2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/workqueue.h>
39 #include <linux/dynamic_queue_limits.h>
41 #include <linux/ethtool.h>
42 #include <net/net_namespace.h>
44 #include <net/dcbnl.h>
46 #include <net/netprio_cgroup.h>
49 #include <linux/netdev_features.h>
50 #include <linux/neighbour.h>
51 #include <uapi/linux/netdevice.h>
52 #include <uapi/linux/if_bonding.h>
53 #include <uapi/linux/pkt_cls.h>
54 #include <linux/hashtable.h>
63 /* 802.15.4 specific */
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info;
71 void netdev_set_default_ethtool_ops(struct net_device *dev,
72 const struct ethtool_ops *ops);
74 /* Backlog congestion levels */
75 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
76 #define NET_RX_DROP 1 /* packet dropped */
79 * Transmit return codes: transmit return codes originate from three different
82 * - qdisc return codes
83 * - driver transmit return codes
86 * Drivers are allowed to return any one of those in their hard_start_xmit()
87 * function. Real network devices commonly used with qdiscs should only return
88 * the driver transmit return codes though - when qdiscs are used, the actual
89 * transmission happens asynchronously, so the value is not propagated to
90 * higher layers. Virtual network devices transmit synchronously; in this case
91 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
92 * others are propagated to higher layers.
95 /* qdisc ->enqueue() return codes. */
96 #define NET_XMIT_SUCCESS 0x00
97 #define NET_XMIT_DROP 0x01 /* skb dropped */
98 #define NET_XMIT_CN 0x02 /* congestion notification */
99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK 0xf0
111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
115 typedef enum netdev_tx netdev_tx_t;
118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121 static inline bool dev_xmit_complete(int rc)
124 * Positive cases with an skb consumed by a driver:
125 * - successful transmission (rc == NETDEV_TX_OK)
126 * - error while transmitting (rc < 0)
127 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 if (likely(rc < NET_XMIT_MASK))
136 * Compute the worst-case header length according to the protocols
140 #if defined(CONFIG_HYPERV_NET)
141 # define LL_MAX_HEADER 128
142 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
143 # if defined(CONFIG_MAC80211_MESH)
144 # define LL_MAX_HEADER 128
146 # define LL_MAX_HEADER 96
149 # define LL_MAX_HEADER 32
152 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
153 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
154 #define MAX_HEADER LL_MAX_HEADER
156 #define MAX_HEADER (LL_MAX_HEADER + 48)
160 * Old network device statistics. Fields are native words
161 * (unsigned long) so they can be read and written atomically.
164 struct net_device_stats {
165 unsigned long rx_packets;
166 unsigned long tx_packets;
167 unsigned long rx_bytes;
168 unsigned long tx_bytes;
169 unsigned long rx_errors;
170 unsigned long tx_errors;
171 unsigned long rx_dropped;
172 unsigned long tx_dropped;
173 unsigned long multicast;
174 unsigned long collisions;
175 unsigned long rx_length_errors;
176 unsigned long rx_over_errors;
177 unsigned long rx_crc_errors;
178 unsigned long rx_frame_errors;
179 unsigned long rx_fifo_errors;
180 unsigned long rx_missed_errors;
181 unsigned long tx_aborted_errors;
182 unsigned long tx_carrier_errors;
183 unsigned long tx_fifo_errors;
184 unsigned long tx_heartbeat_errors;
185 unsigned long tx_window_errors;
186 unsigned long rx_compressed;
187 unsigned long tx_compressed;
191 #include <linux/cache.h>
192 #include <linux/skbuff.h>
195 #include <linux/static_key.h>
196 extern struct static_key rps_needed;
197 extern struct static_key rfs_needed;
204 struct netdev_hw_addr {
205 struct list_head list;
206 unsigned char addr[MAX_ADDR_LEN];
208 #define NETDEV_HW_ADDR_T_LAN 1
209 #define NETDEV_HW_ADDR_T_SAN 2
210 #define NETDEV_HW_ADDR_T_SLAVE 3
211 #define NETDEV_HW_ADDR_T_UNICAST 4
212 #define NETDEV_HW_ADDR_T_MULTICAST 5
217 struct rcu_head rcu_head;
220 struct netdev_hw_addr_list {
221 struct list_head list;
225 #define netdev_hw_addr_list_count(l) ((l)->count)
226 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
227 #define netdev_hw_addr_list_for_each(ha, l) \
228 list_for_each_entry(ha, &(l)->list, list)
230 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
231 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
232 #define netdev_for_each_uc_addr(ha, dev) \
233 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
235 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
236 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
237 #define netdev_for_each_mc_addr(ha, dev) \
238 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
244 /* cached hardware header; allow for machine alignment needs. */
245 #define HH_DATA_MOD 16
246 #define HH_DATA_OFF(__len) \
247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
248 #define HH_DATA_ALIGN(__len) \
249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
253 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
255 * dev->hard_header_len ? (dev->hard_header_len +
256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
258 * We could use other alignment values, but we must maintain the
259 * relationship HH alignment <= LL alignment.
261 #define LL_RESERVED_SPACE(dev) \
262 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
263 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
264 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
267 int (*create) (struct sk_buff *skb, struct net_device *dev,
268 unsigned short type, const void *daddr,
269 const void *saddr, unsigned int len);
270 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
275 bool (*validate)(const char *ll_header, unsigned int len);
278 /* These flag bits are private to the generic network queueing
279 * layer; they may not be explicitly referenced by any other
283 enum netdev_state_t {
285 __LINK_STATE_PRESENT,
286 __LINK_STATE_NOCARRIER,
287 __LINK_STATE_LINKWATCH_PENDING,
288 __LINK_STATE_DORMANT,
293 * This structure holds boot-time configured netdevice settings. They
294 * are then used in the device probing.
296 struct netdev_boot_setup {
300 #define NETDEV_BOOT_SETUP_MAX 8
302 int __init netdev_boot_setup(char *str);
305 * Structure for NAPI scheduling similar to tasklet but with weighting
308 /* The poll_list must only be managed by the entity which
309 * changes the state of the NAPI_STATE_SCHED bit. This means
310 * whoever atomically sets that bit can add this napi_struct
311 * to the per-CPU poll_list, and whoever clears that bit
312 * can remove from the list right before clearing the bit.
314 struct list_head poll_list;
318 unsigned int gro_count;
319 int (*poll)(struct napi_struct *, int);
320 #ifdef CONFIG_NETPOLL
323 struct net_device *dev;
324 struct sk_buff *gro_list;
326 struct hrtimer timer;
327 struct list_head dev_list;
328 struct hlist_node napi_hash_node;
329 unsigned int napi_id;
333 NAPI_STATE_SCHED, /* Poll is scheduled */
334 NAPI_STATE_MISSED, /* reschedule a napi */
335 NAPI_STATE_DISABLE, /* Disable pending */
336 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
337 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
338 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
339 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
343 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
344 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
345 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
346 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
347 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
348 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
349 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
360 typedef enum gro_result gro_result_t;
363 * enum rx_handler_result - Possible return values for rx_handlers.
364 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
366 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
367 * case skb->dev was changed by rx_handler.
368 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
369 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
371 * rx_handlers are functions called from inside __netif_receive_skb(), to do
372 * special processing of the skb, prior to delivery to protocol handlers.
374 * Currently, a net_device can only have a single rx_handler registered. Trying
375 * to register a second rx_handler will return -EBUSY.
377 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
378 * To unregister a rx_handler on a net_device, use
379 * netdev_rx_handler_unregister().
381 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
384 * If the rx_handler consumed the skb in some way, it should return
385 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
386 * the skb to be delivered in some other way.
388 * If the rx_handler changed skb->dev, to divert the skb to another
389 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
390 * new device will be called if it exists.
392 * If the rx_handler decides the skb should be ignored, it should return
393 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
394 * are registered on exact device (ptype->dev == skb->dev).
396 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
397 * delivered, it should return RX_HANDLER_PASS.
399 * A device without a registered rx_handler will behave as if rx_handler
400 * returned RX_HANDLER_PASS.
403 enum rx_handler_result {
409 typedef enum rx_handler_result rx_handler_result_t;
410 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
412 void __napi_schedule(struct napi_struct *n);
413 void __napi_schedule_irqoff(struct napi_struct *n);
415 static inline bool napi_disable_pending(struct napi_struct *n)
417 return test_bit(NAPI_STATE_DISABLE, &n->state);
420 bool napi_schedule_prep(struct napi_struct *n);
423 * napi_schedule - schedule NAPI poll
426 * Schedule NAPI poll routine to be called if it is not already
429 static inline void napi_schedule(struct napi_struct *n)
431 if (napi_schedule_prep(n))
436 * napi_schedule_irqoff - schedule NAPI poll
439 * Variant of napi_schedule(), assuming hard irqs are masked.
441 static inline void napi_schedule_irqoff(struct napi_struct *n)
443 if (napi_schedule_prep(n))
444 __napi_schedule_irqoff(n);
447 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
448 static inline bool napi_reschedule(struct napi_struct *napi)
450 if (napi_schedule_prep(napi)) {
451 __napi_schedule(napi);
457 bool napi_complete_done(struct napi_struct *n, int work_done);
459 * napi_complete - NAPI processing complete
462 * Mark NAPI processing as complete.
463 * Consider using napi_complete_done() instead.
464 * Return false if device should avoid rearming interrupts.
466 static inline bool napi_complete(struct napi_struct *n)
468 return napi_complete_done(n, 0);
472 * napi_hash_del - remove a NAPI from global table
473 * @napi: NAPI context
475 * Warning: caller must observe RCU grace period
476 * before freeing memory containing @napi, if
477 * this function returns true.
478 * Note: core networking stack automatically calls it
479 * from netif_napi_del().
480 * Drivers might want to call this helper to combine all
481 * the needed RCU grace periods into a single one.
483 bool napi_hash_del(struct napi_struct *napi);
486 * napi_disable - prevent NAPI from scheduling
489 * Stop NAPI from being scheduled on this context.
490 * Waits till any outstanding processing completes.
492 void napi_disable(struct napi_struct *n);
495 * napi_enable - enable NAPI scheduling
498 * Resume NAPI from being scheduled on this context.
499 * Must be paired with napi_disable.
501 static inline void napi_enable(struct napi_struct *n)
503 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
504 smp_mb__before_atomic();
505 clear_bit(NAPI_STATE_SCHED, &n->state);
506 clear_bit(NAPI_STATE_NPSVC, &n->state);
510 * napi_synchronize - wait until NAPI is not running
513 * Wait until NAPI is done being scheduled on this context.
514 * Waits till any outstanding processing completes but
515 * does not disable future activations.
517 static inline void napi_synchronize(const struct napi_struct *n)
519 if (IS_ENABLED(CONFIG_SMP))
520 while (test_bit(NAPI_STATE_SCHED, &n->state))
526 enum netdev_queue_state_t {
527 __QUEUE_STATE_DRV_XOFF,
528 __QUEUE_STATE_STACK_XOFF,
529 __QUEUE_STATE_FROZEN,
532 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
533 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
534 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
536 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
537 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
539 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
543 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
544 * netif_tx_* functions below are used to manipulate this flag. The
545 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
546 * queue independently. The netif_xmit_*stopped functions below are called
547 * to check if the queue has been stopped by the driver or stack (either
548 * of the XOFF bits are set in the state). Drivers should not need to call
549 * netif_xmit*stopped functions, they should only be using netif_tx_*.
552 struct netdev_queue {
556 struct net_device *dev;
557 struct Qdisc __rcu *qdisc;
558 struct Qdisc *qdisc_sleeping;
562 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
565 unsigned long tx_maxrate;
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
570 unsigned long trans_timeout;
574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
577 * Time (in jiffies) of last Tx
579 unsigned long trans_start;
586 } ____cacheline_aligned_in_smp;
588 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
590 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
597 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 * This structure holds an RPS map which can be of variable length. The
607 * map is an array of CPUs.
614 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
617 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
618 * tail pointer for that CPU's input queue at the time of last enqueue, and
619 * a hardware filter index.
621 struct rps_dev_flow {
624 unsigned int last_qtail;
626 #define RPS_NO_FILTER 0xffff
629 * The rps_dev_flow_table structure contains a table of flow mappings.
631 struct rps_dev_flow_table {
634 struct rps_dev_flow flows[0];
636 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
637 ((_num) * sizeof(struct rps_dev_flow)))
640 * The rps_sock_flow_table contains mappings of flows to the last CPU
641 * on which they were processed by the application (set in recvmsg).
642 * Each entry is a 32bit value. Upper part is the high-order bits
643 * of flow hash, lower part is CPU number.
644 * rps_cpu_mask is used to partition the space, depending on number of
645 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
646 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
647 * meaning we use 32-6=26 bits for the hash.
649 struct rps_sock_flow_table {
652 u32 ents[0] ____cacheline_aligned_in_smp;
654 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
656 #define RPS_NO_CPU 0xffff
658 extern u32 rps_cpu_mask;
659 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
661 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
665 unsigned int index = hash & table->mask;
666 u32 val = hash & ~rps_cpu_mask;
668 /* We only give a hint, preemption can change CPU under us */
669 val |= raw_smp_processor_id();
671 if (table->ents[index] != val)
672 table->ents[index] = val;
676 #ifdef CONFIG_RFS_ACCEL
677 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
680 #endif /* CONFIG_RPS */
682 /* This structure contains an instance of an RX queue. */
683 struct netdev_rx_queue {
685 struct rps_map __rcu *rps_map;
686 struct rps_dev_flow_table __rcu *rps_flow_table;
689 struct net_device *dev;
690 struct xdp_rxq_info xdp_rxq;
691 } ____cacheline_aligned_in_smp;
694 * RX queue sysfs structures and functions.
696 struct rx_queue_attribute {
697 struct attribute attr;
698 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
699 ssize_t (*store)(struct netdev_rx_queue *queue,
700 const char *buf, size_t len);
705 * This structure holds an XPS map which can be of variable length. The
706 * map is an array of queues.
710 unsigned int alloc_len;
714 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
715 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
716 - sizeof(struct xps_map)) / sizeof(u16))
719 * This structure holds all XPS maps for device. Maps are indexed by CPU.
721 struct xps_dev_maps {
723 struct xps_map __rcu *cpu_map[0];
725 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
726 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
727 #endif /* CONFIG_XPS */
729 #define TC_MAX_QUEUE 16
730 #define TC_BITMASK 15
731 /* HW offloaded queuing disciplines txq count and offset maps */
732 struct netdev_tc_txq {
737 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
739 * This structure is to hold information about the device
740 * configured to run FCoE protocol stack.
742 struct netdev_fcoe_hbainfo {
743 char manufacturer[64];
744 char serial_number[64];
745 char hardware_version[64];
746 char driver_version[64];
747 char optionrom_version[64];
748 char firmware_version[64];
750 char model_description[256];
754 #define MAX_PHYS_ITEM_ID_LEN 32
756 /* This structure holds a unique identifier to identify some
757 * physical item (port for example) used by a netdevice.
759 struct netdev_phys_item_id {
760 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
761 unsigned char id_len;
764 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
765 struct netdev_phys_item_id *b)
767 return a->id_len == b->id_len &&
768 memcmp(a->id, b->id, a->id_len) == 0;
771 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
772 struct sk_buff *skb);
775 TC_SETUP_QDISC_MQPRIO,
778 TC_SETUP_CLSMATCHALL,
786 /* These structures hold the attributes of bpf state that are being passed
787 * to the netdevice through the bpf op.
789 enum bpf_netdev_command {
790 /* Set or clear a bpf program used in the earliest stages of packet
791 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
792 * is responsible for calling bpf_prog_put on any old progs that are
793 * stored. In case of error, the callee need not release the new prog
794 * reference, but on success it takes ownership and must bpf_prog_put
795 * when it is no longer used.
799 /* Check if a bpf program is set on the device. The callee should
800 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
801 * is equivalent to XDP_ATTACHED_DRV.
804 /* BPF program for offload callbacks, invoked at program load time. */
805 BPF_OFFLOAD_VERIFIER_PREP,
806 BPF_OFFLOAD_TRANSLATE,
808 BPF_OFFLOAD_MAP_ALLOC,
809 BPF_OFFLOAD_MAP_FREE,
812 struct bpf_prog_offload_ops;
813 struct netlink_ext_ack;
816 enum bpf_netdev_command command;
821 struct bpf_prog *prog;
822 struct netlink_ext_ack *extack;
828 /* flags with which program was installed */
831 /* BPF_OFFLOAD_VERIFIER_PREP */
833 struct bpf_prog *prog;
834 const struct bpf_prog_offload_ops *ops; /* callee set */
836 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
838 struct bpf_prog *prog;
840 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
842 struct bpf_offloaded_map *offmap;
847 #ifdef CONFIG_XFRM_OFFLOAD
849 int (*xdo_dev_state_add) (struct xfrm_state *x);
850 void (*xdo_dev_state_delete) (struct xfrm_state *x);
851 void (*xdo_dev_state_free) (struct xfrm_state *x);
852 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
853 struct xfrm_state *x);
858 struct rcu_head rcuhead;
863 * This structure defines the management hooks for network devices.
864 * The following hooks can be defined; unless noted otherwise, they are
865 * optional and can be filled with a null pointer.
867 * int (*ndo_init)(struct net_device *dev);
868 * This function is called once when a network device is registered.
869 * The network device can use this for any late stage initialization
870 * or semantic validation. It can fail with an error code which will
871 * be propagated back to register_netdev.
873 * void (*ndo_uninit)(struct net_device *dev);
874 * This function is called when device is unregistered or when registration
875 * fails. It is not called if init fails.
877 * int (*ndo_open)(struct net_device *dev);
878 * This function is called when a network device transitions to the up
881 * int (*ndo_stop)(struct net_device *dev);
882 * This function is called when a network device transitions to the down
885 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
886 * struct net_device *dev);
887 * Called when a packet needs to be transmitted.
888 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
889 * the queue before that can happen; it's for obsolete devices and weird
890 * corner cases, but the stack really does a non-trivial amount
891 * of useless work if you return NETDEV_TX_BUSY.
892 * Required; cannot be NULL.
894 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
895 * struct net_device *dev
896 * netdev_features_t features);
897 * Called by core transmit path to determine if device is capable of
898 * performing offload operations on a given packet. This is to give
899 * the device an opportunity to implement any restrictions that cannot
900 * be otherwise expressed by feature flags. The check is called with
901 * the set of features that the stack has calculated and it returns
902 * those the driver believes to be appropriate.
904 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
905 * void *accel_priv, select_queue_fallback_t fallback);
906 * Called to decide which queue to use when device supports multiple
909 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
910 * This function is called to allow device receiver to make
911 * changes to configuration when multicast or promiscuous is enabled.
913 * void (*ndo_set_rx_mode)(struct net_device *dev);
914 * This function is called device changes address list filtering.
915 * If driver handles unicast address filtering, it should set
916 * IFF_UNICAST_FLT in its priv_flags.
918 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
919 * This function is called when the Media Access Control address
920 * needs to be changed. If this interface is not defined, the
921 * MAC address can not be changed.
923 * int (*ndo_validate_addr)(struct net_device *dev);
924 * Test if Media Access Control address is valid for the device.
926 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
927 * Called when a user requests an ioctl which can't be handled by
928 * the generic interface code. If not defined ioctls return
929 * not supported error code.
931 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
932 * Used to set network devices bus interface parameters. This interface
933 * is retained for legacy reasons; new devices should use the bus
934 * interface (PCI) for low level management.
936 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
937 * Called when a user wants to change the Maximum Transfer Unit
940 * void (*ndo_tx_timeout)(struct net_device *dev);
941 * Callback used when the transmitter has not made any progress
942 * for dev->watchdog ticks.
944 * void (*ndo_get_stats64)(struct net_device *dev,
945 * struct rtnl_link_stats64 *storage);
946 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
947 * Called when a user wants to get the network device usage
948 * statistics. Drivers must do one of the following:
949 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
950 * rtnl_link_stats64 structure passed by the caller.
951 * 2. Define @ndo_get_stats to update a net_device_stats structure
952 * (which should normally be dev->stats) and return a pointer to
953 * it. The structure may be changed asynchronously only if each
954 * field is written atomically.
955 * 3. Update dev->stats asynchronously and atomically, and define
958 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
959 * Return true if this device supports offload stats of this attr_id.
961 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
963 * Get statistics for offload operations by attr_id. Write it into the
966 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
967 * If device supports VLAN filtering this function is called when a
968 * VLAN id is registered.
970 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
971 * If device supports VLAN filtering this function is called when a
972 * VLAN id is unregistered.
974 * void (*ndo_poll_controller)(struct net_device *dev);
976 * SR-IOV management functions.
977 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
978 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
979 * u8 qos, __be16 proto);
980 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
982 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
983 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
984 * int (*ndo_get_vf_config)(struct net_device *dev,
985 * int vf, struct ifla_vf_info *ivf);
986 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
987 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
988 * struct nlattr *port[]);
990 * Enable or disable the VF ability to query its RSS Redirection Table and
991 * Hash Key. This is needed since on some devices VF share this information
992 * with PF and querying it may introduce a theoretical security risk.
993 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
994 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
995 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
997 * Called to setup any 'tc' scheduler, classifier or action on @dev.
998 * This is always called from the stack with the rtnl lock held and netif
999 * tx queues stopped. This allows the netdevice to perform queue
1000 * management safely.
1002 * Fiber Channel over Ethernet (FCoE) offload functions.
1003 * int (*ndo_fcoe_enable)(struct net_device *dev);
1004 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1005 * so the underlying device can perform whatever needed configuration or
1006 * initialization to support acceleration of FCoE traffic.
1008 * int (*ndo_fcoe_disable)(struct net_device *dev);
1009 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1010 * so the underlying device can perform whatever needed clean-ups to
1011 * stop supporting acceleration of FCoE traffic.
1013 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1014 * struct scatterlist *sgl, unsigned int sgc);
1015 * Called when the FCoE Initiator wants to initialize an I/O that
1016 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1017 * perform necessary setup and returns 1 to indicate the device is set up
1018 * successfully to perform DDP on this I/O, otherwise this returns 0.
1020 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1021 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1022 * indicated by the FC exchange id 'xid', so the underlying device can
1023 * clean up and reuse resources for later DDP requests.
1025 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1026 * struct scatterlist *sgl, unsigned int sgc);
1027 * Called when the FCoE Target wants to initialize an I/O that
1028 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1029 * perform necessary setup and returns 1 to indicate the device is set up
1030 * successfully to perform DDP on this I/O, otherwise this returns 0.
1032 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1033 * struct netdev_fcoe_hbainfo *hbainfo);
1034 * Called when the FCoE Protocol stack wants information on the underlying
1035 * device. This information is utilized by the FCoE protocol stack to
1036 * register attributes with Fiber Channel management service as per the
1037 * FC-GS Fabric Device Management Information(FDMI) specification.
1039 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1040 * Called when the underlying device wants to override default World Wide
1041 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1042 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1043 * protocol stack to use.
1046 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1047 * u16 rxq_index, u32 flow_id);
1048 * Set hardware filter for RFS. rxq_index is the target queue index;
1049 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1050 * Return the filter ID on success, or a negative error code.
1052 * Slave management functions (for bridge, bonding, etc).
1053 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1054 * Called to make another netdev an underling.
1056 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1057 * Called to release previously enslaved netdev.
1059 * Feature/offload setting functions.
1060 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1061 * netdev_features_t features);
1062 * Adjusts the requested feature flags according to device-specific
1063 * constraints, and returns the resulting flags. Must not modify
1066 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1067 * Called to update device configuration to new features. Passed
1068 * feature set might be less than what was returned by ndo_fix_features()).
1069 * Must return >0 or -errno if it changed dev->features itself.
1071 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1072 * struct net_device *dev,
1073 * const unsigned char *addr, u16 vid, u16 flags)
1074 * Adds an FDB entry to dev for addr.
1075 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1076 * struct net_device *dev,
1077 * const unsigned char *addr, u16 vid)
1078 * Deletes the FDB entry from dev coresponding to addr.
1079 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1080 * struct net_device *dev, struct net_device *filter_dev,
1082 * Used to add FDB entries to dump requests. Implementers should add
1083 * entries to skb and update idx with the number of entries.
1085 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1087 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1088 * struct net_device *dev, u32 filter_mask,
1090 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1093 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1094 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1095 * which do not represent real hardware may define this to allow their
1096 * userspace components to manage their virtual carrier state. Devices
1097 * that determine carrier state from physical hardware properties (eg
1098 * network cables) or protocol-dependent mechanisms (eg
1099 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1101 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1102 * struct netdev_phys_item_id *ppid);
1103 * Called to get ID of physical port of this device. If driver does
1104 * not implement this, it is assumed that the hw is not able to have
1105 * multiple net devices on single physical port.
1107 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1108 * struct udp_tunnel_info *ti);
1109 * Called by UDP tunnel to notify a driver about the UDP port and socket
1110 * address family that a UDP tunnel is listnening to. It is called only
1111 * when a new port starts listening. The operation is protected by the
1114 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1115 * struct udp_tunnel_info *ti);
1116 * Called by UDP tunnel to notify the driver about a UDP port and socket
1117 * address family that the UDP tunnel is not listening to anymore. The
1118 * operation is protected by the RTNL.
1120 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1121 * struct net_device *dev)
1122 * Called by upper layer devices to accelerate switching or other
1123 * station functionality into hardware. 'pdev is the lowerdev
1124 * to use for the offload and 'dev' is the net device that will
1125 * back the offload. Returns a pointer to the private structure
1126 * the upper layer will maintain.
1127 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1128 * Called by upper layer device to delete the station created
1129 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1130 * the station and priv is the structure returned by the add
1132 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1133 * int queue_index, u32 maxrate);
1134 * Called when a user wants to set a max-rate limitation of specific
1136 * int (*ndo_get_iflink)(const struct net_device *dev);
1137 * Called to get the iflink value of this device.
1138 * void (*ndo_change_proto_down)(struct net_device *dev,
1140 * This function is used to pass protocol port error state information
1141 * to the switch driver. The switch driver can react to the proto_down
1142 * by doing a phys down on the associated switch port.
1143 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1144 * This function is used to get egress tunnel information for given skb.
1145 * This is useful for retrieving outer tunnel header parameters while
1147 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1148 * This function is used to specify the headroom that the skb must
1149 * consider when allocation skb during packet reception. Setting
1150 * appropriate rx headroom value allows avoiding skb head copy on
1151 * forward. Setting a negative value resets the rx headroom to the
1153 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1154 * This function is used to set or query state related to XDP on the
1155 * netdevice and manage BPF offload. See definition of
1156 * enum bpf_netdev_command for details.
1157 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1158 * This function is used to submit a XDP packet for transmit on a
1160 * void (*ndo_xdp_flush)(struct net_device *dev);
1161 * This function is used to inform the driver to flush a particular
1162 * xdp tx queue. Must be called on same CPU as xdp_xmit.
1164 struct net_device_ops {
1165 int (*ndo_init)(struct net_device *dev);
1166 void (*ndo_uninit)(struct net_device *dev);
1167 int (*ndo_open)(struct net_device *dev);
1168 int (*ndo_stop)(struct net_device *dev);
1169 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1170 struct net_device *dev);
1171 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1172 struct net_device *dev,
1173 netdev_features_t features);
1174 u16 (*ndo_select_queue)(struct net_device *dev,
1175 struct sk_buff *skb,
1177 select_queue_fallback_t fallback);
1178 void (*ndo_change_rx_flags)(struct net_device *dev,
1180 void (*ndo_set_rx_mode)(struct net_device *dev);
1181 int (*ndo_set_mac_address)(struct net_device *dev,
1183 int (*ndo_validate_addr)(struct net_device *dev);
1184 int (*ndo_do_ioctl)(struct net_device *dev,
1185 struct ifreq *ifr, int cmd);
1186 int (*ndo_set_config)(struct net_device *dev,
1188 int (*ndo_change_mtu)(struct net_device *dev,
1190 int (*ndo_neigh_setup)(struct net_device *dev,
1191 struct neigh_parms *);
1192 void (*ndo_tx_timeout) (struct net_device *dev);
1194 void (*ndo_get_stats64)(struct net_device *dev,
1195 struct rtnl_link_stats64 *storage);
1196 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1197 int (*ndo_get_offload_stats)(int attr_id,
1198 const struct net_device *dev,
1200 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1202 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1203 __be16 proto, u16 vid);
1204 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1205 __be16 proto, u16 vid);
1206 #ifdef CONFIG_NET_POLL_CONTROLLER
1207 void (*ndo_poll_controller)(struct net_device *dev);
1208 int (*ndo_netpoll_setup)(struct net_device *dev,
1209 struct netpoll_info *info);
1210 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1212 int (*ndo_set_vf_mac)(struct net_device *dev,
1213 int queue, u8 *mac);
1214 int (*ndo_set_vf_vlan)(struct net_device *dev,
1215 int queue, u16 vlan,
1216 u8 qos, __be16 proto);
1217 int (*ndo_set_vf_rate)(struct net_device *dev,
1218 int vf, int min_tx_rate,
1220 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1221 int vf, bool setting);
1222 int (*ndo_set_vf_trust)(struct net_device *dev,
1223 int vf, bool setting);
1224 int (*ndo_get_vf_config)(struct net_device *dev,
1226 struct ifla_vf_info *ivf);
1227 int (*ndo_set_vf_link_state)(struct net_device *dev,
1228 int vf, int link_state);
1229 int (*ndo_get_vf_stats)(struct net_device *dev,
1231 struct ifla_vf_stats
1233 int (*ndo_set_vf_port)(struct net_device *dev,
1235 struct nlattr *port[]);
1236 int (*ndo_get_vf_port)(struct net_device *dev,
1237 int vf, struct sk_buff *skb);
1238 int (*ndo_set_vf_guid)(struct net_device *dev,
1241 int (*ndo_set_vf_rss_query_en)(
1242 struct net_device *dev,
1243 int vf, bool setting);
1244 int (*ndo_setup_tc)(struct net_device *dev,
1245 enum tc_setup_type type,
1247 #if IS_ENABLED(CONFIG_FCOE)
1248 int (*ndo_fcoe_enable)(struct net_device *dev);
1249 int (*ndo_fcoe_disable)(struct net_device *dev);
1250 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1252 struct scatterlist *sgl,
1254 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1256 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1258 struct scatterlist *sgl,
1260 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1261 struct netdev_fcoe_hbainfo *hbainfo);
1264 #if IS_ENABLED(CONFIG_LIBFCOE)
1265 #define NETDEV_FCOE_WWNN 0
1266 #define NETDEV_FCOE_WWPN 1
1267 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1268 u64 *wwn, int type);
1271 #ifdef CONFIG_RFS_ACCEL
1272 int (*ndo_rx_flow_steer)(struct net_device *dev,
1273 const struct sk_buff *skb,
1277 int (*ndo_add_slave)(struct net_device *dev,
1278 struct net_device *slave_dev,
1279 struct netlink_ext_ack *extack);
1280 int (*ndo_del_slave)(struct net_device *dev,
1281 struct net_device *slave_dev);
1282 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1283 netdev_features_t features);
1284 int (*ndo_set_features)(struct net_device *dev,
1285 netdev_features_t features);
1286 int (*ndo_neigh_construct)(struct net_device *dev,
1287 struct neighbour *n);
1288 void (*ndo_neigh_destroy)(struct net_device *dev,
1289 struct neighbour *n);
1291 int (*ndo_fdb_add)(struct ndmsg *ndm,
1292 struct nlattr *tb[],
1293 struct net_device *dev,
1294 const unsigned char *addr,
1297 int (*ndo_fdb_del)(struct ndmsg *ndm,
1298 struct nlattr *tb[],
1299 struct net_device *dev,
1300 const unsigned char *addr,
1302 int (*ndo_fdb_dump)(struct sk_buff *skb,
1303 struct netlink_callback *cb,
1304 struct net_device *dev,
1305 struct net_device *filter_dev,
1308 int (*ndo_bridge_setlink)(struct net_device *dev,
1309 struct nlmsghdr *nlh,
1311 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1313 struct net_device *dev,
1316 int (*ndo_bridge_dellink)(struct net_device *dev,
1317 struct nlmsghdr *nlh,
1319 int (*ndo_change_carrier)(struct net_device *dev,
1321 int (*ndo_get_phys_port_id)(struct net_device *dev,
1322 struct netdev_phys_item_id *ppid);
1323 int (*ndo_get_phys_port_name)(struct net_device *dev,
1324 char *name, size_t len);
1325 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1326 struct udp_tunnel_info *ti);
1327 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1328 struct udp_tunnel_info *ti);
1329 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1330 struct net_device *dev);
1331 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1334 int (*ndo_get_lock_subclass)(struct net_device *dev);
1335 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1338 int (*ndo_get_iflink)(const struct net_device *dev);
1339 int (*ndo_change_proto_down)(struct net_device *dev,
1341 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1342 struct sk_buff *skb);
1343 void (*ndo_set_rx_headroom)(struct net_device *dev,
1344 int needed_headroom);
1345 int (*ndo_bpf)(struct net_device *dev,
1346 struct netdev_bpf *bpf);
1347 int (*ndo_xdp_xmit)(struct net_device *dev,
1348 struct xdp_buff *xdp);
1349 void (*ndo_xdp_flush)(struct net_device *dev);
1353 * enum net_device_priv_flags - &struct net_device priv_flags
1355 * These are the &struct net_device, they are only set internally
1356 * by drivers and used in the kernel. These flags are invisible to
1357 * userspace; this means that the order of these flags can change
1358 * during any kernel release.
1360 * You should have a pretty good reason to be extending these flags.
1362 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1363 * @IFF_EBRIDGE: Ethernet bridging device
1364 * @IFF_BONDING: bonding master or slave
1365 * @IFF_ISATAP: ISATAP interface (RFC4214)
1366 * @IFF_WAN_HDLC: WAN HDLC device
1367 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1369 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1370 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1371 * @IFF_MACVLAN_PORT: device used as macvlan port
1372 * @IFF_BRIDGE_PORT: device used as bridge port
1373 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1374 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1375 * @IFF_UNICAST_FLT: Supports unicast filtering
1376 * @IFF_TEAM_PORT: device used as team port
1377 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1378 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1379 * change when it's running
1380 * @IFF_MACVLAN: Macvlan device
1381 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1382 * underlying stacked devices
1383 * @IFF_IPVLAN_MASTER: IPvlan master device
1384 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1385 * @IFF_L3MDEV_MASTER: device is an L3 master device
1386 * @IFF_NO_QUEUE: device can run without qdisc attached
1387 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1388 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1389 * @IFF_TEAM: device is a team device
1390 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1391 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1392 * entity (i.e. the master device for bridged veth)
1393 * @IFF_MACSEC: device is a MACsec device
1395 enum netdev_priv_flags {
1396 IFF_802_1Q_VLAN = 1<<0,
1400 IFF_WAN_HDLC = 1<<4,
1401 IFF_XMIT_DST_RELEASE = 1<<5,
1402 IFF_DONT_BRIDGE = 1<<6,
1403 IFF_DISABLE_NETPOLL = 1<<7,
1404 IFF_MACVLAN_PORT = 1<<8,
1405 IFF_BRIDGE_PORT = 1<<9,
1406 IFF_OVS_DATAPATH = 1<<10,
1407 IFF_TX_SKB_SHARING = 1<<11,
1408 IFF_UNICAST_FLT = 1<<12,
1409 IFF_TEAM_PORT = 1<<13,
1410 IFF_SUPP_NOFCS = 1<<14,
1411 IFF_LIVE_ADDR_CHANGE = 1<<15,
1412 IFF_MACVLAN = 1<<16,
1413 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1414 IFF_IPVLAN_MASTER = 1<<18,
1415 IFF_IPVLAN_SLAVE = 1<<19,
1416 IFF_L3MDEV_MASTER = 1<<20,
1417 IFF_NO_QUEUE = 1<<21,
1418 IFF_OPENVSWITCH = 1<<22,
1419 IFF_L3MDEV_SLAVE = 1<<23,
1421 IFF_RXFH_CONFIGURED = 1<<25,
1422 IFF_PHONY_HEADROOM = 1<<26,
1426 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1427 #define IFF_EBRIDGE IFF_EBRIDGE
1428 #define IFF_BONDING IFF_BONDING
1429 #define IFF_ISATAP IFF_ISATAP
1430 #define IFF_WAN_HDLC IFF_WAN_HDLC
1431 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1432 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1433 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1434 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1435 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1436 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1437 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1438 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1439 #define IFF_TEAM_PORT IFF_TEAM_PORT
1440 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1441 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1442 #define IFF_MACVLAN IFF_MACVLAN
1443 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1444 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1445 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1446 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1447 #define IFF_NO_QUEUE IFF_NO_QUEUE
1448 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1449 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1450 #define IFF_TEAM IFF_TEAM
1451 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1452 #define IFF_MACSEC IFF_MACSEC
1455 * struct net_device - The DEVICE structure.
1457 * Actually, this whole structure is a big mistake. It mixes I/O
1458 * data with strictly "high-level" data, and it has to know about
1459 * almost every data structure used in the INET module.
1461 * @name: This is the first field of the "visible" part of this structure
1462 * (i.e. as seen by users in the "Space.c" file). It is the name
1465 * @name_hlist: Device name hash chain, please keep it close to name[]
1466 * @ifalias: SNMP alias
1467 * @mem_end: Shared memory end
1468 * @mem_start: Shared memory start
1469 * @base_addr: Device I/O address
1470 * @irq: Device IRQ number
1472 * @state: Generic network queuing layer state, see netdev_state_t
1473 * @dev_list: The global list of network devices
1474 * @napi_list: List entry used for polling NAPI devices
1475 * @unreg_list: List entry when we are unregistering the
1476 * device; see the function unregister_netdev
1477 * @close_list: List entry used when we are closing the device
1478 * @ptype_all: Device-specific packet handlers for all protocols
1479 * @ptype_specific: Device-specific, protocol-specific packet handlers
1481 * @adj_list: Directly linked devices, like slaves for bonding
1482 * @features: Currently active device features
1483 * @hw_features: User-changeable features
1485 * @wanted_features: User-requested features
1486 * @vlan_features: Mask of features inheritable by VLAN devices
1488 * @hw_enc_features: Mask of features inherited by encapsulating devices
1489 * This field indicates what encapsulation
1490 * offloads the hardware is capable of doing,
1491 * and drivers will need to set them appropriately.
1493 * @mpls_features: Mask of features inheritable by MPLS
1495 * @ifindex: interface index
1496 * @group: The group the device belongs to
1498 * @stats: Statistics struct, which was left as a legacy, use
1499 * rtnl_link_stats64 instead
1501 * @rx_dropped: Dropped packets by core network,
1502 * do not use this in drivers
1503 * @tx_dropped: Dropped packets by core network,
1504 * do not use this in drivers
1505 * @rx_nohandler: nohandler dropped packets by core network on
1506 * inactive devices, do not use this in drivers
1507 * @carrier_up_count: Number of times the carrier has been up
1508 * @carrier_down_count: Number of times the carrier has been down
1510 * @wireless_handlers: List of functions to handle Wireless Extensions,
1512 * see <net/iw_handler.h> for details.
1513 * @wireless_data: Instance data managed by the core of wireless extensions
1515 * @netdev_ops: Includes several pointers to callbacks,
1516 * if one wants to override the ndo_*() functions
1517 * @ethtool_ops: Management operations
1518 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1519 * discovery handling. Necessary for e.g. 6LoWPAN.
1520 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1521 * of Layer 2 headers.
1523 * @flags: Interface flags (a la BSD)
1524 * @priv_flags: Like 'flags' but invisible to userspace,
1525 * see if.h for the definitions
1526 * @gflags: Global flags ( kept as legacy )
1527 * @padded: How much padding added by alloc_netdev()
1528 * @operstate: RFC2863 operstate
1529 * @link_mode: Mapping policy to operstate
1530 * @if_port: Selectable AUI, TP, ...
1532 * @mtu: Interface MTU value
1533 * @min_mtu: Interface Minimum MTU value
1534 * @max_mtu: Interface Maximum MTU value
1535 * @type: Interface hardware type
1536 * @hard_header_len: Maximum hardware header length.
1537 * @min_header_len: Minimum hardware header length
1539 * @needed_headroom: Extra headroom the hardware may need, but not in all
1540 * cases can this be guaranteed
1541 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1542 * cases can this be guaranteed. Some cases also use
1543 * LL_MAX_HEADER instead to allocate the skb
1545 * interface address info:
1547 * @perm_addr: Permanent hw address
1548 * @addr_assign_type: Hw address assignment type
1549 * @addr_len: Hardware address length
1550 * @neigh_priv_len: Used in neigh_alloc()
1551 * @dev_id: Used to differentiate devices that share
1552 * the same link layer address
1553 * @dev_port: Used to differentiate devices that share
1555 * @addr_list_lock: XXX: need comments on this one
1556 * @uc_promisc: Counter that indicates promiscuous mode
1557 * has been enabled due to the need to listen to
1558 * additional unicast addresses in a device that
1559 * does not implement ndo_set_rx_mode()
1560 * @uc: unicast mac addresses
1561 * @mc: multicast mac addresses
1562 * @dev_addrs: list of device hw addresses
1563 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1564 * @promiscuity: Number of times the NIC is told to work in
1565 * promiscuous mode; if it becomes 0 the NIC will
1566 * exit promiscuous mode
1567 * @allmulti: Counter, enables or disables allmulticast mode
1569 * @vlan_info: VLAN info
1570 * @dsa_ptr: dsa specific data
1571 * @tipc_ptr: TIPC specific data
1572 * @atalk_ptr: AppleTalk link
1573 * @ip_ptr: IPv4 specific data
1574 * @dn_ptr: DECnet specific data
1575 * @ip6_ptr: IPv6 specific data
1576 * @ax25_ptr: AX.25 specific data
1577 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1579 * @dev_addr: Hw address (before bcast,
1580 * because most packets are unicast)
1582 * @_rx: Array of RX queues
1583 * @num_rx_queues: Number of RX queues
1584 * allocated at register_netdev() time
1585 * @real_num_rx_queues: Number of RX queues currently active in device
1587 * @rx_handler: handler for received packets
1588 * @rx_handler_data: XXX: need comments on this one
1589 * @miniq_ingress: ingress/clsact qdisc specific data for
1590 * ingress processing
1591 * @ingress_queue: XXX: need comments on this one
1592 * @broadcast: hw bcast address
1594 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1595 * indexed by RX queue number. Assigned by driver.
1596 * This must only be set if the ndo_rx_flow_steer
1597 * operation is defined
1598 * @index_hlist: Device index hash chain
1600 * @_tx: Array of TX queues
1601 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1602 * @real_num_tx_queues: Number of TX queues currently active in device
1603 * @qdisc: Root qdisc from userspace point of view
1604 * @tx_queue_len: Max frames per queue allowed
1605 * @tx_global_lock: XXX: need comments on this one
1607 * @xps_maps: XXX: need comments on this one
1608 * @miniq_egress: clsact qdisc specific data for
1610 * @watchdog_timeo: Represents the timeout that is used by
1611 * the watchdog (see dev_watchdog())
1612 * @watchdog_timer: List of timers
1614 * @pcpu_refcnt: Number of references to this device
1615 * @todo_list: Delayed register/unregister
1616 * @link_watch_list: XXX: need comments on this one
1618 * @reg_state: Register/unregister state machine
1619 * @dismantle: Device is going to be freed
1620 * @rtnl_link_state: This enum represents the phases of creating
1623 * @needs_free_netdev: Should unregister perform free_netdev?
1624 * @priv_destructor: Called from unregister
1625 * @npinfo: XXX: need comments on this one
1626 * @nd_net: Network namespace this network device is inside
1628 * @ml_priv: Mid-layer private
1629 * @lstats: Loopback statistics
1630 * @tstats: Tunnel statistics
1631 * @dstats: Dummy statistics
1632 * @vstats: Virtual ethernet statistics
1637 * @dev: Class/net/name entry
1638 * @sysfs_groups: Space for optional device, statistics and wireless
1641 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1642 * @rtnl_link_ops: Rtnl_link_ops
1644 * @gso_max_size: Maximum size of generic segmentation offload
1645 * @gso_max_segs: Maximum number of segments that can be passed to the
1648 * @dcbnl_ops: Data Center Bridging netlink ops
1649 * @num_tc: Number of traffic classes in the net device
1650 * @tc_to_txq: XXX: need comments on this one
1651 * @prio_tc_map: XXX: need comments on this one
1653 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1655 * @priomap: XXX: need comments on this one
1656 * @phydev: Physical device may attach itself
1657 * for hardware timestamping
1659 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1660 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1662 * @proto_down: protocol port state information can be sent to the
1663 * switch driver and used to set the phys state of the
1666 * FIXME: cleanup struct net_device such that network protocol info
1671 char name[IFNAMSIZ];
1672 struct hlist_node name_hlist;
1673 struct dev_ifalias __rcu *ifalias;
1675 * I/O specific fields
1676 * FIXME: Merge these and struct ifmap into one
1678 unsigned long mem_end;
1679 unsigned long mem_start;
1680 unsigned long base_addr;
1684 * Some hardware also needs these fields (state,dev_list,
1685 * napi_list,unreg_list,close_list) but they are not
1686 * part of the usual set specified in Space.c.
1689 unsigned long state;
1691 struct list_head dev_list;
1692 struct list_head napi_list;
1693 struct list_head unreg_list;
1694 struct list_head close_list;
1695 struct list_head ptype_all;
1696 struct list_head ptype_specific;
1699 struct list_head upper;
1700 struct list_head lower;
1703 netdev_features_t features;
1704 netdev_features_t hw_features;
1705 netdev_features_t wanted_features;
1706 netdev_features_t vlan_features;
1707 netdev_features_t hw_enc_features;
1708 netdev_features_t mpls_features;
1709 netdev_features_t gso_partial_features;
1714 struct net_device_stats stats;
1716 atomic_long_t rx_dropped;
1717 atomic_long_t tx_dropped;
1718 atomic_long_t rx_nohandler;
1720 /* Stats to monitor link on/off, flapping */
1721 atomic_t carrier_up_count;
1722 atomic_t carrier_down_count;
1724 #ifdef CONFIG_WIRELESS_EXT
1725 const struct iw_handler_def *wireless_handlers;
1726 struct iw_public_data *wireless_data;
1728 const struct net_device_ops *netdev_ops;
1729 const struct ethtool_ops *ethtool_ops;
1730 #ifdef CONFIG_NET_SWITCHDEV
1731 const struct switchdev_ops *switchdev_ops;
1733 #ifdef CONFIG_NET_L3_MASTER_DEV
1734 const struct l3mdev_ops *l3mdev_ops;
1736 #if IS_ENABLED(CONFIG_IPV6)
1737 const struct ndisc_ops *ndisc_ops;
1740 #ifdef CONFIG_XFRM_OFFLOAD
1741 const struct xfrmdev_ops *xfrmdev_ops;
1744 const struct header_ops *header_ops;
1747 unsigned int priv_flags;
1749 unsigned short gflags;
1750 unsigned short padded;
1752 unsigned char operstate;
1753 unsigned char link_mode;
1755 unsigned char if_port;
1759 unsigned int min_mtu;
1760 unsigned int max_mtu;
1761 unsigned short type;
1762 unsigned short hard_header_len;
1763 unsigned char min_header_len;
1765 unsigned short needed_headroom;
1766 unsigned short needed_tailroom;
1768 /* Interface address info. */
1769 unsigned char perm_addr[MAX_ADDR_LEN];
1770 unsigned char addr_assign_type;
1771 unsigned char addr_len;
1772 unsigned short neigh_priv_len;
1773 unsigned short dev_id;
1774 unsigned short dev_port;
1775 spinlock_t addr_list_lock;
1776 unsigned char name_assign_type;
1778 struct netdev_hw_addr_list uc;
1779 struct netdev_hw_addr_list mc;
1780 struct netdev_hw_addr_list dev_addrs;
1783 struct kset *queues_kset;
1785 unsigned int promiscuity;
1786 unsigned int allmulti;
1789 /* Protocol-specific pointers */
1791 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1792 struct vlan_info __rcu *vlan_info;
1794 #if IS_ENABLED(CONFIG_NET_DSA)
1795 struct dsa_port *dsa_ptr;
1797 #if IS_ENABLED(CONFIG_TIPC)
1798 struct tipc_bearer __rcu *tipc_ptr;
1801 struct in_device __rcu *ip_ptr;
1802 struct dn_dev __rcu *dn_ptr;
1803 struct inet6_dev __rcu *ip6_ptr;
1805 struct wireless_dev *ieee80211_ptr;
1806 struct wpan_dev *ieee802154_ptr;
1807 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1808 struct mpls_dev __rcu *mpls_ptr;
1812 * Cache lines mostly used on receive path (including eth_type_trans())
1814 /* Interface address info used in eth_type_trans() */
1815 unsigned char *dev_addr;
1817 struct netdev_rx_queue *_rx;
1818 unsigned int num_rx_queues;
1819 unsigned int real_num_rx_queues;
1821 struct bpf_prog __rcu *xdp_prog;
1822 unsigned long gro_flush_timeout;
1823 rx_handler_func_t __rcu *rx_handler;
1824 void __rcu *rx_handler_data;
1826 #ifdef CONFIG_NET_CLS_ACT
1827 struct mini_Qdisc __rcu *miniq_ingress;
1829 struct netdev_queue __rcu *ingress_queue;
1830 #ifdef CONFIG_NETFILTER_INGRESS
1831 struct nf_hook_entries __rcu *nf_hooks_ingress;
1834 unsigned char broadcast[MAX_ADDR_LEN];
1835 #ifdef CONFIG_RFS_ACCEL
1836 struct cpu_rmap *rx_cpu_rmap;
1838 struct hlist_node index_hlist;
1841 * Cache lines mostly used on transmit path
1843 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1844 unsigned int num_tx_queues;
1845 unsigned int real_num_tx_queues;
1846 struct Qdisc *qdisc;
1847 #ifdef CONFIG_NET_SCHED
1848 DECLARE_HASHTABLE (qdisc_hash, 4);
1850 unsigned int tx_queue_len;
1851 spinlock_t tx_global_lock;
1855 struct xps_dev_maps __rcu *xps_maps;
1857 #ifdef CONFIG_NET_CLS_ACT
1858 struct mini_Qdisc __rcu *miniq_egress;
1861 /* These may be needed for future network-power-down code. */
1862 struct timer_list watchdog_timer;
1864 int __percpu *pcpu_refcnt;
1865 struct list_head todo_list;
1867 struct list_head link_watch_list;
1869 enum { NETREG_UNINITIALIZED=0,
1870 NETREG_REGISTERED, /* completed register_netdevice */
1871 NETREG_UNREGISTERING, /* called unregister_netdevice */
1872 NETREG_UNREGISTERED, /* completed unregister todo */
1873 NETREG_RELEASED, /* called free_netdev */
1874 NETREG_DUMMY, /* dummy device for NAPI poll */
1880 RTNL_LINK_INITIALIZED,
1881 RTNL_LINK_INITIALIZING,
1882 } rtnl_link_state:16;
1884 bool needs_free_netdev;
1885 void (*priv_destructor)(struct net_device *dev);
1887 #ifdef CONFIG_NETPOLL
1888 struct netpoll_info __rcu *npinfo;
1891 possible_net_t nd_net;
1893 /* mid-layer private */
1896 struct pcpu_lstats __percpu *lstats;
1897 struct pcpu_sw_netstats __percpu *tstats;
1898 struct pcpu_dstats __percpu *dstats;
1899 struct pcpu_vstats __percpu *vstats;
1902 #if IS_ENABLED(CONFIG_GARP)
1903 struct garp_port __rcu *garp_port;
1905 #if IS_ENABLED(CONFIG_MRP)
1906 struct mrp_port __rcu *mrp_port;
1910 const struct attribute_group *sysfs_groups[4];
1911 const struct attribute_group *sysfs_rx_queue_group;
1913 const struct rtnl_link_ops *rtnl_link_ops;
1915 /* for setting kernel sock attribute on TCP connection setup */
1916 #define GSO_MAX_SIZE 65536
1917 unsigned int gso_max_size;
1918 #define GSO_MAX_SEGS 65535
1922 const struct dcbnl_rtnl_ops *dcbnl_ops;
1925 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1926 u8 prio_tc_map[TC_BITMASK + 1];
1928 #if IS_ENABLED(CONFIG_FCOE)
1929 unsigned int fcoe_ddp_xid;
1931 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1932 struct netprio_map __rcu *priomap;
1934 struct phy_device *phydev;
1935 struct lock_class_key *qdisc_tx_busylock;
1936 struct lock_class_key *qdisc_running_key;
1939 #define to_net_dev(d) container_of(d, struct net_device, dev)
1941 static inline bool netif_elide_gro(const struct net_device *dev)
1943 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1948 #define NETDEV_ALIGN 32
1951 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1953 return dev->prio_tc_map[prio & TC_BITMASK];
1957 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1959 if (tc >= dev->num_tc)
1962 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1966 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1967 void netdev_reset_tc(struct net_device *dev);
1968 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1969 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1972 int netdev_get_num_tc(struct net_device *dev)
1978 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1981 return &dev->_tx[index];
1984 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1985 const struct sk_buff *skb)
1987 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1990 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1991 void (*f)(struct net_device *,
1992 struct netdev_queue *,
1998 for (i = 0; i < dev->num_tx_queues; i++)
1999 f(dev, &dev->_tx[i], arg);
2002 #define netdev_lockdep_set_classes(dev) \
2004 static struct lock_class_key qdisc_tx_busylock_key; \
2005 static struct lock_class_key qdisc_running_key; \
2006 static struct lock_class_key qdisc_xmit_lock_key; \
2007 static struct lock_class_key dev_addr_list_lock_key; \
2010 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2011 (dev)->qdisc_running_key = &qdisc_running_key; \
2012 lockdep_set_class(&(dev)->addr_list_lock, \
2013 &dev_addr_list_lock_key); \
2014 for (i = 0; i < (dev)->num_tx_queues; i++) \
2015 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2016 &qdisc_xmit_lock_key); \
2019 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2020 struct sk_buff *skb,
2023 /* returns the headroom that the master device needs to take in account
2024 * when forwarding to this dev
2026 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2028 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2031 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2033 if (dev->netdev_ops->ndo_set_rx_headroom)
2034 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2037 /* set the device rx headroom to the dev's default */
2038 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2040 netdev_set_rx_headroom(dev, -1);
2044 * Net namespace inlines
2047 struct net *dev_net(const struct net_device *dev)
2049 return read_pnet(&dev->nd_net);
2053 void dev_net_set(struct net_device *dev, struct net *net)
2055 write_pnet(&dev->nd_net, net);
2059 * netdev_priv - access network device private data
2060 * @dev: network device
2062 * Get network device private data
2064 static inline void *netdev_priv(const struct net_device *dev)
2066 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2069 /* Set the sysfs physical device reference for the network logical device
2070 * if set prior to registration will cause a symlink during initialization.
2072 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2074 /* Set the sysfs device type for the network logical device to allow
2075 * fine-grained identification of different network device types. For
2076 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2078 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2080 /* Default NAPI poll() weight
2081 * Device drivers are strongly advised to not use bigger value
2083 #define NAPI_POLL_WEIGHT 64
2086 * netif_napi_add - initialize a NAPI context
2087 * @dev: network device
2088 * @napi: NAPI context
2089 * @poll: polling function
2090 * @weight: default weight
2092 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2093 * *any* of the other NAPI-related functions.
2095 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2096 int (*poll)(struct napi_struct *, int), int weight);
2099 * netif_tx_napi_add - initialize a NAPI context
2100 * @dev: network device
2101 * @napi: NAPI context
2102 * @poll: polling function
2103 * @weight: default weight
2105 * This variant of netif_napi_add() should be used from drivers using NAPI
2106 * to exclusively poll a TX queue.
2107 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2109 static inline void netif_tx_napi_add(struct net_device *dev,
2110 struct napi_struct *napi,
2111 int (*poll)(struct napi_struct *, int),
2114 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2115 netif_napi_add(dev, napi, poll, weight);
2119 * netif_napi_del - remove a NAPI context
2120 * @napi: NAPI context
2122 * netif_napi_del() removes a NAPI context from the network device NAPI list
2124 void netif_napi_del(struct napi_struct *napi);
2126 struct napi_gro_cb {
2127 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2130 /* Length of frag0. */
2131 unsigned int frag0_len;
2133 /* This indicates where we are processing relative to skb->data. */
2136 /* This is non-zero if the packet cannot be merged with the new skb. */
2139 /* Save the IP ID here and check when we get to the transport layer */
2142 /* Number of segments aggregated. */
2145 /* Start offset for remote checksum offload */
2146 u16 gro_remcsum_start;
2148 /* jiffies when first packet was created/queued */
2151 /* Used in ipv6_gro_receive() and foo-over-udp */
2154 /* This is non-zero if the packet may be of the same flow. */
2157 /* Used in tunnel GRO receive */
2160 /* GRO checksum is valid */
2163 /* Number of checksums via CHECKSUM_UNNECESSARY */
2168 #define NAPI_GRO_FREE 1
2169 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2171 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2174 /* Used in GRE, set in fou/gue_gro_receive */
2177 /* Used to determine if flush_id can be ignored */
2180 /* Number of gro_receive callbacks this packet already went through */
2181 u8 recursion_counter:4;
2185 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2188 /* used in skb_gro_receive() slow path */
2189 struct sk_buff *last;
2192 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2194 #define GRO_RECURSION_LIMIT 15
2195 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2197 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2200 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2201 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2202 struct sk_buff **head,
2203 struct sk_buff *skb)
2205 if (unlikely(gro_recursion_inc_test(skb))) {
2206 NAPI_GRO_CB(skb)->flush |= 1;
2210 return cb(head, skb);
2213 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2215 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2217 struct sk_buff **head,
2218 struct sk_buff *skb)
2220 if (unlikely(gro_recursion_inc_test(skb))) {
2221 NAPI_GRO_CB(skb)->flush |= 1;
2225 return cb(sk, head, skb);
2228 struct packet_type {
2229 __be16 type; /* This is really htons(ether_type). */
2230 struct net_device *dev; /* NULL is wildcarded here */
2231 int (*func) (struct sk_buff *,
2232 struct net_device *,
2233 struct packet_type *,
2234 struct net_device *);
2235 bool (*id_match)(struct packet_type *ptype,
2237 void *af_packet_priv;
2238 struct list_head list;
2241 struct offload_callbacks {
2242 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2243 netdev_features_t features);
2244 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2245 struct sk_buff *skb);
2246 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2249 struct packet_offload {
2250 __be16 type; /* This is really htons(ether_type). */
2252 struct offload_callbacks callbacks;
2253 struct list_head list;
2256 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2257 struct pcpu_sw_netstats {
2262 struct u64_stats_sync syncp;
2265 #define __netdev_alloc_pcpu_stats(type, gfp) \
2267 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2270 for_each_possible_cpu(__cpu) { \
2271 typeof(type) *stat; \
2272 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2273 u64_stats_init(&stat->syncp); \
2279 #define netdev_alloc_pcpu_stats(type) \
2280 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2282 enum netdev_lag_tx_type {
2283 NETDEV_LAG_TX_TYPE_UNKNOWN,
2284 NETDEV_LAG_TX_TYPE_RANDOM,
2285 NETDEV_LAG_TX_TYPE_BROADCAST,
2286 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2287 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2288 NETDEV_LAG_TX_TYPE_HASH,
2291 struct netdev_lag_upper_info {
2292 enum netdev_lag_tx_type tx_type;
2295 struct netdev_lag_lower_state_info {
2300 #include <linux/notifier.h>
2302 /* netdevice notifier chain. Please remember to update the rtnetlink
2303 * notification exclusion list in rtnetlink_event() when adding new
2306 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2307 #define NETDEV_DOWN 0x0002
2308 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2309 detected a hardware crash and restarted
2310 - we can use this eg to kick tcp sessions
2312 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2313 #define NETDEV_REGISTER 0x0005
2314 #define NETDEV_UNREGISTER 0x0006
2315 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2316 #define NETDEV_CHANGEADDR 0x0008
2317 #define NETDEV_GOING_DOWN 0x0009
2318 #define NETDEV_CHANGENAME 0x000A
2319 #define NETDEV_FEAT_CHANGE 0x000B
2320 #define NETDEV_BONDING_FAILOVER 0x000C
2321 #define NETDEV_PRE_UP 0x000D
2322 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2323 #define NETDEV_POST_TYPE_CHANGE 0x000F
2324 #define NETDEV_POST_INIT 0x0010
2325 #define NETDEV_UNREGISTER_FINAL 0x0011
2326 #define NETDEV_RELEASE 0x0012
2327 #define NETDEV_NOTIFY_PEERS 0x0013
2328 #define NETDEV_JOIN 0x0014
2329 #define NETDEV_CHANGEUPPER 0x0015
2330 #define NETDEV_RESEND_IGMP 0x0016
2331 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2332 #define NETDEV_CHANGEINFODATA 0x0018
2333 #define NETDEV_BONDING_INFO 0x0019
2334 #define NETDEV_PRECHANGEUPPER 0x001A
2335 #define NETDEV_CHANGELOWERSTATE 0x001B
2336 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2337 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
2338 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2340 int register_netdevice_notifier(struct notifier_block *nb);
2341 int unregister_netdevice_notifier(struct notifier_block *nb);
2343 struct netdev_notifier_info {
2344 struct net_device *dev;
2345 struct netlink_ext_ack *extack;
2348 struct netdev_notifier_change_info {
2349 struct netdev_notifier_info info; /* must be first */
2350 unsigned int flags_changed;
2353 struct netdev_notifier_changeupper_info {
2354 struct netdev_notifier_info info; /* must be first */
2355 struct net_device *upper_dev; /* new upper dev */
2356 bool master; /* is upper dev master */
2357 bool linking; /* is the notification for link or unlink */
2358 void *upper_info; /* upper dev info */
2361 struct netdev_notifier_changelowerstate_info {
2362 struct netdev_notifier_info info; /* must be first */
2363 void *lower_state_info; /* is lower dev state */
2366 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2367 struct net_device *dev)
2370 info->extack = NULL;
2373 static inline struct net_device *
2374 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2379 static inline struct netlink_ext_ack *
2380 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2382 return info->extack;
2385 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2388 extern rwlock_t dev_base_lock; /* Device list lock */
2390 #define for_each_netdev(net, d) \
2391 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2392 #define for_each_netdev_reverse(net, d) \
2393 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2394 #define for_each_netdev_rcu(net, d) \
2395 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2396 #define for_each_netdev_safe(net, d, n) \
2397 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2398 #define for_each_netdev_continue(net, d) \
2399 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2400 #define for_each_netdev_continue_rcu(net, d) \
2401 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2402 #define for_each_netdev_in_bond_rcu(bond, slave) \
2403 for_each_netdev_rcu(&init_net, slave) \
2404 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2405 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2407 static inline struct net_device *next_net_device(struct net_device *dev)
2409 struct list_head *lh;
2413 lh = dev->dev_list.next;
2414 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2417 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2419 struct list_head *lh;
2423 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2424 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2427 static inline struct net_device *first_net_device(struct net *net)
2429 return list_empty(&net->dev_base_head) ? NULL :
2430 net_device_entry(net->dev_base_head.next);
2433 static inline struct net_device *first_net_device_rcu(struct net *net)
2435 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2437 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2440 int netdev_boot_setup_check(struct net_device *dev);
2441 unsigned long netdev_boot_base(const char *prefix, int unit);
2442 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2443 const char *hwaddr);
2444 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2445 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2446 void dev_add_pack(struct packet_type *pt);
2447 void dev_remove_pack(struct packet_type *pt);
2448 void __dev_remove_pack(struct packet_type *pt);
2449 void dev_add_offload(struct packet_offload *po);
2450 void dev_remove_offload(struct packet_offload *po);
2452 int dev_get_iflink(const struct net_device *dev);
2453 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2454 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2455 unsigned short mask);
2456 struct net_device *dev_get_by_name(struct net *net, const char *name);
2457 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2458 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2459 int dev_alloc_name(struct net_device *dev, const char *name);
2460 int dev_open(struct net_device *dev);
2461 void dev_close(struct net_device *dev);
2462 void dev_close_many(struct list_head *head, bool unlink);
2463 void dev_disable_lro(struct net_device *dev);
2464 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2465 int dev_queue_xmit(struct sk_buff *skb);
2466 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2467 int register_netdevice(struct net_device *dev);
2468 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2469 void unregister_netdevice_many(struct list_head *head);
2470 static inline void unregister_netdevice(struct net_device *dev)
2472 unregister_netdevice_queue(dev, NULL);
2475 int netdev_refcnt_read(const struct net_device *dev);
2476 void free_netdev(struct net_device *dev);
2477 void netdev_freemem(struct net_device *dev);
2478 void synchronize_net(void);
2479 int init_dummy_netdev(struct net_device *dev);
2481 DECLARE_PER_CPU(int, xmit_recursion);
2482 #define XMIT_RECURSION_LIMIT 10
2484 static inline int dev_recursion_level(void)
2486 return this_cpu_read(xmit_recursion);
2489 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2490 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2491 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2492 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2493 int netdev_get_name(struct net *net, char *name, int ifindex);
2494 int dev_restart(struct net_device *dev);
2495 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2497 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2499 return NAPI_GRO_CB(skb)->data_offset;
2502 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2504 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2507 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2509 NAPI_GRO_CB(skb)->data_offset += len;
2512 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2513 unsigned int offset)
2515 return NAPI_GRO_CB(skb)->frag0 + offset;
2518 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2520 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2523 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2525 NAPI_GRO_CB(skb)->frag0 = NULL;
2526 NAPI_GRO_CB(skb)->frag0_len = 0;
2529 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2530 unsigned int offset)
2532 if (!pskb_may_pull(skb, hlen))
2535 skb_gro_frag0_invalidate(skb);
2536 return skb->data + offset;
2539 static inline void *skb_gro_network_header(struct sk_buff *skb)
2541 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2542 skb_network_offset(skb);
2545 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2546 const void *start, unsigned int len)
2548 if (NAPI_GRO_CB(skb)->csum_valid)
2549 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2550 csum_partial(start, len, 0));
2553 /* GRO checksum functions. These are logical equivalents of the normal
2554 * checksum functions (in skbuff.h) except that they operate on the GRO
2555 * offsets and fields in sk_buff.
2558 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2560 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2562 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2565 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2569 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2570 skb_checksum_start_offset(skb) <
2571 skb_gro_offset(skb)) &&
2572 !skb_at_gro_remcsum_start(skb) &&
2573 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2574 (!zero_okay || check));
2577 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2580 if (NAPI_GRO_CB(skb)->csum_valid &&
2581 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2584 NAPI_GRO_CB(skb)->csum = psum;
2586 return __skb_gro_checksum_complete(skb);
2589 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2591 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2592 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2593 NAPI_GRO_CB(skb)->csum_cnt--;
2595 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2596 * verified a new top level checksum or an encapsulated one
2597 * during GRO. This saves work if we fallback to normal path.
2599 __skb_incr_checksum_unnecessary(skb);
2603 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2606 __sum16 __ret = 0; \
2607 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2608 __ret = __skb_gro_checksum_validate_complete(skb, \
2609 compute_pseudo(skb, proto)); \
2611 skb_gro_incr_csum_unnecessary(skb); \
2615 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2616 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2618 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2620 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2622 #define skb_gro_checksum_simple_validate(skb) \
2623 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2625 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2627 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2628 !NAPI_GRO_CB(skb)->csum_valid);
2631 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2632 __sum16 check, __wsum pseudo)
2634 NAPI_GRO_CB(skb)->csum = ~pseudo;
2635 NAPI_GRO_CB(skb)->csum_valid = 1;
2638 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2640 if (__skb_gro_checksum_convert_check(skb)) \
2641 __skb_gro_checksum_convert(skb, check, \
2642 compute_pseudo(skb, proto)); \
2645 struct gro_remcsum {
2650 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2656 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2657 unsigned int off, size_t hdrlen,
2658 int start, int offset,
2659 struct gro_remcsum *grc,
2663 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2665 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2668 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2672 ptr = skb_gro_header_fast(skb, off);
2673 if (skb_gro_header_hard(skb, off + plen)) {
2674 ptr = skb_gro_header_slow(skb, off + plen, off);
2679 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2682 /* Adjust skb->csum since we changed the packet */
2683 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2685 grc->offset = off + hdrlen + offset;
2691 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2692 struct gro_remcsum *grc)
2695 size_t plen = grc->offset + sizeof(u16);
2700 ptr = skb_gro_header_fast(skb, grc->offset);
2701 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2702 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2707 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2710 #ifdef CONFIG_XFRM_OFFLOAD
2711 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2713 if (PTR_ERR(pp) != -EINPROGRESS)
2714 NAPI_GRO_CB(skb)->flush |= flush;
2717 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2719 NAPI_GRO_CB(skb)->flush |= flush;
2723 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2724 unsigned short type,
2725 const void *daddr, const void *saddr,
2728 if (!dev->header_ops || !dev->header_ops->create)
2731 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2734 static inline int dev_parse_header(const struct sk_buff *skb,
2735 unsigned char *haddr)
2737 const struct net_device *dev = skb->dev;
2739 if (!dev->header_ops || !dev->header_ops->parse)
2741 return dev->header_ops->parse(skb, haddr);
2744 /* ll_header must have at least hard_header_len allocated */
2745 static inline bool dev_validate_header(const struct net_device *dev,
2746 char *ll_header, int len)
2748 if (likely(len >= dev->hard_header_len))
2750 if (len < dev->min_header_len)
2753 if (capable(CAP_SYS_RAWIO)) {
2754 memset(ll_header + len, 0, dev->hard_header_len - len);
2758 if (dev->header_ops && dev->header_ops->validate)
2759 return dev->header_ops->validate(ll_header, len);
2764 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2765 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2766 static inline int unregister_gifconf(unsigned int family)
2768 return register_gifconf(family, NULL);
2771 #ifdef CONFIG_NET_FLOW_LIMIT
2772 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2773 struct sd_flow_limit {
2775 unsigned int num_buckets;
2776 unsigned int history_head;
2777 u16 history[FLOW_LIMIT_HISTORY];
2781 extern int netdev_flow_limit_table_len;
2782 #endif /* CONFIG_NET_FLOW_LIMIT */
2785 * Incoming packets are placed on per-CPU queues
2787 struct softnet_data {
2788 struct list_head poll_list;
2789 struct sk_buff_head process_queue;
2792 unsigned int processed;
2793 unsigned int time_squeeze;
2794 unsigned int received_rps;
2796 struct softnet_data *rps_ipi_list;
2798 #ifdef CONFIG_NET_FLOW_LIMIT
2799 struct sd_flow_limit __rcu *flow_limit;
2801 struct Qdisc *output_queue;
2802 struct Qdisc **output_queue_tailp;
2803 struct sk_buff *completion_queue;
2804 #ifdef CONFIG_XFRM_OFFLOAD
2805 struct sk_buff_head xfrm_backlog;
2808 /* input_queue_head should be written by cpu owning this struct,
2809 * and only read by other cpus. Worth using a cache line.
2811 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2813 /* Elements below can be accessed between CPUs for RPS/RFS */
2814 call_single_data_t csd ____cacheline_aligned_in_smp;
2815 struct softnet_data *rps_ipi_next;
2817 unsigned int input_queue_tail;
2819 unsigned int dropped;
2820 struct sk_buff_head input_pkt_queue;
2821 struct napi_struct backlog;
2825 static inline void input_queue_head_incr(struct softnet_data *sd)
2828 sd->input_queue_head++;
2832 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2833 unsigned int *qtail)
2836 *qtail = ++sd->input_queue_tail;
2840 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2842 void __netif_schedule(struct Qdisc *q);
2843 void netif_schedule_queue(struct netdev_queue *txq);
2845 static inline void netif_tx_schedule_all(struct net_device *dev)
2849 for (i = 0; i < dev->num_tx_queues; i++)
2850 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2853 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2855 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2859 * netif_start_queue - allow transmit
2860 * @dev: network device
2862 * Allow upper layers to call the device hard_start_xmit routine.
2864 static inline void netif_start_queue(struct net_device *dev)
2866 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2869 static inline void netif_tx_start_all_queues(struct net_device *dev)
2873 for (i = 0; i < dev->num_tx_queues; i++) {
2874 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2875 netif_tx_start_queue(txq);
2879 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2882 * netif_wake_queue - restart transmit
2883 * @dev: network device
2885 * Allow upper layers to call the device hard_start_xmit routine.
2886 * Used for flow control when transmit resources are available.
2888 static inline void netif_wake_queue(struct net_device *dev)
2890 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2893 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2897 for (i = 0; i < dev->num_tx_queues; i++) {
2898 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2899 netif_tx_wake_queue(txq);
2903 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2905 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2909 * netif_stop_queue - stop transmitted packets
2910 * @dev: network device
2912 * Stop upper layers calling the device hard_start_xmit routine.
2913 * Used for flow control when transmit resources are unavailable.
2915 static inline void netif_stop_queue(struct net_device *dev)
2917 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2920 void netif_tx_stop_all_queues(struct net_device *dev);
2922 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2924 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2928 * netif_queue_stopped - test if transmit queue is flowblocked
2929 * @dev: network device
2931 * Test if transmit queue on device is currently unable to send.
2933 static inline bool netif_queue_stopped(const struct net_device *dev)
2935 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2938 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2940 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2944 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2946 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2950 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2952 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2956 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2957 * @dev_queue: pointer to transmit queue
2959 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2960 * to give appropriate hint to the CPU.
2962 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2965 prefetchw(&dev_queue->dql.num_queued);
2970 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2971 * @dev_queue: pointer to transmit queue
2973 * BQL enabled drivers might use this helper in their TX completion path,
2974 * to give appropriate hint to the CPU.
2976 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2979 prefetchw(&dev_queue->dql.limit);
2983 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2987 dql_queued(&dev_queue->dql, bytes);
2989 if (likely(dql_avail(&dev_queue->dql) >= 0))
2992 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2995 * The XOFF flag must be set before checking the dql_avail below,
2996 * because in netdev_tx_completed_queue we update the dql_completed
2997 * before checking the XOFF flag.
3001 /* check again in case another CPU has just made room avail */
3002 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3003 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3008 * netdev_sent_queue - report the number of bytes queued to hardware
3009 * @dev: network device
3010 * @bytes: number of bytes queued to the hardware device queue
3012 * Report the number of bytes queued for sending/completion to the network
3013 * device hardware queue. @bytes should be a good approximation and should
3014 * exactly match netdev_completed_queue() @bytes
3016 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3018 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3021 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3022 unsigned int pkts, unsigned int bytes)
3025 if (unlikely(!bytes))
3028 dql_completed(&dev_queue->dql, bytes);
3031 * Without the memory barrier there is a small possiblity that
3032 * netdev_tx_sent_queue will miss the update and cause the queue to
3033 * be stopped forever
3037 if (dql_avail(&dev_queue->dql) < 0)
3040 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3041 netif_schedule_queue(dev_queue);
3046 * netdev_completed_queue - report bytes and packets completed by device
3047 * @dev: network device
3048 * @pkts: actual number of packets sent over the medium
3049 * @bytes: actual number of bytes sent over the medium
3051 * Report the number of bytes and packets transmitted by the network device
3052 * hardware queue over the physical medium, @bytes must exactly match the
3053 * @bytes amount passed to netdev_sent_queue()
3055 static inline void netdev_completed_queue(struct net_device *dev,
3056 unsigned int pkts, unsigned int bytes)
3058 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3061 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3064 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3070 * netdev_reset_queue - reset the packets and bytes count of a network device
3071 * @dev_queue: network device
3073 * Reset the bytes and packet count of a network device and clear the
3074 * software flow control OFF bit for this network device
3076 static inline void netdev_reset_queue(struct net_device *dev_queue)
3078 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3082 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3083 * @dev: network device
3084 * @queue_index: given tx queue index
3086 * Returns 0 if given tx queue index >= number of device tx queues,
3087 * otherwise returns the originally passed tx queue index.
3089 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3091 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3092 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3093 dev->name, queue_index,
3094 dev->real_num_tx_queues);
3102 * netif_running - test if up
3103 * @dev: network device
3105 * Test if the device has been brought up.
3107 static inline bool netif_running(const struct net_device *dev)
3109 return test_bit(__LINK_STATE_START, &dev->state);
3113 * Routines to manage the subqueues on a device. We only need start,
3114 * stop, and a check if it's stopped. All other device management is
3115 * done at the overall netdevice level.
3116 * Also test the device if we're multiqueue.
3120 * netif_start_subqueue - allow sending packets on subqueue
3121 * @dev: network device
3122 * @queue_index: sub queue index
3124 * Start individual transmit queue of a device with multiple transmit queues.
3126 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3128 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3130 netif_tx_start_queue(txq);
3134 * netif_stop_subqueue - stop sending packets on subqueue
3135 * @dev: network device
3136 * @queue_index: sub queue index
3138 * Stop individual transmit queue of a device with multiple transmit queues.
3140 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3142 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3143 netif_tx_stop_queue(txq);
3147 * netif_subqueue_stopped - test status of subqueue
3148 * @dev: network device
3149 * @queue_index: sub queue index
3151 * Check individual transmit queue of a device with multiple transmit queues.
3153 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3156 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3158 return netif_tx_queue_stopped(txq);
3161 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3162 struct sk_buff *skb)
3164 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3168 * netif_wake_subqueue - allow sending packets on subqueue
3169 * @dev: network device
3170 * @queue_index: sub queue index
3172 * Resume individual transmit queue of a device with multiple transmit queues.
3174 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3176 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3178 netif_tx_wake_queue(txq);
3182 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3185 static inline int netif_set_xps_queue(struct net_device *dev,
3186 const struct cpumask *mask,
3193 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3194 unsigned int num_tx_queues);
3197 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3198 * as a distribution range limit for the returned value.
3200 static inline u16 skb_tx_hash(const struct net_device *dev,
3201 struct sk_buff *skb)
3203 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3207 * netif_is_multiqueue - test if device has multiple transmit queues
3208 * @dev: network device
3210 * Check if device has multiple transmit queues
3212 static inline bool netif_is_multiqueue(const struct net_device *dev)
3214 return dev->num_tx_queues > 1;
3217 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3220 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3222 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3230 static inline unsigned int get_netdev_rx_queue_index(
3231 struct netdev_rx_queue *queue)
3233 struct net_device *dev = queue->dev;
3234 int index = queue - dev->_rx;
3236 BUG_ON(index >= dev->num_rx_queues);
3241 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3242 int netif_get_num_default_rss_queues(void);
3244 enum skb_free_reason {
3245 SKB_REASON_CONSUMED,
3249 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3250 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3253 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3254 * interrupt context or with hardware interrupts being disabled.
3255 * (in_irq() || irqs_disabled())
3257 * We provide four helpers that can be used in following contexts :
3259 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3260 * replacing kfree_skb(skb)
3262 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3263 * Typically used in place of consume_skb(skb) in TX completion path
3265 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3266 * replacing kfree_skb(skb)
3268 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3269 * and consumed a packet. Used in place of consume_skb(skb)
3271 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3273 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3276 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3278 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3281 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3283 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3286 static inline void dev_consume_skb_any(struct sk_buff *skb)
3288 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3291 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3292 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3293 int netif_rx(struct sk_buff *skb);
3294 int netif_rx_ni(struct sk_buff *skb);
3295 int netif_receive_skb(struct sk_buff *skb);
3296 int netif_receive_skb_core(struct sk_buff *skb);
3297 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3298 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3299 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3300 gro_result_t napi_gro_frags(struct napi_struct *napi);
3301 struct packet_offload *gro_find_receive_by_type(__be16 type);
3302 struct packet_offload *gro_find_complete_by_type(__be16 type);
3304 static inline void napi_free_frags(struct napi_struct *napi)
3306 kfree_skb(napi->skb);
3310 bool netdev_is_rx_handler_busy(struct net_device *dev);
3311 int netdev_rx_handler_register(struct net_device *dev,
3312 rx_handler_func_t *rx_handler,
3313 void *rx_handler_data);
3314 void netdev_rx_handler_unregister(struct net_device *dev);
3316 bool dev_valid_name(const char *name);
3317 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3318 int dev_ethtool(struct net *net, struct ifreq *);
3319 unsigned int dev_get_flags(const struct net_device *);
3320 int __dev_change_flags(struct net_device *, unsigned int flags);
3321 int dev_change_flags(struct net_device *, unsigned int);
3322 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3323 unsigned int gchanges);
3324 int dev_change_name(struct net_device *, const char *);
3325 int dev_set_alias(struct net_device *, const char *, size_t);
3326 int dev_get_alias(const struct net_device *, char *, size_t);
3327 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3328 int __dev_set_mtu(struct net_device *, int);
3329 int dev_set_mtu(struct net_device *, int);
3330 void dev_set_group(struct net_device *, int);
3331 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3332 int dev_change_carrier(struct net_device *, bool new_carrier);
3333 int dev_get_phys_port_id(struct net_device *dev,
3334 struct netdev_phys_item_id *ppid);
3335 int dev_get_phys_port_name(struct net_device *dev,
3336 char *name, size_t len);
3337 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3338 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3339 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3340 struct netdev_queue *txq, int *ret);
3342 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3343 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3345 void __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3346 struct netdev_bpf *xdp);
3348 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3349 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3350 bool is_skb_forwardable(const struct net_device *dev,
3351 const struct sk_buff *skb);
3353 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3354 struct sk_buff *skb)
3356 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3357 unlikely(!is_skb_forwardable(dev, skb))) {
3358 atomic_long_inc(&dev->rx_dropped);
3363 skb_scrub_packet(skb, true);
3368 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3370 extern int netdev_budget;
3371 extern unsigned int netdev_budget_usecs;
3373 /* Called by rtnetlink.c:rtnl_unlock() */
3374 void netdev_run_todo(void);
3377 * dev_put - release reference to device
3378 * @dev: network device
3380 * Release reference to device to allow it to be freed.
3382 static inline void dev_put(struct net_device *dev)
3384 this_cpu_dec(*dev->pcpu_refcnt);
3388 * dev_hold - get reference to device
3389 * @dev: network device
3391 * Hold reference to device to keep it from being freed.
3393 static inline void dev_hold(struct net_device *dev)
3395 this_cpu_inc(*dev->pcpu_refcnt);
3398 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3399 * and _off may be called from IRQ context, but it is caller
3400 * who is responsible for serialization of these calls.
3402 * The name carrier is inappropriate, these functions should really be
3403 * called netif_lowerlayer_*() because they represent the state of any
3404 * kind of lower layer not just hardware media.
3407 void linkwatch_init_dev(struct net_device *dev);
3408 void linkwatch_fire_event(struct net_device *dev);
3409 void linkwatch_forget_dev(struct net_device *dev);
3412 * netif_carrier_ok - test if carrier present
3413 * @dev: network device
3415 * Check if carrier is present on device
3417 static inline bool netif_carrier_ok(const struct net_device *dev)
3419 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3422 unsigned long dev_trans_start(struct net_device *dev);
3424 void __netdev_watchdog_up(struct net_device *dev);
3426 void netif_carrier_on(struct net_device *dev);
3428 void netif_carrier_off(struct net_device *dev);
3431 * netif_dormant_on - mark device as dormant.
3432 * @dev: network device
3434 * Mark device as dormant (as per RFC2863).
3436 * The dormant state indicates that the relevant interface is not
3437 * actually in a condition to pass packets (i.e., it is not 'up') but is
3438 * in a "pending" state, waiting for some external event. For "on-
3439 * demand" interfaces, this new state identifies the situation where the
3440 * interface is waiting for events to place it in the up state.
3442 static inline void netif_dormant_on(struct net_device *dev)
3444 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3445 linkwatch_fire_event(dev);
3449 * netif_dormant_off - set device as not dormant.
3450 * @dev: network device
3452 * Device is not in dormant state.
3454 static inline void netif_dormant_off(struct net_device *dev)
3456 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3457 linkwatch_fire_event(dev);
3461 * netif_dormant - test if device is dormant
3462 * @dev: network device
3464 * Check if device is dormant.
3466 static inline bool netif_dormant(const struct net_device *dev)
3468 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3473 * netif_oper_up - test if device is operational
3474 * @dev: network device
3476 * Check if carrier is operational
3478 static inline bool netif_oper_up(const struct net_device *dev)
3480 return (dev->operstate == IF_OPER_UP ||
3481 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3485 * netif_device_present - is device available or removed
3486 * @dev: network device
3488 * Check if device has not been removed from system.
3490 static inline bool netif_device_present(struct net_device *dev)
3492 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3495 void netif_device_detach(struct net_device *dev);
3497 void netif_device_attach(struct net_device *dev);
3500 * Network interface message level settings
3504 NETIF_MSG_DRV = 0x0001,
3505 NETIF_MSG_PROBE = 0x0002,
3506 NETIF_MSG_LINK = 0x0004,
3507 NETIF_MSG_TIMER = 0x0008,
3508 NETIF_MSG_IFDOWN = 0x0010,
3509 NETIF_MSG_IFUP = 0x0020,
3510 NETIF_MSG_RX_ERR = 0x0040,
3511 NETIF_MSG_TX_ERR = 0x0080,
3512 NETIF_MSG_TX_QUEUED = 0x0100,
3513 NETIF_MSG_INTR = 0x0200,
3514 NETIF_MSG_TX_DONE = 0x0400,
3515 NETIF_MSG_RX_STATUS = 0x0800,
3516 NETIF_MSG_PKTDATA = 0x1000,
3517 NETIF_MSG_HW = 0x2000,
3518 NETIF_MSG_WOL = 0x4000,
3521 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3522 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3523 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3524 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3525 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3526 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3527 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3528 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3529 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3530 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3531 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3532 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3533 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3534 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3535 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3537 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3540 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3541 return default_msg_enable_bits;
3542 if (debug_value == 0) /* no output */
3544 /* set low N bits */
3545 return (1 << debug_value) - 1;
3548 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3550 spin_lock(&txq->_xmit_lock);
3551 txq->xmit_lock_owner = cpu;
3554 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3556 __acquire(&txq->_xmit_lock);
3560 static inline void __netif_tx_release(struct netdev_queue *txq)
3562 __release(&txq->_xmit_lock);
3565 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3567 spin_lock_bh(&txq->_xmit_lock);
3568 txq->xmit_lock_owner = smp_processor_id();
3571 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3573 bool ok = spin_trylock(&txq->_xmit_lock);
3575 txq->xmit_lock_owner = smp_processor_id();
3579 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3581 txq->xmit_lock_owner = -1;
3582 spin_unlock(&txq->_xmit_lock);
3585 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3587 txq->xmit_lock_owner = -1;
3588 spin_unlock_bh(&txq->_xmit_lock);
3591 static inline void txq_trans_update(struct netdev_queue *txq)
3593 if (txq->xmit_lock_owner != -1)
3594 txq->trans_start = jiffies;
3597 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3598 static inline void netif_trans_update(struct net_device *dev)
3600 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3602 if (txq->trans_start != jiffies)
3603 txq->trans_start = jiffies;
3607 * netif_tx_lock - grab network device transmit lock
3608 * @dev: network device
3610 * Get network device transmit lock
3612 static inline void netif_tx_lock(struct net_device *dev)
3617 spin_lock(&dev->tx_global_lock);
3618 cpu = smp_processor_id();
3619 for (i = 0; i < dev->num_tx_queues; i++) {
3620 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3622 /* We are the only thread of execution doing a
3623 * freeze, but we have to grab the _xmit_lock in
3624 * order to synchronize with threads which are in
3625 * the ->hard_start_xmit() handler and already
3626 * checked the frozen bit.
3628 __netif_tx_lock(txq, cpu);
3629 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3630 __netif_tx_unlock(txq);
3634 static inline void netif_tx_lock_bh(struct net_device *dev)
3640 static inline void netif_tx_unlock(struct net_device *dev)
3644 for (i = 0; i < dev->num_tx_queues; i++) {
3645 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3647 /* No need to grab the _xmit_lock here. If the
3648 * queue is not stopped for another reason, we
3651 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3652 netif_schedule_queue(txq);
3654 spin_unlock(&dev->tx_global_lock);
3657 static inline void netif_tx_unlock_bh(struct net_device *dev)
3659 netif_tx_unlock(dev);
3663 #define HARD_TX_LOCK(dev, txq, cpu) { \
3664 if ((dev->features & NETIF_F_LLTX) == 0) { \
3665 __netif_tx_lock(txq, cpu); \
3667 __netif_tx_acquire(txq); \
3671 #define HARD_TX_TRYLOCK(dev, txq) \
3672 (((dev->features & NETIF_F_LLTX) == 0) ? \
3673 __netif_tx_trylock(txq) : \
3674 __netif_tx_acquire(txq))
3676 #define HARD_TX_UNLOCK(dev, txq) { \
3677 if ((dev->features & NETIF_F_LLTX) == 0) { \
3678 __netif_tx_unlock(txq); \
3680 __netif_tx_release(txq); \
3684 static inline void netif_tx_disable(struct net_device *dev)
3690 cpu = smp_processor_id();
3691 for (i = 0; i < dev->num_tx_queues; i++) {
3692 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3694 __netif_tx_lock(txq, cpu);
3695 netif_tx_stop_queue(txq);
3696 __netif_tx_unlock(txq);
3701 static inline void netif_addr_lock(struct net_device *dev)
3703 spin_lock(&dev->addr_list_lock);
3706 static inline void netif_addr_lock_nested(struct net_device *dev)
3708 int subclass = SINGLE_DEPTH_NESTING;
3710 if (dev->netdev_ops->ndo_get_lock_subclass)
3711 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3713 spin_lock_nested(&dev->addr_list_lock, subclass);
3716 static inline void netif_addr_lock_bh(struct net_device *dev)
3718 spin_lock_bh(&dev->addr_list_lock);
3721 static inline void netif_addr_unlock(struct net_device *dev)
3723 spin_unlock(&dev->addr_list_lock);
3726 static inline void netif_addr_unlock_bh(struct net_device *dev)
3728 spin_unlock_bh(&dev->addr_list_lock);
3732 * dev_addrs walker. Should be used only for read access. Call with
3733 * rcu_read_lock held.
3735 #define for_each_dev_addr(dev, ha) \
3736 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3738 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3740 void ether_setup(struct net_device *dev);
3742 /* Support for loadable net-drivers */
3743 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3744 unsigned char name_assign_type,
3745 void (*setup)(struct net_device *),
3746 unsigned int txqs, unsigned int rxqs);
3747 int dev_get_valid_name(struct net *net, struct net_device *dev,
3750 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3751 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3753 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3754 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3757 int register_netdev(struct net_device *dev);
3758 void unregister_netdev(struct net_device *dev);
3760 /* General hardware address lists handling functions */
3761 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3762 struct netdev_hw_addr_list *from_list, int addr_len);
3763 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3764 struct netdev_hw_addr_list *from_list, int addr_len);
3765 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3766 struct net_device *dev,
3767 int (*sync)(struct net_device *, const unsigned char *),
3768 int (*unsync)(struct net_device *,
3769 const unsigned char *));
3770 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3771 struct net_device *dev,
3772 int (*unsync)(struct net_device *,
3773 const unsigned char *));
3774 void __hw_addr_init(struct netdev_hw_addr_list *list);
3776 /* Functions used for device addresses handling */
3777 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3778 unsigned char addr_type);
3779 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3780 unsigned char addr_type);
3781 void dev_addr_flush(struct net_device *dev);
3782 int dev_addr_init(struct net_device *dev);
3784 /* Functions used for unicast addresses handling */
3785 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3786 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3787 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3788 int dev_uc_sync(struct net_device *to, struct net_device *from);
3789 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3790 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3791 void dev_uc_flush(struct net_device *dev);
3792 void dev_uc_init(struct net_device *dev);
3795 * __dev_uc_sync - Synchonize device's unicast list
3796 * @dev: device to sync
3797 * @sync: function to call if address should be added
3798 * @unsync: function to call if address should be removed
3800 * Add newly added addresses to the interface, and release
3801 * addresses that have been deleted.
3803 static inline int __dev_uc_sync(struct net_device *dev,
3804 int (*sync)(struct net_device *,
3805 const unsigned char *),
3806 int (*unsync)(struct net_device *,
3807 const unsigned char *))
3809 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3813 * __dev_uc_unsync - Remove synchronized addresses from device
3814 * @dev: device to sync
3815 * @unsync: function to call if address should be removed
3817 * Remove all addresses that were added to the device by dev_uc_sync().
3819 static inline void __dev_uc_unsync(struct net_device *dev,
3820 int (*unsync)(struct net_device *,
3821 const unsigned char *))
3823 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3826 /* Functions used for multicast addresses handling */
3827 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3828 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3829 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3830 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3831 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3832 int dev_mc_sync(struct net_device *to, struct net_device *from);
3833 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3834 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3835 void dev_mc_flush(struct net_device *dev);
3836 void dev_mc_init(struct net_device *dev);
3839 * __dev_mc_sync - Synchonize device's multicast list
3840 * @dev: device to sync
3841 * @sync: function to call if address should be added
3842 * @unsync: function to call if address should be removed
3844 * Add newly added addresses to the interface, and release
3845 * addresses that have been deleted.
3847 static inline int __dev_mc_sync(struct net_device *dev,
3848 int (*sync)(struct net_device *,
3849 const unsigned char *),
3850 int (*unsync)(struct net_device *,
3851 const unsigned char *))
3853 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3857 * __dev_mc_unsync - Remove synchronized addresses from device
3858 * @dev: device to sync
3859 * @unsync: function to call if address should be removed
3861 * Remove all addresses that were added to the device by dev_mc_sync().
3863 static inline void __dev_mc_unsync(struct net_device *dev,
3864 int (*unsync)(struct net_device *,
3865 const unsigned char *))
3867 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3870 /* Functions used for secondary unicast and multicast support */
3871 void dev_set_rx_mode(struct net_device *dev);
3872 void __dev_set_rx_mode(struct net_device *dev);
3873 int dev_set_promiscuity(struct net_device *dev, int inc);
3874 int dev_set_allmulti(struct net_device *dev, int inc);
3875 void netdev_state_change(struct net_device *dev);
3876 void netdev_notify_peers(struct net_device *dev);
3877 void netdev_features_change(struct net_device *dev);
3878 /* Load a device via the kmod */
3879 void dev_load(struct net *net, const char *name);
3880 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3881 struct rtnl_link_stats64 *storage);
3882 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3883 const struct net_device_stats *netdev_stats);
3885 extern int netdev_max_backlog;
3886 extern int netdev_tstamp_prequeue;
3887 extern int weight_p;
3888 extern int dev_weight_rx_bias;
3889 extern int dev_weight_tx_bias;
3890 extern int dev_rx_weight;
3891 extern int dev_tx_weight;
3893 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3894 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3895 struct list_head **iter);
3896 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3897 struct list_head **iter);
3899 /* iterate through upper list, must be called under RCU read lock */
3900 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3901 for (iter = &(dev)->adj_list.upper, \
3902 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3904 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3906 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3907 int (*fn)(struct net_device *upper_dev,
3911 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3912 struct net_device *upper_dev);
3914 bool netdev_has_any_upper_dev(struct net_device *dev);
3916 void *netdev_lower_get_next_private(struct net_device *dev,
3917 struct list_head **iter);
3918 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3919 struct list_head **iter);
3921 #define netdev_for_each_lower_private(dev, priv, iter) \
3922 for (iter = (dev)->adj_list.lower.next, \
3923 priv = netdev_lower_get_next_private(dev, &(iter)); \
3925 priv = netdev_lower_get_next_private(dev, &(iter)))
3927 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3928 for (iter = &(dev)->adj_list.lower, \
3929 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3931 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3933 void *netdev_lower_get_next(struct net_device *dev,
3934 struct list_head **iter);
3936 #define netdev_for_each_lower_dev(dev, ldev, iter) \