ptype_specific: Device-specific, protocol-specific packet handlers
1828 * @adj_list: Directly linked devices, like slaves for bonding
1829 * @features: Currently active device features
1830 * @hw_features: User-changeable features
1832 * @wanted_features: User-requested features
1833 * @vlan_features: Mask of features inheritable by VLAN devices
1835 * @hw_enc_features: Mask of features inherited by encapsulating devices
1836 * This field indicates what encapsulation
1837 * offloads the hardware is capable of doing,
1838 * and drivers will need to set them appropriately.
1840 * @mpls_features: Mask of features inheritable by MPLS
1841 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1843 * @ifindex: interface index
1844 * @group: The group the device belongs to
1846 * @stats: Statistics struct, which was left as a legacy, use
1847 * rtnl_link_stats64 instead
1849 * @core_stats: core networking counters,
1850 * do not use this in drivers
1851 * @carrier_up_count: Number of times the carrier has been up
1852 * @carrier_down_count: Number of times the carrier has been down
1854 * @wireless_handlers: List of functions to handle Wireless Extensions,
1856 * see <net/iw_handler.h> for details.
1857 * @wireless_data: Instance data managed by the core of wireless extensions
1859 * @netdev_ops: Includes several pointers to callbacks,
1860 * if one wants to override the ndo_*() functions
1861 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1862 * @ethtool_ops: Management operations
1863 * @l3mdev_ops: Layer 3 master device operations
1864 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1865 * discovery handling. Necessary for e.g. 6LoWPAN.
1866 * @xfrmdev_ops: Transformation offload operations
1867 * @tlsdev_ops: Transport Layer Security offload operations
1868 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1869 * of Layer 2 headers.
1871 * @flags: Interface flags (a la BSD)
1872 * @xdp_features: XDP capability supported by the device
1873 * @priv_flags: Like 'flags' but invisible to userspace,
1874 * see if.h for the definitions
1875 * @gflags: Global flags ( kept as legacy )
1876 * @padded: How much padding added by alloc_netdev()
1877 * @operstate: RFC2863 operstate
1878 * @link_mode: Mapping policy to operstate
1879 * @if_port: Selectable AUI, TP, ...
1881 * @mtu: Interface MTU value
1882 * @min_mtu: Interface Minimum MTU value
1883 * @max_mtu: Interface Maximum MTU value
1884 * @type: Interface hardware type
1885 * @hard_header_len: Maximum hardware header length.
1886 * @min_header_len: Minimum hardware header length
1888 * @needed_headroom: Extra headroom the hardware may need, but not in all
1889 * cases can this be guaranteed
1890 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1891 * cases can this be guaranteed. Some cases also use
1892 * LL_MAX_HEADER instead to allocate the skb
1894 * interface address info:
1896 * @perm_addr: Permanent hw address
1897 * @addr_assign_type: Hw address assignment type
1898 * @addr_len: Hardware address length
1899 * @upper_level: Maximum depth level of upper devices.
1900 * @lower_level: Maximum depth level of lower devices.
1901 * @neigh_priv_len: Used in neigh_alloc()
1902 * @dev_id: Used to differentiate devices that share
1903 * the same link layer address
1904 * @dev_port: Used to differentiate devices that share
1906 * @addr_list_lock: XXX: need comments on this one
1907 * @name_assign_type: network interface name assignment type
1908 * @uc_promisc: Counter that indicates promiscuous mode
1909 * has been enabled due to the need to listen to
1910 * additional unicast addresses in a device that
1911 * does not implement ndo_set_rx_mode()
1912 * @uc: unicast mac addresses
1913 * @mc: multicast mac addresses
1914 * @dev_addrs: list of device hw addresses
1915 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1916 * @promiscuity: Number of times the NIC is told to work in
1917 * promiscuous mode; if it becomes 0 the NIC will
1918 * exit promiscuous mode
1919 * @allmulti: Counter, enables or disables allmulticast mode
1921 * @vlan_info: VLAN info
1922 * @dsa_ptr: dsa specific data
1923 * @tipc_ptr: TIPC specific data
1924 * @atalk_ptr: AppleTalk link
1925 * @ip_ptr: IPv4 specific data
1926 * @ip6_ptr: IPv6 specific data
1927 * @ax25_ptr: AX.25 specific data
1928 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1929 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1931 * @mpls_ptr: mpls_dev struct pointer
1932 * @mctp_ptr: MCTP specific data
1934 * @dev_addr: Hw address (before bcast,
1935 * because most packets are unicast)
1937 * @_rx: Array of RX queues
1938 * @num_rx_queues: Number of RX queues
1939 * allocated at register_netdev() time
1940 * @real_num_rx_queues: Number of RX queues currently active in device
1941 * @xdp_prog: XDP sockets filter program pointer
1942 * @gro_flush_timeout: timeout for GRO layer in NAPI
1943 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1944 * allow to avoid NIC hard IRQ, on busy queues.
1946 * @rx_handler: handler for received packets
1947 * @rx_handler_data: XXX: need comments on this one
1948 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1949 * @ingress_queue: XXX: need comments on this one
1950 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1951 * @broadcast: hw bcast address
1953 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1954 * indexed by RX queue number. Assigned by driver.
1955 * This must only be set if the ndo_rx_flow_steer
1956 * operation is defined
1957 * @index_hlist: Device index hash chain
1959 * @_tx: Array of TX queues
1960 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1961 * @real_num_tx_queues: Number of TX queues currently active in device
1962 * @qdisc: Root qdisc from userspace point of view
1963 * @tx_queue_len: Max frames per queue allowed
1964 * @tx_global_lock: XXX: need comments on this one
1965 * @xdp_bulkq: XDP device bulk queue
1966 * @xps_maps: all CPUs/RXQs maps for XPS device
1968 * @xps_maps: XXX: need comments on this one
1969 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1970 * @nf_hooks_egress: netfilter hooks executed for egress packets
1971 * @qdisc_hash: qdisc hash table
1972 * @watchdog_timeo: Represents the timeout that is used by
1973 * the watchdog (see dev_watchdog())
1974 * @watchdog_timer: List of timers
1976 * @proto_down_reason: reason a netdev interface is held down
1977 * @pcpu_refcnt: Number of references to this device
1978 * @dev_refcnt: Number of references to this device
1979 * @refcnt_tracker: Tracker directory for tracked references to this device
1980 * @todo_list: Delayed register/unregister
1981 * @link_watch_list: XXX: need comments on this one
1983 * @reg_state: Register/unregister state machine
1984 * @dismantle: Device is going to be freed
1985 * @rtnl_link_state: This enum represents the phases of creating
1988 * @needs_free_netdev: Should unregister perform free_netdev?
1989 * @priv_destructor: Called from unregister
1990 * @npinfo: XXX: need comments on this one
1991 * @nd_net: Network namespace this network device is inside
1993 * @ml_priv: Mid-layer private
1994 * @ml_priv_type: Mid-layer private type
1995 * @lstats: Loopback statistics
1996 * @tstats: Tunnel statistics
1997 * @dstats: Dummy statistics
1998 * @vstats: Virtual ethernet statistics
2003 * @dm_private: Drop monitor private
2005 * @dev: Class/net/name entry
2006 * @sysfs_groups: Space for optional device, statistics and wireless
2009 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
2010 * @rtnl_link_ops: Rtnl_link_ops
2012 * @gso_max_size: Maximum size of generic segmentation offload
2013 * @tso_max_size: Device (as in HW) limit on the max TSO request size
2014 * @gso_max_segs: Maximum number of segments that can be passed to the
2016 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
2017 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
2020 * @dcbnl_ops: Data Center Bridging netlink ops
2021 * @num_tc: Number of traffic classes in the net device
2022 * @tc_to_txq: XXX: need comments on this one
2023 * @prio_tc_map: XXX: need comments on this one
2025 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
2027 * @priomap: XXX: need comments on this one
2028 * @phydev: Physical device may attach itself
2029 * for hardware timestamping
2030 * @sfp_bus: attached &struct sfp_bus structure.
2032 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2034 * @proto_down: protocol port state information can be sent to the
2035 * switch driver and used to set the phys state of the
2038 * @wol_enabled: Wake-on-LAN is enabled
2040 * @threaded: napi threaded mode is enabled
2042 * @net_notifier_list: List of per-net netdev notifier block
2043 * that follow this device when it is moved
2044 * to another network namespace.
2046 * @macsec_ops: MACsec offloading ops
2048 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2049 * offload capabilities of the device
2050 * @udp_tunnel_nic: UDP tunnel offload state
2051 * @xdp_state: stores info on attached XDP BPF programs
2053 * @nested_level: Used as a parameter of spin_lock_nested() of
2054 * dev->addr_list_lock.
2055 * @unlink_list: As netif_addr_lock() can be called recursively,
2056 * keep a list of interfaces to be deleted.
2057 * @gro_max_size: Maximum size of aggregated packet in generic
2058 * receive offload (GRO)
2059 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2060 * receive offload (GRO), for IPv4.
2061 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2064 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2065 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2066 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2067 * @dev_registered_tracker: tracker for reference held while
2069 * @offload_xstats_l3: L3 HW stats for this netdevice.
2071 * @devlink_port: Pointer to related devlink port structure.
2072 * Assigned by a driver before netdev registration using
2073 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2074 * during the time netdevice is registered.
2076 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2077 * where the clock is recovered.
2078 * @ts_layer: Tracks which network device
2079 * performs packet time stamping.
2081 * FIXME: cleanup struct net_device such that network protocol info
2086 char name[IFNAMSIZ];
2087 struct netdev_name_node *name_node;
2088 struct dev_ifalias __rcu *ifalias;
2090 * I/O specific fields
2091 * FIXME: Merge these and struct ifmap into one
2093 unsigned long mem_end;
2094 unsigned long mem_start;
2095 unsigned long base_addr;
2098 * Some hardware also needs these fields (state,dev_list,
2099 * napi_list,unreg_list,close_list) but they are not
2100 * part of the usual set specified in Space.c.
2103 unsigned long state;
2105 struct list_head dev_list;
2106 struct list_head napi_list;
2107 struct list_head unreg_list;
2108 struct list_head close_list;
2109 struct list_head ptype_all;
2110 struct list_head ptype_specific;
2113 struct list_head upper;
2114 struct list_head lower;
2117 /* Read-mostly cache-line for fast-path access */
2119 xdp_features_t xdp_features;
2120 unsigned long long priv_flags;
2121 const struct net_device_ops *netdev_ops;
2122 const struct xdp_metadata_ops *xdp_metadata_ops;
2124 unsigned short gflags;
2125 unsigned short hard_header_len;
2127 /* Note : dev->mtu is often read without holding a lock.
2128 * Writers usually hold RTNL.
2129 * It is recommended to use READ_ONCE() to annotate the reads,
2130 * and to use WRITE_ONCE() to annotate the writes.
2133 unsigned short needed_headroom;
2134 unsigned short needed_tailroom;
2136 netdev_features_t features;
2137 netdev_features_t hw_features;
2138 netdev_features_t wanted_features;
2139 netdev_features_t vlan_features;
2140 netdev_features_t hw_enc_features;
2141 netdev_features_t mpls_features;
2142 netdev_features_t gso_partial_features;
2144 unsigned int min_mtu;
2145 unsigned int max_mtu;
2146 unsigned short type;
2147 unsigned char min_header_len;
2148 unsigned char name_assign_type;
2152 struct net_device_stats stats; /* not used by modern drivers */
2154 struct net_device_core_stats __percpu *core_stats;
2156 /* Stats to monitor link on/off, flapping */
2157 atomic_t carrier_up_count;
2158 atomic_t carrier_down_count;
2160 #ifdef CONFIG_WIRELESS_EXT
2161 const struct iw_handler_def *wireless_handlers;
2162 struct iw_public_data *wireless_data;
2164 const struct ethtool_ops *ethtool_ops;
2165 #ifdef CONFIG_NET_L3_MASTER_DEV
2166 const struct l3mdev_ops *l3mdev_ops;
2168 #if IS_ENABLED(CONFIG_IPV6)
2169 const struct ndisc_ops *ndisc_ops;
2172 #ifdef CONFIG_XFRM_OFFLOAD
2173 const struct xfrmdev_ops *xfrmdev_ops;
2176 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2177 const struct tlsdev_ops *tlsdev_ops;
2180 const struct header_ops *header_ops;
2182 unsigned char operstate;
2183 unsigned char link_mode;
2185 unsigned char if_port;
2188 /* Interface address info. */
2189 unsigned char perm_addr[MAX_ADDR_LEN];
2190 unsigned char addr_assign_type;
2191 unsigned char addr_len;
2192 unsigned char upper_level;
2193 unsigned char lower_level;
2195 unsigned short neigh_priv_len;
2196 unsigned short dev_id;
2197 unsigned short dev_port;
2198 unsigned short padded;
2200 spinlock_t addr_list_lock;
2203 struct netdev_hw_addr_list uc;
2204 struct netdev_hw_addr_list mc;
2205 struct netdev_hw_addr_list dev_addrs;
2208 struct kset *queues_kset;
2210 #ifdef CONFIG_LOCKDEP
2211 struct list_head unlink_list;
2213 unsigned int promiscuity;
2214 unsigned int allmulti;
2216 #ifdef CONFIG_LOCKDEP
2217 unsigned char nested_level;
2221 /* Protocol-specific pointers */
2223 struct in_device __rcu *ip_ptr;
2224 struct inet6_dev __rcu *ip6_ptr;
2225 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2226 struct vlan_info __rcu *vlan_info;
2228 #if IS_ENABLED(CONFIG_NET_DSA)
2229 struct dsa_port *dsa_ptr;
2231 #if IS_ENABLED(CONFIG_TIPC)
2232 struct tipc_bearer __rcu *tipc_ptr;
2234 #if IS_ENABLED(CONFIG_ATALK)
2237 #if IS_ENABLED(CONFIG_AX25)
2240 #if IS_ENABLED(CONFIG_CFG80211)
2241 struct wireless_dev *ieee80211_ptr;
2243 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2244 struct wpan_dev *ieee802154_ptr;
2246 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2247 struct mpls_dev __rcu *mpls_ptr;
2249 #if IS_ENABLED(CONFIG_MCTP)
2250 struct mctp_dev __rcu *mctp_ptr;
2254 * Cache lines mostly used on receive path (including eth_type_trans())
2256 /* Interface address info used in eth_type_trans() */
2257 const unsigned char *dev_addr;
2259 struct netdev_rx_queue *_rx;
2260 unsigned int num_rx_queues;
2261 unsigned int real_num_rx_queues;
2263 struct bpf_prog __rcu *xdp_prog;
2264 unsigned long gro_flush_timeout;
2265 int napi_defer_hard_irqs;
2266 #define GRO_LEGACY_MAX_SIZE 65536u
2267 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2268 * and shinfo->gso_segs is a 16bit field.
2270 #define GRO_MAX_SIZE (8 * 65535u)
2271 unsigned int gro_max_size;
2272 unsigned int gro_ipv4_max_size;
2273 unsigned int xdp_zc_max_segs;
2274 rx_handler_func_t __rcu *rx_handler;
2275 void __rcu *rx_handler_data;
2276 #ifdef CONFIG_NET_XGRESS
2277 struct bpf_mprog_entry __rcu *tcx_ingress;
2279 struct netdev_queue __rcu *ingress_queue;
2280 #ifdef CONFIG_NETFILTER_INGRESS
2281 struct nf_hook_entries __rcu *nf_hooks_ingress;
2284 unsigned char broadcast[MAX_ADDR_LEN];
2285 #ifdef CONFIG_RFS_ACCEL
2286 struct cpu_rmap *rx_cpu_rmap;
2288 struct hlist_node index_hlist;
2291 * Cache lines mostly used on transmit path
2293 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2294 unsigned int num_tx_queues;
2295 unsigned int real_num_tx_queues;
2296 struct Qdisc __rcu *qdisc;
2297 unsigned int tx_queue_len;
2298 spinlock_t tx_global_lock;
2300 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2303 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2305 #ifdef CONFIG_NET_XGRESS
2306 struct bpf_mprog_entry __rcu *tcx_egress;
2308 #ifdef CONFIG_NETFILTER_EGRESS
2309 struct nf_hook_entries __rcu *nf_hooks_egress;
2312 #ifdef CONFIG_NET_SCHED
2313 DECLARE_HASHTABLE (qdisc_hash, 4);
2315 /* These may be needed for future network-power-down code. */
2316 struct timer_list watchdog_timer;
2319 u32 proto_down_reason;
2321 struct list_head todo_list;
2323 #ifdef CONFIG_PCPU_DEV_REFCNT
2324 int __percpu *pcpu_refcnt;
2326 refcount_t dev_refcnt;
2328 struct ref_tracker_dir refcnt_tracker;
2330 struct list_head link_watch_list;
2332 enum { NETREG_UNINITIALIZED=0,
2333 NETREG_REGISTERED, /* completed register_netdevice */
2334 NETREG_UNREGISTERING, /* called unregister_netdevice */
2335 NETREG_UNREGISTERED, /* completed unregister todo */
2336 NETREG_RELEASED, /* called free_netdev */
2337 NETREG_DUMMY, /* dummy device for NAPI poll */
2343 RTNL_LINK_INITIALIZED,
2344 RTNL_LINK_INITIALIZING,
2345 } rtnl_link_state:16;
2347 bool needs_free_netdev;
2348 void (*priv_destructor)(struct net_device *dev);
2350 #ifdef CONFIG_NETPOLL
2351 struct netpoll_info __rcu *npinfo;
2354 possible_net_t nd_net;
2356 /* mid-layer private */
2358 enum netdev_ml_priv_type ml_priv_type;
2361 struct pcpu_lstats __percpu *lstats;
2362 struct pcpu_sw_netstats __percpu *tstats;
2363 struct pcpu_dstats __percpu *dstats;
2366 #if IS_ENABLED(CONFIG_GARP)
2367 struct garp_port __rcu *garp_port;
2369 #if IS_ENABLED(CONFIG_MRP)
2370 struct mrp_port __rcu *mrp_port;
2372 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2373 struct dm_hw_stat_delta __rcu *dm_private;
2376 const struct attribute_group *sysfs_groups[4];
2377 const struct attribute_group *sysfs_rx_queue_group;
2379 const struct rtnl_link_ops *rtnl_link_ops;
2381 /* for setting kernel sock attribute on TCP connection setup */
2382 #define GSO_MAX_SEGS 65535u
2383 #define GSO_LEGACY_MAX_SIZE 65536u
2384 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2385 * and shinfo->gso_segs is a 16bit field.
2387 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2389 unsigned int gso_max_size;
2390 #define TSO_LEGACY_MAX_SIZE 65536
2391 #define TSO_MAX_SIZE UINT_MAX
2392 unsigned int tso_max_size;
2394 #define TSO_MAX_SEGS U16_MAX
2396 unsigned int gso_ipv4_max_size;
2399 const struct dcbnl_rtnl_ops *dcbnl_ops;
2402 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2403 u8 prio_tc_map[TC_BITMASK + 1];
2405 #if IS_ENABLED(CONFIG_FCOE)
2406 unsigned int fcoe_ddp_xid;
2408 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2409 struct netprio_map __rcu *priomap;
2411 struct phy_device *phydev;
2412 struct sfp_bus *sfp_bus;
2413 struct lock_class_key *qdisc_tx_busylock;
2415 unsigned wol_enabled:1;
2416 unsigned threaded:1;
2418 struct list_head net_notifier_list;
2420 #if IS_ENABLED(CONFIG_MACSEC)
2421 /* MACsec management functions */
2422 const struct macsec_ops *macsec_ops;
2424 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2425 struct udp_tunnel_nic *udp_tunnel_nic;
2427 /* protected by rtnl_lock */
2428 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2430 u8 dev_addr_shadow[MAX_ADDR_LEN];
2431 netdevice_tracker linkwatch_dev_tracker;
2432 netdevice_tracker watchdog_dev_tracker;
2433 netdevice_tracker dev_registered_tracker;
2434 struct rtnl_hw_stats64 *offload_xstats_l3;
2436 struct devlink_port *devlink_port;
2438 #if IS_ENABLED(CONFIG_DPLL)
2439 struct dpll_pin *dpll_pin;
2442 enum timestamping_layer ts_layer;
2444 #define to_net_dev(d) container_of(d, struct net_device, dev)
2447 * Driver should use this to assign devlink port instance to a netdevice
2448 * before it registers the netdevice. Therefore devlink_port is static
2449 * during the netdev lifetime after it is registered.
2451 #define SET_NETDEV_DEVLINK_PORT(dev, port) \
2453 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2454 ((dev)->devlink_port = (port)); \
2457 static inline bool netif_elide_gro(const struct net_device *dev)
2459 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2464 #define NETDEV_ALIGN 32
2467 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2469 return dev->prio_tc_map[prio & TC_BITMASK];
2473 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2475 if (tc >= dev->num_tc)
2478 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2482 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2483 void netdev_reset_tc(struct net_device *dev);
2484 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2485 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2488 int netdev_get_num_tc(struct net_device *dev)
2493 static inline void net_prefetch(void *p)
2496 #if L1_CACHE_BYTES < 128
2497 prefetch((u8 *)p + L1_CACHE_BYTES);
2501 static inline void net_prefetchw(void *p)
2504 #if L1_CACHE_BYTES < 128
2505 prefetchw((u8 *)p + L1_CACHE_BYTES);
2509 void netdev_unbind_sb_channel(struct net_device *dev,
2510 struct net_device *sb_dev);
2511 int netdev_bind_sb_channel_queue(struct net_device *dev,
2512 struct net_device *sb_dev,
2513 u8 tc, u16 count, u16 offset);
2514 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2515 static inline int netdev_get_sb_channel(struct net_device *dev)
2517 return max_t(int, -dev->num_tc, 0);
2521 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2524 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2525 return &dev->_tx[index];
2528 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2529 const struct sk_buff *skb)
2531 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2534 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2535 void (*f)(struct net_device *,
2536 struct netdev_queue *,
2542 for (i = 0; i < dev->num_tx_queues; i++)
2543 f(dev, &dev->_tx[i], arg);
2546 #define netdev_lockdep_set_classes(dev) \
2548 static struct lock_class_key qdisc_tx_busylock_key; \
2549 static struct lock_class_key qdisc_xmit_lock_key; \
2550 static struct lock_class_key dev_addr_list_lock_key; \
2553 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2554 lockdep_set_class(&(dev)->addr_list_lock, \
2555 &dev_addr_list_lock_key); \
2556 for (i = 0; i < (dev)->num_tx_queues; i++) \
2557 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2558 &qdisc_xmit_lock_key); \
2561 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2562 struct net_device *sb_dev);
2563 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2564 struct sk_buff *skb,
2565 struct net_device *sb_dev);
2567 /* returns the headroom that the master device needs to take in account
2568 * when forwarding to this dev
2570 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2572 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2575 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2577 if (dev->netdev_ops->ndo_set_rx_headroom)
2578 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2581 /* set the device rx headroom to the dev's default */
2582 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2584 netdev_set_rx_headroom(dev, -1);
2587 static inline void *netdev_get_ml_priv(struct net_device *dev,
2588 enum netdev_ml_priv_type type)
2590 if (dev->ml_priv_type != type)
2593 return dev->ml_priv;
2596 static inline void netdev_set_ml_priv(struct net_device *dev,
2598 enum netdev_ml_priv_type type)
2600 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2601 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2602 dev->ml_priv_type, type);
2603 WARN(!dev->ml_priv_type && dev->ml_priv,
2604 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2606 dev->ml_priv = ml_priv;
2607 dev->ml_priv_type = type;
2611 * Net namespace inlines
2614 struct net *dev_net(const struct net_device *dev)
2616 return read_pnet(&dev->nd_net);
2620 void dev_net_set(struct net_device *dev, struct net *net)
2622 write_pnet(&dev->nd_net, net);
2626 * netdev_priv - access network device private data
2627 * @dev: network device
2629 * Get network device private data
2631 static inline void *netdev_priv(const struct net_device *dev)
2633 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2636 /* Set the sysfs physical device reference for the network logical device
2637 * if set prior to registration will cause a symlink during initialization.
2639 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2641 /* Set the sysfs device type for the network logical device to allow
2642 * fine-grained identification of different network device types. For
2643 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2645 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2647 /* Default NAPI poll() weight
2648 * Device drivers are strongly advised to not use bigger value
2650 #define NAPI_POLL_WEIGHT 64
2652 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2653 int (*poll)(struct napi_struct *, int), int weight);
2656 * netif_napi_add() - initialize a NAPI context
2657 * @dev: network device
2658 * @napi: NAPI context
2659 * @poll: polling function
2661 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2662 * *any* of the other NAPI-related functions.
2665 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2666 int (*poll)(struct napi_struct *, int))
2668 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2672 netif_napi_add_tx_weight(struct net_device *dev,
2673 struct napi_struct *napi,
2674 int (*poll)(struct napi_struct *, int),
2677 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2678 netif_napi_add_weight(dev, napi, poll, weight);
2682 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2683 * @dev: network device
2684 * @napi: NAPI context
2685 * @poll: polling function
2687 * This variant of netif_napi_add() should be used from drivers using NAPI
2688 * to exclusively poll a TX queue.
2689 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2691 static inline void netif_napi_add_tx(struct net_device *dev,
2692 struct napi_struct *napi,
2693 int (*poll)(struct napi_struct *, int))
2695 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2699 * __netif_napi_del - remove a NAPI context
2700 * @napi: NAPI context
2702 * Warning: caller must observe RCU grace period before freeing memory
2703 * containing @napi. Drivers might want to call this helper to combine
2704 * all the needed RCU grace periods into a single one.
2706 void __netif_napi_del(struct napi_struct *napi);
2709 * netif_napi_del - remove a NAPI context
2710 * @napi: NAPI context
2712 * netif_napi_del() removes a NAPI context from the network device NAPI list
2714 static inline void netif_napi_del(struct napi_struct *napi)
2716 __netif_napi_del(napi);
2720 struct packet_type {
2721 __be16 type; /* This is really htons(ether_type). */
2722 bool ignore_outgoing;
2723 struct net_device *dev; /* NULL is wildcarded here */
2724 netdevice_tracker dev_tracker;
2725 int (*func) (struct sk_buff *,
2726 struct net_device *,
2727 struct packet_type *,
2728 struct net_device *);
2729 void (*list_func) (struct list_head *,
2730 struct packet_type *,
2731 struct net_device *);
2732 bool (*id_match)(struct packet_type *ptype,
2734 struct net *af_packet_net;
2735 void *af_packet_priv;
2736 struct list_head list;
2739 struct offload_callbacks {
2740 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2741 netdev_features_t features);
2742 struct sk_buff *(*gro_receive)(struct list_head *head,
2743 struct sk_buff *skb);
2744 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2747 struct packet_offload {
2748 __be16 type; /* This is really htons(ether_type). */
2750 struct offload_callbacks callbacks;
2751 struct list_head list;
2754 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2755 struct pcpu_sw_netstats {
2756 u64_stats_t rx_packets;
2757 u64_stats_t rx_bytes;
2758 u64_stats_t tx_packets;
2759 u64_stats_t tx_bytes;
2760 struct u64_stats_sync syncp;
2761 } __aligned(4 * sizeof(u64));
2763 struct pcpu_lstats {
2764 u64_stats_t packets;
2766 struct u64_stats_sync syncp;
2767 } __aligned(2 * sizeof(u64));
2769 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2771 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2773 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2775 u64_stats_update_begin(&tstats->syncp);
2776 u64_stats_add(&tstats->rx_bytes, len);
2777 u64_stats_inc(&tstats->rx_packets);
2778 u64_stats_update_end(&tstats->syncp);
2781 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2782 unsigned int packets,
2785 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2787 u64_stats_update_begin(&tstats->syncp);
2788 u64_stats_add(&tstats->tx_bytes, len);
2789 u64_stats_add(&tstats->tx_packets, packets);
2790 u64_stats_update_end(&tstats->syncp);
2793 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2795 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2797 u64_stats_update_begin(&lstats->syncp);
2798 u64_stats_add(&lstats->bytes, len);
2799 u64_stats_inc(&lstats->packets);
2800 u64_stats_update_end(&lstats->syncp);
2803 #define __netdev_alloc_pcpu_stats(type, gfp) \
2805 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2808 for_each_possible_cpu(__cpu) { \
2809 typeof(type) *stat; \
2810 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2811 u64_stats_init(&stat->syncp); \
2817 #define netdev_alloc_pcpu_stats(type) \
2818 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2820 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2822 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2825 for_each_possible_cpu(__cpu) { \
2826 typeof(type) *stat; \
2827 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2828 u64_stats_init(&stat->syncp); \
2834 enum netdev_lag_tx_type {
2835 NETDEV_LAG_TX_TYPE_UNKNOWN,
2836 NETDEV_LAG_TX_TYPE_RANDOM,
2837 NETDEV_LAG_TX_TYPE_BROADCAST,
2838 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2839 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2840 NETDEV_LAG_TX_TYPE_HASH,
2843 enum netdev_lag_hash {
2844 NETDEV_LAG_HASH_NONE,
2846 NETDEV_LAG_HASH_L34,
2847 NETDEV_LAG_HASH_L23,
2848 NETDEV_LAG_HASH_E23,
2849 NETDEV_LAG_HASH_E34,
2850 NETDEV_LAG_HASH_VLAN_SRCMAC,
2851 NETDEV_LAG_HASH_UNKNOWN,
2854 struct netdev_lag_upper_info {
2855 enum netdev_lag_tx_type tx_type;
2856 enum netdev_lag_hash hash_type;
2859 struct netdev_lag_lower_state_info {
2864 #include <linux/notifier.h>
2866 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2867 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2871 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2873 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2874 detected a hardware crash and restarted
2875 - we can use this eg to kick tcp sessions
2877 NETDEV_CHANGE, /* Notify device state change */
2880 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2881 NETDEV_CHANGEADDR, /* notify after the address change */
2882 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2886 NETDEV_BONDING_FAILOVER,
2888 NETDEV_PRE_TYPE_CHANGE,
2889 NETDEV_POST_TYPE_CHANGE,
2893 NETDEV_NOTIFY_PEERS,
2897 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2898 NETDEV_CHANGEINFODATA,
2899 NETDEV_BONDING_INFO,
2900 NETDEV_PRECHANGEUPPER,
2901 NETDEV_CHANGELOWERSTATE,
2902 NETDEV_UDP_TUNNEL_PUSH_INFO,
2903 NETDEV_UDP_TUNNEL_DROP_INFO,
2904 NETDEV_CHANGE_TX_QUEUE_LEN,
2905 NETDEV_CVLAN_FILTER_PUSH_INFO,
2906 NETDEV_CVLAN_FILTER_DROP_INFO,
2907 NETDEV_SVLAN_FILTER_PUSH_INFO,
2908 NETDEV_SVLAN_FILTER_DROP_INFO,
2909 NETDEV_OFFLOAD_XSTATS_ENABLE,
2910 NETDEV_OFFLOAD_XSTATS_DISABLE,
2911 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2912 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2913 NETDEV_XDP_FEAT_CHANGE,
2915 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2917 int register_netdevice_notifier(struct notifier_block *nb);
2918 int unregister_netdevice_notifier(struct notifier_block *nb);
2919 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2920 int unregister_netdevice_notifier_net(struct net *net,
2921 struct notifier_block *nb);
2922 int register_netdevice_notifier_dev_net(struct net_device *dev,
2923 struct notifier_block *nb,
2924 struct netdev_net_notifier *nn);
2925 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2926 struct notifier_block *nb,
2927 struct netdev_net_notifier *nn);
2929 struct netdev_notifier_info {
2930 struct net_device *dev;
2931 struct netlink_ext_ack *extack;
2934 struct netdev_notifier_info_ext {
2935 struct netdev_notifier_info info; /* must be first */
2941 struct netdev_notifier_change_info {
2942 struct netdev_notifier_info info; /* must be first */
2943 unsigned int flags_changed;
2946 struct netdev_notifier_changeupper_info {
2947 struct netdev_notifier_info info; /* must be first */
2948 struct net_device *upper_dev; /* new upper dev */
2949 bool master; /* is upper dev master */
2950 bool linking; /* is the notification for link or unlink */
2951 void *upper_info; /* upper dev info */
2954 struct netdev_notifier_changelowerstate_info {
2955 struct netdev_notifier_info info; /* must be first */
2956 void *lower_state_info; /* is lower dev state */
2959 struct netdev_notifier_pre_changeaddr_info {
2960 struct netdev_notifier_info info; /* must be first */
2961 const unsigned char *dev_addr;
2964 enum netdev_offload_xstats_type {
2965 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2968 struct netdev_notifier_offload_xstats_info {
2969 struct netdev_notifier_info info; /* must be first */
2970 enum netdev_offload_xstats_type type;
2973 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2974 struct netdev_notifier_offload_xstats_rd *report_delta;
2975 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2976 struct netdev_notifier_offload_xstats_ru *report_used;
2980 int netdev_offload_xstats_enable(struct net_device *dev,
2981 enum netdev_offload_xstats_type type,
2982 struct netlink_ext_ack *extack);
2983 int netdev_offload_xstats_disable(struct net_device *dev,
2984 enum netdev_offload_xstats_type type);
2985 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2986 enum netdev_offload_xstats_type type);
2987 int netdev_offload_xstats_get(struct net_device *dev,
2988 enum netdev_offload_xstats_type type,
2989 struct rtnl_hw_stats64 *stats, bool *used,
2990 struct netlink_ext_ack *extack);
2992 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2993 const struct rtnl_hw_stats64 *stats);
2995 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2996 void netdev_offload_xstats_push_delta(struct net_device *dev,
2997 enum netdev_offload_xstats_type type,
2998 const struct rtnl_hw_stats64 *stats);
3000 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3001 struct net_device *dev)
3004 info->extack = NULL;
3007 static inline struct net_device *
3008 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3013 static inline struct netlink_ext_ack *
3014 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3016 return info->extack;
3019 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3020 int call_netdevice_notifiers_info(unsigned long val,
3021 struct netdev_notifier_info *info);
3023 extern rwlock_t dev_base_lock; /* Device list lock */
3025 #define for_each_netdev(net, d) \
3026 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3027 #define for_each_netdev_reverse(net, d) \
3028 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3029 #define for_each_netdev_rcu(net, d) \
3030 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3031 #define for_each_netdev_safe(net, d, n) \
3032 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3033 #define for_each_netdev_continue(net, d) \
3034 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3035 #define for_each_netdev_continue_reverse(net, d) \
3036 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3038 #define for_each_netdev_continue_rcu(net, d) \
3039 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3040 #define for_each_netdev_in_bond_rcu(bond, slave) \
3041 for_each_netdev_rcu(&init_net, slave) \
3042 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3043 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3045 #define for_each_netdev_dump(net, d, ifindex) \
3046 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3048 static inline struct net_device *next_net_device(struct net_device *dev)
3050 struct list_head *lh;
3054 lh = dev->dev_list.next;
3055 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3058 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3060 struct list_head *lh;
3064 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3065 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3068 static inline struct net_device *first_net_device(struct net *net)
3070 return list_empty(&net->dev_base_head) ? NULL :
3071 net_device_entry(net->dev_base_head.next);
3074 static inline struct net_device *first_net_device_rcu(struct net *net)
3076 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3078 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3081 int netdev_boot_setup_check(struct net_device *dev);
3082 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3083 const char *hwaddr);
3084 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3085 void dev_add_pack(struct packet_type *pt);
3086 void dev_remove_pack(struct packet_type *pt);
3087 void __dev_remove_pack(struct packet_type *pt);
3088 void dev_add_offload(struct packet_offload *po);
3089 void dev_remove_offload(struct packet_offload *po);
3091 int dev_get_iflink(const struct net_device *dev);
3092 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3093 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3094 struct net_device_path_stack *stack);
3095 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3096 unsigned short mask);
3097 struct net_device *dev_get_by_name(struct net *net, const char *name);
3098 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3099 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3100 bool netdev_name_in_use(struct net *net, const char *name);
3101 int dev_alloc_name(struct net_device *dev, const char *name);
3102 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3103 void dev_close(struct net_device *dev);
3104 void dev_close_many(struct list_head *head, bool unlink);
3105 void dev_disable_lro(struct net_device *dev);
3106 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3107 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3108 struct net_device *sb_dev);
3109 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3110 struct net_device *sb_dev);
3112 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3113 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3115 static inline int dev_queue_xmit(struct sk_buff *skb)
3117 return __dev_queue_xmit(skb, NULL);
3120 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3121 struct net_device *sb_dev)
3123 return __dev_queue_xmit(skb, sb_dev);
3126 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3130 ret = __dev_direct_xmit(skb, queue_id);
3131 if (!dev_xmit_complete(ret))
3136 int register_netdevice(struct net_device *dev);
3137 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3138 void unregister_netdevice_many(struct list_head *head);
3139 static inline void unregister_netdevice(struct net_device *dev)
3141 unregister_netdevice_queue(dev, NULL);
3144 int netdev_refcnt_read(const struct net_device *dev);
3145 void free_netdev(struct net_device *dev);
3146 void netdev_freemem(struct net_device *dev);
3147 int init_dummy_netdev(struct net_device *dev);
3149 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3150 struct sk_buff *skb,
3152 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3154 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3155 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3156 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3157 netdevice_tracker *tracker, gfp_t gfp);
3158 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3159 netdevice_tracker *tracker, gfp_t gfp);
3160 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3161 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3163 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3164 unsigned short type,
3165 const void *daddr, const void *saddr,
3168 if (!dev->header_ops || !dev->header_ops->create)
3171 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3174 static inline int dev_parse_header(const struct sk_buff *skb,
3175 unsigned char *haddr)
3177 const struct net_device *dev = skb->dev;
3179 if (!dev->header_ops || !dev->header_ops->parse)
3181 return dev->header_ops->parse(skb, haddr);
3184 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3186 const struct net_device *dev = skb->dev;
3188 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3190 return dev->header_ops->parse_protocol(skb);
3193 /* ll_header must have at least hard_header_len allocated */
3194 static inline bool dev_validate_header(const struct net_device *dev,
3195 char *ll_header, int len)
3197 if (likely(len >= dev->hard_header_len))
3199 if (len < dev->min_header_len)
3202 if (capable(CAP_SYS_RAWIO)) {
3203 memset(ll_header + len, 0, dev->hard_header_len - len);
3207 if (dev->header_ops && dev->header_ops->validate)
3208 return dev->header_ops->validate(ll_header, len);
3213 static inline bool dev_has_header(const struct net_device *dev)
3215 return dev->header_ops && dev->header_ops->create;
3219 * Incoming packets are placed on per-CPU queues
3221 struct softnet_data {
3222 struct list_head poll_list;
3223 struct sk_buff_head process_queue;
3226 unsigned int processed;
3227 unsigned int time_squeeze;
3229 struct softnet_data *rps_ipi_list;
3232 bool in_net_rx_action;
3233 bool in_napi_threaded_poll;
3235 #ifdef CONFIG_NET_FLOW_LIMIT
3236 struct sd_flow_limit __rcu *flow_limit;
3238 struct Qdisc *output_queue;
3239 struct Qdisc **output_queue_tailp;
3240 struct sk_buff *completion_queue;
3241 #ifdef CONFIG_XFRM_OFFLOAD
3242 struct sk_buff_head xfrm_backlog;
3244 /* written and read only by owning cpu: */
3248 #ifdef CONFIG_NET_EGRESS
3253 /* input_queue_head should be written by cpu owning this struct,
3254 * and only read by other cpus. Worth using a cache line.
3256 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3258 /* Elements below can be accessed between CPUs for RPS/RFS */
3259 call_single_data_t csd ____cacheline_aligned_in_smp;
3260 struct softnet_data *rps_ipi_next;
3262 unsigned int input_queue_tail;
3264 unsigned int received_rps;
3265 unsigned int dropped;
3266 struct sk_buff_head input_pkt_queue;
3267 struct napi_struct backlog;
3269 /* Another possibly contended cache line */
3270 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3272 int defer_ipi_scheduled;
3273 struct sk_buff *defer_list;
3274 call_single_data_t defer_csd;
3277 static inline void input_queue_head_incr(struct softnet_data *sd)
3280 sd->input_queue_head++;
3284 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3285 unsigned int *qtail)
3288 *qtail = ++sd->input_queue_tail;
3292 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3294 static inline int dev_recursion_level(void)
3296 return this_cpu_read(softnet_data.xmit.recursion);
3299 #define XMIT_RECURSION_LIMIT 8
3300 static inline bool dev_xmit_recursion(void)
3302 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3303 XMIT_RECURSION_LIMIT);
3306 static inline void dev_xmit_recursion_inc(void)
3308 __this_cpu_inc(softnet_data.xmit.recursion);
3311 static inline void dev_xmit_recursion_dec(void)
3313 __this_cpu_dec(softnet_data.xmit.recursion);
3316 void __netif_schedule(struct Qdisc *q);
3317 void netif_schedule_queue(struct netdev_queue *txq);
3319 static inline void netif_tx_schedule_all(struct net_device *dev)
3323 for (i = 0; i < dev->num_tx_queues; i++)
3324 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3327 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3329 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3333 * netif_start_queue - allow transmit
3334 * @dev: network device
3336 * Allow upper layers to call the device hard_start_xmit routine.
3338 static inline void netif_start_queue(struct net_device *dev)
3340 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3343 static inline void netif_tx_start_all_queues(struct net_device *dev)
3347 for (i = 0; i < dev->num_tx_queues; i++) {
3348 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3349 netif_tx_start_queue(txq);
3353 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3356 * netif_wake_queue - restart transmit
3357 * @dev: network device
3359 * Allow upper layers to call the device hard_start_xmit routine.
3360 * Used for flow control when transmit resources are available.
3362 static inline void netif_wake_queue(struct net_device *dev)
3364 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3367 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3371 for (i = 0; i < dev->num_tx_queues; i++) {
3372 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3373 netif_tx_wake_queue(txq);
3377 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3379 /* Must be an atomic op see netif_txq_try_stop() */
3380 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3384 * netif_stop_queue - stop transmitted packets
3385 * @dev: network device
3387 * Stop upper layers calling the device hard_start_xmit routine.
3388 * Used for flow control when transmit resources are unavailable.
3390 static inline void netif_stop_queue(struct net_device *dev)
3392 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3395 void netif_tx_stop_all_queues(struct net_device *dev);
3397 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3399 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3403 * netif_queue_stopped - test if transmit queue is flowblocked
3404 * @dev: network device
3406 * Test if transmit queue on device is currently unable to send.
3408 static inline bool netif_queue_stopped(const struct net_device *dev)
3410 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3413 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3415 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3419 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3421 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3425 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3427 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3431 * netdev_queue_set_dql_min_limit - set dql minimum limit
3432 * @dev_queue: pointer to transmit queue
3433 * @min_limit: dql minimum limit
3435 * Forces xmit_more() to return true until the minimum threshold
3436 * defined by @min_limit is reached (or until the tx queue is
3437 * empty). Warning: to be use with care, misuse will impact the
3440 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3441 unsigned int min_limit)
3444 dev_queue->dql.min_limit = min_limit;
3449 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3450 * @dev_queue: pointer to transmit queue
3452 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3453 * to give appropriate hint to the CPU.
3455 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3458 prefetchw(&dev_queue->dql.num_queued);
3463 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3464 * @dev_queue: pointer to transmit queue
3466 * BQL enabled drivers might use this helper in their TX completion path,
3467 * to give appropriate hint to the CPU.
3469 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3472 prefetchw(&dev_queue->dql.limit);
3477 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3478 * @dev_queue: network device queue
3479 * @bytes: number of bytes queued to the device queue
3481 * Report the number of bytes queued for sending/completion to the network
3482 * device hardware queue. @bytes should be a good approximation and should
3483 * exactly match netdev_completed_queue() @bytes.
3484 * This is typically called once per packet, from ndo_start_xmit().
3486 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3490 dql_queued(&dev_queue->dql, bytes);
3492 if (likely(dql_avail(&dev_queue->dql) >= 0))
3495 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3498 * The XOFF flag must be set before checking the dql_avail below,
3499 * because in netdev_tx_completed_queue we update the dql_completed
3500 * before checking the XOFF flag.
3504 /* check again in case another CPU has just made room avail */
3505 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3506 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3510 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3511 * that they should not test BQL status themselves.
3512 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3514 * Returns true if the doorbell must be used to kick the NIC.
3516 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3522 dql_queued(&dev_queue->dql, bytes);
3524 return netif_tx_queue_stopped(dev_queue);
3526 netdev_tx_sent_queue(dev_queue, bytes);
3531 * netdev_sent_queue - report the number of bytes queued to hardware
3532 * @dev: network device
3533 * @bytes: number of bytes queued to the hardware device queue
3535 * Report the number of bytes queued for sending/completion to the network
3536 * device hardware queue#0. @bytes should be a good approximation and should
3537 * exactly match netdev_completed_queue() @bytes.
3538 * This is typically called once per packet, from ndo_start_xmit().
3540 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3542 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3545 static inline bool __netdev_sent_queue(struct net_device *dev,
3549 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3554 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3555 * @dev_queue: network device queue
3556 * @pkts: number of packets (currently ignored)
3557 * @bytes: number of bytes dequeued from the device queue
3559 * Must be called at most once per TX completion round (and not per
3560 * individual packet), so that BQL can adjust its limits appropriately.
3562 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3563 unsigned int pkts, unsigned int bytes)
3566 if (unlikely(!bytes))
3569 dql_completed(&dev_queue->dql, bytes);
3572 * Without the memory barrier there is a small possiblity that
3573 * netdev_tx_sent_queue will miss the update and cause the queue to
3574 * be stopped forever
3576 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3578 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3581 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3582 netif_schedule_queue(dev_queue);
3587 * netdev_completed_queue - report bytes and packets completed by device
3588 * @dev: network device
3589 * @pkts: actual number of packets sent over the medium
3590 * @bytes: actual number of bytes sent over the medium
3592 * Report the number of bytes and packets transmitted by the network device
3593 * hardware queue over the physical medium, @bytes must exactly match the
3594 * @bytes amount passed to netdev_sent_queue()
3596 static inline void netdev_completed_queue(struct net_device *dev,
3597 unsigned int pkts, unsigned int bytes)
3599 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3602 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3605 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3611 * netdev_reset_queue - reset the packets and bytes count of a network device
3612 * @dev_queue: network device
3614 * Reset the bytes and packet count of a network device and clear the
3615 * software flow control OFF bit for this network device
3617 static inline void netdev_reset_queue(struct net_device *dev_queue)
3619 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3623 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3624 * @dev: network device
3625 * @queue_index: given tx queue index
3627 * Returns 0 if given tx queue index >= number of device tx queues,
3628 * otherwise returns the originally passed tx queue index.
3630 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3632 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3633 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3634 dev->name, queue_index,
3635 dev->real_num_tx_queues);
3643 * netif_running - test if up
3644 * @dev: network device
3646 * Test if the device has been brought up.
3648 static inline bool netif_running(const struct net_device *dev)
3650 return test_bit(__LINK_STATE_START, &dev->state);
3654 * Routines to manage the subqueues on a device. We only need start,
3655 * stop, and a check if it's stopped. All other device management is
3656 * done at the overall netdevice level.
3657 * Also test the device if we're multiqueue.
3661 * netif_start_subqueue - allow sending packets on subqueue
3662 * @dev: network device
3663 * @queue_index: sub queue index
3665 * Start individual transmit queue of a device with multiple transmit queues.
3667 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3669 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3671 netif_tx_start_queue(txq);
3675 * netif_stop_subqueue - stop sending packets on subqueue
3676 * @dev: network device
3677 * @queue_index: sub queue index
3679 * Stop individual transmit queue of a device with multiple transmit queues.
3681 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3683 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3684 netif_tx_stop_queue(txq);
3688 * __netif_subqueue_stopped - test status of subqueue
3689 * @dev: network device
3690 * @queue_index: sub queue index
3692 * Check individual transmit queue of a device with multiple transmit queues.
3694 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3697 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3699 return netif_tx_queue_stopped(txq);
3703 * netif_subqueue_stopped - test status of subqueue
3704 * @dev: network device
3705 * @skb: sub queue buffer pointer
3707 * Check individual transmit queue of a device with multiple transmit queues.
3709 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3710 struct sk_buff *skb)
3712 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3716 * netif_wake_subqueue - allow sending packets on subqueue
3717 * @dev: network device
3718 * @queue_index: sub queue index
3720 * Resume individual transmit queue of a device with multiple transmit queues.
3722 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3724 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3726 netif_tx_wake_queue(txq);
3730 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3732 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3733 u16 index, enum xps_map_type type);
3736 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3737 * @j: CPU/Rx queue index
3738 * @mask: bitmask of all cpus/rx queues
3739 * @nr_bits: number of bits in the bitmask
3741 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3743 static inline bool netif_attr_test_mask(unsigned long j,
3744 const unsigned long *mask,
3745 unsigned int nr_bits)
3747 cpu_max_bits_warn(j, nr_bits);
3748 return test_bit(j, mask);
3752 * netif_attr_test_online - Test for online CPU/Rx queue
3753 * @j: CPU/Rx queue index
3754 * @online_mask: bitmask for CPUs/Rx queues that are online
3755 * @nr_bits: number of bits in the bitmask
3757 * Returns true if a CPU/Rx queue is online.
3759 static inline bool netif_attr_test_online(unsigned long j,
3760 const unsigned long *online_mask,
3761 unsigned int nr_bits)
3763 cpu_max_bits_warn(j, nr_bits);
3766 return test_bit(j, online_mask);
3768 return (j < nr_bits);
3772 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3773 * @n: CPU/Rx queue index
3774 * @srcp: the cpumask/Rx queue mask pointer
3775 * @nr_bits: number of bits in the bitmask
3777 * Returns >= nr_bits if no further CPUs/Rx queues set.
3779 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3780 unsigned int nr_bits)
3782 /* -1 is a legal arg here. */
3784 cpu_max_bits_warn(n, nr_bits);
3787 return find_next_bit(srcp, nr_bits, n + 1);
3793 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3794 * @n: CPU/Rx queue index
3795 * @src1p: the first CPUs/Rx queues mask pointer
3796 * @src2p: the second CPUs/Rx queues mask pointer
3797 * @nr_bits: number of bits in the bitmask
3799 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3801 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3802 const unsigned long *src2p,
3803 unsigned int nr_bits)
3805 /* -1 is a legal arg here. */
3807 cpu_max_bits_warn(n, nr_bits);
3810 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3812 return find_next_bit(src1p, nr_bits, n + 1);
3814 return find_next_bit(src2p, nr_bits, n + 1);
3819 static inline int netif_set_xps_queue(struct net_device *dev,
3820 const struct cpumask *mask,
3826 static inline int __netif_set_xps_queue(struct net_device *dev,
3827 const unsigned long *mask,
3828 u16 index, enum xps_map_type type)
3835 * netif_is_multiqueue - test if device has multiple transmit queues
3836 * @dev: network device
3838 * Check if device has multiple transmit queues
3840 static inline bool netif_is_multiqueue(const struct net_device *dev)
3842 return dev->num_tx_queues > 1;
3845 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3848 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3850 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3853 dev->real_num_rx_queues = rxqs;
3857 int netif_set_real_num_queues(struct net_device *dev,
3858 unsigned int txq, unsigned int rxq);
3860 int netif_get_num_default_rss_queues(void);
3862 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3863 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3866 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3867 * interrupt context or with hardware interrupts being disabled.
3868 * (in_hardirq() || irqs_disabled())
3870 * We provide four helpers that can be used in following contexts :
3872 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3873 * replacing kfree_skb(skb)
3875 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3876 * Typically used in place of consume_skb(skb) in TX completion path
3878 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3879 * replacing kfree_skb(skb)
3881 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3882 * and consumed a packet. Used in place of consume_skb(skb)
3884 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3886 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3889 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3891 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3894 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3896 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3899 static inline void dev_consume_skb_any(struct sk_buff *skb)
3901 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3904 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3905 struct bpf_prog *xdp_prog);
3906 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3907 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3908 int netif_rx(struct sk_buff *skb);
3909 int __netif_rx(struct sk_buff *skb);
3911 int netif_receive_skb(struct sk_buff *skb);
3912 int netif_receive_skb_core(struct sk_buff *skb);
3913 void netif_receive_skb_list_internal(struct list_head *head);
3914 void netif_receive_skb_list(struct list_head *head);
3915 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3916 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3917 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3918 void napi_get_frags_check(struct napi_struct *napi);
3919 gro_result_t napi_gro_frags(struct napi_struct *napi);
3920 struct packet_offload *gro_find_receive_by_type(__be16 type);
3921 struct packet_offload *gro_find_complete_by_type(__be16 type);
3923 static inline void napi_free_frags(struct napi_struct *napi)
3925 kfree_skb(napi->skb);
3929 bool netdev_is_rx_handler_busy(struct net_device *dev);
3930 int netdev_rx_handler_register(struct net_device *dev,
3931 rx_handler_func_t *rx_handler,
3932 void *rx_handler_data);
3933 void netdev_rx_handler_unregister(struct net_device *dev);
3935 bool dev_valid_name(const char *name);
3936 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3938 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3940 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3941 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3942 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3943 void __user *data, bool *need_copyout);
3944 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3945 int generic_hwtstamp_get_lower(struct net_device *dev,
3946 struct kernel_hwtstamp_config *kernel_cfg);
3947 int generic_hwtstamp_set_lower(struct net_device *dev,
3948 struct kernel_hwtstamp_config *kernel_cfg,
3949 struct netlink_ext_ack *extack);
3950 int dev_set_hwtstamp_phylib(struct net_device *dev,
3951 struct kernel_hwtstamp_config *cfg,
3952 struct netlink_ext_ack *extack);
3953 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3954 unsigned int dev_get_flags(const struct net_device *);
3955 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3956 struct netlink_ext_ack *extack);
3957 int dev_change_flags(struct net_device *dev, unsigned int flags,
3958 struct netlink_ext_ack *extack);
3959 int dev_set_alias(struct net_device *, const char *, size_t);
3960 int dev_get_alias(const struct net_device *, char *, size_t);
3961 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3962 const char *pat, int new_ifindex);
3964 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3967 return __dev_change_net_namespace(dev, net, pat, 0);
3969 int __dev_set_mtu(struct net_device *, int);
3970 int dev_set_mtu(struct net_device *, int);
3971 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3972 struct netlink_ext_ack *extack);
3973 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3974 struct netlink_ext_ack *extack);
3975 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3976 struct netlink_ext_ack *extack);
3977 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3978 int dev_get_port_parent_id(struct net_device *dev,
3979 struct netdev_phys_item_id *ppid, bool recurse);
3980 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3981 void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin);
3982 void netdev_dpll_pin_clear(struct net_device *dev);
3984 static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev)
3986 #if IS_ENABLED(CONFIG_DPLL)
3987 return dev->dpll_pin;
3993 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3994 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3995 struct netdev_queue *txq, int *ret);
3997 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3998 u8 dev_xdp_prog_count(struct net_device *dev);
3999 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4001 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4002 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4003 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4004 bool is_skb_forwardable(const struct net_device *dev,
4005 const struct sk_buff *skb);
4007 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4008 const struct sk_buff *skb,
4009 const bool check_mtu)
4011 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4014 if (!(dev->flags & IFF_UP))
4020 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4021 if (skb->len <= len)
4024 /* if TSO is enabled, we don't care about the length as the packet
4025 * could be forwarded without being segmented before
4027 if (skb_is_gso(skb))
4033 void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4035 #define DEV_CORE_STATS_INC(FIELD) \
4036 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4038 netdev_core_stats_inc(dev, \
4039 offsetof(struct net_device_core_stats, FIELD)); \
4041 DEV_CORE_STATS_INC(rx_dropped)
4042 DEV_CORE_STATS_INC(tx_dropped)
4043 DEV_CORE_STATS_INC(rx_nohandler)
4044 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4045 #undef DEV_CORE_STATS_INC
4047 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4048 struct sk_buff *skb,
4049 const bool check_mtu)
4051 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4052 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4053 dev_core_stats_rx_dropped_inc(dev);
4058 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4063 bool dev_nit_active(struct net_device *dev);
4064 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4066 static inline void __dev_put(struct net_device *dev)
4069 #ifdef CONFIG_PCPU_DEV_REFCNT
4070 this_cpu_dec(*dev->pcpu_refcnt);
4072 refcount_dec(&dev->dev_refcnt);
4077 static inline void __dev_hold(struct net_device *dev)
4080 #ifdef CONFIG_PCPU_DEV_REFCNT
4081 this_cpu_inc(*dev->pcpu_refcnt);
4083 refcount_inc(&dev->dev_refcnt);
4088 static inline void __netdev_tracker_alloc(struct net_device *dev,
4089 netdevice_tracker *tracker,
4092 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4093 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4097 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4098 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4100 static inline void netdev_tracker_alloc(struct net_device *dev,
4101 netdevice_tracker *tracker, gfp_t gfp)
4103 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4104 refcount_dec(&dev->refcnt_tracker.no_tracker);
4105 __netdev_tracker_alloc(dev, tracker, gfp);
4109 static inline void netdev_tracker_free(struct net_device *dev,
4110 netdevice_tracker *tracker)
4112 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4113 ref_tracker_free(&dev->refcnt_tracker, tracker);
4117 static inline void netdev_hold(struct net_device *dev,
4118 netdevice_tracker *tracker, gfp_t gfp)
4122 __netdev_tracker_alloc(dev, tracker, gfp);
4126 static inline void netdev_put(struct net_device *dev,
4127 netdevice_tracker *tracker)
4130 netdev_tracker_free(dev, tracker);
4136 * dev_hold - get reference to device
4137 * @dev: network device
4139 * Hold reference to device to keep it from being freed.
4140 * Try using netdev_hold() instead.
4142 static inline void dev_hold(struct net_device *dev)
4144 netdev_hold(dev, NULL, GFP_ATOMIC);
4148 * dev_put - release reference to device
4149 * @dev: network device
4151 * Release reference to device to allow it to be freed.
4152 * Try using netdev_put() instead.
4154 static inline void dev_put(struct net_device *dev)
4156 netdev_put(dev, NULL);
4159 static inline void netdev_ref_replace(struct net_device *odev,
4160 struct net_device *ndev,
4161 netdevice_tracker *tracker,
4165 netdev_tracker_free(odev, tracker);
4171 __netdev_tracker_alloc(ndev, tracker, gfp);
4174 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4175 * and _off may be called from IRQ context, but it is caller
4176 * who is responsible for serialization of these calls.
4178 * The name carrier is inappropriate, these functions should really be
4179 * called netif_lowerlayer_*() because they represent the state of any
4180 * kind of lower layer not just hardware media.
4182 void linkwatch_fire_event(struct net_device *dev);
4185 * netif_carrier_ok - test if carrier present
4186 * @dev: network device
4188 * Check if carrier is present on device
4190 static inline bool netif_carrier_ok(const struct net_device *dev)
4192 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4195 unsigned long dev_trans_start(struct net_device *dev);
4197 void __netdev_watchdog_up(struct net_device *dev);
4199 void netif_carrier_on(struct net_device *dev);
4200 void netif_carrier_off(struct net_device *dev);
4201 void netif_carrier_event(struct net_device *dev);
4204 * netif_dormant_on - mark device as dormant.
4205 * @dev: network device
4207 * Mark device as dormant (as per RFC2863).
4209 * The dormant state indicates that the relevant interface is not
4210 * actually in a condition to pass packets (i.e., it is not 'up') but is
4211 * in a "pending" state, waiting for some external event. For "on-
4212 * demand" interfaces, this new state identifies the situation where the
4213 * interface is waiting for events to place it in the up state.
4215 static inline void netif_dormant_on(struct net_device *dev)
4217 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4218 linkwatch_fire_event(dev);
4222 * netif_dormant_off - set device as not dormant.
4223 * @dev: network device
4225 * Device is not in dormant state.
4227 static inline void netif_dormant_off(struct net_device *dev)
4229 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4230 linkwatch_fire_event(dev);
4234 * netif_dormant - test if device is dormant
4235 * @dev: network device
4237 * Check if device is dormant.
4239 static inline bool netif_dormant(const struct net_device *dev)
4241 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4246 * netif_testing_on - mark device as under test.
4247 * @dev: network device
4249 * Mark device as under test (as per RFC2863).
4251 * The testing state indicates that some test(s) must be performed on
4252 * the interface. After completion, of the test, the interface state
4253 * will change to up, dormant, or down, as appropriate.
4255 static inline void netif_testing_on(struct net_device *dev)
4257 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4258 linkwatch_fire_event(dev);
4262 * netif_testing_off - set device as not under test.
4263 * @dev: network device
4265 * Device is not in testing state.
4267 static inline void netif_testing_off(struct net_device *dev)
4269 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4270 linkwatch_fire_event(dev);
4274 * netif_testing - test if device is under test
4275 * @dev: network device
4277 * Check if device is under test
4279 static inline bool netif_testing(const struct net_device *dev)
4281 return test_bit(__LINK_STATE_TESTING, &dev->state);
4286 * netif_oper_up - test if device is operational
4287 * @dev: network device
4289 * Check if carrier is operational
4291 static inline bool netif_oper_up(const struct net_device *dev)
4293 return (dev->operstate == IF_OPER_UP ||
4294 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4298 * netif_device_present - is device available or removed
4299 * @dev: network device
4301 * Check if device has not been removed from system.
4303 static inline bool netif_device_present(const struct net_device *dev)
4305 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4308 void netif_device_detach(struct net_device *dev);
4310 void netif_device_attach(struct net_device *dev);
4313 * Network interface message level settings
4318 NETIF_MSG_PROBE_BIT,
4320 NETIF_MSG_TIMER_BIT,
4321 NETIF_MSG_IFDOWN_BIT,
4323 NETIF_MSG_RX_ERR_BIT,
4324 NETIF_MSG_TX_ERR_BIT,
4325 NETIF_MSG_TX_QUEUED_BIT,
4327 NETIF_MSG_TX_DONE_BIT,
4328 NETIF_MSG_RX_STATUS_BIT,
4329 NETIF_MSG_PKTDATA_BIT,
4333 /* When you add a new bit above, update netif_msg_class_names array
4334 * in net/ethtool/common.c
4336 NETIF_MSG_CLASS_COUNT,
4338 /* Both ethtool_ops interface and internal driver implementation use u32 */
4339 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4341 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4342 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4344 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4345 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4346 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4347 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4348 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4349 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4350 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4351 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4352 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4353 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4354 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4355 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4356 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4357 #define NETIF_MSG_HW __NETIF_MSG(HW)
4358 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4360 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4361 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4362 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4363 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4364 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4365 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4366 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4367 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4368 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4369 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4370 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4371 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4372 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4373 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4374 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4376 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4379 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4380 return default_msg_enable_bits;
4381 if (debug_value == 0) /* no output */
4383 /* set low N bits */
4384 return (1U << debug_value) - 1;
4387 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4389 spin_lock(&txq->_xmit_lock);
4390 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4391 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4394 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4396 __acquire(&txq->_xmit_lock);
4400 static inline void __netif_tx_release(struct netdev_queue *txq)
4402 __release(&txq->_xmit_lock);
4405 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4407 spin_lock_bh(&txq->_xmit_lock);
4408 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4409 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4412 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4414 bool ok = spin_trylock(&txq->_xmit_lock);
4417 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4418 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4423 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4425 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4426 WRITE_ONCE(txq->xmit_lock_owner, -1);
4427 spin_unlock(&txq->_xmit_lock);
4430 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4432 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4433 WRITE_ONCE(txq->xmit_lock_owner, -1);
4434 spin_unlock_bh(&txq->_xmit_lock);
4438 * txq->trans_start can be read locklessly from dev_watchdog()
4440 static inline void txq_trans_update(struct netdev_queue *txq)
4442 if (txq->xmit_lock_owner != -1)
4443 WRITE_ONCE(txq->trans_start, jiffies);
4446 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4448 unsigned long now = jiffies;
4450 if (READ_ONCE(txq->trans_start) != now)
4451 WRITE_ONCE(txq->trans_start, now);
4454 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4455 static inline void netif_trans_update(struct net_device *dev)
4457 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4459 txq_trans_cond_update(txq);
4463 * netif_tx_lock - grab network device transmit lock
4464 * @dev: network device
4466 * Get network device transmit lock
4468 void netif_tx_lock(struct net_device *dev);
4470 static inline void netif_tx_lock_bh(struct net_device *dev)
4476 void netif_tx_unlock(struct net_device *dev);
4478 static inline void netif_tx_unlock_bh(struct net_device *dev)
4480 netif_tx_unlock(dev);
4484 #define HARD_TX_LOCK(dev, txq, cpu) { \
4485 if ((dev->features & NETIF_F_LLTX) == 0) { \
4486 __netif_tx_lock(txq, cpu); \
4488 __netif_tx_acquire(txq); \
4492 #define HARD_TX_TRYLOCK(dev, txq) \
4493 (((dev->features & NETIF_F_LLTX) == 0) ? \
4494 __netif_tx_trylock(txq) : \
4495 __netif_tx_acquire(txq))
4497 #define HARD_TX_UNLOCK(dev, txq) { \
4498 if ((dev->features & NETIF_F_LLTX) == 0) { \
4499 __netif_tx_unlock(txq); \
4501 __netif_tx_release(txq); \
4505 static inline void netif_tx_disable(struct net_device *dev)
4511 cpu = smp_processor_id();
4512 spin_lock(&dev->tx_global_lock);
4513 for (i = 0; i < dev->num_tx_queues; i++) {
4514 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4516 __netif_tx_lock(txq, cpu);
4517 netif_tx_stop_queue(txq);
4518 __netif_tx_unlock(txq);
4520 spin_unlock(&dev->tx_global_lock);
4524 static inline void netif_addr_lock(struct net_device *dev)
4526 unsigned char nest_level = 0;
4528 #ifdef CONFIG_LOCKDEP
4529 nest_level = dev->nested_level;
4531 spin_lock_nested(&dev->addr_list_lock, nest_level);
4534 static inline void netif_addr_lock_bh(struct net_device *dev)
4536 unsigned char nest_level = 0;
4538 #ifdef CONFIG_LOCKDEP
4539 nest_level = dev->nested_level;
4542 spin_lock_nested(&dev->addr_list_lock, nest_level);
4545 static inline void netif_addr_unlock(struct net_device *dev)
4547 spin_unlock(&dev->addr_list_lock);
4550 static inline void netif_addr_unlock_bh(struct net_device *dev)
4552 spin_unlock_bh(&dev->addr_list_lock);
4556 * dev_addrs walker. Should be used only for read access. Call with
4557 * rcu_read_lock held.
4559 #define for_each_dev_addr(dev, ha) \
4560 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4562 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4564 void ether_setup(struct net_device *dev);
4566 /* Support for loadable net-drivers */
4567 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4568 unsigned char name_assign_type,
4569 void (*setup)(struct net_device *),
4570 unsigned int txqs, unsigned int rxqs);
4571 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4572 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4574 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4575 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4578 int register_netdev(struct net_device *dev);
4579 void unregister_netdev(struct net_device *dev);
4581 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4583 /* General hardware address lists handling functions */
4584 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4585 struct netdev_hw_addr_list *from_list, int addr_len);
4586 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4587 struct netdev_hw_addr_list *from_list, int addr_len);
4588 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4589 struct net_device *dev,
4590 int (*sync)(struct net_device *, const unsigned char *),
4591 int (*unsync)(struct net_device *,
4592 const unsigned char *));
4593 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4594 struct net_device *dev,
4595 int (*sync)(struct net_device *,
4596 const unsigned char *, int),
4597 int (*unsync)(struct net_device *,
4598 const unsigned char *, int));
4599 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4600 struct net_device *dev,
4601 int (*unsync)(struct net_device *,
4602 const unsigned char *, int));
4603 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4604 struct net_device *dev,
4605 int (*unsync)(struct net_device *,
4606 const unsigned char *));
4607 void __hw_addr_init(struct netdev_hw_addr_list *list);
4609 /* Functions used for device addresses handling */
4610 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4611 const void *addr, size_t len);
4614 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4616 dev_addr_mod(dev, 0, addr, len);
4619 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4621 __dev_addr_set(dev, addr, dev->addr_len);
4624 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4625 unsigned char addr_type);
4626 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4627 unsigned char addr_type);
4629 /* Functions used for unicast addresses handling */
4630 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4631 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4632 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4633 int dev_uc_sync(struct net_device *to, struct net_device *from);
4634 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4635 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4636 void dev_uc_flush(struct net_device *dev);
4637 void dev_uc_init(struct net_device *dev);
4640 * __dev_uc_sync - Synchonize device's unicast list
4641 * @dev: device to sync
4642 * @sync: function to call if address should be added
4643 * @unsync: function to call if address should be removed
4645 * Add newly added addresses to the interface, and release
4646 * addresses that have been deleted.
4648 static inline int __dev_uc_sync(struct net_device *dev,
4649 int (*sync)(struct net_device *,
4650 const unsigned char *),
4651 int (*unsync)(struct net_device *,
4652 const unsigned char *))
4654 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4658 * __dev_uc_unsync - Remove synchronized addresses from device
4659 * @dev: device to sync
4660 * @unsync: function to call if address should be removed
4662 * Remove all addresses that were added to the device by dev_uc_sync().
4664 static inline void __dev_uc_unsync(struct net_device *dev,
4665 int (*unsync)(struct net_device *,
4666 const unsigned char *))
4668 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4671 /* Functions used for multicast addresses handling */
4672 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4673 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4674 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4675 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4676 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4677 int dev_mc_sync(struct net_device *to, struct net_device *from);
4678 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4679 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4680 void dev_mc_flush(struct net_device *dev);
4681 void dev_mc_init(struct net_device *dev);
4684 * __dev_mc_sync - Synchonize device's multicast list
4685 * @dev: device to sync
4686 * @sync: function to call if address should be added
4687 * @unsync: function to call if address should be removed
4689 * Add newly added addresses to the interface, and release
4690 * addresses that have been deleted.
4692 static inline int __dev_mc_sync(struct net_device *dev,
4693 int (*sync)(struct net_device *,
4694 const unsigned char *),
4695 int (*unsync)(struct net_device *,
4696 const unsigned char *))
4698 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4702 * __dev_mc_unsync - Remove synchronized addresses from device
4703 * @dev: device to sync
4704 * @unsync: function to call if address should be removed
4706 * Remove all addresses that were added to the device by dev_mc_sync().
4708 static inline void __dev_mc_unsync(struct net_device *dev,
4709 int (*unsync)(struct net_device *,
4710 const unsigned char *))
4712 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4715 /* Functions used for secondary unicast and multicast support */
4716 void dev_set_rx_mode(struct net_device *dev);
4717 int dev_set_promiscuity(struct net_device *dev, int inc);
4718 int dev_set_allmulti(struct net_device *dev, int inc);
4719 void netdev_state_change(struct net_device *dev);
4720 void __netdev_notify_peers(struct net_device *dev);
4721 void netdev_notify_peers(struct net_device *dev);
4722 void netdev_features_change(struct net_device *dev);
4723 /* Load a device via the kmod */
4724 void dev_load(struct net *net, const char *name);
4725 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4726 struct rtnl_link_stats64 *storage);
4727 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4728 const struct net_device_stats *netdev_stats);
4729 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4730 const struct pcpu_sw_netstats __percpu *netstats);
4731 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4733 extern int netdev_max_backlog;
4734 extern int dev_rx_weight;
4735 extern int dev_tx_weight;
4736 extern int gro_normal_batch;
4739 NESTED_SYNC_IMM_BIT,
4740 NESTED_SYNC_TODO_BIT,
4743 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4744 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4746 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4747 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4749 struct netdev_nested_priv {
4750 unsigned char flags;
4754 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4755 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4756 struct list_head **iter);
4758 /* iterate through upper list, must be called under RCU read lock */
4759 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4760 for (iter = &(dev)->adj_list.upper, \
4761 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4763 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4765 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4766 int (*fn)(struct net_device *upper_dev,
4767 struct netdev_nested_priv *priv),
4768 struct netdev_nested_priv *priv);
4770 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4771 struct net_device *upper_dev);
4773 bool netdev_has_any_upper_dev(struct net_device *dev);
4775 void *netdev_lower_get_next_private(struct net_device *dev,
4776 struct list_head **iter);
4777 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4778 struct list_head **iter);
4780 #define netdev_for_each_lower_private(dev, priv, iter) \
4781 for (iter = (dev)->adj_list.lower.next, \
4782 priv = netdev_lower_get_next_private(dev, &(iter)); \
4784 priv = netdev_lower_get_next_private(dev, &(iter)))
4786 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4787 for (iter = &(dev)->adj_list.lower, \
4788 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4790 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4792 void *netdev_lower_get_next(struct net_device *dev,
4793 struct list_head **iter);
4795 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4796 for (iter = (dev)->adj_list.lower.next, \
4797 ldev = netdev_lower_get_next(dev, &(iter)); \
4799 ldev = netdev_lower_get_next(dev, &(iter)))
4801 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4802 struct list_head **iter);
4803 int netdev_walk_all_lower_dev(struct net_device *dev,
4804 int (*fn)(struct net_device *lower_dev,
4805 struct netdev_nested_priv *priv),
4806 struct netdev_nested_priv *priv);
4807 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4808 int (*fn)(struct net_device *lower_dev,
4809 struct netdev_nested_priv *priv),
4810 struct netdev_nested_priv *priv);
4812 void *netdev_adjacent_get_private(struct list_head *adj_list);
4813 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4814 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4815 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4816 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4817 struct netlink_ext_ack *extack);
4818 int netdev_master_upper_dev_link(struct net_device *dev,
4819 struct net_device *upper_dev,
4820 void *upper_priv, void *upper_info,
4821 struct netlink_ext_ack *extack);
4822 void netdev_upper_dev_unlink(struct net_device *dev,
4823 struct net_device *upper_dev);
4824 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4825 struct net_device *new_dev,
4826 struct net_device *dev,
4827 struct netlink_ext_ack *extack);
4828 void netdev_adjacent_change_commit(struct net_device *old_dev,
4829 struct net_device *new_dev,
4830 struct net_device *dev);
4831 void netdev_adjacent_change_abort(struct net_device *old_dev,
4832 struct net_device *new_dev,
4833 struct net_device *dev);
4834 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4835 void *netdev_lower_dev_get_private(struct net_device *dev,
4836 struct net_device *lower_dev);
4837 void netdev_lower_state_changed(struct net_device *lower_dev,
4838 void *lower_state_info);
4840 /* RSS keys are 40 or 52 bytes long */
4841 #define NETDEV_RSS_KEY_LEN 52
4842 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4843 void netdev_rss_key_fill(void *buffer, size_t len);
4845 int skb_checksum_help(struct sk_buff *skb);
4846 int skb_crc32c_csum_help(struct sk_buff *skb);
4847 int skb_csum_hwoffload_help(struct sk_buff *skb,
4848 const netdev_features_t features);
4850 struct netdev_bonding_info {
4855 struct netdev_notifier_bonding_info {
4856 struct netdev_notifier_info info; /* must be first */
4857 struct netdev_bonding_info bonding_info;
4860 void netdev_bonding_info_change(struct net_device *dev,
4861 struct netdev_bonding_info *bonding_info);
4863 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4864 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4866 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4872 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4874 static inline bool can_checksum_protocol(netdev_features_t features,
4877 if (protocol == htons(ETH_P_FCOE))
4878 return !!(features & NETIF_F_FCOE_CRC);
4880 /* Assume this is an IP checksum (not SCTP CRC) */
4882 if (features & NETIF_F_HW_CSUM) {
4883 /* Can checksum everything */
4888 case htons(ETH_P_IP):
4889 return !!(features & NETIF_F_IP_CSUM);
4890 case htons(ETH_P_IPV6):
4891 return !!(features & NETIF_F_IPV6_CSUM);
4898 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4900 static inline void netdev_rx_csum_fault(struct net_device *dev,
4901 struct sk_buff *skb)
4905 /* rx skb timestamps */
4906 void net_enable_timestamp(void);
4907 void net_disable_timestamp(void);
4909 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4910 const struct skb_shared_hwtstamps *hwtstamps,
4913 const struct net_device_ops *ops = dev->netdev_ops;
4915 if (ops->ndo_get_tstamp)
4916 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4918 return hwtstamps->hwtstamp;
4921 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4922 struct sk_buff *skb, struct net_device *dev,
4925 __this_cpu_write(softnet_data.xmit.more, more);
4926 return ops->ndo_start_xmit(skb, dev);
4929 static inline bool netdev_xmit_more(void)
4931 return __this_cpu_read(softnet_data.xmit.more);
4934 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4935 struct netdev_queue *txq, bool more)
4937 const struct net_device_ops *ops = dev->netdev_ops;
4940 rc = __netdev_start_xmit(ops, skb, dev, more);
4941 if (rc == NETDEV_TX_OK)
4942 txq_trans_update(txq);
4947 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4949 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4952 extern const struct kobj_ns_type_operations net_ns_type_operations;
4954 const char *netdev_drivername(const struct net_device *dev);
4956 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4957 netdev_features_t f2)
4959 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4960 if (f1 & NETIF_F_HW_CSUM)
4961 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4963 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4969 static inline netdev_features_t netdev_get_wanted_features(
4970 struct net_device *dev)
4972 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4974 netdev_features_t netdev_increment_features(netdev_features_t all,
4975 netdev_features_t one, netdev_features_t mask);
4977 /* Allow TSO being used on stacked device :
4978 * Performing the GSO segmentation before last device
4979 * is a performance improvement.
4981 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4982 netdev_features_t mask)
4984 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4987 int __netdev_update_features(struct net_device *dev);
4988 void netdev_update_features(struct net_device *dev);
4989 void netdev_change_features(struct net_device *dev);
4991 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4992 struct net_device *dev);
4994 netdev_features_t passthru_features_check(struct sk_buff *skb,
4995 struct net_device *dev,
4996 netdev_features_t features);
4997 netdev_features_t netif_skb_features(struct sk_buff *skb);
4998 void skb_warn_bad_offload(const struct sk_buff *skb);
5000 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5002 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5004 /* check flags correspondence */
5005 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5006 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5007 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5008 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5009 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5010 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5011 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5012 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5013 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5014 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5015 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5016 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5017 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5018 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5019 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5020 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5021 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5022 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5023 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5025 return (features & feature) == feature;
5028 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5030 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5031 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5034 static inline bool netif_needs_gso(struct sk_buff *skb,
5035 netdev_features_t features)
5037 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5038 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5039 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5042 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5043 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5044 void netif_inherit_tso_max(struct net_device *to,
5045 const struct net_device *from);
5047 static inline bool netif_is_macsec(const struct net_device *dev)
5049 return dev->priv_flags & IFF_MACSEC;
5052 static inline bool netif_is_macvlan(const struct net_device *dev)
5054 return dev->priv_flags & IFF_MACVLAN;
5057 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5059 return dev->priv_flags & IFF_MACVLAN_PORT;
5062 static inline bool netif_is_bond_master(const struct net_device *dev)
5064 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5067 static inline bool netif_is_bond_slave(const struct net_device *dev)
5069 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5072 static inline bool netif_supports_nofcs(struct net_device *dev)
5074 return dev->priv_flags & IFF_SUPP_NOFCS;
5077 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5079 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5082 static inline bool netif_is_l3_master(const struct net_device *dev)
5084 return dev->priv_flags & IFF_L3MDEV_MASTER;
5087 static inline bool netif_is_l3_slave(const struct net_device *dev)
5089 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5092 static inline int dev_sdif(const struct net_device *dev)
5094 #ifdef CONFIG_NET_L3_MASTER_DEV
5095 if (netif_is_l3_slave(dev))
5096 return dev->ifindex;
5101 static inline bool netif_is_bridge_master(const struct net_device *dev)
5103 return dev->priv_flags & IFF_EBRIDGE;
5106 static inline bool netif_is_bridge_port(const struct net_device *dev)
5108 return dev->priv_flags & IFF_BRIDGE_PORT;
5111 static inline bool netif_is_ovs_master(const struct net_device *dev)
5113 return dev->priv_flags & IFF_OPENVSWITCH;
5116 static inline bool netif_is_ovs_port(const struct net_device *dev)
5118 return dev->priv_flags & IFF_OVS_DATAPATH;
5121 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5123 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5126 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5128 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5131 static inline bool netif_is_team_master(const struct net_device *dev)
5133 return dev->priv_flags & IFF_TEAM;
5136 static inline bool netif_is_team_port(const struct net_device *dev)
5138 return dev->priv_flags & IFF_TEAM_PORT;
5141 static inline bool netif_is_lag_master(const struct net_device *dev)
5143 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5146 static inline bool netif_is_lag_port(const struct net_device *dev)
5148 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5151 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5153 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5156 static inline bool netif_is_failover(const struct net_device *dev)
5158 return dev->priv_flags & IFF_FAILOVER;
5161 static inline bool netif_is_failover_slave(const struct net_device *dev)
5163 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5166 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5167 static inline void netif_keep_dst(struct net_device *dev)
5169 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5172 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5173 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5175 /* TODO: reserve and use an additional IFF bit, if we get more users */
5176 return netif_is_macsec(dev);
5179 extern struct pernet_operations __net_initdata loopback_net_ops;
5181 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5183 /* netdev_printk helpers, similar to dev_printk */
5185 static inline const char *netdev_name(const struct net_device *dev)
5187 if (!dev->name[0] || strchr(dev->name, '%'))
5188 return "(unnamed net_device)";
5192 static inline const char *netdev_reg_state(const struct net_device *dev)
5194 switch (dev->reg_state) {
5195 case NETREG_UNINITIALIZED: return " (uninitialized)";
5196 case NETREG_REGISTERED: return "";
5197 case NETREG_UNREGISTERING: return " (unregistering)";
5198 case NETREG_UNREGISTERED: return " (unregistered)";
5199 case NETREG_RELEASED: return " (released)";
5200 case NETREG_DUMMY: return " (dummy)";
5203 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5204 return " (unknown)";
5207 #define MODULE_ALIAS_NETDEV(device) \
5208 MODULE_ALIAS("netdev-" device)
5211 * netdev_WARN() acts like dev_printk(), but with the key difference
5212 * of using a WARN/WARN_ON to get the message out, including the
5213 * file/line information and a backtrace.
5215 #define netdev_WARN(dev, format, args...) \
5216 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5217 netdev_reg_state(dev), ##args)
5219 #define netdev_WARN_ONCE(dev, format, args...) \
5220 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5221 netdev_reg_state(dev), ##args)
5224 * The list of packet types we will receive (as opposed to discard)
5225 * and the routines to invoke.
5227 * Why 16. Because with 16 the only overlap we get on a hash of the
5228 * low nibble of the protocol value is RARP/SNAP/X.25.
5242 #define PTYPE_HASH_SIZE (16)
5243 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5245 extern struct list_head ptype_all __read_mostly;
5246 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5248 extern struct net_device *blackhole_netdev;
5250 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5251 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5252 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5253 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5254 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5256 #endif /* _LINUX_NETDEVICE_H */