2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
55 #include <linux/hashtable.h>
62 /* 802.15.4 specific */
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
69 void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
72 /* Backlog congestion levels */
73 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74 #define NET_RX_DROP 1 /* packet dropped */
77 * Transmit return codes: transmit return codes originate from three different
80 * - qdisc return codes
81 * - driver transmit return codes
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
90 * others are propagated to higher layers.
93 /* qdisc ->enqueue() return codes. */
94 #define NET_XMIT_SUCCESS 0x00
95 #define NET_XMIT_DROP 0x01 /* skb dropped */
96 #define NET_XMIT_CN 0x02 /* congestion notification */
97 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
99 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
102 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
103 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105 /* Driver transmit return codes */
106 #define NETDEV_TX_MASK 0xf0
109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
113 typedef enum netdev_tx netdev_tx_t;
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 static inline bool dev_xmit_complete(int rc)
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 if (likely(rc < NET_XMIT_MASK))
134 * Compute the worst-case header length according to the protocols
138 #if defined(CONFIG_HYPERV_NET)
139 # define LL_MAX_HEADER 128
140 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
141 # if defined(CONFIG_MAC80211_MESH)
142 # define LL_MAX_HEADER 128
144 # define LL_MAX_HEADER 96
147 # define LL_MAX_HEADER 32
150 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
152 #define MAX_HEADER LL_MAX_HEADER
154 #define MAX_HEADER (LL_MAX_HEADER + 48)
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
162 struct net_device_stats {
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
172 unsigned long collisions;
173 unsigned long rx_length_errors;
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
189 #include <linux/cache.h>
190 #include <linux/skbuff.h>
193 #include <linux/static_key.h>
194 extern struct static_key rps_needed;
201 struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
205 #define NETDEV_HW_ADDR_T_LAN 1
206 #define NETDEV_HW_ADDR_T_SAN 2
207 #define NETDEV_HW_ADDR_T_SLAVE 3
208 #define NETDEV_HW_ADDR_T_UNICAST 4
209 #define NETDEV_HW_ADDR_T_MULTICAST 5
214 struct rcu_head rcu_head;
217 struct netdev_hw_addr_list {
218 struct list_head list;
222 #define netdev_hw_addr_list_count(l) ((l)->count)
223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224 #define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229 #define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234 #define netdev_for_each_mc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
242 /* cached hardware header; allow for machine alignment needs. */
243 #define HH_DATA_MOD 16
244 #define HH_DATA_OFF(__len) \
245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
246 #define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
259 #define LL_RESERVED_SPACE(dev) \
260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
261 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
267 const void *saddr, unsigned int len);
268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
270 void (*cache_update)(struct hh_cache *hh,
271 const struct net_device *dev,
272 const unsigned char *haddr);
273 bool (*validate)(const char *ll_header, unsigned int len);
276 /* These flag bits are private to the generic network queueing
277 * layer; they may not be explicitly referenced by any other
281 enum netdev_state_t {
283 __LINK_STATE_PRESENT,
284 __LINK_STATE_NOCARRIER,
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
291 * This structure holds boot-time configured netdevice settings. They
292 * are then used in the device probing.
294 struct netdev_boot_setup {
298 #define NETDEV_BOOT_SETUP_MAX 8
300 int __init netdev_boot_setup(char *str);
303 * Structure for NAPI scheduling similar to tasklet but with weighting
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
309 * to the per-CPU poll_list, and whoever clears that bit
310 * can remove from the list right before clearing the bit.
312 struct list_head poll_list;
316 unsigned int gro_count;
317 int (*poll)(struct napi_struct *, int);
318 #ifdef CONFIG_NETPOLL
319 spinlock_t poll_lock;
322 struct net_device *dev;
323 struct sk_buff *gro_list;
325 struct hrtimer timer;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_DISABLE, /* Disable pending */
334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
346 typedef enum gro_result gro_result_t;
349 * enum rx_handler_result - Possible return values for rx_handlers.
350 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
352 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
353 * case skb->dev was changed by rx_handler.
354 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
355 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
357 * rx_handlers are functions called from inside __netif_receive_skb(), to do
358 * special processing of the skb, prior to delivery to protocol handlers.
360 * Currently, a net_device can only have a single rx_handler registered. Trying
361 * to register a second rx_handler will return -EBUSY.
363 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
364 * To unregister a rx_handler on a net_device, use
365 * netdev_rx_handler_unregister().
367 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
370 * If the rx_handler consumed the skb in some way, it should return
371 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
372 * the skb to be delivered in some other way.
374 * If the rx_handler changed skb->dev, to divert the skb to another
375 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
376 * new device will be called if it exists.
378 * If the rx_handler decides the skb should be ignored, it should return
379 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
380 * are registered on exact device (ptype->dev == skb->dev).
382 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
383 * delivered, it should return RX_HANDLER_PASS.
385 * A device without a registered rx_handler will behave as if rx_handler
386 * returned RX_HANDLER_PASS.
389 enum rx_handler_result {
395 typedef enum rx_handler_result rx_handler_result_t;
396 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
398 void __napi_schedule(struct napi_struct *n);
399 void __napi_schedule_irqoff(struct napi_struct *n);
401 static inline bool napi_disable_pending(struct napi_struct *n)
403 return test_bit(NAPI_STATE_DISABLE, &n->state);
407 * napi_schedule_prep - check if NAPI can be scheduled
410 * Test if NAPI routine is already running, and if not mark
411 * it as running. This is used as a condition variable to
412 * insure only one NAPI poll instance runs. We also make
413 * sure there is no pending NAPI disable.
415 static inline bool napi_schedule_prep(struct napi_struct *n)
417 return !napi_disable_pending(n) &&
418 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
422 * napi_schedule - schedule NAPI poll
425 * Schedule NAPI poll routine to be called if it is not already
428 static inline void napi_schedule(struct napi_struct *n)
430 if (napi_schedule_prep(n))
435 * napi_schedule_irqoff - schedule NAPI poll
438 * Variant of napi_schedule(), assuming hard irqs are masked.
440 static inline void napi_schedule_irqoff(struct napi_struct *n)
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
446 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
447 static inline bool napi_reschedule(struct napi_struct *napi)
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
456 void __napi_complete(struct napi_struct *n);
457 void napi_complete_done(struct napi_struct *n, int work_done);
459 * napi_complete - NAPI processing complete
462 * Mark NAPI processing as complete.
463 * Consider using napi_complete_done() instead.
465 static inline void napi_complete(struct napi_struct *n)
467 return napi_complete_done(n, 0);
471 * napi_hash_add - add a NAPI to global hashtable
472 * @napi: NAPI context
474 * Generate a new napi_id and store a @napi under it in napi_hash.
475 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
476 * Note: This is normally automatically done from netif_napi_add(),
477 * so might disappear in a future Linux version.
479 void napi_hash_add(struct napi_struct *napi);
482 * napi_hash_del - remove a NAPI from global table
483 * @napi: NAPI context
485 * Warning: caller must observe RCU grace period
486 * before freeing memory containing @napi, if
487 * this function returns true.
488 * Note: core networking stack automatically calls it
489 * from netif_napi_del().
490 * Drivers might want to call this helper to combine all
491 * the needed RCU grace periods into a single one.
493 bool napi_hash_del(struct napi_struct *napi);
496 * napi_disable - prevent NAPI from scheduling
499 * Stop NAPI from being scheduled on this context.
500 * Waits till any outstanding processing completes.
502 void napi_disable(struct napi_struct *n);
505 * napi_enable - enable NAPI scheduling
508 * Resume NAPI from being scheduled on this context.
509 * Must be paired with napi_disable.
511 static inline void napi_enable(struct napi_struct *n)
513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
514 smp_mb__before_atomic();
515 clear_bit(NAPI_STATE_SCHED, &n->state);
516 clear_bit(NAPI_STATE_NPSVC, &n->state);
520 * napi_synchronize - wait until NAPI is not running
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
527 static inline void napi_synchronize(const struct napi_struct *n)
529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
536 enum netdev_queue_state_t {
537 __QUEUE_STATE_DRV_XOFF,
538 __QUEUE_STATE_STACK_XOFF,
539 __QUEUE_STATE_FROZEN,
542 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
543 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
544 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
546 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
549 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
554 * netif_tx_* functions below are used to manipulate this flag. The
555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556 * queue independently. The netif_xmit_*stopped functions below are called
557 * to check if the queue has been stopped by the driver or stack (either
558 * of the XOFF bits are set in the state). Drivers should not need to call
559 * netif_xmit*stopped functions, they should only be using netif_tx_*.
562 struct netdev_queue {
566 struct net_device *dev;
567 struct Qdisc __rcu *qdisc;
568 struct Qdisc *qdisc_sleeping;
572 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
575 unsigned long tx_maxrate;
577 * Number of TX timeouts for this queue
578 * (/sys/class/net/DEV/Q/trans_timeout)
580 unsigned long trans_timeout;
584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
587 * Time (in jiffies) of last Tx
589 unsigned long trans_start;
596 } ____cacheline_aligned_in_smp;
598 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
607 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
609 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
624 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628 * tail pointer for that CPU's input queue at the time of last enqueue, and
629 * a hardware filter index.
631 struct rps_dev_flow {
634 unsigned int last_qtail;
636 #define RPS_NO_FILTER 0xffff
639 * The rps_dev_flow_table structure contains a table of flow mappings.
641 struct rps_dev_flow_table {
644 struct rps_dev_flow flows[0];
646 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
647 ((_num) * sizeof(struct rps_dev_flow)))
650 * The rps_sock_flow_table contains mappings of flows to the last CPU
651 * on which they were processed by the application (set in recvmsg).
652 * Each entry is a 32bit value. Upper part is the high-order bits
653 * of flow hash, lower part is CPU number.
654 * rps_cpu_mask is used to partition the space, depending on number of
655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
657 * meaning we use 32-6=26 bits for the hash.
659 struct rps_sock_flow_table {
662 u32 ents[0] ____cacheline_aligned_in_smp;
664 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
666 #define RPS_NO_CPU 0xffff
668 extern u32 rps_cpu_mask;
669 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
671 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
678 /* We only give a hint, preemption can change CPU under us */
679 val |= raw_smp_processor_id();
681 if (table->ents[index] != val)
682 table->ents[index] = val;
686 #ifdef CONFIG_RFS_ACCEL
687 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
690 #endif /* CONFIG_RPS */
692 /* This structure contains an instance of an RX queue. */
693 struct netdev_rx_queue {
695 struct rps_map __rcu *rps_map;
696 struct rps_dev_flow_table __rcu *rps_flow_table;
699 struct net_device *dev;
700 } ____cacheline_aligned_in_smp;
703 * RX queue sysfs structures and functions.
705 struct rx_queue_attribute {
706 struct attribute attr;
707 ssize_t (*show)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, char *buf);
709 ssize_t (*store)(struct netdev_rx_queue *queue,
710 struct rx_queue_attribute *attr, const char *buf, size_t len);
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
720 unsigned int alloc_len;
724 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
725 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726 - sizeof(struct xps_map)) / sizeof(u16))
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
731 struct xps_dev_maps {
733 struct xps_map __rcu *cpu_map[0];
735 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
736 (nr_cpu_ids * sizeof(struct xps_map *)))
737 #endif /* CONFIG_XPS */
739 #define TC_MAX_QUEUE 16
740 #define TC_BITMASK 15
741 /* HW offloaded queuing disciplines txq count and offset maps */
742 struct netdev_tc_txq {
747 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
752 struct netdev_fcoe_hbainfo {
753 char manufacturer[64];
754 char serial_number[64];
755 char hardware_version[64];
756 char driver_version[64];
757 char optionrom_version[64];
758 char firmware_version[64];
760 char model_description[256];
764 #define MAX_PHYS_ITEM_ID_LEN 32
766 /* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
769 struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
771 unsigned char id_len;
774 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
781 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
784 /* These structures hold the attributes of qdisc and classifiers
785 * that are being passed to the netdevice through the setup_tc op.
795 struct tc_cls_u32_offload;
797 struct tc_to_netdev {
801 struct tc_cls_u32_offload *cls_u32;
802 struct tc_cls_flower_offload *cls_flower;
803 struct tc_cls_matchall_offload *cls_mall;
804 struct tc_cls_bpf_offload *cls_bpf;
808 /* These structures hold the attributes of xdp state that are being passed
809 * to the netdevice through the xdp op.
811 enum xdp_netdev_command {
812 /* Set or clear a bpf program used in the earliest stages of packet
813 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
814 * is responsible for calling bpf_prog_put on any old progs that are
815 * stored. In case of error, the callee need not release the new prog
816 * reference, but on success it takes ownership and must bpf_prog_put
817 * when it is no longer used.
820 /* Check if a bpf program is set on the device. The callee should
821 * return true if a program is currently attached and running.
827 enum xdp_netdev_command command;
830 struct bpf_prog *prog;
837 * This structure defines the management hooks for network devices.
838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
841 * int (*ndo_init)(struct net_device *dev);
842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
851 * int (*ndo_open)(struct net_device *dev);
852 * This function is called when a network device transitions to the up
855 * int (*ndo_stop)(struct net_device *dev);
856 * This function is called when a network device transitions to the down
859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
861 * Called when a packet needs to be transmitted.
862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
866 * Required; cannot be NULL.
868 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
869 * netdev_features_t features);
870 * Adjusts the requested feature flags according to device-specific
871 * constraints, and returns the resulting flags. Must not modify
874 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
875 * void *accel_priv, select_queue_fallback_t fallback);
876 * Called to decide which queue to use when device supports multiple
879 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
880 * This function is called to allow device receiver to make
881 * changes to configuration when multicast or promiscuous is enabled.
883 * void (*ndo_set_rx_mode)(struct net_device *dev);
884 * This function is called device changes address list filtering.
885 * If driver handles unicast address filtering, it should set
886 * IFF_UNICAST_FLT in its priv_flags.
888 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
889 * This function is called when the Media Access Control address
890 * needs to be changed. If this interface is not defined, the
891 * MAC address can not be changed.
893 * int (*ndo_validate_addr)(struct net_device *dev);
894 * Test if Media Access Control address is valid for the device.
896 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
897 * Called when a user requests an ioctl which can't be handled by
898 * the generic interface code. If not defined ioctls return
899 * not supported error code.
901 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
902 * Used to set network devices bus interface parameters. This interface
903 * is retained for legacy reasons; new devices should use the bus
904 * interface (PCI) for low level management.
906 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
907 * Called when a user wants to change the Maximum Transfer Unit
908 * of a device. If not defined, any request to change MTU will
909 * will return an error.
911 * void (*ndo_tx_timeout)(struct net_device *dev);
912 * Callback used when the transmitter has not made any progress
913 * for dev->watchdog ticks.
915 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
916 * struct rtnl_link_stats64 *storage);
917 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
918 * Called when a user wants to get the network device usage
919 * statistics. Drivers must do one of the following:
920 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
921 * rtnl_link_stats64 structure passed by the caller.
922 * 2. Define @ndo_get_stats to update a net_device_stats structure
923 * (which should normally be dev->stats) and return a pointer to
924 * it. The structure may be changed asynchronously only if each
925 * field is written atomically.
926 * 3. Update dev->stats asynchronously and atomically, and define
929 * bool (*ndo_has_offload_stats)(int attr_id)
930 * Return true if this device supports offload stats of this attr_id.
932 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
934 * Get statistics for offload operations by attr_id. Write it into the
937 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
938 * If device supports VLAN filtering this function is called when a
939 * VLAN id is registered.
941 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
942 * If device supports VLAN filtering this function is called when a
943 * VLAN id is unregistered.
945 * void (*ndo_poll_controller)(struct net_device *dev);
947 * SR-IOV management functions.
948 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
949 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
950 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
952 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
953 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
954 * int (*ndo_get_vf_config)(struct net_device *dev,
955 * int vf, struct ifla_vf_info *ivf);
956 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
957 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
958 * struct nlattr *port[]);
960 * Enable or disable the VF ability to query its RSS Redirection Table and
961 * Hash Key. This is needed since on some devices VF share this information
962 * with PF and querying it may introduce a theoretical security risk.
963 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
964 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
965 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
966 * Called to setup 'tc' number of traffic classes in the net device. This
967 * is always called from the stack with the rtnl lock held and netif tx
968 * queues stopped. This allows the netdevice to perform queue management
971 * Fiber Channel over Ethernet (FCoE) offload functions.
972 * int (*ndo_fcoe_enable)(struct net_device *dev);
973 * Called when the FCoE protocol stack wants to start using LLD for FCoE
974 * so the underlying device can perform whatever needed configuration or
975 * initialization to support acceleration of FCoE traffic.
977 * int (*ndo_fcoe_disable)(struct net_device *dev);
978 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
979 * so the underlying device can perform whatever needed clean-ups to
980 * stop supporting acceleration of FCoE traffic.
982 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
983 * struct scatterlist *sgl, unsigned int sgc);
984 * Called when the FCoE Initiator wants to initialize an I/O that
985 * is a possible candidate for Direct Data Placement (DDP). The LLD can
986 * perform necessary setup and returns 1 to indicate the device is set up
987 * successfully to perform DDP on this I/O, otherwise this returns 0.
989 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
990 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
991 * indicated by the FC exchange id 'xid', so the underlying device can
992 * clean up and reuse resources for later DDP requests.
994 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
995 * struct scatterlist *sgl, unsigned int sgc);
996 * Called when the FCoE Target wants to initialize an I/O that
997 * is a possible candidate for Direct Data Placement (DDP). The LLD can
998 * perform necessary setup and returns 1 to indicate the device is set up
999 * successfully to perform DDP on this I/O, otherwise this returns 0.
1001 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1002 * struct netdev_fcoe_hbainfo *hbainfo);
1003 * Called when the FCoE Protocol stack wants information on the underlying
1004 * device. This information is utilized by the FCoE protocol stack to
1005 * register attributes with Fiber Channel management service as per the
1006 * FC-GS Fabric Device Management Information(FDMI) specification.
1008 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1009 * Called when the underlying device wants to override default World Wide
1010 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1011 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1012 * protocol stack to use.
1015 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1016 * u16 rxq_index, u32 flow_id);
1017 * Set hardware filter for RFS. rxq_index is the target queue index;
1018 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1019 * Return the filter ID on success, or a negative error code.
1021 * Slave management functions (for bridge, bonding, etc).
1022 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1023 * Called to make another netdev an underling.
1025 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1026 * Called to release previously enslaved netdev.
1028 * Feature/offload setting functions.
1029 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1030 * Called to update device configuration to new features. Passed
1031 * feature set might be less than what was returned by ndo_fix_features()).
1032 * Must return >0 or -errno if it changed dev->features itself.
1034 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1035 * struct net_device *dev,
1036 * const unsigned char *addr, u16 vid, u16 flags)
1037 * Adds an FDB entry to dev for addr.
1038 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1039 * struct net_device *dev,
1040 * const unsigned char *addr, u16 vid)
1041 * Deletes the FDB entry from dev coresponding to addr.
1042 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1043 * struct net_device *dev, struct net_device *filter_dev,
1045 * Used to add FDB entries to dump requests. Implementers should add
1046 * entries to skb and update idx with the number of entries.
1048 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1050 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1051 * struct net_device *dev, u32 filter_mask,
1053 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1056 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1057 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1058 * which do not represent real hardware may define this to allow their
1059 * userspace components to manage their virtual carrier state. Devices
1060 * that determine carrier state from physical hardware properties (eg
1061 * network cables) or protocol-dependent mechanisms (eg
1062 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1064 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1065 * struct netdev_phys_item_id *ppid);
1066 * Called to get ID of physical port of this device. If driver does
1067 * not implement this, it is assumed that the hw is not able to have
1068 * multiple net devices on single physical port.
1070 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1071 * struct udp_tunnel_info *ti);
1072 * Called by UDP tunnel to notify a driver about the UDP port and socket
1073 * address family that a UDP tunnel is listnening to. It is called only
1074 * when a new port starts listening. The operation is protected by the
1077 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1078 * struct udp_tunnel_info *ti);
1079 * Called by UDP tunnel to notify the driver about a UDP port and socket
1080 * address family that the UDP tunnel is not listening to anymore. The
1081 * operation is protected by the RTNL.
1083 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1084 * struct net_device *dev)
1085 * Called by upper layer devices to accelerate switching or other
1086 * station functionality into hardware. 'pdev is the lowerdev
1087 * to use for the offload and 'dev' is the net device that will
1088 * back the offload. Returns a pointer to the private structure
1089 * the upper layer will maintain.
1090 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1091 * Called by upper layer device to delete the station created
1092 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1093 * the station and priv is the structure returned by the add
1095 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1096 * struct net_device *dev,
1098 * Callback to use for xmit over the accelerated station. This
1099 * is used in place of ndo_start_xmit on accelerated net
1101 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1102 * struct net_device *dev
1103 * netdev_features_t features);
1104 * Called by core transmit path to determine if device is capable of
1105 * performing offload operations on a given packet. This is to give
1106 * the device an opportunity to implement any restrictions that cannot
1107 * be otherwise expressed by feature flags. The check is called with
1108 * the set of features that the stack has calculated and it returns
1109 * those the driver believes to be appropriate.
1110 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1111 * int queue_index, u32 maxrate);
1112 * Called when a user wants to set a max-rate limitation of specific
1114 * int (*ndo_get_iflink)(const struct net_device *dev);
1115 * Called to get the iflink value of this device.
1116 * void (*ndo_change_proto_down)(struct net_device *dev,
1118 * This function is used to pass protocol port error state information
1119 * to the switch driver. The switch driver can react to the proto_down
1120 * by doing a phys down on the associated switch port.
1121 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1122 * This function is used to get egress tunnel information for given skb.
1123 * This is useful for retrieving outer tunnel header parameters while
1125 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1126 * This function is used to specify the headroom that the skb must
1127 * consider when allocation skb during packet reception. Setting
1128 * appropriate rx headroom value allows avoiding skb head copy on
1129 * forward. Setting a negative value resets the rx headroom to the
1131 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1132 * This function is used to set or query state related to XDP on the
1133 * netdevice. See definition of enum xdp_netdev_command for details.
1136 struct net_device_ops {
1137 int (*ndo_init)(struct net_device *dev);
1138 void (*ndo_uninit)(struct net_device *dev);
1139 int (*ndo_open)(struct net_device *dev);
1140 int (*ndo_stop)(struct net_device *dev);
1141 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1142 struct net_device *dev);
1143 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1144 struct net_device *dev,
1145 netdev_features_t features);
1146 u16 (*ndo_select_queue)(struct net_device *dev,
1147 struct sk_buff *skb,
1149 select_queue_fallback_t fallback);
1150 void (*ndo_change_rx_flags)(struct net_device *dev,
1152 void (*ndo_set_rx_mode)(struct net_device *dev);
1153 int (*ndo_set_mac_address)(struct net_device *dev,
1155 int (*ndo_validate_addr)(struct net_device *dev);
1156 int (*ndo_do_ioctl)(struct net_device *dev,
1157 struct ifreq *ifr, int cmd);
1158 int (*ndo_set_config)(struct net_device *dev,
1160 int (*ndo_change_mtu)(struct net_device *dev,
1162 int (*ndo_neigh_setup)(struct net_device *dev,
1163 struct neigh_parms *);
1164 void (*ndo_tx_timeout) (struct net_device *dev);
1166 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1167 struct rtnl_link_stats64 *storage);
1168 bool (*ndo_has_offload_stats)(int attr_id);
1169 int (*ndo_get_offload_stats)(int attr_id,
1170 const struct net_device *dev,
1172 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1174 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1175 __be16 proto, u16 vid);
1176 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1177 __be16 proto, u16 vid);
1178 #ifdef CONFIG_NET_POLL_CONTROLLER
1179 void (*ndo_poll_controller)(struct net_device *dev);
1180 int (*ndo_netpoll_setup)(struct net_device *dev,
1181 struct netpoll_info *info);
1182 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1184 #ifdef CONFIG_NET_RX_BUSY_POLL
1185 int (*ndo_busy_poll)(struct napi_struct *dev);
1187 int (*ndo_set_vf_mac)(struct net_device *dev,
1188 int queue, u8 *mac);
1189 int (*ndo_set_vf_vlan)(struct net_device *dev,
1190 int queue, u16 vlan, u8 qos);
1191 int (*ndo_set_vf_rate)(struct net_device *dev,
1192 int vf, int min_tx_rate,
1194 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1195 int vf, bool setting);
1196 int (*ndo_set_vf_trust)(struct net_device *dev,
1197 int vf, bool setting);
1198 int (*ndo_get_vf_config)(struct net_device *dev,
1200 struct ifla_vf_info *ivf);
1201 int (*ndo_set_vf_link_state)(struct net_device *dev,
1202 int vf, int link_state);
1203 int (*ndo_get_vf_stats)(struct net_device *dev,
1205 struct ifla_vf_stats
1207 int (*ndo_set_vf_port)(struct net_device *dev,
1209 struct nlattr *port[]);
1210 int (*ndo_get_vf_port)(struct net_device *dev,
1211 int vf, struct sk_buff *skb);
1212 int (*ndo_set_vf_guid)(struct net_device *dev,
1215 int (*ndo_set_vf_rss_query_en)(
1216 struct net_device *dev,
1217 int vf, bool setting);
1218 int (*ndo_setup_tc)(struct net_device *dev,
1221 struct tc_to_netdev *tc);
1222 #if IS_ENABLED(CONFIG_FCOE)
1223 int (*ndo_fcoe_enable)(struct net_device *dev);
1224 int (*ndo_fcoe_disable)(struct net_device *dev);
1225 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1227 struct scatterlist *sgl,
1229 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1231 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1233 struct scatterlist *sgl,
1235 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1236 struct netdev_fcoe_hbainfo *hbainfo);
1239 #if IS_ENABLED(CONFIG_LIBFCOE)
1240 #define NETDEV_FCOE_WWNN 0
1241 #define NETDEV_FCOE_WWPN 1
1242 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1243 u64 *wwn, int type);
1246 #ifdef CONFIG_RFS_ACCEL
1247 int (*ndo_rx_flow_steer)(struct net_device *dev,
1248 const struct sk_buff *skb,
1252 int (*ndo_add_slave)(struct net_device *dev,
1253 struct net_device *slave_dev);
1254 int (*ndo_del_slave)(struct net_device *dev,
1255 struct net_device *slave_dev);
1256 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1257 netdev_features_t features);
1258 int (*ndo_set_features)(struct net_device *dev,
1259 netdev_features_t features);
1260 int (*ndo_neigh_construct)(struct net_device *dev,
1261 struct neighbour *n);
1262 void (*ndo_neigh_destroy)(struct net_device *dev,
1263 struct neighbour *n);
1265 int (*ndo_fdb_add)(struct ndmsg *ndm,
1266 struct nlattr *tb[],
1267 struct net_device *dev,
1268 const unsigned char *addr,
1271 int (*ndo_fdb_del)(struct ndmsg *ndm,
1272 struct nlattr *tb[],
1273 struct net_device *dev,
1274 const unsigned char *addr,
1276 int (*ndo_fdb_dump)(struct sk_buff *skb,
1277 struct netlink_callback *cb,
1278 struct net_device *dev,
1279 struct net_device *filter_dev,
1282 int (*ndo_bridge_setlink)(struct net_device *dev,
1283 struct nlmsghdr *nlh,
1285 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1287 struct net_device *dev,
1290 int (*ndo_bridge_dellink)(struct net_device *dev,
1291 struct nlmsghdr *nlh,
1293 int (*ndo_change_carrier)(struct net_device *dev,
1295 int (*ndo_get_phys_port_id)(struct net_device *dev,
1296 struct netdev_phys_item_id *ppid);
1297 int (*ndo_get_phys_port_name)(struct net_device *dev,
1298 char *name, size_t len);
1299 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1300 struct udp_tunnel_info *ti);
1301 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1302 struct udp_tunnel_info *ti);
1303 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1304 struct net_device *dev);
1305 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1308 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1309 struct net_device *dev,
1311 int (*ndo_get_lock_subclass)(struct net_device *dev);
1312 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1315 int (*ndo_get_iflink)(const struct net_device *dev);
1316 int (*ndo_change_proto_down)(struct net_device *dev,
1318 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1319 struct sk_buff *skb);
1320 void (*ndo_set_rx_headroom)(struct net_device *dev,
1321 int needed_headroom);
1322 int (*ndo_xdp)(struct net_device *dev,
1323 struct netdev_xdp *xdp);
1327 * enum net_device_priv_flags - &struct net_device priv_flags
1329 * These are the &struct net_device, they are only set internally
1330 * by drivers and used in the kernel. These flags are invisible to
1331 * userspace; this means that the order of these flags can change
1332 * during any kernel release.
1334 * You should have a pretty good reason to be extending these flags.
1336 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1337 * @IFF_EBRIDGE: Ethernet bridging device
1338 * @IFF_BONDING: bonding master or slave
1339 * @IFF_ISATAP: ISATAP interface (RFC4214)
1340 * @IFF_WAN_HDLC: WAN HDLC device
1341 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1343 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1344 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1345 * @IFF_MACVLAN_PORT: device used as macvlan port
1346 * @IFF_BRIDGE_PORT: device used as bridge port
1347 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1348 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1349 * @IFF_UNICAST_FLT: Supports unicast filtering
1350 * @IFF_TEAM_PORT: device used as team port
1351 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1352 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1353 * change when it's running
1354 * @IFF_MACVLAN: Macvlan device
1355 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1356 * underlying stacked devices
1357 * @IFF_IPVLAN_MASTER: IPvlan master device
1358 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1359 * @IFF_L3MDEV_MASTER: device is an L3 master device
1360 * @IFF_NO_QUEUE: device can run without qdisc attached
1361 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1362 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1363 * @IFF_TEAM: device is a team device
1364 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1365 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1366 * entity (i.e. the master device for bridged veth)
1367 * @IFF_MACSEC: device is a MACsec device
1369 enum netdev_priv_flags {
1370 IFF_802_1Q_VLAN = 1<<0,
1374 IFF_WAN_HDLC = 1<<4,
1375 IFF_XMIT_DST_RELEASE = 1<<5,
1376 IFF_DONT_BRIDGE = 1<<6,
1377 IFF_DISABLE_NETPOLL = 1<<7,
1378 IFF_MACVLAN_PORT = 1<<8,
1379 IFF_BRIDGE_PORT = 1<<9,
1380 IFF_OVS_DATAPATH = 1<<10,
1381 IFF_TX_SKB_SHARING = 1<<11,
1382 IFF_UNICAST_FLT = 1<<12,
1383 IFF_TEAM_PORT = 1<<13,
1384 IFF_SUPP_NOFCS = 1<<14,
1385 IFF_LIVE_ADDR_CHANGE = 1<<15,
1386 IFF_MACVLAN = 1<<16,
1387 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1388 IFF_IPVLAN_MASTER = 1<<18,
1389 IFF_IPVLAN_SLAVE = 1<<19,
1390 IFF_L3MDEV_MASTER = 1<<20,
1391 IFF_NO_QUEUE = 1<<21,
1392 IFF_OPENVSWITCH = 1<<22,
1393 IFF_L3MDEV_SLAVE = 1<<23,
1395 IFF_RXFH_CONFIGURED = 1<<25,
1396 IFF_PHONY_HEADROOM = 1<<26,
1400 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1401 #define IFF_EBRIDGE IFF_EBRIDGE
1402 #define IFF_BONDING IFF_BONDING
1403 #define IFF_ISATAP IFF_ISATAP
1404 #define IFF_WAN_HDLC IFF_WAN_HDLC
1405 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1406 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1407 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1408 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1409 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1410 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1411 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1412 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1413 #define IFF_TEAM_PORT IFF_TEAM_PORT
1414 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1415 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1416 #define IFF_MACVLAN IFF_MACVLAN
1417 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1418 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1419 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1420 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1421 #define IFF_NO_QUEUE IFF_NO_QUEUE
1422 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1423 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1424 #define IFF_TEAM IFF_TEAM
1425 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1426 #define IFF_MACSEC IFF_MACSEC
1429 * struct net_device - The DEVICE structure.
1430 * Actually, this whole structure is a big mistake. It mixes I/O
1431 * data with strictly "high-level" data, and it has to know about
1432 * almost every data structure used in the INET module.
1434 * @name: This is the first field of the "visible" part of this structure
1435 * (i.e. as seen by users in the "Space.c" file). It is the name
1438 * @name_hlist: Device name hash chain, please keep it close to name[]
1439 * @ifalias: SNMP alias
1440 * @mem_end: Shared memory end
1441 * @mem_start: Shared memory start
1442 * @base_addr: Device I/O address
1443 * @irq: Device IRQ number
1445 * @carrier_changes: Stats to monitor carrier on<->off transitions
1447 * @state: Generic network queuing layer state, see netdev_state_t
1448 * @dev_list: The global list of network devices
1449 * @napi_list: List entry used for polling NAPI devices
1450 * @unreg_list: List entry when we are unregistering the
1451 * device; see the function unregister_netdev
1452 * @close_list: List entry used when we are closing the device
1453 * @ptype_all: Device-specific packet handlers for all protocols
1454 * @ptype_specific: Device-specific, protocol-specific packet handlers
1456 * @adj_list: Directly linked devices, like slaves for bonding
1457 * @all_adj_list: All linked devices, *including* neighbours
1458 * @features: Currently active device features
1459 * @hw_features: User-changeable features
1461 * @wanted_features: User-requested features
1462 * @vlan_features: Mask of features inheritable by VLAN devices
1464 * @hw_enc_features: Mask of features inherited by encapsulating devices
1465 * This field indicates what encapsulation
1466 * offloads the hardware is capable of doing,
1467 * and drivers will need to set them appropriately.
1469 * @mpls_features: Mask of features inheritable by MPLS
1471 * @ifindex: interface index
1472 * @group: The group the device belongs to
1474 * @stats: Statistics struct, which was left as a legacy, use
1475 * rtnl_link_stats64 instead
1477 * @rx_dropped: Dropped packets by core network,
1478 * do not use this in drivers
1479 * @tx_dropped: Dropped packets by core network,
1480 * do not use this in drivers
1481 * @rx_nohandler: nohandler dropped packets by core network on
1482 * inactive devices, do not use this in drivers
1484 * @wireless_handlers: List of functions to handle Wireless Extensions,
1486 * see <net/iw_handler.h> for details.
1487 * @wireless_data: Instance data managed by the core of wireless extensions
1489 * @netdev_ops: Includes several pointers to callbacks,
1490 * if one wants to override the ndo_*() functions
1491 * @ethtool_ops: Management operations
1492 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1493 * discovery handling. Necessary for e.g. 6LoWPAN.
1494 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1495 * of Layer 2 headers.
1497 * @flags: Interface flags (a la BSD)
1498 * @priv_flags: Like 'flags' but invisible to userspace,
1499 * see if.h for the definitions
1500 * @gflags: Global flags ( kept as legacy )
1501 * @padded: How much padding added by alloc_netdev()
1502 * @operstate: RFC2863 operstate
1503 * @link_mode: Mapping policy to operstate
1504 * @if_port: Selectable AUI, TP, ...
1506 * @mtu: Interface MTU value
1507 * @type: Interface hardware type
1508 * @hard_header_len: Maximum hardware header length.
1510 * @needed_headroom: Extra headroom the hardware may need, but not in all
1511 * cases can this be guaranteed
1512 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1513 * cases can this be guaranteed. Some cases also use
1514 * LL_MAX_HEADER instead to allocate the skb
1516 * interface address info:
1518 * @perm_addr: Permanent hw address
1519 * @addr_assign_type: Hw address assignment type
1520 * @addr_len: Hardware address length
1521 * @neigh_priv_len: Used in neigh_alloc()
1522 * @dev_id: Used to differentiate devices that share
1523 * the same link layer address
1524 * @dev_port: Used to differentiate devices that share
1526 * @addr_list_lock: XXX: need comments on this one
1527 * @uc_promisc: Counter that indicates promiscuous mode
1528 * has been enabled due to the need to listen to
1529 * additional unicast addresses in a device that
1530 * does not implement ndo_set_rx_mode()
1531 * @uc: unicast mac addresses
1532 * @mc: multicast mac addresses
1533 * @dev_addrs: list of device hw addresses
1534 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1535 * @promiscuity: Number of times the NIC is told to work in
1536 * promiscuous mode; if it becomes 0 the NIC will
1537 * exit promiscuous mode
1538 * @allmulti: Counter, enables or disables allmulticast mode
1540 * @vlan_info: VLAN info
1541 * @dsa_ptr: dsa specific data
1542 * @tipc_ptr: TIPC specific data
1543 * @atalk_ptr: AppleTalk link
1544 * @ip_ptr: IPv4 specific data
1545 * @dn_ptr: DECnet specific data
1546 * @ip6_ptr: IPv6 specific data
1547 * @ax25_ptr: AX.25 specific data
1548 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1550 * @last_rx: Time of last Rx
1551 * @dev_addr: Hw address (before bcast,
1552 * because most packets are unicast)
1554 * @_rx: Array of RX queues
1555 * @num_rx_queues: Number of RX queues
1556 * allocated at register_netdev() time
1557 * @real_num_rx_queues: Number of RX queues currently active in device
1559 * @rx_handler: handler for received packets
1560 * @rx_handler_data: XXX: need comments on this one
1561 * @ingress_queue: XXX: need comments on this one
1562 * @broadcast: hw bcast address
1564 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1565 * indexed by RX queue number. Assigned by driver.
1566 * This must only be set if the ndo_rx_flow_steer
1567 * operation is defined
1568 * @index_hlist: Device index hash chain
1570 * @_tx: Array of TX queues
1571 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1572 * @real_num_tx_queues: Number of TX queues currently active in device
1573 * @qdisc: Root qdisc from userspace point of view
1574 * @tx_queue_len: Max frames per queue allowed
1575 * @tx_global_lock: XXX: need comments on this one
1577 * @xps_maps: XXX: need comments on this one
1579 * @watchdog_timeo: Represents the timeout that is used by
1580 * the watchdog (see dev_watchdog())
1581 * @watchdog_timer: List of timers
1583 * @pcpu_refcnt: Number of references to this device
1584 * @todo_list: Delayed register/unregister
1585 * @link_watch_list: XXX: need comments on this one
1587 * @reg_state: Register/unregister state machine
1588 * @dismantle: Device is going to be freed
1589 * @rtnl_link_state: This enum represents the phases of creating
1592 * @destructor: Called from unregister,
1593 * can be used to call free_netdev
1594 * @npinfo: XXX: need comments on this one
1595 * @nd_net: Network namespace this network device is inside
1597 * @ml_priv: Mid-layer private
1598 * @lstats: Loopback statistics
1599 * @tstats: Tunnel statistics
1600 * @dstats: Dummy statistics
1601 * @vstats: Virtual ethernet statistics
1606 * @dev: Class/net/name entry
1607 * @sysfs_groups: Space for optional device, statistics and wireless
1610 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1611 * @rtnl_link_ops: Rtnl_link_ops
1613 * @gso_max_size: Maximum size of generic segmentation offload
1614 * @gso_max_segs: Maximum number of segments that can be passed to the
1617 * @dcbnl_ops: Data Center Bridging netlink ops
1618 * @num_tc: Number of traffic classes in the net device
1619 * @tc_to_txq: XXX: need comments on this one
1620 * @prio_tc_map XXX: need comments on this one
1622 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1624 * @priomap: XXX: need comments on this one
1625 * @phydev: Physical device may attach itself
1626 * for hardware timestamping
1628 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1629 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1631 * @proto_down: protocol port state information can be sent to the
1632 * switch driver and used to set the phys state of the
1635 * FIXME: cleanup struct net_device such that network protocol info
1640 char name[IFNAMSIZ];
1641 struct hlist_node name_hlist;
1644 * I/O specific fields
1645 * FIXME: Merge these and struct ifmap into one
1647 unsigned long mem_end;
1648 unsigned long mem_start;
1649 unsigned long base_addr;
1652 atomic_t carrier_changes;
1655 * Some hardware also needs these fields (state,dev_list,
1656 * napi_list,unreg_list,close_list) but they are not
1657 * part of the usual set specified in Space.c.
1660 unsigned long state;
1662 struct list_head dev_list;
1663 struct list_head napi_list;
1664 struct list_head unreg_list;
1665 struct list_head close_list;
1666 struct list_head ptype_all;
1667 struct list_head ptype_specific;
1670 struct list_head upper;
1671 struct list_head lower;
1675 struct list_head upper;
1676 struct list_head lower;
1679 netdev_features_t features;
1680 netdev_features_t hw_features;
1681 netdev_features_t wanted_features;
1682 netdev_features_t vlan_features;
1683 netdev_features_t hw_enc_features;
1684 netdev_features_t mpls_features;
1685 netdev_features_t gso_partial_features;
1690 struct net_device_stats stats;
1692 atomic_long_t rx_dropped;
1693 atomic_long_t tx_dropped;
1694 atomic_long_t rx_nohandler;
1696 #ifdef CONFIG_WIRELESS_EXT
1697 const struct iw_handler_def *wireless_handlers;
1698 struct iw_public_data *wireless_data;
1700 const struct net_device_ops *netdev_ops;
1701 const struct ethtool_ops *ethtool_ops;
1702 #ifdef CONFIG_NET_SWITCHDEV
1703 const struct switchdev_ops *switchdev_ops;
1705 #ifdef CONFIG_NET_L3_MASTER_DEV
1706 const struct l3mdev_ops *l3mdev_ops;
1708 #if IS_ENABLED(CONFIG_IPV6)
1709 const struct ndisc_ops *ndisc_ops;
1712 const struct header_ops *header_ops;
1715 unsigned int priv_flags;
1717 unsigned short gflags;
1718 unsigned short padded;
1720 unsigned char operstate;
1721 unsigned char link_mode;
1723 unsigned char if_port;
1727 unsigned short type;
1728 unsigned short hard_header_len;
1730 unsigned short needed_headroom;
1731 unsigned short needed_tailroom;
1733 /* Interface address info. */
1734 unsigned char perm_addr[MAX_ADDR_LEN];
1735 unsigned char addr_assign_type;
1736 unsigned char addr_len;
1737 unsigned short neigh_priv_len;
1738 unsigned short dev_id;
1739 unsigned short dev_port;
1740 spinlock_t addr_list_lock;
1741 unsigned char name_assign_type;
1743 struct netdev_hw_addr_list uc;
1744 struct netdev_hw_addr_list mc;
1745 struct netdev_hw_addr_list dev_addrs;
1748 struct kset *queues_kset;
1750 unsigned int promiscuity;
1751 unsigned int allmulti;
1754 /* Protocol-specific pointers */
1756 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1757 struct vlan_info __rcu *vlan_info;
1759 #if IS_ENABLED(CONFIG_NET_DSA)
1760 struct dsa_switch_tree *dsa_ptr;
1762 #if IS_ENABLED(CONFIG_TIPC)
1763 struct tipc_bearer __rcu *tipc_ptr;
1766 struct in_device __rcu *ip_ptr;
1767 struct dn_dev __rcu *dn_ptr;
1768 struct inet6_dev __rcu *ip6_ptr;
1770 struct wireless_dev *ieee80211_ptr;
1771 struct wpan_dev *ieee802154_ptr;
1772 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1773 struct mpls_dev __rcu *mpls_ptr;
1777 * Cache lines mostly used on receive path (including eth_type_trans())
1779 unsigned long last_rx;
1781 /* Interface address info used in eth_type_trans() */
1782 unsigned char *dev_addr;
1785 struct netdev_rx_queue *_rx;
1787 unsigned int num_rx_queues;
1788 unsigned int real_num_rx_queues;
1791 unsigned long gro_flush_timeout;
1792 rx_handler_func_t __rcu *rx_handler;
1793 void __rcu *rx_handler_data;
1795 #ifdef CONFIG_NET_CLS_ACT
1796 struct tcf_proto __rcu *ingress_cl_list;
1798 struct netdev_queue __rcu *ingress_queue;
1799 #ifdef CONFIG_NETFILTER_INGRESS
1800 struct list_head nf_hooks_ingress;
1803 unsigned char broadcast[MAX_ADDR_LEN];
1804 #ifdef CONFIG_RFS_ACCEL
1805 struct cpu_rmap *rx_cpu_rmap;
1807 struct hlist_node index_hlist;
1810 * Cache lines mostly used on transmit path
1812 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1813 unsigned int num_tx_queues;
1814 unsigned int real_num_tx_queues;
1815 struct Qdisc *qdisc;
1816 #ifdef CONFIG_NET_SCHED
1817 DECLARE_HASHTABLE (qdisc_hash, 4);
1819 unsigned long tx_queue_len;
1820 spinlock_t tx_global_lock;
1824 struct xps_dev_maps __rcu *xps_maps;
1826 #ifdef CONFIG_NET_CLS_ACT
1827 struct tcf_proto __rcu *egress_cl_list;
1830 /* These may be needed for future network-power-down code. */
1831 struct timer_list watchdog_timer;
1833 int __percpu *pcpu_refcnt;
1834 struct list_head todo_list;
1836 struct list_head link_watch_list;
1838 enum { NETREG_UNINITIALIZED=0,
1839 NETREG_REGISTERED, /* completed register_netdevice */
1840 NETREG_UNREGISTERING, /* called unregister_netdevice */
1841 NETREG_UNREGISTERED, /* completed unregister todo */
1842 NETREG_RELEASED, /* called free_netdev */
1843 NETREG_DUMMY, /* dummy device for NAPI poll */
1849 RTNL_LINK_INITIALIZED,
1850 RTNL_LINK_INITIALIZING,
1851 } rtnl_link_state:16;
1853 void (*destructor)(struct net_device *dev);
1855 #ifdef CONFIG_NETPOLL
1856 struct netpoll_info __rcu *npinfo;
1859 possible_net_t nd_net;
1861 /* mid-layer private */
1864 struct pcpu_lstats __percpu *lstats;
1865 struct pcpu_sw_netstats __percpu *tstats;
1866 struct pcpu_dstats __percpu *dstats;
1867 struct pcpu_vstats __percpu *vstats;
1870 struct garp_port __rcu *garp_port;
1871 struct mrp_port __rcu *mrp_port;
1874 const struct attribute_group *sysfs_groups[4];
1875 const struct attribute_group *sysfs_rx_queue_group;
1877 const struct rtnl_link_ops *rtnl_link_ops;
1879 /* for setting kernel sock attribute on TCP connection setup */
1880 #define GSO_MAX_SIZE 65536
1881 unsigned int gso_max_size;
1882 #define GSO_MAX_SEGS 65535
1886 const struct dcbnl_rtnl_ops *dcbnl_ops;
1889 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1890 u8 prio_tc_map[TC_BITMASK + 1];
1892 #if IS_ENABLED(CONFIG_FCOE)
1893 unsigned int fcoe_ddp_xid;
1895 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1896 struct netprio_map __rcu *priomap;
1898 struct phy_device *phydev;
1899 struct lock_class_key *qdisc_tx_busylock;
1900 struct lock_class_key *qdisc_running_key;
1903 #define to_net_dev(d) container_of(d, struct net_device, dev)
1905 #define NETDEV_ALIGN 32
1908 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1910 return dev->prio_tc_map[prio & TC_BITMASK];
1914 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1916 if (tc >= dev->num_tc)
1919 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1924 void netdev_reset_tc(struct net_device *dev)
1927 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1928 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1932 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1934 if (tc >= dev->num_tc)
1937 dev->tc_to_txq[tc].count = count;
1938 dev->tc_to_txq[tc].offset = offset;
1943 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1945 if (num_tc > TC_MAX_QUEUE)
1948 dev->num_tc = num_tc;
1953 int netdev_get_num_tc(struct net_device *dev)
1959 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1962 return &dev->_tx[index];
1965 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1966 const struct sk_buff *skb)
1968 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1971 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1972 void (*f)(struct net_device *,
1973 struct netdev_queue *,
1979 for (i = 0; i < dev->num_tx_queues; i++)
1980 f(dev, &dev->_tx[i], arg);
1983 #define netdev_lockdep_set_classes(dev) \
1985 static struct lock_class_key qdisc_tx_busylock_key; \
1986 static struct lock_class_key qdisc_running_key; \
1987 static struct lock_class_key qdisc_xmit_lock_key; \
1988 static struct lock_class_key dev_addr_list_lock_key; \
1991 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1992 (dev)->qdisc_running_key = &qdisc_running_key; \
1993 lockdep_set_class(&(dev)->addr_list_lock, \
1994 &dev_addr_list_lock_key); \
1995 for (i = 0; i < (dev)->num_tx_queues; i++) \
1996 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1997 &qdisc_xmit_lock_key); \
2000 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2001 struct sk_buff *skb,
2004 /* returns the headroom that the master device needs to take in account
2005 * when forwarding to this dev
2007 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2009 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2012 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2014 if (dev->netdev_ops->ndo_set_rx_headroom)
2015 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2018 /* set the device rx headroom to the dev's default */
2019 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2021 netdev_set_rx_headroom(dev, -1);
2025 * Net namespace inlines
2028 struct net *dev_net(const struct net_device *dev)
2030 return read_pnet(&dev->nd_net);
2034 void dev_net_set(struct net_device *dev, struct net *net)
2036 write_pnet(&dev->nd_net, net);
2039 static inline bool netdev_uses_dsa(struct net_device *dev)
2041 #if IS_ENABLED(CONFIG_NET_DSA)
2042 if (dev->dsa_ptr != NULL)
2043 return dsa_uses_tagged_protocol(dev->dsa_ptr);
2049 * netdev_priv - access network device private data
2050 * @dev: network device
2052 * Get network device private data
2054 static inline void *netdev_priv(const struct net_device *dev)
2056 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2059 /* Set the sysfs physical device reference for the network logical device
2060 * if set prior to registration will cause a symlink during initialization.
2062 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2064 /* Set the sysfs device type for the network logical device to allow
2065 * fine-grained identification of different network device types. For
2066 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2068 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2070 /* Default NAPI poll() weight
2071 * Device drivers are strongly advised to not use bigger value
2073 #define NAPI_POLL_WEIGHT 64
2076 * netif_napi_add - initialize a NAPI context
2077 * @dev: network device
2078 * @napi: NAPI context
2079 * @poll: polling function
2080 * @weight: default weight
2082 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2083 * *any* of the other NAPI-related functions.
2085 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2086 int (*poll)(struct napi_struct *, int), int weight);
2089 * netif_tx_napi_add - initialize a NAPI context
2090 * @dev: network device
2091 * @napi: NAPI context
2092 * @poll: polling function
2093 * @weight: default weight
2095 * This variant of netif_napi_add() should be used from drivers using NAPI
2096 * to exclusively poll a TX queue.
2097 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2099 static inline void netif_tx_napi_add(struct net_device *dev,
2100 struct napi_struct *napi,
2101 int (*poll)(struct napi_struct *, int),
2104 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2105 netif_napi_add(dev, napi, poll, weight);
2109 * netif_napi_del - remove a NAPI context
2110 * @napi: NAPI context
2112 * netif_napi_del() removes a NAPI context from the network device NAPI list
2114 void netif_napi_del(struct napi_struct *napi);
2116 struct napi_gro_cb {
2117 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2120 /* Length of frag0. */
2121 unsigned int frag0_len;
2123 /* This indicates where we are processing relative to skb->data. */
2126 /* This is non-zero if the packet cannot be merged with the new skb. */
2129 /* Save the IP ID here and check when we get to the transport layer */
2132 /* Number of segments aggregated. */
2135 /* Start offset for remote checksum offload */
2136 u16 gro_remcsum_start;
2138 /* jiffies when first packet was created/queued */
2141 /* Used in ipv6_gro_receive() and foo-over-udp */
2144 /* This is non-zero if the packet may be of the same flow. */
2147 /* Used in tunnel GRO receive */
2150 /* GRO checksum is valid */
2153 /* Number of checksums via CHECKSUM_UNNECESSARY */
2158 #define NAPI_GRO_FREE 1
2159 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2161 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2164 /* Used in GRE, set in fou/gue_gro_receive */
2167 /* Used to determine if flush_id can be ignored */
2172 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2175 /* used in skb_gro_receive() slow path */
2176 struct sk_buff *last;
2179 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2181 struct packet_type {
2182 __be16 type; /* This is really htons(ether_type). */
2183 struct net_device *dev; /* NULL is wildcarded here */
2184 int (*func) (struct sk_buff *,
2185 struct net_device *,
2186 struct packet_type *,
2187 struct net_device *);
2188 bool (*id_match)(struct packet_type *ptype,
2190 void *af_packet_priv;
2191 struct list_head list;
2194 struct offload_callbacks {
2195 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2196 netdev_features_t features);
2197 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2198 struct sk_buff *skb);
2199 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2202 struct packet_offload {
2203 __be16 type; /* This is really htons(ether_type). */
2205 struct offload_callbacks callbacks;
2206 struct list_head list;
2209 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2210 struct pcpu_sw_netstats {
2215 struct u64_stats_sync syncp;
2218 #define __netdev_alloc_pcpu_stats(type, gfp) \
2220 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2223 for_each_possible_cpu(__cpu) { \
2224 typeof(type) *stat; \
2225 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2226 u64_stats_init(&stat->syncp); \
2232 #define netdev_alloc_pcpu_stats(type) \
2233 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2235 enum netdev_lag_tx_type {
2236 NETDEV_LAG_TX_TYPE_UNKNOWN,
2237 NETDEV_LAG_TX_TYPE_RANDOM,
2238 NETDEV_LAG_TX_TYPE_BROADCAST,
2239 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2240 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2241 NETDEV_LAG_TX_TYPE_HASH,
2244 struct netdev_lag_upper_info {
2245 enum netdev_lag_tx_type tx_type;
2248 struct netdev_lag_lower_state_info {
2253 #include <linux/notifier.h>
2255 /* netdevice notifier chain. Please remember to update the rtnetlink
2256 * notification exclusion list in rtnetlink_event() when adding new
2259 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2260 #define NETDEV_DOWN 0x0002
2261 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2262 detected a hardware crash and restarted
2263 - we can use this eg to kick tcp sessions
2265 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2266 #define NETDEV_REGISTER 0x0005
2267 #define NETDEV_UNREGISTER 0x0006
2268 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2269 #define NETDEV_CHANGEADDR 0x0008
2270 #define NETDEV_GOING_DOWN 0x0009
2271 #define NETDEV_CHANGENAME 0x000A
2272 #define NETDEV_FEAT_CHANGE 0x000B
2273 #define NETDEV_BONDING_FAILOVER 0x000C
2274 #define NETDEV_PRE_UP 0x000D
2275 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2276 #define NETDEV_POST_TYPE_CHANGE 0x000F
2277 #define NETDEV_POST_INIT 0x0010
2278 #define NETDEV_UNREGISTER_FINAL 0x0011
2279 #define NETDEV_RELEASE 0x0012
2280 #define NETDEV_NOTIFY_PEERS 0x0013
2281 #define NETDEV_JOIN 0x0014
2282 #define NETDEV_CHANGEUPPER 0x0015
2283 #define NETDEV_RESEND_IGMP 0x0016
2284 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2285 #define NETDEV_CHANGEINFODATA 0x0018
2286 #define NETDEV_BONDING_INFO 0x0019
2287 #define NETDEV_PRECHANGEUPPER 0x001A
2288 #define NETDEV_CHANGELOWERSTATE 0x001B
2289 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2290 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2292 int register_netdevice_notifier(struct notifier_block *nb);
2293 int unregister_netdevice_notifier(struct notifier_block *nb);
2295 struct netdev_notifier_info {
2296 struct net_device *dev;
2299 struct netdev_notifier_change_info {
2300 struct netdev_notifier_info info; /* must be first */
2301 unsigned int flags_changed;
2304 struct netdev_notifier_changeupper_info {
2305 struct netdev_notifier_info info; /* must be first */
2306 struct net_device *upper_dev; /* new upper dev */
2307 bool master; /* is upper dev master */
2308 bool linking; /* is the notification for link or unlink */
2309 void *upper_info; /* upper dev info */
2312 struct netdev_notifier_changelowerstate_info {
2313 struct netdev_notifier_info info; /* must be first */
2314 void *lower_state_info; /* is lower dev state */
2317 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2318 struct net_device *dev)
2323 static inline struct net_device *
2324 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2329 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2332 extern rwlock_t dev_base_lock; /* Device list lock */
2334 #define for_each_netdev(net, d) \
2335 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2336 #define for_each_netdev_reverse(net, d) \
2337 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2338 #define for_each_netdev_rcu(net, d) \
2339 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2340 #define for_each_netdev_safe(net, d, n) \
2341 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2342 #define for_each_netdev_continue(net, d) \
2343 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2344 #define for_each_netdev_continue_rcu(net, d) \
2345 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2346 #define for_each_netdev_in_bond_rcu(bond, slave) \
2347 for_each_netdev_rcu(&init_net, slave) \
2348 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2349 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2351 static inline struct net_device *next_net_device(struct net_device *dev)
2353 struct list_head *lh;
2357 lh = dev->dev_list.next;
2358 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2361 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2363 struct list_head *lh;
2367 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2368 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2371 static inline struct net_device *first_net_device(struct net *net)
2373 return list_empty(&net->dev_base_head) ? NULL :
2374 net_device_entry(net->dev_base_head.next);
2377 static inline struct net_device *first_net_device_rcu(struct net *net)
2379 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2381 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2384 int netdev_boot_setup_check(struct net_device *dev);
2385 unsigned long netdev_boot_base(const char *prefix, int unit);
2386 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2387 const char *hwaddr);
2388 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2389 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2390 void dev_add_pack(struct packet_type *pt);
2391 void dev_remove_pack(struct packet_type *pt);
2392 void __dev_remove_pack(struct packet_type *pt);
2393 void dev_add_offload(struct packet_offload *po);
2394 void dev_remove_offload(struct packet_offload *po);
2396 int dev_get_iflink(const struct net_device *dev);
2397 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2398 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2399 unsigned short mask);
2400 struct net_device *dev_get_by_name(struct net *net, const char *name);
2401 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2402 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2403 int dev_alloc_name(struct net_device *dev, const char *name);
2404 int dev_open(struct net_device *dev);
2405 int dev_close(struct net_device *dev);
2406 int dev_close_many(struct list_head *head, bool unlink);
2407 void dev_disable_lro(struct net_device *dev);
2408 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2409 int dev_queue_xmit(struct sk_buff *skb);
2410 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2411 int register_netdevice(struct net_device *dev);
2412 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2413 void unregister_netdevice_many(struct list_head *head);
2414 static inline void unregister_netdevice(struct net_device *dev)
2416 unregister_netdevice_queue(dev, NULL);
2419 int netdev_refcnt_read(const struct net_device *dev);
2420 void free_netdev(struct net_device *dev);
2421 void netdev_freemem(struct net_device *dev);
2422 void synchronize_net(void);
2423 int init_dummy_netdev(struct net_device *dev);
2425 DECLARE_PER_CPU(int, xmit_recursion);
2426 #define XMIT_RECURSION_LIMIT 10
2428 static inline int dev_recursion_level(void)
2430 return this_cpu_read(xmit_recursion);
2433 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2434 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2435 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2436 int netdev_get_name(struct net *net, char *name, int ifindex);
2437 int dev_restart(struct net_device *dev);
2438 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2440 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2442 return NAPI_GRO_CB(skb)->data_offset;
2445 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2447 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2450 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2452 NAPI_GRO_CB(skb)->data_offset += len;
2455 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2456 unsigned int offset)
2458 return NAPI_GRO_CB(skb)->frag0 + offset;
2461 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2463 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2466 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2467 unsigned int offset)
2469 if (!pskb_may_pull(skb, hlen))
2472 NAPI_GRO_CB(skb)->frag0 = NULL;
2473 NAPI_GRO_CB(skb)->frag0_len = 0;
2474 return skb->data + offset;
2477 static inline void *skb_gro_network_header(struct sk_buff *skb)
2479 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2480 skb_network_offset(skb);
2483 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2484 const void *start, unsigned int len)
2486 if (NAPI_GRO_CB(skb)->csum_valid)
2487 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2488 csum_partial(start, len, 0));
2491 /* GRO checksum functions. These are logical equivalents of the normal
2492 * checksum functions (in skbuff.h) except that they operate on the GRO
2493 * offsets and fields in sk_buff.
2496 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2498 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2500 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2503 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2507 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2508 skb_checksum_start_offset(skb) <
2509 skb_gro_offset(skb)) &&
2510 !skb_at_gro_remcsum_start(skb) &&
2511 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2512 (!zero_okay || check));
2515 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2518 if (NAPI_GRO_CB(skb)->csum_valid &&
2519 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2522 NAPI_GRO_CB(skb)->csum = psum;
2524 return __skb_gro_checksum_complete(skb);
2527 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2529 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2530 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2531 NAPI_GRO_CB(skb)->csum_cnt--;
2533 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2534 * verified a new top level checksum or an encapsulated one
2535 * during GRO. This saves work if we fallback to normal path.
2537 __skb_incr_checksum_unnecessary(skb);
2541 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2544 __sum16 __ret = 0; \
2545 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2546 __ret = __skb_gro_checksum_validate_complete(skb, \
2547 compute_pseudo(skb, proto)); \
2549 __skb_mark_checksum_bad(skb); \
2551 skb_gro_incr_csum_unnecessary(skb); \
2555 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2556 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2558 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2560 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2562 #define skb_gro_checksum_simple_validate(skb) \
2563 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2565 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2567 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2568 !NAPI_GRO_CB(skb)->csum_valid);
2571 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2572 __sum16 check, __wsum pseudo)
2574 NAPI_GRO_CB(skb)->csum = ~pseudo;
2575 NAPI_GRO_CB(skb)->csum_valid = 1;
2578 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2580 if (__skb_gro_checksum_convert_check(skb)) \
2581 __skb_gro_checksum_convert(skb, check, \
2582 compute_pseudo(skb, proto)); \
2585 struct gro_remcsum {
2590 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2596 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2597 unsigned int off, size_t hdrlen,
2598 int start, int offset,
2599 struct gro_remcsum *grc,
2603 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2605 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2608 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2612 ptr = skb_gro_header_fast(skb, off);
2613 if (skb_gro_header_hard(skb, off + plen)) {
2614 ptr = skb_gro_header_slow(skb, off + plen, off);
2619 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2622 /* Adjust skb->csum since we changed the packet */
2623 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2625 grc->offset = off + hdrlen + offset;
2631 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2632 struct gro_remcsum *grc)
2635 size_t plen = grc->offset + sizeof(u16);
2640 ptr = skb_gro_header_fast(skb, grc->offset);
2641 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2642 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2647 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2650 struct skb_csum_offl_spec {
2664 bool __skb_csum_offload_chk(struct sk_buff *skb,
2665 const struct skb_csum_offl_spec *spec,
2666 bool *csum_encapped,
2669 static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2670 const struct skb_csum_offl_spec *spec,
2671 bool *csum_encapped,
2674 if (skb->ip_summed != CHECKSUM_PARTIAL)
2677 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2680 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2681 const struct skb_csum_offl_spec *spec)
2685 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2688 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2690 static const struct skb_csum_offl_spec csum_offl_spec = {
2692 .ip_options_okay = 1,
2699 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2702 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2704 static const struct skb_csum_offl_spec csum_offl_spec = {
2706 .ip_options_okay = 1,
2712 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2715 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2716 unsigned short type,
2717 const void *daddr, const void *saddr,
2720 if (!dev->header_ops || !dev->header_ops->create)
2723 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2726 static inline int dev_parse_header(const struct sk_buff *skb,
2727 unsigned char *haddr)
2729 const struct net_device *dev = skb->dev;
2731 if (!dev->header_ops || !dev->header_ops->parse)
2733 return dev->header_ops->parse(skb, haddr);
2736 /* ll_header must have at least hard_header_len allocated */
2737 static inline bool dev_validate_header(const struct net_device *dev,
2738 char *ll_header, int len)
2740 if (likely(len >= dev->hard_header_len))
2743 if (capable(CAP_SYS_RAWIO)) {
2744 memset(ll_header + len, 0, dev->hard_header_len - len);
2748 if (dev->header_ops && dev->header_ops->validate)
2749 return dev->header_ops->validate(ll_header, len);
2754 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2755 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2756 static inline int unregister_gifconf(unsigned int family)
2758 return register_gifconf(family, NULL);
2761 #ifdef CONFIG_NET_FLOW_LIMIT
2762 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2763 struct sd_flow_limit {
2765 unsigned int num_buckets;
2766 unsigned int history_head;
2767 u16 history[FLOW_LIMIT_HISTORY];
2771 extern int netdev_flow_limit_table_len;
2772 #endif /* CONFIG_NET_FLOW_LIMIT */
2775 * Incoming packets are placed on per-CPU queues
2777 struct softnet_data {
2778 struct list_head poll_list;
2779 struct sk_buff_head process_queue;
2782 unsigned int processed;
2783 unsigned int time_squeeze;
2784 unsigned int received_rps;
2786 struct softnet_data *rps_ipi_list;
2788 #ifdef CONFIG_NET_FLOW_LIMIT
2789 struct sd_flow_limit __rcu *flow_limit;
2791 struct Qdisc *output_queue;
2792 struct Qdisc **output_queue_tailp;
2793 struct sk_buff *completion_queue;
2796 /* input_queue_head should be written by cpu owning this struct,
2797 * and only read by other cpus. Worth using a cache line.
2799 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2801 /* Elements below can be accessed between CPUs for RPS/RFS */
2802 struct call_single_data csd ____cacheline_aligned_in_smp;
2803 struct softnet_data *rps_ipi_next;
2805 unsigned int input_queue_tail;
2807 unsigned int dropped;
2808 struct sk_buff_head input_pkt_queue;
2809 struct napi_struct backlog;
2813 static inline void input_queue_head_incr(struct softnet_data *sd)
2816 sd->input_queue_head++;
2820 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2821 unsigned int *qtail)
2824 *qtail = ++sd->input_queue_tail;
2828 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2830 void __netif_schedule(struct Qdisc *q);
2831 void netif_schedule_queue(struct netdev_queue *txq);
2833 static inline void netif_tx_schedule_all(struct net_device *dev)
2837 for (i = 0; i < dev->num_tx_queues; i++)
2838 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2841 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2843 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2847 * netif_start_queue - allow transmit
2848 * @dev: network device
2850 * Allow upper layers to call the device hard_start_xmit routine.
2852 static inline void netif_start_queue(struct net_device *dev)
2854 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2857 static inline void netif_tx_start_all_queues(struct net_device *dev)
2861 for (i = 0; i < dev->num_tx_queues; i++) {
2862 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2863 netif_tx_start_queue(txq);
2867 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2870 * netif_wake_queue - restart transmit
2871 * @dev: network device
2873 * Allow upper layers to call the device hard_start_xmit routine.
2874 * Used for flow control when transmit resources are available.
2876 static inline void netif_wake_queue(struct net_device *dev)
2878 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2881 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2885 for (i = 0; i < dev->num_tx_queues; i++) {
2886 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2887 netif_tx_wake_queue(txq);
2891 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2893 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2897 * netif_stop_queue - stop transmitted packets
2898 * @dev: network device
2900 * Stop upper layers calling the device hard_start_xmit routine.
2901 * Used for flow control when transmit resources are unavailable.
2903 static inline void netif_stop_queue(struct net_device *dev)
2905 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2908 void netif_tx_stop_all_queues(struct net_device *dev);
2910 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2912 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2916 * netif_queue_stopped - test if transmit queue is flowblocked
2917 * @dev: network device
2919 * Test if transmit queue on device is currently unable to send.
2921 static inline bool netif_queue_stopped(const struct net_device *dev)
2923 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2926 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2928 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2932 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2934 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2938 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2940 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2944 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2945 * @dev_queue: pointer to transmit queue
2947 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2948 * to give appropriate hint to the CPU.
2950 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2953 prefetchw(&dev_queue->dql.num_queued);
2958 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2959 * @dev_queue: pointer to transmit queue
2961 * BQL enabled drivers might use this helper in their TX completion path,
2962 * to give appropriate hint to the CPU.
2964 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2967 prefetchw(&dev_queue->dql.limit);
2971 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2975 dql_queued(&dev_queue->dql, bytes);
2977 if (likely(dql_avail(&dev_queue->dql) >= 0))
2980 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2983 * The XOFF flag must be set before checking the dql_avail below,
2984 * because in netdev_tx_completed_queue we update the dql_completed
2985 * before checking the XOFF flag.
2989 /* check again in case another CPU has just made room avail */
2990 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2991 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2996 * netdev_sent_queue - report the number of bytes queued to hardware
2997 * @dev: network device
2998 * @bytes: number of bytes queued to the hardware device queue
3000 * Report the number of bytes queued for sending/completion to the network
3001 * device hardware queue. @bytes should be a good approximation and should
3002 * exactly match netdev_completed_queue() @bytes
3004 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3006 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3009 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3010 unsigned int pkts, unsigned int bytes)
3013 if (unlikely(!bytes))
3016 dql_completed(&dev_queue->dql, bytes);
3019 * Without the memory barrier there is a small possiblity that
3020 * netdev_tx_sent_queue will miss the update and cause the queue to
3021 * be stopped forever
3025 if (dql_avail(&dev_queue->dql) < 0)
3028 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3029 netif_schedule_queue(dev_queue);
3034 * netdev_completed_queue - report bytes and packets completed by device
3035 * @dev: network device
3036 * @pkts: actual number of packets sent over the medium
3037 * @bytes: actual number of bytes sent over the medium
3039 * Report the number of bytes and packets transmitted by the network device
3040 * hardware queue over the physical medium, @bytes must exactly match the
3041 * @bytes amount passed to netdev_sent_queue()
3043 static inline void netdev_completed_queue(struct net_device *dev,
3044 unsigned int pkts, unsigned int bytes)
3046 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3049 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3052 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3058 * netdev_reset_queue - reset the packets and bytes count of a network device
3059 * @dev_queue: network device
3061 * Reset the bytes and packet count of a network device and clear the
3062 * software flow control OFF bit for this network device
3064 static inline void netdev_reset_queue(struct net_device *dev_queue)
3066 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3070 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3071 * @dev: network device
3072 * @queue_index: given tx queue index
3074 * Returns 0 if given tx queue index >= number of device tx queues,
3075 * otherwise returns the originally passed tx queue index.
3077 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3079 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3080 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3081 dev->name, queue_index,
3082 dev->real_num_tx_queues);
3090 * netif_running - test if up
3091 * @dev: network device
3093 * Test if the device has been brought up.
3095 static inline bool netif_running(const struct net_device *dev)
3097 return test_bit(__LINK_STATE_START, &dev->state);
3101 * Routines to manage the subqueues on a device. We only need start,
3102 * stop, and a check if it's stopped. All other device management is
3103 * done at the overall netdevice level.
3104 * Also test the device if we're multiqueue.
3108 * netif_start_subqueue - allow sending packets on subqueue
3109 * @dev: network device
3110 * @queue_index: sub queue index
3112 * Start individual transmit queue of a device with multiple transmit queues.
3114 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3116 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3118 netif_tx_start_queue(txq);
3122 * netif_stop_subqueue - stop sending packets on subqueue
3123 * @dev: network device
3124 * @queue_index: sub queue index
3126 * Stop individual transmit queue of a device with multiple transmit queues.
3128 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3130 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3131 netif_tx_stop_queue(txq);
3135 * netif_subqueue_stopped - test status of subqueue
3136 * @dev: network device
3137 * @queue_index: sub queue index
3139 * Check individual transmit queue of a device with multiple transmit queues.
3141 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3144 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3146 return netif_tx_queue_stopped(txq);
3149 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3150 struct sk_buff *skb)
3152 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3155 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3158 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3161 static inline int netif_set_xps_queue(struct net_device *dev,
3162 const struct cpumask *mask,
3169 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3170 unsigned int num_tx_queues);
3173 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3174 * as a distribution range limit for the returned value.
3176 static inline u16 skb_tx_hash(const struct net_device *dev,
3177 struct sk_buff *skb)
3179 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3183 * netif_is_multiqueue - test if device has multiple transmit queues
3184 * @dev: network device
3186 * Check if device has multiple transmit queues
3188 static inline bool netif_is_multiqueue(const struct net_device *dev)
3190 return dev->num_tx_queues > 1;
3193 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3196 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3198 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3206 static inline unsigned int get_netdev_rx_queue_index(
3207 struct netdev_rx_queue *queue)
3209 struct net_device *dev = queue->dev;
3210 int index = queue - dev->_rx;
3212 BUG_ON(index >= dev->num_rx_queues);
3217 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3218 int netif_get_num_default_rss_queues(void);
3220 enum skb_free_reason {
3221 SKB_REASON_CONSUMED,
3225 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3226 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3229 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3230 * interrupt context or with hardware interrupts being disabled.
3231 * (in_irq() || irqs_disabled())
3233 * We provide four helpers that can be used in following contexts :
3235 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3236 * replacing kfree_skb(skb)
3238 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3239 * Typically used in place of consume_skb(skb) in TX completion path
3241 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3242 * replacing kfree_skb(skb)
3244 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3245 * and consumed a packet. Used in place of consume_skb(skb)
3247 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3249 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3252 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3254 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3257 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3259 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3262 static inline void dev_consume_skb_any(struct sk_buff *skb)
3264 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3267 int netif_rx(struct sk_buff *skb);
3268 int netif_rx_ni(struct sk_buff *skb);
3269 int netif_receive_skb(struct sk_buff *skb);
3270 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3271 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3272 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3273 gro_result_t napi_gro_frags(struct napi_struct *napi);
3274 struct packet_offload *gro_find_receive_by_type(__be16 type);
3275 struct packet_offload *gro_find_complete_by_type(__be16 type);
3277 static inline void napi_free_frags(struct napi_struct *napi)
3279 kfree_skb(napi->skb);
3283 bool netdev_is_rx_handler_busy(struct net_device *dev);
3284 int netdev_rx_handler_register(struct net_device *dev,
3285 rx_handler_func_t *rx_handler,
3286 void *rx_handler_data);
3287 void netdev_rx_handler_unregister(struct net_device *dev);
3289 bool dev_valid_name(const char *name);
3290 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3291 int dev_ethtool(struct net *net, struct ifreq *);
3292 unsigned int dev_get_flags(const struct net_device *);
3293 int __dev_change_flags(struct net_device *, unsigned int flags);
3294 int dev_change_flags(struct net_device *, unsigned int);
3295 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3296 unsigned int gchanges);
3297 int dev_change_name(struct net_device *, const char *);
3298 int dev_set_alias(struct net_device *, const char *, size_t);
3299 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3300 int dev_set_mtu(struct net_device *, int);
3301 void dev_set_group(struct net_device *, int);
3302 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3303 int dev_change_carrier(struct net_device *, bool new_carrier);
3304 int dev_get_phys_port_id(struct net_device *dev,
3305 struct netdev_phys_item_id *ppid);
3306 int dev_get_phys_port_name(struct net_device *dev,
3307 char *name, size_t len);
3308 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3309 int dev_change_xdp_fd(struct net_device *dev, int fd);
3310 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3311 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3312 struct netdev_queue *txq, int *ret);
3313 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3314 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3315 bool is_skb_forwardable(const struct net_device *dev,
3316 const struct sk_buff *skb);
3318 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3320 extern int netdev_budget;
3322 /* Called by rtnetlink.c:rtnl_unlock() */
3323 void netdev_run_todo(void);
3326 * dev_put - release reference to device
3327 * @dev: network device
3329 * Release reference to device to allow it to be freed.
3331 static inline void dev_put(struct net_device *dev)
3333 this_cpu_dec(*dev->pcpu_refcnt);
3337 * dev_hold - get reference to device
3338 * @dev: network device
3340 * Hold reference to device to keep it from being freed.
3342 static inline void dev_hold(struct net_device *dev)
3344 this_cpu_inc(*dev->pcpu_refcnt);
3347 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3348 * and _off may be called from IRQ context, but it is caller
3349 * who is responsible for serialization of these calls.
3351 * The name carrier is inappropriate, these functions should really be
3352 * called netif_lowerlayer_*() because they represent the state of any
3353 * kind of lower layer not just hardware media.
3356 void linkwatch_init_dev(struct net_device *dev);
3357 void linkwatch_fire_event(struct net_device *dev);
3358 void linkwatch_forget_dev(struct net_device *dev);
3361 * netif_carrier_ok - test if carrier present
3362 * @dev: network device
3364 * Check if carrier is present on device
3366 static inline bool netif_carrier_ok(const struct net_device *dev)
3368 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3371 unsigned long dev_trans_start(struct net_device *dev);
3373 void __netdev_watchdog_up(struct net_device *dev);
3375 void netif_carrier_on(struct net_device *dev);
3377 void netif_carrier_off(struct net_device *dev);
3380 * netif_dormant_on - mark device as dormant.
3381 * @dev: network device
3383 * Mark device as dormant (as per RFC2863).
3385 * The dormant state indicates that the relevant interface is not
3386 * actually in a condition to pass packets (i.e., it is not 'up') but is
3387 * in a "pending" state, waiting for some external event. For "on-
3388 * demand" interfaces, this new state identifies the situation where the
3389 * interface is waiting for events to place it in the up state.
3391 static inline void netif_dormant_on(struct net_device *dev)
3393 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3394 linkwatch_fire_event(dev);
3398 * netif_dormant_off - set device as not dormant.
3399 * @dev: network device
3401 * Device is not in dormant state.
3403 static inline void netif_dormant_off(struct net_device *dev)
3405 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3406 linkwatch_fire_event(dev);
3410 * netif_dormant - test if carrier present
3411 * @dev: network device
3413 * Check if carrier is present on device
3415 static inline bool netif_dormant(const struct net_device *dev)
3417 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3422 * netif_oper_up - test if device is operational
3423 * @dev: network device
3425 * Check if carrier is operational
3427 static inline bool netif_oper_up(const struct net_device *dev)
3429 return (dev->operstate == IF_OPER_UP ||
3430 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3434 * netif_device_present - is device available or removed
3435 * @dev: network device
3437 * Check if device has not been removed from system.
3439 static inline bool netif_device_present(struct net_device *dev)
3441 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3444 void netif_device_detach(struct net_device *dev);
3446 void netif_device_attach(struct net_device *dev);
3449 * Network interface message level settings
3453 NETIF_MSG_DRV = 0x0001,
3454 NETIF_MSG_PROBE = 0x0002,
3455 NETIF_MSG_LINK = 0x0004,
3456 NETIF_MSG_TIMER = 0x0008,
3457 NETIF_MSG_IFDOWN = 0x0010,
3458 NETIF_MSG_IFUP = 0x0020,
3459 NETIF_MSG_RX_ERR = 0x0040,
3460 NETIF_MSG_TX_ERR = 0x0080,
3461 NETIF_MSG_TX_QUEUED = 0x0100,
3462 NETIF_MSG_INTR = 0x0200,
3463 NETIF_MSG_TX_DONE = 0x0400,
3464 NETIF_MSG_RX_STATUS = 0x0800,
3465 NETIF_MSG_PKTDATA = 0x1000,
3466 NETIF_MSG_HW = 0x2000,
3467 NETIF_MSG_WOL = 0x4000,
3470 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3471 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3472 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3473 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3474 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3475 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3476 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3477 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3478 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3479 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3480 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3481 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3482 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3483 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3484 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3486 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3489 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3490 return default_msg_enable_bits;
3491 if (debug_value == 0) /* no output */
3493 /* set low N bits */
3494 return (1 << debug_value) - 1;
3497 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3499 spin_lock(&txq->_xmit_lock);
3500 txq->xmit_lock_owner = cpu;
3503 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3505 spin_lock_bh(&txq->_xmit_lock);
3506 txq->xmit_lock_owner = smp_processor_id();
3509 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3511 bool ok = spin_trylock(&txq->_xmit_lock);
3513 txq->xmit_lock_owner = smp_processor_id();
3517 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3519 txq->xmit_lock_owner = -1;
3520 spin_unlock(&txq->_xmit_lock);
3523 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3525 txq->xmit_lock_owner = -1;
3526 spin_unlock_bh(&txq->_xmit_lock);
3529 static inline void txq_trans_update(struct netdev_queue *txq)
3531 if (txq->xmit_lock_owner != -1)
3532 txq->trans_start = jiffies;
3535 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3536 static inline void netif_trans_update(struct net_device *dev)
3538 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3540 if (txq->trans_start != jiffies)
3541 txq->trans_start = jiffies;
3545 * netif_tx_lock - grab network device transmit lock
3546 * @dev: network device
3548 * Get network device transmit lock
3550 static inline void netif_tx_lock(struct net_device *dev)
3555 spin_lock(&dev->tx_global_lock);
3556 cpu = smp_processor_id();
3557 for (i = 0; i < dev->num_tx_queues; i++) {
3558 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3560 /* We are the only thread of execution doing a
3561 * freeze, but we have to grab the _xmit_lock in
3562 * order to synchronize with threads which are in
3563 * the ->hard_start_xmit() handler and already
3564 * checked the frozen bit.
3566 __netif_tx_lock(txq, cpu);
3567 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3568 __netif_tx_unlock(txq);
3572 static inline void netif_tx_lock_bh(struct net_device *dev)
3578 static inline void netif_tx_unlock(struct net_device *dev)
3582 for (i = 0; i < dev->num_tx_queues; i++) {
3583 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3585 /* No need to grab the _xmit_lock here. If the
3586 * queue is not stopped for another reason, we
3589 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3590 netif_schedule_queue(txq);
3592 spin_unlock(&dev->tx_global_lock);
3595 static inline void netif_tx_unlock_bh(struct net_device *dev)
3597 netif_tx_unlock(dev);
3601 #define HARD_TX_LOCK(dev, txq, cpu) { \
3602 if ((dev->features & NETIF_F_LLTX) == 0) { \
3603 __netif_tx_lock(txq, cpu); \
3607 #define HARD_TX_TRYLOCK(dev, txq) \
3608 (((dev->features & NETIF_F_LLTX) == 0) ? \
3609 __netif_tx_trylock(txq) : \
3612 #define HARD_TX_UNLOCK(dev, txq) { \
3613 if ((dev->features & NETIF_F_LLTX) == 0) { \
3614 __netif_tx_unlock(txq); \
3618 static inline void netif_tx_disable(struct net_device *dev)
3624 cpu = smp_processor_id();
3625 for (i = 0; i < dev->num_tx_queues; i++) {
3626 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3628 __netif_tx_lock(txq, cpu);
3629 netif_tx_stop_queue(txq);
3630 __netif_tx_unlock(txq);
3635 static inline void netif_addr_lock(struct net_device *dev)
3637 spin_lock(&dev->addr_list_lock);
3640 static inline void netif_addr_lock_nested(struct net_device *dev)
3642 int subclass = SINGLE_DEPTH_NESTING;
3644 if (dev->netdev_ops->ndo_get_lock_subclass)
3645 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3647 spin_lock_nested(&dev->addr_list_lock, subclass);
3650 static inline void netif_addr_lock_bh(struct net_device *dev)
3652 spin_lock_bh(&dev->addr_list_lock);
3655 static inline void netif_addr_unlock(struct net_device *dev)
3657 spin_unlock(&dev->addr_list_lock);
3660 static inline void netif_addr_unlock_bh(struct net_device *dev)
3662 spin_unlock_bh(&dev->addr_list_lock);
3666 * dev_addrs walker. Should be used only for read access. Call with
3667 * rcu_read_lock held.
3669 #define for_each_dev_addr(dev, ha) \
3670 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3672 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3674 void ether_setup(struct net_device *dev);
3676 /* Support for loadable net-drivers */
3677 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3678 unsigned char name_assign_type,
3679 void (*setup)(struct net_device *),
3680 unsigned int txqs, unsigned int rxqs);
3681 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3682 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3684 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3685 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3688 int register_netdev(struct net_device *dev);
3689 void unregister_netdev(struct net_device *dev);
3691 /* General hardware address lists handling functions */
3692 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3693 struct netdev_hw_addr_list *from_list, int addr_len);
3694 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3695 struct netdev_hw_addr_list *from_list, int addr_len);
3696 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3697 struct net_device *dev,
3698 int (*sync)(struct net_device *, const unsigned char *),
3699 int (*unsync)(struct net_device *,
3700 const unsigned char *));
3701 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3702 struct net_device *dev,
3703 int (*unsync)(struct net_device *,
3704 const unsigned char *));
3705 void __hw_addr_init(struct netdev_hw_addr_list *list);
3707 /* Functions used for device addresses handling */
3708 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3709 unsigned char addr_type);
3710 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3711 unsigned char addr_type);
3712 void dev_addr_flush(struct net_device *dev);
3713 int dev_addr_init(struct net_device *dev);
3715 /* Functions used for unicast addresses handling */
3716 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3717 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3718 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3719 int dev_uc_sync(struct net_device *to, struct net_device *from);
3720 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3721 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3722 void dev_uc_flush(struct net_device *dev);
3723 void dev_uc_init(struct net_device *dev);
3726 * __dev_uc_sync - Synchonize device's unicast list
3727 * @dev: device to sync
3728 * @sync: function to call if address should be added
3729 * @unsync: function to call if address should be removed
3731 * Add newly added addresses to the interface, and release
3732 * addresses that have been deleted.
3734 static inline int __dev_uc_sync(struct net_device *dev,
3735 int (*sync)(struct net_device *,
3736 const unsigned char *),
3737 int (*unsync)(struct net_device *,
3738 const unsigned char *))
3740 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3744 * __dev_uc_unsync - Remove synchronized addresses from device
3745 * @dev: device to sync
3746 * @unsync: function to call if address should be removed
3748 * Remove all addresses that were added to the device by dev_uc_sync().
3750 static inline void __dev_uc_unsync(struct net_device *dev,
3751 int (*unsync)(struct net_device *,
3752 const unsigned char *))
3754 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3757 /* Functions used for multicast addresses handling */
3758 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3759 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3760 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3761 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3762 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3763 int dev_mc_sync(struct net_device *to, struct net_device *from);
3764 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3765 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3766 void dev_mc_flush(struct net_device *dev);
3767 void dev_mc_init(struct net_device *dev);
3770 * __dev_mc_sync - Synchonize device's multicast list
3771 * @dev: device to sync
3772 * @sync: function to call if address should be added
3773 * @unsync: function to call if address should be removed
3775 * Add newly added addresses to the interface, and release
3776 * addresses that have been deleted.
3778 static inline int __dev_mc_sync(struct net_device *dev,
3779 int (*sync)(struct net_device *,
3780 const unsigned char *),
3781 int (*unsync)(struct net_device *,
3782 const unsigned char *))
3784 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3788 * __dev_mc_unsync - Remove synchronized addresses from device
3789 * @dev: device to sync
3790 * @unsync: function to call if address should be removed
3792 * Remove all addresses that were added to the device by dev_mc_sync().
3794 static inline void __dev_mc_unsync(struct net_device *dev,
3795 int (*unsync)(struct net_device *,
3796 const unsigned char *))
3798 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3801 /* Functions used for secondary unicast and multicast support */
3802 void dev_set_rx_mode(struct net_device *dev);
3803 void __dev_set_rx_mode(struct net_device *dev);
3804 int dev_set_promiscuity(struct net_device *dev, int inc);
3805 int dev_set_allmulti(struct net_device *dev, int inc);
3806 void netdev_state_change(struct net_device *dev);
3807 void netdev_notify_peers(struct net_device *dev);
3808 void netdev_features_change(struct net_device *dev);
3809 /* Load a device via the kmod */
3810 void dev_load(struct net *net, const char *name);
3811 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3812 struct rtnl_link_stats64 *storage);
3813 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3814 const struct net_device_stats *netdev_stats);
3816 extern int netdev_max_backlog;
3817 extern int netdev_tstamp_prequeue;
3818 extern int weight_p;
3820 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3821 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3822 struct list_head **iter);
3823 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3824 struct list_head **iter);
3826 /* iterate through upper list, must be called under RCU read lock */
3827 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3828 for (iter = &(dev)->adj_list.upper, \
3829 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3831 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3833 /* iterate through upper list, must be called under RCU read lock */
3834 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3835 for (iter = &(dev)->all_adj_list.upper, \
3836 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3838 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3840 void *netdev_lower_get_next_private(struct net_device *dev,
3841 struct list_head **iter);
3842 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3843 struct list_head **iter);
3845 #define netdev_for_each_lower_private(dev, priv, iter) \
3846 for (iter = (dev)->adj_list.lower.next, \
3847 priv = netdev_lower_get_next_private(dev, &(iter)); \
3849 priv = netdev_lower_get_next_private(dev, &(iter)))
3851 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3852 for (iter = &(dev)->adj_list.lower, \
3853 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3855 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3857 void *netdev_lower_get_next(struct net_device *dev,
3858 struct list_head **iter);
3860 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3861 for (iter = (dev)->adj_list.lower.next, \
3862 ldev = netdev_lower_get_next(dev, &(iter)); \
3864 ldev = netdev_lower_get_next(dev, &(iter)))
3866 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3867 struct list_head **iter);
3868 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3869 struct list_head **iter);
3871 #define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3872 for (iter = (dev)->all_adj_list.lower.next, \
3873 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3875 ldev = netdev_all_lower_get_next(dev, &(iter)))
3877 #define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
3878 for (iter = (dev)->all_adj_list.lower.next, \
3879 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3881 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3883 void *netdev_adjacent_get_private(struct list_head *adj_list);
3884 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3885 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3886 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3887 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3888 int netdev_master_upper_dev_link(struct net_device *dev,
3889 struct net_device *upper_dev,
3890 void *upper_priv, void *upper_info);
3891 void netdev_upper_dev_unlink(struct net_device *dev,
3892 struct net_device *upper_dev);
3893 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3894 void *netdev_lower_dev_get_private(struct net_device *dev,
3895 struct net_device *lower_dev);
3896 void netdev_lower_state_changed(struct net_device *lower_dev,
3897 void *lower_state_info);
3898 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3899 struct neighbour *n);
3900 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3901 struct neighbour *n);
3903 /* RSS keys are 40 or 52 bytes long */
3904 #define NETDEV_RSS_KEY_LEN 52
3905 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3906 void netdev_rss_key_fill(void *buffer, size_t len);
3908 int dev_get_nest_level(struct net_device *dev);
3909 int skb_checksum_help(struct sk_buff *skb);
3910 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3911 netdev_features_t features, bool tx_path);
3912 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3913 netdev_features_t features);
3915 struct netdev_bonding_info {
3920 struct netdev_notifier_bonding_info {
3921 struct netdev_notifier_info info; /* must be first */
3922 struct netdev_bonding_info bonding_info;
3925 void netdev_bonding_info_change(struct net_device *dev,
3926 struct netdev_bonding_info *bonding_info);
3929 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3931 return __skb_gso_segment(skb, features, true);
3933 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3935 static inline bool can_checksum_protocol(netdev_features_t features,
3938 if (protocol == htons(ETH_P_FCOE))
3939 return !!(features & NETIF_F_FCOE_CRC);
3941 /* Assume this is an IP checksum (not SCTP CRC) */
3943 if (features & NETIF_F_HW_CSUM) {
3944 /* Can checksum everything */
3949 case htons(ETH_P_IP):
3950 return !!(features & NETIF_F_IP_CSUM);
3951 case htons(ETH_P_IPV6):
3952 return !!(features & NETIF_F_IPV6_CSUM);
3958 /* Map an ethertype into IP protocol if possible */
3959 static inline int eproto_to_ipproto(int eproto)
3962 case htons(ETH_P_IP):
3964 case htons(ETH_P_IPV6):
3965 return IPPROTO_IPV6;
3972 void netdev_rx_csum_fault(struct net_device *dev);
3974 static inline void netdev_rx_csum_fault(struct net_device *dev)
3978 /* rx skb timestamps */
3979 void net_enable_timestamp(void);
3980 void net_disable_timestamp(void);
3982 #ifdef CONFIG_PROC_FS
3983 int __init dev_proc_init(void);
3985 #define dev_proc_init() 0
3988 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3989 struct sk_buff *skb, struct net_device *dev,
3992 skb->xmit_more = more ? 1 : 0;
3993 return ops->ndo_start_xmit(skb, dev);
3996 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3997 struct netdev_queue *txq, bool more)
3999 const struct net_device_ops *ops = dev->netdev_ops;
4002 rc = __netdev_start_xmit(ops, skb, dev, more);
4003 if (rc == NETDEV_TX_OK)
4004 txq_trans_update(txq);
4009 int netdev_class_create_file_ns(struct class_attribute *class_attr,
4011 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4014 static inline int netdev_class_create_file(struct class_attribute *class_attr)
4016 return netdev_class_create_file_ns(class_attr, NULL);
4019 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4021 netdev_class_remove_file_ns(class_attr, NULL);
4024 extern struct kobj_ns_type_operations net_ns_type_operations;
4026 const char *netdev_drivername(const struct net_device *dev);
4028 void linkwatch_run_queue(void);
4030 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4031 netdev_features_t f2)
4033 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4034 if (f1 & NETIF_F_HW_CSUM)
4035 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4037 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4043 static inline netdev_features_t netdev_get_wanted_features(
4044 struct net_device *dev)
4046 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4048 netdev_features_t netdev_increment_features(netdev_features_t all,
4049 netdev_features_t one, netdev_features_t mask);
4051 /* Allow TSO being used on stacked device :
4052 * Performing the GSO segmentation before last device
4053 * is a performance improvement.
4055 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4056 netdev_features_t mask)
4058 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4061 int __netdev_update_features(struct net_device *dev);
4062 void netdev_update_features(struct net_device *dev);
4063 void netdev_change_features(struct net_device *dev);
4065 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4066 struct net_device *dev);
4068 netdev_features_t passthru_features_check(struct sk_buff *skb,
4069 struct net_device *dev,
4070 netdev_features_t features);
4071 netdev_features_t netif_skb_features(struct sk_buff *skb);
4073 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4075 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4077 /* check flags correspondence */
4078 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4079 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4080 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4081 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4082 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4083 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4084 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4085 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4086 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4087 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4088 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4089 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4090 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4091 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4092 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4093 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4095 return (features & feature) == feature;
4098 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4100 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4101 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4104 static inline bool netif_needs_gso(struct sk_buff *skb,
4105 netdev_features_t features)
4107 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4108 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4109 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4112 static inline void netif_set_gso_max_size(struct net_device *dev,
4115 dev->gso_max_size = size;
4118 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4119 int pulled_hlen, u16 mac_offset,
4122 skb->protocol = protocol;
4123 skb->encapsulation = 1;
4124 skb_push(skb, pulled_hlen);
4125 skb_reset_transport_header(skb);
4126 skb->mac_header = mac_offset;
4127 skb->network_header = skb->mac_header + mac_len;
4128 skb->mac_len = mac_len;
4131 static inline bool netif_is_macsec(const struct net_device *dev)
4133 return dev->priv_flags & IFF_MACSEC;
4136 static inline bool netif_is_macvlan(const struct net_device *dev)
4138 return dev->priv_flags & IFF_MACVLAN;
4141 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4143 return dev->priv_flags & IFF_MACVLAN_PORT;
4146 static inline bool netif_is_ipvlan(const struct net_device *dev)
4148 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4151 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4153 return dev->priv_flags & IFF_IPVLAN_MASTER;
4156 static inline bool netif_is_bond_master(const struct net_device *dev)
4158 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4161 static inline bool netif_is_bond_slave(const struct net_device *dev)
4163 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4166 static inline bool netif_supports_nofcs(struct net_device *dev)
4168 return dev->priv_flags & IFF_SUPP_NOFCS;
4171 static inline bool netif_is_l3_master(const struct net_device *dev)
4173 return dev->priv_flags & IFF_L3MDEV_MASTER;
4176 static inline bool netif_is_l3_slave(const struct net_device *dev)
4178 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4181 static inline bool netif_is_bridge_master(const struct net_device *dev)
4183 return dev->priv_flags & IFF_EBRIDGE;
4186 static inline bool netif_is_bridge_port(const struct net_device *dev)
4188 return dev->priv_flags & IFF_BRIDGE_PORT;
4191 static inline bool netif_is_ovs_master(const struct net_device *dev)
4193 return dev->priv_flags & IFF_OPENVSWITCH;
4196 static inline bool netif_is_team_master(const struct net_device *dev)
4198 return dev->priv_flags & IFF_TEAM;
4201 static inline bool netif_is_team_port(const struct net_device *dev)
4203 return dev->priv_flags & IFF_TEAM_PORT;
4206 static inline bool netif_is_lag_master(const struct net_device *dev)
4208 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4211 static inline bool netif_is_lag_port(const struct net_device *dev)
4213 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4216 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4218 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4221 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4222 static inline void netif_keep_dst(struct net_device *dev)
4224 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4227 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4228 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4230 /* TODO: reserve and use an additional IFF bit, if we get more users */
4231 return dev->priv_flags & IFF_MACSEC;
4234 extern struct pernet_operations __net_initdata loopback_net_ops;
4236 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4238 /* netdev_printk helpers, similar to dev_printk */
4240 static inline const char *netdev_name(const struct net_device *dev)
4242 if (!dev->name[0] || strchr(dev->name, '%'))
4243 return "(unnamed net_device)";
4247 static inline const char *netdev_reg_state(const struct net_device *dev)
4249 switch (dev->reg_state) {
4250 case NETREG_UNINITIALIZED: return " (uninitialized)";
4251 case NETREG_REGISTERED: return "";
4252 case NETREG_UNREGISTERING: return " (unregistering)";
4253 case NETREG_UNREGISTERED: return " (unregistered)";
4254 case NETREG_RELEASED: return " (released)";
4255 case NETREG_DUMMY: return " (dummy)";
4258 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4259 return " (unknown)";
4263 void netdev_printk(const char *level, const struct net_device *dev,
4264 const char *format, ...);
4266 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4268 void netdev_alert(const struct net_device *dev, const char *format, ...);
4270 void netdev_crit(const struct net_device *dev, const char *format, ...);
4272 void netdev_err(const struct net_device *dev, const char *format, ...);
4274 void netdev_warn(const struct net_device *dev, const char *format, ...);
4276 void netdev_notice(const struct net_device *dev, const char *format, ...);
4278 void netdev_info(const struct net_device *dev, const char *format, ...);
4280 #define MODULE_ALIAS_NETDEV(device) \
4281 MODULE_ALIAS("netdev-" device)
4283 #if defined(CONFIG_DYNAMIC_DEBUG)
4284 #define netdev_dbg(__dev, format, args...) \
4286 dynamic_netdev_dbg(__dev, format, ##args); \
4288 #elif defined(DEBUG)
4289 #define netdev_dbg(__dev, format, args...) \
4290 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4292 #define netdev_dbg(__dev, format, args...) \
4295 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4299 #if defined(VERBOSE_DEBUG)
4300 #define netdev_vdbg netdev_dbg
4303 #define netdev_vdbg(dev, format, args...) \
4306 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4312 * netdev_WARN() acts like dev_printk(), but with the key difference
4313 * of using a WARN/WARN_ON to get the message out, including the
4314 * file/line information and a backtrace.
4316 #define netdev_WARN(dev, format, args...) \
4317 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4318 netdev_reg_state(dev), ##args)
4320 /* netif printk helpers, similar to netdev_printk */
4322 #define netif_printk(priv, type, level, dev, fmt, args...) \
4324 if (netif_msg_##type(priv)) \
4325 netdev_printk(level, (dev), fmt, ##args); \
4328 #define netif_level(level, priv, type, dev, fmt, args...) \
4330 if (netif_msg_##type(priv)) \
4331 netdev_##level(dev, fmt, ##args); \
4334 #define netif_emerg(priv, type, dev, fmt, args...) \
4335 netif_level(emerg, priv, type, dev, fmt, ##args)
4336 #define netif_alert(priv, type, dev, fmt, args...) \
4337 netif_level(alert, priv, type, dev, fmt, ##args)
4338 #define netif_crit(priv, type, dev, fmt, args...) \
4339 netif_level(crit, priv, type, dev, fmt, ##args)
4340 #define netif_err(priv, type, dev, fmt, args...) \
4341 netif_level(err, priv, type, dev, fmt, ##args)
4342 #define netif_warn(priv, type, dev, fmt, args...) \
4343 netif_level(warn, priv, type, dev, fmt, ##args)
4344 #define netif_notice(priv, type, dev, fmt, args...) \
4345 netif_level(notice, priv, type, dev, fmt, ##args)
4346 #define netif_info(priv, type, dev, fmt, args...) \
4347 netif_level(info, priv, type, dev, fmt, ##args)
4349 #if defined(CONFIG_DYNAMIC_DEBUG)
4350 #define netif_dbg(priv, type, netdev, format, args...) \
4352 if (netif_msg_##type(priv)) \
4353 dynamic_netdev_dbg(netdev, format, ##args); \
4355 #elif defined(DEBUG)
4356 #define netif_dbg(priv, type, dev, format, args...) \
4357 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4359 #define netif_dbg(priv, type, dev, format, args...) \
4362 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4367 #if defined(VERBOSE_DEBUG)
4368 #define netif_vdbg netif_dbg
4370 #define netif_vdbg(priv, type, dev, format, args...) \
4373 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4379 * The list of packet types we will receive (as opposed to discard)
4380 * and the routines to invoke.
4382 * Why 16. Because with 16 the only overlap we get on a hash of the
4383 * low nibble of the protocol value is RARP/SNAP/X.25.
4385 * NOTE: That is no longer true with the addition of VLAN tags. Not
4386 * sure which should go first, but I bet it won't make much
4387 * difference if we are running VLANs. The good news is that
4388 * this protocol won't be in the list unless compiled in, so
4389 * the average user (w/out VLANs) will not be adversely affected.
4405 #define PTYPE_HASH_SIZE (16)
4406 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4408 #endif /* _LINUX_NETDEVICE_H */