1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Definitions for the Interfaces handler.
9 * Version: @(#)dev.h 1.0.10 08/12/93
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
19 * Moved to /usr/include/linux for NET3
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
32 #include <linux/percpu.h>
33 #include <linux/rculist.h>
34 #include <linux/workqueue.h>
35 #include <linux/dynamic_queue_limits.h>
37 #include <linux/ethtool.h>
38 #include <net/net_namespace.h>
40 #include <net/dcbnl.h>
42 #include <net/netprio_cgroup.h>
45 #include <linux/netdev_features.h>
46 #include <linux/neighbour.h>
47 #include <uapi/linux/netdevice.h>
48 #include <uapi/linux/if_bonding.h>
49 #include <uapi/linux/pkt_cls.h>
50 #include <linux/hashtable.h>
56 struct ip_tunnel_parm;
57 struct macsec_context;
63 /* 802.15.4 specific */
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info;
68 struct udp_tunnel_nic_info;
69 struct udp_tunnel_nic;
73 void synchronize_net(void);
74 void netdev_set_default_ethtool_ops(struct net_device *dev,
75 const struct ethtool_ops *ops);
77 /* Backlog congestion levels */
78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
79 #define NET_RX_DROP 1 /* packet dropped */
81 #define MAX_NEST_DEV 8
84 * Transmit return codes: transmit return codes originate from three different
87 * - qdisc return codes
88 * - driver transmit return codes
91 * Drivers are allowed to return any one of those in their hard_start_xmit()
92 * function. Real network devices commonly used with qdiscs should only return
93 * the driver transmit return codes though - when qdiscs are used, the actual
94 * transmission happens asynchronously, so the value is not propagated to
95 * higher layers. Virtual network devices transmit synchronously; in this case
96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
97 * others are propagated to higher layers.
100 /* qdisc ->enqueue() return codes. */
101 #define NET_XMIT_SUCCESS 0x00
102 #define NET_XMIT_DROP 0x01 /* skb dropped */
103 #define NET_XMIT_CN 0x02 /* congestion notification */
104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
107 * indicates that the device will soon be dropping packets, or already drops
108 * some packets of the same priority; prompting us to send less aggressively. */
109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
112 /* Driver transmit return codes */
113 #define NETDEV_TX_MASK 0xf0
116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
117 NETDEV_TX_OK = 0x00, /* driver took care of packet */
118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
120 typedef enum netdev_tx netdev_tx_t;
123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
126 static inline bool dev_xmit_complete(int rc)
129 * Positive cases with an skb consumed by a driver:
130 * - successful transmission (rc == NETDEV_TX_OK)
131 * - error while transmitting (rc < 0)
132 * - error while queueing to a different device (rc & NET_XMIT_MASK)
134 if (likely(rc < NET_XMIT_MASK))
141 * Compute the worst-case header length according to the protocols
145 #if defined(CONFIG_HYPERV_NET)
146 # define LL_MAX_HEADER 128
147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
148 # if defined(CONFIG_MAC80211_MESH)
149 # define LL_MAX_HEADER 128
151 # define LL_MAX_HEADER 96
154 # define LL_MAX_HEADER 32
157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
159 #define MAX_HEADER LL_MAX_HEADER
161 #define MAX_HEADER (LL_MAX_HEADER + 48)
165 * Old network device statistics. Fields are native words
166 * (unsigned long) so they can be read and written atomically.
169 struct net_device_stats {
170 unsigned long rx_packets;
171 unsigned long tx_packets;
172 unsigned long rx_bytes;
173 unsigned long tx_bytes;
174 unsigned long rx_errors;
175 unsigned long tx_errors;
176 unsigned long rx_dropped;
177 unsigned long tx_dropped;
178 unsigned long multicast;
179 unsigned long collisions;
180 unsigned long rx_length_errors;
181 unsigned long rx_over_errors;
182 unsigned long rx_crc_errors;
183 unsigned long rx_frame_errors;
184 unsigned long rx_fifo_errors;
185 unsigned long rx_missed_errors;
186 unsigned long tx_aborted_errors;
187 unsigned long tx_carrier_errors;
188 unsigned long tx_fifo_errors;
189 unsigned long tx_heartbeat_errors;
190 unsigned long tx_window_errors;
191 unsigned long rx_compressed;
192 unsigned long tx_compressed;
196 #include <linux/cache.h>
197 #include <linux/skbuff.h>
200 #include <linux/static_key.h>
201 extern struct static_key_false rps_needed;
202 extern struct static_key_false rfs_needed;
209 struct netdev_hw_addr {
210 struct list_head list;
211 unsigned char addr[MAX_ADDR_LEN];
213 #define NETDEV_HW_ADDR_T_LAN 1
214 #define NETDEV_HW_ADDR_T_SAN 2
215 #define NETDEV_HW_ADDR_T_SLAVE 3
216 #define NETDEV_HW_ADDR_T_UNICAST 4
217 #define NETDEV_HW_ADDR_T_MULTICAST 5
222 struct rcu_head rcu_head;
225 struct netdev_hw_addr_list {
226 struct list_head list;
230 #define netdev_hw_addr_list_count(l) ((l)->count)
231 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
232 #define netdev_hw_addr_list_for_each(ha, l) \
233 list_for_each_entry(ha, &(l)->list, list)
235 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
236 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
237 #define netdev_for_each_uc_addr(ha, dev) \
238 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
240 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
241 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
242 #define netdev_for_each_mc_addr(ha, dev) \
243 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
249 /* cached hardware header; allow for machine alignment needs. */
250 #define HH_DATA_MOD 16
251 #define HH_DATA_OFF(__len) \
252 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
253 #define HH_DATA_ALIGN(__len) \
254 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
255 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
258 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
260 * dev->hard_header_len ? (dev->hard_header_len +
261 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
263 * We could use other alignment values, but we must maintain the
264 * relationship HH alignment <= LL alignment.
266 #define LL_RESERVED_SPACE(dev) \
267 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
268 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
269 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
272 int (*create) (struct sk_buff *skb, struct net_device *dev,
273 unsigned short type, const void *daddr,
274 const void *saddr, unsigned int len);
275 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
276 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
277 void (*cache_update)(struct hh_cache *hh,
278 const struct net_device *dev,
279 const unsigned char *haddr);
280 bool (*validate)(const char *ll_header, unsigned int len);
281 __be16 (*parse_protocol)(const struct sk_buff *skb);
284 /* These flag bits are private to the generic network queueing
285 * layer; they may not be explicitly referenced by any other
289 enum netdev_state_t {
291 __LINK_STATE_PRESENT,
292 __LINK_STATE_NOCARRIER,
293 __LINK_STATE_LINKWATCH_PENDING,
294 __LINK_STATE_DORMANT,
295 __LINK_STATE_TESTING,
300 * This structure holds boot-time configured netdevice settings. They
301 * are then used in the device probing.
303 struct netdev_boot_setup {
307 #define NETDEV_BOOT_SETUP_MAX 8
309 int __init netdev_boot_setup(char *str);
312 struct list_head list;
317 * size of gro hash buckets, must less than bit number of
318 * napi_struct::gro_bitmask
320 #define GRO_HASH_BUCKETS 8
323 * Structure for NAPI scheduling similar to tasklet but with weighting
326 /* The poll_list must only be managed by the entity which
327 * changes the state of the NAPI_STATE_SCHED bit. This means
328 * whoever atomically sets that bit can add this napi_struct
329 * to the per-CPU poll_list, and whoever clears that bit
330 * can remove from the list right before clearing the bit.
332 struct list_head poll_list;
336 int defer_hard_irqs_count;
337 unsigned long gro_bitmask;
338 int (*poll)(struct napi_struct *, int);
339 #ifdef CONFIG_NETPOLL
342 struct net_device *dev;
343 struct gro_list gro_hash[GRO_HASH_BUCKETS];
345 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
346 int rx_count; /* length of rx_list */
347 struct hrtimer timer;
348 struct list_head dev_list;
349 struct hlist_node napi_hash_node;
350 unsigned int napi_id;
354 NAPI_STATE_SCHED, /* Poll is scheduled */
355 NAPI_STATE_MISSED, /* reschedule a napi */
356 NAPI_STATE_DISABLE, /* Disable pending */
357 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
358 NAPI_STATE_LISTED, /* NAPI added to system lists */
359 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
360 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
364 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
365 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
366 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
367 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
368 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
369 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
370 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
381 typedef enum gro_result gro_result_t;
384 * enum rx_handler_result - Possible return values for rx_handlers.
385 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
387 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
388 * case skb->dev was changed by rx_handler.
389 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
390 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
392 * rx_handlers are functions called from inside __netif_receive_skb(), to do
393 * special processing of the skb, prior to delivery to protocol handlers.
395 * Currently, a net_device can only have a single rx_handler registered. Trying
396 * to register a second rx_handler will return -EBUSY.
398 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
399 * To unregister a rx_handler on a net_device, use
400 * netdev_rx_handler_unregister().
402 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
405 * If the rx_handler consumed the skb in some way, it should return
406 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
407 * the skb to be delivered in some other way.
409 * If the rx_handler changed skb->dev, to divert the skb to another
410 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
411 * new device will be called if it exists.
413 * If the rx_handler decides the skb should be ignored, it should return
414 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
415 * are registered on exact device (ptype->dev == skb->dev).
417 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
418 * delivered, it should return RX_HANDLER_PASS.
420 * A device without a registered rx_handler will behave as if rx_handler
421 * returned RX_HANDLER_PASS.
424 enum rx_handler_result {
430 typedef enum rx_handler_result rx_handler_result_t;
431 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
433 void __napi_schedule(struct napi_struct *n);
434 void __napi_schedule_irqoff(struct napi_struct *n);
436 static inline bool napi_disable_pending(struct napi_struct *n)
438 return test_bit(NAPI_STATE_DISABLE, &n->state);
441 bool napi_schedule_prep(struct napi_struct *n);
444 * napi_schedule - schedule NAPI poll
447 * Schedule NAPI poll routine to be called if it is not already
450 static inline void napi_schedule(struct napi_struct *n)
452 if (napi_schedule_prep(n))
457 * napi_schedule_irqoff - schedule NAPI poll
460 * Variant of napi_schedule(), assuming hard irqs are masked.
462 static inline void napi_schedule_irqoff(struct napi_struct *n)
464 if (napi_schedule_prep(n))
465 __napi_schedule_irqoff(n);
468 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
469 static inline bool napi_reschedule(struct napi_struct *napi)
471 if (napi_schedule_prep(napi)) {
472 __napi_schedule(napi);
478 bool napi_complete_done(struct napi_struct *n, int work_done);
480 * napi_complete - NAPI processing complete
483 * Mark NAPI processing as complete.
484 * Consider using napi_complete_done() instead.
485 * Return false if device should avoid rearming interrupts.
487 static inline bool napi_complete(struct napi_struct *n)
489 return napi_complete_done(n, 0);
493 * napi_disable - prevent NAPI from scheduling
496 * Stop NAPI from being scheduled on this context.
497 * Waits till any outstanding processing completes.
499 void napi_disable(struct napi_struct *n);
502 * napi_enable - enable NAPI scheduling
505 * Resume NAPI from being scheduled on this context.
506 * Must be paired with napi_disable.
508 static inline void napi_enable(struct napi_struct *n)
510 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
511 smp_mb__before_atomic();
512 clear_bit(NAPI_STATE_SCHED, &n->state);
513 clear_bit(NAPI_STATE_NPSVC, &n->state);
517 * napi_synchronize - wait until NAPI is not running
520 * Wait until NAPI is done being scheduled on this context.
521 * Waits till any outstanding processing completes but
522 * does not disable future activations.
524 static inline void napi_synchronize(const struct napi_struct *n)
526 if (IS_ENABLED(CONFIG_SMP))
527 while (test_bit(NAPI_STATE_SCHED, &n->state))
534 * napi_if_scheduled_mark_missed - if napi is running, set the
538 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
541 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
543 unsigned long val, new;
546 val = READ_ONCE(n->state);
547 if (val & NAPIF_STATE_DISABLE)
550 if (!(val & NAPIF_STATE_SCHED))
553 new = val | NAPIF_STATE_MISSED;
554 } while (cmpxchg(&n->state, val, new) != val);
559 enum netdev_queue_state_t {
560 __QUEUE_STATE_DRV_XOFF,
561 __QUEUE_STATE_STACK_XOFF,
562 __QUEUE_STATE_FROZEN,
565 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
566 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
567 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
569 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
570 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
572 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
576 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
577 * netif_tx_* functions below are used to manipulate this flag. The
578 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
579 * queue independently. The netif_xmit_*stopped functions below are called
580 * to check if the queue has been stopped by the driver or stack (either
581 * of the XOFF bits are set in the state). Drivers should not need to call
582 * netif_xmit*stopped functions, they should only be using netif_tx_*.
585 struct netdev_queue {
589 struct net_device *dev;
590 struct Qdisc __rcu *qdisc;
591 struct Qdisc *qdisc_sleeping;
595 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
598 unsigned long tx_maxrate;
600 * Number of TX timeouts for this queue
601 * (/sys/class/net/DEV/Q/trans_timeout)
603 unsigned long trans_timeout;
605 /* Subordinate device that the queue has been assigned to */
606 struct net_device *sb_dev;
607 #ifdef CONFIG_XDP_SOCKETS
608 struct xsk_buff_pool *pool;
613 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
616 * Time (in jiffies) of last Tx
618 unsigned long trans_start;
625 } ____cacheline_aligned_in_smp;
627 extern int sysctl_fb_tunnels_only_for_init_net;
628 extern int sysctl_devconf_inherit_init_net;
631 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
632 * == 1 : For initns only
635 static inline bool net_has_fallback_tunnels(const struct net *net)
637 return (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1) ||
638 !sysctl_fb_tunnels_only_for_init_net;
641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
659 * This structure holds an RPS map which can be of variable length. The
660 * map is an array of CPUs.
667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
671 * tail pointer for that CPU's input queue at the time of last enqueue, and
672 * a hardware filter index.
674 struct rps_dev_flow {
677 unsigned int last_qtail;
679 #define RPS_NO_FILTER 0xffff
682 * The rps_dev_flow_table structure contains a table of flow mappings.
684 struct rps_dev_flow_table {
687 struct rps_dev_flow flows[];
689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
690 ((_num) * sizeof(struct rps_dev_flow)))
693 * The rps_sock_flow_table contains mappings of flows to the last CPU
694 * on which they were processed by the application (set in recvmsg).
695 * Each entry is a 32bit value. Upper part is the high-order bits
696 * of flow hash, lower part is CPU number.
697 * rps_cpu_mask is used to partition the space, depending on number of
698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
700 * meaning we use 32-6=26 bits for the hash.
702 struct rps_sock_flow_table {
705 u32 ents[] ____cacheline_aligned_in_smp;
707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
709 #define RPS_NO_CPU 0xffff
711 extern u32 rps_cpu_mask;
712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
718 unsigned int index = hash & table->mask;
719 u32 val = hash & ~rps_cpu_mask;
721 /* We only give a hint, preemption can change CPU under us */
722 val |= raw_smp_processor_id();
724 if (table->ents[index] != val)
725 table->ents[index] = val;
729 #ifdef CONFIG_RFS_ACCEL
730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
733 #endif /* CONFIG_RPS */
735 /* This structure contains an instance of an RX queue. */
736 struct netdev_rx_queue {
738 struct rps_map __rcu *rps_map;
739 struct rps_dev_flow_table __rcu *rps_flow_table;
742 struct net_device *dev;
743 struct xdp_rxq_info xdp_rxq;
744 #ifdef CONFIG_XDP_SOCKETS
745 struct xsk_buff_pool *pool;
747 } ____cacheline_aligned_in_smp;
750 * RX queue sysfs structures and functions.
752 struct rx_queue_attribute {
753 struct attribute attr;
754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
755 ssize_t (*store)(struct netdev_rx_queue *queue,
756 const char *buf, size_t len);
761 * This structure holds an XPS map which can be of variable length. The
762 * map is an array of queues.
766 unsigned int alloc_len;
770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
772 - sizeof(struct xps_map)) / sizeof(u16))
775 * This structure holds all XPS maps for device. Maps are indexed by CPU.
777 struct xps_dev_maps {
779 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
782 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
785 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
786 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
788 #endif /* CONFIG_XPS */
790 #define TC_MAX_QUEUE 16
791 #define TC_BITMASK 15
792 /* HW offloaded queuing disciplines txq count and offset maps */
793 struct netdev_tc_txq {
798 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
800 * This structure is to hold information about the device
801 * configured to run FCoE protocol stack.
803 struct netdev_fcoe_hbainfo {
804 char manufacturer[64];
805 char serial_number[64];
806 char hardware_version[64];
807 char driver_version[64];
808 char optionrom_version[64];
809 char firmware_version[64];
811 char model_description[256];
815 #define MAX_PHYS_ITEM_ID_LEN 32
817 /* This structure holds a unique identifier to identify some
818 * physical item (port for example) used by a netdevice.
820 struct netdev_phys_item_id {
821 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
822 unsigned char id_len;
825 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
826 struct netdev_phys_item_id *b)
828 return a->id_len == b->id_len &&
829 memcmp(a->id, b->id, a->id_len) == 0;
832 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
834 struct net_device *sb_dev);
837 TC_SETUP_QDISC_MQPRIO,
840 TC_SETUP_CLSMATCHALL,
850 TC_SETUP_QDISC_TAPRIO,
857 /* These structures hold the attributes of bpf state that are being passed
858 * to the netdevice through the bpf op.
860 enum bpf_netdev_command {
861 /* Set or clear a bpf program used in the earliest stages of packet
862 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
863 * is responsible for calling bpf_prog_put on any old progs that are
864 * stored. In case of error, the callee need not release the new prog
865 * reference, but on success it takes ownership and must bpf_prog_put
866 * when it is no longer used.
870 /* BPF program for offload callbacks, invoked at program load time. */
871 BPF_OFFLOAD_MAP_ALLOC,
872 BPF_OFFLOAD_MAP_FREE,
876 struct bpf_prog_offload_ops;
877 struct netlink_ext_ack;
879 struct xdp_dev_bulk_queue;
889 struct bpf_xdp_entity {
890 struct bpf_prog *prog;
891 struct bpf_xdp_link *link;
895 enum bpf_netdev_command command;
900 struct bpf_prog *prog;
901 struct netlink_ext_ack *extack;
903 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
905 struct bpf_offloaded_map *offmap;
907 /* XDP_SETUP_XSK_POOL */
909 struct xsk_buff_pool *pool;
915 /* Flags for ndo_xsk_wakeup. */
916 #define XDP_WAKEUP_RX (1 << 0)
917 #define XDP_WAKEUP_TX (1 << 1)
919 #ifdef CONFIG_XFRM_OFFLOAD
921 int (*xdo_dev_state_add) (struct xfrm_state *x);
922 void (*xdo_dev_state_delete) (struct xfrm_state *x);
923 void (*xdo_dev_state_free) (struct xfrm_state *x);
924 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
925 struct xfrm_state *x);
926 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
931 struct rcu_head rcuhead;
938 struct netdev_name_node {
939 struct hlist_node hlist;
940 struct list_head list;
941 struct net_device *dev;
945 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
946 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
948 struct netdev_net_notifier {
949 struct list_head list;
950 struct notifier_block *nb;
954 * This structure defines the management hooks for network devices.
955 * The following hooks can be defined; unless noted otherwise, they are
956 * optional and can be filled with a null pointer.
958 * int (*ndo_init)(struct net_device *dev);
959 * This function is called once when a network device is registered.
960 * The network device can use this for any late stage initialization
961 * or semantic validation. It can fail with an error code which will
962 * be propagated back to register_netdev.
964 * void (*ndo_uninit)(struct net_device *dev);
965 * This function is called when device is unregistered or when registration
966 * fails. It is not called if init fails.
968 * int (*ndo_open)(struct net_device *dev);
969 * This function is called when a network device transitions to the up
972 * int (*ndo_stop)(struct net_device *dev);
973 * This function is called when a network device transitions to the down
976 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
977 * struct net_device *dev);
978 * Called when a packet needs to be transmitted.
979 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
980 * the queue before that can happen; it's for obsolete devices and weird
981 * corner cases, but the stack really does a non-trivial amount
982 * of useless work if you return NETDEV_TX_BUSY.
983 * Required; cannot be NULL.
985 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
986 * struct net_device *dev
987 * netdev_features_t features);
988 * Called by core transmit path to determine if device is capable of
989 * performing offload operations on a given packet. This is to give
990 * the device an opportunity to implement any restrictions that cannot
991 * be otherwise expressed by feature flags. The check is called with
992 * the set of features that the stack has calculated and it returns
993 * those the driver believes to be appropriate.
995 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
996 * struct net_device *sb_dev);
997 * Called to decide which queue to use when device supports multiple
1000 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1001 * This function is called to allow device receiver to make
1002 * changes to configuration when multicast or promiscuous is enabled.
1004 * void (*ndo_set_rx_mode)(struct net_device *dev);
1005 * This function is called device changes address list filtering.
1006 * If driver handles unicast address filtering, it should set
1007 * IFF_UNICAST_FLT in its priv_flags.
1009 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1010 * This function is called when the Media Access Control address
1011 * needs to be changed. If this interface is not defined, the
1012 * MAC address can not be changed.
1014 * int (*ndo_validate_addr)(struct net_device *dev);
1015 * Test if Media Access Control address is valid for the device.
1017 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1018 * Called when a user requests an ioctl which can't be handled by
1019 * the generic interface code. If not defined ioctls return
1020 * not supported error code.
1022 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1023 * Used to set network devices bus interface parameters. This interface
1024 * is retained for legacy reasons; new devices should use the bus
1025 * interface (PCI) for low level management.
1027 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1028 * Called when a user wants to change the Maximum Transfer Unit
1031 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1032 * Callback used when the transmitter has not made any progress
1033 * for dev->watchdog ticks.
1035 * void (*ndo_get_stats64)(struct net_device *dev,
1036 * struct rtnl_link_stats64 *storage);
1037 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1038 * Called when a user wants to get the network device usage
1039 * statistics. Drivers must do one of the following:
1040 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1041 * rtnl_link_stats64 structure passed by the caller.
1042 * 2. Define @ndo_get_stats to update a net_device_stats structure
1043 * (which should normally be dev->stats) and return a pointer to
1044 * it. The structure may be changed asynchronously only if each
1045 * field is written atomically.
1046 * 3. Update dev->stats asynchronously and atomically, and define
1047 * neither operation.
1049 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1050 * Return true if this device supports offload stats of this attr_id.
1052 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1054 * Get statistics for offload operations by attr_id. Write it into the
1055 * attr_data pointer.
1057 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1058 * If device supports VLAN filtering this function is called when a
1059 * VLAN id is registered.
1061 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1062 * If device supports VLAN filtering this function is called when a
1063 * VLAN id is unregistered.
1065 * void (*ndo_poll_controller)(struct net_device *dev);
1067 * SR-IOV management functions.
1068 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1069 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1070 * u8 qos, __be16 proto);
1071 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1073 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1074 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1075 * int (*ndo_get_vf_config)(struct net_device *dev,
1076 * int vf, struct ifla_vf_info *ivf);
1077 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1078 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1079 * struct nlattr *port[]);
1081 * Enable or disable the VF ability to query its RSS Redirection Table and
1082 * Hash Key. This is needed since on some devices VF share this information
1083 * with PF and querying it may introduce a theoretical security risk.
1084 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1085 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1086 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1088 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1089 * This is always called from the stack with the rtnl lock held and netif
1090 * tx queues stopped. This allows the netdevice to perform queue
1091 * management safely.
1093 * Fiber Channel over Ethernet (FCoE) offload functions.
1094 * int (*ndo_fcoe_enable)(struct net_device *dev);
1095 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1096 * so the underlying device can perform whatever needed configuration or
1097 * initialization to support acceleration of FCoE traffic.
1099 * int (*ndo_fcoe_disable)(struct net_device *dev);
1100 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1101 * so the underlying device can perform whatever needed clean-ups to
1102 * stop supporting acceleration of FCoE traffic.
1104 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1105 * struct scatterlist *sgl, unsigned int sgc);
1106 * Called when the FCoE Initiator wants to initialize an I/O that
1107 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1108 * perform necessary setup and returns 1 to indicate the device is set up
1109 * successfully to perform DDP on this I/O, otherwise this returns 0.
1111 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1112 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1113 * indicated by the FC exchange id 'xid', so the underlying device can
1114 * clean up and reuse resources for later DDP requests.
1116 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1117 * struct scatterlist *sgl, unsigned int sgc);
1118 * Called when the FCoE Target wants to initialize an I/O that
1119 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1120 * perform necessary setup and returns 1 to indicate the device is set up
1121 * successfully to perform DDP on this I/O, otherwise this returns 0.
1123 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1124 * struct netdev_fcoe_hbainfo *hbainfo);
1125 * Called when the FCoE Protocol stack wants information on the underlying
1126 * device. This information is utilized by the FCoE protocol stack to
1127 * register attributes with Fiber Channel management service as per the
1128 * FC-GS Fabric Device Management Information(FDMI) specification.
1130 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1131 * Called when the underlying device wants to override default World Wide
1132 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1133 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1134 * protocol stack to use.
1137 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1138 * u16 rxq_index, u32 flow_id);
1139 * Set hardware filter for RFS. rxq_index is the target queue index;
1140 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1141 * Return the filter ID on success, or a negative error code.
1143 * Slave management functions (for bridge, bonding, etc).
1144 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1145 * Called to make another netdev an underling.
1147 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1148 * Called to release previously enslaved netdev.
1150 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1151 * struct sk_buff *skb,
1153 * Get the xmit slave of master device. If all_slaves is true, function
1154 * assume all the slaves can transmit.
1156 * Feature/offload setting functions.
1157 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1158 * netdev_features_t features);
1159 * Adjusts the requested feature flags according to device-specific
1160 * constraints, and returns the resulting flags. Must not modify
1163 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1164 * Called to update device configuration to new features. Passed
1165 * feature set might be less than what was returned by ndo_fix_features()).
1166 * Must return >0 or -errno if it changed dev->features itself.
1168 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1169 * struct net_device *dev,
1170 * const unsigned char *addr, u16 vid, u16 flags,
1171 * struct netlink_ext_ack *extack);
1172 * Adds an FDB entry to dev for addr.
1173 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1174 * struct net_device *dev,
1175 * const unsigned char *addr, u16 vid)
1176 * Deletes the FDB entry from dev coresponding to addr.
1177 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1178 * struct net_device *dev, struct net_device *filter_dev,
1180 * Used to add FDB entries to dump requests. Implementers should add
1181 * entries to skb and update idx with the number of entries.
1183 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1184 * u16 flags, struct netlink_ext_ack *extack)
1185 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1186 * struct net_device *dev, u32 filter_mask,
1188 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1191 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1192 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1193 * which do not represent real hardware may define this to allow their
1194 * userspace components to manage their virtual carrier state. Devices
1195 * that determine carrier state from physical hardware properties (eg
1196 * network cables) or protocol-dependent mechanisms (eg
1197 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1199 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1200 * struct netdev_phys_item_id *ppid);
1201 * Called to get ID of physical port of this device. If driver does
1202 * not implement this, it is assumed that the hw is not able to have
1203 * multiple net devices on single physical port.
1205 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1206 * struct netdev_phys_item_id *ppid)
1207 * Called to get the parent ID of the physical port of this device.
1209 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1210 * struct udp_tunnel_info *ti);
1211 * Called by UDP tunnel to notify a driver about the UDP port and socket
1212 * address family that a UDP tunnel is listnening to. It is called only
1213 * when a new port starts listening. The operation is protected by the
1216 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1217 * struct udp_tunnel_info *ti);
1218 * Called by UDP tunnel to notify the driver about a UDP port and socket
1219 * address family that the UDP tunnel is not listening to anymore. The
1220 * operation is protected by the RTNL.
1222 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1223 * struct net_device *dev)
1224 * Called by upper layer devices to accelerate switching or other
1225 * station functionality into hardware. 'pdev is the lowerdev
1226 * to use for the offload and 'dev' is the net device that will
1227 * back the offload. Returns a pointer to the private structure
1228 * the upper layer will maintain.
1229 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1230 * Called by upper layer device to delete the station created
1231 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1232 * the station and priv is the structure returned by the add
1234 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1235 * int queue_index, u32 maxrate);
1236 * Called when a user wants to set a max-rate limitation of specific
1238 * int (*ndo_get_iflink)(const struct net_device *dev);
1239 * Called to get the iflink value of this device.
1240 * void (*ndo_change_proto_down)(struct net_device *dev,
1242 * This function is used to pass protocol port error state information
1243 * to the switch driver. The switch driver can react to the proto_down
1244 * by doing a phys down on the associated switch port.
1245 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1246 * This function is used to get egress tunnel information for given skb.
1247 * This is useful for retrieving outer tunnel header parameters while
1249 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1250 * This function is used to specify the headroom that the skb must
1251 * consider when allocation skb during packet reception. Setting
1252 * appropriate rx headroom value allows avoiding skb head copy on
1253 * forward. Setting a negative value resets the rx headroom to the
1255 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1256 * This function is used to set or query state related to XDP on the
1257 * netdevice and manage BPF offload. See definition of
1258 * enum bpf_netdev_command for details.
1259 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1261 * This function is used to submit @n XDP packets for transmit on a
1262 * netdevice. Returns number of frames successfully transmitted, frames
1263 * that got dropped are freed/returned via xdp_return_frame().
1264 * Returns negative number, means general error invoking ndo, meaning
1265 * no frames were xmit'ed and core-caller will free all frames.
1266 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1267 * This function is used to wake up the softirq, ksoftirqd or kthread
1268 * responsible for sending and/or receiving packets on a specific
1269 * queue id bound to an AF_XDP socket. The flags field specifies if
1270 * only RX, only Tx, or both should be woken up using the flags
1271 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1272 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1273 * Get devlink port instance associated with a given netdev.
1274 * Called with a reference on the netdevice and devlink locks only,
1275 * rtnl_lock is not held.
1276 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1278 * Add, change, delete or get information on an IPv4 tunnel.
1280 struct net_device_ops {
1281 int (*ndo_init)(struct net_device *dev);
1282 void (*ndo_uninit)(struct net_device *dev);
1283 int (*ndo_open)(struct net_device *dev);
1284 int (*ndo_stop)(struct net_device *dev);
1285 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1286 struct net_device *dev);
1287 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1288 struct net_device *dev,
1289 netdev_features_t features);
1290 u16 (*ndo_select_queue)(struct net_device *dev,
1291 struct sk_buff *skb,
1292 struct net_device *sb_dev);
1293 void (*ndo_change_rx_flags)(struct net_device *dev,
1295 void (*ndo_set_rx_mode)(struct net_device *dev);
1296 int (*ndo_set_mac_address)(struct net_device *dev,
1298 int (*ndo_validate_addr)(struct net_device *dev);
1299 int (*ndo_do_ioctl)(struct net_device *dev,
1300 struct ifreq *ifr, int cmd);
1301 int (*ndo_set_config)(struct net_device *dev,
1303 int (*ndo_change_mtu)(struct net_device *dev,
1305 int (*ndo_neigh_setup)(struct net_device *dev,
1306 struct neigh_parms *);
1307 void (*ndo_tx_timeout) (struct net_device *dev,
1308 unsigned int txqueue);
1310 void (*ndo_get_stats64)(struct net_device *dev,
1311 struct rtnl_link_stats64 *storage);
1312 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1313 int (*ndo_get_offload_stats)(int attr_id,
1314 const struct net_device *dev,
1316 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1318 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1319 __be16 proto, u16 vid);
1320 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1321 __be16 proto, u16 vid);
1322 #ifdef CONFIG_NET_POLL_CONTROLLER
1323 void (*ndo_poll_controller)(struct net_device *dev);
1324 int (*ndo_netpoll_setup)(struct net_device *dev,
1325 struct netpoll_info *info);
1326 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1328 int (*ndo_set_vf_mac)(struct net_device *dev,
1329 int queue, u8 *mac);
1330 int (*ndo_set_vf_vlan)(struct net_device *dev,
1331 int queue, u16 vlan,
1332 u8 qos, __be16 proto);
1333 int (*ndo_set_vf_rate)(struct net_device *dev,
1334 int vf, int min_tx_rate,
1336 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1337 int vf, bool setting);
1338 int (*ndo_set_vf_trust)(struct net_device *dev,
1339 int vf, bool setting);
1340 int (*ndo_get_vf_config)(struct net_device *dev,
1342 struct ifla_vf_info *ivf);
1343 int (*ndo_set_vf_link_state)(struct net_device *dev,
1344 int vf, int link_state);
1345 int (*ndo_get_vf_stats)(struct net_device *dev,
1347 struct ifla_vf_stats
1349 int (*ndo_set_vf_port)(struct net_device *dev,
1351 struct nlattr *port[]);
1352 int (*ndo_get_vf_port)(struct net_device *dev,
1353 int vf, struct sk_buff *skb);
1354 int (*ndo_get_vf_guid)(struct net_device *dev,
1356 struct ifla_vf_guid *node_guid,
1357 struct ifla_vf_guid *port_guid);
1358 int (*ndo_set_vf_guid)(struct net_device *dev,
1361 int (*ndo_set_vf_rss_query_en)(
1362 struct net_device *dev,
1363 int vf, bool setting);
1364 int (*ndo_setup_tc)(struct net_device *dev,
1365 enum tc_setup_type type,
1367 #if IS_ENABLED(CONFIG_FCOE)
1368 int (*ndo_fcoe_enable)(struct net_device *dev);
1369 int (*ndo_fcoe_disable)(struct net_device *dev);
1370 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1372 struct scatterlist *sgl,
1374 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1376 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1378 struct scatterlist *sgl,
1380 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1381 struct netdev_fcoe_hbainfo *hbainfo);
1384 #if IS_ENABLED(CONFIG_LIBFCOE)
1385 #define NETDEV_FCOE_WWNN 0
1386 #define NETDEV_FCOE_WWPN 1
1387 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1388 u64 *wwn, int type);
1391 #ifdef CONFIG_RFS_ACCEL
1392 int (*ndo_rx_flow_steer)(struct net_device *dev,
1393 const struct sk_buff *skb,
1397 int (*ndo_add_slave)(struct net_device *dev,
1398 struct net_device *slave_dev,
1399 struct netlink_ext_ack *extack);
1400 int (*ndo_del_slave)(struct net_device *dev,
1401 struct net_device *slave_dev);
1402 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1403 struct sk_buff *skb,
1405 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1406 netdev_features_t features);
1407 int (*ndo_set_features)(struct net_device *dev,
1408 netdev_features_t features);
1409 int (*ndo_neigh_construct)(struct net_device *dev,
1410 struct neighbour *n);
1411 void (*ndo_neigh_destroy)(struct net_device *dev,
1412 struct neighbour *n);
1414 int (*ndo_fdb_add)(struct ndmsg *ndm,
1415 struct nlattr *tb[],
1416 struct net_device *dev,
1417 const unsigned char *addr,
1420 struct netlink_ext_ack *extack);
1421 int (*ndo_fdb_del)(struct ndmsg *ndm,
1422 struct nlattr *tb[],
1423 struct net_device *dev,
1424 const unsigned char *addr,
1426 int (*ndo_fdb_dump)(struct sk_buff *skb,
1427 struct netlink_callback *cb,
1428 struct net_device *dev,
1429 struct net_device *filter_dev,
1431 int (*ndo_fdb_get)(struct sk_buff *skb,
1432 struct nlattr *tb[],
1433 struct net_device *dev,
1434 const unsigned char *addr,
1435 u16 vid, u32 portid, u32 seq,
1436 struct netlink_ext_ack *extack);
1437 int (*ndo_bridge_setlink)(struct net_device *dev,
1438 struct nlmsghdr *nlh,
1440 struct netlink_ext_ack *extack);
1441 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1443 struct net_device *dev,
1446 int (*ndo_bridge_dellink)(struct net_device *dev,
1447 struct nlmsghdr *nlh,
1449 int (*ndo_change_carrier)(struct net_device *dev,
1451 int (*ndo_get_phys_port_id)(struct net_device *dev,
1452 struct netdev_phys_item_id *ppid);
1453 int (*ndo_get_port_parent_id)(struct net_device *dev,
1454 struct netdev_phys_item_id *ppid);
1455 int (*ndo_get_phys_port_name)(struct net_device *dev,
1456 char *name, size_t len);
1457 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1458 struct udp_tunnel_info *ti);
1459 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1460 struct udp_tunnel_info *ti);
1461 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1462 struct net_device *dev);
1463 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1466 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1469 int (*ndo_get_iflink)(const struct net_device *dev);
1470 int (*ndo_change_proto_down)(struct net_device *dev,
1472 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1473 struct sk_buff *skb);
1474 void (*ndo_set_rx_headroom)(struct net_device *dev,
1475 int needed_headroom);
1476 int (*ndo_bpf)(struct net_device *dev,
1477 struct netdev_bpf *bpf);
1478 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1479 struct xdp_frame **xdp,
1481 int (*ndo_xsk_wakeup)(struct net_device *dev,
1482 u32 queue_id, u32 flags);
1483 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1484 int (*ndo_tunnel_ctl)(struct net_device *dev,
1485 struct ip_tunnel_parm *p, int cmd);
1489 * enum net_device_priv_flags - &struct net_device priv_flags
1491 * These are the &struct net_device, they are only set internally
1492 * by drivers and used in the kernel. These flags are invisible to
1493 * userspace; this means that the order of these flags can change
1494 * during any kernel release.
1496 * You should have a pretty good reason to be extending these flags.
1498 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1499 * @IFF_EBRIDGE: Ethernet bridging device
1500 * @IFF_BONDING: bonding master or slave
1501 * @IFF_ISATAP: ISATAP interface (RFC4214)
1502 * @IFF_WAN_HDLC: WAN HDLC device
1503 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1505 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1506 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1507 * @IFF_MACVLAN_PORT: device used as macvlan port
1508 * @IFF_BRIDGE_PORT: device used as bridge port
1509 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1510 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1511 * @IFF_UNICAST_FLT: Supports unicast filtering
1512 * @IFF_TEAM_PORT: device used as team port
1513 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1514 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1515 * change when it's running
1516 * @IFF_MACVLAN: Macvlan device
1517 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1518 * underlying stacked devices
1519 * @IFF_L3MDEV_MASTER: device is an L3 master device
1520 * @IFF_NO_QUEUE: device can run without qdisc attached
1521 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1522 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1523 * @IFF_TEAM: device is a team device
1524 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1525 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1526 * entity (i.e. the master device for bridged veth)
1527 * @IFF_MACSEC: device is a MACsec device
1528 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1529 * @IFF_FAILOVER: device is a failover master device
1530 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1531 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1532 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1534 enum netdev_priv_flags {
1535 IFF_802_1Q_VLAN = 1<<0,
1539 IFF_WAN_HDLC = 1<<4,
1540 IFF_XMIT_DST_RELEASE = 1<<5,
1541 IFF_DONT_BRIDGE = 1<<6,
1542 IFF_DISABLE_NETPOLL = 1<<7,
1543 IFF_MACVLAN_PORT = 1<<8,
1544 IFF_BRIDGE_PORT = 1<<9,
1545 IFF_OVS_DATAPATH = 1<<10,
1546 IFF_TX_SKB_SHARING = 1<<11,
1547 IFF_UNICAST_FLT = 1<<12,
1548 IFF_TEAM_PORT = 1<<13,
1549 IFF_SUPP_NOFCS = 1<<14,
1550 IFF_LIVE_ADDR_CHANGE = 1<<15,
1551 IFF_MACVLAN = 1<<16,
1552 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1553 IFF_L3MDEV_MASTER = 1<<18,
1554 IFF_NO_QUEUE = 1<<19,
1555 IFF_OPENVSWITCH = 1<<20,
1556 IFF_L3MDEV_SLAVE = 1<<21,
1558 IFF_RXFH_CONFIGURED = 1<<23,
1559 IFF_PHONY_HEADROOM = 1<<24,
1561 IFF_NO_RX_HANDLER = 1<<26,
1562 IFF_FAILOVER = 1<<27,
1563 IFF_FAILOVER_SLAVE = 1<<28,
1564 IFF_L3MDEV_RX_HANDLER = 1<<29,
1565 IFF_LIVE_RENAME_OK = 1<<30,
1568 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1569 #define IFF_EBRIDGE IFF_EBRIDGE
1570 #define IFF_BONDING IFF_BONDING
1571 #define IFF_ISATAP IFF_ISATAP
1572 #define IFF_WAN_HDLC IFF_WAN_HDLC
1573 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1574 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1575 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1576 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1577 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1578 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1579 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1580 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1581 #define IFF_TEAM_PORT IFF_TEAM_PORT
1582 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1583 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1584 #define IFF_MACVLAN IFF_MACVLAN
1585 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1586 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1587 #define IFF_NO_QUEUE IFF_NO_QUEUE
1588 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1589 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1590 #define IFF_TEAM IFF_TEAM
1591 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1592 #define IFF_MACSEC IFF_MACSEC
1593 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1594 #define IFF_FAILOVER IFF_FAILOVER
1595 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1596 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1597 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1600 * struct net_device - The DEVICE structure.
1602 * Actually, this whole structure is a big mistake. It mixes I/O
1603 * data with strictly "high-level" data, and it has to know about
1604 * almost every data structure used in the INET module.
1606 * @name: This is the first field of the "visible" part of this structure
1607 * (i.e. as seen by users in the "Space.c" file). It is the name
1610 * @name_node: Name hashlist node
1611 * @ifalias: SNMP alias
1612 * @mem_end: Shared memory end
1613 * @mem_start: Shared memory start
1614 * @base_addr: Device I/O address
1615 * @irq: Device IRQ number
1617 * @state: Generic network queuing layer state, see netdev_state_t
1618 * @dev_list: The global list of network devices
1619 * @napi_list: List entry used for polling NAPI devices
1620 * @unreg_list: List entry when we are unregistering the
1621 * device; see the function unregister_netdev
1622 * @close_list: List entry used when we are closing the device
1623 * @ptype_all: Device-specific packet handlers for all protocols
1624 * @ptype_specific: Device-specific, protocol-specific packet handlers
1626 * @adj_list: Directly linked devices, like slaves for bonding
1627 * @features: Currently active device features
1628 * @hw_features: User-changeable features
1630 * @wanted_features: User-requested features
1631 * @vlan_features: Mask of features inheritable by VLAN devices
1633 * @hw_enc_features: Mask of features inherited by encapsulating devices
1634 * This field indicates what encapsulation
1635 * offloads the hardware is capable of doing,
1636 * and drivers will need to set them appropriately.
1638 * @mpls_features: Mask of features inheritable by MPLS
1639 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1641 * @ifindex: interface index
1642 * @group: The group the device belongs to
1644 * @stats: Statistics struct, which was left as a legacy, use
1645 * rtnl_link_stats64 instead
1647 * @rx_dropped: Dropped packets by core network,
1648 * do not use this in drivers
1649 * @tx_dropped: Dropped packets by core network,
1650 * do not use this in drivers
1651 * @rx_nohandler: nohandler dropped packets by core network on
1652 * inactive devices, do not use this in drivers
1653 * @carrier_up_count: Number of times the carrier has been up
1654 * @carrier_down_count: Number of times the carrier has been down
1656 * @wireless_handlers: List of functions to handle Wireless Extensions,
1658 * see <net/iw_handler.h> for details.
1659 * @wireless_data: Instance data managed by the core of wireless extensions
1661 * @netdev_ops: Includes several pointers to callbacks,
1662 * if one wants to override the ndo_*() functions
1663 * @ethtool_ops: Management operations
1664 * @l3mdev_ops: Layer 3 master device operations
1665 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1666 * discovery handling. Necessary for e.g. 6LoWPAN.
1667 * @xfrmdev_ops: Transformation offload operations
1668 * @tlsdev_ops: Transport Layer Security offload operations
1669 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1670 * of Layer 2 headers.
1672 * @flags: Interface flags (a la BSD)
1673 * @priv_flags: Like 'flags' but invisible to userspace,
1674 * see if.h for the definitions
1675 * @gflags: Global flags ( kept as legacy )
1676 * @padded: How much padding added by alloc_netdev()
1677 * @operstate: RFC2863 operstate
1678 * @link_mode: Mapping policy to operstate
1679 * @if_port: Selectable AUI, TP, ...
1681 * @mtu: Interface MTU value
1682 * @min_mtu: Interface Minimum MTU value
1683 * @max_mtu: Interface Maximum MTU value
1684 * @type: Interface hardware type
1685 * @hard_header_len: Maximum hardware header length.
1686 * @min_header_len: Minimum hardware header length
1688 * @needed_headroom: Extra headroom the hardware may need, but not in all
1689 * cases can this be guaranteed
1690 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1691 * cases can this be guaranteed. Some cases also use
1692 * LL_MAX_HEADER instead to allocate the skb
1694 * interface address info:
1696 * @perm_addr: Permanent hw address
1697 * @addr_assign_type: Hw address assignment type
1698 * @addr_len: Hardware address length
1699 * @upper_level: Maximum depth level of upper devices.
1700 * @lower_level: Maximum depth level of lower devices.
1701 * @neigh_priv_len: Used in neigh_alloc()
1702 * @dev_id: Used to differentiate devices that share
1703 * the same link layer address
1704 * @dev_port: Used to differentiate devices that share
1706 * @addr_list_lock: XXX: need comments on this one
1707 * @name_assign_type: network interface name assignment type
1708 * @uc_promisc: Counter that indicates promiscuous mode
1709 * has been enabled due to the need to listen to
1710 * additional unicast addresses in a device that
1711 * does not implement ndo_set_rx_mode()
1712 * @uc: unicast mac addresses
1713 * @mc: multicast mac addresses
1714 * @dev_addrs: list of device hw addresses
1715 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1716 * @promiscuity: Number of times the NIC is told to work in
1717 * promiscuous mode; if it becomes 0 the NIC will
1718 * exit promiscuous mode
1719 * @allmulti: Counter, enables or disables allmulticast mode
1721 * @vlan_info: VLAN info
1722 * @dsa_ptr: dsa specific data
1723 * @tipc_ptr: TIPC specific data
1724 * @atalk_ptr: AppleTalk link
1725 * @ip_ptr: IPv4 specific data
1726 * @dn_ptr: DECnet specific data
1727 * @ip6_ptr: IPv6 specific data
1728 * @ax25_ptr: AX.25 specific data
1729 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1730 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1732 * @mpls_ptr: mpls_dev struct pointer
1734 * @dev_addr: Hw address (before bcast,
1735 * because most packets are unicast)
1737 * @_rx: Array of RX queues
1738 * @num_rx_queues: Number of RX queues
1739 * allocated at register_netdev() time
1740 * @real_num_rx_queues: Number of RX queues currently active in device
1741 * @xdp_prog: XDP sockets filter program pointer
1742 * @gro_flush_timeout: timeout for GRO layer in NAPI
1743 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1744 * allow to avoid NIC hard IRQ, on busy queues.
1746 * @rx_handler: handler for received packets
1747 * @rx_handler_data: XXX: need comments on this one
1748 * @miniq_ingress: ingress/clsact qdisc specific data for
1749 * ingress processing
1750 * @ingress_queue: XXX: need comments on this one
1751 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1752 * @broadcast: hw bcast address
1754 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1755 * indexed by RX queue number. Assigned by driver.
1756 * This must only be set if the ndo_rx_flow_steer
1757 * operation is defined
1758 * @index_hlist: Device index hash chain
1760 * @_tx: Array of TX queues
1761 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1762 * @real_num_tx_queues: Number of TX queues currently active in device
1763 * @qdisc: Root qdisc from userspace point of view
1764 * @tx_queue_len: Max frames per queue allowed
1765 * @tx_global_lock: XXX: need comments on this one
1766 * @xdp_bulkq: XDP device bulk queue
1767 * @xps_cpus_map: all CPUs map for XPS device
1768 * @xps_rxqs_map: all RXQs map for XPS device
1770 * @xps_maps: XXX: need comments on this one
1771 * @miniq_egress: clsact qdisc specific data for
1773 * @qdisc_hash: qdisc hash table
1774 * @watchdog_timeo: Represents the timeout that is used by
1775 * the watchdog (see dev_watchdog())
1776 * @watchdog_timer: List of timers
1778 * @pcpu_refcnt: Number of references to this device
1779 * @todo_list: Delayed register/unregister
1780 * @link_watch_list: XXX: need comments on this one
1782 * @reg_state: Register/unregister state machine
1783 * @dismantle: Device is going to be freed
1784 * @rtnl_link_state: This enum represents the phases of creating
1787 * @needs_free_netdev: Should unregister perform free_netdev?
1788 * @priv_destructor: Called from unregister
1789 * @npinfo: XXX: need comments on this one
1790 * @nd_net: Network namespace this network device is inside
1792 * @ml_priv: Mid-layer private
1793 * @lstats: Loopback statistics
1794 * @tstats: Tunnel statistics
1795 * @dstats: Dummy statistics
1796 * @vstats: Virtual ethernet statistics
1801 * @dev: Class/net/name entry
1802 * @sysfs_groups: Space for optional device, statistics and wireless
1805 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1806 * @rtnl_link_ops: Rtnl_link_ops
1808 * @gso_max_size: Maximum size of generic segmentation offload
1809 * @gso_max_segs: Maximum number of segments that can be passed to the
1812 * @dcbnl_ops: Data Center Bridging netlink ops
1813 * @num_tc: Number of traffic classes in the net device
1814 * @tc_to_txq: XXX: need comments on this one
1815 * @prio_tc_map: XXX: need comments on this one
1817 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1819 * @priomap: XXX: need comments on this one
1820 * @phydev: Physical device may attach itself
1821 * for hardware timestamping
1822 * @sfp_bus: attached &struct sfp_bus structure.
1824 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1825 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1827 * @proto_down: protocol port state information can be sent to the
1828 * switch driver and used to set the phys state of the
1831 * @wol_enabled: Wake-on-LAN is enabled
1833 * @net_notifier_list: List of per-net netdev notifier block
1834 * that follow this device when it is moved
1835 * to another network namespace.
1837 * @macsec_ops: MACsec offloading ops
1839 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1840 * offload capabilities of the device
1841 * @udp_tunnel_nic: UDP tunnel offload state
1843 * FIXME: cleanup struct net_device such that network protocol info
1848 char name[IFNAMSIZ];
1849 struct netdev_name_node *name_node;
1850 struct dev_ifalias __rcu *ifalias;
1852 * I/O specific fields
1853 * FIXME: Merge these and struct ifmap into one
1855 unsigned long mem_end;
1856 unsigned long mem_start;
1857 unsigned long base_addr;
1861 * Some hardware also needs these fields (state,dev_list,
1862 * napi_list,unreg_list,close_list) but they are not
1863 * part of the usual set specified in Space.c.
1866 unsigned long state;
1868 struct list_head dev_list;
1869 struct list_head napi_list;
1870 struct list_head unreg_list;
1871 struct list_head close_list;
1872 struct list_head ptype_all;
1873 struct list_head ptype_specific;
1876 struct list_head upper;
1877 struct list_head lower;
1880 netdev_features_t features;
1881 netdev_features_t hw_features;
1882 netdev_features_t wanted_features;
1883 netdev_features_t vlan_features;
1884 netdev_features_t hw_enc_features;
1885 netdev_features_t mpls_features;
1886 netdev_features_t gso_partial_features;
1891 struct net_device_stats stats;
1893 atomic_long_t rx_dropped;
1894 atomic_long_t tx_dropped;
1895 atomic_long_t rx_nohandler;
1897 /* Stats to monitor link on/off, flapping */
1898 atomic_t carrier_up_count;
1899 atomic_t carrier_down_count;
1901 #ifdef CONFIG_WIRELESS_EXT
1902 const struct iw_handler_def *wireless_handlers;
1903 struct iw_public_data *wireless_data;
1905 const struct net_device_ops *netdev_ops;
1906 const struct ethtool_ops *ethtool_ops;
1907 #ifdef CONFIG_NET_L3_MASTER_DEV
1908 const struct l3mdev_ops *l3mdev_ops;
1910 #if IS_ENABLED(CONFIG_IPV6)
1911 const struct ndisc_ops *ndisc_ops;
1914 #ifdef CONFIG_XFRM_OFFLOAD
1915 const struct xfrmdev_ops *xfrmdev_ops;
1918 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1919 const struct tlsdev_ops *tlsdev_ops;
1922 const struct header_ops *header_ops;
1925 unsigned int priv_flags;
1927 unsigned short gflags;
1928 unsigned short padded;
1930 unsigned char operstate;
1931 unsigned char link_mode;
1933 unsigned char if_port;
1936 /* Note : dev->mtu is often read without holding a lock.
1937 * Writers usually hold RTNL.
1938 * It is recommended to use READ_ONCE() to annotate the reads,
1939 * and to use WRITE_ONCE() to annotate the writes.
1942 unsigned int min_mtu;
1943 unsigned int max_mtu;
1944 unsigned short type;
1945 unsigned short hard_header_len;
1946 unsigned char min_header_len;
1948 unsigned short needed_headroom;
1949 unsigned short needed_tailroom;
1951 /* Interface address info. */
1952 unsigned char perm_addr[MAX_ADDR_LEN];
1953 unsigned char addr_assign_type;
1954 unsigned char addr_len;
1955 unsigned char upper_level;
1956 unsigned char lower_level;
1957 unsigned short neigh_priv_len;
1958 unsigned short dev_id;
1959 unsigned short dev_port;
1960 spinlock_t addr_list_lock;
1961 unsigned char name_assign_type;
1963 struct netdev_hw_addr_list uc;
1964 struct netdev_hw_addr_list mc;
1965 struct netdev_hw_addr_list dev_addrs;
1968 struct kset *queues_kset;
1970 unsigned int promiscuity;
1971 unsigned int allmulti;
1974 /* Protocol-specific pointers */
1976 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1977 struct vlan_info __rcu *vlan_info;
1979 #if IS_ENABLED(CONFIG_NET_DSA)
1980 struct dsa_port *dsa_ptr;
1982 #if IS_ENABLED(CONFIG_TIPC)
1983 struct tipc_bearer __rcu *tipc_ptr;
1985 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1988 struct in_device __rcu *ip_ptr;
1989 #if IS_ENABLED(CONFIG_DECNET)
1990 struct dn_dev __rcu *dn_ptr;
1992 struct inet6_dev __rcu *ip6_ptr;
1993 #if IS_ENABLED(CONFIG_AX25)
1996 struct wireless_dev *ieee80211_ptr;
1997 struct wpan_dev *ieee802154_ptr;
1998 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1999 struct mpls_dev __rcu *mpls_ptr;
2003 * Cache lines mostly used on receive path (including eth_type_trans())
2005 /* Interface address info used in eth_type_trans() */
2006 unsigned char *dev_addr;
2008 struct netdev_rx_queue *_rx;
2009 unsigned int num_rx_queues;
2010 unsigned int real_num_rx_queues;
2012 struct bpf_prog __rcu *xdp_prog;
2013 unsigned long gro_flush_timeout;
2014 int napi_defer_hard_irqs;
2015 rx_handler_func_t __rcu *rx_handler;
2016 void __rcu *rx_handler_data;
2018 #ifdef CONFIG_NET_CLS_ACT
2019 struct mini_Qdisc __rcu *miniq_ingress;
2021 struct netdev_queue __rcu *ingress_queue;
2022 #ifdef CONFIG_NETFILTER_INGRESS
2023 struct nf_hook_entries __rcu *nf_hooks_ingress;
2026 unsigned char broadcast[MAX_ADDR_LEN];
2027 #ifdef CONFIG_RFS_ACCEL
2028 struct cpu_rmap *rx_cpu_rmap;
2030 struct hlist_node index_hlist;
2033 * Cache lines mostly used on transmit path
2035 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2036 unsigned int num_tx_queues;
2037 unsigned int real_num_tx_queues;
2038 struct Qdisc *qdisc;
2039 unsigned int tx_queue_len;
2040 spinlock_t tx_global_lock;
2042 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2045 struct xps_dev_maps __rcu *xps_cpus_map;
2046 struct xps_dev_maps __rcu *xps_rxqs_map;
2048 #ifdef CONFIG_NET_CLS_ACT
2049 struct mini_Qdisc __rcu *miniq_egress;
2052 #ifdef CONFIG_NET_SCHED
2053 DECLARE_HASHTABLE (qdisc_hash, 4);
2055 /* These may be needed for future network-power-down code. */
2056 struct timer_list watchdog_timer;
2059 u32 proto_down_reason;
2061 struct list_head todo_list;
2062 int __percpu *pcpu_refcnt;
2064 struct list_head link_watch_list;
2066 enum { NETREG_UNINITIALIZED=0,
2067 NETREG_REGISTERED, /* completed register_netdevice */
2068 NETREG_UNREGISTERING, /* called unregister_netdevice */
2069 NETREG_UNREGISTERED, /* completed unregister todo */
2070 NETREG_RELEASED, /* called free_netdev */
2071 NETREG_DUMMY, /* dummy device for NAPI poll */
2077 RTNL_LINK_INITIALIZED,
2078 RTNL_LINK_INITIALIZING,
2079 } rtnl_link_state:16;
2081 bool needs_free_netdev;
2082 void (*priv_destructor)(struct net_device *dev);
2084 #ifdef CONFIG_NETPOLL
2085 struct netpoll_info __rcu *npinfo;
2088 possible_net_t nd_net;
2090 /* mid-layer private */
2093 struct pcpu_lstats __percpu *lstats;
2094 struct pcpu_sw_netstats __percpu *tstats;
2095 struct pcpu_dstats __percpu *dstats;
2098 #if IS_ENABLED(CONFIG_GARP)
2099 struct garp_port __rcu *garp_port;
2101 #if IS_ENABLED(CONFIG_MRP)
2102 struct mrp_port __rcu *mrp_port;
2106 const struct attribute_group *sysfs_groups[4];
2107 const struct attribute_group *sysfs_rx_queue_group;
2109 const struct rtnl_link_ops *rtnl_link_ops;
2111 /* for setting kernel sock attribute on TCP connection setup */
2112 #define GSO_MAX_SIZE 65536
2113 unsigned int gso_max_size;
2114 #define GSO_MAX_SEGS 65535
2118 const struct dcbnl_rtnl_ops *dcbnl_ops;
2121 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2122 u8 prio_tc_map[TC_BITMASK + 1];
2124 #if IS_ENABLED(CONFIG_FCOE)
2125 unsigned int fcoe_ddp_xid;
2127 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2128 struct netprio_map __rcu *priomap;
2130 struct phy_device *phydev;
2131 struct sfp_bus *sfp_bus;
2132 struct lock_class_key *qdisc_tx_busylock;
2133 struct lock_class_key *qdisc_running_key;
2135 unsigned wol_enabled:1;
2137 struct list_head net_notifier_list;
2139 #if IS_ENABLED(CONFIG_MACSEC)
2140 /* MACsec management functions */
2141 const struct macsec_ops *macsec_ops;
2143 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2144 struct udp_tunnel_nic *udp_tunnel_nic;
2146 /* protected by rtnl_lock */
2147 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2149 #define to_net_dev(d) container_of(d, struct net_device, dev)
2151 static inline bool netif_elide_gro(const struct net_device *dev)
2153 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2158 #define NETDEV_ALIGN 32
2161 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2163 return dev->prio_tc_map[prio & TC_BITMASK];
2167 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2169 if (tc >= dev->num_tc)
2172 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2176 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2177 void netdev_reset_tc(struct net_device *dev);
2178 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2179 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2182 int netdev_get_num_tc(struct net_device *dev)
2187 static inline void net_prefetch(void *p)
2190 #if L1_CACHE_BYTES < 128
2191 prefetch((u8 *)p + L1_CACHE_BYTES);
2195 static inline void net_prefetchw(void *p)
2198 #if L1_CACHE_BYTES < 128
2199 prefetchw((u8 *)p + L1_CACHE_BYTES);
2203 void netdev_unbind_sb_channel(struct net_device *dev,
2204 struct net_device *sb_dev);
2205 int netdev_bind_sb_channel_queue(struct net_device *dev,
2206 struct net_device *sb_dev,
2207 u8 tc, u16 count, u16 offset);
2208 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2209 static inline int netdev_get_sb_channel(struct net_device *dev)
2211 return max_t(int, -dev->num_tc, 0);
2215 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2218 return &dev->_tx[index];
2221 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2222 const struct sk_buff *skb)
2224 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2227 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2228 void (*f)(struct net_device *,
2229 struct netdev_queue *,
2235 for (i = 0; i < dev->num_tx_queues; i++)
2236 f(dev, &dev->_tx[i], arg);
2239 #define netdev_lockdep_set_classes(dev) \
2241 static struct lock_class_key qdisc_tx_busylock_key; \
2242 static struct lock_class_key qdisc_running_key; \
2243 static struct lock_class_key qdisc_xmit_lock_key; \
2244 static struct lock_class_key dev_addr_list_lock_key; \
2247 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2248 (dev)->qdisc_running_key = &qdisc_running_key; \
2249 lockdep_set_class(&(dev)->addr_list_lock, \
2250 &dev_addr_list_lock_key); \
2251 for (i = 0; i < (dev)->num_tx_queues; i++) \
2252 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2253 &qdisc_xmit_lock_key); \
2256 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2257 struct net_device *sb_dev);
2258 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2259 struct sk_buff *skb,
2260 struct net_device *sb_dev);
2262 /* returns the headroom that the master device needs to take in account
2263 * when forwarding to this dev
2265 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2267 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2270 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2272 if (dev->netdev_ops->ndo_set_rx_headroom)
2273 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2276 /* set the device rx headroom to the dev's default */
2277 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2279 netdev_set_rx_headroom(dev, -1);
2283 * Net namespace inlines
2286 struct net *dev_net(const struct net_device *dev)
2288 return read_pnet(&dev->nd_net);
2292 void dev_net_set(struct net_device *dev, struct net *net)
2294 write_pnet(&dev->nd_net, net);
2298 * netdev_priv - access network device private data
2299 * @dev: network device
2301 * Get network device private data
2303 static inline void *netdev_priv(const struct net_device *dev)
2305 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2308 /* Set the sysfs physical device reference for the network logical device
2309 * if set prior to registration will cause a symlink during initialization.
2311 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2313 /* Set the sysfs device type for the network logical device to allow
2314 * fine-grained identification of different network device types. For
2315 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2317 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2319 /* Default NAPI poll() weight
2320 * Device drivers are strongly advised to not use bigger value
2322 #define NAPI_POLL_WEIGHT 64
2325 * netif_napi_add - initialize a NAPI context
2326 * @dev: network device
2327 * @napi: NAPI context
2328 * @poll: polling function
2329 * @weight: default weight
2331 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2332 * *any* of the other NAPI-related functions.
2334 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2335 int (*poll)(struct napi_struct *, int), int weight);
2338 * netif_tx_napi_add - initialize a NAPI context
2339 * @dev: network device
2340 * @napi: NAPI context
2341 * @poll: polling function
2342 * @weight: default weight
2344 * This variant of netif_napi_add() should be used from drivers using NAPI
2345 * to exclusively poll a TX queue.
2346 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2348 static inline void netif_tx_napi_add(struct net_device *dev,
2349 struct napi_struct *napi,
2350 int (*poll)(struct napi_struct *, int),
2353 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2354 netif_napi_add(dev, napi, poll, weight);
2358 * __netif_napi_del - remove a NAPI context
2359 * @napi: NAPI context
2361 * Warning: caller must observe RCU grace period before freeing memory
2362 * containing @napi. Drivers might want to call this helper to combine
2363 * all the needed RCU grace periods into a single one.
2365 void __netif_napi_del(struct napi_struct *napi);
2368 * netif_napi_del - remove a NAPI context
2369 * @napi: NAPI context
2371 * netif_napi_del() removes a NAPI context from the network device NAPI list
2373 static inline void netif_napi_del(struct napi_struct *napi)
2375 __netif_napi_del(napi);
2379 struct napi_gro_cb {
2380 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2383 /* Length of frag0. */
2384 unsigned int frag0_len;
2386 /* This indicates where we are processing relative to skb->data. */
2389 /* This is non-zero if the packet cannot be merged with the new skb. */
2392 /* Save the IP ID here and check when we get to the transport layer */
2395 /* Number of segments aggregated. */
2398 /* Start offset for remote checksum offload */
2399 u16 gro_remcsum_start;
2401 /* jiffies when first packet was created/queued */
2404 /* Used in ipv6_gro_receive() and foo-over-udp */
2407 /* This is non-zero if the packet may be of the same flow. */
2410 /* Used in tunnel GRO receive */
2413 /* GRO checksum is valid */
2416 /* Number of checksums via CHECKSUM_UNNECESSARY */
2421 #define NAPI_GRO_FREE 1
2422 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2424 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2427 /* Used in GRE, set in fou/gue_gro_receive */
2430 /* Used to determine if flush_id can be ignored */
2433 /* Number of gro_receive callbacks this packet already went through */
2434 u8 recursion_counter:4;
2436 /* GRO is done by frag_list pointer chaining. */
2439 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2442 /* used in skb_gro_receive() slow path */
2443 struct sk_buff *last;
2446 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2448 #define GRO_RECURSION_LIMIT 15
2449 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2451 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2454 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2455 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2456 struct list_head *head,
2457 struct sk_buff *skb)
2459 if (unlikely(gro_recursion_inc_test(skb))) {
2460 NAPI_GRO_CB(skb)->flush |= 1;
2464 return cb(head, skb);
2467 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2469 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2471 struct list_head *head,
2472 struct sk_buff *skb)
2474 if (unlikely(gro_recursion_inc_test(skb))) {
2475 NAPI_GRO_CB(skb)->flush |= 1;
2479 return cb(sk, head, skb);
2482 struct packet_type {
2483 __be16 type; /* This is really htons(ether_type). */
2484 bool ignore_outgoing;
2485 struct net_device *dev; /* NULL is wildcarded here */
2486 int (*func) (struct sk_buff *,
2487 struct net_device *,
2488 struct packet_type *,
2489 struct net_device *);
2490 void (*list_func) (struct list_head *,
2491 struct packet_type *,
2492 struct net_device *);
2493 bool (*id_match)(struct packet_type *ptype,
2495 void *af_packet_priv;
2496 struct list_head list;
2499 struct offload_callbacks {
2500 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2501 netdev_features_t features);
2502 struct sk_buff *(*gro_receive)(struct list_head *head,
2503 struct sk_buff *skb);
2504 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2507 struct packet_offload {
2508 __be16 type; /* This is really htons(ether_type). */
2510 struct offload_callbacks callbacks;
2511 struct list_head list;
2514 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2515 struct pcpu_sw_netstats {
2520 struct u64_stats_sync syncp;
2521 } __aligned(4 * sizeof(u64));
2523 struct pcpu_lstats {
2524 u64_stats_t packets;
2526 struct u64_stats_sync syncp;
2527 } __aligned(2 * sizeof(u64));
2529 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2531 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2533 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2535 u64_stats_update_begin(&lstats->syncp);
2536 u64_stats_add(&lstats->bytes, len);
2537 u64_stats_inc(&lstats->packets);
2538 u64_stats_update_end(&lstats->syncp);
2541 #define __netdev_alloc_pcpu_stats(type, gfp) \
2543 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2546 for_each_possible_cpu(__cpu) { \
2547 typeof(type) *stat; \
2548 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2549 u64_stats_init(&stat->syncp); \
2555 #define netdev_alloc_pcpu_stats(type) \
2556 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2558 enum netdev_lag_tx_type {
2559 NETDEV_LAG_TX_TYPE_UNKNOWN,
2560 NETDEV_LAG_TX_TYPE_RANDOM,
2561 NETDEV_LAG_TX_TYPE_BROADCAST,
2562 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2563 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2564 NETDEV_LAG_TX_TYPE_HASH,
2567 enum netdev_lag_hash {
2568 NETDEV_LAG_HASH_NONE,
2570 NETDEV_LAG_HASH_L34,
2571 NETDEV_LAG_HASH_L23,
2572 NETDEV_LAG_HASH_E23,
2573 NETDEV_LAG_HASH_E34,
2574 NETDEV_LAG_HASH_UNKNOWN,
2577 struct netdev_lag_upper_info {
2578 enum netdev_lag_tx_type tx_type;
2579 enum netdev_lag_hash hash_type;
2582 struct netdev_lag_lower_state_info {
2587 #include <linux/notifier.h>
2589 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2590 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2594 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2596 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2597 detected a hardware crash and restarted
2598 - we can use this eg to kick tcp sessions
2600 NETDEV_CHANGE, /* Notify device state change */
2603 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2604 NETDEV_CHANGEADDR, /* notify after the address change */
2605 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2609 NETDEV_BONDING_FAILOVER,
2611 NETDEV_PRE_TYPE_CHANGE,
2612 NETDEV_POST_TYPE_CHANGE,
2615 NETDEV_NOTIFY_PEERS,
2619 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2620 NETDEV_CHANGEINFODATA,
2621 NETDEV_BONDING_INFO,
2622 NETDEV_PRECHANGEUPPER,
2623 NETDEV_CHANGELOWERSTATE,
2624 NETDEV_UDP_TUNNEL_PUSH_INFO,
2625 NETDEV_UDP_TUNNEL_DROP_INFO,
2626 NETDEV_CHANGE_TX_QUEUE_LEN,
2627 NETDEV_CVLAN_FILTER_PUSH_INFO,
2628 NETDEV_CVLAN_FILTER_DROP_INFO,
2629 NETDEV_SVLAN_FILTER_PUSH_INFO,
2630 NETDEV_SVLAN_FILTER_DROP_INFO,
2632 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2634 int register_netdevice_notifier(struct notifier_block *nb);
2635 int unregister_netdevice_notifier(struct notifier_block *nb);
2636 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2637 int unregister_netdevice_notifier_net(struct net *net,
2638 struct notifier_block *nb);
2639 int register_netdevice_notifier_dev_net(struct net_device *dev,
2640 struct notifier_block *nb,
2641 struct netdev_net_notifier *nn);
2642 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2643 struct notifier_block *nb,
2644 struct netdev_net_notifier *nn);
2646 struct netdev_notifier_info {
2647 struct net_device *dev;
2648 struct netlink_ext_ack *extack;
2651 struct netdev_notifier_info_ext {
2652 struct netdev_notifier_info info; /* must be first */
2658 struct netdev_notifier_change_info {
2659 struct netdev_notifier_info info; /* must be first */
2660 unsigned int flags_changed;
2663 struct netdev_notifier_changeupper_info {
2664 struct netdev_notifier_info info; /* must be first */
2665 struct net_device *upper_dev; /* new upper dev */
2666 bool master; /* is upper dev master */
2667 bool linking; /* is the notification for link or unlink */
2668 void *upper_info; /* upper dev info */
2671 struct netdev_notifier_changelowerstate_info {
2672 struct netdev_notifier_info info; /* must be first */
2673 void *lower_state_info; /* is lower dev state */
2676 struct netdev_notifier_pre_changeaddr_info {
2677 struct netdev_notifier_info info; /* must be first */
2678 const unsigned char *dev_addr;
2681 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2682 struct net_device *dev)
2685 info->extack = NULL;
2688 static inline struct net_device *
2689 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2694 static inline struct netlink_ext_ack *
2695 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2697 return info->extack;
2700 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2703 extern rwlock_t dev_base_lock; /* Device list lock */
2705 #define for_each_netdev(net, d) \
2706 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2707 #define for_each_netdev_reverse(net, d) \
2708 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2709 #define for_each_netdev_rcu(net, d) \
2710 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2711 #define for_each_netdev_safe(net, d, n) \
2712 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2713 #define for_each_netdev_continue(net, d) \
2714 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2715 #define for_each_netdev_continue_reverse(net, d) \
2716 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2718 #define for_each_netdev_continue_rcu(net, d) \
2719 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2720 #define for_each_netdev_in_bond_rcu(bond, slave) \
2721 for_each_netdev_rcu(&init_net, slave) \
2722 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2723 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2725 static inline struct net_device *next_net_device(struct net_device *dev)
2727 struct list_head *lh;
2731 lh = dev->dev_list.next;
2732 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2735 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2737 struct list_head *lh;
2741 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2742 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2745 static inline struct net_device *first_net_device(struct net *net)
2747 return list_empty(&net->dev_base_head) ? NULL :
2748 net_device_entry(net->dev_base_head.next);
2751 static inline struct net_device *first_net_device_rcu(struct net *net)
2753 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2755 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2758 int netdev_boot_setup_check(struct net_device *dev);
2759 unsigned long netdev_boot_base(const char *prefix, int unit);
2760 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2761 const char *hwaddr);
2762 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2763 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2764 void dev_add_pack(struct packet_type *pt);
2765 void dev_remove_pack(struct packet_type *pt);
2766 void __dev_remove_pack(struct packet_type *pt);
2767 void dev_add_offload(struct packet_offload *po);
2768 void dev_remove_offload(struct packet_offload *po);
2770 int dev_get_iflink(const struct net_device *dev);
2771 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2772 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2773 unsigned short mask);
2774 struct net_device *dev_get_by_name(struct net *net, const char *name);
2775 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2776 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2777 int dev_alloc_name(struct net_device *dev, const char *name);
2778 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2779 void dev_close(struct net_device *dev);
2780 void dev_close_many(struct list_head *head, bool unlink);
2781 void dev_disable_lro(struct net_device *dev);
2782 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2783 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2784 struct net_device *sb_dev);
2785 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2786 struct net_device *sb_dev);
2787 int dev_queue_xmit(struct sk_buff *skb);
2788 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2789 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2790 int register_netdevice(struct net_device *dev);
2791 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2792 void unregister_netdevice_many(struct list_head *head);
2793 static inline void unregister_netdevice(struct net_device *dev)
2795 unregister_netdevice_queue(dev, NULL);
2798 int netdev_refcnt_read(const struct net_device *dev);
2799 void free_netdev(struct net_device *dev);
2800 void netdev_freemem(struct net_device *dev);
2801 int init_dummy_netdev(struct net_device *dev);
2803 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2804 struct sk_buff *skb,
2806 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2807 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2808 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2809 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2810 int netdev_get_name(struct net *net, char *name, int ifindex);
2811 int dev_restart(struct net_device *dev);
2812 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2813 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2815 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2817 return NAPI_GRO_CB(skb)->data_offset;
2820 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2822 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2825 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2827 NAPI_GRO_CB(skb)->data_offset += len;
2830 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2831 unsigned int offset)
2833 return NAPI_GRO_CB(skb)->frag0 + offset;
2836 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2838 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2841 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2843 NAPI_GRO_CB(skb)->frag0 = NULL;
2844 NAPI_GRO_CB(skb)->frag0_len = 0;
2847 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2848 unsigned int offset)
2850 if (!pskb_may_pull(skb, hlen))
2853 skb_gro_frag0_invalidate(skb);
2854 return skb->data + offset;
2857 static inline void *skb_gro_network_header(struct sk_buff *skb)
2859 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2860 skb_network_offset(skb);
2863 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2864 const void *start, unsigned int len)
2866 if (NAPI_GRO_CB(skb)->csum_valid)
2867 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2868 csum_partial(start, len, 0));
2871 /* GRO checksum functions. These are logical equivalents of the normal
2872 * checksum functions (in skbuff.h) except that they operate on the GRO
2873 * offsets and fields in sk_buff.
2876 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2878 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2880 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2883 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2887 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2888 skb_checksum_start_offset(skb) <
2889 skb_gro_offset(skb)) &&
2890 !skb_at_gro_remcsum_start(skb) &&
2891 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2892 (!zero_okay || check));
2895 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2898 if (NAPI_GRO_CB(skb)->csum_valid &&
2899 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2902 NAPI_GRO_CB(skb)->csum = psum;
2904 return __skb_gro_checksum_complete(skb);
2907 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2909 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2910 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2911 NAPI_GRO_CB(skb)->csum_cnt--;
2913 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2914 * verified a new top level checksum or an encapsulated one
2915 * during GRO. This saves work if we fallback to normal path.
2917 __skb_incr_checksum_unnecessary(skb);
2921 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2924 __sum16 __ret = 0; \
2925 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2926 __ret = __skb_gro_checksum_validate_complete(skb, \
2927 compute_pseudo(skb, proto)); \
2929 skb_gro_incr_csum_unnecessary(skb); \
2933 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2934 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2936 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2938 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2940 #define skb_gro_checksum_simple_validate(skb) \
2941 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2943 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2945 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2946 !NAPI_GRO_CB(skb)->csum_valid);
2949 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2952 NAPI_GRO_CB(skb)->csum = ~pseudo;
2953 NAPI_GRO_CB(skb)->csum_valid = 1;
2956 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
2958 if (__skb_gro_checksum_convert_check(skb)) \
2959 __skb_gro_checksum_convert(skb, \
2960 compute_pseudo(skb, proto)); \
2963 struct gro_remcsum {
2968 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2974 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2975 unsigned int off, size_t hdrlen,
2976 int start, int offset,
2977 struct gro_remcsum *grc,
2981 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2983 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2986 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2990 ptr = skb_gro_header_fast(skb, off);
2991 if (skb_gro_header_hard(skb, off + plen)) {
2992 ptr = skb_gro_header_slow(skb, off + plen, off);
2997 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3000 /* Adjust skb->csum since we changed the packet */
3001 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3003 grc->offset = off + hdrlen + offset;
3009 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3010 struct gro_remcsum *grc)
3013 size_t plen = grc->offset + sizeof(u16);
3018 ptr = skb_gro_header_fast(skb, grc->offset);
3019 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3020 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3025 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3028 #ifdef CONFIG_XFRM_OFFLOAD
3029 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3031 if (PTR_ERR(pp) != -EINPROGRESS)
3032 NAPI_GRO_CB(skb)->flush |= flush;
3034 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3037 struct gro_remcsum *grc)
3039 if (PTR_ERR(pp) != -EINPROGRESS) {
3040 NAPI_GRO_CB(skb)->flush |= flush;
3041 skb_gro_remcsum_cleanup(skb, grc);
3042 skb->remcsum_offload = 0;
3046 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3048 NAPI_GRO_CB(skb)->flush |= flush;
3050 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3053 struct gro_remcsum *grc)
3055 NAPI_GRO_CB(skb)->flush |= flush;
3056 skb_gro_remcsum_cleanup(skb, grc);
3057 skb->remcsum_offload = 0;
3061 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3062 unsigned short type,
3063 const void *daddr, const void *saddr,
3066 if (!dev->header_ops || !dev->header_ops->create)
3069 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3072 static inline int dev_parse_header(const struct sk_buff *skb,
3073 unsigned char *haddr)
3075 const struct net_device *dev = skb->dev;
3077 if (!dev->header_ops || !dev->header_ops->parse)
3079 return dev->header_ops->parse(skb, haddr);
3082 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3084 const struct net_device *dev = skb->dev;
3086 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3088 return dev->header_ops->parse_protocol(skb);
3091 /* ll_header must have at least hard_header_len allocated */
3092 static inline bool dev_validate_header(const struct net_device *dev,
3093 char *ll_header, int len)
3095 if (likely(len >= dev->hard_header_len))
3097 if (len < dev->min_header_len)
3100 if (capable(CAP_SYS_RAWIO)) {
3101 memset(ll_header + len, 0, dev->hard_header_len - len);
3105 if (dev->header_ops && dev->header_ops->validate)
3106 return dev->header_ops->validate(ll_header, len);
3111 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3113 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3114 static inline int unregister_gifconf(unsigned int family)
3116 return register_gifconf(family, NULL);
3119 #ifdef CONFIG_NET_FLOW_LIMIT
3120 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3121 struct sd_flow_limit {
3123 unsigned int num_buckets;
3124 unsigned int history_head;
3125 u16 history[FLOW_LIMIT_HISTORY];
3129 extern int netdev_flow_limit_table_len;
3130 #endif /* CONFIG_NET_FLOW_LIMIT */
3133 * Incoming packets are placed on per-CPU queues
3135 struct softnet_data {
3136 struct list_head poll_list;
3137 struct sk_buff_head process_queue;
3140 unsigned int processed;
3141 unsigned int time_squeeze;
3142 unsigned int received_rps;
3144 struct softnet_data *rps_ipi_list;
3146 #ifdef CONFIG_NET_FLOW_LIMIT
3147 struct sd_flow_limit __rcu *flow_limit;
3149 struct Qdisc *output_queue;
3150 struct Qdisc **output_queue_tailp;
3151 struct sk_buff *completion_queue;
3152 #ifdef CONFIG_XFRM_OFFLOAD
3153 struct sk_buff_head xfrm_backlog;
3155 /* written and read only by owning cpu: */
3161 /* input_queue_head should be written by cpu owning this struct,
3162 * and only read by other cpus. Worth using a cache line.
3164 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3166 /* Elements below can be accessed between CPUs for RPS/RFS */
3167 call_single_data_t csd ____cacheline_aligned_in_smp;
3168 struct softnet_data *rps_ipi_next;
3170 unsigned int input_queue_tail;
3172 unsigned int dropped;
3173 struct sk_buff_head input_pkt_queue;
3174 struct napi_struct backlog;
3178 static inline void input_queue_head_incr(struct softnet_data *sd)
3181 sd->input_queue_head++;
3185 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3186 unsigned int *qtail)
3189 *qtail = ++sd->input_queue_tail;
3193 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3195 static inline int dev_recursion_level(void)
3197 return this_cpu_read(softnet_data.xmit.recursion);
3200 #define XMIT_RECURSION_LIMIT 8
3201 static inline bool dev_xmit_recursion(void)
3203 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3204 XMIT_RECURSION_LIMIT);
3207 static inline void dev_xmit_recursion_inc(void)
3209 __this_cpu_inc(softnet_data.xmit.recursion);
3212 static inline void dev_xmit_recursion_dec(void)
3214 __this_cpu_dec(softnet_data.xmit.recursion);
3217 void __netif_schedule(struct Qdisc *q);
3218 void netif_schedule_queue(struct netdev_queue *txq);
3220 static inline void netif_tx_schedule_all(struct net_device *dev)
3224 for (i = 0; i < dev->num_tx_queues; i++)
3225 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3228 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3230 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3234 * netif_start_queue - allow transmit
3235 * @dev: network device
3237 * Allow upper layers to call the device hard_start_xmit routine.
3239 static inline void netif_start_queue(struct net_device *dev)
3241 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3244 static inline void netif_tx_start_all_queues(struct net_device *dev)
3248 for (i = 0; i < dev->num_tx_queues; i++) {
3249 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3250 netif_tx_start_queue(txq);
3254 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3257 * netif_wake_queue - restart transmit
3258 * @dev: network device
3260 * Allow upper layers to call the device hard_start_xmit routine.
3261 * Used for flow control when transmit resources are available.
3263 static inline void netif_wake_queue(struct net_device *dev)
3265 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3268 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3272 for (i = 0; i < dev->num_tx_queues; i++) {
3273 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3274 netif_tx_wake_queue(txq);
3278 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3280 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3284 * netif_stop_queue - stop transmitted packets
3285 * @dev: network device
3287 * Stop upper layers calling the device hard_start_xmit routine.
3288 * Used for flow control when transmit resources are unavailable.
3290 static inline void netif_stop_queue(struct net_device *dev)
3292 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3295 void netif_tx_stop_all_queues(struct net_device *dev);
3297 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3299 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3303 * netif_queue_stopped - test if transmit queue is flowblocked
3304 * @dev: network device
3306 * Test if transmit queue on device is currently unable to send.
3308 static inline bool netif_queue_stopped(const struct net_device *dev)
3310 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3313 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3315 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3319 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3321 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3325 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3327 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3331 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3332 * @dev_queue: pointer to transmit queue
3334 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3335 * to give appropriate hint to the CPU.
3337 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3340 prefetchw(&dev_queue->dql.num_queued);
3345 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3346 * @dev_queue: pointer to transmit queue
3348 * BQL enabled drivers might use this helper in their TX completion path,
3349 * to give appropriate hint to the CPU.
3351 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3354 prefetchw(&dev_queue->dql.limit);
3358 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3362 dql_queued(&dev_queue->dql, bytes);
3364 if (likely(dql_avail(&dev_queue->dql) >= 0))
3367 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3370 * The XOFF flag must be set before checking the dql_avail below,
3371 * because in netdev_tx_completed_queue we update the dql_completed
3372 * before checking the XOFF flag.
3376 /* check again in case another CPU has just made room avail */
3377 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3378 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3382 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3383 * that they should not test BQL status themselves.
3384 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3386 * Returns true if the doorbell must be used to kick the NIC.
3388 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3394 dql_queued(&dev_queue->dql, bytes);
3396 return netif_tx_queue_stopped(dev_queue);
3398 netdev_tx_sent_queue(dev_queue, bytes);
3403 * netdev_sent_queue - report the number of bytes queued to hardware
3404 * @dev: network device
3405 * @bytes: number of bytes queued to the hardware device queue
3407 * Report the number of bytes queued for sending/completion to the network
3408 * device hardware queue. @bytes should be a good approximation and should
3409 * exactly match netdev_completed_queue() @bytes
3411 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3413 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3416 static inline bool __netdev_sent_queue(struct net_device *dev,
3420 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3424 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3425 unsigned int pkts, unsigned int bytes)
3428 if (unlikely(!bytes))
3431 dql_completed(&dev_queue->dql, bytes);
3434 * Without the memory barrier there is a small possiblity that
3435 * netdev_tx_sent_queue will miss the update and cause the queue to
3436 * be stopped forever
3440 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3443 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3444 netif_schedule_queue(dev_queue);
3449 * netdev_completed_queue - report bytes and packets completed by device
3450 * @dev: network device
3451 * @pkts: actual number of packets sent over the medium
3452 * @bytes: actual number of bytes sent over the medium
3454 * Report the number of bytes and packets transmitted by the network device
3455 * hardware queue over the physical medium, @bytes must exactly match the
3456 * @bytes amount passed to netdev_sent_queue()
3458 static inline void netdev_completed_queue(struct net_device *dev,
3459 unsigned int pkts, unsigned int bytes)
3461 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3464 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3467 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3473 * netdev_reset_queue - reset the packets and bytes count of a network device
3474 * @dev_queue: network device
3476 * Reset the bytes and packet count of a network device and clear the
3477 * software flow control OFF bit for this network device
3479 static inline void netdev_reset_queue(struct net_device *dev_queue)
3481 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3485 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3486 * @dev: network device
3487 * @queue_index: given tx queue index
3489 * Returns 0 if given tx queue index >= number of device tx queues,
3490 * otherwise returns the originally passed tx queue index.
3492 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3494 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3495 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3496 dev->name, queue_index,
3497 dev->real_num_tx_queues);
3505 * netif_running - test if up
3506 * @dev: network device
3508 * Test if the device has been brought up.
3510 static inline bool netif_running(const struct net_device *dev)
3512 return test_bit(__LINK_STATE_START, &dev->state);
3516 * Routines to manage the subqueues on a device. We only need start,
3517 * stop, and a check if it's stopped. All other device management is
3518 * done at the overall netdevice level.
3519 * Also test the device if we're multiqueue.
3523 * netif_start_subqueue - allow sending packets on subqueue
3524 * @dev: network device
3525 * @queue_index: sub queue index
3527 * Start individual transmit queue of a device with multiple transmit queues.
3529 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3531 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3533 netif_tx_start_queue(txq);
3537 * netif_stop_subqueue - stop sending packets on subqueue
3538 * @dev: network device
3539 * @queue_index: sub queue index
3541 * Stop individual transmit queue of a device with multiple transmit queues.
3543 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3545 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3546 netif_tx_stop_queue(txq);
3550 * netif_subqueue_stopped - test status of subqueue
3551 * @dev: network device
3552 * @queue_index: sub queue index
3554 * Check individual transmit queue of a device with multiple transmit queues.
3556 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3559 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3561 return netif_tx_queue_stopped(txq);
3564 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3565 struct sk_buff *skb)
3567 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3571 * netif_wake_subqueue - allow sending packets on subqueue
3572 * @dev: network device
3573 * @queue_index: sub queue index
3575 * Resume individual transmit queue of a device with multiple transmit queues.
3577 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3579 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3581 netif_tx_wake_queue(txq);
3585 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3587 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3588 u16 index, bool is_rxqs_map);
3591 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3592 * @j: CPU/Rx queue index
3593 * @mask: bitmask of all cpus/rx queues
3594 * @nr_bits: number of bits in the bitmask
3596 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3598 static inline bool netif_attr_test_mask(unsigned long j,
3599 const unsigned long *mask,
3600 unsigned int nr_bits)
3602 cpu_max_bits_warn(j, nr_bits);
3603 return test_bit(j, mask);
3607 * netif_attr_test_online - Test for online CPU/Rx queue
3608 * @j: CPU/Rx queue index
3609 * @online_mask: bitmask for CPUs/Rx queues that are online
3610 * @nr_bits: number of bits in the bitmask
3612 * Returns true if a CPU/Rx queue is online.
3614 static inline bool netif_attr_test_online(unsigned long j,
3615 const unsigned long *online_mask,
3616 unsigned int nr_bits)
3618 cpu_max_bits_warn(j, nr_bits);
3621 return test_bit(j, online_mask);
3623 return (j < nr_bits);
3627 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3628 * @n: CPU/Rx queue index
3629 * @srcp: the cpumask/Rx queue mask pointer
3630 * @nr_bits: number of bits in the bitmask
3632 * Returns >= nr_bits if no further CPUs/Rx queues set.
3634 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3635 unsigned int nr_bits)
3637 /* -1 is a legal arg here. */
3639 cpu_max_bits_warn(n, nr_bits);
3642 return find_next_bit(srcp, nr_bits, n + 1);
3648 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3649 * @n: CPU/Rx queue index
3650 * @src1p: the first CPUs/Rx queues mask pointer
3651 * @src2p: the second CPUs/Rx queues mask pointer
3652 * @nr_bits: number of bits in the bitmask
3654 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3656 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3657 const unsigned long *src2p,
3658 unsigned int nr_bits)
3660 /* -1 is a legal arg here. */
3662 cpu_max_bits_warn(n, nr_bits);
3665 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3667 return find_next_bit(src1p, nr_bits, n + 1);
3669 return find_next_bit(src2p, nr_bits, n + 1);
3674 static inline int netif_set_xps_queue(struct net_device *dev,
3675 const struct cpumask *mask,
3681 static inline int __netif_set_xps_queue(struct net_device *dev,
3682 const unsigned long *mask,
3683 u16 index, bool is_rxqs_map)
3690 * netif_is_multiqueue - test if device has multiple transmit queues
3691 * @dev: network device
3693 * Check if device has multiple transmit queues
3695 static inline bool netif_is_multiqueue(const struct net_device *dev)
3697 return dev->num_tx_queues > 1;
3700 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3703 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3705 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3708 dev->real_num_rx_queues = rxqs;
3713 static inline struct netdev_rx_queue *
3714 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3716 return dev->_rx + rxq;
3720 static inline unsigned int get_netdev_rx_queue_index(
3721 struct netdev_rx_queue *queue)
3723 struct net_device *dev = queue->dev;
3724 int index = queue - dev->_rx;
3726 BUG_ON(index >= dev->num_rx_queues);
3731 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3732 int netif_get_num_default_rss_queues(void);
3734 enum skb_free_reason {
3735 SKB_REASON_CONSUMED,
3739 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3740 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3743 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3744 * interrupt context or with hardware interrupts being disabled.
3745 * (in_irq() || irqs_disabled())
3747 * We provide four helpers that can be used in following contexts :
3749 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3750 * replacing kfree_skb(skb)
3752 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3753 * Typically used in place of consume_skb(skb) in TX completion path
3755 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3756 * replacing kfree_skb(skb)
3758 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3759 * and consumed a packet. Used in place of consume_skb(skb)
3761 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3763 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3766 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3768 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3771 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3773 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3776 static inline void dev_consume_skb_any(struct sk_buff *skb)
3778 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3781 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3782 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3783 int netif_rx(struct sk_buff *skb);
3784 int netif_rx_ni(struct sk_buff *skb);
3785 int netif_receive_skb(struct sk_buff *skb);
3786 int netif_receive_skb_core(struct sk_buff *skb);
3787 void netif_receive_skb_list(struct list_head *head);
3788 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3789 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3790 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3791 gro_result_t napi_gro_frags(struct napi_struct *napi);
3792 struct packet_offload *gro_find_receive_by_type(__be16 type);
3793 struct packet_offload *gro_find_complete_by_type(__be16 type);
3795 static inline void napi_free_frags(struct napi_struct *napi)
3797 kfree_skb(napi->skb);
3801 bool netdev_is_rx_handler_busy(struct net_device *dev);
3802 int netdev_rx_handler_register(struct net_device *dev,
3803 rx_handler_func_t *rx_handler,
3804 void *rx_handler_data);
3805 void netdev_rx_handler_unregister(struct net_device *dev);
3807 bool dev_valid_name(const char *name);
3808 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3809 bool *need_copyout);
3810 int dev_ifconf(struct net *net, struct ifconf *, int);
3811 int dev_ethtool(struct net *net, struct ifreq *);
3812 unsigned int dev_get_flags(const struct net_device *);
3813 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3814 struct netlink_ext_ack *extack);
3815 int dev_change_flags(struct net_device *dev, unsigned int flags,
3816 struct netlink_ext_ack *extack);
3817 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3818 unsigned int gchanges);
3819 int dev_change_name(struct net_device *, const char *);
3820 int dev_set_alias(struct net_device *, const char *, size_t);
3821 int dev_get_alias(const struct net_device *, char *, size_t);
3822 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3823 int __dev_set_mtu(struct net_device *, int);
3824 int dev_validate_mtu(struct net_device *dev, int mtu,
3825 struct netlink_ext_ack *extack);
3826 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3827 struct netlink_ext_ack *extack);
3828 int dev_set_mtu(struct net_device *, int);
3829 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3830 void dev_set_group(struct net_device *, int);
3831 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3832 struct netlink_ext_ack *extack);
3833 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3834 struct netlink_ext_ack *extack);
3835 int dev_change_carrier(struct net_device *, bool new_carrier);
3836 int dev_get_phys_port_id(struct net_device *dev,
3837 struct netdev_phys_item_id *ppid);
3838 int dev_get_phys_port_name(struct net_device *dev,
3839 char *name, size_t len);
3840 int dev_get_port_parent_id(struct net_device *dev,
3841 struct netdev_phys_item_id *ppid, bool recurse);
3842 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3843 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3844 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3845 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3847 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3848 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3849 struct netdev_queue *txq, int *ret);
3851 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3852 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3853 int fd, int expected_fd, u32 flags);
3854 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3855 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3857 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3859 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3860 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3861 bool is_skb_forwardable(const struct net_device *dev,
3862 const struct sk_buff *skb);
3864 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3865 struct sk_buff *skb)
3867 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3868 unlikely(!is_skb_forwardable(dev, skb))) {
3869 atomic_long_inc(&dev->rx_dropped);
3874 skb_scrub_packet(skb, true);
3879 bool dev_nit_active(struct net_device *dev);
3880 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3882 extern int netdev_budget;
3883 extern unsigned int netdev_budget_usecs;
3885 /* Called by rtnetlink.c:rtnl_unlock() */
3886 void netdev_run_todo(void);
3889 * dev_put - release reference to device
3890 * @dev: network device
3892 * Release reference to device to allow it to be freed.
3894 static inline void dev_put(struct net_device *dev)
3896 this_cpu_dec(*dev->pcpu_refcnt);
3900 * dev_hold - get reference to device
3901 * @dev: network device
3903 * Hold reference to device to keep it from being freed.
3905 static inline void dev_hold(struct net_device *dev)
3907 this_cpu_inc(*dev->pcpu_refcnt);
3910 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3911 * and _off may be called from IRQ context, but it is caller
3912 * who is responsible for serialization of these calls.
3914 * The name carrier is inappropriate, these functions should really be
3915 * called netif_lowerlayer_*() because they represent the state of any
3916 * kind of lower layer not just hardware media.
3919 void linkwatch_init_dev(struct net_device *dev);
3920 void linkwatch_fire_event(struct net_device *dev);
3921 void linkwatch_forget_dev(struct net_device *dev);
3924 * netif_carrier_ok - test if carrier present
3925 * @dev: network device
3927 * Check if carrier is present on device
3929 static inline bool netif_carrier_ok(const struct net_device *dev)
3931 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3934 unsigned long dev_trans_start(struct net_device *dev);
3936 void __netdev_watchdog_up(struct net_device *dev);
3938 void netif_carrier_on(struct net_device *dev);
3940 void netif_carrier_off(struct net_device *dev);
3943 * netif_dormant_on - mark device as dormant.
3944 * @dev: network device
3946 * Mark device as dormant (as per RFC2863).
3948 * The dormant state indicates that the relevant interface is not
3949 * actually in a condition to pass packets (i.e., it is not 'up') but is
3950 * in a "pending" state, waiting for some external event. For "on-
3951 * demand" interfaces, this new state identifies the situation where the
3952 * interface is waiting for events to place it in the up state.
3954 static inline void netif_dormant_on(struct net_device *dev)
3956 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3957 linkwatch_fire_event(dev);
3961 * netif_dormant_off - set device as not dormant.
3962 * @dev: network device
3964 * Device is not in dormant state.
3966 static inline void netif_dormant_off(struct net_device *dev)
3968 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3969 linkwatch_fire_event(dev);
3973 * netif_dormant - test if device is dormant
3974 * @dev: network device
3976 * Check if device is dormant.
3978 static inline bool netif_dormant(const struct net_device *dev)
3980 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3985 * netif_testing_on - mark device as under test.
3986 * @dev: network device
3988 * Mark device as under test (as per RFC2863).
3990 * The testing state indicates that some test(s) must be performed on
3991 * the interface. After completion, of the test, the interface state
3992 * will change to up, dormant, or down, as appropriate.
3994 static inline void netif_testing_on(struct net_device *dev)
3996 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
3997 linkwatch_fire_event(dev);
4001 * netif_testing_off - set device as not under test.
4002 * @dev: network device
4004 * Device is not in testing state.
4006 static inline void netif_testing_off(struct net_device *dev)
4008 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4009 linkwatch_fire_event(dev);
4013 * netif_testing - test if device is under test
4014 * @dev: network device
4016 * Check if device is under test
4018 static inline bool netif_testing(const struct net_device *dev)
4020 return test_bit(__LINK_STATE_TESTING, &dev->state);
4025 * netif_oper_up - test if device is operational
4026 * @dev: network device
4028 * Check if carrier is operational
4030 static inline bool netif_oper_up(const struct net_device *dev)
4032 return (dev->operstate == IF_OPER_UP ||
4033 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4037 * netif_device_present - is device available or removed
4038 * @dev: network device
4040 * Check if device has not been removed from system.
4042 static inline bool netif_device_present(struct net_device *dev)
4044 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4047 void netif_device_detach(struct net_device *dev);
4049 void netif_device_attach(struct net_device *dev);
4052 * Network interface message level settings
4057 NETIF_MSG_PROBE_BIT,
4059 NETIF_MSG_TIMER_BIT,
4060 NETIF_MSG_IFDOWN_BIT,
4062 NETIF_MSG_RX_ERR_BIT,
4063 NETIF_MSG_TX_ERR_BIT,
4064 NETIF_MSG_TX_QUEUED_BIT,
4066 NETIF_MSG_TX_DONE_BIT,
4067 NETIF_MSG_RX_STATUS_BIT,
4068 NETIF_MSG_PKTDATA_BIT,
4072 /* When you add a new bit above, update netif_msg_class_names array
4073 * in net/ethtool/common.c
4075 NETIF_MSG_CLASS_COUNT,
4077 /* Both ethtool_ops interface and internal driver implementation use u32 */
4078 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4080 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4081 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4083 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4084 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4085 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4086 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4087 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4088 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4089 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4090 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4091 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4092 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4093 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4094 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4095 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4096 #define NETIF_MSG_HW __NETIF_MSG(HW)
4097 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4099 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4100 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4101 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4102 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4103 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4104 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4105 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4106 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4107 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4108 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4109 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4110 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4111 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4112 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4113 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4115 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4118 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4119 return default_msg_enable_bits;
4120 if (debug_value == 0) /* no output */
4122 /* set low N bits */
4123 return (1U << debug_value) - 1;
4126 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4128 spin_lock(&txq->_xmit_lock);
4129 txq->xmit_lock_owner = cpu;
4132 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4134 __acquire(&txq->_xmit_lock);
4138 static inline void __netif_tx_release(struct netdev_queue *txq)
4140 __release(&txq->_xmit_lock);
4143 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4145 spin_lock_bh(&txq->_xmit_lock);
4146 txq->xmit_lock_owner = smp_processor_id();
4149 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4151 bool ok = spin_trylock(&txq->_xmit_lock);
4153 txq->xmit_lock_owner = smp_processor_id();
4157 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4159 txq->xmit_lock_owner = -1;
4160 spin_unlock(&txq->_xmit_lock);
4163 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4165 txq->xmit_lock_owner = -1;
4166 spin_unlock_bh(&txq->_xmit_lock);
4169 static inline void txq_trans_update(struct netdev_queue *txq)
4171 if (txq->xmit_lock_owner != -1)
4172 txq->trans_start = jiffies;
4175 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4176 static inline void netif_trans_update(struct net_device *dev)
4178 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4180 if (txq->trans_start != jiffies)
4181 txq->trans_start = jiffies;
4185 * netif_tx_lock - grab network device transmit lock
4186 * @dev: network device
4188 * Get network device transmit lock
4190 static inline void netif_tx_lock(struct net_device *dev)
4195 spin_lock(&dev->tx_global_lock);
4196 cpu = smp_processor_id();
4197 for (i = 0; i < dev->num_tx_queues; i++) {
4198 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4200 /* We are the only thread of execution doing a
4201 * freeze, but we have to grab the _xmit_lock in
4202 * order to synchronize with threads which are in
4203 * the ->hard_start_xmit() handler and already
4204 * checked the frozen bit.
4206 __netif_tx_lock(txq, cpu);
4207 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4208 __netif_tx_unlock(txq);
4212 static inline void netif_tx_lock_bh(struct net_device *dev)
4218 static inline void netif_tx_unlock(struct net_device *dev)
4222 for (i = 0; i < dev->num_tx_queues; i++) {
4223 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4225 /* No need to grab the _xmit_lock here. If the
4226 * queue is not stopped for another reason, we
4229 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4230 netif_schedule_queue(txq);
4232 spin_unlock(&dev->tx_global_lock);
4235 static inline void netif_tx_unlock_bh(struct net_device *dev)
4237 netif_tx_unlock(dev);
4241 #define HARD_TX_LOCK(dev, txq, cpu) { \
4242 if ((dev->features & NETIF_F_LLTX) == 0) { \
4243 __netif_tx_lock(txq, cpu); \
4245 __netif_tx_acquire(txq); \
4249 #define HARD_TX_TRYLOCK(dev, txq) \
4250 (((dev->features & NETIF_F_LLTX) == 0) ? \
4251 __netif_tx_trylock(txq) : \
4252 __netif_tx_acquire(txq))
4254 #define HARD_TX_UNLOCK(dev, txq) { \
4255 if ((dev->features & NETIF_F_LLTX) == 0) { \
4256 __netif_tx_unlock(txq); \
4258 __netif_tx_release(txq); \
4262 static inline void netif_tx_disable(struct net_device *dev)
4268 cpu = smp_processor_id();
4269 for (i = 0; i < dev->num_tx_queues; i++) {
4270 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4272 __netif_tx_lock(txq, cpu);
4273 netif_tx_stop_queue(txq);
4274 __netif_tx_unlock(txq);
4279 static inline void netif_addr_lock(struct net_device *dev)
4281 spin_lock(&dev->addr_list_lock);
4284 static inline void netif_addr_lock_nested(struct net_device *dev)
4286 spin_lock_nested(&dev->addr_list_lock, dev->lower_level);
4289 static inline void netif_addr_lock_bh(struct net_device *dev)
4291 spin_lock_bh(&dev->addr_list_lock);
4294 static inline void netif_addr_unlock(struct net_device *dev)
4296 spin_unlock(&dev->addr_list_lock);
4299 static inline void netif_addr_unlock_bh(struct net_device *dev)
4301 spin_unlock_bh(&dev->addr_list_lock);
4305 * dev_addrs walker. Should be used only for read access. Call with
4306 * rcu_read_lock held.
4308 #define for_each_dev_addr(dev, ha) \
4309 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4311 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4313 void ether_setup(struct net_device *dev);
4315 /* Support for loadable net-drivers */
4316 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4317 unsigned char name_assign_type,
4318 void (*setup)(struct net_device *),
4319 unsigned int txqs, unsigned int rxqs);
4320 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4321 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4323 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4324 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4327 int register_netdev(struct net_device *dev);
4328 void unregister_netdev(struct net_device *dev);
4330 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4332 /* General hardware address lists handling functions */
4333 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4334 struct netdev_hw_addr_list *from_list, int addr_len);
4335 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4336 struct netdev_hw_addr_list *from_list, int addr_len);
4337 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4338 struct net_device *dev,
4339 int (*sync)(struct net_device *, const unsigned char *),
4340 int (*unsync)(struct net_device *,
4341 const unsigned char *));
4342 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4343 struct net_device *dev,
4344 int (*sync)(struct net_device *,
4345 const unsigned char *, int),
4346 int (*unsync)(struct net_device *,
4347 const unsigned char *, int));
4348 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4349 struct net_device *dev,
4350 int (*unsync)(struct net_device *,
4351 const unsigned char *, int));
4352 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4353 struct net_device *dev,
4354 int (*unsync)(struct net_device *,
4355 const unsigned char *));
4356 void __hw_addr_init(struct netdev_hw_addr_list *list);
4358 /* Functions used for device addresses handling */
4359 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4360 unsigned char addr_type);
4361 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4362 unsigned char addr_type);
4363 void dev_addr_flush(struct net_device *dev);
4364 int dev_addr_init(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_features_change(struct net_device *dev);
4460 /* Load a device via the kmod */
4461 void dev_load(struct net *net, const char *name);
4462 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4463 struct rtnl_link_stats64 *storage);
4464 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4465 const struct net_device_stats *netdev_stats);
4467 extern int netdev_max_backlog;
4468 extern int netdev_tstamp_prequeue;
4469 extern int weight_p;
4470 extern int dev_weight_rx_bias;
4471 extern int dev_weight_tx_bias;
4472 extern int dev_rx_weight;
4473 extern int dev_tx_weight;
4474 extern int gro_normal_batch;
4476 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4477 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4478 struct list_head **iter);
4479 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4480 struct list_head **iter);
4482 /* iterate through upper list, must be called under RCU read lock */
4483 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4484 for (iter = &(dev)->adj_list.upper, \
4485 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4487 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4489 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4490 int (*fn)(struct net_device *upper_dev,
4494 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4495 struct net_device *upper_dev);
4497 bool netdev_has_any_upper_dev(struct net_device *dev);
4499 void *netdev_lower_get_next_private(struct net_device *dev,
4500 struct list_head **iter);
4501 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4502 struct list_head **iter);
4504 #define netdev_for_each_lower_private(dev, priv, iter) \
4505 for (iter = (dev)->adj_list.lower.next, \
4506 priv = netdev_lower_get_next_private(dev, &(iter)); \
4508 priv = netdev_lower_get_next_private(dev, &(iter)))
4510 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4511 for (iter = &(dev)->adj_list.lower, \
4512 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4514 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4516 void *netdev_lower_get_next(struct net_device *dev,
4517 struct list_head **iter);
4519 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4520 for (iter = (dev)->adj_list.lower.next, \
4521 ldev = netdev_lower_get_next(dev, &(iter)); \
4523 ldev = netdev_lower_get_next(dev, &(iter)))
4525 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4526 struct list_head **iter);
4527 int netdev_walk_all_lower_dev(struct net_device *dev,
4528 int (*fn)(struct net_device *lower_dev,
4531 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4532 int (*fn)(struct net_device *lower_dev,
4536 void *netdev_adjacent_get_private(struct list_head *adj_list);
4537 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4538 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4539 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4540 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4541 struct netlink_ext_ack *extack);
4542 int netdev_master_upper_dev_link(struct net_device *dev,
4543 struct net_device *upper_dev,
4544 void *upper_priv, void *upper_info,
4545 struct netlink_ext_ack *extack);
4546 void netdev_upper_dev_unlink(struct net_device *dev,
4547 struct net_device *upper_dev);
4548 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4549 struct net_device *new_dev,
4550 struct net_device *dev,
4551 struct netlink_ext_ack *extack);
4552 void netdev_adjacent_change_commit(struct net_device *old_dev,
4553 struct net_device *new_dev,
4554 struct net_device *dev);
4555 void netdev_adjacent_change_abort(struct net_device *old_dev,
4556 struct net_device *new_dev,
4557 struct net_device *dev);
4558 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4559 void *netdev_lower_dev_get_private(struct net_device *dev,
4560 struct net_device *lower_dev);
4561 void netdev_lower_state_changed(struct net_device *lower_dev,
4562 void *lower_state_info);
4564 /* RSS keys are 40 or 52 bytes long */
4565 #define NETDEV_RSS_KEY_LEN 52
4566 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4567 void netdev_rss_key_fill(void *buffer, size_t len);
4569 int skb_checksum_help(struct sk_buff *skb);
4570 int skb_crc32c_csum_help(struct sk_buff *skb);
4571 int skb_csum_hwoffload_help(struct sk_buff *skb,
4572 const netdev_features_t features);
4574 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4575 netdev_features_t features, bool tx_path);
4576 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4577 netdev_features_t features);
4579 struct netdev_bonding_info {
4584 struct netdev_notifier_bonding_info {
4585 struct netdev_notifier_info info; /* must be first */
4586 struct netdev_bonding_info bonding_info;
4589 void netdev_bonding_info_change(struct net_device *dev,
4590 struct netdev_bonding_info *bonding_info);
4592 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4593 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4595 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4602 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4604 return __skb_gso_segment(skb, features, true);
4606 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4608 static inline bool can_checksum_protocol(netdev_features_t features,
4611 if (protocol == htons(ETH_P_FCOE))
4612 return !!(features & NETIF_F_FCOE_CRC);
4614 /* Assume this is an IP checksum (not SCTP CRC) */
4616 if (features & NETIF_F_HW_CSUM) {
4617 /* Can checksum everything */
4622 case htons(ETH_P_IP):
4623 return !!(features & NETIF_F_IP_CSUM);
4624 case htons(ETH_P_IPV6):
4625 return !!(features & NETIF_F_IPV6_CSUM);
4632 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4634 static inline void netdev_rx_csum_fault(struct net_device *dev,
4635 struct sk_buff *skb)
4639 /* rx skb timestamps */
4640 void net_enable_timestamp(void);
4641 void net_disable_timestamp(void);
4643 #ifdef CONFIG_PROC_FS
4644 int __init dev_proc_init(void);
4646 #define dev_proc_init() 0
4649 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4650 struct sk_buff *skb, struct net_device *dev,
4653 __this_cpu_write(softnet_data.xmit.more, more);
4654 return ops->ndo_start_xmit(skb, dev);
4657 static inline bool netdev_xmit_more(void)
4659 return __this_cpu_read(softnet_data.xmit.more);
4662 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4663 struct netdev_queue *txq, bool more)
4665 const struct net_device_ops *ops = dev->netdev_ops;
4668 rc = __netdev_start_xmit(ops, skb, dev, more);
4669 if (rc == NETDEV_TX_OK)
4670 txq_trans_update(txq);
4675 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4677 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4680 extern const struct kobj_ns_type_operations net_ns_type_operations;
4682 const char *netdev_drivername(const struct net_device *dev);
4684 void linkwatch_run_queue(void);
4686 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4687 netdev_features_t f2)
4689 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4690 if (f1 & NETIF_F_HW_CSUM)
4691 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4693 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4699 static inline netdev_features_t netdev_get_wanted_features(
4700 struct net_device *dev)
4702 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4704 netdev_features_t netdev_increment_features(netdev_features_t all,
4705 netdev_features_t one, netdev_features_t mask);
4707 /* Allow TSO being used on stacked device :
4708 * Performing the GSO segmentation before last device
4709 * is a performance improvement.
4711 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4712 netdev_features_t mask)
4714 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4717 int __netdev_update_features(struct net_device *dev);
4718 void netdev_update_features(struct net_device *dev);
4719 void netdev_change_features(struct net_device *dev);
4721 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4722 struct net_device *dev);
4724 netdev_features_t passthru_features_check(struct sk_buff *skb,
4725 struct net_device *dev,
4726 netdev_features_t features);
4727 netdev_features_t netif_skb_features(struct sk_buff *skb);
4729 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4731 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4733 /* check flags correspondence */
4734 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4735 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4736 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4737 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4738 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4739 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4740 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4741 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4742 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4743 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4744 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4745 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4746 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4747 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4748 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4749 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4750 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4751 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4752 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4754 return (features & feature) == feature;
4757 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4759 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4760 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4763 static inline bool netif_needs_gso(struct sk_buff *skb,
4764 netdev_features_t features)
4766 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4767 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4768 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4771 static inline void netif_set_gso_max_size(struct net_device *dev,
4774 dev->gso_max_size = size;
4777 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4778 int pulled_hlen, u16 mac_offset,
4781 skb->protocol = protocol;
4782 skb->encapsulation = 1;
4783 skb_push(skb, pulled_hlen);
4784 skb_reset_transport_header(skb);
4785 skb->mac_header = mac_offset;
4786 skb->network_header = skb->mac_header + mac_len;
4787 skb->mac_len = mac_len;
4790 static inline bool netif_is_macsec(const struct net_device *dev)
4792 return dev->priv_flags & IFF_MACSEC;
4795 static inline bool netif_is_macvlan(const struct net_device *dev)
4797 return dev->priv_flags & IFF_MACVLAN;
4800 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4802 return dev->priv_flags & IFF_MACVLAN_PORT;
4805 static inline bool netif_is_bond_master(const struct net_device *dev)
4807 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4810 static inline bool netif_is_bond_slave(const struct net_device *dev)
4812 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4815 static inline bool netif_supports_nofcs(struct net_device *dev)
4817 return dev->priv_flags & IFF_SUPP_NOFCS;
4820 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4822 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4825 static inline bool netif_is_l3_master(const struct net_device *dev)
4827 return dev->priv_flags & IFF_L3MDEV_MASTER;
4830 static inline bool netif_is_l3_slave(const struct net_device *dev)
4832 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4835 static inline bool netif_is_bridge_master(const struct net_device *dev)
4837 return dev->priv_flags & IFF_EBRIDGE;
4840 static inline bool netif_is_bridge_port(const struct net_device *dev)
4842 return dev->priv_flags & IFF_BRIDGE_PORT;
4845 static inline bool netif_is_ovs_master(const struct net_device *dev)
4847 return dev->priv_flags & IFF_OPENVSWITCH;
4850 static inline bool netif_is_ovs_port(const struct net_device *dev)
4852 return dev->priv_flags & IFF_OVS_DATAPATH;
4855 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
4857 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
4860 static inline bool netif_is_team_master(const struct net_device *dev)
4862 return dev->priv_flags & IFF_TEAM;
4865 static inline bool netif_is_team_port(const struct net_device *dev)
4867 return dev->priv_flags & IFF_TEAM_PORT;
4870 static inline bool netif_is_lag_master(const struct net_device *dev)
4872 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4875 static inline bool netif_is_lag_port(const struct net_device *dev)
4877 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4880 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4882 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4885 static inline bool netif_is_failover(const struct net_device *dev)
4887 return dev->priv_flags & IFF_FAILOVER;
4890 static inline bool netif_is_failover_slave(const struct net_device *dev)
4892 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4895 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4896 static inline void netif_keep_dst(struct net_device *dev)
4898 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4901 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4902 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4904 /* TODO: reserve and use an additional IFF bit, if we get more users */
4905 return dev->priv_flags & IFF_MACSEC;
4908 extern struct pernet_operations __net_initdata loopback_net_ops;
4910 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4912 /* netdev_printk helpers, similar to dev_printk */
4914 static inline const char *netdev_name(const struct net_device *dev)
4916 if (!dev->name[0] || strchr(dev->name, '%'))
4917 return "(unnamed net_device)";
4921 static inline bool netdev_unregistering(const struct net_device *dev)
4923 return dev->reg_state == NETREG_UNREGISTERING;
4926 static inline const char *netdev_reg_state(const struct net_device *dev)
4928 switch (dev->reg_state) {
4929 case NETREG_UNINITIALIZED: return " (uninitialized)";
4930 case NETREG_REGISTERED: return "";
4931 case NETREG_UNREGISTERING: return " (unregistering)";
4932 case NETREG_UNREGISTERED: return " (unregistered)";
4933 case NETREG_RELEASED: return " (released)";
4934 case NETREG_DUMMY: return " (dummy)";
4937 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4938 return " (unknown)";
4941 __printf(3, 4) __cold
4942 void netdev_printk(const char *level, const struct net_device *dev,
4943 const char *format, ...);
4944 __printf(2, 3) __cold
4945 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4946 __printf(2, 3) __cold
4947 void netdev_alert(const struct net_device *dev, const char *format, ...);
4948 __printf(2, 3) __cold
4949 void netdev_crit(const struct net_device *dev, const char *format, ...);
4950 __printf(2, 3) __cold
4951 void netdev_err(const struct net_device *dev, const char *format, ...);
4952 __printf(2, 3) __cold
4953 void netdev_warn(const struct net_device *dev, const char *format, ...);
4954 __printf(2, 3) __cold
4955 void netdev_notice(const struct net_device *dev, const char *format, ...);
4956 __printf(2, 3) __cold
4957 void netdev_info(const struct net_device *dev, const char *format, ...);
4959 #define netdev_level_once(level, dev, fmt, ...) \
4961 static bool __print_once __read_mostly; \
4963 if (!__print_once) { \
4964 __print_once = true; \
4965 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4969 #define netdev_emerg_once(dev, fmt, ...) \
4970 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4971 #define netdev_alert_once(dev, fmt, ...) \
4972 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4973 #define netdev_crit_once(dev, fmt, ...) \
4974 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4975 #define netdev_err_once(dev, fmt, ...) \
4976 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4977 #define netdev_warn_once(dev, fmt, ...) \
4978 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4979 #define netdev_notice_once(dev, fmt, ...) \
4980 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4981 #define netdev_info_once(dev, fmt, ...) \
4982 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4984 #define MODULE_ALIAS_NETDEV(device) \
4985 MODULE_ALIAS("netdev-" device)
4987 #if defined(CONFIG_DYNAMIC_DEBUG) || \
4988 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
4989 #define netdev_dbg(__dev, format, args...) \
4991 dynamic_netdev_dbg(__dev, format, ##args); \
4993 #elif defined(DEBUG)
4994 #define netdev_dbg(__dev, format, args...) \
4995 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4997 #define netdev_dbg(__dev, format, args...) \
5000 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5004 #if defined(VERBOSE_DEBUG)
5005 #define netdev_vdbg netdev_dbg
5008 #define netdev_vdbg(dev, format, args...) \
5011 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5017 * netdev_WARN() acts like dev_printk(), but with the key difference
5018 * of using a WARN/WARN_ON to get the message out, including the
5019 * file/line information and a backtrace.
5021 #define netdev_WARN(dev, format, args...) \
5022 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5023 netdev_reg_state(dev), ##args)
5025 #define netdev_WARN_ONCE(dev, format, args...) \
5026 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5027 netdev_reg_state(dev), ##args)
5029 /* netif printk helpers, similar to netdev_printk */
5031 #define netif_printk(priv, type, level, dev, fmt, args...) \
5033 if (netif_msg_##type(priv)) \
5034 netdev_printk(level, (dev), fmt, ##args); \
5037 #define netif_level(level, priv, type, dev, fmt, args...) \
5039 if (netif_msg_##type(priv)) \
5040 netdev_##level(dev, fmt, ##args); \
5043 #define netif_emerg(priv, type, dev, fmt, args...) \
5044 netif_level(emerg, priv, type, dev, fmt, ##args)
5045 #define netif_alert(priv, type, dev, fmt, args...) \
5046 netif_level(alert, priv, type, dev, fmt, ##args)
5047 #define netif_crit(priv, type, dev, fmt, args...) \
5048 netif_level(crit, priv, type, dev, fmt, ##args)
5049 #define netif_err(priv, type, dev, fmt, args...) \
5050 netif_level(err, priv, type, dev, fmt, ##args)
5051 #define netif_warn(priv, type, dev, fmt, args...) \
5052 netif_level(warn, priv, type, dev, fmt, ##args)
5053 #define netif_notice(priv, type, dev, fmt, args...) \
5054 netif_level(notice, priv, type, dev, fmt, ##args)
5055 #define netif_info(priv, type, dev, fmt, args...) \
5056 netif_level(info, priv, type, dev, fmt, ##args)
5058 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5059 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5060 #define netif_dbg(priv, type, netdev, format, args...) \
5062 if (netif_msg_##type(priv)) \
5063 dynamic_netdev_dbg(netdev, format, ##args); \
5065 #elif defined(DEBUG)
5066 #define netif_dbg(priv, type, dev, format, args...) \
5067 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5069 #define netif_dbg(priv, type, dev, format, args...) \
5072 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5077 /* if @cond then downgrade to debug, else print at @level */
5078 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5081 netif_dbg(priv, type, netdev, fmt, ##args); \
5083 netif_ ## level(priv, type, netdev, fmt, ##args); \
5086 #if defined(VERBOSE_DEBUG)
5087 #define netif_vdbg netif_dbg
5089 #define netif_vdbg(priv, type, dev, format, args...) \
5092 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5098 * The list of packet types we will receive (as opposed to discard)
5099 * and the routines to invoke.
5101 * Why 16. Because with 16 the only overlap we get on a hash of the
5102 * low nibble of the protocol value is RARP/SNAP/X.25.
5116 #define PTYPE_HASH_SIZE (16)
5117 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5119 extern struct net_device *blackhole_netdev;
5121 #endif /* _LINUX_NETDEVICE_H */