1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Definitions for the Interfaces handler.
9 * Version: @(#)dev.h 1.0.10 08/12/93
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
19 * Moved to /usr/include/linux for NET3
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
32 #include <linux/percpu.h>
33 #include <linux/rculist.h>
34 #include <linux/workqueue.h>
35 #include <linux/dynamic_queue_limits.h>
37 #include <net/net_namespace.h>
39 #include <net/dcbnl.h>
41 #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 <linux/hashtable.h>
50 #include <linux/rbtree.h>
51 #include <net/net_trackers.h>
58 struct ip_tunnel_parm;
59 struct macsec_context;
65 /* 802.15.4 specific */
68 /* UDP Tunnel offloads */
69 struct udp_tunnel_info;
70 struct udp_tunnel_nic_info;
71 struct udp_tunnel_nic;
75 void synchronize_net(void);
76 void netdev_set_default_ethtool_ops(struct net_device *dev,
77 const struct ethtool_ops *ops);
79 /* Backlog congestion levels */
80 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
81 #define NET_RX_DROP 1 /* packet dropped */
83 #define MAX_NEST_DEV 8
86 * Transmit return codes: transmit return codes originate from three different
89 * - qdisc return codes
90 * - driver transmit return codes
93 * Drivers are allowed to return any one of those in their hard_start_xmit()
94 * function. Real network devices commonly used with qdiscs should only return
95 * the driver transmit return codes though - when qdiscs are used, the actual
96 * transmission happens asynchronously, so the value is not propagated to
97 * higher layers. Virtual network devices transmit synchronously; in this case
98 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
99 * others are propagated to higher layers.
102 /* qdisc ->enqueue() return codes. */
103 #define NET_XMIT_SUCCESS 0x00
104 #define NET_XMIT_DROP 0x01 /* skb dropped */
105 #define NET_XMIT_CN 0x02 /* congestion notification */
106 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
108 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
109 * indicates that the device will soon be dropping packets, or already drops
110 * some packets of the same priority; prompting us to send less aggressively. */
111 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
112 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
114 /* Driver transmit return codes */
115 #define NETDEV_TX_MASK 0xf0
118 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
119 NETDEV_TX_OK = 0x00, /* driver took care of packet */
120 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
122 typedef enum netdev_tx netdev_tx_t;
125 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
126 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
128 static inline bool dev_xmit_complete(int rc)
131 * Positive cases with an skb consumed by a driver:
132 * - successful transmission (rc == NETDEV_TX_OK)
133 * - error while transmitting (rc < 0)
134 * - error while queueing to a different device (rc & NET_XMIT_MASK)
136 if (likely(rc < NET_XMIT_MASK))
143 * Compute the worst-case header length according to the protocols
147 #if defined(CONFIG_HYPERV_NET)
148 # define LL_MAX_HEADER 128
149 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
150 # if defined(CONFIG_MAC80211_MESH)
151 # define LL_MAX_HEADER 128
153 # define LL_MAX_HEADER 96
156 # define LL_MAX_HEADER 32
159 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
160 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
161 #define MAX_HEADER LL_MAX_HEADER
163 #define MAX_HEADER (LL_MAX_HEADER + 48)
167 * Old network device statistics. Fields are native words
168 * (unsigned long) so they can be read and written atomically.
171 struct net_device_stats {
172 unsigned long rx_packets;
173 unsigned long tx_packets;
174 unsigned long rx_bytes;
175 unsigned long tx_bytes;
176 unsigned long rx_errors;
177 unsigned long tx_errors;
178 unsigned long rx_dropped;
179 unsigned long tx_dropped;
180 unsigned long multicast;
181 unsigned long collisions;
182 unsigned long rx_length_errors;
183 unsigned long rx_over_errors;
184 unsigned long rx_crc_errors;
185 unsigned long rx_frame_errors;
186 unsigned long rx_fifo_errors;
187 unsigned long rx_missed_errors;
188 unsigned long tx_aborted_errors;
189 unsigned long tx_carrier_errors;
190 unsigned long tx_fifo_errors;
191 unsigned long tx_heartbeat_errors;
192 unsigned long tx_window_errors;
193 unsigned long rx_compressed;
194 unsigned long tx_compressed;
198 #include <linux/cache.h>
199 #include <linux/skbuff.h>
202 #include <linux/static_key.h>
203 extern struct static_key_false rps_needed;
204 extern struct static_key_false rfs_needed;
211 struct netdev_hw_addr {
212 struct list_head list;
214 unsigned char addr[MAX_ADDR_LEN];
216 #define NETDEV_HW_ADDR_T_LAN 1
217 #define NETDEV_HW_ADDR_T_SAN 2
218 #define NETDEV_HW_ADDR_T_UNICAST 3
219 #define NETDEV_HW_ADDR_T_MULTICAST 4
224 struct rcu_head rcu_head;
227 struct netdev_hw_addr_list {
228 struct list_head list;
231 /* Auxiliary tree for faster lookup on addition and deletion */
235 #define netdev_hw_addr_list_count(l) ((l)->count)
236 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
237 #define netdev_hw_addr_list_for_each(ha, l) \
238 list_for_each_entry(ha, &(l)->list, list)
240 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
241 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
242 #define netdev_for_each_uc_addr(ha, dev) \
243 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
245 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
246 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
247 #define netdev_for_each_mc_addr(ha, dev) \
248 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
254 /* cached hardware header; allow for machine alignment needs. */
255 #define HH_DATA_MOD 16
256 #define HH_DATA_OFF(__len) \
257 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
258 #define HH_DATA_ALIGN(__len) \
259 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
260 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
263 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
265 * dev->hard_header_len ? (dev->hard_header_len +
266 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
268 * We could use other alignment values, but we must maintain the
269 * relationship HH alignment <= LL alignment.
271 #define LL_RESERVED_SPACE(dev) \
272 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
273 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
274 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
277 int (*create) (struct sk_buff *skb, struct net_device *dev,
278 unsigned short type, const void *daddr,
279 const void *saddr, unsigned int len);
280 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
281 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
282 void (*cache_update)(struct hh_cache *hh,
283 const struct net_device *dev,
284 const unsigned char *haddr);
285 bool (*validate)(const char *ll_header, unsigned int len);
286 __be16 (*parse_protocol)(const struct sk_buff *skb);
289 /* These flag bits are private to the generic network queueing
290 * layer; they may not be explicitly referenced by any other
294 enum netdev_state_t {
296 __LINK_STATE_PRESENT,
297 __LINK_STATE_NOCARRIER,
298 __LINK_STATE_LINKWATCH_PENDING,
299 __LINK_STATE_DORMANT,
300 __LINK_STATE_TESTING,
304 struct list_head list;
309 * size of gro hash buckets, must less than bit number of
310 * napi_struct::gro_bitmask
312 #define GRO_HASH_BUCKETS 8
315 * Structure for NAPI scheduling similar to tasklet but with weighting
318 /* The poll_list must only be managed by the entity which
319 * changes the state of the NAPI_STATE_SCHED bit. This means
320 * whoever atomically sets that bit can add this napi_struct
321 * to the per-CPU poll_list, and whoever clears that bit
322 * can remove from the list right before clearing the bit.
324 struct list_head poll_list;
328 int defer_hard_irqs_count;
329 unsigned long gro_bitmask;
330 int (*poll)(struct napi_struct *, int);
331 #ifdef CONFIG_NETPOLL
334 struct net_device *dev;
335 struct gro_list gro_hash[GRO_HASH_BUCKETS];
337 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
338 int rx_count; /* length of rx_list */
339 struct hrtimer timer;
340 struct list_head dev_list;
341 struct hlist_node napi_hash_node;
342 unsigned int napi_id;
343 struct task_struct *thread;
347 NAPI_STATE_SCHED, /* Poll is scheduled */
348 NAPI_STATE_MISSED, /* reschedule a napi */
349 NAPI_STATE_DISABLE, /* Disable pending */
350 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
351 NAPI_STATE_LISTED, /* NAPI added to system lists */
352 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
353 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
354 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
355 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
356 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
360 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
361 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
362 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
363 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
364 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
365 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
366 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
367 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
368 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
369 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
379 typedef enum gro_result gro_result_t;
382 * enum rx_handler_result - Possible return values for rx_handlers.
383 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
385 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
386 * case skb->dev was changed by rx_handler.
387 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
388 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
390 * rx_handlers are functions called from inside __netif_receive_skb(), to do
391 * special processing of the skb, prior to delivery to protocol handlers.
393 * Currently, a net_device can only have a single rx_handler registered. Trying
394 * to register a second rx_handler will return -EBUSY.
396 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
397 * To unregister a rx_handler on a net_device, use
398 * netdev_rx_handler_unregister().
400 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
403 * If the rx_handler consumed the skb in some way, it should return
404 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
405 * the skb to be delivered in some other way.
407 * If the rx_handler changed skb->dev, to divert the skb to another
408 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
409 * new device will be called if it exists.
411 * If the rx_handler decides the skb should be ignored, it should return
412 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
413 * are registered on exact device (ptype->dev == skb->dev).
415 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
416 * delivered, it should return RX_HANDLER_PASS.
418 * A device without a registered rx_handler will behave as if rx_handler
419 * returned RX_HANDLER_PASS.
422 enum rx_handler_result {
428 typedef enum rx_handler_result rx_handler_result_t;
429 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
431 void __napi_schedule(struct napi_struct *n);
432 void __napi_schedule_irqoff(struct napi_struct *n);
434 static inline bool napi_disable_pending(struct napi_struct *n)
436 return test_bit(NAPI_STATE_DISABLE, &n->state);
439 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
441 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
444 bool napi_schedule_prep(struct napi_struct *n);
447 * napi_schedule - schedule NAPI poll
450 * Schedule NAPI poll routine to be called if it is not already
453 static inline void napi_schedule(struct napi_struct *n)
455 if (napi_schedule_prep(n))
460 * napi_schedule_irqoff - schedule NAPI poll
463 * Variant of napi_schedule(), assuming hard irqs are masked.
465 static inline void napi_schedule_irqoff(struct napi_struct *n)
467 if (napi_schedule_prep(n))
468 __napi_schedule_irqoff(n);
471 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
472 static inline bool napi_reschedule(struct napi_struct *napi)
474 if (napi_schedule_prep(napi)) {
475 __napi_schedule(napi);
481 bool napi_complete_done(struct napi_struct *n, int work_done);
483 * napi_complete - NAPI processing complete
486 * Mark NAPI processing as complete.
487 * Consider using napi_complete_done() instead.
488 * Return false if device should avoid rearming interrupts.
490 static inline bool napi_complete(struct napi_struct *n)
492 return napi_complete_done(n, 0);
495 int dev_set_threaded(struct net_device *dev, bool threaded);
498 * napi_disable - prevent NAPI from scheduling
501 * Stop NAPI from being scheduled on this context.
502 * Waits till any outstanding processing completes.
504 void napi_disable(struct napi_struct *n);
506 void napi_enable(struct napi_struct *n);
509 * napi_synchronize - wait until NAPI is not running
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
516 static inline void napi_synchronize(const struct napi_struct *n)
518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
526 * napi_if_scheduled_mark_missed - if napi is running, set the
530 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
533 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
535 unsigned long val, new;
538 val = READ_ONCE(n->state);
539 if (val & NAPIF_STATE_DISABLE)
542 if (!(val & NAPIF_STATE_SCHED))
545 new = val | NAPIF_STATE_MISSED;
546 } while (cmpxchg(&n->state, val, new) != val);
551 enum netdev_queue_state_t {
552 __QUEUE_STATE_DRV_XOFF,
553 __QUEUE_STATE_STACK_XOFF,
554 __QUEUE_STATE_FROZEN,
557 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
558 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
559 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
561 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
562 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
564 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
568 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
569 * netif_tx_* functions below are used to manipulate this flag. The
570 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
571 * queue independently. The netif_xmit_*stopped functions below are called
572 * to check if the queue has been stopped by the driver or stack (either
573 * of the XOFF bits are set in the state). Drivers should not need to call
574 * netif_xmit*stopped functions, they should only be using netif_tx_*.
577 struct netdev_queue {
581 struct net_device *dev;
582 netdevice_tracker dev_tracker;
584 struct Qdisc __rcu *qdisc;
585 struct Qdisc *qdisc_sleeping;
589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
592 unsigned long tx_maxrate;
594 * Number of TX timeouts for this queue
595 * (/sys/class/net/DEV/Q/trans_timeout)
597 atomic_long_t trans_timeout;
599 /* Subordinate device that the queue has been assigned to */
600 struct net_device *sb_dev;
601 #ifdef CONFIG_XDP_SOCKETS
602 struct xsk_buff_pool *pool;
607 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
610 * Time (in jiffies) of last Tx
612 unsigned long trans_start;
619 } ____cacheline_aligned_in_smp;
621 extern int sysctl_fb_tunnels_only_for_init_net;
622 extern int sysctl_devconf_inherit_init_net;
625 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
626 * == 1 : For initns only
629 static inline bool net_has_fallback_tunnels(const struct net *net)
631 return !IS_ENABLED(CONFIG_SYSCTL) ||
632 !sysctl_fb_tunnels_only_for_init_net ||
633 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
636 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
638 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
645 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
647 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
654 * This structure holds an RPS map which can be of variable length. The
655 * map is an array of CPUs.
662 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
665 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
666 * tail pointer for that CPU's input queue at the time of last enqueue, and
667 * a hardware filter index.
669 struct rps_dev_flow {
672 unsigned int last_qtail;
674 #define RPS_NO_FILTER 0xffff
677 * The rps_dev_flow_table structure contains a table of flow mappings.
679 struct rps_dev_flow_table {
682 struct rps_dev_flow flows[];
684 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
685 ((_num) * sizeof(struct rps_dev_flow)))
688 * The rps_sock_flow_table contains mappings of flows to the last CPU
689 * on which they were processed by the application (set in recvmsg).
690 * Each entry is a 32bit value. Upper part is the high-order bits
691 * of flow hash, lower part is CPU number.
692 * rps_cpu_mask is used to partition the space, depending on number of
693 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
694 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
695 * meaning we use 32-6=26 bits for the hash.
697 struct rps_sock_flow_table {
700 u32 ents[] ____cacheline_aligned_in_smp;
702 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
704 #define RPS_NO_CPU 0xffff
706 extern u32 rps_cpu_mask;
707 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
709 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
713 unsigned int index = hash & table->mask;
714 u32 val = hash & ~rps_cpu_mask;
716 /* We only give a hint, preemption can change CPU under us */
717 val |= raw_smp_processor_id();
719 if (table->ents[index] != val)
720 table->ents[index] = val;
724 #ifdef CONFIG_RFS_ACCEL
725 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
728 #endif /* CONFIG_RPS */
730 /* This structure contains an instance of an RX queue. */
731 struct netdev_rx_queue {
732 struct xdp_rxq_info xdp_rxq;
734 struct rps_map __rcu *rps_map;
735 struct rps_dev_flow_table __rcu *rps_flow_table;
738 struct net_device *dev;
739 netdevice_tracker dev_tracker;
741 #ifdef CONFIG_XDP_SOCKETS
742 struct xsk_buff_pool *pool;
744 } ____cacheline_aligned_in_smp;
747 * RX queue sysfs structures and functions.
749 struct rx_queue_attribute {
750 struct attribute attr;
751 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
752 ssize_t (*store)(struct netdev_rx_queue *queue,
753 const char *buf, size_t len);
756 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
765 * This structure holds an XPS map which can be of variable length. The
766 * map is an array of queues.
770 unsigned int alloc_len;
774 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
775 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
776 - sizeof(struct xps_map)) / sizeof(u16))
779 * This structure holds all XPS maps for device. Maps are indexed by CPU.
781 * We keep track of the number of cpus/rxqs used when the struct is allocated,
782 * in nr_ids. This will help not accessing out-of-bound memory.
784 * We keep track of the number of traffic classes used when the struct is
785 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
786 * not crossing its upper bound, as the original dev->num_tc can be updated in
789 struct xps_dev_maps {
793 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
796 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
797 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
799 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
800 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
802 #endif /* CONFIG_XPS */
804 #define TC_MAX_QUEUE 16
805 #define TC_BITMASK 15
806 /* HW offloaded queuing disciplines txq count and offset maps */
807 struct netdev_tc_txq {
812 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
814 * This structure is to hold information about the device
815 * configured to run FCoE protocol stack.
817 struct netdev_fcoe_hbainfo {
818 char manufacturer[64];
819 char serial_number[64];
820 char hardware_version[64];
821 char driver_version[64];
822 char optionrom_version[64];
823 char firmware_version[64];
825 char model_description[256];
829 #define MAX_PHYS_ITEM_ID_LEN 32
831 /* This structure holds a unique identifier to identify some
832 * physical item (port for example) used by a netdevice.
834 struct netdev_phys_item_id {
835 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
836 unsigned char id_len;
839 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
840 struct netdev_phys_item_id *b)
842 return a->id_len == b->id_len &&
843 memcmp(a->id, b->id, a->id_len) == 0;
846 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
848 struct net_device *sb_dev);
850 enum net_device_path_type {
851 DEV_PATH_ETHERNET = 0,
858 struct net_device_path {
859 enum net_device_path_type type;
860 const struct net_device *dev;
869 DEV_PATH_BR_VLAN_KEEP,
870 DEV_PATH_BR_VLAN_TAG,
871 DEV_PATH_BR_VLAN_UNTAG,
872 DEV_PATH_BR_VLAN_UNTAG_HW,
884 #define NET_DEVICE_PATH_STACK_MAX 5
885 #define NET_DEVICE_PATH_VLAN_MAX 2
887 struct net_device_path_stack {
889 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
892 struct net_device_path_ctx {
893 const struct net_device *dev;
900 } vlan[NET_DEVICE_PATH_VLAN_MAX];
904 TC_SETUP_QDISC_MQPRIO,
907 TC_SETUP_CLSMATCHALL,
917 TC_SETUP_QDISC_TAPRIO,
926 /* These structures hold the attributes of bpf state that are being passed
927 * to the netdevice through the bpf op.
929 enum bpf_netdev_command {
930 /* Set or clear a bpf program used in the earliest stages of packet
931 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
932 * is responsible for calling bpf_prog_put on any old progs that are
933 * stored. In case of error, the callee need not release the new prog
934 * reference, but on success it takes ownership and must bpf_prog_put
935 * when it is no longer used.
939 /* BPF program for offload callbacks, invoked at program load time. */
940 BPF_OFFLOAD_MAP_ALLOC,
941 BPF_OFFLOAD_MAP_FREE,
945 struct bpf_prog_offload_ops;
946 struct netlink_ext_ack;
948 struct xdp_dev_bulk_queue;
958 struct bpf_xdp_entity {
959 struct bpf_prog *prog;
960 struct bpf_xdp_link *link;
964 enum bpf_netdev_command command;
969 struct bpf_prog *prog;
970 struct netlink_ext_ack *extack;
972 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
974 struct bpf_offloaded_map *offmap;
976 /* XDP_SETUP_XSK_POOL */
978 struct xsk_buff_pool *pool;
984 /* Flags for ndo_xsk_wakeup. */
985 #define XDP_WAKEUP_RX (1 << 0)
986 #define XDP_WAKEUP_TX (1 << 1)
988 #ifdef CONFIG_XFRM_OFFLOAD
990 int (*xdo_dev_state_add) (struct xfrm_state *x);
991 void (*xdo_dev_state_delete) (struct xfrm_state *x);
992 void (*xdo_dev_state_free) (struct xfrm_state *x);
993 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
994 struct xfrm_state *x);
995 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1000 struct rcu_head rcuhead;
1007 struct netdev_name_node {
1008 struct hlist_node hlist;
1009 struct list_head list;
1010 struct net_device *dev;
1014 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
1015 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
1017 struct netdev_net_notifier {
1018 struct list_head list;
1019 struct notifier_block *nb;
1023 * This structure defines the management hooks for network devices.
1024 * The following hooks can be defined; unless noted otherwise, they are
1025 * optional and can be filled with a null pointer.
1027 * int (*ndo_init)(struct net_device *dev);
1028 * This function is called once when a network device is registered.
1029 * The network device can use this for any late stage initialization
1030 * or semantic validation. It can fail with an error code which will
1031 * be propagated back to register_netdev.
1033 * void (*ndo_uninit)(struct net_device *dev);
1034 * This function is called when device is unregistered or when registration
1035 * fails. It is not called if init fails.
1037 * int (*ndo_open)(struct net_device *dev);
1038 * This function is called when a network device transitions to the up
1041 * int (*ndo_stop)(struct net_device *dev);
1042 * This function is called when a network device transitions to the down
1045 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1046 * struct net_device *dev);
1047 * Called when a packet needs to be transmitted.
1048 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1049 * the queue before that can happen; it's for obsolete devices and weird
1050 * corner cases, but the stack really does a non-trivial amount
1051 * of useless work if you return NETDEV_TX_BUSY.
1052 * Required; cannot be NULL.
1054 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1055 * struct net_device *dev
1056 * netdev_features_t features);
1057 * Called by core transmit path to determine if device is capable of
1058 * performing offload operations on a given packet. This is to give
1059 * the device an opportunity to implement any restrictions that cannot
1060 * be otherwise expressed by feature flags. The check is called with
1061 * the set of features that the stack has calculated and it returns
1062 * those the driver believes to be appropriate.
1064 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1065 * struct net_device *sb_dev);
1066 * Called to decide which queue to use when device supports multiple
1069 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1070 * This function is called to allow device receiver to make
1071 * changes to configuration when multicast or promiscuous is enabled.
1073 * void (*ndo_set_rx_mode)(struct net_device *dev);
1074 * This function is called device changes address list filtering.
1075 * If driver handles unicast address filtering, it should set
1076 * IFF_UNICAST_FLT in its priv_flags.
1078 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1079 * This function is called when the Media Access Control address
1080 * needs to be changed. If this interface is not defined, the
1081 * MAC address can not be changed.
1083 * int (*ndo_validate_addr)(struct net_device *dev);
1084 * Test if Media Access Control address is valid for the device.
1086 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1087 * Old-style ioctl entry point. This is used internally by the
1088 * appletalk and ieee802154 subsystems but is no longer called by
1089 * the device ioctl handler.
1091 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1092 * Used by the bonding driver for its device specific ioctls:
1093 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1094 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1096 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1097 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1098 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1100 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1101 * Used to set network devices bus interface parameters. This interface
1102 * is retained for legacy reasons; new devices should use the bus
1103 * interface (PCI) for low level management.
1105 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1106 * Called when a user wants to change the Maximum Transfer Unit
1109 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1110 * Callback used when the transmitter has not made any progress
1111 * for dev->watchdog ticks.
1113 * void (*ndo_get_stats64)(struct net_device *dev,
1114 * struct rtnl_link_stats64 *storage);
1115 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1116 * Called when a user wants to get the network device usage
1117 * statistics. Drivers must do one of the following:
1118 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1119 * rtnl_link_stats64 structure passed by the caller.
1120 * 2. Define @ndo_get_stats to update a net_device_stats structure
1121 * (which should normally be dev->stats) and return a pointer to
1122 * it. The structure may be changed asynchronously only if each
1123 * field is written atomically.
1124 * 3. Update dev->stats asynchronously and atomically, and define
1125 * neither operation.
1127 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1128 * Return true if this device supports offload stats of this attr_id.
1130 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1132 * Get statistics for offload operations by attr_id. Write it into the
1133 * attr_data pointer.
1135 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1136 * If device supports VLAN filtering this function is called when a
1137 * VLAN id is registered.
1139 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1140 * If device supports VLAN filtering this function is called when a
1141 * VLAN id is unregistered.
1143 * void (*ndo_poll_controller)(struct net_device *dev);
1145 * SR-IOV management functions.
1146 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1147 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1148 * u8 qos, __be16 proto);
1149 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1151 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1152 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1153 * int (*ndo_get_vf_config)(struct net_device *dev,
1154 * int vf, struct ifla_vf_info *ivf);
1155 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1156 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1157 * struct nlattr *port[]);
1159 * Enable or disable the VF ability to query its RSS Redirection Table and
1160 * Hash Key. This is needed since on some devices VF share this information
1161 * with PF and querying it may introduce a theoretical security risk.
1162 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1163 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1164 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1166 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1167 * This is always called from the stack with the rtnl lock held and netif
1168 * tx queues stopped. This allows the netdevice to perform queue
1169 * management safely.
1171 * Fiber Channel over Ethernet (FCoE) offload functions.
1172 * int (*ndo_fcoe_enable)(struct net_device *dev);
1173 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1174 * so the underlying device can perform whatever needed configuration or
1175 * initialization to support acceleration of FCoE traffic.
1177 * int (*ndo_fcoe_disable)(struct net_device *dev);
1178 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1179 * so the underlying device can perform whatever needed clean-ups to
1180 * stop supporting acceleration of FCoE traffic.
1182 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1183 * struct scatterlist *sgl, unsigned int sgc);
1184 * Called when the FCoE Initiator wants to initialize an I/O that
1185 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1186 * perform necessary setup and returns 1 to indicate the device is set up
1187 * successfully to perform DDP on this I/O, otherwise this returns 0.
1189 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1190 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1191 * indicated by the FC exchange id 'xid', so the underlying device can
1192 * clean up and reuse resources for later DDP requests.
1194 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1195 * struct scatterlist *sgl, unsigned int sgc);
1196 * Called when the FCoE Target wants to initialize an I/O that
1197 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1198 * perform necessary setup and returns 1 to indicate the device is set up
1199 * successfully to perform DDP on this I/O, otherwise this returns 0.
1201 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1202 * struct netdev_fcoe_hbainfo *hbainfo);
1203 * Called when the FCoE Protocol stack wants information on the underlying
1204 * device. This information is utilized by the FCoE protocol stack to
1205 * register attributes with Fiber Channel management service as per the
1206 * FC-GS Fabric Device Management Information(FDMI) specification.
1208 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1209 * Called when the underlying device wants to override default World Wide
1210 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1211 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1212 * protocol stack to use.
1215 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1216 * u16 rxq_index, u32 flow_id);
1217 * Set hardware filter for RFS. rxq_index is the target queue index;
1218 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1219 * Return the filter ID on success, or a negative error code.
1221 * Slave management functions (for bridge, bonding, etc).
1222 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1223 * Called to make another netdev an underling.
1225 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1226 * Called to release previously enslaved netdev.
1228 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1229 * struct sk_buff *skb,
1231 * Get the xmit slave of master device. If all_slaves is true, function
1232 * assume all the slaves can transmit.
1234 * Feature/offload setting functions.
1235 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1236 * netdev_features_t features);
1237 * Adjusts the requested feature flags according to device-specific
1238 * constraints, and returns the resulting flags. Must not modify
1241 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1242 * Called to update device configuration to new features. Passed
1243 * feature set might be less than what was returned by ndo_fix_features()).
1244 * Must return >0 or -errno if it changed dev->features itself.
1246 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1247 * struct net_device *dev,
1248 * const unsigned char *addr, u16 vid, u16 flags,
1249 * struct netlink_ext_ack *extack);
1250 * Adds an FDB entry to dev for addr.
1251 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1252 * struct net_device *dev,
1253 * const unsigned char *addr, u16 vid)
1254 * Deletes the FDB entry from dev coresponding to addr.
1255 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1256 * struct net_device *dev, struct net_device *filter_dev,
1258 * Used to add FDB entries to dump requests. Implementers should add
1259 * entries to skb and update idx with the number of entries.
1261 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1262 * u16 flags, struct netlink_ext_ack *extack)
1263 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1264 * struct net_device *dev, u32 filter_mask,
1266 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1269 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1270 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1271 * which do not represent real hardware may define this to allow their
1272 * userspace components to manage their virtual carrier state. Devices
1273 * that determine carrier state from physical hardware properties (eg
1274 * network cables) or protocol-dependent mechanisms (eg
1275 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1277 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1278 * struct netdev_phys_item_id *ppid);
1279 * Called to get ID of physical port of this device. If driver does
1280 * not implement this, it is assumed that the hw is not able to have
1281 * multiple net devices on single physical port.
1283 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1284 * struct netdev_phys_item_id *ppid)
1285 * Called to get the parent ID of the physical port of this device.
1287 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1288 * struct net_device *dev)
1289 * Called by upper layer devices to accelerate switching or other
1290 * station functionality into hardware. 'pdev is the lowerdev
1291 * to use for the offload and 'dev' is the net device that will
1292 * back the offload. Returns a pointer to the private structure
1293 * the upper layer will maintain.
1294 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1295 * Called by upper layer device to delete the station created
1296 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1297 * the station and priv is the structure returned by the add
1299 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1300 * int queue_index, u32 maxrate);
1301 * Called when a user wants to set a max-rate limitation of specific
1303 * int (*ndo_get_iflink)(const struct net_device *dev);
1304 * Called to get the iflink value of this device.
1305 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1306 * This function is used to get egress tunnel information for given skb.
1307 * This is useful for retrieving outer tunnel header parameters while
1309 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1310 * This function is used to specify the headroom that the skb must
1311 * consider when allocation skb during packet reception. Setting
1312 * appropriate rx headroom value allows avoiding skb head copy on
1313 * forward. Setting a negative value resets the rx headroom to the
1315 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1316 * This function is used to set or query state related to XDP on the
1317 * netdevice and manage BPF offload. See definition of
1318 * enum bpf_netdev_command for details.
1319 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1321 * This function is used to submit @n XDP packets for transmit on a
1322 * netdevice. Returns number of frames successfully transmitted, frames
1323 * that got dropped are freed/returned via xdp_return_frame().
1324 * Returns negative number, means general error invoking ndo, meaning
1325 * no frames were xmit'ed and core-caller will free all frames.
1326 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1327 * struct xdp_buff *xdp);
1328 * Get the xmit slave of master device based on the xdp_buff.
1329 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1330 * This function is used to wake up the softirq, ksoftirqd or kthread
1331 * responsible for sending and/or receiving packets on a specific
1332 * queue id bound to an AF_XDP socket. The flags field specifies if
1333 * only RX, only Tx, or both should be woken up using the flags
1334 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1335 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1336 * Get devlink port instance associated with a given netdev.
1337 * Called with a reference on the netdevice and devlink locks only,
1338 * rtnl_lock is not held.
1339 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1341 * Add, change, delete or get information on an IPv4 tunnel.
1342 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1343 * If a device is paired with a peer device, return the peer instance.
1344 * The caller must be under RCU read context.
1345 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1346 * Get the forwarding path to reach the real device from the HW destination address
1348 struct net_device_ops {
1349 int (*ndo_init)(struct net_device *dev);
1350 void (*ndo_uninit)(struct net_device *dev);
1351 int (*ndo_open)(struct net_device *dev);
1352 int (*ndo_stop)(struct net_device *dev);
1353 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1354 struct net_device *dev);
1355 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1356 struct net_device *dev,
1357 netdev_features_t features);
1358 u16 (*ndo_select_queue)(struct net_device *dev,
1359 struct sk_buff *skb,
1360 struct net_device *sb_dev);
1361 void (*ndo_change_rx_flags)(struct net_device *dev,
1363 void (*ndo_set_rx_mode)(struct net_device *dev);
1364 int (*ndo_set_mac_address)(struct net_device *dev,
1366 int (*ndo_validate_addr)(struct net_device *dev);
1367 int (*ndo_do_ioctl)(struct net_device *dev,
1368 struct ifreq *ifr, int cmd);
1369 int (*ndo_eth_ioctl)(struct net_device *dev,
1370 struct ifreq *ifr, int cmd);
1371 int (*ndo_siocbond)(struct net_device *dev,
1372 struct ifreq *ifr, int cmd);
1373 int (*ndo_siocwandev)(struct net_device *dev,
1374 struct if_settings *ifs);
1375 int (*ndo_siocdevprivate)(struct net_device *dev,
1377 void __user *data, int cmd);
1378 int (*ndo_set_config)(struct net_device *dev,
1380 int (*ndo_change_mtu)(struct net_device *dev,
1382 int (*ndo_neigh_setup)(struct net_device *dev,
1383 struct neigh_parms *);
1384 void (*ndo_tx_timeout) (struct net_device *dev,
1385 unsigned int txqueue);
1387 void (*ndo_get_stats64)(struct net_device *dev,
1388 struct rtnl_link_stats64 *storage);
1389 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1390 int (*ndo_get_offload_stats)(int attr_id,
1391 const struct net_device *dev,
1393 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1395 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1396 __be16 proto, u16 vid);
1397 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1398 __be16 proto, u16 vid);
1399 #ifdef CONFIG_NET_POLL_CONTROLLER
1400 void (*ndo_poll_controller)(struct net_device *dev);
1401 int (*ndo_netpoll_setup)(struct net_device *dev,
1402 struct netpoll_info *info);
1403 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1405 int (*ndo_set_vf_mac)(struct net_device *dev,
1406 int queue, u8 *mac);
1407 int (*ndo_set_vf_vlan)(struct net_device *dev,
1408 int queue, u16 vlan,
1409 u8 qos, __be16 proto);
1410 int (*ndo_set_vf_rate)(struct net_device *dev,
1411 int vf, int min_tx_rate,
1413 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1414 int vf, bool setting);
1415 int (*ndo_set_vf_trust)(struct net_device *dev,
1416 int vf, bool setting);
1417 int (*ndo_get_vf_config)(struct net_device *dev,
1419 struct ifla_vf_info *ivf);
1420 int (*ndo_set_vf_link_state)(struct net_device *dev,
1421 int vf, int link_state);
1422 int (*ndo_get_vf_stats)(struct net_device *dev,
1424 struct ifla_vf_stats
1426 int (*ndo_set_vf_port)(struct net_device *dev,
1428 struct nlattr *port[]);
1429 int (*ndo_get_vf_port)(struct net_device *dev,
1430 int vf, struct sk_buff *skb);
1431 int (*ndo_get_vf_guid)(struct net_device *dev,
1433 struct ifla_vf_guid *node_guid,
1434 struct ifla_vf_guid *port_guid);
1435 int (*ndo_set_vf_guid)(struct net_device *dev,
1438 int (*ndo_set_vf_rss_query_en)(
1439 struct net_device *dev,
1440 int vf, bool setting);
1441 int (*ndo_setup_tc)(struct net_device *dev,
1442 enum tc_setup_type type,
1444 #if IS_ENABLED(CONFIG_FCOE)
1445 int (*ndo_fcoe_enable)(struct net_device *dev);
1446 int (*ndo_fcoe_disable)(struct net_device *dev);
1447 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1449 struct scatterlist *sgl,
1451 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1453 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1455 struct scatterlist *sgl,
1457 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1458 struct netdev_fcoe_hbainfo *hbainfo);
1461 #if IS_ENABLED(CONFIG_LIBFCOE)
1462 #define NETDEV_FCOE_WWNN 0
1463 #define NETDEV_FCOE_WWPN 1
1464 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1465 u64 *wwn, int type);
1468 #ifdef CONFIG_RFS_ACCEL
1469 int (*ndo_rx_flow_steer)(struct net_device *dev,
1470 const struct sk_buff *skb,
1474 int (*ndo_add_slave)(struct net_device *dev,
1475 struct net_device *slave_dev,
1476 struct netlink_ext_ack *extack);
1477 int (*ndo_del_slave)(struct net_device *dev,
1478 struct net_device *slave_dev);
1479 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1480 struct sk_buff *skb,
1482 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1484 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1485 netdev_features_t features);
1486 int (*ndo_set_features)(struct net_device *dev,
1487 netdev_features_t features);
1488 int (*ndo_neigh_construct)(struct net_device *dev,
1489 struct neighbour *n);
1490 void (*ndo_neigh_destroy)(struct net_device *dev,
1491 struct neighbour *n);
1493 int (*ndo_fdb_add)(struct ndmsg *ndm,
1494 struct nlattr *tb[],
1495 struct net_device *dev,
1496 const unsigned char *addr,
1499 struct netlink_ext_ack *extack);
1500 int (*ndo_fdb_del)(struct ndmsg *ndm,
1501 struct nlattr *tb[],
1502 struct net_device *dev,
1503 const unsigned char *addr,
1505 int (*ndo_fdb_dump)(struct sk_buff *skb,
1506 struct netlink_callback *cb,
1507 struct net_device *dev,
1508 struct net_device *filter_dev,
1510 int (*ndo_fdb_get)(struct sk_buff *skb,
1511 struct nlattr *tb[],
1512 struct net_device *dev,
1513 const unsigned char *addr,
1514 u16 vid, u32 portid, u32 seq,
1515 struct netlink_ext_ack *extack);
1516 int (*ndo_bridge_setlink)(struct net_device *dev,
1517 struct nlmsghdr *nlh,
1519 struct netlink_ext_ack *extack);
1520 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1522 struct net_device *dev,
1525 int (*ndo_bridge_dellink)(struct net_device *dev,
1526 struct nlmsghdr *nlh,
1528 int (*ndo_change_carrier)(struct net_device *dev,
1530 int (*ndo_get_phys_port_id)(struct net_device *dev,
1531 struct netdev_phys_item_id *ppid);
1532 int (*ndo_get_port_parent_id)(struct net_device *dev,
1533 struct netdev_phys_item_id *ppid);
1534 int (*ndo_get_phys_port_name)(struct net_device *dev,
1535 char *name, size_t len);
1536 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1537 struct net_device *dev);
1538 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1541 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1544 int (*ndo_get_iflink)(const struct net_device *dev);
1545 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1546 struct sk_buff *skb);
1547 void (*ndo_set_rx_headroom)(struct net_device *dev,
1548 int needed_headroom);
1549 int (*ndo_bpf)(struct net_device *dev,
1550 struct netdev_bpf *bpf);
1551 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1552 struct xdp_frame **xdp,
1554 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1555 struct xdp_buff *xdp);
1556 int (*ndo_xsk_wakeup)(struct net_device *dev,
1557 u32 queue_id, u32 flags);
1558 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1559 int (*ndo_tunnel_ctl)(struct net_device *dev,
1560 struct ip_tunnel_parm *p, int cmd);
1561 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1562 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1563 struct net_device_path *path);
1567 * enum netdev_priv_flags - &struct net_device priv_flags
1569 * These are the &struct net_device, they are only set internally
1570 * by drivers and used in the kernel. These flags are invisible to
1571 * userspace; this means that the order of these flags can change
1572 * during any kernel release.
1574 * You should have a pretty good reason to be extending these flags.
1576 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1577 * @IFF_EBRIDGE: Ethernet bridging device
1578 * @IFF_BONDING: bonding master or slave
1579 * @IFF_ISATAP: ISATAP interface (RFC4214)
1580 * @IFF_WAN_HDLC: WAN HDLC device
1581 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1583 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1584 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1585 * @IFF_MACVLAN_PORT: device used as macvlan port
1586 * @IFF_BRIDGE_PORT: device used as bridge port
1587 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1588 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1589 * @IFF_UNICAST_FLT: Supports unicast filtering
1590 * @IFF_TEAM_PORT: device used as team port
1591 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1592 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1593 * change when it's running
1594 * @IFF_MACVLAN: Macvlan device
1595 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1596 * underlying stacked devices
1597 * @IFF_L3MDEV_MASTER: device is an L3 master device
1598 * @IFF_NO_QUEUE: device can run without qdisc attached
1599 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1600 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1601 * @IFF_TEAM: device is a team device
1602 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1603 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1604 * entity (i.e. the master device for bridged veth)
1605 * @IFF_MACSEC: device is a MACsec device
1606 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1607 * @IFF_FAILOVER: device is a failover master device
1608 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1609 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1610 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1611 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1612 * skb_headlen(skb) == 0 (data starts from frag0)
1613 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1615 enum netdev_priv_flags {
1616 IFF_802_1Q_VLAN = 1<<0,
1620 IFF_WAN_HDLC = 1<<4,
1621 IFF_XMIT_DST_RELEASE = 1<<5,
1622 IFF_DONT_BRIDGE = 1<<6,
1623 IFF_DISABLE_NETPOLL = 1<<7,
1624 IFF_MACVLAN_PORT = 1<<8,
1625 IFF_BRIDGE_PORT = 1<<9,
1626 IFF_OVS_DATAPATH = 1<<10,
1627 IFF_TX_SKB_SHARING = 1<<11,
1628 IFF_UNICAST_FLT = 1<<12,
1629 IFF_TEAM_PORT = 1<<13,
1630 IFF_SUPP_NOFCS = 1<<14,
1631 IFF_LIVE_ADDR_CHANGE = 1<<15,
1632 IFF_MACVLAN = 1<<16,
1633 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1634 IFF_L3MDEV_MASTER = 1<<18,
1635 IFF_NO_QUEUE = 1<<19,
1636 IFF_OPENVSWITCH = 1<<20,
1637 IFF_L3MDEV_SLAVE = 1<<21,
1639 IFF_RXFH_CONFIGURED = 1<<23,
1640 IFF_PHONY_HEADROOM = 1<<24,
1642 IFF_NO_RX_HANDLER = 1<<26,
1643 IFF_FAILOVER = 1<<27,
1644 IFF_FAILOVER_SLAVE = 1<<28,
1645 IFF_L3MDEV_RX_HANDLER = 1<<29,
1646 IFF_LIVE_RENAME_OK = 1<<30,
1647 IFF_TX_SKB_NO_LINEAR = 1<<31,
1648 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1651 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1652 #define IFF_EBRIDGE IFF_EBRIDGE
1653 #define IFF_BONDING IFF_BONDING
1654 #define IFF_ISATAP IFF_ISATAP
1655 #define IFF_WAN_HDLC IFF_WAN_HDLC
1656 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1657 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1658 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1659 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1660 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1661 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1662 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1663 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1664 #define IFF_TEAM_PORT IFF_TEAM_PORT
1665 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1666 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1667 #define IFF_MACVLAN IFF_MACVLAN
1668 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1669 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1670 #define IFF_NO_QUEUE IFF_NO_QUEUE
1671 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1672 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1673 #define IFF_TEAM IFF_TEAM
1674 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1675 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1676 #define IFF_MACSEC IFF_MACSEC
1677 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1678 #define IFF_FAILOVER IFF_FAILOVER
1679 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1680 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1681 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1682 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1684 /* Specifies the type of the struct net_device::ml_priv pointer */
1685 enum netdev_ml_priv_type {
1691 * struct net_device - The DEVICE structure.
1693 * Actually, this whole structure is a big mistake. It mixes I/O
1694 * data with strictly "high-level" data, and it has to know about
1695 * almost every data structure used in the INET module.
1697 * @name: This is the first field of the "visible" part of this structure
1698 * (i.e. as seen by users in the "Space.c" file). It is the name
1701 * @name_node: Name hashlist node
1702 * @ifalias: SNMP alias
1703 * @mem_end: Shared memory end
1704 * @mem_start: Shared memory start
1705 * @base_addr: Device I/O address
1706 * @irq: Device IRQ number
1708 * @state: Generic network queuing layer state, see netdev_state_t
1709 * @dev_list: The global list of network devices
1710 * @napi_list: List entry used for polling NAPI devices
1711 * @unreg_list: List entry when we are unregistering the
1712 * device; see the function unregister_netdev
1713 * @close_list: List entry used when we are closing the device
1714 * @ptype_all: Device-specific packet handlers for all protocols
1715 * @ptype_specific: Device-specific, protocol-specific packet handlers
1717 * @adj_list: Directly linked devices, like slaves for bonding
1718 * @features: Currently active device features
1719 * @hw_features: User-changeable features
1721 * @wanted_features: User-requested features
1722 * @vlan_features: Mask of features inheritable by VLAN devices
1724 * @hw_enc_features: Mask of features inherited by encapsulating devices
1725 * This field indicates what encapsulation
1726 * offloads the hardware is capable of doing,
1727 * and drivers will need to set them appropriately.
1729 * @mpls_features: Mask of features inheritable by MPLS
1730 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1732 * @ifindex: interface index
1733 * @group: The group the device belongs to
1735 * @stats: Statistics struct, which was left as a legacy, use
1736 * rtnl_link_stats64 instead
1738 * @rx_dropped: Dropped packets by core network,
1739 * do not use this in drivers
1740 * @tx_dropped: Dropped packets by core network,
1741 * do not use this in drivers
1742 * @rx_nohandler: nohandler dropped packets by core network on
1743 * inactive devices, do not use this in drivers
1744 * @carrier_up_count: Number of times the carrier has been up
1745 * @carrier_down_count: Number of times the carrier has been down
1747 * @wireless_handlers: List of functions to handle Wireless Extensions,
1749 * see <net/iw_handler.h> for details.
1750 * @wireless_data: Instance data managed by the core of wireless extensions
1752 * @netdev_ops: Includes several pointers to callbacks,
1753 * if one wants to override the ndo_*() functions
1754 * @ethtool_ops: Management operations
1755 * @l3mdev_ops: Layer 3 master device operations
1756 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1757 * discovery handling. Necessary for e.g. 6LoWPAN.
1758 * @xfrmdev_ops: Transformation offload operations
1759 * @tlsdev_ops: Transport Layer Security offload operations
1760 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1761 * of Layer 2 headers.
1763 * @flags: Interface flags (a la BSD)
1764 * @priv_flags: Like 'flags' but invisible to userspace,
1765 * see if.h for the definitions
1766 * @gflags: Global flags ( kept as legacy )
1767 * @padded: How much padding added by alloc_netdev()
1768 * @operstate: RFC2863 operstate
1769 * @link_mode: Mapping policy to operstate
1770 * @if_port: Selectable AUI, TP, ...
1772 * @mtu: Interface MTU value
1773 * @min_mtu: Interface Minimum MTU value
1774 * @max_mtu: Interface Maximum MTU value
1775 * @type: Interface hardware type
1776 * @hard_header_len: Maximum hardware header length.
1777 * @min_header_len: Minimum hardware header length
1779 * @needed_headroom: Extra headroom the hardware may need, but not in all
1780 * cases can this be guaranteed
1781 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1782 * cases can this be guaranteed. Some cases also use
1783 * LL_MAX_HEADER instead to allocate the skb
1785 * interface address info:
1787 * @perm_addr: Permanent hw address
1788 * @addr_assign_type: Hw address assignment type
1789 * @addr_len: Hardware address length
1790 * @upper_level: Maximum depth level of upper devices.
1791 * @lower_level: Maximum depth level of lower devices.
1792 * @neigh_priv_len: Used in neigh_alloc()
1793 * @dev_id: Used to differentiate devices that share
1794 * the same link layer address
1795 * @dev_port: Used to differentiate devices that share
1797 * @addr_list_lock: XXX: need comments on this one
1798 * @name_assign_type: network interface name assignment type
1799 * @uc_promisc: Counter that indicates promiscuous mode
1800 * has been enabled due to the need to listen to
1801 * additional unicast addresses in a device that
1802 * does not implement ndo_set_rx_mode()
1803 * @uc: unicast mac addresses
1804 * @mc: multicast mac addresses
1805 * @dev_addrs: list of device hw addresses
1806 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1807 * @promiscuity: Number of times the NIC is told to work in
1808 * promiscuous mode; if it becomes 0 the NIC will
1809 * exit promiscuous mode
1810 * @allmulti: Counter, enables or disables allmulticast mode
1812 * @vlan_info: VLAN info
1813 * @dsa_ptr: dsa specific data
1814 * @tipc_ptr: TIPC specific data
1815 * @atalk_ptr: AppleTalk link
1816 * @ip_ptr: IPv4 specific data
1817 * @dn_ptr: DECnet specific data
1818 * @ip6_ptr: IPv6 specific data
1819 * @ax25_ptr: AX.25 specific data
1820 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1821 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1823 * @mpls_ptr: mpls_dev struct pointer
1824 * @mctp_ptr: MCTP specific data
1826 * @dev_addr: Hw address (before bcast,
1827 * because most packets are unicast)
1829 * @_rx: Array of RX queues
1830 * @num_rx_queues: Number of RX queues
1831 * allocated at register_netdev() time
1832 * @real_num_rx_queues: Number of RX queues currently active in device
1833 * @xdp_prog: XDP sockets filter program pointer
1834 * @gro_flush_timeout: timeout for GRO layer in NAPI
1835 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1836 * allow to avoid NIC hard IRQ, on busy queues.
1838 * @rx_handler: handler for received packets
1839 * @rx_handler_data: XXX: need comments on this one
1840 * @miniq_ingress: ingress/clsact qdisc specific data for
1841 * ingress processing
1842 * @ingress_queue: XXX: need comments on this one
1843 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1844 * @broadcast: hw bcast address
1846 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1847 * indexed by RX queue number. Assigned by driver.
1848 * This must only be set if the ndo_rx_flow_steer
1849 * operation is defined
1850 * @index_hlist: Device index hash chain
1852 * @_tx: Array of TX queues
1853 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1854 * @real_num_tx_queues: Number of TX queues currently active in device
1855 * @qdisc: Root qdisc from userspace point of view
1856 * @tx_queue_len: Max frames per queue allowed
1857 * @tx_global_lock: XXX: need comments on this one
1858 * @xdp_bulkq: XDP device bulk queue
1859 * @xps_maps: all CPUs/RXQs maps for XPS device
1861 * @xps_maps: XXX: need comments on this one
1862 * @miniq_egress: clsact qdisc specific data for
1864 * @nf_hooks_egress: netfilter hooks executed for egress packets
1865 * @qdisc_hash: qdisc hash table
1866 * @watchdog_timeo: Represents the timeout that is used by
1867 * the watchdog (see dev_watchdog())
1868 * @watchdog_timer: List of timers
1870 * @proto_down_reason: reason a netdev interface is held down
1871 * @pcpu_refcnt: Number of references to this device
1872 * @dev_refcnt: Number of references to this device
1873 * @refcnt_tracker: Tracker directory for tracked references to this device
1874 * @todo_list: Delayed register/unregister
1875 * @link_watch_list: XXX: need comments on this one
1877 * @reg_state: Register/unregister state machine
1878 * @dismantle: Device is going to be freed
1879 * @rtnl_link_state: This enum represents the phases of creating
1882 * @needs_free_netdev: Should unregister perform free_netdev?
1883 * @priv_destructor: Called from unregister
1884 * @npinfo: XXX: need comments on this one
1885 * @nd_net: Network namespace this network device is inside
1887 * @ml_priv: Mid-layer private
1888 * @ml_priv_type: Mid-layer private type
1889 * @lstats: Loopback statistics
1890 * @tstats: Tunnel statistics
1891 * @dstats: Dummy statistics
1892 * @vstats: Virtual ethernet statistics
1897 * @dev: Class/net/name entry
1898 * @sysfs_groups: Space for optional device, statistics and wireless
1901 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1902 * @rtnl_link_ops: Rtnl_link_ops
1904 * @gso_max_size: Maximum size of generic segmentation offload
1905 * @gso_max_segs: Maximum number of segments that can be passed to the
1908 * @dcbnl_ops: Data Center Bridging netlink ops
1909 * @num_tc: Number of traffic classes in the net device
1910 * @tc_to_txq: XXX: need comments on this one
1911 * @prio_tc_map: XXX: need comments on this one
1913 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1915 * @priomap: XXX: need comments on this one
1916 * @phydev: Physical device may attach itself
1917 * for hardware timestamping
1918 * @sfp_bus: attached &struct sfp_bus structure.
1920 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1922 * @proto_down: protocol port state information can be sent to the
1923 * switch driver and used to set the phys state of the
1926 * @wol_enabled: Wake-on-LAN is enabled
1928 * @threaded: napi threaded mode is enabled
1930 * @net_notifier_list: List of per-net netdev notifier block
1931 * that follow this device when it is moved
1932 * to another network namespace.
1934 * @macsec_ops: MACsec offloading ops
1936 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1937 * offload capabilities of the device
1938 * @udp_tunnel_nic: UDP tunnel offload state
1939 * @xdp_state: stores info on attached XDP BPF programs
1941 * @nested_level: Used as a parameter of spin_lock_nested() of
1942 * dev->addr_list_lock.
1943 * @unlink_list: As netif_addr_lock() can be called recursively,
1944 * keep a list of interfaces to be deleted.
1945 * @gro_max_size: Maximum size of aggregated packet in generic
1946 * receive offload (GRO)
1948 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
1949 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
1950 * @watchdog_dev_tracker: refcount tracker used by watchdog.
1952 * FIXME: cleanup struct net_device such that network protocol info
1957 char name[IFNAMSIZ];
1958 struct netdev_name_node *name_node;
1959 struct dev_ifalias __rcu *ifalias;
1961 * I/O specific fields
1962 * FIXME: Merge these and struct ifmap into one
1964 unsigned long mem_end;
1965 unsigned long mem_start;
1966 unsigned long base_addr;
1969 * Some hardware also needs these fields (state,dev_list,
1970 * napi_list,unreg_list,close_list) but they are not
1971 * part of the usual set specified in Space.c.
1974 unsigned long state;
1976 struct list_head dev_list;
1977 struct list_head napi_list;
1978 struct list_head unreg_list;
1979 struct list_head close_list;
1980 struct list_head ptype_all;
1981 struct list_head ptype_specific;
1984 struct list_head upper;
1985 struct list_head lower;
1988 /* Read-mostly cache-line for fast-path access */
1990 unsigned long long priv_flags;
1991 const struct net_device_ops *netdev_ops;
1993 unsigned short gflags;
1994 unsigned short hard_header_len;
1996 /* Note : dev->mtu is often read without holding a lock.
1997 * Writers usually hold RTNL.
1998 * It is recommended to use READ_ONCE() to annotate the reads,
1999 * and to use WRITE_ONCE() to annotate the writes.
2002 unsigned short needed_headroom;
2003 unsigned short needed_tailroom;
2005 netdev_features_t features;
2006 netdev_features_t hw_features;
2007 netdev_features_t wanted_features;
2008 netdev_features_t vlan_features;
2009 netdev_features_t hw_enc_features;
2010 netdev_features_t mpls_features;
2011 netdev_features_t gso_partial_features;
2013 unsigned int min_mtu;
2014 unsigned int max_mtu;
2015 unsigned short type;
2016 unsigned char min_header_len;
2017 unsigned char name_assign_type;
2021 struct net_device_stats stats; /* not used by modern drivers */
2023 atomic_long_t rx_dropped;
2024 atomic_long_t tx_dropped;
2025 atomic_long_t rx_nohandler;
2027 /* Stats to monitor link on/off, flapping */
2028 atomic_t carrier_up_count;
2029 atomic_t carrier_down_count;
2031 #ifdef CONFIG_WIRELESS_EXT
2032 const struct iw_handler_def *wireless_handlers;
2033 struct iw_public_data *wireless_data;
2035 const struct ethtool_ops *ethtool_ops;
2036 #ifdef CONFIG_NET_L3_MASTER_DEV
2037 const struct l3mdev_ops *l3mdev_ops;
2039 #if IS_ENABLED(CONFIG_IPV6)
2040 const struct ndisc_ops *ndisc_ops;
2043 #ifdef CONFIG_XFRM_OFFLOAD
2044 const struct xfrmdev_ops *xfrmdev_ops;
2047 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2048 const struct tlsdev_ops *tlsdev_ops;
2051 const struct header_ops *header_ops;
2053 unsigned char operstate;
2054 unsigned char link_mode;
2056 unsigned char if_port;
2059 /* Interface address info. */
2060 unsigned char perm_addr[MAX_ADDR_LEN];
2061 unsigned char addr_assign_type;
2062 unsigned char addr_len;
2063 unsigned char upper_level;
2064 unsigned char lower_level;
2066 unsigned short neigh_priv_len;
2067 unsigned short dev_id;
2068 unsigned short dev_port;
2069 unsigned short padded;
2071 spinlock_t addr_list_lock;
2074 struct netdev_hw_addr_list uc;
2075 struct netdev_hw_addr_list mc;
2076 struct netdev_hw_addr_list dev_addrs;
2079 struct kset *queues_kset;
2081 #ifdef CONFIG_LOCKDEP
2082 struct list_head unlink_list;
2084 unsigned int promiscuity;
2085 unsigned int allmulti;
2087 #ifdef CONFIG_LOCKDEP
2088 unsigned char nested_level;
2092 /* Protocol-specific pointers */
2094 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2095 struct vlan_info __rcu *vlan_info;
2097 #if IS_ENABLED(CONFIG_NET_DSA)
2098 struct dsa_port *dsa_ptr;
2100 #if IS_ENABLED(CONFIG_TIPC)
2101 struct tipc_bearer __rcu *tipc_ptr;
2103 #if IS_ENABLED(CONFIG_ATALK)
2106 struct in_device __rcu *ip_ptr;
2107 #if IS_ENABLED(CONFIG_DECNET)
2108 struct dn_dev __rcu *dn_ptr;
2110 struct inet6_dev __rcu *ip6_ptr;
2111 #if IS_ENABLED(CONFIG_AX25)
2114 struct wireless_dev *ieee80211_ptr;
2115 struct wpan_dev *ieee802154_ptr;
2116 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2117 struct mpls_dev __rcu *mpls_ptr;
2119 #if IS_ENABLED(CONFIG_MCTP)
2120 struct mctp_dev __rcu *mctp_ptr;
2124 * Cache lines mostly used on receive path (including eth_type_trans())
2126 /* Interface address info used in eth_type_trans() */
2127 const unsigned char *dev_addr;
2129 struct netdev_rx_queue *_rx;
2130 unsigned int num_rx_queues;
2131 unsigned int real_num_rx_queues;
2133 struct bpf_prog __rcu *xdp_prog;
2134 unsigned long gro_flush_timeout;
2135 int napi_defer_hard_irqs;
2136 #define GRO_MAX_SIZE 65536
2137 unsigned int gro_max_size;
2138 rx_handler_func_t __rcu *rx_handler;
2139 void __rcu *rx_handler_data;
2141 #ifdef CONFIG_NET_CLS_ACT
2142 struct mini_Qdisc __rcu *miniq_ingress;
2144 struct netdev_queue __rcu *ingress_queue;
2145 #ifdef CONFIG_NETFILTER_INGRESS
2146 struct nf_hook_entries __rcu *nf_hooks_ingress;
2149 unsigned char broadcast[MAX_ADDR_LEN];
2150 #ifdef CONFIG_RFS_ACCEL
2151 struct cpu_rmap *rx_cpu_rmap;
2153 struct hlist_node index_hlist;
2156 * Cache lines mostly used on transmit path
2158 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2159 unsigned int num_tx_queues;
2160 unsigned int real_num_tx_queues;
2161 struct Qdisc *qdisc;
2162 unsigned int tx_queue_len;
2163 spinlock_t tx_global_lock;
2165 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2168 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2170 #ifdef CONFIG_NET_CLS_ACT
2171 struct mini_Qdisc __rcu *miniq_egress;
2173 #ifdef CONFIG_NETFILTER_EGRESS
2174 struct nf_hook_entries __rcu *nf_hooks_egress;
2177 #ifdef CONFIG_NET_SCHED
2178 DECLARE_HASHTABLE (qdisc_hash, 4);
2180 /* These may be needed for future network-power-down code. */
2181 struct timer_list watchdog_timer;
2184 u32 proto_down_reason;
2186 struct list_head todo_list;
2188 #ifdef CONFIG_PCPU_DEV_REFCNT
2189 int __percpu *pcpu_refcnt;
2191 refcount_t dev_refcnt;
2193 struct ref_tracker_dir refcnt_tracker;
2195 struct list_head link_watch_list;
2197 enum { NETREG_UNINITIALIZED=0,
2198 NETREG_REGISTERED, /* completed register_netdevice */
2199 NETREG_UNREGISTERING, /* called unregister_netdevice */
2200 NETREG_UNREGISTERED, /* completed unregister todo */
2201 NETREG_RELEASED, /* called free_netdev */
2202 NETREG_DUMMY, /* dummy device for NAPI poll */
2208 RTNL_LINK_INITIALIZED,
2209 RTNL_LINK_INITIALIZING,
2210 } rtnl_link_state:16;
2212 bool needs_free_netdev;
2213 void (*priv_destructor)(struct net_device *dev);
2215 #ifdef CONFIG_NETPOLL
2216 struct netpoll_info __rcu *npinfo;
2219 possible_net_t nd_net;
2221 /* mid-layer private */
2223 enum netdev_ml_priv_type ml_priv_type;
2226 struct pcpu_lstats __percpu *lstats;
2227 struct pcpu_sw_netstats __percpu *tstats;
2228 struct pcpu_dstats __percpu *dstats;
2231 #if IS_ENABLED(CONFIG_GARP)
2232 struct garp_port __rcu *garp_port;
2234 #if IS_ENABLED(CONFIG_MRP)
2235 struct mrp_port __rcu *mrp_port;
2239 const struct attribute_group *sysfs_groups[4];
2240 const struct attribute_group *sysfs_rx_queue_group;
2242 const struct rtnl_link_ops *rtnl_link_ops;
2244 /* for setting kernel sock attribute on TCP connection setup */
2245 #define GSO_MAX_SIZE 65536
2246 unsigned int gso_max_size;
2247 #define GSO_MAX_SEGS 65535
2251 const struct dcbnl_rtnl_ops *dcbnl_ops;
2254 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2255 u8 prio_tc_map[TC_BITMASK + 1];
2257 #if IS_ENABLED(CONFIG_FCOE)
2258 unsigned int fcoe_ddp_xid;
2260 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2261 struct netprio_map __rcu *priomap;
2263 struct phy_device *phydev;
2264 struct sfp_bus *sfp_bus;
2265 struct lock_class_key *qdisc_tx_busylock;
2267 unsigned wol_enabled:1;
2268 unsigned threaded:1;
2270 struct list_head net_notifier_list;
2272 #if IS_ENABLED(CONFIG_MACSEC)
2273 /* MACsec management functions */
2274 const struct macsec_ops *macsec_ops;
2276 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2277 struct udp_tunnel_nic *udp_tunnel_nic;
2279 /* protected by rtnl_lock */
2280 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2282 u8 dev_addr_shadow[MAX_ADDR_LEN];
2283 netdevice_tracker linkwatch_dev_tracker;
2284 netdevice_tracker watchdog_dev_tracker;
2286 #define to_net_dev(d) container_of(d, struct net_device, dev)
2288 static inline bool netif_elide_gro(const struct net_device *dev)
2290 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2295 #define NETDEV_ALIGN 32
2298 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2300 return dev->prio_tc_map[prio & TC_BITMASK];
2304 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2306 if (tc >= dev->num_tc)
2309 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2313 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2314 void netdev_reset_tc(struct net_device *dev);
2315 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2316 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2319 int netdev_get_num_tc(struct net_device *dev)
2324 static inline void net_prefetch(void *p)
2327 #if L1_CACHE_BYTES < 128
2328 prefetch((u8 *)p + L1_CACHE_BYTES);
2332 static inline void net_prefetchw(void *p)
2335 #if L1_CACHE_BYTES < 128
2336 prefetchw((u8 *)p + L1_CACHE_BYTES);
2340 void netdev_unbind_sb_channel(struct net_device *dev,
2341 struct net_device *sb_dev);
2342 int netdev_bind_sb_channel_queue(struct net_device *dev,
2343 struct net_device *sb_dev,
2344 u8 tc, u16 count, u16 offset);
2345 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2346 static inline int netdev_get_sb_channel(struct net_device *dev)
2348 return max_t(int, -dev->num_tc, 0);
2352 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2355 return &dev->_tx[index];
2358 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2359 const struct sk_buff *skb)
2361 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2364 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2365 void (*f)(struct net_device *,
2366 struct netdev_queue *,
2372 for (i = 0; i < dev->num_tx_queues; i++)
2373 f(dev, &dev->_tx[i], arg);
2376 #define netdev_lockdep_set_classes(dev) \
2378 static struct lock_class_key qdisc_tx_busylock_key; \
2379 static struct lock_class_key qdisc_xmit_lock_key; \
2380 static struct lock_class_key dev_addr_list_lock_key; \
2383 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2384 lockdep_set_class(&(dev)->addr_list_lock, \
2385 &dev_addr_list_lock_key); \
2386 for (i = 0; i < (dev)->num_tx_queues; i++) \
2387 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2388 &qdisc_xmit_lock_key); \
2391 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2392 struct net_device *sb_dev);
2393 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2394 struct sk_buff *skb,
2395 struct net_device *sb_dev);
2397 /* returns the headroom that the master device needs to take in account
2398 * when forwarding to this dev
2400 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2402 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2405 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2407 if (dev->netdev_ops->ndo_set_rx_headroom)
2408 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2411 /* set the device rx headroom to the dev's default */
2412 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2414 netdev_set_rx_headroom(dev, -1);
2417 static inline void *netdev_get_ml_priv(struct net_device *dev,
2418 enum netdev_ml_priv_type type)
2420 if (dev->ml_priv_type != type)
2423 return dev->ml_priv;
2426 static inline void netdev_set_ml_priv(struct net_device *dev,
2428 enum netdev_ml_priv_type type)
2430 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2431 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2432 dev->ml_priv_type, type);
2433 WARN(!dev->ml_priv_type && dev->ml_priv,
2434 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2436 dev->ml_priv = ml_priv;
2437 dev->ml_priv_type = type;
2441 * Net namespace inlines
2444 struct net *dev_net(const struct net_device *dev)
2446 return read_pnet(&dev->nd_net);
2450 void dev_net_set(struct net_device *dev, struct net *net)
2452 write_pnet(&dev->nd_net, net);
2456 * netdev_priv - access network device private data
2457 * @dev: network device
2459 * Get network device private data
2461 static inline void *netdev_priv(const struct net_device *dev)
2463 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2466 /* Set the sysfs physical device reference for the network logical device
2467 * if set prior to registration will cause a symlink during initialization.
2469 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2471 /* Set the sysfs device type for the network logical device to allow
2472 * fine-grained identification of different network device types. For
2473 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2475 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2477 /* Default NAPI poll() weight
2478 * Device drivers are strongly advised to not use bigger value
2480 #define NAPI_POLL_WEIGHT 64
2483 * netif_napi_add - initialize a NAPI context
2484 * @dev: network device
2485 * @napi: NAPI context
2486 * @poll: polling function
2487 * @weight: default weight
2489 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2490 * *any* of the other NAPI-related functions.
2492 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2493 int (*poll)(struct napi_struct *, int), int weight);
2496 * netif_tx_napi_add - initialize a NAPI context
2497 * @dev: network device
2498 * @napi: NAPI context
2499 * @poll: polling function
2500 * @weight: default weight
2502 * This variant of netif_napi_add() should be used from drivers using NAPI
2503 * to exclusively poll a TX queue.
2504 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2506 static inline void netif_tx_napi_add(struct net_device *dev,
2507 struct napi_struct *napi,
2508 int (*poll)(struct napi_struct *, int),
2511 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2512 netif_napi_add(dev, napi, poll, weight);
2516 * __netif_napi_del - remove a NAPI context
2517 * @napi: NAPI context
2519 * Warning: caller must observe RCU grace period before freeing memory
2520 * containing @napi. Drivers might want to call this helper to combine
2521 * all the needed RCU grace periods into a single one.
2523 void __netif_napi_del(struct napi_struct *napi);
2526 * netif_napi_del - remove a NAPI context
2527 * @napi: NAPI context
2529 * netif_napi_del() removes a NAPI context from the network device NAPI list
2531 static inline void netif_napi_del(struct napi_struct *napi)
2533 __netif_napi_del(napi);
2537 struct packet_type {
2538 __be16 type; /* This is really htons(ether_type). */
2539 bool ignore_outgoing;
2540 struct net_device *dev; /* NULL is wildcarded here */
2541 netdevice_tracker dev_tracker;
2542 int (*func) (struct sk_buff *,
2543 struct net_device *,
2544 struct packet_type *,
2545 struct net_device *);
2546 void (*list_func) (struct list_head *,
2547 struct packet_type *,
2548 struct net_device *);
2549 bool (*id_match)(struct packet_type *ptype,
2551 struct net *af_packet_net;
2552 void *af_packet_priv;
2553 struct list_head list;
2556 struct offload_callbacks {
2557 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2558 netdev_features_t features);
2559 struct sk_buff *(*gro_receive)(struct list_head *head,
2560 struct sk_buff *skb);
2561 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2564 struct packet_offload {
2565 __be16 type; /* This is really htons(ether_type). */
2567 struct offload_callbacks callbacks;
2568 struct list_head list;
2571 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2572 struct pcpu_sw_netstats {
2577 struct u64_stats_sync syncp;
2578 } __aligned(4 * sizeof(u64));
2580 struct pcpu_lstats {
2581 u64_stats_t packets;
2583 struct u64_stats_sync syncp;
2584 } __aligned(2 * sizeof(u64));
2586 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2588 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2590 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2592 u64_stats_update_begin(&tstats->syncp);
2593 tstats->rx_bytes += len;
2594 tstats->rx_packets++;
2595 u64_stats_update_end(&tstats->syncp);
2598 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2599 unsigned int packets,
2602 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2604 u64_stats_update_begin(&tstats->syncp);
2605 tstats->tx_bytes += len;
2606 tstats->tx_packets += packets;
2607 u64_stats_update_end(&tstats->syncp);
2610 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2612 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2614 u64_stats_update_begin(&lstats->syncp);
2615 u64_stats_add(&lstats->bytes, len);
2616 u64_stats_inc(&lstats->packets);
2617 u64_stats_update_end(&lstats->syncp);
2620 #define __netdev_alloc_pcpu_stats(type, gfp) \
2622 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2625 for_each_possible_cpu(__cpu) { \
2626 typeof(type) *stat; \
2627 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2628 u64_stats_init(&stat->syncp); \
2634 #define netdev_alloc_pcpu_stats(type) \
2635 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2637 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2639 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2642 for_each_possible_cpu(__cpu) { \
2643 typeof(type) *stat; \
2644 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2645 u64_stats_init(&stat->syncp); \
2651 enum netdev_lag_tx_type {
2652 NETDEV_LAG_TX_TYPE_UNKNOWN,
2653 NETDEV_LAG_TX_TYPE_RANDOM,
2654 NETDEV_LAG_TX_TYPE_BROADCAST,
2655 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2656 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2657 NETDEV_LAG_TX_TYPE_HASH,
2660 enum netdev_lag_hash {
2661 NETDEV_LAG_HASH_NONE,
2663 NETDEV_LAG_HASH_L34,
2664 NETDEV_LAG_HASH_L23,
2665 NETDEV_LAG_HASH_E23,
2666 NETDEV_LAG_HASH_E34,
2667 NETDEV_LAG_HASH_VLAN_SRCMAC,
2668 NETDEV_LAG_HASH_UNKNOWN,
2671 struct netdev_lag_upper_info {
2672 enum netdev_lag_tx_type tx_type;
2673 enum netdev_lag_hash hash_type;
2676 struct netdev_lag_lower_state_info {
2681 #include <linux/notifier.h>
2683 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2684 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2688 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2690 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2691 detected a hardware crash and restarted
2692 - we can use this eg to kick tcp sessions
2694 NETDEV_CHANGE, /* Notify device state change */
2697 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2698 NETDEV_CHANGEADDR, /* notify after the address change */
2699 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2703 NETDEV_BONDING_FAILOVER,
2705 NETDEV_PRE_TYPE_CHANGE,
2706 NETDEV_POST_TYPE_CHANGE,
2709 NETDEV_NOTIFY_PEERS,
2713 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2714 NETDEV_CHANGEINFODATA,
2715 NETDEV_BONDING_INFO,
2716 NETDEV_PRECHANGEUPPER,
2717 NETDEV_CHANGELOWERSTATE,
2718 NETDEV_UDP_TUNNEL_PUSH_INFO,
2719 NETDEV_UDP_TUNNEL_DROP_INFO,
2720 NETDEV_CHANGE_TX_QUEUE_LEN,
2721 NETDEV_CVLAN_FILTER_PUSH_INFO,
2722 NETDEV_CVLAN_FILTER_DROP_INFO,
2723 NETDEV_SVLAN_FILTER_PUSH_INFO,
2724 NETDEV_SVLAN_FILTER_DROP_INFO,
2726 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2728 int register_netdevice_notifier(struct notifier_block *nb);
2729 int unregister_netdevice_notifier(struct notifier_block *nb);
2730 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2731 int unregister_netdevice_notifier_net(struct net *net,
2732 struct notifier_block *nb);
2733 int register_netdevice_notifier_dev_net(struct net_device *dev,
2734 struct notifier_block *nb,
2735 struct netdev_net_notifier *nn);
2736 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2737 struct notifier_block *nb,
2738 struct netdev_net_notifier *nn);
2740 struct netdev_notifier_info {
2741 struct net_device *dev;
2742 struct netlink_ext_ack *extack;
2745 struct netdev_notifier_info_ext {
2746 struct netdev_notifier_info info; /* must be first */
2752 struct netdev_notifier_change_info {
2753 struct netdev_notifier_info info; /* must be first */
2754 unsigned int flags_changed;
2757 struct netdev_notifier_changeupper_info {
2758 struct netdev_notifier_info info; /* must be first */
2759 struct net_device *upper_dev; /* new upper dev */
2760 bool master; /* is upper dev master */
2761 bool linking; /* is the notification for link or unlink */
2762 void *upper_info; /* upper dev info */
2765 struct netdev_notifier_changelowerstate_info {
2766 struct netdev_notifier_info info; /* must be first */
2767 void *lower_state_info; /* is lower dev state */
2770 struct netdev_notifier_pre_changeaddr_info {
2771 struct netdev_notifier_info info; /* must be first */
2772 const unsigned char *dev_addr;
2775 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2776 struct net_device *dev)
2779 info->extack = NULL;
2782 static inline struct net_device *
2783 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2788 static inline struct netlink_ext_ack *
2789 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2791 return info->extack;
2794 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2797 extern rwlock_t dev_base_lock; /* Device list lock */
2799 #define for_each_netdev(net, d) \
2800 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2801 #define for_each_netdev_reverse(net, d) \
2802 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2803 #define for_each_netdev_rcu(net, d) \
2804 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2805 #define for_each_netdev_safe(net, d, n) \
2806 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2807 #define for_each_netdev_continue(net, d) \
2808 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2809 #define for_each_netdev_continue_reverse(net, d) \
2810 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2812 #define for_each_netdev_continue_rcu(net, d) \
2813 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2814 #define for_each_netdev_in_bond_rcu(bond, slave) \
2815 for_each_netdev_rcu(&init_net, slave) \
2816 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2817 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2819 static inline struct net_device *next_net_device(struct net_device *dev)
2821 struct list_head *lh;
2825 lh = dev->dev_list.next;
2826 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2829 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2831 struct list_head *lh;
2835 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2836 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2839 static inline struct net_device *first_net_device(struct net *net)
2841 return list_empty(&net->dev_base_head) ? NULL :
2842 net_device_entry(net->dev_base_head.next);
2845 static inline struct net_device *first_net_device_rcu(struct net *net)
2847 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2849 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2852 int netdev_boot_setup_check(struct net_device *dev);
2853 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2854 const char *hwaddr);
2855 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2856 void dev_add_pack(struct packet_type *pt);
2857 void dev_remove_pack(struct packet_type *pt);
2858 void __dev_remove_pack(struct packet_type *pt);
2859 void dev_add_offload(struct packet_offload *po);
2860 void dev_remove_offload(struct packet_offload *po);
2862 int dev_get_iflink(const struct net_device *dev);
2863 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2864 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2865 struct net_device_path_stack *stack);
2866 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2867 unsigned short mask);
2868 struct net_device *dev_get_by_name(struct net *net, const char *name);
2869 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2870 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2871 bool netdev_name_in_use(struct net *net, const char *name);
2872 int dev_alloc_name(struct net_device *dev, const char *name);
2873 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2874 void dev_close(struct net_device *dev);
2875 void dev_close_many(struct list_head *head, bool unlink);
2876 void dev_disable_lro(struct net_device *dev);
2877 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2878 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2879 struct net_device *sb_dev);
2880 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2881 struct net_device *sb_dev);
2883 int dev_queue_xmit(struct sk_buff *skb);
2884 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2885 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2887 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2891 ret = __dev_direct_xmit(skb, queue_id);
2892 if (!dev_xmit_complete(ret))
2897 int register_netdevice(struct net_device *dev);
2898 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2899 void unregister_netdevice_many(struct list_head *head);
2900 static inline void unregister_netdevice(struct net_device *dev)
2902 unregister_netdevice_queue(dev, NULL);
2905 int netdev_refcnt_read(const struct net_device *dev);
2906 void free_netdev(struct net_device *dev);
2907 void netdev_freemem(struct net_device *dev);
2908 int init_dummy_netdev(struct net_device *dev);
2910 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2911 struct sk_buff *skb,
2913 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2915 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2916 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2917 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2918 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2919 int netdev_get_name(struct net *net, char *name, int ifindex);
2920 int dev_restart(struct net_device *dev);
2923 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2924 unsigned short type,
2925 const void *daddr, const void *saddr,
2928 if (!dev->header_ops || !dev->header_ops->create)
2931 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2934 static inline int dev_parse_header(const struct sk_buff *skb,
2935 unsigned char *haddr)
2937 const struct net_device *dev = skb->dev;
2939 if (!dev->header_ops || !dev->header_ops->parse)
2941 return dev->header_ops->parse(skb, haddr);
2944 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2946 const struct net_device *dev = skb->dev;
2948 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2950 return dev->header_ops->parse_protocol(skb);
2953 /* ll_header must have at least hard_header_len allocated */
2954 static inline bool dev_validate_header(const struct net_device *dev,
2955 char *ll_header, int len)
2957 if (likely(len >= dev->hard_header_len))
2959 if (len < dev->min_header_len)
2962 if (capable(CAP_SYS_RAWIO)) {
2963 memset(ll_header + len, 0, dev->hard_header_len - len);
2967 if (dev->header_ops && dev->header_ops->validate)
2968 return dev->header_ops->validate(ll_header, len);
2973 static inline bool dev_has_header(const struct net_device *dev)
2975 return dev->header_ops && dev->header_ops->create;
2978 #ifdef CONFIG_NET_FLOW_LIMIT
2979 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2980 struct sd_flow_limit {
2982 unsigned int num_buckets;
2983 unsigned int history_head;
2984 u16 history[FLOW_LIMIT_HISTORY];
2988 extern int netdev_flow_limit_table_len;
2989 #endif /* CONFIG_NET_FLOW_LIMIT */
2992 * Incoming packets are placed on per-CPU queues
2994 struct softnet_data {
2995 struct list_head poll_list;
2996 struct sk_buff_head process_queue;
2999 unsigned int processed;
3000 unsigned int time_squeeze;
3001 unsigned int received_rps;
3003 struct softnet_data *rps_ipi_list;
3005 #ifdef CONFIG_NET_FLOW_LIMIT
3006 struct sd_flow_limit __rcu *flow_limit;
3008 struct Qdisc *output_queue;
3009 struct Qdisc **output_queue_tailp;
3010 struct sk_buff *completion_queue;
3011 #ifdef CONFIG_XFRM_OFFLOAD
3012 struct sk_buff_head xfrm_backlog;
3014 /* written and read only by owning cpu: */
3020 /* input_queue_head should be written by cpu owning this struct,
3021 * and only read by other cpus. Worth using a cache line.
3023 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3025 /* Elements below can be accessed between CPUs for RPS/RFS */
3026 call_single_data_t csd ____cacheline_aligned_in_smp;
3027 struct softnet_data *rps_ipi_next;
3029 unsigned int input_queue_tail;
3031 unsigned int dropped;
3032 struct sk_buff_head input_pkt_queue;
3033 struct napi_struct backlog;
3037 static inline void input_queue_head_incr(struct softnet_data *sd)
3040 sd->input_queue_head++;
3044 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3045 unsigned int *qtail)
3048 *qtail = ++sd->input_queue_tail;
3052 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3054 static inline int dev_recursion_level(void)
3056 return this_cpu_read(softnet_data.xmit.recursion);
3059 #define XMIT_RECURSION_LIMIT 8
3060 static inline bool dev_xmit_recursion(void)
3062 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3063 XMIT_RECURSION_LIMIT);
3066 static inline void dev_xmit_recursion_inc(void)
3068 __this_cpu_inc(softnet_data.xmit.recursion);
3071 static inline void dev_xmit_recursion_dec(void)
3073 __this_cpu_dec(softnet_data.xmit.recursion);
3076 void __netif_schedule(struct Qdisc *q);
3077 void netif_schedule_queue(struct netdev_queue *txq);
3079 static inline void netif_tx_schedule_all(struct net_device *dev)
3083 for (i = 0; i < dev->num_tx_queues; i++)
3084 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3087 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3089 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3093 * netif_start_queue - allow transmit
3094 * @dev: network device
3096 * Allow upper layers to call the device hard_start_xmit routine.
3098 static inline void netif_start_queue(struct net_device *dev)
3100 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3103 static inline void netif_tx_start_all_queues(struct net_device *dev)
3107 for (i = 0; i < dev->num_tx_queues; i++) {
3108 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3109 netif_tx_start_queue(txq);
3113 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3116 * netif_wake_queue - restart transmit
3117 * @dev: network device
3119 * Allow upper layers to call the device hard_start_xmit routine.
3120 * Used for flow control when transmit resources are available.
3122 static inline void netif_wake_queue(struct net_device *dev)
3124 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3127 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3131 for (i = 0; i < dev->num_tx_queues; i++) {
3132 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3133 netif_tx_wake_queue(txq);
3137 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3139 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3143 * netif_stop_queue - stop transmitted packets
3144 * @dev: network device
3146 * Stop upper layers calling the device hard_start_xmit routine.
3147 * Used for flow control when transmit resources are unavailable.
3149 static inline void netif_stop_queue(struct net_device *dev)
3151 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3154 void netif_tx_stop_all_queues(struct net_device *dev);
3156 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3158 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3162 * netif_queue_stopped - test if transmit queue is flowblocked
3163 * @dev: network device
3165 * Test if transmit queue on device is currently unable to send.
3167 static inline bool netif_queue_stopped(const struct net_device *dev)
3169 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3172 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3174 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3178 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3180 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3184 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3186 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3190 * netdev_queue_set_dql_min_limit - set dql minimum limit
3191 * @dev_queue: pointer to transmit queue
3192 * @min_limit: dql minimum limit
3194 * Forces xmit_more() to return true until the minimum threshold
3195 * defined by @min_limit is reached (or until the tx queue is
3196 * empty). Warning: to be use with care, misuse will impact the
3199 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3200 unsigned int min_limit)
3203 dev_queue->dql.min_limit = min_limit;
3208 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3209 * @dev_queue: pointer to transmit queue
3211 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3212 * to give appropriate hint to the CPU.
3214 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3217 prefetchw(&dev_queue->dql.num_queued);
3222 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3223 * @dev_queue: pointer to transmit queue
3225 * BQL enabled drivers might use this helper in their TX completion path,
3226 * to give appropriate hint to the CPU.
3228 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3231 prefetchw(&dev_queue->dql.limit);
3235 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3239 dql_queued(&dev_queue->dql, bytes);
3241 if (likely(dql_avail(&dev_queue->dql) >= 0))
3244 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3247 * The XOFF flag must be set before checking the dql_avail below,
3248 * because in netdev_tx_completed_queue we update the dql_completed
3249 * before checking the XOFF flag.
3253 /* check again in case another CPU has just made room avail */
3254 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3255 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3259 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3260 * that they should not test BQL status themselves.
3261 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3263 * Returns true if the doorbell must be used to kick the NIC.
3265 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3271 dql_queued(&dev_queue->dql, bytes);
3273 return netif_tx_queue_stopped(dev_queue);
3275 netdev_tx_sent_queue(dev_queue, bytes);
3280 * netdev_sent_queue - report the number of bytes queued to hardware
3281 * @dev: network device
3282 * @bytes: number of bytes queued to the hardware device queue
3284 * Report the number of bytes queued for sending/completion to the network
3285 * device hardware queue. @bytes should be a good approximation and should
3286 * exactly match netdev_completed_queue() @bytes
3288 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3290 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3293 static inline bool __netdev_sent_queue(struct net_device *dev,
3297 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3301 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3302 unsigned int pkts, unsigned int bytes)
3305 if (unlikely(!bytes))
3308 dql_completed(&dev_queue->dql, bytes);
3311 * Without the memory barrier there is a small possiblity that
3312 * netdev_tx_sent_queue will miss the update and cause the queue to
3313 * be stopped forever
3317 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3320 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3321 netif_schedule_queue(dev_queue);
3326 * netdev_completed_queue - report bytes and packets completed by device
3327 * @dev: network device
3328 * @pkts: actual number of packets sent over the medium
3329 * @bytes: actual number of bytes sent over the medium
3331 * Report the number of bytes and packets transmitted by the network device
3332 * hardware queue over the physical medium, @bytes must exactly match the
3333 * @bytes amount passed to netdev_sent_queue()
3335 static inline void netdev_completed_queue(struct net_device *dev,
3336 unsigned int pkts, unsigned int bytes)
3338 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3341 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3344 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3350 * netdev_reset_queue - reset the packets and bytes count of a network device
3351 * @dev_queue: network device
3353 * Reset the bytes and packet count of a network device and clear the
3354 * software flow control OFF bit for this network device
3356 static inline void netdev_reset_queue(struct net_device *dev_queue)
3358 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3362 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3363 * @dev: network device
3364 * @queue_index: given tx queue index
3366 * Returns 0 if given tx queue index >= number of device tx queues,
3367 * otherwise returns the originally passed tx queue index.
3369 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3371 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3372 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3373 dev->name, queue_index,
3374 dev->real_num_tx_queues);
3382 * netif_running - test if up
3383 * @dev: network device
3385 * Test if the device has been brought up.
3387 static inline bool netif_running(const struct net_device *dev)
3389 return test_bit(__LINK_STATE_START, &dev->state);
3393 * Routines to manage the subqueues on a device. We only need start,
3394 * stop, and a check if it's stopped. All other device management is
3395 * done at the overall netdevice level.
3396 * Also test the device if we're multiqueue.
3400 * netif_start_subqueue - allow sending packets on subqueue
3401 * @dev: network device
3402 * @queue_index: sub queue index
3404 * Start individual transmit queue of a device with multiple transmit queues.
3406 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3408 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3410 netif_tx_start_queue(txq);
3414 * netif_stop_subqueue - stop sending packets on subqueue
3415 * @dev: network device
3416 * @queue_index: sub queue index
3418 * Stop individual transmit queue of a device with multiple transmit queues.
3420 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3422 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3423 netif_tx_stop_queue(txq);
3427 * __netif_subqueue_stopped - test status of subqueue
3428 * @dev: network device
3429 * @queue_index: sub queue index
3431 * Check individual transmit queue of a device with multiple transmit queues.
3433 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3436 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3438 return netif_tx_queue_stopped(txq);
3442 * netif_subqueue_stopped - test status of subqueue
3443 * @dev: network device
3444 * @skb: sub queue buffer pointer
3446 * Check individual transmit queue of a device with multiple transmit queues.
3448 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3449 struct sk_buff *skb)
3451 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3455 * netif_wake_subqueue - allow sending packets on subqueue
3456 * @dev: network device
3457 * @queue_index: sub queue index
3459 * Resume individual transmit queue of a device with multiple transmit queues.
3461 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3463 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3465 netif_tx_wake_queue(txq);
3469 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3471 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3472 u16 index, enum xps_map_type type);
3475 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3476 * @j: CPU/Rx queue index
3477 * @mask: bitmask of all cpus/rx queues
3478 * @nr_bits: number of bits in the bitmask
3480 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3482 static inline bool netif_attr_test_mask(unsigned long j,
3483 const unsigned long *mask,
3484 unsigned int nr_bits)
3486 cpu_max_bits_warn(j, nr_bits);
3487 return test_bit(j, mask);
3491 * netif_attr_test_online - Test for online CPU/Rx queue
3492 * @j: CPU/Rx queue index
3493 * @online_mask: bitmask for CPUs/Rx queues that are online
3494 * @nr_bits: number of bits in the bitmask
3496 * Returns true if a CPU/Rx queue is online.
3498 static inline bool netif_attr_test_online(unsigned long j,
3499 const unsigned long *online_mask,
3500 unsigned int nr_bits)
3502 cpu_max_bits_warn(j, nr_bits);
3505 return test_bit(j, online_mask);
3507 return (j < nr_bits);
3511 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3512 * @n: CPU/Rx queue index
3513 * @srcp: the cpumask/Rx queue mask pointer
3514 * @nr_bits: number of bits in the bitmask
3516 * Returns >= nr_bits if no further CPUs/Rx queues set.
3518 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3519 unsigned int nr_bits)
3521 /* -1 is a legal arg here. */
3523 cpu_max_bits_warn(n, nr_bits);
3526 return find_next_bit(srcp, nr_bits, n + 1);
3532 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3533 * @n: CPU/Rx queue index
3534 * @src1p: the first CPUs/Rx queues mask pointer
3535 * @src2p: the second CPUs/Rx queues mask pointer
3536 * @nr_bits: number of bits in the bitmask
3538 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3540 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3541 const unsigned long *src2p,
3542 unsigned int nr_bits)
3544 /* -1 is a legal arg here. */
3546 cpu_max_bits_warn(n, nr_bits);
3549 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3551 return find_next_bit(src1p, nr_bits, n + 1);
3553 return find_next_bit(src2p, nr_bits, n + 1);
3558 static inline int netif_set_xps_queue(struct net_device *dev,
3559 const struct cpumask *mask,
3565 static inline int __netif_set_xps_queue(struct net_device *dev,
3566 const unsigned long *mask,
3567 u16 index, enum xps_map_type type)
3574 * netif_is_multiqueue - test if device has multiple transmit queues
3575 * @dev: network device
3577 * Check if device has multiple transmit queues
3579 static inline bool netif_is_multiqueue(const struct net_device *dev)
3581 return dev->num_tx_queues > 1;
3584 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3587 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3589 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3592 dev->real_num_rx_queues = rxqs;
3596 int netif_set_real_num_queues(struct net_device *dev,
3597 unsigned int txq, unsigned int rxq);
3599 static inline struct netdev_rx_queue *
3600 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3602 return dev->_rx + rxq;
3606 static inline unsigned int get_netdev_rx_queue_index(
3607 struct netdev_rx_queue *queue)
3609 struct net_device *dev = queue->dev;
3610 int index = queue - dev->_rx;
3612 BUG_ON(index >= dev->num_rx_queues);
3617 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3618 int netif_get_num_default_rss_queues(void);
3620 enum skb_free_reason {
3621 SKB_REASON_CONSUMED,
3625 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3626 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3629 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3630 * interrupt context or with hardware interrupts being disabled.
3631 * (in_hardirq() || irqs_disabled())
3633 * We provide four helpers that can be used in following contexts :
3635 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3636 * replacing kfree_skb(skb)
3638 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3639 * Typically used in place of consume_skb(skb) in TX completion path
3641 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3642 * replacing kfree_skb(skb)
3644 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3645 * and consumed a packet. Used in place of consume_skb(skb)
3647 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3649 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3652 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3654 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3657 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3659 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3662 static inline void dev_consume_skb_any(struct sk_buff *skb)
3664 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3667 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3668 struct bpf_prog *xdp_prog);
3669 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3670 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3671 int netif_rx(struct sk_buff *skb);
3672 int netif_rx_ni(struct sk_buff *skb);
3673 int netif_rx_any_context(struct sk_buff *skb);
3674 int netif_receive_skb(struct sk_buff *skb);
3675 int netif_receive_skb_core(struct sk_buff *skb);
3676 void netif_receive_skb_list_internal(struct list_head *head);
3677 void netif_receive_skb_list(struct list_head *head);
3678 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3679 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3680 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3681 gro_result_t napi_gro_frags(struct napi_struct *napi);
3682 struct packet_offload *gro_find_receive_by_type(__be16 type);
3683 struct packet_offload *gro_find_complete_by_type(__be16 type);
3685 static inline void napi_free_frags(struct napi_struct *napi)
3687 kfree_skb(napi->skb);
3691 bool netdev_is_rx_handler_busy(struct net_device *dev);
3692 int netdev_rx_handler_register(struct net_device *dev,
3693 rx_handler_func_t *rx_handler,
3694 void *rx_handler_data);
3695 void netdev_rx_handler_unregister(struct net_device *dev);
3697 bool dev_valid_name(const char *name);
3698 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3700 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3702 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3703 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3704 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3705 void __user *data, bool *need_copyout);
3706 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3707 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3708 unsigned int dev_get_flags(const struct net_device *);
3709 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3710 struct netlink_ext_ack *extack);
3711 int dev_change_flags(struct net_device *dev, unsigned int flags,
3712 struct netlink_ext_ack *extack);
3713 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3714 unsigned int gchanges);
3715 int dev_change_name(struct net_device *, const char *);
3716 int dev_set_alias(struct net_device *, const char *, size_t);
3717 int dev_get_alias(const struct net_device *, char *, size_t);
3718 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3719 const char *pat, int new_ifindex);
3721 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3724 return __dev_change_net_namespace(dev, net, pat, 0);
3726 int __dev_set_mtu(struct net_device *, int);
3727 int dev_validate_mtu(struct net_device *dev, int mtu,
3728 struct netlink_ext_ack *extack);
3729 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3730 struct netlink_ext_ack *extack);
3731 int dev_set_mtu(struct net_device *, int);
3732 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3733 void dev_set_group(struct net_device *, int);
3734 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3735 struct netlink_ext_ack *extack);
3736 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3737 struct netlink_ext_ack *extack);
3738 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3739 struct netlink_ext_ack *extack);
3740 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3741 int dev_change_carrier(struct net_device *, bool new_carrier);
3742 int dev_get_phys_port_id(struct net_device *dev,
3743 struct netdev_phys_item_id *ppid);
3744 int dev_get_phys_port_name(struct net_device *dev,
3745 char *name, size_t len);
3746 int dev_get_port_parent_id(struct net_device *dev,
3747 struct netdev_phys_item_id *ppid, bool recurse);
3748 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3749 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3750 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3752 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3753 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3754 struct netdev_queue *txq, int *ret);
3756 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3757 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3758 int fd, int expected_fd, u32 flags);
3759 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3760 u8 dev_xdp_prog_count(struct net_device *dev);
3761 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3763 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3764 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3765 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3766 bool is_skb_forwardable(const struct net_device *dev,
3767 const struct sk_buff *skb);
3769 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3770 const struct sk_buff *skb,
3771 const bool check_mtu)
3773 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3776 if (!(dev->flags & IFF_UP))
3782 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3783 if (skb->len <= len)
3786 /* if TSO is enabled, we don't care about the length as the packet
3787 * could be forwarded without being segmented before
3789 if (skb_is_gso(skb))
3795 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3796 struct sk_buff *skb,
3797 const bool check_mtu)
3799 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3800 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3801 atomic_long_inc(&dev->rx_dropped);
3806 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
3811 bool dev_nit_active(struct net_device *dev);
3812 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3814 extern int netdev_budget;
3815 extern unsigned int netdev_budget_usecs;
3817 /* Called by rtnetlink.c:rtnl_unlock() */
3818 void netdev_run_todo(void);
3821 * dev_put - release reference to device
3822 * @dev: network device
3824 * Release reference to device to allow it to be freed.
3825 * Try using dev_put_track() instead.
3827 static inline void dev_put(struct net_device *dev)
3830 #ifdef CONFIG_PCPU_DEV_REFCNT
3831 this_cpu_dec(*dev->pcpu_refcnt);
3833 refcount_dec(&dev->dev_refcnt);
3839 * dev_hold - get reference to device
3840 * @dev: network device
3842 * Hold reference to device to keep it from being freed.
3843 * Try using dev_hold_track() instead.
3845 static inline void dev_hold(struct net_device *dev)
3848 #ifdef CONFIG_PCPU_DEV_REFCNT
3849 this_cpu_inc(*dev->pcpu_refcnt);
3851 refcount_inc(&dev->dev_refcnt);
3856 static inline void netdev_tracker_alloc(struct net_device *dev,
3857 netdevice_tracker *tracker, gfp_t gfp)
3859 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3860 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
3864 static inline void netdev_tracker_free(struct net_device *dev,
3865 netdevice_tracker *tracker)
3867 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3868 ref_tracker_free(&dev->refcnt_tracker, tracker);
3872 static inline void dev_hold_track(struct net_device *dev,
3873 netdevice_tracker *tracker, gfp_t gfp)
3877 netdev_tracker_alloc(dev, tracker, gfp);
3881 static inline void dev_put_track(struct net_device *dev,
3882 netdevice_tracker *tracker)
3885 netdev_tracker_free(dev, tracker);
3890 static inline void dev_replace_track(struct net_device *odev,
3891 struct net_device *ndev,
3892 netdevice_tracker *tracker,
3896 netdev_tracker_free(odev, tracker);
3902 netdev_tracker_alloc(ndev, tracker, gfp);
3905 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3906 * and _off may be called from IRQ context, but it is caller
3907 * who is responsible for serialization of these calls.
3909 * The name carrier is inappropriate, these functions should really be
3910 * called netif_lowerlayer_*() because they represent the state of any
3911 * kind of lower layer not just hardware media.
3914 void linkwatch_init_dev(struct net_device *dev);
3915 void linkwatch_fire_event(struct net_device *dev);
3916 void linkwatch_forget_dev(struct net_device *dev);
3919 * netif_carrier_ok - test if carrier present
3920 * @dev: network device
3922 * Check if carrier is present on device
3924 static inline bool netif_carrier_ok(const struct net_device *dev)
3926 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3929 unsigned long dev_trans_start(struct net_device *dev);
3931 void __netdev_watchdog_up(struct net_device *dev);
3933 void netif_carrier_on(struct net_device *dev);
3934 void netif_carrier_off(struct net_device *dev);
3935 void netif_carrier_event(struct net_device *dev);
3938 * netif_dormant_on - mark device as dormant.
3939 * @dev: network device
3941 * Mark device as dormant (as per RFC2863).
3943 * The dormant state indicates that the relevant interface is not
3944 * actually in a condition to pass packets (i.e., it is not 'up') but is
3945 * in a "pending" state, waiting for some external event. For "on-
3946 * demand" interfaces, this new state identifies the situation where the
3947 * interface is waiting for events to place it in the up state.
3949 static inline void netif_dormant_on(struct net_device *dev)
3951 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3952 linkwatch_fire_event(dev);
3956 * netif_dormant_off - set device as not dormant.
3957 * @dev: network device
3959 * Device is not in dormant state.
3961 static inline void netif_dormant_off(struct net_device *dev)
3963 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3964 linkwatch_fire_event(dev);
3968 * netif_dormant - test if device is dormant
3969 * @dev: network device
3971 * Check if device is dormant.
3973 static inline bool netif_dormant(const struct net_device *dev)
3975 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3980 * netif_testing_on - mark device as under test.
3981 * @dev: network device
3983 * Mark device as under test (as per RFC2863).
3985 * The testing state indicates that some test(s) must be performed on
3986 * the interface. After completion, of the test, the interface state
3987 * will change to up, dormant, or down, as appropriate.
3989 static inline void netif_testing_on(struct net_device *dev)
3991 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
3992 linkwatch_fire_event(dev);
3996 * netif_testing_off - set device as not under test.
3997 * @dev: network device
3999 * Device is not in testing state.
4001 static inline void netif_testing_off(struct net_device *dev)
4003 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4004 linkwatch_fire_event(dev);
4008 * netif_testing - test if device is under test
4009 * @dev: network device
4011 * Check if device is under test
4013 static inline bool netif_testing(const struct net_device *dev)
4015 return test_bit(__LINK_STATE_TESTING, &dev->state);
4020 * netif_oper_up - test if device is operational
4021 * @dev: network device
4023 * Check if carrier is operational
4025 static inline bool netif_oper_up(const struct net_device *dev)
4027 return (dev->operstate == IF_OPER_UP ||
4028 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4032 * netif_device_present - is device available or removed
4033 * @dev: network device
4035 * Check if device has not been removed from system.
4037 static inline bool netif_device_present(const struct net_device *dev)
4039 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4042 void netif_device_detach(struct net_device *dev);
4044 void netif_device_attach(struct net_device *dev);
4047 * Network interface message level settings
4052 NETIF_MSG_PROBE_BIT,
4054 NETIF_MSG_TIMER_BIT,
4055 NETIF_MSG_IFDOWN_BIT,
4057 NETIF_MSG_RX_ERR_BIT,
4058 NETIF_MSG_TX_ERR_BIT,
4059 NETIF_MSG_TX_QUEUED_BIT,
4061 NETIF_MSG_TX_DONE_BIT,
4062 NETIF_MSG_RX_STATUS_BIT,
4063 NETIF_MSG_PKTDATA_BIT,
4067 /* When you add a new bit above, update netif_msg_class_names array
4068 * in net/ethtool/common.c
4070 NETIF_MSG_CLASS_COUNT,
4072 /* Both ethtool_ops interface and internal driver implementation use u32 */
4073 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4075 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4076 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4078 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4079 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4080 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4081 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4082 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4083 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4084 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4085 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4086 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4087 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4088 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4089 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4090 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4091 #define NETIF_MSG_HW __NETIF_MSG(HW)
4092 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4094 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4095 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4096 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4097 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4098 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4099 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4100 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4101 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4102 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4103 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4104 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4105 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4106 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4107 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4108 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4110 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4113 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4114 return default_msg_enable_bits;
4115 if (debug_value == 0) /* no output */
4117 /* set low N bits */
4118 return (1U << debug_value) - 1;
4121 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4123 spin_lock(&txq->_xmit_lock);
4124 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4125 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4128 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4130 __acquire(&txq->_xmit_lock);
4134 static inline void __netif_tx_release(struct netdev_queue *txq)
4136 __release(&txq->_xmit_lock);
4139 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4141 spin_lock_bh(&txq->_xmit_lock);
4142 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4143 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4146 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4148 bool ok = spin_trylock(&txq->_xmit_lock);
4151 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4152 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4157 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4159 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4160 WRITE_ONCE(txq->xmit_lock_owner, -1);
4161 spin_unlock(&txq->_xmit_lock);
4164 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4166 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4167 WRITE_ONCE(txq->xmit_lock_owner, -1);
4168 spin_unlock_bh(&txq->_xmit_lock);
4172 * txq->trans_start can be read locklessly from dev_watchdog()
4174 static inline void txq_trans_update(struct netdev_queue *txq)
4176 if (txq->xmit_lock_owner != -1)
4177 WRITE_ONCE(txq->trans_start, jiffies);
4180 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4182 unsigned long now = jiffies;
4184 if (READ_ONCE(txq->trans_start) != now)
4185 WRITE_ONCE(txq->trans_start, now);
4188 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4189 static inline void netif_trans_update(struct net_device *dev)
4191 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4193 txq_trans_cond_update(txq);
4197 * netif_tx_lock - grab network device transmit lock
4198 * @dev: network device
4200 * Get network device transmit lock
4202 void netif_tx_lock(struct net_device *dev);
4204 static inline void netif_tx_lock_bh(struct net_device *dev)
4210 void netif_tx_unlock(struct net_device *dev);
4212 static inline void netif_tx_unlock_bh(struct net_device *dev)
4214 netif_tx_unlock(dev);
4218 #define HARD_TX_LOCK(dev, txq, cpu) { \
4219 if ((dev->features & NETIF_F_LLTX) == 0) { \
4220 __netif_tx_lock(txq, cpu); \
4222 __netif_tx_acquire(txq); \
4226 #define HARD_TX_TRYLOCK(dev, txq) \
4227 (((dev->features & NETIF_F_LLTX) == 0) ? \
4228 __netif_tx_trylock(txq) : \
4229 __netif_tx_acquire(txq))
4231 #define HARD_TX_UNLOCK(dev, txq) { \
4232 if ((dev->features & NETIF_F_LLTX) == 0) { \
4233 __netif_tx_unlock(txq); \
4235 __netif_tx_release(txq); \
4239 static inline void netif_tx_disable(struct net_device *dev)
4245 cpu = smp_processor_id();
4246 spin_lock(&dev->tx_global_lock);
4247 for (i = 0; i < dev->num_tx_queues; i++) {
4248 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4250 __netif_tx_lock(txq, cpu);
4251 netif_tx_stop_queue(txq);
4252 __netif_tx_unlock(txq);
4254 spin_unlock(&dev->tx_global_lock);
4258 static inline void netif_addr_lock(struct net_device *dev)
4260 unsigned char nest_level = 0;
4262 #ifdef CONFIG_LOCKDEP
4263 nest_level = dev->nested_level;
4265 spin_lock_nested(&dev->addr_list_lock, nest_level);
4268 static inline void netif_addr_lock_bh(struct net_device *dev)
4270 unsigned char nest_level = 0;
4272 #ifdef CONFIG_LOCKDEP
4273 nest_level = dev->nested_level;
4276 spin_lock_nested(&dev->addr_list_lock, nest_level);
4279 static inline void netif_addr_unlock(struct net_device *dev)
4281 spin_unlock(&dev->addr_list_lock);
4284 static inline void netif_addr_unlock_bh(struct net_device *dev)
4286 spin_unlock_bh(&dev->addr_list_lock);
4290 * dev_addrs walker. Should be used only for read access. Call with
4291 * rcu_read_lock held.
4293 #define for_each_dev_addr(dev, ha) \
4294 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4296 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4298 void ether_setup(struct net_device *dev);
4300 /* Support for loadable net-drivers */
4301 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4302 unsigned char name_assign_type,
4303 void (*setup)(struct net_device *),
4304 unsigned int txqs, unsigned int rxqs);
4305 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4306 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4308 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4309 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4312 int register_netdev(struct net_device *dev);
4313 void unregister_netdev(struct net_device *dev);
4315 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4317 /* General hardware address lists handling functions */
4318 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4319 struct netdev_hw_addr_list *from_list, int addr_len);
4320 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4321 struct netdev_hw_addr_list *from_list, int addr_len);
4322 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4323 struct net_device *dev,
4324 int (*sync)(struct net_device *, const unsigned char *),
4325 int (*unsync)(struct net_device *,
4326 const unsigned char *));
4327 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4328 struct net_device *dev,
4329 int (*sync)(struct net_device *,
4330 const unsigned char *, int),
4331 int (*unsync)(struct net_device *,
4332 const unsigned char *, int));
4333 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4334 struct net_device *dev,
4335 int (*unsync)(struct net_device *,
4336 const unsigned char *, int));
4337 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4338 struct net_device *dev,
4339 int (*unsync)(struct net_device *,
4340 const unsigned char *));
4341 void __hw_addr_init(struct netdev_hw_addr_list *list);
4343 /* Functions used for device addresses handling */
4344 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4345 const void *addr, size_t len);
4348 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4350 dev_addr_mod(dev, 0, addr, len);
4353 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4355 __dev_addr_set(dev, addr, dev->addr_len);
4358 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4359 unsigned char addr_type);
4360 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4361 unsigned char addr_type);
4362 void dev_addr_flush(struct net_device *dev);
4363 int dev_addr_init(struct net_device *dev);
4364 void dev_addr_check(struct net_device *dev);
4366 /* Functions used for unicast addresses handling */
4367 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4368 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4369 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4370 int dev_uc_sync(struct net_device *to, struct net_device *from);
4371 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4372 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4373 void dev_uc_flush(struct net_device *dev);
4374 void dev_uc_init(struct net_device *dev);
4377 * __dev_uc_sync - Synchonize device's unicast list
4378 * @dev: device to sync
4379 * @sync: function to call if address should be added
4380 * @unsync: function to call if address should be removed
4382 * Add newly added addresses to the interface, and release
4383 * addresses that have been deleted.
4385 static inline int __dev_uc_sync(struct net_device *dev,
4386 int (*sync)(struct net_device *,
4387 const unsigned char *),
4388 int (*unsync)(struct net_device *,
4389 const unsigned char *))
4391 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4395 * __dev_uc_unsync - Remove synchronized addresses from device
4396 * @dev: device to sync
4397 * @unsync: function to call if address should be removed
4399 * Remove all addresses that were added to the device by dev_uc_sync().
4401 static inline void __dev_uc_unsync(struct net_device *dev,
4402 int (*unsync)(struct net_device *,
4403 const unsigned char *))
4405 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4408 /* Functions used for multicast addresses handling */
4409 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4410 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4411 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4412 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4413 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4414 int dev_mc_sync(struct net_device *to, struct net_device *from);
4415 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4416 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4417 void dev_mc_flush(struct net_device *dev);
4418 void dev_mc_init(struct net_device *dev);
4421 * __dev_mc_sync - Synchonize device's multicast list
4422 * @dev: device to sync
4423 * @sync: function to call if address should be added
4424 * @unsync: function to call if address should be removed
4426 * Add newly added addresses to the interface, and release
4427 * addresses that have been deleted.
4429 static inline int __dev_mc_sync(struct net_device *dev,
4430 int (*sync)(struct net_device *,
4431 const unsigned char *),
4432 int (*unsync)(struct net_device *,
4433 const unsigned char *))
4435 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4439 * __dev_mc_unsync - Remove synchronized addresses from device
4440 * @dev: device to sync
4441 * @unsync: function to call if address should be removed
4443 * Remove all addresses that were added to the device by dev_mc_sync().
4445 static inline void __dev_mc_unsync(struct net_device *dev,
4446 int (*unsync)(struct net_device *,
4447 const unsigned char *))
4449 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4452 /* Functions used for secondary unicast and multicast support */
4453 void dev_set_rx_mode(struct net_device *dev);
4454 void __dev_set_rx_mode(struct net_device *dev);
4455 int dev_set_promiscuity(struct net_device *dev, int inc);
4456 int dev_set_allmulti(struct net_device *dev, int inc);
4457 void netdev_state_change(struct net_device *dev);
4458 void __netdev_notify_peers(struct net_device *dev);
4459 void netdev_notify_peers(struct net_device *dev);
4460 void netdev_features_change(struct net_device *dev);
4461 /* Load a device via the kmod */
4462 void dev_load(struct net *net, const char *name);
4463 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4464 struct rtnl_link_stats64 *storage);
4465 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4466 const struct net_device_stats *netdev_stats);
4467 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4468 const struct pcpu_sw_netstats __percpu *netstats);
4469 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4471 extern int netdev_max_backlog;
4472 extern int netdev_tstamp_prequeue;
4473 extern int netdev_unregister_timeout_secs;
4474 extern int weight_p;
4475 extern int dev_weight_rx_bias;
4476 extern int dev_weight_tx_bias;
4477 extern int dev_rx_weight;
4478 extern int dev_tx_weight;
4479 extern int gro_normal_batch;
4482 NESTED_SYNC_IMM_BIT,
4483 NESTED_SYNC_TODO_BIT,
4486 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4487 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4489 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4490 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4492 struct netdev_nested_priv {
4493 unsigned char flags;
4497 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4498 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4499 struct list_head **iter);
4501 #ifdef CONFIG_LOCKDEP
4502 static LIST_HEAD(net_unlink_list);
4504 static inline void net_unlink_todo(struct net_device *dev)
4506 if (list_empty(&dev->unlink_list))
4507 list_add_tail(&dev->unlink_list, &net_unlink_list);
4511 /* iterate through upper list, must be called under RCU read lock */
4512 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4513 for (iter = &(dev)->adj_list.upper, \
4514 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4516 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4518 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4519 int (*fn)(struct net_device *upper_dev,
4520 struct netdev_nested_priv *priv),
4521 struct netdev_nested_priv *priv);
4523 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4524 struct net_device *upper_dev);
4526 bool netdev_has_any_upper_dev(struct net_device *dev);
4528 void *netdev_lower_get_next_private(struct net_device *dev,
4529 struct list_head **iter);
4530 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4531 struct list_head **iter);
4533 #define netdev_for_each_lower_private(dev, priv, iter) \
4534 for (iter = (dev)->adj_list.lower.next, \
4535 priv = netdev_lower_get_next_private(dev, &(iter)); \
4537 priv = netdev_lower_get_next_private(dev, &(iter)))
4539 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4540 for (iter = &(dev)->adj_list.lower, \
4541 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4543 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4545 void *netdev_lower_get_next(struct net_device *dev,
4546 struct list_head **iter);
4548 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4549 for (iter = (dev)->adj_list.lower.next, \
4550 ldev = netdev_lower_get_next(dev, &(iter)); \
4552 ldev = netdev_lower_get_next(dev, &(iter)))
4554 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4555 struct list_head **iter);
4556 int netdev_walk_all_lower_dev(struct net_device *dev,
4557 int (*fn)(struct net_device *lower_dev,
4558 struct netdev_nested_priv *priv),
4559 struct netdev_nested_priv *priv);
4560 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4561 int (*fn)(struct net_device *lower_dev,
4562 struct netdev_nested_priv *priv),
4563 struct netdev_nested_priv *priv);
4565 void *netdev_adjacent_get_private(struct list_head *adj_list);
4566 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4567 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4568 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4569 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4570 struct netlink_ext_ack *extack);
4571 int netdev_master_upper_dev_link(struct net_device *dev,
4572 struct net_device *upper_dev,
4573 void *upper_priv, void *upper_info,
4574 struct netlink_ext_ack *extack);
4575 void netdev_upper_dev_unlink(struct net_device *dev,
4576 struct net_device *upper_dev);
4577 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4578 struct net_device *new_dev,
4579 struct net_device *dev,
4580 struct netlink_ext_ack *extack);
4581 void netdev_adjacent_change_commit(struct net_device *old_dev,
4582 struct net_device *new_dev,
4583 struct net_device *dev);
4584 void netdev_adjacent_change_abort(struct net_device *old_dev,
4585 struct net_device *new_dev,
4586 struct net_device *dev);
4587 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4588 void *netdev_lower_dev_get_private(struct net_device *dev,
4589 struct net_device *lower_dev);
4590 void netdev_lower_state_changed(struct net_device *lower_dev,
4591 void *lower_state_info);
4593 /* RSS keys are 40 or 52 bytes long */
4594 #define NETDEV_RSS_KEY_LEN 52
4595 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4596 void netdev_rss_key_fill(void *buffer, size_t len);
4598 int skb_checksum_help(struct sk_buff *skb);
4599 int skb_crc32c_csum_help(struct sk_buff *skb);
4600 int skb_csum_hwoffload_help(struct sk_buff *skb,
4601 const netdev_features_t features);
4603 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4604 netdev_features_t features, bool tx_path);
4605 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4606 netdev_features_t features);
4608 struct netdev_bonding_info {
4613 struct netdev_notifier_bonding_info {
4614 struct netdev_notifier_info info; /* must be first */
4615 struct netdev_bonding_info bonding_info;
4618 void netdev_bonding_info_change(struct net_device *dev,
4619 struct netdev_bonding_info *bonding_info);
4621 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4622 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4624 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4631 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4633 return __skb_gso_segment(skb, features, true);
4635 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4637 static inline bool can_checksum_protocol(netdev_features_t features,
4640 if (protocol == htons(ETH_P_FCOE))
4641 return !!(features & NETIF_F_FCOE_CRC);
4643 /* Assume this is an IP checksum (not SCTP CRC) */
4645 if (features & NETIF_F_HW_CSUM) {
4646 /* Can checksum everything */
4651 case htons(ETH_P_IP):
4652 return !!(features & NETIF_F_IP_CSUM);
4653 case htons(ETH_P_IPV6):
4654 return !!(features & NETIF_F_IPV6_CSUM);
4661 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4663 static inline void netdev_rx_csum_fault(struct net_device *dev,
4664 struct sk_buff *skb)
4668 /* rx skb timestamps */
4669 void net_enable_timestamp(void);
4670 void net_disable_timestamp(void);
4672 #ifdef CONFIG_PROC_FS
4673 int __init dev_proc_init(void);
4675 #define dev_proc_init() 0
4678 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4679 struct sk_buff *skb, struct net_device *dev,
4682 __this_cpu_write(softnet_data.xmit.more, more);
4683 return ops->ndo_start_xmit(skb, dev);
4686 static inline bool netdev_xmit_more(void)
4688 return __this_cpu_read(softnet_data.xmit.more);
4691 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4692 struct netdev_queue *txq, bool more)
4694 const struct net_device_ops *ops = dev->netdev_ops;
4697 rc = __netdev_start_xmit(ops, skb, dev, more);
4698 if (rc == NETDEV_TX_OK)
4699 txq_trans_update(txq);
4704 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4706 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4709 extern const struct kobj_ns_type_operations net_ns_type_operations;
4711 const char *netdev_drivername(const struct net_device *dev);
4713 void linkwatch_run_queue(void);
4715 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4716 netdev_features_t f2)
4718 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4719 if (f1 & NETIF_F_HW_CSUM)
4720 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4722 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4728 static inline netdev_features_t netdev_get_wanted_features(
4729 struct net_device *dev)
4731 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4733 netdev_features_t netdev_increment_features(netdev_features_t all,
4734 netdev_features_t one, netdev_features_t mask);
4736 /* Allow TSO being used on stacked device :
4737 * Performing the GSO segmentation before last device
4738 * is a performance improvement.
4740 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4741 netdev_features_t mask)
4743 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4746 int __netdev_update_features(struct net_device *dev);
4747 void netdev_update_features(struct net_device *dev);
4748 void netdev_change_features(struct net_device *dev);
4750 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4751 struct net_device *dev);
4753 netdev_features_t passthru_features_check(struct sk_buff *skb,
4754 struct net_device *dev,
4755 netdev_features_t features);
4756 netdev_features_t netif_skb_features(struct sk_buff *skb);
4758 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4760 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4762 /* check flags correspondence */
4763 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4764 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4765 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4766 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4767 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4768 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4769 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4770 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4771 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4772 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4773 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4774 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4775 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4776 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4777 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4778 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4779 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4780 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4781 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4783 return (features & feature) == feature;
4786 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4788 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4789 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4792 static inline bool netif_needs_gso(struct sk_buff *skb,
4793 netdev_features_t features)
4795 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4796 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4797 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4800 static inline void netif_set_gso_max_size(struct net_device *dev,
4803 /* dev->gso_max_size is read locklessly from sk_setup_caps() */
4804 WRITE_ONCE(dev->gso_max_size, size);
4807 static inline void netif_set_gso_max_segs(struct net_device *dev,
4810 /* dev->gso_max_segs is read locklessly from sk_setup_caps() */
4811 WRITE_ONCE(dev->gso_max_segs, segs);
4814 static inline void netif_set_gro_max_size(struct net_device *dev,
4817 /* This pairs with the READ_ONCE() in skb_gro_receive() */
4818 WRITE_ONCE(dev->gro_max_size, size);
4821 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4822 int pulled_hlen, u16 mac_offset,
4825 skb->protocol = protocol;
4826 skb->encapsulation = 1;
4827 skb_push(skb, pulled_hlen);
4828 skb_reset_transport_header(skb);
4829 skb->mac_header = mac_offset;
4830 skb->network_header = skb->mac_header + mac_len;
4831 skb->mac_len = mac_len;
4834 static inline bool netif_is_macsec(const struct net_device *dev)
4836 return dev->priv_flags & IFF_MACSEC;
4839 static inline bool netif_is_macvlan(const struct net_device *dev)
4841 return dev->priv_flags & IFF_MACVLAN;
4844 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4846 return dev->priv_flags & IFF_MACVLAN_PORT;
4849 static inline bool netif_is_bond_master(const struct net_device *dev)
4851 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4854 static inline bool netif_is_bond_slave(const struct net_device *dev)
4856 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4859 static inline bool netif_supports_nofcs(struct net_device *dev)
4861 return dev->priv_flags & IFF_SUPP_NOFCS;
4864 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4866 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4869 static inline bool netif_is_l3_master(const struct net_device *dev)
4871 return dev->priv_flags & IFF_L3MDEV_MASTER;
4874 static inline bool netif_is_l3_slave(const struct net_device *dev)
4876 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4879 static inline bool netif_is_bridge_master(const struct net_device *dev)
4881 return dev->priv_flags & IFF_EBRIDGE;
4884 static inline bool netif_is_bridge_port(const struct net_device *dev)
4886 return dev->priv_flags & IFF_BRIDGE_PORT;
4889 static inline bool netif_is_ovs_master(const struct net_device *dev)
4891 return dev->priv_flags & IFF_OPENVSWITCH;
4894 static inline bool netif_is_ovs_port(const struct net_device *dev)
4896 return dev->priv_flags & IFF_OVS_DATAPATH;
4899 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
4901 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
4904 static inline bool netif_is_team_master(const struct net_device *dev)
4906 return dev->priv_flags & IFF_TEAM;
4909 static inline bool netif_is_team_port(const struct net_device *dev)
4911 return dev->priv_flags & IFF_TEAM_PORT;
4914 static inline bool netif_is_lag_master(const struct net_device *dev)
4916 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4919 static inline bool netif_is_lag_port(const struct net_device *dev)
4921 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4924 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4926 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4929 static inline bool netif_is_failover(const struct net_device *dev)
4931 return dev->priv_flags & IFF_FAILOVER;
4934 static inline bool netif_is_failover_slave(const struct net_device *dev)
4936 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4939 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4940 static inline void netif_keep_dst(struct net_device *dev)
4942 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4945 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4946 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4948 /* TODO: reserve and use an additional IFF bit, if we get more users */
4949 return netif_is_macsec(dev);
4952 extern struct pernet_operations __net_initdata loopback_net_ops;
4954 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4956 /* netdev_printk helpers, similar to dev_printk */
4958 static inline const char *netdev_name(const struct net_device *dev)
4960 if (!dev->name[0] || strchr(dev->name, '%'))
4961 return "(unnamed net_device)";
4965 static inline bool netdev_unregistering(const struct net_device *dev)
4967 return dev->reg_state == NETREG_UNREGISTERING;
4970 static inline const char *netdev_reg_state(const struct net_device *dev)
4972 switch (dev->reg_state) {
4973 case NETREG_UNINITIALIZED: return " (uninitialized)";
4974 case NETREG_REGISTERED: return "";
4975 case NETREG_UNREGISTERING: return " (unregistering)";
4976 case NETREG_UNREGISTERED: return " (unregistered)";
4977 case NETREG_RELEASED: return " (released)";
4978 case NETREG_DUMMY: return " (dummy)";
4981 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4982 return " (unknown)";
4985 __printf(3, 4) __cold
4986 void netdev_printk(const char *level, const struct net_device *dev,
4987 const char *format, ...);
4988 __printf(2, 3) __cold
4989 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4990 __printf(2, 3) __cold
4991 void netdev_alert(const struct net_device *dev, const char *format, ...);
4992 __printf(2, 3) __cold
4993 void netdev_crit(const struct net_device *dev, const char *format, ...);
4994 __printf(2, 3) __cold
4995 void netdev_err(const struct net_device *dev, const char *format, ...);
4996 __printf(2, 3) __cold
4997 void netdev_warn(const struct net_device *dev, const char *format, ...);
4998 __printf(2, 3) __cold
4999 void netdev_notice(const struct net_device *dev, const char *format, ...);
5000 __printf(2, 3) __cold
5001 void netdev_info(const struct net_device *dev, const char *format, ...);
5003 #define netdev_level_once(level, dev, fmt, ...) \
5005 static bool __section(".data.once") __print_once; \
5007 if (!__print_once) { \
5008 __print_once = true; \
5009 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5013 #define netdev_emerg_once(dev, fmt, ...) \
5014 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5015 #define netdev_alert_once(dev, fmt, ...) \
5016 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5017 #define netdev_crit_once(dev, fmt, ...) \
5018 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5019 #define netdev_err_once(dev, fmt, ...) \
5020 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5021 #define netdev_warn_once(dev, fmt, ...) \
5022 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5023 #define netdev_notice_once(dev, fmt, ...) \
5024 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5025 #define netdev_info_once(dev, fmt, ...) \
5026 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5028 #define MODULE_ALIAS_NETDEV(device) \
5029 MODULE_ALIAS("netdev-" device)
5031 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5032 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5033 #define netdev_dbg(__dev, format, args...) \
5035 dynamic_netdev_dbg(__dev, format, ##args); \
5037 #elif defined(DEBUG)
5038 #define netdev_dbg(__dev, format, args...) \
5039 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5041 #define netdev_dbg(__dev, format, args...) \
5044 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5048 #if defined(VERBOSE_DEBUG)
5049 #define netdev_vdbg netdev_dbg
5052 #define netdev_vdbg(dev, format, args...) \
5055 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5061 * netdev_WARN() acts like dev_printk(), but with the key difference
5062 * of using a WARN/WARN_ON to get the message out, including the
5063 * file/line information and a backtrace.
5065 #define netdev_WARN(dev, format, args...) \
5066 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5067 netdev_reg_state(dev), ##args)
5069 #define netdev_WARN_ONCE(dev, format, args...) \
5070 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5071 netdev_reg_state(dev), ##args)
5073 /* netif printk helpers, similar to netdev_printk */
5075 #define netif_printk(priv, type, level, dev, fmt, args...) \
5077 if (netif_msg_##type(priv)) \
5078 netdev_printk(level, (dev), fmt, ##args); \
5081 #define netif_level(level, priv, type, dev, fmt, args...) \
5083 if (netif_msg_##type(priv)) \
5084 netdev_##level(dev, fmt, ##args); \
5087 #define netif_emerg(priv, type, dev, fmt, args...) \
5088 netif_level(emerg, priv, type, dev, fmt, ##args)
5089 #define netif_alert(priv, type, dev, fmt, args...) \
5090 netif_level(alert, priv, type, dev, fmt, ##args)
5091 #define netif_crit(priv, type, dev, fmt, args...) \
5092 netif_level(crit, priv, type, dev, fmt, ##args)
5093 #define netif_err(priv, type, dev, fmt, args...) \
5094 netif_level(err, priv, type, dev, fmt, ##args)
5095 #define netif_warn(priv, type, dev, fmt, args...) \
5096 netif_level(warn, priv, type, dev, fmt, ##args)
5097 #define netif_notice(priv, type, dev, fmt, args...) \
5098 netif_level(notice, priv, type, dev, fmt, ##args)
5099 #define netif_info(priv, type, dev, fmt, args...) \
5100 netif_level(info, priv, type, dev, fmt, ##args)
5102 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5103 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5104 #define netif_dbg(priv, type, netdev, format, args...) \
5106 if (netif_msg_##type(priv)) \
5107 dynamic_netdev_dbg(netdev, format, ##args); \
5109 #elif defined(DEBUG)
5110 #define netif_dbg(priv, type, dev, format, args...) \
5111 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5113 #define netif_dbg(priv, type, dev, format, args...) \
5116 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5121 /* if @cond then downgrade to debug, else print at @level */
5122 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5125 netif_dbg(priv, type, netdev, fmt, ##args); \
5127 netif_ ## level(priv, type, netdev, fmt, ##args); \
5130 #if defined(VERBOSE_DEBUG)
5131 #define netif_vdbg netif_dbg
5133 #define netif_vdbg(priv, type, dev, format, args...) \
5136 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5142 * The list of packet types we will receive (as opposed to discard)
5143 * and the routines to invoke.
5145 * Why 16. Because with 16 the only overlap we get on a hash of the
5146 * low nibble of the protocol value is RARP/SNAP/X.25.
5160 #define PTYPE_HASH_SIZE (16)
5161 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5163 extern struct list_head ptype_all __read_mostly;
5164 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5166 extern struct net_device *blackhole_netdev;
5168 #endif /* _LINUX_NETDEVICE_H */