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
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 * Adds an FDB entry to dev for addr.
1157 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1158 * struct net_device *dev,
1159 * const unsigned char *addr, u16 vid)
1160 * Deletes the FDB entry from dev coresponding to addr.
1161 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1162 * struct net_device *dev, struct net_device *filter_dev,
1164 * Used to add FDB entries to dump requests. Implementers should add
1165 * entries to skb and update idx with the number of entries.
1167 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1168 * u16 flags, struct netlink_ext_ack *extack)
1169 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1170 * struct net_device *dev, u32 filter_mask,
1172 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1175 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1176 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1177 * which do not represent real hardware may define this to allow their
1178 * userspace components to manage their virtual carrier state. Devices
1179 * that determine carrier state from physical hardware properties (eg
1180 * network cables) or protocol-dependent mechanisms (eg
1181 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1183 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1184 * struct netdev_phys_item_id *ppid);
1185 * Called to get ID of physical port of this device. If driver does
1186 * not implement this, it is assumed that the hw is not able to have
1187 * multiple net devices on single physical port.
1189 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1190 * struct udp_tunnel_info *ti);
1191 * Called by UDP tunnel to notify a driver about the UDP port and socket
1192 * address family that a UDP tunnel is listnening to. It is called only
1193 * when a new port starts listening. The operation is protected by the
1196 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1197 * struct udp_tunnel_info *ti);
1198 * Called by UDP tunnel to notify the driver about a UDP port and socket
1199 * address family that the UDP tunnel is not listening to anymore. The
1200 * operation is protected by the RTNL.
1202 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1203 * struct net_device *dev)
1204 * Called by upper layer devices to accelerate switching or other
1205 * station functionality into hardware. 'pdev is the lowerdev
1206 * to use for the offload and 'dev' is the net device that will
1207 * back the offload. Returns a pointer to the private structure
1208 * the upper layer will maintain.
1209 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1210 * Called by upper layer device to delete the station created
1211 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1212 * the station and priv is the structure returned by the add
1214 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1215 * int queue_index, u32 maxrate);
1216 * Called when a user wants to set a max-rate limitation of specific
1218 * int (*ndo_get_iflink)(const struct net_device *dev);
1219 * Called to get the iflink value of this device.
1220 * void (*ndo_change_proto_down)(struct net_device *dev,
1222 * This function is used to pass protocol port error state information
1223 * to the switch driver. The switch driver can react to the proto_down
1224 * by doing a phys down on the associated switch port.
1225 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1226 * This function is used to get egress tunnel information for given skb.
1227 * This is useful for retrieving outer tunnel header parameters while
1229 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1230 * This function is used to specify the headroom that the skb must
1231 * consider when allocation skb during packet reception. Setting
1232 * appropriate rx headroom value allows avoiding skb head copy on
1233 * forward. Setting a negative value resets the rx headroom to the
1235 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1236 * This function is used to set or query state related to XDP on the
1237 * netdevice and manage BPF offload. See definition of
1238 * enum bpf_netdev_command for details.
1239 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1241 * This function is used to submit @n XDP packets for transmit on a
1242 * netdevice. Returns number of frames successfully transmitted, frames
1243 * that got dropped are freed/returned via xdp_return_frame().
1244 * Returns negative number, means general error invoking ndo, meaning
1245 * no frames were xmit'ed and core-caller will free all frames.
1247 struct net_device_ops {
1248 int (*ndo_init)(struct net_device *dev);
1249 void (*ndo_uninit)(struct net_device *dev);
1250 int (*ndo_open)(struct net_device *dev);
1251 int (*ndo_stop)(struct net_device *dev);
1252 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1253 struct net_device *dev);
1254 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1255 struct net_device *dev,
1256 netdev_features_t features);
1257 u16 (*ndo_select_queue)(struct net_device *dev,
1258 struct sk_buff *skb,
1259 struct net_device *sb_dev,
1260 select_queue_fallback_t fallback);
1261 void (*ndo_change_rx_flags)(struct net_device *dev,
1263 void (*ndo_set_rx_mode)(struct net_device *dev);
1264 int (*ndo_set_mac_address)(struct net_device *dev,
1266 int (*ndo_validate_addr)(struct net_device *dev);
1267 int (*ndo_do_ioctl)(struct net_device *dev,
1268 struct ifreq *ifr, int cmd);
1269 int (*ndo_set_config)(struct net_device *dev,
1271 int (*ndo_change_mtu)(struct net_device *dev,
1273 int (*ndo_neigh_setup)(struct net_device *dev,
1274 struct neigh_parms *);
1275 void (*ndo_tx_timeout) (struct net_device *dev);
1277 void (*ndo_get_stats64)(struct net_device *dev,
1278 struct rtnl_link_stats64 *storage);
1279 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1280 int (*ndo_get_offload_stats)(int attr_id,
1281 const struct net_device *dev,
1283 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1285 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1286 __be16 proto, u16 vid);
1287 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1288 __be16 proto, u16 vid);
1289 #ifdef CONFIG_NET_POLL_CONTROLLER
1290 void (*ndo_poll_controller)(struct net_device *dev);
1291 int (*ndo_netpoll_setup)(struct net_device *dev,
1292 struct netpoll_info *info);
1293 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1295 int (*ndo_set_vf_mac)(struct net_device *dev,
1296 int queue, u8 *mac);
1297 int (*ndo_set_vf_vlan)(struct net_device *dev,
1298 int queue, u16 vlan,
1299 u8 qos, __be16 proto);
1300 int (*ndo_set_vf_rate)(struct net_device *dev,
1301 int vf, int min_tx_rate,
1303 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1304 int vf, bool setting);
1305 int (*ndo_set_vf_trust)(struct net_device *dev,
1306 int vf, bool setting);
1307 int (*ndo_get_vf_config)(struct net_device *dev,
1309 struct ifla_vf_info *ivf);
1310 int (*ndo_set_vf_link_state)(struct net_device *dev,
1311 int vf, int link_state);
1312 int (*ndo_get_vf_stats)(struct net_device *dev,
1314 struct ifla_vf_stats
1316 int (*ndo_set_vf_port)(struct net_device *dev,
1318 struct nlattr *port[]);
1319 int (*ndo_get_vf_port)(struct net_device *dev,
1320 int vf, struct sk_buff *skb);
1321 int (*ndo_set_vf_guid)(struct net_device *dev,
1324 int (*ndo_set_vf_rss_query_en)(
1325 struct net_device *dev,
1326 int vf, bool setting);
1327 int (*ndo_setup_tc)(struct net_device *dev,
1328 enum tc_setup_type type,
1330 #if IS_ENABLED(CONFIG_FCOE)
1331 int (*ndo_fcoe_enable)(struct net_device *dev);
1332 int (*ndo_fcoe_disable)(struct net_device *dev);
1333 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1335 struct scatterlist *sgl,
1337 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1339 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1341 struct scatterlist *sgl,
1343 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1344 struct netdev_fcoe_hbainfo *hbainfo);
1347 #if IS_ENABLED(CONFIG_LIBFCOE)
1348 #define NETDEV_FCOE_WWNN 0
1349 #define NETDEV_FCOE_WWPN 1
1350 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1351 u64 *wwn, int type);
1354 #ifdef CONFIG_RFS_ACCEL
1355 int (*ndo_rx_flow_steer)(struct net_device *dev,
1356 const struct sk_buff *skb,
1360 int (*ndo_add_slave)(struct net_device *dev,
1361 struct net_device *slave_dev,
1362 struct netlink_ext_ack *extack);
1363 int (*ndo_del_slave)(struct net_device *dev,
1364 struct net_device *slave_dev);
1365 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1366 netdev_features_t features);
1367 int (*ndo_set_features)(struct net_device *dev,
1368 netdev_features_t features);
1369 int (*ndo_neigh_construct)(struct net_device *dev,
1370 struct neighbour *n);
1371 void (*ndo_neigh_destroy)(struct net_device *dev,
1372 struct neighbour *n);
1374 int (*ndo_fdb_add)(struct ndmsg *ndm,
1375 struct nlattr *tb[],
1376 struct net_device *dev,
1377 const unsigned char *addr,
1380 int (*ndo_fdb_del)(struct ndmsg *ndm,
1381 struct nlattr *tb[],
1382 struct net_device *dev,
1383 const unsigned char *addr,
1385 int (*ndo_fdb_dump)(struct sk_buff *skb,
1386 struct netlink_callback *cb,
1387 struct net_device *dev,
1388 struct net_device *filter_dev,
1390 int (*ndo_fdb_get)(struct sk_buff *skb,
1391 struct nlattr *tb[],
1392 struct net_device *dev,
1393 const unsigned char *addr,
1394 u16 vid, u32 portid, u32 seq,
1395 struct netlink_ext_ack *extack);
1396 int (*ndo_bridge_setlink)(struct net_device *dev,
1397 struct nlmsghdr *nlh,
1399 struct netlink_ext_ack *extack);
1400 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1402 struct net_device *dev,
1405 int (*ndo_bridge_dellink)(struct net_device *dev,
1406 struct nlmsghdr *nlh,
1408 int (*ndo_change_carrier)(struct net_device *dev,
1410 int (*ndo_get_phys_port_id)(struct net_device *dev,
1411 struct netdev_phys_item_id *ppid);
1412 int (*ndo_get_phys_port_name)(struct net_device *dev,
1413 char *name, size_t len);
1414 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1415 struct udp_tunnel_info *ti);
1416 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1417 struct udp_tunnel_info *ti);
1418 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1419 struct net_device *dev);
1420 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1423 int (*ndo_get_lock_subclass)(struct net_device *dev);
1424 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1427 int (*ndo_get_iflink)(const struct net_device *dev);
1428 int (*ndo_change_proto_down)(struct net_device *dev,
1430 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1431 struct sk_buff *skb);
1432 void (*ndo_set_rx_headroom)(struct net_device *dev,
1433 int needed_headroom);
1434 int (*ndo_bpf)(struct net_device *dev,
1435 struct netdev_bpf *bpf);
1436 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1437 struct xdp_frame **xdp,
1439 int (*ndo_xsk_async_xmit)(struct net_device *dev,
1444 * enum net_device_priv_flags - &struct net_device priv_flags
1446 * These are the &struct net_device, they are only set internally
1447 * by drivers and used in the kernel. These flags are invisible to
1448 * userspace; this means that the order of these flags can change
1449 * during any kernel release.
1451 * You should have a pretty good reason to be extending these flags.
1453 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1454 * @IFF_EBRIDGE: Ethernet bridging device
1455 * @IFF_BONDING: bonding master or slave
1456 * @IFF_ISATAP: ISATAP interface (RFC4214)
1457 * @IFF_WAN_HDLC: WAN HDLC device
1458 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1460 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1461 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1462 * @IFF_MACVLAN_PORT: device used as macvlan port
1463 * @IFF_BRIDGE_PORT: device used as bridge port
1464 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1465 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1466 * @IFF_UNICAST_FLT: Supports unicast filtering
1467 * @IFF_TEAM_PORT: device used as team port
1468 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1469 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1470 * change when it's running
1471 * @IFF_MACVLAN: Macvlan device
1472 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1473 * underlying stacked devices
1474 * @IFF_L3MDEV_MASTER: device is an L3 master device
1475 * @IFF_NO_QUEUE: device can run without qdisc attached
1476 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1477 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1478 * @IFF_TEAM: device is a team device
1479 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1480 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1481 * entity (i.e. the master device for bridged veth)
1482 * @IFF_MACSEC: device is a MACsec device
1483 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1484 * @IFF_FAILOVER: device is a failover master device
1485 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1487 enum netdev_priv_flags {
1488 IFF_802_1Q_VLAN = 1<<0,
1492 IFF_WAN_HDLC = 1<<4,
1493 IFF_XMIT_DST_RELEASE = 1<<5,
1494 IFF_DONT_BRIDGE = 1<<6,
1495 IFF_DISABLE_NETPOLL = 1<<7,
1496 IFF_MACVLAN_PORT = 1<<8,
1497 IFF_BRIDGE_PORT = 1<<9,
1498 IFF_OVS_DATAPATH = 1<<10,
1499 IFF_TX_SKB_SHARING = 1<<11,
1500 IFF_UNICAST_FLT = 1<<12,
1501 IFF_TEAM_PORT = 1<<13,
1502 IFF_SUPP_NOFCS = 1<<14,
1503 IFF_LIVE_ADDR_CHANGE = 1<<15,
1504 IFF_MACVLAN = 1<<16,
1505 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1506 IFF_L3MDEV_MASTER = 1<<18,
1507 IFF_NO_QUEUE = 1<<19,
1508 IFF_OPENVSWITCH = 1<<20,
1509 IFF_L3MDEV_SLAVE = 1<<21,
1511 IFF_RXFH_CONFIGURED = 1<<23,
1512 IFF_PHONY_HEADROOM = 1<<24,
1514 IFF_NO_RX_HANDLER = 1<<26,
1515 IFF_FAILOVER = 1<<27,
1516 IFF_FAILOVER_SLAVE = 1<<28,
1519 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1520 #define IFF_EBRIDGE IFF_EBRIDGE
1521 #define IFF_BONDING IFF_BONDING
1522 #define IFF_ISATAP IFF_ISATAP
1523 #define IFF_WAN_HDLC IFF_WAN_HDLC
1524 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1525 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1526 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1527 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1528 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1529 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1530 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1531 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1532 #define IFF_TEAM_PORT IFF_TEAM_PORT
1533 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1534 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1535 #define IFF_MACVLAN IFF_MACVLAN
1536 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1537 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1538 #define IFF_NO_QUEUE IFF_NO_QUEUE
1539 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1540 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1541 #define IFF_TEAM IFF_TEAM
1542 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1543 #define IFF_MACSEC IFF_MACSEC
1544 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1545 #define IFF_FAILOVER IFF_FAILOVER
1546 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1549 * struct net_device - The DEVICE structure.
1551 * Actually, this whole structure is a big mistake. It mixes I/O
1552 * data with strictly "high-level" data, and it has to know about
1553 * almost every data structure used in the INET module.
1555 * @name: This is the first field of the "visible" part of this structure
1556 * (i.e. as seen by users in the "Space.c" file). It is the name
1559 * @name_hlist: Device name hash chain, please keep it close to name[]
1560 * @ifalias: SNMP alias
1561 * @mem_end: Shared memory end
1562 * @mem_start: Shared memory start
1563 * @base_addr: Device I/O address
1564 * @irq: Device IRQ number
1566 * @state: Generic network queuing layer state, see netdev_state_t
1567 * @dev_list: The global list of network devices
1568 * @napi_list: List entry used for polling NAPI devices
1569 * @unreg_list: List entry when we are unregistering the
1570 * device; see the function unregister_netdev
1571 * @close_list: List entry used when we are closing the device
1572 * @ptype_all: Device-specific packet handlers for all protocols
1573 * @ptype_specific: Device-specific, protocol-specific packet handlers
1575 * @adj_list: Directly linked devices, like slaves for bonding
1576 * @features: Currently active device features
1577 * @hw_features: User-changeable features
1579 * @wanted_features: User-requested features
1580 * @vlan_features: Mask of features inheritable by VLAN devices
1582 * @hw_enc_features: Mask of features inherited by encapsulating devices
1583 * This field indicates what encapsulation
1584 * offloads the hardware is capable of doing,
1585 * and drivers will need to set them appropriately.
1587 * @mpls_features: Mask of features inheritable by MPLS
1589 * @ifindex: interface index
1590 * @group: The group the device belongs to
1592 * @stats: Statistics struct, which was left as a legacy, use
1593 * rtnl_link_stats64 instead
1595 * @rx_dropped: Dropped packets by core network,
1596 * do not use this in drivers
1597 * @tx_dropped: Dropped packets by core network,
1598 * do not use this in drivers
1599 * @rx_nohandler: nohandler dropped packets by core network on
1600 * inactive devices, do not use this in drivers
1601 * @carrier_up_count: Number of times the carrier has been up
1602 * @carrier_down_count: Number of times the carrier has been down
1604 * @wireless_handlers: List of functions to handle Wireless Extensions,
1606 * see <net/iw_handler.h> for details.
1607 * @wireless_data: Instance data managed by the core of wireless extensions
1609 * @netdev_ops: Includes several pointers to callbacks,
1610 * if one wants to override the ndo_*() functions
1611 * @ethtool_ops: Management operations
1612 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1613 * discovery handling. Necessary for e.g. 6LoWPAN.
1614 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1615 * of Layer 2 headers.
1617 * @flags: Interface flags (a la BSD)
1618 * @priv_flags: Like 'flags' but invisible to userspace,
1619 * see if.h for the definitions
1620 * @gflags: Global flags ( kept as legacy )
1621 * @padded: How much padding added by alloc_netdev()
1622 * @operstate: RFC2863 operstate
1623 * @link_mode: Mapping policy to operstate
1624 * @if_port: Selectable AUI, TP, ...
1626 * @mtu: Interface MTU value
1627 * @min_mtu: Interface Minimum MTU value
1628 * @max_mtu: Interface Maximum MTU value
1629 * @type: Interface hardware type
1630 * @hard_header_len: Maximum hardware header length.
1631 * @min_header_len: Minimum hardware header length
1633 * @needed_headroom: Extra headroom the hardware may need, but not in all
1634 * cases can this be guaranteed
1635 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1636 * cases can this be guaranteed. Some cases also use
1637 * LL_MAX_HEADER instead to allocate the skb
1639 * interface address info:
1641 * @perm_addr: Permanent hw address
1642 * @addr_assign_type: Hw address assignment type
1643 * @addr_len: Hardware address length
1644 * @neigh_priv_len: Used in neigh_alloc()
1645 * @dev_id: Used to differentiate devices that share
1646 * the same link layer address
1647 * @dev_port: Used to differentiate devices that share
1649 * @addr_list_lock: XXX: need comments on this one
1650 * @uc_promisc: Counter that indicates promiscuous mode
1651 * has been enabled due to the need to listen to
1652 * additional unicast addresses in a device that
1653 * does not implement ndo_set_rx_mode()
1654 * @uc: unicast mac addresses
1655 * @mc: multicast mac addresses
1656 * @dev_addrs: list of device hw addresses
1657 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1658 * @promiscuity: Number of times the NIC is told to work in
1659 * promiscuous mode; if it becomes 0 the NIC will
1660 * exit promiscuous mode
1661 * @allmulti: Counter, enables or disables allmulticast mode
1663 * @vlan_info: VLAN info
1664 * @dsa_ptr: dsa specific data
1665 * @tipc_ptr: TIPC specific data
1666 * @atalk_ptr: AppleTalk link
1667 * @ip_ptr: IPv4 specific data
1668 * @dn_ptr: DECnet specific data
1669 * @ip6_ptr: IPv6 specific data
1670 * @ax25_ptr: AX.25 specific data
1671 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1673 * @dev_addr: Hw address (before bcast,
1674 * because most packets are unicast)
1676 * @_rx: Array of RX queues
1677 * @num_rx_queues: Number of RX queues
1678 * allocated at register_netdev() time
1679 * @real_num_rx_queues: Number of RX queues currently active in device
1681 * @rx_handler: handler for received packets
1682 * @rx_handler_data: XXX: need comments on this one
1683 * @miniq_ingress: ingress/clsact qdisc specific data for
1684 * ingress processing
1685 * @ingress_queue: XXX: need comments on this one
1686 * @broadcast: hw bcast address
1688 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1689 * indexed by RX queue number. Assigned by driver.
1690 * This must only be set if the ndo_rx_flow_steer
1691 * operation is defined
1692 * @index_hlist: Device index hash chain
1694 * @_tx: Array of TX queues
1695 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1696 * @real_num_tx_queues: Number of TX queues currently active in device
1697 * @qdisc: Root qdisc from userspace point of view
1698 * @tx_queue_len: Max frames per queue allowed
1699 * @tx_global_lock: XXX: need comments on this one
1701 * @xps_maps: XXX: need comments on this one
1702 * @miniq_egress: clsact qdisc specific data for
1704 * @watchdog_timeo: Represents the timeout that is used by
1705 * the watchdog (see dev_watchdog())
1706 * @watchdog_timer: List of timers
1708 * @pcpu_refcnt: Number of references to this device
1709 * @todo_list: Delayed register/unregister
1710 * @link_watch_list: XXX: need comments on this one
1712 * @reg_state: Register/unregister state machine
1713 * @dismantle: Device is going to be freed
1714 * @rtnl_link_state: This enum represents the phases of creating
1717 * @needs_free_netdev: Should unregister perform free_netdev?
1718 * @priv_destructor: Called from unregister
1719 * @npinfo: XXX: need comments on this one
1720 * @nd_net: Network namespace this network device is inside
1722 * @ml_priv: Mid-layer private
1723 * @lstats: Loopback statistics
1724 * @tstats: Tunnel statistics
1725 * @dstats: Dummy statistics
1726 * @vstats: Virtual ethernet statistics
1731 * @dev: Class/net/name entry
1732 * @sysfs_groups: Space for optional device, statistics and wireless
1735 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1736 * @rtnl_link_ops: Rtnl_link_ops
1738 * @gso_max_size: Maximum size of generic segmentation offload
1739 * @gso_max_segs: Maximum number of segments that can be passed to the
1742 * @dcbnl_ops: Data Center Bridging netlink ops
1743 * @num_tc: Number of traffic classes in the net device
1744 * @tc_to_txq: XXX: need comments on this one
1745 * @prio_tc_map: XXX: need comments on this one
1747 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1749 * @priomap: XXX: need comments on this one
1750 * @phydev: Physical device may attach itself
1751 * for hardware timestamping
1752 * @sfp_bus: attached &struct sfp_bus structure.
1754 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1755 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1757 * @proto_down: protocol port state information can be sent to the
1758 * switch driver and used to set the phys state of the
1761 * @wol_enabled: Wake-on-LAN is enabled
1763 * FIXME: cleanup struct net_device such that network protocol info
1768 char name[IFNAMSIZ];
1769 struct hlist_node name_hlist;
1770 struct dev_ifalias __rcu *ifalias;
1772 * I/O specific fields
1773 * FIXME: Merge these and struct ifmap into one
1775 unsigned long mem_end;
1776 unsigned long mem_start;
1777 unsigned long base_addr;
1781 * Some hardware also needs these fields (state,dev_list,
1782 * napi_list,unreg_list,close_list) but they are not
1783 * part of the usual set specified in Space.c.
1786 unsigned long state;
1788 struct list_head dev_list;
1789 struct list_head napi_list;
1790 struct list_head unreg_list;
1791 struct list_head close_list;
1792 struct list_head ptype_all;
1793 struct list_head ptype_specific;
1796 struct list_head upper;
1797 struct list_head lower;
1800 netdev_features_t features;
1801 netdev_features_t hw_features;
1802 netdev_features_t wanted_features;
1803 netdev_features_t vlan_features;
1804 netdev_features_t hw_enc_features;
1805 netdev_features_t mpls_features;
1806 netdev_features_t gso_partial_features;
1811 struct net_device_stats stats;
1813 atomic_long_t rx_dropped;
1814 atomic_long_t tx_dropped;
1815 atomic_long_t rx_nohandler;
1817 /* Stats to monitor link on/off, flapping */
1818 atomic_t carrier_up_count;
1819 atomic_t carrier_down_count;
1821 #ifdef CONFIG_WIRELESS_EXT
1822 const struct iw_handler_def *wireless_handlers;
1823 struct iw_public_data *wireless_data;
1825 const struct net_device_ops *netdev_ops;
1826 const struct ethtool_ops *ethtool_ops;
1827 #ifdef CONFIG_NET_SWITCHDEV
1828 const struct switchdev_ops *switchdev_ops;
1830 #ifdef CONFIG_NET_L3_MASTER_DEV
1831 const struct l3mdev_ops *l3mdev_ops;
1833 #if IS_ENABLED(CONFIG_IPV6)
1834 const struct ndisc_ops *ndisc_ops;
1837 #ifdef CONFIG_XFRM_OFFLOAD
1838 const struct xfrmdev_ops *xfrmdev_ops;
1841 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1842 const struct tlsdev_ops *tlsdev_ops;
1845 const struct header_ops *header_ops;
1848 unsigned int priv_flags;
1850 unsigned short gflags;
1851 unsigned short padded;
1853 unsigned char operstate;
1854 unsigned char link_mode;
1856 unsigned char if_port;
1860 unsigned int min_mtu;
1861 unsigned int max_mtu;
1862 unsigned short type;
1863 unsigned short hard_header_len;
1864 unsigned char min_header_len;
1866 unsigned short needed_headroom;
1867 unsigned short needed_tailroom;
1869 /* Interface address info. */
1870 unsigned char perm_addr[MAX_ADDR_LEN];
1871 unsigned char addr_assign_type;
1872 unsigned char addr_len;
1873 unsigned short neigh_priv_len;
1874 unsigned short dev_id;
1875 unsigned short dev_port;
1876 spinlock_t addr_list_lock;
1877 unsigned char name_assign_type;
1879 struct netdev_hw_addr_list uc;
1880 struct netdev_hw_addr_list mc;
1881 struct netdev_hw_addr_list dev_addrs;
1884 struct kset *queues_kset;
1886 unsigned int promiscuity;
1887 unsigned int allmulti;
1890 /* Protocol-specific pointers */
1892 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1893 struct vlan_info __rcu *vlan_info;
1895 #if IS_ENABLED(CONFIG_NET_DSA)
1896 struct dsa_port *dsa_ptr;
1898 #if IS_ENABLED(CONFIG_TIPC)
1899 struct tipc_bearer __rcu *tipc_ptr;
1901 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1904 struct in_device __rcu *ip_ptr;
1905 #if IS_ENABLED(CONFIG_DECNET)
1906 struct dn_dev __rcu *dn_ptr;
1908 struct inet6_dev __rcu *ip6_ptr;
1909 #if IS_ENABLED(CONFIG_AX25)
1912 struct wireless_dev *ieee80211_ptr;
1913 struct wpan_dev *ieee802154_ptr;
1914 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1915 struct mpls_dev __rcu *mpls_ptr;
1919 * Cache lines mostly used on receive path (including eth_type_trans())
1921 /* Interface address info used in eth_type_trans() */
1922 unsigned char *dev_addr;
1924 struct netdev_rx_queue *_rx;
1925 unsigned int num_rx_queues;
1926 unsigned int real_num_rx_queues;
1928 struct bpf_prog __rcu *xdp_prog;
1929 unsigned long gro_flush_timeout;
1930 rx_handler_func_t __rcu *rx_handler;
1931 void __rcu *rx_handler_data;
1933 #ifdef CONFIG_NET_CLS_ACT
1934 struct mini_Qdisc __rcu *miniq_ingress;
1936 struct netdev_queue __rcu *ingress_queue;
1937 #ifdef CONFIG_NETFILTER_INGRESS
1938 struct nf_hook_entries __rcu *nf_hooks_ingress;
1941 unsigned char broadcast[MAX_ADDR_LEN];
1942 #ifdef CONFIG_RFS_ACCEL
1943 struct cpu_rmap *rx_cpu_rmap;
1945 struct hlist_node index_hlist;
1948 * Cache lines mostly used on transmit path
1950 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1951 unsigned int num_tx_queues;
1952 unsigned int real_num_tx_queues;
1953 struct Qdisc *qdisc;
1954 #ifdef CONFIG_NET_SCHED
1955 DECLARE_HASHTABLE (qdisc_hash, 4);
1957 unsigned int tx_queue_len;
1958 spinlock_t tx_global_lock;
1962 struct xps_dev_maps __rcu *xps_cpus_map;
1963 struct xps_dev_maps __rcu *xps_rxqs_map;
1965 #ifdef CONFIG_NET_CLS_ACT
1966 struct mini_Qdisc __rcu *miniq_egress;
1969 /* These may be needed for future network-power-down code. */
1970 struct timer_list watchdog_timer;
1972 int __percpu *pcpu_refcnt;
1973 struct list_head todo_list;
1975 struct list_head link_watch_list;
1977 enum { NETREG_UNINITIALIZED=0,
1978 NETREG_REGISTERED, /* completed register_netdevice */
1979 NETREG_UNREGISTERING, /* called unregister_netdevice */
1980 NETREG_UNREGISTERED, /* completed unregister todo */
1981 NETREG_RELEASED, /* called free_netdev */
1982 NETREG_DUMMY, /* dummy device for NAPI poll */
1988 RTNL_LINK_INITIALIZED,
1989 RTNL_LINK_INITIALIZING,
1990 } rtnl_link_state:16;
1992 bool needs_free_netdev;
1993 void (*priv_destructor)(struct net_device *dev);
1995 #ifdef CONFIG_NETPOLL
1996 struct netpoll_info __rcu *npinfo;
1999 possible_net_t nd_net;
2001 /* mid-layer private */
2004 struct pcpu_lstats __percpu *lstats;
2005 struct pcpu_sw_netstats __percpu *tstats;
2006 struct pcpu_dstats __percpu *dstats;
2009 #if IS_ENABLED(CONFIG_GARP)
2010 struct garp_port __rcu *garp_port;
2012 #if IS_ENABLED(CONFIG_MRP)
2013 struct mrp_port __rcu *mrp_port;
2017 const struct attribute_group *sysfs_groups[4];
2018 const struct attribute_group *sysfs_rx_queue_group;
2020 const struct rtnl_link_ops *rtnl_link_ops;
2022 /* for setting kernel sock attribute on TCP connection setup */
2023 #define GSO_MAX_SIZE 65536
2024 unsigned int gso_max_size;
2025 #define GSO_MAX_SEGS 65535
2029 const struct dcbnl_rtnl_ops *dcbnl_ops;
2032 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2033 u8 prio_tc_map[TC_BITMASK + 1];
2035 #if IS_ENABLED(CONFIG_FCOE)
2036 unsigned int fcoe_ddp_xid;
2038 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2039 struct netprio_map __rcu *priomap;
2041 struct phy_device *phydev;
2042 struct sfp_bus *sfp_bus;
2043 struct lock_class_key *qdisc_tx_busylock;
2044 struct lock_class_key *qdisc_running_key;
2046 unsigned wol_enabled:1;
2048 #define to_net_dev(d) container_of(d, struct net_device, dev)
2050 static inline bool netif_elide_gro(const struct net_device *dev)
2052 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2057 #define NETDEV_ALIGN 32
2060 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2062 return dev->prio_tc_map[prio & TC_BITMASK];
2066 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2068 if (tc >= dev->num_tc)
2071 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2075 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2076 void netdev_reset_tc(struct net_device *dev);
2077 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2078 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2081 int netdev_get_num_tc(struct net_device *dev)
2086 void netdev_unbind_sb_channel(struct net_device *dev,
2087 struct net_device *sb_dev);
2088 int netdev_bind_sb_channel_queue(struct net_device *dev,
2089 struct net_device *sb_dev,
2090 u8 tc, u16 count, u16 offset);
2091 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2092 static inline int netdev_get_sb_channel(struct net_device *dev)
2094 return max_t(int, -dev->num_tc, 0);
2098 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2101 return &dev->_tx[index];
2104 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2105 const struct sk_buff *skb)
2107 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2110 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2111 void (*f)(struct net_device *,
2112 struct netdev_queue *,
2118 for (i = 0; i < dev->num_tx_queues; i++)
2119 f(dev, &dev->_tx[i], arg);
2122 #define netdev_lockdep_set_classes(dev) \
2124 static struct lock_class_key qdisc_tx_busylock_key; \
2125 static struct lock_class_key qdisc_running_key; \
2126 static struct lock_class_key qdisc_xmit_lock_key; \
2127 static struct lock_class_key dev_addr_list_lock_key; \
2130 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2131 (dev)->qdisc_running_key = &qdisc_running_key; \
2132 lockdep_set_class(&(dev)->addr_list_lock, \
2133 &dev_addr_list_lock_key); \
2134 for (i = 0; i < (dev)->num_tx_queues; i++) \
2135 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2136 &qdisc_xmit_lock_key); \
2139 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2140 struct sk_buff *skb,
2141 struct net_device *sb_dev);
2143 /* returns the headroom that the master device needs to take in account
2144 * when forwarding to this dev
2146 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2148 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2151 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2153 if (dev->netdev_ops->ndo_set_rx_headroom)
2154 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2157 /* set the device rx headroom to the dev's default */
2158 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2160 netdev_set_rx_headroom(dev, -1);
2164 * Net namespace inlines
2167 struct net *dev_net(const struct net_device *dev)
2169 return read_pnet(&dev->nd_net);
2173 void dev_net_set(struct net_device *dev, struct net *net)
2175 write_pnet(&dev->nd_net, net);
2179 * netdev_priv - access network device private data
2180 * @dev: network device
2182 * Get network device private data
2184 static inline void *netdev_priv(const struct net_device *dev)
2186 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2189 /* Set the sysfs physical device reference for the network logical device
2190 * if set prior to registration will cause a symlink during initialization.
2192 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2194 /* Set the sysfs device type for the network logical device to allow
2195 * fine-grained identification of different network device types. For
2196 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2198 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2200 /* Default NAPI poll() weight
2201 * Device drivers are strongly advised to not use bigger value
2203 #define NAPI_POLL_WEIGHT 64
2206 * netif_napi_add - initialize a NAPI context
2207 * @dev: network device
2208 * @napi: NAPI context
2209 * @poll: polling function
2210 * @weight: default weight
2212 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2213 * *any* of the other NAPI-related functions.
2215 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2216 int (*poll)(struct napi_struct *, int), int weight);
2219 * netif_tx_napi_add - initialize a NAPI context
2220 * @dev: network device
2221 * @napi: NAPI context
2222 * @poll: polling function
2223 * @weight: default weight
2225 * This variant of netif_napi_add() should be used from drivers using NAPI
2226 * to exclusively poll a TX queue.
2227 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2229 static inline void netif_tx_napi_add(struct net_device *dev,
2230 struct napi_struct *napi,
2231 int (*poll)(struct napi_struct *, int),
2234 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2235 netif_napi_add(dev, napi, poll, weight);
2239 * netif_napi_del - remove a NAPI context
2240 * @napi: NAPI context
2242 * netif_napi_del() removes a NAPI context from the network device NAPI list
2244 void netif_napi_del(struct napi_struct *napi);
2246 struct napi_gro_cb {
2247 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2250 /* Length of frag0. */
2251 unsigned int frag0_len;
2253 /* This indicates where we are processing relative to skb->data. */
2256 /* This is non-zero if the packet cannot be merged with the new skb. */
2259 /* Save the IP ID here and check when we get to the transport layer */
2262 /* Number of segments aggregated. */
2265 /* Start offset for remote checksum offload */
2266 u16 gro_remcsum_start;
2268 /* jiffies when first packet was created/queued */
2271 /* Used in ipv6_gro_receive() and foo-over-udp */
2274 /* This is non-zero if the packet may be of the same flow. */
2277 /* Used in tunnel GRO receive */
2280 /* GRO checksum is valid */
2283 /* Number of checksums via CHECKSUM_UNNECESSARY */
2288 #define NAPI_GRO_FREE 1
2289 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2291 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2294 /* Used in GRE, set in fou/gue_gro_receive */
2297 /* Used to determine if flush_id can be ignored */
2300 /* Number of gro_receive callbacks this packet already went through */
2301 u8 recursion_counter:4;
2305 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2308 /* used in skb_gro_receive() slow path */
2309 struct sk_buff *last;
2312 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2314 #define GRO_RECURSION_LIMIT 15
2315 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2317 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2320 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2321 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2322 struct list_head *head,
2323 struct sk_buff *skb)
2325 if (unlikely(gro_recursion_inc_test(skb))) {
2326 NAPI_GRO_CB(skb)->flush |= 1;
2330 return cb(head, skb);
2333 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2335 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2337 struct list_head *head,
2338 struct sk_buff *skb)
2340 if (unlikely(gro_recursion_inc_test(skb))) {
2341 NAPI_GRO_CB(skb)->flush |= 1;
2345 return cb(sk, head, skb);
2348 struct packet_type {
2349 __be16 type; /* This is really htons(ether_type). */
2350 bool ignore_outgoing;
2351 struct net_device *dev; /* NULL is wildcarded here */
2352 int (*func) (struct sk_buff *,
2353 struct net_device *,
2354 struct packet_type *,
2355 struct net_device *);
2356 void (*list_func) (struct list_head *,
2357 struct packet_type *,
2358 struct net_device *);
2359 bool (*id_match)(struct packet_type *ptype,
2361 void *af_packet_priv;
2362 struct list_head list;
2365 struct offload_callbacks {
2366 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2367 netdev_features_t features);
2368 struct sk_buff *(*gro_receive)(struct list_head *head,
2369 struct sk_buff *skb);
2370 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2373 struct packet_offload {
2374 __be16 type; /* This is really htons(ether_type). */
2376 struct offload_callbacks callbacks;
2377 struct list_head list;
2380 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2381 struct pcpu_sw_netstats {
2386 struct u64_stats_sync syncp;
2387 } __aligned(4 * sizeof(u64));
2389 struct pcpu_lstats {
2392 struct u64_stats_sync syncp;
2393 } __aligned(2 * sizeof(u64));
2395 #define __netdev_alloc_pcpu_stats(type, gfp) \
2397 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2400 for_each_possible_cpu(__cpu) { \
2401 typeof(type) *stat; \
2402 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2403 u64_stats_init(&stat->syncp); \
2409 #define netdev_alloc_pcpu_stats(type) \
2410 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2412 enum netdev_lag_tx_type {
2413 NETDEV_LAG_TX_TYPE_UNKNOWN,
2414 NETDEV_LAG_TX_TYPE_RANDOM,
2415 NETDEV_LAG_TX_TYPE_BROADCAST,
2416 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2417 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2418 NETDEV_LAG_TX_TYPE_HASH,
2421 enum netdev_lag_hash {
2422 NETDEV_LAG_HASH_NONE,
2424 NETDEV_LAG_HASH_L34,
2425 NETDEV_LAG_HASH_L23,
2426 NETDEV_LAG_HASH_E23,
2427 NETDEV_LAG_HASH_E34,
2428 NETDEV_LAG_HASH_UNKNOWN,
2431 struct netdev_lag_upper_info {
2432 enum netdev_lag_tx_type tx_type;
2433 enum netdev_lag_hash hash_type;
2436 struct netdev_lag_lower_state_info {
2441 #include <linux/notifier.h>
2443 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2444 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2448 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2450 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2451 detected a hardware crash and restarted
2452 - we can use this eg to kick tcp sessions
2454 NETDEV_CHANGE, /* Notify device state change */
2457 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2458 NETDEV_CHANGEADDR, /* notify after the address change */
2459 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2463 NETDEV_BONDING_FAILOVER,
2465 NETDEV_PRE_TYPE_CHANGE,
2466 NETDEV_POST_TYPE_CHANGE,
2469 NETDEV_NOTIFY_PEERS,
2473 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2474 NETDEV_CHANGEINFODATA,
2475 NETDEV_BONDING_INFO,
2476 NETDEV_PRECHANGEUPPER,
2477 NETDEV_CHANGELOWERSTATE,
2478 NETDEV_UDP_TUNNEL_PUSH_INFO,
2479 NETDEV_UDP_TUNNEL_DROP_INFO,
2480 NETDEV_CHANGE_TX_QUEUE_LEN,
2481 NETDEV_CVLAN_FILTER_PUSH_INFO,
2482 NETDEV_CVLAN_FILTER_DROP_INFO,
2483 NETDEV_SVLAN_FILTER_PUSH_INFO,
2484 NETDEV_SVLAN_FILTER_DROP_INFO,
2486 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2488 int register_netdevice_notifier(struct notifier_block *nb);
2489 int unregister_netdevice_notifier(struct notifier_block *nb);
2491 struct netdev_notifier_info {
2492 struct net_device *dev;
2493 struct netlink_ext_ack *extack;
2496 struct netdev_notifier_info_ext {
2497 struct netdev_notifier_info info; /* must be first */
2503 struct netdev_notifier_change_info {
2504 struct netdev_notifier_info info; /* must be first */
2505 unsigned int flags_changed;
2508 struct netdev_notifier_changeupper_info {
2509 struct netdev_notifier_info info; /* must be first */
2510 struct net_device *upper_dev; /* new upper dev */
2511 bool master; /* is upper dev master */
2512 bool linking; /* is the notification for link or unlink */
2513 void *upper_info; /* upper dev info */
2516 struct netdev_notifier_changelowerstate_info {
2517 struct netdev_notifier_info info; /* must be first */
2518 void *lower_state_info; /* is lower dev state */
2521 struct netdev_notifier_pre_changeaddr_info {
2522 struct netdev_notifier_info info; /* must be first */
2523 const unsigned char *dev_addr;
2526 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2527 struct net_device *dev)
2530 info->extack = NULL;
2533 static inline struct net_device *
2534 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2539 static inline struct netlink_ext_ack *
2540 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2542 return info->extack;
2545 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2548 extern rwlock_t dev_base_lock; /* Device list lock */
2550 #define for_each_netdev(net, d) \
2551 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2552 #define for_each_netdev_reverse(net, d) \
2553 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2554 #define for_each_netdev_rcu(net, d) \
2555 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2556 #define for_each_netdev_safe(net, d, n) \
2557 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2558 #define for_each_netdev_continue(net, d) \
2559 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2560 #define for_each_netdev_continue_rcu(net, d) \
2561 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2562 #define for_each_netdev_in_bond_rcu(bond, slave) \
2563 for_each_netdev_rcu(&init_net, slave) \
2564 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2565 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2567 static inline struct net_device *next_net_device(struct net_device *dev)
2569 struct list_head *lh;
2573 lh = dev->dev_list.next;
2574 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2577 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2579 struct list_head *lh;
2583 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2584 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2587 static inline struct net_device *first_net_device(struct net *net)
2589 return list_empty(&net->dev_base_head) ? NULL :
2590 net_device_entry(net->dev_base_head.next);
2593 static inline struct net_device *first_net_device_rcu(struct net *net)
2595 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2597 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2600 int netdev_boot_setup_check(struct net_device *dev);
2601 unsigned long netdev_boot_base(const char *prefix, int unit);
2602 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2603 const char *hwaddr);
2604 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2605 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2606 void dev_add_pack(struct packet_type *pt);
2607 void dev_remove_pack(struct packet_type *pt);
2608 void __dev_remove_pack(struct packet_type *pt);
2609 void dev_add_offload(struct packet_offload *po);
2610 void dev_remove_offload(struct packet_offload *po);
2612 int dev_get_iflink(const struct net_device *dev);
2613 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2614 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2615 unsigned short mask);
2616 struct net_device *dev_get_by_name(struct net *net, const char *name);
2617 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2618 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2619 int dev_alloc_name(struct net_device *dev, const char *name);
2620 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2621 void dev_close(struct net_device *dev);
2622 void dev_close_many(struct list_head *head, bool unlink);
2623 void dev_disable_lro(struct net_device *dev);
2624 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2625 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2626 struct net_device *sb_dev,
2627 select_queue_fallback_t fallback);
2628 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2629 struct net_device *sb_dev,
2630 select_queue_fallback_t fallback);
2631 int dev_queue_xmit(struct sk_buff *skb);
2632 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2633 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2634 int register_netdevice(struct net_device *dev);
2635 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2636 void unregister_netdevice_many(struct list_head *head);
2637 static inline void unregister_netdevice(struct net_device *dev)
2639 unregister_netdevice_queue(dev, NULL);
2642 int netdev_refcnt_read(const struct net_device *dev);
2643 void free_netdev(struct net_device *dev);
2644 void netdev_freemem(struct net_device *dev);
2645 void synchronize_net(void);
2646 int init_dummy_netdev(struct net_device *dev);
2648 DECLARE_PER_CPU(int, xmit_recursion);
2649 #define XMIT_RECURSION_LIMIT 10
2651 static inline int dev_recursion_level(void)
2653 return this_cpu_read(xmit_recursion);
2656 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2657 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2658 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2659 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2660 int netdev_get_name(struct net *net, char *name, int ifindex);
2661 int dev_restart(struct net_device *dev);
2662 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2664 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2666 return NAPI_GRO_CB(skb)->data_offset;
2669 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2671 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2674 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2676 NAPI_GRO_CB(skb)->data_offset += len;
2679 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2680 unsigned int offset)
2682 return NAPI_GRO_CB(skb)->frag0 + offset;
2685 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2687 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2690 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2692 NAPI_GRO_CB(skb)->frag0 = NULL;
2693 NAPI_GRO_CB(skb)->frag0_len = 0;
2696 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2697 unsigned int offset)
2699 if (!pskb_may_pull(skb, hlen))
2702 skb_gro_frag0_invalidate(skb);
2703 return skb->data + offset;
2706 static inline void *skb_gro_network_header(struct sk_buff *skb)
2708 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2709 skb_network_offset(skb);
2712 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2713 const void *start, unsigned int len)
2715 if (NAPI_GRO_CB(skb)->csum_valid)
2716 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2717 csum_partial(start, len, 0));
2720 /* GRO checksum functions. These are logical equivalents of the normal
2721 * checksum functions (in skbuff.h) except that they operate on the GRO
2722 * offsets and fields in sk_buff.
2725 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2727 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2729 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2732 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2736 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2737 skb_checksum_start_offset(skb) <
2738 skb_gro_offset(skb)) &&
2739 !skb_at_gro_remcsum_start(skb) &&
2740 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2741 (!zero_okay || check));
2744 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2747 if (NAPI_GRO_CB(skb)->csum_valid &&
2748 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2751 NAPI_GRO_CB(skb)->csum = psum;
2753 return __skb_gro_checksum_complete(skb);
2756 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2758 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2759 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2760 NAPI_GRO_CB(skb)->csum_cnt--;
2762 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2763 * verified a new top level checksum or an encapsulated one
2764 * during GRO. This saves work if we fallback to normal path.
2766 __skb_incr_checksum_unnecessary(skb);
2770 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2773 __sum16 __ret = 0; \
2774 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2775 __ret = __skb_gro_checksum_validate_complete(skb, \
2776 compute_pseudo(skb, proto)); \
2778 skb_gro_incr_csum_unnecessary(skb); \
2782 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2783 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2785 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2787 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2789 #define skb_gro_checksum_simple_validate(skb) \
2790 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2792 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2794 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2795 !NAPI_GRO_CB(skb)->csum_valid);
2798 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2799 __sum16 check, __wsum pseudo)
2801 NAPI_GRO_CB(skb)->csum = ~pseudo;
2802 NAPI_GRO_CB(skb)->csum_valid = 1;
2805 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2807 if (__skb_gro_checksum_convert_check(skb)) \
2808 __skb_gro_checksum_convert(skb, check, \
2809 compute_pseudo(skb, proto)); \
2812 struct gro_remcsum {
2817 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2823 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2824 unsigned int off, size_t hdrlen,
2825 int start, int offset,
2826 struct gro_remcsum *grc,
2830 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2832 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2835 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2839 ptr = skb_gro_header_fast(skb, off);
2840 if (skb_gro_header_hard(skb, off + plen)) {
2841 ptr = skb_gro_header_slow(skb, off + plen, off);
2846 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2849 /* Adjust skb->csum since we changed the packet */
2850 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2852 grc->offset = off + hdrlen + offset;
2858 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2859 struct gro_remcsum *grc)
2862 size_t plen = grc->offset + sizeof(u16);
2867 ptr = skb_gro_header_fast(skb, grc->offset);
2868 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2869 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2874 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2877 #ifdef CONFIG_XFRM_OFFLOAD
2878 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2880 if (PTR_ERR(pp) != -EINPROGRESS)
2881 NAPI_GRO_CB(skb)->flush |= flush;
2883 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2886 struct gro_remcsum *grc)
2888 if (PTR_ERR(pp) != -EINPROGRESS) {
2889 NAPI_GRO_CB(skb)->flush |= flush;
2890 skb_gro_remcsum_cleanup(skb, grc);
2891 skb->remcsum_offload = 0;
2895 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2897 NAPI_GRO_CB(skb)->flush |= flush;
2899 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2902 struct gro_remcsum *grc)
2904 NAPI_GRO_CB(skb)->flush |= flush;
2905 skb_gro_remcsum_cleanup(skb, grc);
2906 skb->remcsum_offload = 0;
2910 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2911 unsigned short type,
2912 const void *daddr, const void *saddr,
2915 if (!dev->header_ops || !dev->header_ops->create)
2918 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2921 static inline int dev_parse_header(const struct sk_buff *skb,
2922 unsigned char *haddr)
2924 const struct net_device *dev = skb->dev;
2926 if (!dev->header_ops || !dev->header_ops->parse)
2928 return dev->header_ops->parse(skb, haddr);
2931 /* ll_header must have at least hard_header_len allocated */
2932 static inline bool dev_validate_header(const struct net_device *dev,
2933 char *ll_header, int len)
2935 if (likely(len >= dev->hard_header_len))
2937 if (len < dev->min_header_len)
2940 if (capable(CAP_SYS_RAWIO)) {
2941 memset(ll_header + len, 0, dev->hard_header_len - len);
2945 if (dev->header_ops && dev->header_ops->validate)
2946 return dev->header_ops->validate(ll_header, len);
2951 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2953 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2954 static inline int unregister_gifconf(unsigned int family)
2956 return register_gifconf(family, NULL);
2959 #ifdef CONFIG_NET_FLOW_LIMIT
2960 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2961 struct sd_flow_limit {
2963 unsigned int num_buckets;
2964 unsigned int history_head;
2965 u16 history[FLOW_LIMIT_HISTORY];
2969 extern int netdev_flow_limit_table_len;
2970 #endif /* CONFIG_NET_FLOW_LIMIT */
2973 * Incoming packets are placed on per-CPU queues
2975 struct softnet_data {
2976 struct list_head poll_list;
2977 struct sk_buff_head process_queue;
2980 unsigned int processed;
2981 unsigned int time_squeeze;
2982 unsigned int received_rps;
2984 struct softnet_data *rps_ipi_list;
2986 #ifdef CONFIG_NET_FLOW_LIMIT
2987 struct sd_flow_limit __rcu *flow_limit;
2989 struct Qdisc *output_queue;
2990 struct Qdisc **output_queue_tailp;
2991 struct sk_buff *completion_queue;
2992 #ifdef CONFIG_XFRM_OFFLOAD
2993 struct sk_buff_head xfrm_backlog;
2996 /* input_queue_head should be written by cpu owning this struct,
2997 * and only read by other cpus. Worth using a cache line.
2999 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3001 /* Elements below can be accessed between CPUs for RPS/RFS */
3002 call_single_data_t csd ____cacheline_aligned_in_smp;
3003 struct softnet_data *rps_ipi_next;
3005 unsigned int input_queue_tail;
3007 unsigned int dropped;
3008 struct sk_buff_head input_pkt_queue;
3009 struct napi_struct backlog;
3013 static inline void input_queue_head_incr(struct softnet_data *sd)
3016 sd->input_queue_head++;
3020 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3021 unsigned int *qtail)
3024 *qtail = ++sd->input_queue_tail;
3028 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3030 void __netif_schedule(struct Qdisc *q);
3031 void netif_schedule_queue(struct netdev_queue *txq);
3033 static inline void netif_tx_schedule_all(struct net_device *dev)
3037 for (i = 0; i < dev->num_tx_queues; i++)
3038 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3041 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3043 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3047 * netif_start_queue - allow transmit
3048 * @dev: network device
3050 * Allow upper layers to call the device hard_start_xmit routine.
3052 static inline void netif_start_queue(struct net_device *dev)
3054 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3057 static inline void netif_tx_start_all_queues(struct net_device *dev)
3061 for (i = 0; i < dev->num_tx_queues; i++) {
3062 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3063 netif_tx_start_queue(txq);
3067 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3070 * netif_wake_queue - restart transmit
3071 * @dev: network device
3073 * Allow upper layers to call the device hard_start_xmit routine.
3074 * Used for flow control when transmit resources are available.
3076 static inline void netif_wake_queue(struct net_device *dev)
3078 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3081 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3085 for (i = 0; i < dev->num_tx_queues; i++) {
3086 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3087 netif_tx_wake_queue(txq);
3091 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3093 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3097 * netif_stop_queue - stop transmitted packets
3098 * @dev: network device
3100 * Stop upper layers calling the device hard_start_xmit routine.
3101 * Used for flow control when transmit resources are unavailable.
3103 static inline void netif_stop_queue(struct net_device *dev)
3105 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3108 void netif_tx_stop_all_queues(struct net_device *dev);
3110 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3112 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3116 * netif_queue_stopped - test if transmit queue is flowblocked
3117 * @dev: network device
3119 * Test if transmit queue on device is currently unable to send.
3121 static inline bool netif_queue_stopped(const struct net_device *dev)
3123 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3126 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3128 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3132 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3134 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3138 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3140 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3144 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3145 * @dev_queue: pointer to transmit queue
3147 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3148 * to give appropriate hint to the CPU.
3150 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3153 prefetchw(&dev_queue->dql.num_queued);
3158 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3159 * @dev_queue: pointer to transmit queue
3161 * BQL enabled drivers might use this helper in their TX completion path,
3162 * to give appropriate hint to the CPU.
3164 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3167 prefetchw(&dev_queue->dql.limit);
3171 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3175 dql_queued(&dev_queue->dql, bytes);
3177 if (likely(dql_avail(&dev_queue->dql) >= 0))
3180 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3183 * The XOFF flag must be set before checking the dql_avail below,
3184 * because in netdev_tx_completed_queue we update the dql_completed
3185 * before checking the XOFF flag.
3189 /* check again in case another CPU has just made room avail */
3190 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3191 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3195 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3196 * that they should not test BQL status themselves.
3197 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3199 * Returns true if the doorbell must be used to kick the NIC.
3201 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3207 dql_queued(&dev_queue->dql, bytes);
3209 return netif_tx_queue_stopped(dev_queue);
3211 netdev_tx_sent_queue(dev_queue, bytes);
3216 * netdev_sent_queue - report the number of bytes queued to hardware
3217 * @dev: network device
3218 * @bytes: number of bytes queued to the hardware device queue
3220 * Report the number of bytes queued for sending/completion to the network
3221 * device hardware queue. @bytes should be a good approximation and should
3222 * exactly match netdev_completed_queue() @bytes
3224 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3226 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3229 static inline bool __netdev_sent_queue(struct net_device *dev,
3233 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3237 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3238 unsigned int pkts, unsigned int bytes)
3241 if (unlikely(!bytes))
3244 dql_completed(&dev_queue->dql, bytes);
3247 * Without the memory barrier there is a small possiblity that
3248 * netdev_tx_sent_queue will miss the update and cause the queue to
3249 * be stopped forever
3253 if (dql_avail(&dev_queue->dql) < 0)
3256 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3257 netif_schedule_queue(dev_queue);
3262 * netdev_completed_queue - report bytes and packets completed by device
3263 * @dev: network device
3264 * @pkts: actual number of packets sent over the medium
3265 * @bytes: actual number of bytes sent over the medium
3267 * Report the number of bytes and packets transmitted by the network device
3268 * hardware queue over the physical medium, @bytes must exactly match the
3269 * @bytes amount passed to netdev_sent_queue()
3271 static inline void netdev_completed_queue(struct net_device *dev,
3272 unsigned int pkts, unsigned int bytes)
3274 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3277 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3280 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3286 * netdev_reset_queue - reset the packets and bytes count of a network device
3287 * @dev_queue: network device
3289 * Reset the bytes and packet count of a network device and clear the
3290 * software flow control OFF bit for this network device
3292 static inline void netdev_reset_queue(struct net_device *dev_queue)
3294 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3298 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3299 * @dev: network device
3300 * @queue_index: given tx queue index
3302 * Returns 0 if given tx queue index >= number of device tx queues,
3303 * otherwise returns the originally passed tx queue index.
3305 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3307 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3308 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3309 dev->name, queue_index,
3310 dev->real_num_tx_queues);
3318 * netif_running - test if up
3319 * @dev: network device
3321 * Test if the device has been brought up.
3323 static inline bool netif_running(const struct net_device *dev)
3325 return test_bit(__LINK_STATE_START, &dev->state);
3329 * Routines to manage the subqueues on a device. We only need start,
3330 * stop, and a check if it's stopped. All other device management is
3331 * done at the overall netdevice level.
3332 * Also test the device if we're multiqueue.
3336 * netif_start_subqueue - allow sending packets on subqueue
3337 * @dev: network device
3338 * @queue_index: sub queue index
3340 * Start individual transmit queue of a device with multiple transmit queues.
3342 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3344 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3346 netif_tx_start_queue(txq);
3350 * netif_stop_subqueue - stop sending packets on subqueue
3351 * @dev: network device
3352 * @queue_index: sub queue index
3354 * Stop individual transmit queue of a device with multiple transmit queues.
3356 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3358 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3359 netif_tx_stop_queue(txq);
3363 * netif_subqueue_stopped - test status of subqueue
3364 * @dev: network device
3365 * @queue_index: sub queue index
3367 * Check individual transmit queue of a device with multiple transmit queues.
3369 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3372 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3374 return netif_tx_queue_stopped(txq);
3377 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3378 struct sk_buff *skb)
3380 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3384 * netif_wake_subqueue - allow sending packets on subqueue
3385 * @dev: network device
3386 * @queue_index: sub queue index
3388 * Resume individual transmit queue of a device with multiple transmit queues.
3390 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3392 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3394 netif_tx_wake_queue(txq);
3398 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3400 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3401 u16 index, bool is_rxqs_map);
3404 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3405 * @j: CPU/Rx queue index
3406 * @mask: bitmask of all cpus/rx queues
3407 * @nr_bits: number of bits in the bitmask
3409 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3411 static inline bool netif_attr_test_mask(unsigned long j,
3412 const unsigned long *mask,
3413 unsigned int nr_bits)
3415 cpu_max_bits_warn(j, nr_bits);
3416 return test_bit(j, mask);
3420 * netif_attr_test_online - Test for online CPU/Rx queue
3421 * @j: CPU/Rx queue index
3422 * @online_mask: bitmask for CPUs/Rx queues that are online
3423 * @nr_bits: number of bits in the bitmask
3425 * Returns true if a CPU/Rx queue is online.
3427 static inline bool netif_attr_test_online(unsigned long j,
3428 const unsigned long *online_mask,
3429 unsigned int nr_bits)
3431 cpu_max_bits_warn(j, nr_bits);
3434 return test_bit(j, online_mask);
3436 return (j < nr_bits);
3440 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3441 * @n: CPU/Rx queue index
3442 * @srcp: the cpumask/Rx queue mask pointer
3443 * @nr_bits: number of bits in the bitmask
3445 * Returns >= nr_bits if no further CPUs/Rx queues set.
3447 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3448 unsigned int nr_bits)
3450 /* -1 is a legal arg here. */
3452 cpu_max_bits_warn(n, nr_bits);
3455 return find_next_bit(srcp, nr_bits, n + 1);
3461 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3462 * @n: CPU/Rx queue index
3463 * @src1p: the first CPUs/Rx queues mask pointer
3464 * @src2p: the second CPUs/Rx queues mask pointer
3465 * @nr_bits: number of bits in the bitmask
3467 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3469 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3470 const unsigned long *src2p,
3471 unsigned int nr_bits)
3473 /* -1 is a legal arg here. */
3475 cpu_max_bits_warn(n, nr_bits);
3478 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3480 return find_next_bit(src1p, nr_bits, n + 1);
3482 return find_next_bit(src2p, nr_bits, n + 1);
3487 static inline int netif_set_xps_queue(struct net_device *dev,
3488 const struct cpumask *mask,
3494 static inline int __netif_set_xps_queue(struct net_device *dev,
3495 const unsigned long *mask,
3496 u16 index, bool is_rxqs_map)
3503 * netif_is_multiqueue - test if device has multiple transmit queues
3504 * @dev: network device
3506 * Check if device has multiple transmit queues
3508 static inline bool netif_is_multiqueue(const struct net_device *dev)
3510 return dev->num_tx_queues > 1;
3513 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3516 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3518 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3521 dev->real_num_rx_queues = rxqs;
3526 static inline struct netdev_rx_queue *
3527 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3529 return dev->_rx + rxq;
3533 static inline unsigned int get_netdev_rx_queue_index(
3534 struct netdev_rx_queue *queue)
3536 struct net_device *dev = queue->dev;
3537 int index = queue - dev->_rx;
3539 BUG_ON(index >= dev->num_rx_queues);
3544 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3545 int netif_get_num_default_rss_queues(void);
3547 enum skb_free_reason {
3548 SKB_REASON_CONSUMED,
3552 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3553 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3556 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3557 * interrupt context or with hardware interrupts being disabled.
3558 * (in_irq() || irqs_disabled())
3560 * We provide four helpers that can be used in following contexts :
3562 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3563 * replacing kfree_skb(skb)
3565 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3566 * Typically used in place of consume_skb(skb) in TX completion path
3568 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3569 * replacing kfree_skb(skb)
3571 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3572 * and consumed a packet. Used in place of consume_skb(skb)
3574 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3576 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3579 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3581 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3584 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3586 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3589 static inline void dev_consume_skb_any(struct sk_buff *skb)
3591 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3594 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3595 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3596 int netif_rx(struct sk_buff *skb);
3597 int netif_rx_ni(struct sk_buff *skb);
3598 int netif_receive_skb(struct sk_buff *skb);
3599 int netif_receive_skb_core(struct sk_buff *skb);
3600 void netif_receive_skb_list(struct list_head *head);
3601 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3602 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3603 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3604 gro_result_t napi_gro_frags(struct napi_struct *napi);
3605 struct packet_offload *gro_find_receive_by_type(__be16 type);
3606 struct packet_offload *gro_find_complete_by_type(__be16 type);
3608 static inline void napi_free_frags(struct napi_struct *napi)
3610 kfree_skb(napi->skb);
3614 bool netdev_is_rx_handler_busy(struct net_device *dev);
3615 int netdev_rx_handler_register(struct net_device *dev,
3616 rx_handler_func_t *rx_handler,
3617 void *rx_handler_data);
3618 void netdev_rx_handler_unregister(struct net_device *dev);
3620 bool dev_valid_name(const char *name);
3621 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3622 bool *need_copyout);
3623 int dev_ifconf(struct net *net, struct ifconf *, int);
3624 int dev_ethtool(struct net *net, struct ifreq *);
3625 unsigned int dev_get_flags(const struct net_device *);
3626 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3627 struct netlink_ext_ack *extack);
3628 int dev_change_flags(struct net_device *dev, unsigned int flags,
3629 struct netlink_ext_ack *extack);
3630 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3631 unsigned int gchanges);
3632 int dev_change_name(struct net_device *, const char *);
3633 int dev_set_alias(struct net_device *, const char *, size_t);
3634 int dev_get_alias(const struct net_device *, char *, size_t);
3635 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3636 int __dev_set_mtu(struct net_device *, int);
3637 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3638 struct netlink_ext_ack *extack);
3639 int dev_set_mtu(struct net_device *, int);
3640 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3641 void dev_set_group(struct net_device *, int);
3642 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3643 struct netlink_ext_ack *extack);
3644 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3645 struct netlink_ext_ack *extack);
3646 int dev_change_carrier(struct net_device *, bool new_carrier);
3647 int dev_get_phys_port_id(struct net_device *dev,
3648 struct netdev_phys_item_id *ppid);
3649 int dev_get_phys_port_name(struct net_device *dev,
3650 char *name, size_t len);
3651 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3652 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3653 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3654 struct netdev_queue *txq, int *ret);
3656 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3657 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3659 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3660 enum bpf_netdev_command cmd);
3661 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3663 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3664 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3665 bool is_skb_forwardable(const struct net_device *dev,
3666 const struct sk_buff *skb);
3668 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3669 struct sk_buff *skb)
3671 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3672 unlikely(!is_skb_forwardable(dev, skb))) {
3673 atomic_long_inc(&dev->rx_dropped);
3678 skb_scrub_packet(skb, true);
3683 bool dev_nit_active(struct net_device *dev);
3684 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3686 extern int netdev_budget;
3687 extern unsigned int netdev_budget_usecs;
3689 /* Called by rtnetlink.c:rtnl_unlock() */
3690 void netdev_run_todo(void);
3693 * dev_put - release reference to device
3694 * @dev: network device
3696 * Release reference to device to allow it to be freed.
3698 static inline void dev_put(struct net_device *dev)
3700 this_cpu_dec(*dev->pcpu_refcnt);
3704 * dev_hold - get reference to device
3705 * @dev: network device
3707 * Hold reference to device to keep it from being freed.
3709 static inline void dev_hold(struct net_device *dev)
3711 this_cpu_inc(*dev->pcpu_refcnt);
3714 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3715 * and _off may be called from IRQ context, but it is caller
3716 * who is responsible for serialization of these calls.
3718 * The name carrier is inappropriate, these functions should really be
3719 * called netif_lowerlayer_*() because they represent the state of any
3720 * kind of lower layer not just hardware media.
3723 void linkwatch_init_dev(struct net_device *dev);
3724 void linkwatch_fire_event(struct net_device *dev);
3725 void linkwatch_forget_dev(struct net_device *dev);
3728 * netif_carrier_ok - test if carrier present
3729 * @dev: network device
3731 * Check if carrier is present on device
3733 static inline bool netif_carrier_ok(const struct net_device *dev)
3735 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3738 unsigned long dev_trans_start(struct net_device *dev);
3740 void __netdev_watchdog_up(struct net_device *dev);
3742 void netif_carrier_on(struct net_device *dev);
3744 void netif_carrier_off(struct net_device *dev);
3747 * netif_dormant_on - mark device as dormant.
3748 * @dev: network device
3750 * Mark device as dormant (as per RFC2863).
3752 * The dormant state indicates that the relevant interface is not
3753 * actually in a condition to pass packets (i.e., it is not 'up') but is
3754 * in a "pending" state, waiting for some external event. For "on-
3755 * demand" interfaces, this new state identifies the situation where the
3756 * interface is waiting for events to place it in the up state.
3758 static inline void netif_dormant_on(struct net_device *dev)
3760 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3761 linkwatch_fire_event(dev);
3765 * netif_dormant_off - set device as not dormant.
3766 * @dev: network device
3768 * Device is not in dormant state.
3770 static inline void netif_dormant_off(struct net_device *dev)
3772 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3773 linkwatch_fire_event(dev);
3777 * netif_dormant - test if device is dormant
3778 * @dev: network device
3780 * Check if device is dormant.
3782 static inline bool netif_dormant(const struct net_device *dev)
3784 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3789 * netif_oper_up - test if device is operational
3790 * @dev: network device
3792 * Check if carrier is operational
3794 static inline bool netif_oper_up(const struct net_device *dev)
3796 return (dev->operstate == IF_OPER_UP ||
3797 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3801 * netif_device_present - is device available or removed
3802 * @dev: network device
3804 * Check if device has not been removed from system.
3806 static inline bool netif_device_present(struct net_device *dev)
3808 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3811 void netif_device_detach(struct net_device *dev);
3813 void netif_device_attach(struct net_device *dev);
3816 * Network interface message level settings
3820 NETIF_MSG_DRV = 0x0001,
3821 NETIF_MSG_PROBE = 0x0002,
3822 NETIF_MSG_LINK = 0x0004,
3823 NETIF_MSG_TIMER = 0x0008,
3824 NETIF_MSG_IFDOWN = 0x0010,
3825 NETIF_MSG_IFUP = 0x0020,
3826 NETIF_MSG_RX_ERR = 0x0040,
3827 NETIF_MSG_TX_ERR = 0x0080,
3828 NETIF_MSG_TX_QUEUED = 0x0100,
3829 NETIF_MSG_INTR = 0x0200,
3830 NETIF_MSG_TX_DONE = 0x0400,
3831 NETIF_MSG_RX_STATUS = 0x0800,
3832 NETIF_MSG_PKTDATA = 0x1000,
3833 NETIF_MSG_HW = 0x2000,
3834 NETIF_MSG_WOL = 0x4000,
3837 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3838 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3839 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3840 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3841 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3842 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3843 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3844 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3845 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3846 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3847 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3848 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3849 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3850 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3851 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3853 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3856 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3857 return default_msg_enable_bits;
3858 if (debug_value == 0) /* no output */
3860 /* set low N bits */
3861 return (1 << debug_value) - 1;
3864 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3866 spin_lock(&txq->_xmit_lock);
3867 txq->xmit_lock_owner = cpu;
3870 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3872 __acquire(&txq->_xmit_lock);
3876 static inline void __netif_tx_release(struct netdev_queue *txq)
3878 __release(&txq->_xmit_lock);
3881 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3883 spin_lock_bh(&txq->_xmit_lock);
3884 txq->xmit_lock_owner = smp_processor_id();
3887 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3889 bool ok = spin_trylock(&txq->_xmit_lock);
3891 txq->xmit_lock_owner = smp_processor_id();
3895 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3897 txq->xmit_lock_owner = -1;
3898 spin_unlock(&txq->_xmit_lock);
3901 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3903 txq->xmit_lock_owner = -1;
3904 spin_unlock_bh(&txq->_xmit_lock);
3907 static inline void txq_trans_update(struct netdev_queue *txq)
3909 if (txq->xmit_lock_owner != -1)
3910 txq->trans_start = jiffies;
3913 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3914 static inline void netif_trans_update(struct net_device *dev)
3916 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3918 if (txq->trans_start != jiffies)
3919 txq->trans_start = jiffies;
3923 * netif_tx_lock - grab network device transmit lock
3924 * @dev: network device
3926 * Get network device transmit lock
3928 static inline void netif_tx_lock(struct net_device *dev)
3933 spin_lock(&dev->tx_global_lock);
3934 cpu = smp_processor_id();
3935 for (i = 0; i < dev->num_tx_queues; i++) {
3936 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3938 /* We are the only thread of execution doing a
3939 * freeze, but we have to grab the _xmit_lock in
3940 * order to synchronize with threads which are in
3941 * the ->hard_start_xmit() handler and already
3942 * checked the frozen bit.
3944 __netif_tx_lock(txq, cpu);
3945 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3946 __netif_tx_unlock(txq);
3950 static inline void netif_tx_lock_bh(struct net_device *dev)
3956 static inline void netif_tx_unlock(struct net_device *dev)
3960 for (i = 0; i < dev->num_tx_queues; i++) {
3961 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3963 /* No need to grab the _xmit_lock here. If the
3964 * queue is not stopped for another reason, we
3967 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3968 netif_schedule_queue(txq);
3970 spin_unlock(&dev->tx_global_lock);
3973 static inline void netif_tx_unlock_bh(struct net_device *dev)
3975 netif_tx_unlock(dev);
3979 #define HARD_TX_LOCK(dev, txq, cpu) { \
3980 if ((dev->features & NETIF_F_LLTX) == 0) { \
3981 __netif_tx_lock(txq, cpu); \
3983 __netif_tx_acquire(txq); \
3987 #define HARD_TX_TRYLOCK(dev, txq) \
3988 (((dev->features & NETIF_F_LLTX) == 0) ? \
3989 __netif_tx_trylock(txq) : \
3990 __netif_tx_acquire(txq))
3992 #define HARD_TX_UNLOCK(dev, txq) { \
3993 if ((dev->features & NETIF_F_LLTX) == 0) { \
3994 __netif_tx_unlock(txq); \
3996 __netif_tx_release(txq); \
4000 static inline void netif_tx_disable(struct net_device *dev)
4006 cpu = smp_processor_id();
4007 for (i = 0; i < dev->num_tx_queues; i++) {
4008 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4010 __netif_tx_lock(txq, cpu);
4011 netif_tx_stop_queue(txq);
4012 __netif_tx_unlock(txq);
4017 static inline void netif_addr_lock(struct net_device *dev)
4019 spin_lock(&dev->addr_list_lock);
4022 static inline void netif_addr_lock_nested(struct net_device *dev)
4024 int subclass = SINGLE_DEPTH_NESTING;
4026 if (dev->netdev_ops->ndo_get_lock_subclass)
4027 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
4029 spin_lock_nested(&dev->addr_list_lock, subclass);
4032 static inline void netif_addr_lock_bh(struct net_device *dev)
4034 spin_lock_bh(&dev->addr_list_lock);
4037 static inline void netif_addr_unlock(struct net_device *dev)
4039 spin_unlock(&dev->addr_list_lock);
4042 static inline void netif_addr_unlock_bh(struct net_device *dev)
4044 spin_unlock_bh(&dev->addr_list_lock);
4048 * dev_addrs walker. Should be used only for read access. Call with
4049 * rcu_read_lock held.
4051 #define for_each_dev_addr(dev, ha) \
4052 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4054 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4056 void ether_setup(struct net_device *dev);
4058 /* Support for loadable net-drivers */
4059 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4060 unsigned char name_assign_type,
4061 void (*setup)(struct net_device *),
4062 unsigned int txqs, unsigned int rxqs);
4063 int dev_get_valid_name(struct net *net, struct net_device *dev,
4066 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4067 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4069 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4070 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4073 int register_netdev(struct net_device *dev);
4074 void unregister_netdev(struct net_device *dev);
4076 /* General hardware address lists handling functions */
4077 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4078 struct netdev_hw_addr_list *from_list, int addr_len);
4079 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4080 struct netdev_hw_addr_list *from_list, int addr_len);
4081 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4082 struct net_device *dev,
4083 int (*sync)(struct net_device *, const unsigned char *),
4084 int (*unsync)(struct net_device *,
4085 const unsigned char *));
4086 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4087 struct net_device *dev,
4088 int (*sync)(struct net_device *,
4089 const unsigned char *, int),
4090 int (*unsync)(struct net_device *,
4091 const unsigned char *, int));
4092 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4093 struct net_device *dev,
4094 int (*unsync)(struct net_device *,
4095 const unsigned char *, int));
4096 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4097 struct net_device *dev,
4098 int (*unsync)(struct net_device *,
4099 const unsigned char *));
4100 void __hw_addr_init(struct netdev_hw_addr_list *list);
4102 /* Functions used for device addresses handling */
4103 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4104 unsigned char addr_type);
4105 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4106 unsigned char addr_type);
4107 void dev_addr_flush(struct net_device *dev);
4108 int dev_addr_init(struct net_device *dev);
4110 /* Functions used for unicast addresses handling */
4111 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4112 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4113 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4114 int dev_uc_sync(struct net_device *to, struct net_device *from);
4115 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4116 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4117 void dev_uc_flush(struct net_device *dev);
4118 void dev_uc_init(struct net_device *dev);
4121 * __dev_uc_sync - Synchonize device's unicast list
4122 * @dev: device to sync
4123 * @sync: function to call if address should be added
4124 * @unsync: function to call if address should be removed
4126 * Add newly added addresses to the interface, and release
4127 * addresses that have been deleted.
4129 static inline int __dev_uc_sync(struct net_device *dev,
4130 int (*sync)(struct net_device *,
4131 const unsigned char *),
4132 int (*unsync)(struct net_device *,
4133 const unsigned char *))
4135 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4139 * __dev_uc_unsync - Remove synchronized addresses from device
4140 * @dev: device to sync
4141 * @unsync: function to call if address should be removed
4143 * Remove all addresses that were added to the device by dev_uc_sync().
4145 static inline void __dev_uc_unsync(struct net_device *dev,
4146 int (*unsync)(struct net_device *,
4147 const unsigned char *))
4149 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4152 /* Functions used for multicast addresses handling */
4153 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4154 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4155 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4156 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4157 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4158 int dev_mc_sync(struct net_device *to, struct net_device *from);
4159 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4160 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4161 void dev_mc_flush(struct net_device *dev);
4162 void dev_mc_init(struct net_device *dev);
4165 * __dev_mc_sync - Synchonize device's multicast list
4166 * @dev: device to sync
4167 * @sync: function to call if address should be added
4168 * @unsync: function to call if address should be removed
4170 * Add newly added addresses to the interface, and release
4171 * addresses that have been deleted.
4173 static inline int __dev_mc_sync(struct net_device *dev,
4174 int (*sync)(struct net_device *,
4175 const unsigned char *),
4176 int (*unsync)(struct net_device *,
4177 const unsigned char *))
4179 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4183 * __dev_mc_unsync - Remove synchronized addresses from device
4184 * @dev: device to sync
4185 * @unsync: function to call if address should be removed
4187 * Remove all addresses that were added to the device by dev_mc_sync().
4189 static inline void __dev_mc_unsync(struct net_device *dev,
4190 int (*unsync)(struct net_device *,
4191 const unsigned char *))
4193 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4196 /* Functions used for secondary unicast and multicast support */
4197 void dev_set_rx_mode(struct net_device *dev);
4198 void __dev_set_rx_mode(struct net_device *dev);
4199 int dev_set_promiscuity(struct net_device *dev, int inc);
4200 int dev_set_allmulti(struct net_device *dev, int inc);
4201 void netdev_state_change(struct net_device *dev);
4202 void netdev_notify_peers(struct net_device *dev);
4203 void netdev_features_change(struct net_device *dev);
4204 /* Load a device via the kmod */
4205 void dev_load(struct net *net, const char *name);
4206 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4207 struct rtnl_link_stats64 *storage);
4208 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4209 const struct net_device_stats *netdev_stats);
4211 extern int netdev_max_backlog;
4212 extern int netdev_tstamp_prequeue;
4213 extern int weight_p;
4214 extern int dev_weight_rx_bias;
4215 extern int dev_weight_tx_bias;
4216 extern int dev_rx_weight;
4217 extern int dev_tx_weight;
4219 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4220 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4221 struct list_head **iter);
4222 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4223 struct list_head **iter);
4225 /* iterate through upper list, must be called under RCU read lock */
4226 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4227 for (iter = &(dev)->adj_list.upper, \
4228 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4230 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4232 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4233 int (*fn)(struct net_device *upper_dev,
4237 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4238 struct net_device *upper_dev);
4240 bool netdev_has_any_upper_dev(struct net_device *dev);
4242 void *netdev_lower_get_next_private(struct net_device *dev,
4243 struct list_head **iter);
4244 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4245 struct list_head **iter);
4247 #define netdev_for_each_lower_private(dev, priv, iter) \
4248 for (iter = (dev)->adj_list.lower.next, \
4249 priv = netdev_lower_get_next_private(dev, &(iter)); \
4251 priv = netdev_lower_get_next_private(dev, &(iter)))
4253 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4254 for (iter = &(dev)->adj_list.lower, \
4255 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4257 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4259 void *netdev_lower_get_next(struct net_device *dev,
4260 struct list_head **iter);
4262 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4263 for (iter = (dev)->adj_list.lower.next, \
4264 ldev = netdev_lower_get_next(dev, &(iter)); \
4266 ldev = netdev_lower_get_next(dev, &(iter)))
4268 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4269 struct list_head **iter);
4270 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4271 struct list_head **iter);
4273 int netdev_walk_all_lower_dev(struct net_device *dev,
4274 int (*fn)(struct net_device *lower_dev,
4277 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4278 int (*fn)(struct net_device *lower_dev,
4282 void *netdev_adjacent_get_private(struct list_head *adj_list);
4283 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4284 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4285 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4286 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4287 struct netlink_ext_ack *extack);
4288 int netdev_master_upper_dev_link(struct net_device *dev,
4289 struct net_device *upper_dev,
4290 void *upper_priv, void *upper_info,
4291 struct netlink_ext_ack *extack);
4292 void netdev_upper_dev_unlink(struct net_device *dev,
4293 struct net_device *upper_dev);
4294 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4295 void *netdev_lower_dev_get_private(struct net_device *dev,
4296 struct net_device *lower_dev);
4297 void netdev_lower_state_changed(struct net_device *lower_dev,
4298 void *lower_state_info);
4300 /* RSS keys are 40 or 52 bytes long */
4301 #define NETDEV_RSS_KEY_LEN 52
4302 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4303 void netdev_rss_key_fill(void *buffer, size_t len);
4305 int dev_get_nest_level(struct net_device *dev);
4306 int skb_checksum_help(struct sk_buff *skb);
4307 int skb_crc32c_csum_help(struct sk_buff *skb);
4308 int skb_csum_hwoffload_help(struct sk_buff *skb,
4309 const netdev_features_t features);
4311 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4312 netdev_features_t features, bool tx_path);
4313 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4314 netdev_features_t features);
4316 struct netdev_bonding_info {
4321 struct netdev_notifier_bonding_info {
4322 struct netdev_notifier_info info; /* must be first */
4323 struct netdev_bonding_info bonding_info;
4326 void netdev_bonding_info_change(struct net_device *dev,
4327 struct netdev_bonding_info *bonding_info);
4330 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4332 return __skb_gso_segment(skb, features, true);
4334 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4336 static inline bool can_checksum_protocol(netdev_features_t features,
4339 if (protocol == htons(ETH_P_FCOE))
4340 return !!(features & NETIF_F_FCOE_CRC);
4342 /* Assume this is an IP checksum (not SCTP CRC) */
4344 if (features & NETIF_F_HW_CSUM) {
4345 /* Can checksum everything */
4350 case htons(ETH_P_IP):
4351 return !!(features & NETIF_F_IP_CSUM);
4352 case htons(ETH_P_IPV6):
4353 return !!(features & NETIF_F_IPV6_CSUM);
4360 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4362 static inline void netdev_rx_csum_fault(struct net_device *dev,
4363 struct sk_buff *skb)
4367 /* rx skb timestamps */
4368 void net_enable_timestamp(void);
4369 void net_disable_timestamp(void);
4371 #ifdef CONFIG_PROC_FS
4372 int __init dev_proc_init(void);
4374 #define dev_proc_init() 0
4377 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4378 struct sk_buff *skb, struct net_device *dev,
4381 skb->xmit_more = more ? 1 : 0;
4382 return ops->ndo_start_xmit(skb, dev);
4385 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4386 struct netdev_queue *txq, bool more)
4388 const struct net_device_ops *ops = dev->netdev_ops;
4391 rc = __netdev_start_xmit(ops, skb, dev, more);
4392 if (rc == NETDEV_TX_OK)
4393 txq_trans_update(txq);
4398 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4400 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4403 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4405 return netdev_class_create_file_ns(class_attr, NULL);
4408 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4410 netdev_class_remove_file_ns(class_attr, NULL);
4413 extern const struct kobj_ns_type_operations net_ns_type_operations;
4415 const char *netdev_drivername(const struct net_device *dev);
4417 void linkwatch_run_queue(void);
4419 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4420 netdev_features_t f2)
4422 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4423 if (f1 & NETIF_F_HW_CSUM)
4424 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4426 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4432 static inline netdev_features_t netdev_get_wanted_features(
4433 struct net_device *dev)
4435 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4437 netdev_features_t netdev_increment_features(netdev_features_t all,
4438 netdev_features_t one, netdev_features_t mask);
4440 /* Allow TSO being used on stacked device :
4441 * Performing the GSO segmentation before last device
4442 * is a performance improvement.
4444 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4445 netdev_features_t mask)
4447 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4450 int __netdev_update_features(struct net_device *dev);
4451 void netdev_update_features(struct net_device *dev);
4452 void netdev_change_features(struct net_device *dev);
4454 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4455 struct net_device *dev);
4457 netdev_features_t passthru_features_check(struct sk_buff *skb,
4458 struct net_device *dev,
4459 netdev_features_t features);
4460 netdev_features_t netif_skb_features(struct sk_buff *skb);
4462 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4464 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4466 /* check flags correspondence */
4467 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4468 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4469 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4470 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4471 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4472 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4473 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4474 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4475 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4476 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4477 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4478 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4479 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4480 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4481 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4482 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4483 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4484 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4486 return (features & feature) == feature;
4489 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4491 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4492 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4495 static inline bool netif_needs_gso(struct sk_buff *skb,
4496 netdev_features_t features)
4498 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4499 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4500 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4503 static inline void netif_set_gso_max_size(struct net_device *dev,
4506 dev->gso_max_size = size;
4509 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4510 int pulled_hlen, u16 mac_offset,
4513 skb->protocol = protocol;
4514 skb->encapsulation = 1;
4515 skb_push(skb, pulled_hlen);
4516 skb_reset_transport_header(skb);
4517 skb->mac_header = mac_offset;
4518 skb->network_header = skb->mac_header + mac_len;
4519 skb->mac_len = mac_len;
4522 static inline bool netif_is_macsec(const struct net_device *dev)
4524 return dev->priv_flags & IFF_MACSEC;
4527 static inline bool netif_is_macvlan(const struct net_device *dev)
4529 return dev->priv_flags & IFF_MACVLAN;
4532 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4534 return dev->priv_flags & IFF_MACVLAN_PORT;
4537 static inline bool netif_is_bond_master(const struct net_device *dev)
4539 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4542 static inline bool netif_is_bond_slave(const struct net_device *dev)
4544 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4547 static inline bool netif_supports_nofcs(struct net_device *dev)
4549 return dev->priv_flags & IFF_SUPP_NOFCS;
4552 static inline bool netif_is_l3_master(const struct net_device *dev)
4554 return dev->priv_flags & IFF_L3MDEV_MASTER;
4557 static inline bool netif_is_l3_slave(const struct net_device *dev)
4559 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4562 static inline bool netif_is_bridge_master(const struct net_device *dev)
4564 return dev->priv_flags & IFF_EBRIDGE;
4567 static inline bool netif_is_bridge_port(const struct net_device *dev)
4569 return dev->priv_flags & IFF_BRIDGE_PORT;
4572 static inline bool netif_is_ovs_master(const struct net_device *dev)
4574 return dev->priv_flags & IFF_OPENVSWITCH;
4577 static inline bool netif_is_ovs_port(const struct net_device *dev)
4579 return dev->priv_flags & IFF_OVS_DATAPATH;
4582 static inline bool netif_is_team_master(const struct net_device *dev)
4584 return dev->priv_flags & IFF_TEAM;
4587 static inline bool netif_is_team_port(const struct net_device *dev)
4589 return dev->priv_flags & IFF_TEAM_PORT;
4592 static inline bool netif_is_lag_master(const struct net_device *dev)
4594 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4597 static inline bool netif_is_lag_port(const struct net_device *dev)
4599 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4602 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4604 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4607 static inline bool netif_is_failover(const struct net_device *dev)
4609 return dev->priv_flags & IFF_FAILOVER;
4612 static inline bool netif_is_failover_slave(const struct net_device *dev)
4614 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4617 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4618 static inline void netif_keep_dst(struct net_device *dev)
4620 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4623 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4624 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4626 /* TODO: reserve and use an additional IFF bit, if we get more users */
4627 return dev->priv_flags & IFF_MACSEC;
4630 extern struct pernet_operations __net_initdata loopback_net_ops;
4632 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4634 /* netdev_printk helpers, similar to dev_printk */
4636 static inline const char *netdev_name(const struct net_device *dev)
4638 if (!dev->name[0] || strchr(dev->name, '%'))
4639 return "(unnamed net_device)";
4643 static inline bool netdev_unregistering(const struct net_device *dev)
4645 return dev->reg_state == NETREG_UNREGISTERING;
4648 static inline const char *netdev_reg_state(const struct net_device *dev)
4650 switch (dev->reg_state) {
4651 case NETREG_UNINITIALIZED: return " (uninitialized)";
4652 case NETREG_REGISTERED: return "";
4653 case NETREG_UNREGISTERING: return " (unregistering)";
4654 case NETREG_UNREGISTERED: return " (unregistered)";
4655 case NETREG_RELEASED: return " (released)";
4656 case NETREG_DUMMY: return " (dummy)";
4659 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4660 return " (unknown)";
4664 void netdev_printk(const char *level, const struct net_device *dev,
4665 const char *format, ...);
4667 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4669 void netdev_alert(const struct net_device *dev, const char *format, ...);
4671 void netdev_crit(const struct net_device *dev, const char *format, ...);
4673 void netdev_err(const struct net_device *dev, const char *format, ...);
4675 void netdev_warn(const struct net_device *dev, const char *format, ...);
4677 void netdev_notice(const struct net_device *dev, const char *format, ...);
4679 void netdev_info(const struct net_device *dev, const char *format, ...);
4681 #define netdev_level_once(level, dev, fmt, ...) \
4683 static bool __print_once __read_mostly; \
4685 if (!__print_once) { \
4686 __print_once = true; \
4687 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4691 #define netdev_emerg_once(dev, fmt, ...) \
4692 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4693 #define netdev_alert_once(dev, fmt, ...) \
4694 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4695 #define netdev_crit_once(dev, fmt, ...) \
4696 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4697 #define netdev_err_once(dev, fmt, ...) \
4698 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4699 #define netdev_warn_once(dev, fmt, ...) \
4700 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4701 #define netdev_notice_once(dev, fmt, ...) \
4702 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4703 #define netdev_info_once(dev, fmt, ...) \
4704 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4706 #define MODULE_ALIAS_NETDEV(device) \
4707 MODULE_ALIAS("netdev-" device)
4709 #if defined(CONFIG_DYNAMIC_DEBUG)
4710 #define netdev_dbg(__dev, format, args...) \
4712 dynamic_netdev_dbg(__dev, format, ##args); \
4714 #elif defined(DEBUG)
4715 #define netdev_dbg(__dev, format, args...) \
4716 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4718 #define netdev_dbg(__dev, format, args...) \
4721 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4725 #if defined(VERBOSE_DEBUG)
4726 #define netdev_vdbg netdev_dbg
4729 #define netdev_vdbg(dev, format, args...) \
4732 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4738 * netdev_WARN() acts like dev_printk(), but with the key difference
4739 * of using a WARN/WARN_ON to get the message out, including the
4740 * file/line information and a backtrace.
4742 #define netdev_WARN(dev, format, args...) \
4743 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4744 netdev_reg_state(dev), ##args)
4746 #define netdev_WARN_ONCE(dev, format, args...) \
4747 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4748 netdev_reg_state(dev), ##args)
4750 /* netif printk helpers, similar to netdev_printk */
4752 #define netif_printk(priv, type, level, dev, fmt, args...) \
4754 if (netif_msg_##type(priv)) \
4755 netdev_printk(level, (dev), fmt, ##args); \
4758 #define netif_level(level, priv, type, dev, fmt, args...) \
4760 if (netif_msg_##type(priv)) \
4761 netdev_##level(dev, fmt, ##args); \
4764 #define netif_emerg(priv, type, dev, fmt, args...) \
4765 netif_level(emerg, priv, type, dev, fmt, ##args)
4766 #define netif_alert(priv, type, dev, fmt, args...) \
4767 netif_level(alert, priv, type, dev, fmt, ##args)
4768 #define netif_crit(priv, type, dev, fmt, args...) \
4769 netif_level(crit, priv, type, dev, fmt, ##args)
4770 #define netif_err(priv, type, dev, fmt, args...) \
4771 netif_level(err, priv, type, dev, fmt, ##args)
4772 #define netif_warn(priv, type, dev, fmt, args...) \
4773 netif_level(warn, priv, type, dev, fmt, ##args)
4774 #define netif_notice(priv, type, dev, fmt, args...) \
4775 netif_level(notice, priv, type, dev, fmt, ##args)
4776 #define netif_info(priv, type, dev, fmt, args...) \
4777 netif_level(info, priv, type, dev, fmt, ##args)
4779 #if defined(CONFIG_DYNAMIC_DEBUG)
4780 #define netif_dbg(priv, type, netdev, format, args...) \
4782 if (netif_msg_##type(priv)) \
4783 dynamic_netdev_dbg(netdev, format, ##args); \
4785 #elif defined(DEBUG)
4786 #define netif_dbg(priv, type, dev, format, args...) \
4787 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4789 #define netif_dbg(priv, type, dev, format, args...) \
4792 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4797 /* if @cond then downgrade to debug, else print at @level */
4798 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4801 netif_dbg(priv, type, netdev, fmt, ##args); \
4803 netif_ ## level(priv, type, netdev, fmt, ##args); \
4806 #if defined(VERBOSE_DEBUG)
4807 #define netif_vdbg netif_dbg
4809 #define netif_vdbg(priv, type, dev, format, args...) \
4812 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4818 * The list of packet types we will receive (as opposed to discard)
4819 * and the routines to invoke.
4821 * Why 16. Because with 16 the only overlap we get on a hash of the
4822 * low nibble of the protocol value is RARP/SNAP/X.25.
4836 #define PTYPE_HASH_SIZE (16)
4837 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4839 #endif /* _LINUX_NETDEVICE_H */