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>
64 /* 802.15.4 specific */
67 /* UDP Tunnel offloads */
68 struct udp_tunnel_info;
72 void netdev_set_default_ethtool_ops(struct net_device *dev,
73 const struct ethtool_ops *ops);
75 /* Backlog congestion levels */
76 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
77 #define NET_RX_DROP 1 /* packet dropped */
80 * Transmit return codes: transmit return codes originate from three different
83 * - qdisc return codes
84 * - driver transmit return codes
87 * Drivers are allowed to return any one of those in their hard_start_xmit()
88 * function. Real network devices commonly used with qdiscs should only return
89 * the driver transmit return codes though - when qdiscs are used, the actual
90 * transmission happens asynchronously, so the value is not propagated to
91 * higher layers. Virtual network devices transmit synchronously; in this case
92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
93 * others are propagated to higher layers.
96 /* qdisc ->enqueue() return codes. */
97 #define NET_XMIT_SUCCESS 0x00
98 #define NET_XMIT_DROP 0x01 /* skb dropped */
99 #define NET_XMIT_CN 0x02 /* congestion notification */
100 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
102 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
103 * indicates that the device will soon be dropping packets, or already drops
104 * some packets of the same priority; prompting us to send less aggressively. */
105 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
106 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
108 /* Driver transmit return codes */
109 #define NETDEV_TX_MASK 0xf0
112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
113 NETDEV_TX_OK = 0x00, /* driver took care of packet */
114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
116 typedef enum netdev_tx netdev_tx_t;
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
122 static inline bool dev_xmit_complete(int rc)
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
130 if (likely(rc < NET_XMIT_MASK))
137 * Compute the worst-case header length according to the protocols
141 #if defined(CONFIG_HYPERV_NET)
142 # define LL_MAX_HEADER 128
143 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
144 # if defined(CONFIG_MAC80211_MESH)
145 # define LL_MAX_HEADER 128
147 # define LL_MAX_HEADER 96
150 # define LL_MAX_HEADER 32
153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
155 #define MAX_HEADER LL_MAX_HEADER
157 #define MAX_HEADER (LL_MAX_HEADER + 48)
161 * Old network device statistics. Fields are native words
162 * (unsigned long) so they can be read and written atomically.
165 struct net_device_stats {
166 unsigned long rx_packets;
167 unsigned long tx_packets;
168 unsigned long rx_bytes;
169 unsigned long tx_bytes;
170 unsigned long rx_errors;
171 unsigned long tx_errors;
172 unsigned long rx_dropped;
173 unsigned long tx_dropped;
174 unsigned long multicast;
175 unsigned long collisions;
176 unsigned long rx_length_errors;
177 unsigned long rx_over_errors;
178 unsigned long rx_crc_errors;
179 unsigned long rx_frame_errors;
180 unsigned long rx_fifo_errors;
181 unsigned long rx_missed_errors;
182 unsigned long tx_aborted_errors;
183 unsigned long tx_carrier_errors;
184 unsigned long tx_fifo_errors;
185 unsigned long tx_heartbeat_errors;
186 unsigned long tx_window_errors;
187 unsigned long rx_compressed;
188 unsigned long tx_compressed;
192 #include <linux/cache.h>
193 #include <linux/skbuff.h>
196 #include <linux/static_key.h>
197 extern struct static_key rps_needed;
198 extern struct static_key rfs_needed;
205 struct netdev_hw_addr {
206 struct list_head list;
207 unsigned char addr[MAX_ADDR_LEN];
209 #define NETDEV_HW_ADDR_T_LAN 1
210 #define NETDEV_HW_ADDR_T_SAN 2
211 #define NETDEV_HW_ADDR_T_SLAVE 3
212 #define NETDEV_HW_ADDR_T_UNICAST 4
213 #define NETDEV_HW_ADDR_T_MULTICAST 5
218 struct rcu_head rcu_head;
221 struct netdev_hw_addr_list {
222 struct list_head list;
226 #define netdev_hw_addr_list_count(l) ((l)->count)
227 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
228 #define netdev_hw_addr_list_for_each(ha, l) \
229 list_for_each_entry(ha, &(l)->list, list)
231 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
232 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
233 #define netdev_for_each_uc_addr(ha, dev) \
234 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
236 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
237 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
238 #define netdev_for_each_mc_addr(ha, dev) \
239 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
245 /* cached hardware header; allow for machine alignment needs. */
246 #define HH_DATA_MOD 16
247 #define HH_DATA_OFF(__len) \
248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
249 #define HH_DATA_ALIGN(__len) \
250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
254 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
256 * dev->hard_header_len ? (dev->hard_header_len +
257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
259 * We could use other alignment values, but we must maintain the
260 * relationship HH alignment <= LL alignment.
262 #define LL_RESERVED_SPACE(dev) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
268 int (*create) (struct sk_buff *skb, struct net_device *dev,
269 unsigned short type, const void *daddr,
270 const void *saddr, unsigned int len);
271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
273 void (*cache_update)(struct hh_cache *hh,
274 const struct net_device *dev,
275 const unsigned char *haddr);
276 bool (*validate)(const char *ll_header, unsigned int len);
279 /* These flag bits are private to the generic network queueing
280 * layer; they may not be explicitly referenced by any other
284 enum netdev_state_t {
286 __LINK_STATE_PRESENT,
287 __LINK_STATE_NOCARRIER,
288 __LINK_STATE_LINKWATCH_PENDING,
289 __LINK_STATE_DORMANT,
294 * This structure holds boot-time configured netdevice settings. They
295 * are then used in the device probing.
297 struct netdev_boot_setup {
301 #define NETDEV_BOOT_SETUP_MAX 8
303 int __init netdev_boot_setup(char *str);
306 struct list_head list;
311 * size of gro hash buckets, must less than bit number of
312 * napi_struct::gro_bitmask
314 #define GRO_HASH_BUCKETS 8
317 * Structure for NAPI scheduling similar to tasklet but with weighting
320 /* The poll_list must only be managed by the entity which
321 * changes the state of the NAPI_STATE_SCHED bit. This means
322 * whoever atomically sets that bit can add this napi_struct
323 * to the per-CPU poll_list, and whoever clears that bit
324 * can remove from the list right before clearing the bit.
326 struct list_head poll_list;
330 unsigned long gro_bitmask;
331 int (*poll)(struct napi_struct *, int);
332 #ifdef CONFIG_NETPOLL
335 struct net_device *dev;
336 struct gro_list gro_hash[GRO_HASH_BUCKETS];
338 struct hrtimer timer;
339 struct list_head dev_list;
340 struct hlist_node napi_hash_node;
341 unsigned int napi_id;
345 NAPI_STATE_SCHED, /* Poll is scheduled */
346 NAPI_STATE_MISSED, /* reschedule a napi */
347 NAPI_STATE_DISABLE, /* Disable pending */
348 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
349 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
350 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
351 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
355 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
356 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
357 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
358 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
359 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
360 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
361 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
372 typedef enum gro_result gro_result_t;
375 * enum rx_handler_result - Possible return values for rx_handlers.
376 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
378 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
379 * case skb->dev was changed by rx_handler.
380 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
381 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
383 * rx_handlers are functions called from inside __netif_receive_skb(), to do
384 * special processing of the skb, prior to delivery to protocol handlers.
386 * Currently, a net_device can only have a single rx_handler registered. Trying
387 * to register a second rx_handler will return -EBUSY.
389 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
390 * To unregister a rx_handler on a net_device, use
391 * netdev_rx_handler_unregister().
393 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
396 * If the rx_handler consumed the skb in some way, it should return
397 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
398 * the skb to be delivered in some other way.
400 * If the rx_handler changed skb->dev, to divert the skb to another
401 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
402 * new device will be called if it exists.
404 * If the rx_handler decides the skb should be ignored, it should return
405 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
406 * are registered on exact device (ptype->dev == skb->dev).
408 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
409 * delivered, it should return RX_HANDLER_PASS.
411 * A device without a registered rx_handler will behave as if rx_handler
412 * returned RX_HANDLER_PASS.
415 enum rx_handler_result {
421 typedef enum rx_handler_result rx_handler_result_t;
422 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
424 void __napi_schedule(struct napi_struct *n);
425 void __napi_schedule_irqoff(struct napi_struct *n);
427 static inline bool napi_disable_pending(struct napi_struct *n)
429 return test_bit(NAPI_STATE_DISABLE, &n->state);
432 bool napi_schedule_prep(struct napi_struct *n);
435 * napi_schedule - schedule NAPI poll
438 * Schedule NAPI poll routine to be called if it is not already
441 static inline void napi_schedule(struct napi_struct *n)
443 if (napi_schedule_prep(n))
448 * napi_schedule_irqoff - schedule NAPI poll
451 * Variant of napi_schedule(), assuming hard irqs are masked.
453 static inline void napi_schedule_irqoff(struct napi_struct *n)
455 if (napi_schedule_prep(n))
456 __napi_schedule_irqoff(n);
459 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
460 static inline bool napi_reschedule(struct napi_struct *napi)
462 if (napi_schedule_prep(napi)) {
463 __napi_schedule(napi);
469 bool napi_complete_done(struct napi_struct *n, int work_done);
471 * napi_complete - NAPI processing complete
474 * Mark NAPI processing as complete.
475 * Consider using napi_complete_done() instead.
476 * Return false if device should avoid rearming interrupts.
478 static inline bool napi_complete(struct napi_struct *n)
480 return napi_complete_done(n, 0);
484 * napi_hash_del - remove a NAPI from global table
485 * @napi: NAPI context
487 * Warning: caller must observe RCU grace period
488 * before freeing memory containing @napi, if
489 * this function returns true.
490 * Note: core networking stack automatically calls it
491 * from netif_napi_del().
492 * Drivers might want to call this helper to combine all
493 * the needed RCU grace periods into a single one.
495 bool napi_hash_del(struct napi_struct *napi);
498 * napi_disable - prevent NAPI from scheduling
501 * Stop NAPI from being scheduled on this context.
502 * Waits till any outstanding processing completes.
504 void napi_disable(struct napi_struct *n);
507 * napi_enable - enable NAPI scheduling
510 * Resume NAPI from being scheduled on this context.
511 * Must be paired with napi_disable.
513 static inline void napi_enable(struct napi_struct *n)
515 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
516 smp_mb__before_atomic();
517 clear_bit(NAPI_STATE_SCHED, &n->state);
518 clear_bit(NAPI_STATE_NPSVC, &n->state);
522 * napi_synchronize - wait until NAPI is not running
525 * Wait until NAPI is done being scheduled on this context.
526 * Waits till any outstanding processing completes but
527 * does not disable future activations.
529 static inline void napi_synchronize(const struct napi_struct *n)
531 if (IS_ENABLED(CONFIG_SMP))
532 while (test_bit(NAPI_STATE_SCHED, &n->state))
539 * napi_if_scheduled_mark_missed - if napi is running, set the
543 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
546 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
548 unsigned long val, new;
551 val = READ_ONCE(n->state);
552 if (val & NAPIF_STATE_DISABLE)
555 if (!(val & NAPIF_STATE_SCHED))
558 new = val | NAPIF_STATE_MISSED;
559 } while (cmpxchg(&n->state, val, new) != val);
564 enum netdev_queue_state_t {
565 __QUEUE_STATE_DRV_XOFF,
566 __QUEUE_STATE_STACK_XOFF,
567 __QUEUE_STATE_FROZEN,
570 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
571 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
572 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
574 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
575 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
577 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
581 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
582 * netif_tx_* functions below are used to manipulate this flag. The
583 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
584 * queue independently. The netif_xmit_*stopped functions below are called
585 * to check if the queue has been stopped by the driver or stack (either
586 * of the XOFF bits are set in the state). Drivers should not need to call
587 * netif_xmit*stopped functions, they should only be using netif_tx_*.
590 struct netdev_queue {
594 struct net_device *dev;
595 struct Qdisc __rcu *qdisc;
596 struct Qdisc *qdisc_sleeping;
600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
603 unsigned long tx_maxrate;
605 * Number of TX timeouts for this queue
606 * (/sys/class/net/DEV/Q/trans_timeout)
608 unsigned long trans_timeout;
610 /* Subordinate device that the queue has been assigned to */
611 struct net_device *sb_dev;
612 #ifdef CONFIG_XDP_SOCKETS
613 struct xdp_umem *umem;
618 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
621 * Time (in jiffies) of last Tx
623 unsigned long trans_start;
630 } ____cacheline_aligned_in_smp;
632 extern int sysctl_fb_tunnels_only_for_init_net;
634 static inline bool net_has_fallback_tunnels(const struct net *net)
636 return net == &init_net ||
637 !IS_ENABLED(CONFIG_SYSCTL) ||
638 !sysctl_fb_tunnels_only_for_init_net;
641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
659 * This structure holds an RPS map which can be of variable length. The
660 * map is an array of CPUs.
667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
671 * tail pointer for that CPU's input queue at the time of last enqueue, and
672 * a hardware filter index.
674 struct rps_dev_flow {
677 unsigned int last_qtail;
679 #define RPS_NO_FILTER 0xffff
682 * The rps_dev_flow_table structure contains a table of flow mappings.
684 struct rps_dev_flow_table {
687 struct rps_dev_flow flows[0];
689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
690 ((_num) * sizeof(struct rps_dev_flow)))
693 * The rps_sock_flow_table contains mappings of flows to the last CPU
694 * on which they were processed by the application (set in recvmsg).
695 * Each entry is a 32bit value. Upper part is the high-order bits
696 * of flow hash, lower part is CPU number.
697 * rps_cpu_mask is used to partition the space, depending on number of
698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
700 * meaning we use 32-6=26 bits for the hash.
702 struct rps_sock_flow_table {
705 u32 ents[0] ____cacheline_aligned_in_smp;
707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
709 #define RPS_NO_CPU 0xffff
711 extern u32 rps_cpu_mask;
712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
718 unsigned int index = hash & table->mask;
719 u32 val = hash & ~rps_cpu_mask;
721 /* We only give a hint, preemption can change CPU under us */
722 val |= raw_smp_processor_id();
724 if (table->ents[index] != val)
725 table->ents[index] = val;
729 #ifdef CONFIG_RFS_ACCEL
730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
733 #endif /* CONFIG_RPS */
735 /* This structure contains an instance of an RX queue. */
736 struct netdev_rx_queue {
738 struct rps_map __rcu *rps_map;
739 struct rps_dev_flow_table __rcu *rps_flow_table;
742 struct net_device *dev;
743 struct xdp_rxq_info xdp_rxq;
744 #ifdef CONFIG_XDP_SOCKETS
745 struct xdp_umem *umem;
747 } ____cacheline_aligned_in_smp;
750 * RX queue sysfs structures and functions.
752 struct rx_queue_attribute {
753 struct attribute attr;
754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
755 ssize_t (*store)(struct netdev_rx_queue *queue,
756 const char *buf, size_t len);
761 * This structure holds an XPS map which can be of variable length. The
762 * map is an array of queues.
766 unsigned int alloc_len;
770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
772 - sizeof(struct xps_map)) / sizeof(u16))
775 * This structure holds all XPS maps for device. Maps are indexed by CPU.
777 struct xps_dev_maps {
779 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
782 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
785 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
786 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
788 #endif /* CONFIG_XPS */
790 #define TC_MAX_QUEUE 16
791 #define TC_BITMASK 15
792 /* HW offloaded queuing disciplines txq count and offset maps */
793 struct netdev_tc_txq {
798 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
800 * This structure is to hold information about the device
801 * configured to run FCoE protocol stack.
803 struct netdev_fcoe_hbainfo {
804 char manufacturer[64];
805 char serial_number[64];
806 char hardware_version[64];
807 char driver_version[64];
808 char optionrom_version[64];
809 char firmware_version[64];
811 char model_description[256];
815 #define MAX_PHYS_ITEM_ID_LEN 32
817 /* This structure holds a unique identifier to identify some
818 * physical item (port for example) used by a netdevice.
820 struct netdev_phys_item_id {
821 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
822 unsigned char id_len;
825 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
826 struct netdev_phys_item_id *b)
828 return a->id_len == b->id_len &&
829 memcmp(a->id, b->id, a->id_len) == 0;
832 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
834 struct net_device *sb_dev);
837 TC_SETUP_QDISC_MQPRIO,
840 TC_SETUP_CLSMATCHALL,
852 /* These structures hold the attributes of bpf state that are being passed
853 * to the netdevice through the bpf op.
855 enum bpf_netdev_command {
856 /* Set or clear a bpf program used in the earliest stages of packet
857 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
858 * is responsible for calling bpf_prog_put on any old progs that are
859 * stored. In case of error, the callee need not release the new prog
860 * reference, but on success it takes ownership and must bpf_prog_put
861 * when it is no longer used.
867 /* BPF program for offload callbacks, invoked at program load time. */
868 BPF_OFFLOAD_MAP_ALLOC,
869 BPF_OFFLOAD_MAP_FREE,
874 struct bpf_prog_offload_ops;
875 struct netlink_ext_ack;
879 enum bpf_netdev_command command;
884 struct bpf_prog *prog;
885 struct netlink_ext_ack *extack;
887 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
890 /* flags with which program was installed */
893 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
895 struct bpf_offloaded_map *offmap;
897 /* XDP_QUERY_XSK_UMEM, XDP_SETUP_XSK_UMEM */
899 struct xdp_umem *umem; /* out for query*/
900 u16 queue_id; /* in for query */
905 #ifdef CONFIG_XFRM_OFFLOAD
907 int (*xdo_dev_state_add) (struct xfrm_state *x);
908 void (*xdo_dev_state_delete) (struct xfrm_state *x);
909 void (*xdo_dev_state_free) (struct xfrm_state *x);
910 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
911 struct xfrm_state *x);
912 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
916 #if IS_ENABLED(CONFIG_TLS_DEVICE)
917 enum tls_offload_ctx_dir {
918 TLS_OFFLOAD_CTX_DIR_RX,
919 TLS_OFFLOAD_CTX_DIR_TX,
922 struct tls_crypto_info;
926 int (*tls_dev_add)(struct net_device *netdev, struct sock *sk,
927 enum tls_offload_ctx_dir direction,
928 struct tls_crypto_info *crypto_info,
929 u32 start_offload_tcp_sn);
930 void (*tls_dev_del)(struct net_device *netdev,
931 struct tls_context *ctx,
932 enum tls_offload_ctx_dir direction);
933 void (*tls_dev_resync_rx)(struct net_device *netdev,
934 struct sock *sk, u32 seq, u64 rcd_sn);
939 struct rcu_head rcuhead;
944 * This structure defines the management hooks for network devices.
945 * The following hooks can be defined; unless noted otherwise, they are
946 * optional and can be filled with a null pointer.
948 * int (*ndo_init)(struct net_device *dev);
949 * This function is called once when a network device is registered.
950 * The network device can use this for any late stage initialization
951 * or semantic validation. It can fail with an error code which will
952 * be propagated back to register_netdev.
954 * void (*ndo_uninit)(struct net_device *dev);
955 * This function is called when device is unregistered or when registration
956 * fails. It is not called if init fails.
958 * int (*ndo_open)(struct net_device *dev);
959 * This function is called when a network device transitions to the up
962 * int (*ndo_stop)(struct net_device *dev);
963 * This function is called when a network device transitions to the down
966 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
967 * struct net_device *dev);
968 * Called when a packet needs to be transmitted.
969 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
970 * the queue before that can happen; it's for obsolete devices and weird
971 * corner cases, but the stack really does a non-trivial amount
972 * of useless work if you return NETDEV_TX_BUSY.
973 * Required; cannot be NULL.
975 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
976 * struct net_device *dev
977 * netdev_features_t features);
978 * Called by core transmit path to determine if device is capable of
979 * performing offload operations on a given packet. This is to give
980 * the device an opportunity to implement any restrictions that cannot
981 * be otherwise expressed by feature flags. The check is called with
982 * the set of features that the stack has calculated and it returns
983 * those the driver believes to be appropriate.
985 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
986 * struct net_device *sb_dev,
987 * select_queue_fallback_t fallback);
988 * Called to decide which queue to use when device supports multiple
991 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
992 * This function is called to allow device receiver to make
993 * changes to configuration when multicast or promiscuous is enabled.
995 * void (*ndo_set_rx_mode)(struct net_device *dev);
996 * This function is called device changes address list filtering.
997 * If driver handles unicast address filtering, it should set
998 * IFF_UNICAST_FLT in its priv_flags.
1000 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1001 * This function is called when the Media Access Control address
1002 * needs to be changed. If this interface is not defined, the
1003 * MAC address can not be changed.
1005 * int (*ndo_validate_addr)(struct net_device *dev);
1006 * Test if Media Access Control address is valid for the device.
1008 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1009 * Called when a user requests an ioctl which can't be handled by
1010 * the generic interface code. If not defined ioctls return
1011 * not supported error code.
1013 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1014 * Used to set network devices bus interface parameters. This interface
1015 * is retained for legacy reasons; new devices should use the bus
1016 * interface (PCI) for low level management.
1018 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1019 * Called when a user wants to change the Maximum Transfer Unit
1022 * void (*ndo_tx_timeout)(struct net_device *dev);
1023 * Callback used when the transmitter has not made any progress
1024 * for dev->watchdog ticks.
1026 * void (*ndo_get_stats64)(struct net_device *dev,
1027 * struct rtnl_link_stats64 *storage);
1028 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1029 * Called when a user wants to get the network device usage
1030 * statistics. Drivers must do one of the following:
1031 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1032 * rtnl_link_stats64 structure passed by the caller.
1033 * 2. Define @ndo_get_stats to update a net_device_stats structure
1034 * (which should normally be dev->stats) and return a pointer to
1035 * it. The structure may be changed asynchronously only if each
1036 * field is written atomically.
1037 * 3. Update dev->stats asynchronously and atomically, and define
1038 * neither operation.
1040 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1041 * Return true if this device supports offload stats of this attr_id.
1043 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1045 * Get statistics for offload operations by attr_id. Write it into the
1046 * attr_data pointer.
1048 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1049 * If device supports VLAN filtering this function is called when a
1050 * VLAN id is registered.
1052 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1053 * If device supports VLAN filtering this function is called when a
1054 * VLAN id is unregistered.
1056 * void (*ndo_poll_controller)(struct net_device *dev);
1058 * SR-IOV management functions.
1059 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1060 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1061 * u8 qos, __be16 proto);
1062 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1064 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1065 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1066 * int (*ndo_get_vf_config)(struct net_device *dev,
1067 * int vf, struct ifla_vf_info *ivf);
1068 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1069 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1070 * struct nlattr *port[]);
1072 * Enable or disable the VF ability to query its RSS Redirection Table and
1073 * Hash Key. This is needed since on some devices VF share this information
1074 * with PF and querying it may introduce a theoretical security risk.
1075 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1076 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1077 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1079 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1080 * This is always called from the stack with the rtnl lock held and netif
1081 * tx queues stopped. This allows the netdevice to perform queue
1082 * management safely.
1084 * Fiber Channel over Ethernet (FCoE) offload functions.
1085 * int (*ndo_fcoe_enable)(struct net_device *dev);
1086 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1087 * so the underlying device can perform whatever needed configuration or
1088 * initialization to support acceleration of FCoE traffic.
1090 * int (*ndo_fcoe_disable)(struct net_device *dev);
1091 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1092 * so the underlying device can perform whatever needed clean-ups to
1093 * stop supporting acceleration of FCoE traffic.
1095 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1096 * struct scatterlist *sgl, unsigned int sgc);
1097 * Called when the FCoE Initiator wants to initialize an I/O that
1098 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1099 * perform necessary setup and returns 1 to indicate the device is set up
1100 * successfully to perform DDP on this I/O, otherwise this returns 0.
1102 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1103 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1104 * indicated by the FC exchange id 'xid', so the underlying device can
1105 * clean up and reuse resources for later DDP requests.
1107 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1108 * struct scatterlist *sgl, unsigned int sgc);
1109 * Called when the FCoE Target wants to initialize an I/O that
1110 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1111 * perform necessary setup and returns 1 to indicate the device is set up
1112 * successfully to perform DDP on this I/O, otherwise this returns 0.
1114 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1115 * struct netdev_fcoe_hbainfo *hbainfo);
1116 * Called when the FCoE Protocol stack wants information on the underlying
1117 * device. This information is utilized by the FCoE protocol stack to
1118 * register attributes with Fiber Channel management service as per the
1119 * FC-GS Fabric Device Management Information(FDMI) specification.
1121 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1122 * Called when the underlying device wants to override default World Wide
1123 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1124 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1125 * protocol stack to use.
1128 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1129 * u16 rxq_index, u32 flow_id);
1130 * Set hardware filter for RFS. rxq_index is the target queue index;
1131 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1132 * Return the filter ID on success, or a negative error code.
1134 * Slave management functions (for bridge, bonding, etc).
1135 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1136 * Called to make another netdev an underling.
1138 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1139 * Called to release previously enslaved netdev.
1141 * Feature/offload setting functions.
1142 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1143 * netdev_features_t features);
1144 * Adjusts the requested feature flags according to device-specific
1145 * constraints, and returns the resulting flags. Must not modify
1148 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1149 * Called to update device configuration to new features. Passed
1150 * feature set might be less than what was returned by ndo_fix_features()).
1151 * Must return >0 or -errno if it changed dev->features itself.
1153 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1154 * struct net_device *dev,
1155 * const unsigned char *addr, u16 vid, u16 flags,
1156 * struct netlink_ext_ack *extack);
1157 * Adds an FDB entry to dev for addr.
1158 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1159 * struct net_device *dev,
1160 * const unsigned char *addr, u16 vid)
1161 * Deletes the FDB entry from dev coresponding to addr.
1162 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1163 * struct net_device *dev, struct net_device *filter_dev,
1165 * Used to add FDB entries to dump requests. Implementers should add
1166 * entries to skb and update idx with the number of entries.
1168 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1169 * u16 flags, struct netlink_ext_ack *extack)
1170 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1171 * struct net_device *dev, u32 filter_mask,
1173 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1176 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1177 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1178 * which do not represent real hardware may define this to allow their
1179 * userspace components to manage their virtual carrier state. Devices
1180 * that determine carrier state from physical hardware properties (eg
1181 * network cables) or protocol-dependent mechanisms (eg
1182 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1184 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1185 * struct netdev_phys_item_id *ppid);
1186 * Called to get ID of physical port of this device. If driver does
1187 * not implement this, it is assumed that the hw is not able to have
1188 * multiple net devices on single physical port.
1190 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1191 * struct udp_tunnel_info *ti);
1192 * Called by UDP tunnel to notify a driver about the UDP port and socket
1193 * address family that a UDP tunnel is listnening to. It is called only
1194 * when a new port starts listening. The operation is protected by the
1197 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1198 * struct udp_tunnel_info *ti);
1199 * Called by UDP tunnel to notify the driver about a UDP port and socket
1200 * address family that the UDP tunnel is not listening to anymore. The
1201 * operation is protected by the RTNL.
1203 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1204 * struct net_device *dev)
1205 * Called by upper layer devices to accelerate switching or other
1206 * station functionality into hardware. 'pdev is the lowerdev
1207 * to use for the offload and 'dev' is the net device that will
1208 * back the offload. Returns a pointer to the private structure
1209 * the upper layer will maintain.
1210 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1211 * Called by upper layer device to delete the station created
1212 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1213 * the station and priv is the structure returned by the add
1215 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1216 * int queue_index, u32 maxrate);
1217 * Called when a user wants to set a max-rate limitation of specific
1219 * int (*ndo_get_iflink)(const struct net_device *dev);
1220 * Called to get the iflink value of this device.
1221 * void (*ndo_change_proto_down)(struct net_device *dev,
1223 * This function is used to pass protocol port error state information
1224 * to the switch driver. The switch driver can react to the proto_down
1225 * by doing a phys down on the associated switch port.
1226 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1227 * This function is used to get egress tunnel information for given skb.
1228 * This is useful for retrieving outer tunnel header parameters while
1230 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1231 * This function is used to specify the headroom that the skb must
1232 * consider when allocation skb during packet reception. Setting
1233 * appropriate rx headroom value allows avoiding skb head copy on
1234 * forward. Setting a negative value resets the rx headroom to the
1236 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1237 * This function is used to set or query state related to XDP on the
1238 * netdevice and manage BPF offload. See definition of
1239 * enum bpf_netdev_command for details.
1240 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1242 * This function is used to submit @n XDP packets for transmit on a
1243 * netdevice. Returns number of frames successfully transmitted, frames
1244 * that got dropped are freed/returned via xdp_return_frame().
1245 * Returns negative number, means general error invoking ndo, meaning
1246 * no frames were xmit'ed and core-caller will free all frames.
1248 struct net_device_ops {
1249 int (*ndo_init)(struct net_device *dev);
1250 void (*ndo_uninit)(struct net_device *dev);
1251 int (*ndo_open)(struct net_device *dev);
1252 int (*ndo_stop)(struct net_device *dev);
1253 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1254 struct net_device *dev);
1255 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1256 struct net_device *dev,
1257 netdev_features_t features);
1258 u16 (*ndo_select_queue)(struct net_device *dev,
1259 struct sk_buff *skb,
1260 struct net_device *sb_dev,
1261 select_queue_fallback_t fallback);
1262 void (*ndo_change_rx_flags)(struct net_device *dev,
1264 void (*ndo_set_rx_mode)(struct net_device *dev);
1265 int (*ndo_set_mac_address)(struct net_device *dev,
1267 int (*ndo_validate_addr)(struct net_device *dev);
1268 int (*ndo_do_ioctl)(struct net_device *dev,
1269 struct ifreq *ifr, int cmd);
1270 int (*ndo_set_config)(struct net_device *dev,
1272 int (*ndo_change_mtu)(struct net_device *dev,
1274 int (*ndo_neigh_setup)(struct net_device *dev,
1275 struct neigh_parms *);
1276 void (*ndo_tx_timeout) (struct net_device *dev);
1278 void (*ndo_get_stats64)(struct net_device *dev,
1279 struct rtnl_link_stats64 *storage);
1280 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1281 int (*ndo_get_offload_stats)(int attr_id,
1282 const struct net_device *dev,
1284 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1286 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1287 __be16 proto, u16 vid);
1288 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1289 __be16 proto, u16 vid);
1290 #ifdef CONFIG_NET_POLL_CONTROLLER
1291 void (*ndo_poll_controller)(struct net_device *dev);
1292 int (*ndo_netpoll_setup)(struct net_device *dev,
1293 struct netpoll_info *info);
1294 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1296 int (*ndo_set_vf_mac)(struct net_device *dev,
1297 int queue, u8 *mac);
1298 int (*ndo_set_vf_vlan)(struct net_device *dev,
1299 int queue, u16 vlan,
1300 u8 qos, __be16 proto);
1301 int (*ndo_set_vf_rate)(struct net_device *dev,
1302 int vf, int min_tx_rate,
1304 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1305 int vf, bool setting);
1306 int (*ndo_set_vf_trust)(struct net_device *dev,
1307 int vf, bool setting);
1308 int (*ndo_get_vf_config)(struct net_device *dev,
1310 struct ifla_vf_info *ivf);
1311 int (*ndo_set_vf_link_state)(struct net_device *dev,
1312 int vf, int link_state);
1313 int (*ndo_get_vf_stats)(struct net_device *dev,
1315 struct ifla_vf_stats
1317 int (*ndo_set_vf_port)(struct net_device *dev,
1319 struct nlattr *port[]);
1320 int (*ndo_get_vf_port)(struct net_device *dev,
1321 int vf, struct sk_buff *skb);
1322 int (*ndo_set_vf_guid)(struct net_device *dev,
1325 int (*ndo_set_vf_rss_query_en)(
1326 struct net_device *dev,
1327 int vf, bool setting);
1328 int (*ndo_setup_tc)(struct net_device *dev,
1329 enum tc_setup_type type,
1331 #if IS_ENABLED(CONFIG_FCOE)
1332 int (*ndo_fcoe_enable)(struct net_device *dev);
1333 int (*ndo_fcoe_disable)(struct net_device *dev);
1334 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1336 struct scatterlist *sgl,
1338 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1340 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1342 struct scatterlist *sgl,
1344 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1345 struct netdev_fcoe_hbainfo *hbainfo);
1348 #if IS_ENABLED(CONFIG_LIBFCOE)
1349 #define NETDEV_FCOE_WWNN 0
1350 #define NETDEV_FCOE_WWPN 1
1351 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1352 u64 *wwn, int type);
1355 #ifdef CONFIG_RFS_ACCEL
1356 int (*ndo_rx_flow_steer)(struct net_device *dev,
1357 const struct sk_buff *skb,
1361 int (*ndo_add_slave)(struct net_device *dev,
1362 struct net_device *slave_dev,
1363 struct netlink_ext_ack *extack);
1364 int (*ndo_del_slave)(struct net_device *dev,
1365 struct net_device *slave_dev);
1366 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1367 netdev_features_t features);
1368 int (*ndo_set_features)(struct net_device *dev,
1369 netdev_features_t features);
1370 int (*ndo_neigh_construct)(struct net_device *dev,
1371 struct neighbour *n);
1372 void (*ndo_neigh_destroy)(struct net_device *dev,
1373 struct neighbour *n);
1375 int (*ndo_fdb_add)(struct ndmsg *ndm,
1376 struct nlattr *tb[],
1377 struct net_device *dev,
1378 const unsigned char *addr,
1381 struct netlink_ext_ack *extack);
1382 int (*ndo_fdb_del)(struct ndmsg *ndm,
1383 struct nlattr *tb[],
1384 struct net_device *dev,
1385 const unsigned char *addr,
1387 int (*ndo_fdb_dump)(struct sk_buff *skb,
1388 struct netlink_callback *cb,
1389 struct net_device *dev,
1390 struct net_device *filter_dev,
1392 int (*ndo_fdb_get)(struct sk_buff *skb,
1393 struct nlattr *tb[],
1394 struct net_device *dev,
1395 const unsigned char *addr,
1396 u16 vid, u32 portid, u32 seq,
1397 struct netlink_ext_ack *extack);
1398 int (*ndo_bridge_setlink)(struct net_device *dev,
1399 struct nlmsghdr *nlh,
1401 struct netlink_ext_ack *extack);
1402 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1404 struct net_device *dev,
1407 int (*ndo_bridge_dellink)(struct net_device *dev,
1408 struct nlmsghdr *nlh,
1410 int (*ndo_change_carrier)(struct net_device *dev,
1412 int (*ndo_get_phys_port_id)(struct net_device *dev,
1413 struct netdev_phys_item_id *ppid);
1414 int (*ndo_get_phys_port_name)(struct net_device *dev,
1415 char *name, size_t len);
1416 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1417 struct udp_tunnel_info *ti);
1418 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1419 struct udp_tunnel_info *ti);
1420 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1421 struct net_device *dev);
1422 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1425 int (*ndo_get_lock_subclass)(struct net_device *dev);
1426 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1429 int (*ndo_get_iflink)(const struct net_device *dev);
1430 int (*ndo_change_proto_down)(struct net_device *dev,
1432 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1433 struct sk_buff *skb);
1434 void (*ndo_set_rx_headroom)(struct net_device *dev,
1435 int needed_headroom);
1436 int (*ndo_bpf)(struct net_device *dev,
1437 struct netdev_bpf *bpf);
1438 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1439 struct xdp_frame **xdp,
1441 int (*ndo_xsk_async_xmit)(struct net_device *dev,
1446 * enum net_device_priv_flags - &struct net_device priv_flags
1448 * These are the &struct net_device, they are only set internally
1449 * by drivers and used in the kernel. These flags are invisible to
1450 * userspace; this means that the order of these flags can change
1451 * during any kernel release.
1453 * You should have a pretty good reason to be extending these flags.
1455 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1456 * @IFF_EBRIDGE: Ethernet bridging device
1457 * @IFF_BONDING: bonding master or slave
1458 * @IFF_ISATAP: ISATAP interface (RFC4214)
1459 * @IFF_WAN_HDLC: WAN HDLC device
1460 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1462 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1463 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1464 * @IFF_MACVLAN_PORT: device used as macvlan port
1465 * @IFF_BRIDGE_PORT: device used as bridge port
1466 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1467 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1468 * @IFF_UNICAST_FLT: Supports unicast filtering
1469 * @IFF_TEAM_PORT: device used as team port
1470 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1471 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1472 * change when it's running
1473 * @IFF_MACVLAN: Macvlan device
1474 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1475 * underlying stacked devices
1476 * @IFF_L3MDEV_MASTER: device is an L3 master device
1477 * @IFF_NO_QUEUE: device can run without qdisc attached
1478 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1479 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1480 * @IFF_TEAM: device is a team device
1481 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1482 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1483 * entity (i.e. the master device for bridged veth)
1484 * @IFF_MACSEC: device is a MACsec device
1485 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1486 * @IFF_FAILOVER: device is a failover master device
1487 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1489 enum netdev_priv_flags {
1490 IFF_802_1Q_VLAN = 1<<0,
1494 IFF_WAN_HDLC = 1<<4,
1495 IFF_XMIT_DST_RELEASE = 1<<5,
1496 IFF_DONT_BRIDGE = 1<<6,
1497 IFF_DISABLE_NETPOLL = 1<<7,
1498 IFF_MACVLAN_PORT = 1<<8,
1499 IFF_BRIDGE_PORT = 1<<9,
1500 IFF_OVS_DATAPATH = 1<<10,
1501 IFF_TX_SKB_SHARING = 1<<11,
1502 IFF_UNICAST_FLT = 1<<12,
1503 IFF_TEAM_PORT = 1<<13,
1504 IFF_SUPP_NOFCS = 1<<14,
1505 IFF_LIVE_ADDR_CHANGE = 1<<15,
1506 IFF_MACVLAN = 1<<16,
1507 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1508 IFF_L3MDEV_MASTER = 1<<18,
1509 IFF_NO_QUEUE = 1<<19,
1510 IFF_OPENVSWITCH = 1<<20,
1511 IFF_L3MDEV_SLAVE = 1<<21,
1513 IFF_RXFH_CONFIGURED = 1<<23,
1514 IFF_PHONY_HEADROOM = 1<<24,
1516 IFF_NO_RX_HANDLER = 1<<26,
1517 IFF_FAILOVER = 1<<27,
1518 IFF_FAILOVER_SLAVE = 1<<28,
1521 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1522 #define IFF_EBRIDGE IFF_EBRIDGE
1523 #define IFF_BONDING IFF_BONDING
1524 #define IFF_ISATAP IFF_ISATAP
1525 #define IFF_WAN_HDLC IFF_WAN_HDLC
1526 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1527 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1528 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1529 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1530 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1531 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1532 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1533 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1534 #define IFF_TEAM_PORT IFF_TEAM_PORT
1535 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1536 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1537 #define IFF_MACVLAN IFF_MACVLAN
1538 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1539 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1540 #define IFF_NO_QUEUE IFF_NO_QUEUE
1541 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1542 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1543 #define IFF_TEAM IFF_TEAM
1544 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1545 #define IFF_MACSEC IFF_MACSEC
1546 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1547 #define IFF_FAILOVER IFF_FAILOVER
1548 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1551 * struct net_device - The DEVICE structure.
1553 * Actually, this whole structure is a big mistake. It mixes I/O
1554 * data with strictly "high-level" data, and it has to know about
1555 * almost every data structure used in the INET module.
1557 * @name: This is the first field of the "visible" part of this structure
1558 * (i.e. as seen by users in the "Space.c" file). It is the name
1561 * @name_hlist: Device name hash chain, please keep it close to name[]
1562 * @ifalias: SNMP alias
1563 * @mem_end: Shared memory end
1564 * @mem_start: Shared memory start
1565 * @base_addr: Device I/O address
1566 * @irq: Device IRQ number
1568 * @state: Generic network queuing layer state, see netdev_state_t
1569 * @dev_list: The global list of network devices
1570 * @napi_list: List entry used for polling NAPI devices
1571 * @unreg_list: List entry when we are unregistering the
1572 * device; see the function unregister_netdev
1573 * @close_list: List entry used when we are closing the device
1574 * @ptype_all: Device-specific packet handlers for all protocols
1575 * @ptype_specific: Device-specific, protocol-specific packet handlers
1577 * @adj_list: Directly linked devices, like slaves for bonding
1578 * @features: Currently active device features
1579 * @hw_features: User-changeable features
1581 * @wanted_features: User-requested features
1582 * @vlan_features: Mask of features inheritable by VLAN devices
1584 * @hw_enc_features: Mask of features inherited by encapsulating devices
1585 * This field indicates what encapsulation
1586 * offloads the hardware is capable of doing,
1587 * and drivers will need to set them appropriately.
1589 * @mpls_features: Mask of features inheritable by MPLS
1591 * @ifindex: interface index
1592 * @group: The group the device belongs to
1594 * @stats: Statistics struct, which was left as a legacy, use
1595 * rtnl_link_stats64 instead
1597 * @rx_dropped: Dropped packets by core network,
1598 * do not use this in drivers
1599 * @tx_dropped: Dropped packets by core network,
1600 * do not use this in drivers
1601 * @rx_nohandler: nohandler dropped packets by core network on
1602 * inactive devices, do not use this in drivers
1603 * @carrier_up_count: Number of times the carrier has been up
1604 * @carrier_down_count: Number of times the carrier has been down
1606 * @wireless_handlers: List of functions to handle Wireless Extensions,
1608 * see <net/iw_handler.h> for details.
1609 * @wireless_data: Instance data managed by the core of wireless extensions
1611 * @netdev_ops: Includes several pointers to callbacks,
1612 * if one wants to override the ndo_*() functions
1613 * @ethtool_ops: Management operations
1614 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1615 * discovery handling. Necessary for e.g. 6LoWPAN.
1616 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1617 * of Layer 2 headers.
1619 * @flags: Interface flags (a la BSD)
1620 * @priv_flags: Like 'flags' but invisible to userspace,
1621 * see if.h for the definitions
1622 * @gflags: Global flags ( kept as legacy )
1623 * @padded: How much padding added by alloc_netdev()
1624 * @operstate: RFC2863 operstate
1625 * @link_mode: Mapping policy to operstate
1626 * @if_port: Selectable AUI, TP, ...
1628 * @mtu: Interface MTU value
1629 * @min_mtu: Interface Minimum MTU value
1630 * @max_mtu: Interface Maximum MTU value
1631 * @type: Interface hardware type
1632 * @hard_header_len: Maximum hardware header length.
1633 * @min_header_len: Minimum hardware header length
1635 * @needed_headroom: Extra headroom the hardware may need, but not in all
1636 * cases can this be guaranteed
1637 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1638 * cases can this be guaranteed. Some cases also use
1639 * LL_MAX_HEADER instead to allocate the skb
1641 * interface address info:
1643 * @perm_addr: Permanent hw address
1644 * @addr_assign_type: Hw address assignment type
1645 * @addr_len: Hardware address length
1646 * @neigh_priv_len: Used in neigh_alloc()
1647 * @dev_id: Used to differentiate devices that share
1648 * the same link layer address
1649 * @dev_port: Used to differentiate devices that share
1651 * @addr_list_lock: XXX: need comments on this one
1652 * @uc_promisc: Counter that indicates promiscuous mode
1653 * has been enabled due to the need to listen to
1654 * additional unicast addresses in a device that
1655 * does not implement ndo_set_rx_mode()
1656 * @uc: unicast mac addresses
1657 * @mc: multicast mac addresses
1658 * @dev_addrs: list of device hw addresses
1659 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1660 * @promiscuity: Number of times the NIC is told to work in
1661 * promiscuous mode; if it becomes 0 the NIC will
1662 * exit promiscuous mode
1663 * @allmulti: Counter, enables or disables allmulticast mode
1665 * @vlan_info: VLAN info
1666 * @dsa_ptr: dsa specific data
1667 * @tipc_ptr: TIPC specific data
1668 * @atalk_ptr: AppleTalk link
1669 * @ip_ptr: IPv4 specific data
1670 * @dn_ptr: DECnet specific data
1671 * @ip6_ptr: IPv6 specific data
1672 * @ax25_ptr: AX.25 specific data
1673 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1675 * @dev_addr: Hw address (before bcast,
1676 * because most packets are unicast)
1678 * @_rx: Array of RX queues
1679 * @num_rx_queues: Number of RX queues
1680 * allocated at register_netdev() time
1681 * @real_num_rx_queues: Number of RX queues currently active in device
1683 * @rx_handler: handler for received packets
1684 * @rx_handler_data: XXX: need comments on this one
1685 * @miniq_ingress: ingress/clsact qdisc specific data for
1686 * ingress processing
1687 * @ingress_queue: XXX: need comments on this one
1688 * @broadcast: hw bcast address
1690 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1691 * indexed by RX queue number. Assigned by driver.
1692 * This must only be set if the ndo_rx_flow_steer
1693 * operation is defined
1694 * @index_hlist: Device index hash chain
1696 * @_tx: Array of TX queues
1697 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1698 * @real_num_tx_queues: Number of TX queues currently active in device
1699 * @qdisc: Root qdisc from userspace point of view
1700 * @tx_queue_len: Max frames per queue allowed
1701 * @tx_global_lock: XXX: need comments on this one
1703 * @xps_maps: XXX: need comments on this one
1704 * @miniq_egress: clsact qdisc specific data for
1706 * @watchdog_timeo: Represents the timeout that is used by
1707 * the watchdog (see dev_watchdog())
1708 * @watchdog_timer: List of timers
1710 * @pcpu_refcnt: Number of references to this device
1711 * @todo_list: Delayed register/unregister
1712 * @link_watch_list: XXX: need comments on this one
1714 * @reg_state: Register/unregister state machine
1715 * @dismantle: Device is going to be freed
1716 * @rtnl_link_state: This enum represents the phases of creating
1719 * @needs_free_netdev: Should unregister perform free_netdev?
1720 * @priv_destructor: Called from unregister
1721 * @npinfo: XXX: need comments on this one
1722 * @nd_net: Network namespace this network device is inside
1724 * @ml_priv: Mid-layer private
1725 * @lstats: Loopback statistics
1726 * @tstats: Tunnel statistics
1727 * @dstats: Dummy statistics
1728 * @vstats: Virtual ethernet statistics
1733 * @dev: Class/net/name entry
1734 * @sysfs_groups: Space for optional device, statistics and wireless
1737 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1738 * @rtnl_link_ops: Rtnl_link_ops
1740 * @gso_max_size: Maximum size of generic segmentation offload
1741 * @gso_max_segs: Maximum number of segments that can be passed to the
1744 * @dcbnl_ops: Data Center Bridging netlink ops
1745 * @num_tc: Number of traffic classes in the net device
1746 * @tc_to_txq: XXX: need comments on this one
1747 * @prio_tc_map: XXX: need comments on this one
1749 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1751 * @priomap: XXX: need comments on this one
1752 * @phydev: Physical device may attach itself
1753 * for hardware timestamping
1754 * @sfp_bus: attached &struct sfp_bus structure.
1756 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1757 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1759 * @proto_down: protocol port state information can be sent to the
1760 * switch driver and used to set the phys state of the
1763 * @wol_enabled: Wake-on-LAN is enabled
1765 * FIXME: cleanup struct net_device such that network protocol info
1770 char name[IFNAMSIZ];
1771 struct hlist_node name_hlist;
1772 struct dev_ifalias __rcu *ifalias;
1774 * I/O specific fields
1775 * FIXME: Merge these and struct ifmap into one
1777 unsigned long mem_end;
1778 unsigned long mem_start;
1779 unsigned long base_addr;
1783 * Some hardware also needs these fields (state,dev_list,
1784 * napi_list,unreg_list,close_list) but they are not
1785 * part of the usual set specified in Space.c.
1788 unsigned long state;
1790 struct list_head dev_list;
1791 struct list_head napi_list;
1792 struct list_head unreg_list;
1793 struct list_head close_list;
1794 struct list_head ptype_all;
1795 struct list_head ptype_specific;
1798 struct list_head upper;
1799 struct list_head lower;
1802 netdev_features_t features;
1803 netdev_features_t hw_features;
1804 netdev_features_t wanted_features;
1805 netdev_features_t vlan_features;
1806 netdev_features_t hw_enc_features;
1807 netdev_features_t mpls_features;
1808 netdev_features_t gso_partial_features;
1813 struct net_device_stats stats;
1815 atomic_long_t rx_dropped;
1816 atomic_long_t tx_dropped;
1817 atomic_long_t rx_nohandler;
1819 /* Stats to monitor link on/off, flapping */
1820 atomic_t carrier_up_count;
1821 atomic_t carrier_down_count;
1823 #ifdef CONFIG_WIRELESS_EXT
1824 const struct iw_handler_def *wireless_handlers;
1825 struct iw_public_data *wireless_data;
1827 const struct net_device_ops *netdev_ops;
1828 const struct ethtool_ops *ethtool_ops;
1829 #ifdef CONFIG_NET_SWITCHDEV
1830 const struct switchdev_ops *switchdev_ops;
1832 #ifdef CONFIG_NET_L3_MASTER_DEV
1833 const struct l3mdev_ops *l3mdev_ops;
1835 #if IS_ENABLED(CONFIG_IPV6)
1836 const struct ndisc_ops *ndisc_ops;
1839 #ifdef CONFIG_XFRM_OFFLOAD
1840 const struct xfrmdev_ops *xfrmdev_ops;
1843 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1844 const struct tlsdev_ops *tlsdev_ops;
1847 const struct header_ops *header_ops;
1850 unsigned int priv_flags;
1852 unsigned short gflags;
1853 unsigned short padded;
1855 unsigned char operstate;
1856 unsigned char link_mode;
1858 unsigned char if_port;
1862 unsigned int min_mtu;
1863 unsigned int max_mtu;
1864 unsigned short type;
1865 unsigned short hard_header_len;
1866 unsigned char min_header_len;
1868 unsigned short needed_headroom;
1869 unsigned short needed_tailroom;
1871 /* Interface address info. */
1872 unsigned char perm_addr[MAX_ADDR_LEN];
1873 unsigned char addr_assign_type;
1874 unsigned char addr_len;
1875 unsigned short neigh_priv_len;
1876 unsigned short dev_id;
1877 unsigned short dev_port;
1878 spinlock_t addr_list_lock;
1879 unsigned char name_assign_type;
1881 struct netdev_hw_addr_list uc;
1882 struct netdev_hw_addr_list mc;
1883 struct netdev_hw_addr_list dev_addrs;
1886 struct kset *queues_kset;
1888 unsigned int promiscuity;
1889 unsigned int allmulti;
1892 /* Protocol-specific pointers */
1894 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1895 struct vlan_info __rcu *vlan_info;
1897 #if IS_ENABLED(CONFIG_NET_DSA)
1898 struct dsa_port *dsa_ptr;
1900 #if IS_ENABLED(CONFIG_TIPC)
1901 struct tipc_bearer __rcu *tipc_ptr;
1903 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1906 struct in_device __rcu *ip_ptr;
1907 #if IS_ENABLED(CONFIG_DECNET)
1908 struct dn_dev __rcu *dn_ptr;
1910 struct inet6_dev __rcu *ip6_ptr;
1911 #if IS_ENABLED(CONFIG_AX25)
1914 struct wireless_dev *ieee80211_ptr;
1915 struct wpan_dev *ieee802154_ptr;
1916 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1917 struct mpls_dev __rcu *mpls_ptr;
1921 * Cache lines mostly used on receive path (including eth_type_trans())
1923 /* Interface address info used in eth_type_trans() */
1924 unsigned char *dev_addr;
1926 struct netdev_rx_queue *_rx;
1927 unsigned int num_rx_queues;
1928 unsigned int real_num_rx_queues;
1930 struct bpf_prog __rcu *xdp_prog;
1931 unsigned long gro_flush_timeout;
1932 rx_handler_func_t __rcu *rx_handler;
1933 void __rcu *rx_handler_data;
1935 #ifdef CONFIG_NET_CLS_ACT
1936 struct mini_Qdisc __rcu *miniq_ingress;
1938 struct netdev_queue __rcu *ingress_queue;
1939 #ifdef CONFIG_NETFILTER_INGRESS
1940 struct nf_hook_entries __rcu *nf_hooks_ingress;
1943 unsigned char broadcast[MAX_ADDR_LEN];
1944 #ifdef CONFIG_RFS_ACCEL
1945 struct cpu_rmap *rx_cpu_rmap;
1947 struct hlist_node index_hlist;
1950 * Cache lines mostly used on transmit path
1952 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1953 unsigned int num_tx_queues;
1954 unsigned int real_num_tx_queues;
1955 struct Qdisc *qdisc;
1956 #ifdef CONFIG_NET_SCHED
1957 DECLARE_HASHTABLE (qdisc_hash, 4);
1959 unsigned int tx_queue_len;
1960 spinlock_t tx_global_lock;
1964 struct xps_dev_maps __rcu *xps_cpus_map;
1965 struct xps_dev_maps __rcu *xps_rxqs_map;
1967 #ifdef CONFIG_NET_CLS_ACT
1968 struct mini_Qdisc __rcu *miniq_egress;
1971 /* These may be needed for future network-power-down code. */
1972 struct timer_list watchdog_timer;
1974 int __percpu *pcpu_refcnt;
1975 struct list_head todo_list;
1977 struct list_head link_watch_list;
1979 enum { NETREG_UNINITIALIZED=0,
1980 NETREG_REGISTERED, /* completed register_netdevice */
1981 NETREG_UNREGISTERING, /* called unregister_netdevice */
1982 NETREG_UNREGISTERED, /* completed unregister todo */
1983 NETREG_RELEASED, /* called free_netdev */
1984 NETREG_DUMMY, /* dummy device for NAPI poll */
1990 RTNL_LINK_INITIALIZED,
1991 RTNL_LINK_INITIALIZING,
1992 } rtnl_link_state:16;
1994 bool needs_free_netdev;
1995 void (*priv_destructor)(struct net_device *dev);
1997 #ifdef CONFIG_NETPOLL
1998 struct netpoll_info __rcu *npinfo;
2001 possible_net_t nd_net;
2003 /* mid-layer private */
2006 struct pcpu_lstats __percpu *lstats;
2007 struct pcpu_sw_netstats __percpu *tstats;
2008 struct pcpu_dstats __percpu *dstats;
2011 #if IS_ENABLED(CONFIG_GARP)
2012 struct garp_port __rcu *garp_port;
2014 #if IS_ENABLED(CONFIG_MRP)
2015 struct mrp_port __rcu *mrp_port;
2019 const struct attribute_group *sysfs_groups[4];
2020 const struct attribute_group *sysfs_rx_queue_group;
2022 const struct rtnl_link_ops *rtnl_link_ops;
2024 /* for setting kernel sock attribute on TCP connection setup */
2025 #define GSO_MAX_SIZE 65536
2026 unsigned int gso_max_size;
2027 #define GSO_MAX_SEGS 65535
2031 const struct dcbnl_rtnl_ops *dcbnl_ops;
2034 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2035 u8 prio_tc_map[TC_BITMASK + 1];
2037 #if IS_ENABLED(CONFIG_FCOE)
2038 unsigned int fcoe_ddp_xid;
2040 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2041 struct netprio_map __rcu *priomap;
2043 struct phy_device *phydev;
2044 struct sfp_bus *sfp_bus;
2045 struct lock_class_key *qdisc_tx_busylock;
2046 struct lock_class_key *qdisc_running_key;
2048 unsigned wol_enabled:1;
2050 #define to_net_dev(d) container_of(d, struct net_device, dev)
2052 static inline bool netif_elide_gro(const struct net_device *dev)
2054 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2059 #define NETDEV_ALIGN 32
2062 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2064 return dev->prio_tc_map[prio & TC_BITMASK];
2068 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2070 if (tc >= dev->num_tc)
2073 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2077 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2078 void netdev_reset_tc(struct net_device *dev);
2079 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2080 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2083 int netdev_get_num_tc(struct net_device *dev)
2088 void netdev_unbind_sb_channel(struct net_device *dev,
2089 struct net_device *sb_dev);
2090 int netdev_bind_sb_channel_queue(struct net_device *dev,
2091 struct net_device *sb_dev,
2092 u8 tc, u16 count, u16 offset);
2093 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2094 static inline int netdev_get_sb_channel(struct net_device *dev)
2096 return max_t(int, -dev->num_tc, 0);
2100 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2103 return &dev->_tx[index];
2106 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2107 const struct sk_buff *skb)
2109 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2112 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2113 void (*f)(struct net_device *,
2114 struct netdev_queue *,
2120 for (i = 0; i < dev->num_tx_queues; i++)
2121 f(dev, &dev->_tx[i], arg);
2124 #define netdev_lockdep_set_classes(dev) \
2126 static struct lock_class_key qdisc_tx_busylock_key; \
2127 static struct lock_class_key qdisc_running_key; \
2128 static struct lock_class_key qdisc_xmit_lock_key; \
2129 static struct lock_class_key dev_addr_list_lock_key; \
2132 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2133 (dev)->qdisc_running_key = &qdisc_running_key; \
2134 lockdep_set_class(&(dev)->addr_list_lock, \
2135 &dev_addr_list_lock_key); \
2136 for (i = 0; i < (dev)->num_tx_queues; i++) \
2137 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2138 &qdisc_xmit_lock_key); \
2141 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2142 struct sk_buff *skb,
2143 struct net_device *sb_dev);
2145 /* returns the headroom that the master device needs to take in account
2146 * when forwarding to this dev
2148 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2150 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2153 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2155 if (dev->netdev_ops->ndo_set_rx_headroom)
2156 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2159 /* set the device rx headroom to the dev's default */
2160 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2162 netdev_set_rx_headroom(dev, -1);
2166 * Net namespace inlines
2169 struct net *dev_net(const struct net_device *dev)
2171 return read_pnet(&dev->nd_net);
2175 void dev_net_set(struct net_device *dev, struct net *net)
2177 write_pnet(&dev->nd_net, net);
2181 * netdev_priv - access network device private data
2182 * @dev: network device
2184 * Get network device private data
2186 static inline void *netdev_priv(const struct net_device *dev)
2188 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2191 /* Set the sysfs physical device reference for the network logical device
2192 * if set prior to registration will cause a symlink during initialization.
2194 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2196 /* Set the sysfs device type for the network logical device to allow
2197 * fine-grained identification of different network device types. For
2198 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2200 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2202 /* Default NAPI poll() weight
2203 * Device drivers are strongly advised to not use bigger value
2205 #define NAPI_POLL_WEIGHT 64
2208 * netif_napi_add - initialize a NAPI context
2209 * @dev: network device
2210 * @napi: NAPI context
2211 * @poll: polling function
2212 * @weight: default weight
2214 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2215 * *any* of the other NAPI-related functions.
2217 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2218 int (*poll)(struct napi_struct *, int), int weight);
2221 * netif_tx_napi_add - initialize a NAPI context
2222 * @dev: network device
2223 * @napi: NAPI context
2224 * @poll: polling function
2225 * @weight: default weight
2227 * This variant of netif_napi_add() should be used from drivers using NAPI
2228 * to exclusively poll a TX queue.
2229 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2231 static inline void netif_tx_napi_add(struct net_device *dev,
2232 struct napi_struct *napi,
2233 int (*poll)(struct napi_struct *, int),
2236 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2237 netif_napi_add(dev, napi, poll, weight);
2241 * netif_napi_del - remove a NAPI context
2242 * @napi: NAPI context
2244 * netif_napi_del() removes a NAPI context from the network device NAPI list
2246 void netif_napi_del(struct napi_struct *napi);
2248 struct napi_gro_cb {
2249 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2252 /* Length of frag0. */
2253 unsigned int frag0_len;
2255 /* This indicates where we are processing relative to skb->data. */
2258 /* This is non-zero if the packet cannot be merged with the new skb. */
2261 /* Save the IP ID here and check when we get to the transport layer */
2264 /* Number of segments aggregated. */
2267 /* Start offset for remote checksum offload */
2268 u16 gro_remcsum_start;
2270 /* jiffies when first packet was created/queued */
2273 /* Used in ipv6_gro_receive() and foo-over-udp */
2276 /* This is non-zero if the packet may be of the same flow. */
2279 /* Used in tunnel GRO receive */
2282 /* GRO checksum is valid */
2285 /* Number of checksums via CHECKSUM_UNNECESSARY */
2290 #define NAPI_GRO_FREE 1
2291 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2293 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2296 /* Used in GRE, set in fou/gue_gro_receive */
2299 /* Used to determine if flush_id can be ignored */
2302 /* Number of gro_receive callbacks this packet already went through */
2303 u8 recursion_counter:4;
2307 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2310 /* used in skb_gro_receive() slow path */
2311 struct sk_buff *last;
2314 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2316 #define GRO_RECURSION_LIMIT 15
2317 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2319 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2322 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2323 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2324 struct list_head *head,
2325 struct sk_buff *skb)
2327 if (unlikely(gro_recursion_inc_test(skb))) {
2328 NAPI_GRO_CB(skb)->flush |= 1;
2332 return cb(head, skb);
2335 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2337 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2339 struct list_head *head,
2340 struct sk_buff *skb)
2342 if (unlikely(gro_recursion_inc_test(skb))) {
2343 NAPI_GRO_CB(skb)->flush |= 1;
2347 return cb(sk, head, skb);
2350 struct packet_type {
2351 __be16 type; /* This is really htons(ether_type). */
2352 bool ignore_outgoing;
2353 struct net_device *dev; /* NULL is wildcarded here */
2354 int (*func) (struct sk_buff *,
2355 struct net_device *,
2356 struct packet_type *,
2357 struct net_device *);
2358 void (*list_func) (struct list_head *,
2359 struct packet_type *,
2360 struct net_device *);
2361 bool (*id_match)(struct packet_type *ptype,
2363 void *af_packet_priv;
2364 struct list_head list;
2367 struct offload_callbacks {
2368 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2369 netdev_features_t features);
2370 struct sk_buff *(*gro_receive)(struct list_head *head,
2371 struct sk_buff *skb);
2372 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2375 struct packet_offload {
2376 __be16 type; /* This is really htons(ether_type). */
2378 struct offload_callbacks callbacks;
2379 struct list_head list;
2382 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2383 struct pcpu_sw_netstats {
2388 struct u64_stats_sync syncp;
2389 } __aligned(4 * sizeof(u64));
2391 struct pcpu_lstats {
2394 struct u64_stats_sync syncp;
2395 } __aligned(2 * sizeof(u64));
2397 #define __netdev_alloc_pcpu_stats(type, gfp) \
2399 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2402 for_each_possible_cpu(__cpu) { \
2403 typeof(type) *stat; \
2404 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2405 u64_stats_init(&stat->syncp); \
2411 #define netdev_alloc_pcpu_stats(type) \
2412 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2414 enum netdev_lag_tx_type {
2415 NETDEV_LAG_TX_TYPE_UNKNOWN,
2416 NETDEV_LAG_TX_TYPE_RANDOM,
2417 NETDEV_LAG_TX_TYPE_BROADCAST,
2418 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2419 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2420 NETDEV_LAG_TX_TYPE_HASH,
2423 enum netdev_lag_hash {
2424 NETDEV_LAG_HASH_NONE,
2426 NETDEV_LAG_HASH_L34,
2427 NETDEV_LAG_HASH_L23,
2428 NETDEV_LAG_HASH_E23,
2429 NETDEV_LAG_HASH_E34,
2430 NETDEV_LAG_HASH_UNKNOWN,
2433 struct netdev_lag_upper_info {
2434 enum netdev_lag_tx_type tx_type;
2435 enum netdev_lag_hash hash_type;
2438 struct netdev_lag_lower_state_info {
2443 #include <linux/notifier.h>
2445 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2446 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2450 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2452 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2453 detected a hardware crash and restarted
2454 - we can use this eg to kick tcp sessions
2456 NETDEV_CHANGE, /* Notify device state change */
2459 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2460 NETDEV_CHANGEADDR, /* notify after the address change */
2461 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2465 NETDEV_BONDING_FAILOVER,
2467 NETDEV_PRE_TYPE_CHANGE,
2468 NETDEV_POST_TYPE_CHANGE,
2471 NETDEV_NOTIFY_PEERS,
2475 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2476 NETDEV_CHANGEINFODATA,
2477 NETDEV_BONDING_INFO,
2478 NETDEV_PRECHANGEUPPER,
2479 NETDEV_CHANGELOWERSTATE,
2480 NETDEV_UDP_TUNNEL_PUSH_INFO,
2481 NETDEV_UDP_TUNNEL_DROP_INFO,
2482 NETDEV_CHANGE_TX_QUEUE_LEN,
2483 NETDEV_CVLAN_FILTER_PUSH_INFO,
2484 NETDEV_CVLAN_FILTER_DROP_INFO,
2485 NETDEV_SVLAN_FILTER_PUSH_INFO,
2486 NETDEV_SVLAN_FILTER_DROP_INFO,
2488 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2490 int register_netdevice_notifier(struct notifier_block *nb);
2491 int unregister_netdevice_notifier(struct notifier_block *nb);
2493 struct netdev_notifier_info {
2494 struct net_device *dev;
2495 struct netlink_ext_ack *extack;
2498 struct netdev_notifier_info_ext {
2499 struct netdev_notifier_info info; /* must be first */
2505 struct netdev_notifier_change_info {
2506 struct netdev_notifier_info info; /* must be first */
2507 unsigned int flags_changed;
2510 struct netdev_notifier_changeupper_info {
2511 struct netdev_notifier_info info; /* must be first */
2512 struct net_device *upper_dev; /* new upper dev */
2513 bool master; /* is upper dev master */
2514 bool linking; /* is the notification for link or unlink */
2515 void *upper_info; /* upper dev info */
2518 struct netdev_notifier_changelowerstate_info {
2519 struct netdev_notifier_info info; /* must be first */
2520 void *lower_state_info; /* is lower dev state */
2523 struct netdev_notifier_pre_changeaddr_info {
2524 struct netdev_notifier_info info; /* must be first */
2525 const unsigned char *dev_addr;
2528 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2529 struct net_device *dev)
2532 info->extack = NULL;
2535 static inline struct net_device *
2536 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2541 static inline struct netlink_ext_ack *
2542 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2544 return info->extack;
2547 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2550 extern rwlock_t dev_base_lock; /* Device list lock */
2552 #define for_each_netdev(net, d) \
2553 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2554 #define for_each_netdev_reverse(net, d) \
2555 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2556 #define for_each_netdev_rcu(net, d) \
2557 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2558 #define for_each_netdev_safe(net, d, n) \
2559 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2560 #define for_each_netdev_continue(net, d) \
2561 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2562 #define for_each_netdev_continue_rcu(net, d) \
2563 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2564 #define for_each_netdev_in_bond_rcu(bond, slave) \
2565 for_each_netdev_rcu(&init_net, slave) \
2566 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2567 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2569 static inline struct net_device *next_net_device(struct net_device *dev)
2571 struct list_head *lh;
2575 lh = dev->dev_list.next;
2576 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2579 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2581 struct list_head *lh;
2585 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2586 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2589 static inline struct net_device *first_net_device(struct net *net)
2591 return list_empty(&net->dev_base_head) ? NULL :
2592 net_device_entry(net->dev_base_head.next);
2595 static inline struct net_device *first_net_device_rcu(struct net *net)
2597 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2599 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2602 int netdev_boot_setup_check(struct net_device *dev);
2603 unsigned long netdev_boot_base(const char *prefix, int unit);
2604 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2605 const char *hwaddr);
2606 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2607 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2608 void dev_add_pack(struct packet_type *pt);
2609 void dev_remove_pack(struct packet_type *pt);
2610 void __dev_remove_pack(struct packet_type *pt);
2611 void dev_add_offload(struct packet_offload *po);
2612 void dev_remove_offload(struct packet_offload *po);
2614 int dev_get_iflink(const struct net_device *dev);
2615 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2616 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2617 unsigned short mask);
2618 struct net_device *dev_get_by_name(struct net *net, const char *name);
2619 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2620 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2621 int dev_alloc_name(struct net_device *dev, const char *name);
2622 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2623 void dev_close(struct net_device *dev);
2624 void dev_close_many(struct list_head *head, bool unlink);
2625 void dev_disable_lro(struct net_device *dev);
2626 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2627 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2628 struct net_device *sb_dev,
2629 select_queue_fallback_t fallback);
2630 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2631 struct net_device *sb_dev,
2632 select_queue_fallback_t fallback);
2633 int dev_queue_xmit(struct sk_buff *skb);
2634 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2635 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2636 int register_netdevice(struct net_device *dev);
2637 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2638 void unregister_netdevice_many(struct list_head *head);
2639 static inline void unregister_netdevice(struct net_device *dev)
2641 unregister_netdevice_queue(dev, NULL);
2644 int netdev_refcnt_read(const struct net_device *dev);
2645 void free_netdev(struct net_device *dev);
2646 void netdev_freemem(struct net_device *dev);
2647 void synchronize_net(void);
2648 int init_dummy_netdev(struct net_device *dev);
2650 DECLARE_PER_CPU(int, xmit_recursion);
2651 #define XMIT_RECURSION_LIMIT 10
2653 static inline int dev_recursion_level(void)
2655 return this_cpu_read(xmit_recursion);
2658 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2659 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2660 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2661 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2662 int netdev_get_name(struct net *net, char *name, int ifindex);
2663 int dev_restart(struct net_device *dev);
2664 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2666 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2668 return NAPI_GRO_CB(skb)->data_offset;
2671 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2673 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2676 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2678 NAPI_GRO_CB(skb)->data_offset += len;
2681 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2682 unsigned int offset)
2684 return NAPI_GRO_CB(skb)->frag0 + offset;
2687 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2689 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2692 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2694 NAPI_GRO_CB(skb)->frag0 = NULL;
2695 NAPI_GRO_CB(skb)->frag0_len = 0;
2698 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2699 unsigned int offset)
2701 if (!pskb_may_pull(skb, hlen))
2704 skb_gro_frag0_invalidate(skb);
2705 return skb->data + offset;
2708 static inline void *skb_gro_network_header(struct sk_buff *skb)
2710 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2711 skb_network_offset(skb);
2714 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2715 const void *start, unsigned int len)
2717 if (NAPI_GRO_CB(skb)->csum_valid)
2718 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2719 csum_partial(start, len, 0));
2722 /* GRO checksum functions. These are logical equivalents of the normal
2723 * checksum functions (in skbuff.h) except that they operate on the GRO
2724 * offsets and fields in sk_buff.
2727 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2729 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2731 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2734 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2738 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2739 skb_checksum_start_offset(skb) <
2740 skb_gro_offset(skb)) &&
2741 !skb_at_gro_remcsum_start(skb) &&
2742 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2743 (!zero_okay || check));
2746 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2749 if (NAPI_GRO_CB(skb)->csum_valid &&
2750 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2753 NAPI_GRO_CB(skb)->csum = psum;
2755 return __skb_gro_checksum_complete(skb);
2758 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2760 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2761 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2762 NAPI_GRO_CB(skb)->csum_cnt--;
2764 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2765 * verified a new top level checksum or an encapsulated one
2766 * during GRO. This saves work if we fallback to normal path.
2768 __skb_incr_checksum_unnecessary(skb);
2772 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2775 __sum16 __ret = 0; \
2776 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2777 __ret = __skb_gro_checksum_validate_complete(skb, \
2778 compute_pseudo(skb, proto)); \
2780 skb_gro_incr_csum_unnecessary(skb); \
2784 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2785 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2787 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2789 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2791 #define skb_gro_checksum_simple_validate(skb) \
2792 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2794 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2796 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2797 !NAPI_GRO_CB(skb)->csum_valid);
2800 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2801 __sum16 check, __wsum pseudo)
2803 NAPI_GRO_CB(skb)->csum = ~pseudo;
2804 NAPI_GRO_CB(skb)->csum_valid = 1;
2807 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2809 if (__skb_gro_checksum_convert_check(skb)) \
2810 __skb_gro_checksum_convert(skb, check, \
2811 compute_pseudo(skb, proto)); \
2814 struct gro_remcsum {
2819 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2825 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2826 unsigned int off, size_t hdrlen,
2827 int start, int offset,
2828 struct gro_remcsum *grc,
2832 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2834 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2837 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2841 ptr = skb_gro_header_fast(skb, off);
2842 if (skb_gro_header_hard(skb, off + plen)) {
2843 ptr = skb_gro_header_slow(skb, off + plen, off);
2848 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2851 /* Adjust skb->csum since we changed the packet */
2852 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2854 grc->offset = off + hdrlen + offset;
2860 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2861 struct gro_remcsum *grc)
2864 size_t plen = grc->offset + sizeof(u16);
2869 ptr = skb_gro_header_fast(skb, grc->offset);
2870 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2871 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2876 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2879 #ifdef CONFIG_XFRM_OFFLOAD
2880 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2882 if (PTR_ERR(pp) != -EINPROGRESS)
2883 NAPI_GRO_CB(skb)->flush |= flush;
2885 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2888 struct gro_remcsum *grc)
2890 if (PTR_ERR(pp) != -EINPROGRESS) {
2891 NAPI_GRO_CB(skb)->flush |= flush;
2892 skb_gro_remcsum_cleanup(skb, grc);
2893 skb->remcsum_offload = 0;
2897 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2899 NAPI_GRO_CB(skb)->flush |= flush;
2901 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2904 struct gro_remcsum *grc)
2906 NAPI_GRO_CB(skb)->flush |= flush;
2907 skb_gro_remcsum_cleanup(skb, grc);
2908 skb->remcsum_offload = 0;
2912 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2913 unsigned short type,
2914 const void *daddr, const void *saddr,
2917 if (!dev->header_ops || !dev->header_ops->create)
2920 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2923 static inline int dev_parse_header(const struct sk_buff *skb,
2924 unsigned char *haddr)
2926 const struct net_device *dev = skb->dev;
2928 if (!dev->header_ops || !dev->header_ops->parse)
2930 return dev->header_ops->parse(skb, haddr);
2933 /* ll_header must have at least hard_header_len allocated */
2934 static inline bool dev_validate_header(const struct net_device *dev,
2935 char *ll_header, int len)
2937 if (likely(len >= dev->hard_header_len))
2939 if (len < dev->min_header_len)
2942 if (capable(CAP_SYS_RAWIO)) {
2943 memset(ll_header + len, 0, dev->hard_header_len - len);
2947 if (dev->header_ops && dev->header_ops->validate)
2948 return dev->header_ops->validate(ll_header, len);
2953 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2955 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2956 static inline int unregister_gifconf(unsigned int family)
2958 return register_gifconf(family, NULL);
2961 #ifdef CONFIG_NET_FLOW_LIMIT
2962 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2963 struct sd_flow_limit {
2965 unsigned int num_buckets;
2966 unsigned int history_head;
2967 u16 history[FLOW_LIMIT_HISTORY];
2971 extern int netdev_flow_limit_table_len;
2972 #endif /* CONFIG_NET_FLOW_LIMIT */
2975 * Incoming packets are placed on per-CPU queues
2977 struct softnet_data {
2978 struct list_head poll_list;
2979 struct sk_buff_head process_queue;
2982 unsigned int processed;
2983 unsigned int time_squeeze;
2984 unsigned int received_rps;
2986 struct softnet_data *rps_ipi_list;
2988 #ifdef CONFIG_NET_FLOW_LIMIT
2989 struct sd_flow_limit __rcu *flow_limit;
2991 struct Qdisc *output_queue;
2992 struct Qdisc **output_queue_tailp;
2993 struct sk_buff *completion_queue;
2994 #ifdef CONFIG_XFRM_OFFLOAD
2995 struct sk_buff_head xfrm_backlog;
2998 /* input_queue_head should be written by cpu owning this struct,
2999 * and only read by other cpus. Worth using a cache line.
3001 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3003 /* Elements below can be accessed between CPUs for RPS/RFS */
3004 call_single_data_t csd ____cacheline_aligned_in_smp;
3005 struct softnet_data *rps_ipi_next;
3007 unsigned int input_queue_tail;
3009 unsigned int dropped;
3010 struct sk_buff_head input_pkt_queue;
3011 struct napi_struct backlog;
3015 static inline void input_queue_head_incr(struct softnet_data *sd)
3018 sd->input_queue_head++;
3022 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3023 unsigned int *qtail)
3026 *qtail = ++sd->input_queue_tail;
3030 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3032 void __netif_schedule(struct Qdisc *q);
3033 void netif_schedule_queue(struct netdev_queue *txq);
3035 static inline void netif_tx_schedule_all(struct net_device *dev)
3039 for (i = 0; i < dev->num_tx_queues; i++)
3040 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3043 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3045 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3049 * netif_start_queue - allow transmit
3050 * @dev: network device
3052 * Allow upper layers to call the device hard_start_xmit routine.
3054 static inline void netif_start_queue(struct net_device *dev)
3056 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3059 static inline void netif_tx_start_all_queues(struct net_device *dev)
3063 for (i = 0; i < dev->num_tx_queues; i++) {
3064 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3065 netif_tx_start_queue(txq);
3069 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3072 * netif_wake_queue - restart transmit
3073 * @dev: network device
3075 * Allow upper layers to call the device hard_start_xmit routine.
3076 * Used for flow control when transmit resources are available.
3078 static inline void netif_wake_queue(struct net_device *dev)
3080 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3083 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3087 for (i = 0; i < dev->num_tx_queues; i++) {
3088 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3089 netif_tx_wake_queue(txq);
3093 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3095 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3099 * netif_stop_queue - stop transmitted packets
3100 * @dev: network device
3102 * Stop upper layers calling the device hard_start_xmit routine.
3103 * Used for flow control when transmit resources are unavailable.
3105 static inline void netif_stop_queue(struct net_device *dev)
3107 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3110 void netif_tx_stop_all_queues(struct net_device *dev);
3112 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3114 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3118 * netif_queue_stopped - test if transmit queue is flowblocked
3119 * @dev: network device
3121 * Test if transmit queue on device is currently unable to send.
3123 static inline bool netif_queue_stopped(const struct net_device *dev)
3125 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3128 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3130 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3134 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3136 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3140 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3142 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3146 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3147 * @dev_queue: pointer to transmit queue
3149 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3150 * to give appropriate hint to the CPU.
3152 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3155 prefetchw(&dev_queue->dql.num_queued);
3160 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3161 * @dev_queue: pointer to transmit queue
3163 * BQL enabled drivers might use this helper in their TX completion path,
3164 * to give appropriate hint to the CPU.
3166 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3169 prefetchw(&dev_queue->dql.limit);
3173 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3177 dql_queued(&dev_queue->dql, bytes);
3179 if (likely(dql_avail(&dev_queue->dql) >= 0))
3182 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3185 * The XOFF flag must be set before checking the dql_avail below,
3186 * because in netdev_tx_completed_queue we update the dql_completed
3187 * before checking the XOFF flag.
3191 /* check again in case another CPU has just made room avail */
3192 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3193 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3197 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3198 * that they should not test BQL status themselves.
3199 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3201 * Returns true if the doorbell must be used to kick the NIC.
3203 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3209 dql_queued(&dev_queue->dql, bytes);
3211 return netif_tx_queue_stopped(dev_queue);
3213 netdev_tx_sent_queue(dev_queue, bytes);
3218 * netdev_sent_queue - report the number of bytes queued to hardware
3219 * @dev: network device
3220 * @bytes: number of bytes queued to the hardware device queue
3222 * Report the number of bytes queued for sending/completion to the network
3223 * device hardware queue. @bytes should be a good approximation and should
3224 * exactly match netdev_completed_queue() @bytes
3226 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3228 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3231 static inline bool __netdev_sent_queue(struct net_device *dev,
3235 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3239 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3240 unsigned int pkts, unsigned int bytes)
3243 if (unlikely(!bytes))
3246 dql_completed(&dev_queue->dql, bytes);
3249 * Without the memory barrier there is a small possiblity that
3250 * netdev_tx_sent_queue will miss the update and cause the queue to
3251 * be stopped forever
3255 if (dql_avail(&dev_queue->dql) < 0)
3258 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3259 netif_schedule_queue(dev_queue);
3264 * netdev_completed_queue - report bytes and packets completed by device
3265 * @dev: network device
3266 * @pkts: actual number of packets sent over the medium
3267 * @bytes: actual number of bytes sent over the medium
3269 * Report the number of bytes and packets transmitted by the network device
3270 * hardware queue over the physical medium, @bytes must exactly match the
3271 * @bytes amount passed to netdev_sent_queue()
3273 static inline void netdev_completed_queue(struct net_device *dev,
3274 unsigned int pkts, unsigned int bytes)
3276 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3279 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3282 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3288 * netdev_reset_queue - reset the packets and bytes count of a network device
3289 * @dev_queue: network device
3291 * Reset the bytes and packet count of a network device and clear the
3292 * software flow control OFF bit for this network device
3294 static inline void netdev_reset_queue(struct net_device *dev_queue)
3296 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3300 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3301 * @dev: network device
3302 * @queue_index: given tx queue index
3304 * Returns 0 if given tx queue index >= number of device tx queues,
3305 * otherwise returns the originally passed tx queue index.
3307 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3309 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3310 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3311 dev->name, queue_index,
3312 dev->real_num_tx_queues);
3320 * netif_running - test if up
3321 * @dev: network device
3323 * Test if the device has been brought up.
3325 static inline bool netif_running(const struct net_device *dev)
3327 return test_bit(__LINK_STATE_START, &dev->state);
3331 * Routines to manage the subqueues on a device. We only need start,
3332 * stop, and a check if it's stopped. All other device management is
3333 * done at the overall netdevice level.
3334 * Also test the device if we're multiqueue.
3338 * netif_start_subqueue - allow sending packets on subqueue
3339 * @dev: network device
3340 * @queue_index: sub queue index
3342 * Start individual transmit queue of a device with multiple transmit queues.
3344 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3346 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3348 netif_tx_start_queue(txq);
3352 * netif_stop_subqueue - stop sending packets on subqueue
3353 * @dev: network device
3354 * @queue_index: sub queue index
3356 * Stop individual transmit queue of a device with multiple transmit queues.
3358 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3360 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3361 netif_tx_stop_queue(txq);
3365 * netif_subqueue_stopped - test status of subqueue
3366 * @dev: network device
3367 * @queue_index: sub queue index
3369 * Check individual transmit queue of a device with multiple transmit queues.
3371 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3374 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3376 return netif_tx_queue_stopped(txq);
3379 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3380 struct sk_buff *skb)
3382 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3386 * netif_wake_subqueue - allow sending packets on subqueue
3387 * @dev: network device
3388 * @queue_index: sub queue index
3390 * Resume individual transmit queue of a device with multiple transmit queues.
3392 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3394 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3396 netif_tx_wake_queue(txq);
3400 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3402 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3403 u16 index, bool is_rxqs_map);
3406 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3407 * @j: CPU/Rx queue index
3408 * @mask: bitmask of all cpus/rx queues
3409 * @nr_bits: number of bits in the bitmask
3411 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3413 static inline bool netif_attr_test_mask(unsigned long j,
3414 const unsigned long *mask,
3415 unsigned int nr_bits)
3417 cpu_max_bits_warn(j, nr_bits);
3418 return test_bit(j, mask);
3422 * netif_attr_test_online - Test for online CPU/Rx queue
3423 * @j: CPU/Rx queue index
3424 * @online_mask: bitmask for CPUs/Rx queues that are online
3425 * @nr_bits: number of bits in the bitmask
3427 * Returns true if a CPU/Rx queue is online.
3429 static inline bool netif_attr_test_online(unsigned long j,
3430 const unsigned long *online_mask,
3431 unsigned int nr_bits)
3433 cpu_max_bits_warn(j, nr_bits);
3436 return test_bit(j, online_mask);
3438 return (j < nr_bits);
3442 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3443 * @n: CPU/Rx queue index
3444 * @srcp: the cpumask/Rx queue mask pointer
3445 * @nr_bits: number of bits in the bitmask
3447 * Returns >= nr_bits if no further CPUs/Rx queues set.
3449 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3450 unsigned int nr_bits)
3452 /* -1 is a legal arg here. */
3454 cpu_max_bits_warn(n, nr_bits);
3457 return find_next_bit(srcp, nr_bits, n + 1);
3463 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3464 * @n: CPU/Rx queue index
3465 * @src1p: the first CPUs/Rx queues mask pointer
3466 * @src2p: the second CPUs/Rx queues mask pointer
3467 * @nr_bits: number of bits in the bitmask
3469 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3471 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3472 const unsigned long *src2p,
3473 unsigned int nr_bits)
3475 /* -1 is a legal arg here. */
3477 cpu_max_bits_warn(n, nr_bits);
3480 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3482 return find_next_bit(src1p, nr_bits, n + 1);
3484 return find_next_bit(src2p, nr_bits, n + 1);
3489 static inline int netif_set_xps_queue(struct net_device *dev,
3490 const struct cpumask *mask,
3496 static inline int __netif_set_xps_queue(struct net_device *dev,
3497 const unsigned long *mask,
3498 u16 index, bool is_rxqs_map)
3505 * netif_is_multiqueue - test if device has multiple transmit queues
3506 * @dev: network device
3508 * Check if device has multiple transmit queues
3510 static inline bool netif_is_multiqueue(const struct net_device *dev)
3512 return dev->num_tx_queues > 1;
3515 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3518 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3520 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3523 dev->real_num_rx_queues = rxqs;
3528 static inline struct netdev_rx_queue *
3529 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3531 return dev->_rx + rxq;
3535 static inline unsigned int get_netdev_rx_queue_index(
3536 struct netdev_rx_queue *queue)
3538 struct net_device *dev = queue->dev;
3539 int index = queue - dev->_rx;
3541 BUG_ON(index >= dev->num_rx_queues);
3546 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3547 int netif_get_num_default_rss_queues(void);
3549 enum skb_free_reason {
3550 SKB_REASON_CONSUMED,
3554 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3555 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3558 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3559 * interrupt context or with hardware interrupts being disabled.
3560 * (in_irq() || irqs_disabled())
3562 * We provide four helpers that can be used in following contexts :
3564 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3565 * replacing kfree_skb(skb)
3567 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3568 * Typically used in place of consume_skb(skb) in TX completion path
3570 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3571 * replacing kfree_skb(skb)
3573 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3574 * and consumed a packet. Used in place of consume_skb(skb)
3576 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3578 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3581 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3583 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3586 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3588 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3591 static inline void dev_consume_skb_any(struct sk_buff *skb)
3593 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3596 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3597 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3598 int netif_rx(struct sk_buff *skb);
3599 int netif_rx_ni(struct sk_buff *skb);
3600 int netif_receive_skb(struct sk_buff *skb);
3601 int netif_receive_skb_core(struct sk_buff *skb);
3602 void netif_receive_skb_list(struct list_head *head);
3603 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3604 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3605 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3606 gro_result_t napi_gro_frags(struct napi_struct *napi);
3607 struct packet_offload *gro_find_receive_by_type(__be16 type);
3608 struct packet_offload *gro_find_complete_by_type(__be16 type);
3610 static inline void napi_free_frags(struct napi_struct *napi)
3612 kfree_skb(napi->skb);
3616 bool netdev_is_rx_handler_busy(struct net_device *dev);
3617 int netdev_rx_handler_register(struct net_device *dev,
3618 rx_handler_func_t *rx_handler,
3619 void *rx_handler_data);
3620 void netdev_rx_handler_unregister(struct net_device *dev);
3622 bool dev_valid_name(const char *name);
3623 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3624 bool *need_copyout);
3625 int dev_ifconf(struct net *net, struct ifconf *, int);
3626 int dev_ethtool(struct net *net, struct ifreq *);
3627 unsigned int dev_get_flags(const struct net_device *);
3628 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3629 struct netlink_ext_ack *extack);
3630 int dev_change_flags(struct net_device *dev, unsigned int flags,
3631 struct netlink_ext_ack *extack);
3632 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3633 unsigned int gchanges);
3634 int dev_change_name(struct net_device *, const char *);
3635 int dev_set_alias(struct net_device *, const char *, size_t);
3636 int dev_get_alias(const struct net_device *, char *, size_t);
3637 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3638 int __dev_set_mtu(struct net_device *, int);
3639 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3640 struct netlink_ext_ack *extack);
3641 int dev_set_mtu(struct net_device *, int);
3642 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3643 void dev_set_group(struct net_device *, int);
3644 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3645 struct netlink_ext_ack *extack);
3646 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3647 struct netlink_ext_ack *extack);
3648 int dev_change_carrier(struct net_device *, bool new_carrier);
3649 int dev_get_phys_port_id(struct net_device *dev,
3650 struct netdev_phys_item_id *ppid);
3651 int dev_get_phys_port_name(struct net_device *dev,
3652 char *name, size_t len);
3653 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3654 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3655 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3656 struct netdev_queue *txq, int *ret);
3658 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3659 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3661 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3662 enum bpf_netdev_command cmd);
3663 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3665 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3666 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3667 bool is_skb_forwardable(const struct net_device *dev,
3668 const struct sk_buff *skb);
3670 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3671 struct sk_buff *skb)
3673 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3674 unlikely(!is_skb_forwardable(dev, skb))) {
3675 atomic_long_inc(&dev->rx_dropped);
3680 skb_scrub_packet(skb, true);
3685 bool dev_nit_active(struct net_device *dev);
3686 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3688 extern int netdev_budget;
3689 extern unsigned int netdev_budget_usecs;
3691 /* Called by rtnetlink.c:rtnl_unlock() */
3692 void netdev_run_todo(void);
3695 * dev_put - release reference to device
3696 * @dev: network device
3698 * Release reference to device to allow it to be freed.
3700 static inline void dev_put(struct net_device *dev)
3702 this_cpu_dec(*dev->pcpu_refcnt);
3706 * dev_hold - get reference to device
3707 * @dev: network device
3709 * Hold reference to device to keep it from being freed.
3711 static inline void dev_hold(struct net_device *dev)
3713 this_cpu_inc(*dev->pcpu_refcnt);
3716 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3717 * and _off may be called from IRQ context, but it is caller
3718 * who is responsible for serialization of these calls.
3720 * The name carrier is inappropriate, these functions should really be
3721 * called netif_lowerlayer_*() because they represent the state of any
3722 * kind of lower layer not just hardware media.
3725 void linkwatch_init_dev(struct net_device *dev);
3726 void linkwatch_fire_event(struct net_device *dev);
3727 void linkwatch_forget_dev(struct net_device *dev);
3730 * netif_carrier_ok - test if carrier present
3731 * @dev: network device
3733 * Check if carrier is present on device
3735 static inline bool netif_carrier_ok(const struct net_device *dev)
3737 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3740 unsigned long dev_trans_start(struct net_device *dev);
3742 void __netdev_watchdog_up(struct net_device *dev);
3744 void netif_carrier_on(struct net_device *dev);
3746 void netif_carrier_off(struct net_device *dev);
3749 * netif_dormant_on - mark device as dormant.
3750 * @dev: network device
3752 * Mark device as dormant (as per RFC2863).
3754 * The dormant state indicates that the relevant interface is not
3755 * actually in a condition to pass packets (i.e., it is not 'up') but is
3756 * in a "pending" state, waiting for some external event. For "on-
3757 * demand" interfaces, this new state identifies the situation where the
3758 * interface is waiting for events to place it in the up state.
3760 static inline void netif_dormant_on(struct net_device *dev)
3762 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3763 linkwatch_fire_event(dev);
3767 * netif_dormant_off - set device as not dormant.
3768 * @dev: network device
3770 * Device is not in dormant state.
3772 static inline void netif_dormant_off(struct net_device *dev)
3774 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3775 linkwatch_fire_event(dev);
3779 * netif_dormant - test if device is dormant
3780 * @dev: network device
3782 * Check if device is dormant.
3784 static inline bool netif_dormant(const struct net_device *dev)
3786 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3791 * netif_oper_up - test if device is operational
3792 * @dev: network device
3794 * Check if carrier is operational
3796 static inline bool netif_oper_up(const struct net_device *dev)
3798 return (dev->operstate == IF_OPER_UP ||
3799 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3803 * netif_device_present - is device available or removed
3804 * @dev: network device
3806 * Check if device has not been removed from system.
3808 static inline bool netif_device_present(struct net_device *dev)
3810 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3813 void netif_device_detach(struct net_device *dev);
3815 void netif_device_attach(struct net_device *dev);
3818 * Network interface message level settings
3822 NETIF_MSG_DRV = 0x0001,
3823 NETIF_MSG_PROBE = 0x0002,
3824 NETIF_MSG_LINK = 0x0004,
3825 NETIF_MSG_TIMER = 0x0008,
3826 NETIF_MSG_IFDOWN = 0x0010,
3827 NETIF_MSG_IFUP = 0x0020,
3828 NETIF_MSG_RX_ERR = 0x0040,
3829 NETIF_MSG_TX_ERR = 0x0080,
3830 NETIF_MSG_TX_QUEUED = 0x0100,
3831 NETIF_MSG_INTR = 0x0200,
3832 NETIF_MSG_TX_DONE = 0x0400,
3833 NETIF_MSG_RX_STATUS = 0x0800,
3834 NETIF_MSG_PKTDATA = 0x1000,
3835 NETIF_MSG_HW = 0x2000,
3836 NETIF_MSG_WOL = 0x4000,
3839 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3840 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3841 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3842 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3843 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3844 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3845 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3846 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3847 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3848 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3849 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3850 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3851 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3852 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3853 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3855 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3858 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3859 return default_msg_enable_bits;
3860 if (debug_value == 0) /* no output */
3862 /* set low N bits */
3863 return (1 << debug_value) - 1;
3866 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3868 spin_lock(&txq->_xmit_lock);
3869 txq->xmit_lock_owner = cpu;
3872 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3874 __acquire(&txq->_xmit_lock);
3878 static inline void __netif_tx_release(struct netdev_queue *txq)
3880 __release(&txq->_xmit_lock);
3883 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3885 spin_lock_bh(&txq->_xmit_lock);
3886 txq->xmit_lock_owner = smp_processor_id();
3889 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3891 bool ok = spin_trylock(&txq->_xmit_lock);
3893 txq->xmit_lock_owner = smp_processor_id();
3897 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3899 txq->xmit_lock_owner = -1;
3900 spin_unlock(&txq->_xmit_lock);
3903 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3905 txq->xmit_lock_owner = -1;
3906 spin_unlock_bh(&txq->_xmit_lock);
3909 static inline void txq_trans_update(struct netdev_queue *txq)
3911 if (txq->xmit_lock_owner != -1)
3912 txq->trans_start = jiffies;
3915 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3916 static inline void netif_trans_update(struct net_device *dev)
3918 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3920 if (txq->trans_start != jiffies)
3921 txq->trans_start = jiffies;
3925 * netif_tx_lock - grab network device transmit lock
3926 * @dev: network device
3928 * Get network device transmit lock
3930 static inline void netif_tx_lock(struct net_device *dev)
3935 spin_lock(&dev->tx_global_lock);
3936 cpu = smp_processor_id();
3937 for (i = 0; i < dev->num_tx_queues; i++) {
3938 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3940 /* We are the only thread of execution doing a
3941 * freeze, but we have to grab the _xmit_lock in
3942 * order to synchronize with threads which are in
3943 * the ->hard_start_xmit() handler and already
3944 * checked the frozen bit.
3946 __netif_tx_lock(txq, cpu);
3947 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3948 __netif_tx_unlock(txq);
3952 static inline void netif_tx_lock_bh(struct net_device *dev)
3958 static inline void netif_tx_unlock(struct net_device *dev)
3962 for (i = 0; i < dev->num_tx_queues; i++) {
3963 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3965 /* No need to grab the _xmit_lock here. If the
3966 * queue is not stopped for another reason, we
3969 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3970 netif_schedule_queue(txq);
3972 spin_unlock(&dev->tx_global_lock);
3975 static inline void netif_tx_unlock_bh(struct net_device *dev)
3977 netif_tx_unlock(dev);
3981 #define HARD_TX_LOCK(dev, txq, cpu) { \
3982 if ((dev->features & NETIF_F_LLTX) == 0) { \
3983 __netif_tx_lock(txq, cpu); \
3985 __netif_tx_acquire(txq); \
3989 #define HARD_TX_TRYLOCK(dev, txq) \
3990 (((dev->features & NETIF_F_LLTX) == 0) ? \
3991 __netif_tx_trylock(txq) : \
3992 __netif_tx_acquire(txq))
3994 #define HARD_TX_UNLOCK(dev, txq) { \
3995 if ((dev->features & NETIF_F_LLTX) == 0) { \
3996 __netif_tx_unlock(txq); \
3998 __netif_tx_release(txq); \
4002 static inline void netif_tx_disable(struct net_device *dev)
4008 cpu = smp_processor_id();
4009 for (i = 0; i < dev->num_tx_queues; i++) {
4010 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4012 __netif_tx_lock(txq, cpu);
4013 netif_tx_stop_queue(txq);
4014 __netif_tx_unlock(txq);
4019 static inline void netif_addr_lock(struct net_device *dev)
4021 spin_lock(&dev->addr_list_lock);
4024 static inline void netif_addr_lock_nested(struct net_device *dev)
4026 int subclass = SINGLE_DEPTH_NESTING;
4028 if (dev->netdev_ops->ndo_get_lock_subclass)
4029 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
4031 spin_lock_nested(&dev->addr_list_lock, subclass);
4034 static inline void netif_addr_lock_bh(struct net_device *dev)
4036 spin_lock_bh(&dev->addr_list_lock);
4039 static inline void netif_addr_unlock(struct net_device *dev)
4041 spin_unlock(&dev->addr_list_lock);
4044 static inline void netif_addr_unlock_bh(struct net_device *dev)
4046 spin_unlock_bh(&dev->addr_list_lock);
4050 * dev_addrs walker. Should be used only for read access. Call with
4051 * rcu_read_lock held.
4053 #define for_each_dev_addr(dev, ha) \
4054 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4056 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4058 void ether_setup(struct net_device *dev);
4060 /* Support for loadable net-drivers */
4061 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4062 unsigned char name_assign_type,
4063 void (*setup)(struct net_device *),
4064 unsigned int txqs, unsigned int rxqs);
4065 int dev_get_valid_name(struct net *net, struct net_device *dev,
4068 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4069 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4071 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4072 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4075 int register_netdev(struct net_device *dev);
4076 void unregister_netdev(struct net_device *dev);
4078 /* General hardware address lists handling functions */
4079 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4080 struct netdev_hw_addr_list *from_list, int addr_len);
4081 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4082 struct netdev_hw_addr_list *from_list, int addr_len);
4083 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4084 struct net_device *dev,
4085 int (*sync)(struct net_device *, const unsigned char *),
4086 int (*unsync)(struct net_device *,
4087 const unsigned char *));
4088 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4089 struct net_device *dev,
4090 int (*sync)(struct net_device *,
4091 const unsigned char *, int),
4092 int (*unsync)(struct net_device *,
4093 const unsigned char *, int));
4094 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4095 struct net_device *dev,
4096 int (*unsync)(struct net_device *,
4097 const unsigned char *, int));
4098 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4099 struct net_device *dev,
4100 int (*unsync)(struct net_device *,
4101 const unsigned char *));
4102 void __hw_addr_init(struct netdev_hw_addr_list *list);
4104 /* Functions used for device addresses handling */
4105 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4106 unsigned char addr_type);
4107 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4108 unsigned char addr_type);
4109 void dev_addr_flush(struct net_device *dev);
4110 int dev_addr_init(struct net_device *dev);
4112 /* Functions used for unicast addresses handling */
4113 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4114 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4115 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4116 int dev_uc_sync(struct net_device *to, struct net_device *from);
4117 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4118 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4119 void dev_uc_flush(struct net_device *dev);
4120 void dev_uc_init(struct net_device *dev);
4123 * __dev_uc_sync - Synchonize device's unicast list
4124 * @dev: device to sync
4125 * @sync: function to call if address should be added
4126 * @unsync: function to call if address should be removed
4128 * Add newly added addresses to the interface, and release
4129 * addresses that have been deleted.
4131 static inline int __dev_uc_sync(struct net_device *dev,
4132 int (*sync)(struct net_device *,
4133 const unsigned char *),
4134 int (*unsync)(struct net_device *,
4135 const unsigned char *))
4137 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4141 * __dev_uc_unsync - Remove synchronized addresses from device
4142 * @dev: device to sync
4143 * @unsync: function to call if address should be removed
4145 * Remove all addresses that were added to the device by dev_uc_sync().
4147 static inline void __dev_uc_unsync(struct net_device *dev,
4148 int (*unsync)(struct net_device *,
4149 const unsigned char *))
4151 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4154 /* Functions used for multicast addresses handling */
4155 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4156 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4157 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4158 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4159 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4160 int dev_mc_sync(struct net_device *to, struct net_device *from);
4161 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4162 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4163 void dev_mc_flush(struct net_device *dev);
4164 void dev_mc_init(struct net_device *dev);
4167 * __dev_mc_sync - Synchonize device's multicast list
4168 * @dev: device to sync
4169 * @sync: function to call if address should be added
4170 * @unsync: function to call if address should be removed
4172 * Add newly added addresses to the interface, and release
4173 * addresses that have been deleted.
4175 static inline int __dev_mc_sync(struct net_device *dev,
4176 int (*sync)(struct net_device *,
4177 const unsigned char *),
4178 int (*unsync)(struct net_device *,
4179 const unsigned char *))
4181 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4185 * __dev_mc_unsync - Remove synchronized addresses from device
4186 * @dev: device to sync
4187 * @unsync: function to call if address should be removed
4189 * Remove all addresses that were added to the device by dev_mc_sync().
4191 static inline void __dev_mc_unsync(struct net_device *dev,
4192 int (*unsync)(struct net_device *,
4193 const unsigned char *))
4195 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4198 /* Functions used for secondary unicast and multicast support */
4199 void dev_set_rx_mode(struct net_device *dev);
4200 void __dev_set_rx_mode(struct net_device *dev);
4201 int dev_set_promiscuity(struct net_device *dev, int inc);
4202 int dev_set_allmulti(struct net_device *dev, int inc);
4203 void netdev_state_change(struct net_device *dev);
4204 void netdev_notify_peers(struct net_device *dev);
4205 void netdev_features_change(struct net_device *dev);
4206 /* Load a device via the kmod */
4207 void dev_load(struct net *net, const char *name);
4208 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4209 struct rtnl_link_stats64 *storage);
4210 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4211 const struct net_device_stats *netdev_stats);
4213 extern int netdev_max_backlog;
4214 extern int netdev_tstamp_prequeue;
4215 extern int weight_p;
4216 extern int dev_weight_rx_bias;
4217 extern int dev_weight_tx_bias;
4218 extern int dev_rx_weight;
4219 extern int dev_tx_weight;
4221 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4222 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4223 struct list_head **iter);
4224 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4225 struct list_head **iter);
4227 /* iterate through upper list, must be called under RCU read lock */
4228 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4229 for (iter = &(dev)->adj_list.upper, \
4230 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4232 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4234 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4235 int (*fn)(struct net_device *upper_dev,
4239 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4240 struct net_device *upper_dev);
4242 bool netdev_has_any_upper_dev(struct net_device *dev);
4244 void *netdev_lower_get_next_private(struct net_device *dev,
4245 struct list_head **iter);
4246 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4247 struct list_head **iter);
4249 #define netdev_for_each_lower_private(dev, priv, iter) \
4250 for (iter = (dev)->adj_list.lower.next, \
4251 priv = netdev_lower_get_next_private(dev, &(iter)); \
4253 priv = netdev_lower_get_next_private(dev, &(iter)))
4255 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4256 for (iter = &(dev)->adj_list.lower, \
4257 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4259 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4261 void *netdev_lower_get_next(struct net_device *dev,
4262 struct list_head **iter);
4264 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4265 for (iter = (dev)->adj_list.lower.next, \
4266 ldev = netdev_lower_get_next(dev, &(iter)); \
4268 ldev = netdev_lower_get_next(dev, &(iter)))
4270 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4271 struct list_head **iter);
4272 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4273 struct list_head **iter);
4275 int netdev_walk_all_lower_dev(struct net_device *dev,
4276 int (*fn)(struct net_device *lower_dev,
4279 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4280 int (*fn)(struct net_device *lower_dev,
4284 void *netdev_adjacent_get_private(struct list_head *adj_list);
4285 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4286 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4287 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4288 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4289 struct netlink_ext_ack *extack);
4290 int netdev_master_upper_dev_link(struct net_device *dev,
4291 struct net_device *upper_dev,
4292 void *upper_priv, void *upper_info,
4293 struct netlink_ext_ack *extack);
4294 void netdev_upper_dev_unlink(struct net_device *dev,
4295 struct net_device *upper_dev);
4296 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4297 void *netdev_lower_dev_get_private(struct net_device *dev,
4298 struct net_device *lower_dev);
4299 void netdev_lower_state_changed(struct net_device *lower_dev,
4300 void *lower_state_info);
4302 /* RSS keys are 40 or 52 bytes long */
4303 #define NETDEV_RSS_KEY_LEN 52
4304 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4305 void netdev_rss_key_fill(void *buffer, size_t len);
4307 int dev_get_nest_level(struct net_device *dev);
4308 int skb_checksum_help(struct sk_buff *skb);
4309 int skb_crc32c_csum_help(struct sk_buff *skb);
4310 int skb_csum_hwoffload_help(struct sk_buff *skb,
4311 const netdev_features_t features);
4313 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4314 netdev_features_t features, bool tx_path);
4315 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4316 netdev_features_t features);
4318 struct netdev_bonding_info {
4323 struct netdev_notifier_bonding_info {
4324 struct netdev_notifier_info info; /* must be first */
4325 struct netdev_bonding_info bonding_info;
4328 void netdev_bonding_info_change(struct net_device *dev,
4329 struct netdev_bonding_info *bonding_info);
4332 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4334 return __skb_gso_segment(skb, features, true);
4336 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4338 static inline bool can_checksum_protocol(netdev_features_t features,
4341 if (protocol == htons(ETH_P_FCOE))
4342 return !!(features & NETIF_F_FCOE_CRC);
4344 /* Assume this is an IP checksum (not SCTP CRC) */
4346 if (features & NETIF_F_HW_CSUM) {
4347 /* Can checksum everything */
4352 case htons(ETH_P_IP):
4353 return !!(features & NETIF_F_IP_CSUM);
4354 case htons(ETH_P_IPV6):
4355 return !!(features & NETIF_F_IPV6_CSUM);
4362 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4364 static inline void netdev_rx_csum_fault(struct net_device *dev,
4365 struct sk_buff *skb)
4369 /* rx skb timestamps */
4370 void net_enable_timestamp(void);
4371 void net_disable_timestamp(void);
4373 #ifdef CONFIG_PROC_FS
4374 int __init dev_proc_init(void);
4376 #define dev_proc_init() 0
4379 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4380 struct sk_buff *skb, struct net_device *dev,
4383 skb->xmit_more = more ? 1 : 0;
4384 return ops->ndo_start_xmit(skb, dev);
4387 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4388 struct netdev_queue *txq, bool more)
4390 const struct net_device_ops *ops = dev->netdev_ops;
4393 rc = __netdev_start_xmit(ops, skb, dev, more);
4394 if (rc == NETDEV_TX_OK)
4395 txq_trans_update(txq);
4400 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4402 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4405 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4407 return netdev_class_create_file_ns(class_attr, NULL);
4410 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4412 netdev_class_remove_file_ns(class_attr, NULL);
4415 extern const struct kobj_ns_type_operations net_ns_type_operations;
4417 const char *netdev_drivername(const struct net_device *dev);
4419 void linkwatch_run_queue(void);
4421 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4422 netdev_features_t f2)
4424 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4425 if (f1 & NETIF_F_HW_CSUM)
4426 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4428 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4434 static inline netdev_features_t netdev_get_wanted_features(
4435 struct net_device *dev)
4437 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4439 netdev_features_t netdev_increment_features(netdev_features_t all,
4440 netdev_features_t one, netdev_features_t mask);
4442 /* Allow TSO being used on stacked device :
4443 * Performing the GSO segmentation before last device
4444 * is a performance improvement.
4446 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4447 netdev_features_t mask)
4449 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4452 int __netdev_update_features(struct net_device *dev);
4453 void netdev_update_features(struct net_device *dev);
4454 void netdev_change_features(struct net_device *dev);
4456 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4457 struct net_device *dev);
4459 netdev_features_t passthru_features_check(struct sk_buff *skb,
4460 struct net_device *dev,
4461 netdev_features_t features);
4462 netdev_features_t netif_skb_features(struct sk_buff *skb);
4464 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4466 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4468 /* check flags correspondence */
4469 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4470 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4471 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4472 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4473 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4474 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4475 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4476 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4477 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4478 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4479 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4480 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4481 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4482 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4483 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4484 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4485 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4486 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4488 return (features & feature) == feature;
4491 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4493 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4494 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4497 static inline bool netif_needs_gso(struct sk_buff *skb,
4498 netdev_features_t features)
4500 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4501 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4502 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4505 static inline void netif_set_gso_max_size(struct net_device *dev,
4508 dev->gso_max_size = size;
4511 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4512 int pulled_hlen, u16 mac_offset,
4515 skb->protocol = protocol;
4516 skb->encapsulation = 1;
4517 skb_push(skb, pulled_hlen);
4518 skb_reset_transport_header(skb);
4519 skb->mac_header = mac_offset;
4520 skb->network_header = skb->mac_header + mac_len;
4521 skb->mac_len = mac_len;
4524 static inline bool netif_is_macsec(const struct net_device *dev)
4526 return dev->priv_flags & IFF_MACSEC;
4529 static inline bool netif_is_macvlan(const struct net_device *dev)
4531 return dev->priv_flags & IFF_MACVLAN;
4534 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4536 return dev->priv_flags & IFF_MACVLAN_PORT;
4539 static inline bool netif_is_bond_master(const struct net_device *dev)
4541 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4544 static inline bool netif_is_bond_slave(const struct net_device *dev)
4546 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4549 static inline bool netif_supports_nofcs(struct net_device *dev)
4551 return dev->priv_flags & IFF_SUPP_NOFCS;
4554 static inline bool netif_is_l3_master(const struct net_device *dev)
4556 return dev->priv_flags & IFF_L3MDEV_MASTER;
4559 static inline bool netif_is_l3_slave(const struct net_device *dev)
4561 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4564 static inline bool netif_is_bridge_master(const struct net_device *dev)
4566 return dev->priv_flags & IFF_EBRIDGE;
4569 static inline bool netif_is_bridge_port(const struct net_device *dev)
4571 return dev->priv_flags & IFF_BRIDGE_PORT;
4574 static inline bool netif_is_ovs_master(const struct net_device *dev)
4576 return dev->priv_flags & IFF_OPENVSWITCH;
4579 static inline bool netif_is_ovs_port(const struct net_device *dev)
4581 return dev->priv_flags & IFF_OVS_DATAPATH;
4584 static inline bool netif_is_team_master(const struct net_device *dev)
4586 return dev->priv_flags & IFF_TEAM;
4589 static inline bool netif_is_team_port(const struct net_device *dev)
4591 return dev->priv_flags & IFF_TEAM_PORT;
4594 static inline bool netif_is_lag_master(const struct net_device *dev)
4596 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4599 static inline bool netif_is_lag_port(const struct net_device *dev)
4601 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4604 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4606 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4609 static inline bool netif_is_failover(const struct net_device *dev)
4611 return dev->priv_flags & IFF_FAILOVER;
4614 static inline bool netif_is_failover_slave(const struct net_device *dev)
4616 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4619 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4620 static inline void netif_keep_dst(struct net_device *dev)
4622 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4625 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4626 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4628 /* TODO: reserve and use an additional IFF bit, if we get more users */
4629 return dev->priv_flags & IFF_MACSEC;
4632 extern struct pernet_operations __net_initdata loopback_net_ops;
4634 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4636 /* netdev_printk helpers, similar to dev_printk */
4638 static inline const char *netdev_name(const struct net_device *dev)
4640 if (!dev->name[0] || strchr(dev->name, '%'))
4641 return "(unnamed net_device)";
4645 static inline bool netdev_unregistering(const struct net_device *dev)
4647 return dev->reg_state == NETREG_UNREGISTERING;
4650 static inline const char *netdev_reg_state(const struct net_device *dev)
4652 switch (dev->reg_state) {
4653 case NETREG_UNINITIALIZED: return " (uninitialized)";
4654 case NETREG_REGISTERED: return "";
4655 case NETREG_UNREGISTERING: return " (unregistering)";
4656 case NETREG_UNREGISTERED: return " (unregistered)";
4657 case NETREG_RELEASED: return " (released)";
4658 case NETREG_DUMMY: return " (dummy)";
4661 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4662 return " (unknown)";
4666 void netdev_printk(const char *level, const struct net_device *dev,
4667 const char *format, ...);
4669 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4671 void netdev_alert(const struct net_device *dev, const char *format, ...);
4673 void netdev_crit(const struct net_device *dev, const char *format, ...);
4675 void netdev_err(const struct net_device *dev, const char *format, ...);
4677 void netdev_warn(const struct net_device *dev, const char *format, ...);
4679 void netdev_notice(const struct net_device *dev, const char *format, ...);
4681 void netdev_info(const struct net_device *dev, const char *format, ...);
4683 #define netdev_level_once(level, dev, fmt, ...) \
4685 static bool __print_once __read_mostly; \
4687 if (!__print_once) { \
4688 __print_once = true; \
4689 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4693 #define netdev_emerg_once(dev, fmt, ...) \
4694 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4695 #define netdev_alert_once(dev, fmt, ...) \
4696 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4697 #define netdev_crit_once(dev, fmt, ...) \
4698 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4699 #define netdev_err_once(dev, fmt, ...) \
4700 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4701 #define netdev_warn_once(dev, fmt, ...) \
4702 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4703 #define netdev_notice_once(dev, fmt, ...) \
4704 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4705 #define netdev_info_once(dev, fmt, ...) \
4706 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4708 #define MODULE_ALIAS_NETDEV(device) \
4709 MODULE_ALIAS("netdev-" device)
4711 #if defined(CONFIG_DYNAMIC_DEBUG)
4712 #define netdev_dbg(__dev, format, args...) \
4714 dynamic_netdev_dbg(__dev, format, ##args); \
4716 #elif defined(DEBUG)
4717 #define netdev_dbg(__dev, format, args...) \
4718 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4720 #define netdev_dbg(__dev, format, args...) \
4723 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4727 #if defined(VERBOSE_DEBUG)
4728 #define netdev_vdbg netdev_dbg
4731 #define netdev_vdbg(dev, format, args...) \
4734 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4740 * netdev_WARN() acts like dev_printk(), but with the key difference
4741 * of using a WARN/WARN_ON to get the message out, including the
4742 * file/line information and a backtrace.
4744 #define netdev_WARN(dev, format, args...) \
4745 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4746 netdev_reg_state(dev), ##args)
4748 #define netdev_WARN_ONCE(dev, format, args...) \
4749 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4750 netdev_reg_state(dev), ##args)
4752 /* netif printk helpers, similar to netdev_printk */
4754 #define netif_printk(priv, type, level, dev, fmt, args...) \
4756 if (netif_msg_##type(priv)) \
4757 netdev_printk(level, (dev), fmt, ##args); \
4760 #define netif_level(level, priv, type, dev, fmt, args...) \
4762 if (netif_msg_##type(priv)) \
4763 netdev_##level(dev, fmt, ##args); \
4766 #define netif_emerg(priv, type, dev, fmt, args...) \
4767 netif_level(emerg, priv, type, dev, fmt, ##args)
4768 #define netif_alert(priv, type, dev, fmt, args...) \
4769 netif_level(alert, priv, type, dev, fmt, ##args)
4770 #define netif_crit(priv, type, dev, fmt, args...) \
4771 netif_level(crit, priv, type, dev, fmt, ##args)
4772 #define netif_err(priv, type, dev, fmt, args...) \
4773 netif_level(err, priv, type, dev, fmt, ##args)
4774 #define netif_warn(priv, type, dev, fmt, args...) \
4775 netif_level(warn, priv, type, dev, fmt, ##args)
4776 #define netif_notice(priv, type, dev, fmt, args...) \
4777 netif_level(notice, priv, type, dev, fmt, ##args)
4778 #define netif_info(priv, type, dev, fmt, args...) \
4779 netif_level(info, priv, type, dev, fmt, ##args)
4781 #if defined(CONFIG_DYNAMIC_DEBUG)
4782 #define netif_dbg(priv, type, netdev, format, args...) \
4784 if (netif_msg_##type(priv)) \
4785 dynamic_netdev_dbg(netdev, format, ##args); \
4787 #elif defined(DEBUG)
4788 #define netif_dbg(priv, type, dev, format, args...) \
4789 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4791 #define netif_dbg(priv, type, dev, format, args...) \
4794 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4799 /* if @cond then downgrade to debug, else print at @level */
4800 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4803 netif_dbg(priv, type, netdev, fmt, ##args); \
4805 netif_ ## level(priv, type, netdev, fmt, ##args); \
4808 #if defined(VERBOSE_DEBUG)
4809 #define netif_vdbg netif_dbg
4811 #define netif_vdbg(priv, type, dev, format, args...) \
4814 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4820 * The list of packet types we will receive (as opposed to discard)
4821 * and the routines to invoke.
4823 * Why 16. Because with 16 the only overlap we get on a hash of the
4824 * low nibble of the protocol value is RARP/SNAP/X.25.
4838 #define PTYPE_HASH_SIZE (16)
4839 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4841 #endif /* _LINUX_NETDEVICE_H */