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
321 struct net_device *dev;
322 struct sk_buff *gro_list;
324 struct hrtimer timer;
325 struct list_head dev_list;
326 struct hlist_node napi_hash_node;
327 unsigned int napi_id;
331 NAPI_STATE_SCHED, /* Poll is scheduled */
332 NAPI_STATE_DISABLE, /* Disable pending */
333 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
334 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
335 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
336 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
340 NAPIF_STATE_SCHED = (1UL << NAPI_STATE_SCHED),
341 NAPIF_STATE_DISABLE = (1UL << NAPI_STATE_DISABLE),
342 NAPIF_STATE_NPSVC = (1UL << NAPI_STATE_NPSVC),
343 NAPIF_STATE_HASHED = (1UL << NAPI_STATE_HASHED),
344 NAPIF_STATE_NO_BUSY_POLL = (1UL << NAPI_STATE_NO_BUSY_POLL),
345 NAPIF_STATE_IN_BUSY_POLL = (1UL << NAPI_STATE_IN_BUSY_POLL),
355 typedef enum gro_result gro_result_t;
358 * enum rx_handler_result - Possible return values for rx_handlers.
359 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
361 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
362 * case skb->dev was changed by rx_handler.
363 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
364 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
366 * rx_handlers are functions called from inside __netif_receive_skb(), to do
367 * special processing of the skb, prior to delivery to protocol handlers.
369 * Currently, a net_device can only have a single rx_handler registered. Trying
370 * to register a second rx_handler will return -EBUSY.
372 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
373 * To unregister a rx_handler on a net_device, use
374 * netdev_rx_handler_unregister().
376 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
379 * If the rx_handler consumed the skb in some way, it should return
380 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
381 * the skb to be delivered in some other way.
383 * If the rx_handler changed skb->dev, to divert the skb to another
384 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
385 * new device will be called if it exists.
387 * If the rx_handler decides the skb should be ignored, it should return
388 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
389 * are registered on exact device (ptype->dev == skb->dev).
391 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
392 * delivered, it should return RX_HANDLER_PASS.
394 * A device without a registered rx_handler will behave as if rx_handler
395 * returned RX_HANDLER_PASS.
398 enum rx_handler_result {
404 typedef enum rx_handler_result rx_handler_result_t;
405 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
407 void __napi_schedule(struct napi_struct *n);
408 void __napi_schedule_irqoff(struct napi_struct *n);
410 static inline bool napi_disable_pending(struct napi_struct *n)
412 return test_bit(NAPI_STATE_DISABLE, &n->state);
416 * napi_schedule_prep - check if NAPI can be scheduled
419 * Test if NAPI routine is already running, and if not mark
420 * it as running. This is used as a condition variable to
421 * insure only one NAPI poll instance runs. We also make
422 * sure there is no pending NAPI disable.
424 static inline bool napi_schedule_prep(struct napi_struct *n)
426 return !napi_disable_pending(n) &&
427 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
431 * napi_schedule - schedule NAPI poll
434 * Schedule NAPI poll routine to be called if it is not already
437 static inline void napi_schedule(struct napi_struct *n)
439 if (napi_schedule_prep(n))
444 * napi_schedule_irqoff - schedule NAPI poll
447 * Variant of napi_schedule(), assuming hard irqs are masked.
449 static inline void napi_schedule_irqoff(struct napi_struct *n)
451 if (napi_schedule_prep(n))
452 __napi_schedule_irqoff(n);
455 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
456 static inline bool napi_reschedule(struct napi_struct *napi)
458 if (napi_schedule_prep(napi)) {
459 __napi_schedule(napi);
465 bool __napi_complete(struct napi_struct *n);
466 bool napi_complete_done(struct napi_struct *n, int work_done);
468 * napi_complete - NAPI processing complete
471 * Mark NAPI processing as complete.
472 * Consider using napi_complete_done() instead.
473 * Return false if device should avoid rearming interrupts.
475 static inline bool napi_complete(struct napi_struct *n)
477 return napi_complete_done(n, 0);
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: NAPI context
484 * Warning: caller must observe RCU grace period
485 * before freeing memory containing @napi, if
486 * this function returns true.
487 * Note: core networking stack automatically calls it
488 * from netif_napi_del().
489 * Drivers might want to call this helper to combine all
490 * the needed RCU grace periods into a single one.
492 bool napi_hash_del(struct napi_struct *napi);
495 * napi_disable - prevent NAPI from scheduling
498 * Stop NAPI from being scheduled on this context.
499 * Waits till any outstanding processing completes.
501 void napi_disable(struct napi_struct *n);
504 * napi_enable - enable NAPI scheduling
507 * Resume NAPI from being scheduled on this context.
508 * Must be paired with napi_disable.
510 static inline void napi_enable(struct napi_struct *n)
512 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
513 smp_mb__before_atomic();
514 clear_bit(NAPI_STATE_SCHED, &n->state);
515 clear_bit(NAPI_STATE_NPSVC, &n->state);
519 * napi_synchronize - wait until NAPI is not running
522 * Wait until NAPI is done being scheduled on this context.
523 * Waits till any outstanding processing completes but
524 * does not disable future activations.
526 static inline void napi_synchronize(const struct napi_struct *n)
528 if (IS_ENABLED(CONFIG_SMP))
529 while (test_bit(NAPI_STATE_SCHED, &n->state))
535 enum netdev_queue_state_t {
536 __QUEUE_STATE_DRV_XOFF,
537 __QUEUE_STATE_STACK_XOFF,
538 __QUEUE_STATE_FROZEN,
541 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
542 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
543 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
545 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
546 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
552 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
553 * netif_tx_* functions below are used to manipulate this flag. The
554 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
555 * queue independently. The netif_xmit_*stopped functions below are called
556 * to check if the queue has been stopped by the driver or stack (either
557 * of the XOFF bits are set in the state). Drivers should not need to call
558 * netif_xmit*stopped functions, they should only be using netif_tx_*.
561 struct netdev_queue {
565 struct net_device *dev;
566 struct Qdisc __rcu *qdisc;
567 struct Qdisc *qdisc_sleeping;
571 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
574 unsigned long tx_maxrate;
576 * Number of TX timeouts for this queue
577 * (/sys/class/net/DEV/Q/trans_timeout)
579 unsigned long trans_timeout;
583 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
586 * Time (in jiffies) of last Tx
588 unsigned long trans_start;
595 } ____cacheline_aligned_in_smp;
597 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
615 * This structure holds an RPS map which can be of variable length. The
616 * map is an array of CPUs.
623 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627 * tail pointer for that CPU's input queue at the time of last enqueue, and
628 * a hardware filter index.
630 struct rps_dev_flow {
633 unsigned int last_qtail;
635 #define RPS_NO_FILTER 0xffff
638 * The rps_dev_flow_table structure contains a table of flow mappings.
640 struct rps_dev_flow_table {
643 struct rps_dev_flow flows[0];
645 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
646 ((_num) * sizeof(struct rps_dev_flow)))
649 * The rps_sock_flow_table contains mappings of flows to the last CPU
650 * on which they were processed by the application (set in recvmsg).
651 * Each entry is a 32bit value. Upper part is the high-order bits
652 * of flow hash, lower part is CPU number.
653 * rps_cpu_mask is used to partition the space, depending on number of
654 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
656 * meaning we use 32-6=26 bits for the hash.
658 struct rps_sock_flow_table {
661 u32 ents[0] ____cacheline_aligned_in_smp;
663 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
665 #define RPS_NO_CPU 0xffff
667 extern u32 rps_cpu_mask;
668 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
674 unsigned int index = hash & table->mask;
675 u32 val = hash & ~rps_cpu_mask;
677 /* We only give a hint, preemption can change CPU under us */
678 val |= raw_smp_processor_id();
680 if (table->ents[index] != val)
681 table->ents[index] = val;
685 #ifdef CONFIG_RFS_ACCEL
686 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
689 #endif /* CONFIG_RPS */
691 /* This structure contains an instance of an RX queue. */
692 struct netdev_rx_queue {
694 struct rps_map __rcu *rps_map;
695 struct rps_dev_flow_table __rcu *rps_flow_table;
698 struct net_device *dev;
699 } ____cacheline_aligned_in_smp;
702 * RX queue sysfs structures and functions.
704 struct rx_queue_attribute {
705 struct attribute attr;
706 ssize_t (*show)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, char *buf);
708 ssize_t (*store)(struct netdev_rx_queue *queue,
709 struct rx_queue_attribute *attr, const char *buf, size_t len);
714 * This structure holds an XPS map which can be of variable length. The
715 * map is an array of queues.
719 unsigned int alloc_len;
723 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
724 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725 - sizeof(struct xps_map)) / sizeof(u16))
728 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 struct xps_dev_maps {
732 struct xps_map __rcu *cpu_map[0];
734 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
735 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
736 #endif /* CONFIG_XPS */
738 #define TC_MAX_QUEUE 16
739 #define TC_BITMASK 15
740 /* HW offloaded queuing disciplines txq count and offset maps */
741 struct netdev_tc_txq {
746 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748 * This structure is to hold information about the device
749 * configured to run FCoE protocol stack.
751 struct netdev_fcoe_hbainfo {
752 char manufacturer[64];
753 char serial_number[64];
754 char hardware_version[64];
755 char driver_version[64];
756 char optionrom_version[64];
757 char firmware_version[64];
759 char model_description[256];
763 #define MAX_PHYS_ITEM_ID_LEN 32
765 /* This structure holds a unique identifier to identify some
766 * physical item (port for example) used by a netdevice.
768 struct netdev_phys_item_id {
769 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
770 unsigned char id_len;
773 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
774 struct netdev_phys_item_id *b)
776 return a->id_len == b->id_len &&
777 memcmp(a->id, b->id, a->id_len) == 0;
780 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
781 struct sk_buff *skb);
783 /* These structures hold the attributes of qdisc and classifiers
784 * that are being passed to the netdevice through the setup_tc op.
794 struct tc_cls_u32_offload;
796 struct tc_to_netdev {
800 struct tc_cls_u32_offload *cls_u32;
801 struct tc_cls_flower_offload *cls_flower;
802 struct tc_cls_matchall_offload *cls_mall;
803 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)(const struct net_device *dev, 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,
950 * u8 qos, __be16 proto);
951 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
953 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
954 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
955 * int (*ndo_get_vf_config)(struct net_device *dev,
956 * int vf, struct ifla_vf_info *ivf);
957 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
958 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
959 * struct nlattr *port[]);
961 * Enable or disable the VF ability to query its RSS Redirection Table and
962 * Hash Key. This is needed since on some devices VF share this information
963 * with PF and querying it may introduce a theoretical security risk.
964 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
965 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
966 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
967 * Called to setup 'tc' number of traffic classes in the net device. This
968 * is always called from the stack with the rtnl lock held and netif tx
969 * queues stopped. This allows the netdevice to perform queue management
972 * Fiber Channel over Ethernet (FCoE) offload functions.
973 * int (*ndo_fcoe_enable)(struct net_device *dev);
974 * Called when the FCoE protocol stack wants to start using LLD for FCoE
975 * so the underlying device can perform whatever needed configuration or
976 * initialization to support acceleration of FCoE traffic.
978 * int (*ndo_fcoe_disable)(struct net_device *dev);
979 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
980 * so the underlying device can perform whatever needed clean-ups to
981 * stop supporting acceleration of FCoE traffic.
983 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
984 * struct scatterlist *sgl, unsigned int sgc);
985 * Called when the FCoE Initiator wants to initialize an I/O that
986 * is a possible candidate for Direct Data Placement (DDP). The LLD can
987 * perform necessary setup and returns 1 to indicate the device is set up
988 * successfully to perform DDP on this I/O, otherwise this returns 0.
990 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
991 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
992 * indicated by the FC exchange id 'xid', so the underlying device can
993 * clean up and reuse resources for later DDP requests.
995 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
996 * struct scatterlist *sgl, unsigned int sgc);
997 * Called when the FCoE Target wants to initialize an I/O that
998 * is a possible candidate for Direct Data Placement (DDP). The LLD can
999 * perform necessary setup and returns 1 to indicate the device is set up
1000 * successfully to perform DDP on this I/O, otherwise this returns 0.
1002 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1003 * struct netdev_fcoe_hbainfo *hbainfo);
1004 * Called when the FCoE Protocol stack wants information on the underlying
1005 * device. This information is utilized by the FCoE protocol stack to
1006 * register attributes with Fiber Channel management service as per the
1007 * FC-GS Fabric Device Management Information(FDMI) specification.
1009 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1010 * Called when the underlying device wants to override default World Wide
1011 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1012 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1013 * protocol stack to use.
1016 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1017 * u16 rxq_index, u32 flow_id);
1018 * Set hardware filter for RFS. rxq_index is the target queue index;
1019 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1020 * Return the filter ID on success, or a negative error code.
1022 * Slave management functions (for bridge, bonding, etc).
1023 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1024 * Called to make another netdev an underling.
1026 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1027 * Called to release previously enslaved netdev.
1029 * Feature/offload setting functions.
1030 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1031 * Called to update device configuration to new features. Passed
1032 * feature set might be less than what was returned by ndo_fix_features()).
1033 * Must return >0 or -errno if it changed dev->features itself.
1035 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1036 * struct net_device *dev,
1037 * const unsigned char *addr, u16 vid, u16 flags)
1038 * Adds an FDB entry to dev for addr.
1039 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1040 * struct net_device *dev,
1041 * const unsigned char *addr, u16 vid)
1042 * Deletes the FDB entry from dev coresponding to addr.
1043 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1044 * struct net_device *dev, struct net_device *filter_dev,
1046 * Used to add FDB entries to dump requests. Implementers should add
1047 * entries to skb and update idx with the number of entries.
1049 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1051 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1052 * struct net_device *dev, u32 filter_mask,
1054 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1057 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1058 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1059 * which do not represent real hardware may define this to allow their
1060 * userspace components to manage their virtual carrier state. Devices
1061 * that determine carrier state from physical hardware properties (eg
1062 * network cables) or protocol-dependent mechanisms (eg
1063 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1065 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1066 * struct netdev_phys_item_id *ppid);
1067 * Called to get ID of physical port of this device. If driver does
1068 * not implement this, it is assumed that the hw is not able to have
1069 * multiple net devices on single physical port.
1071 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1072 * struct udp_tunnel_info *ti);
1073 * Called by UDP tunnel to notify a driver about the UDP port and socket
1074 * address family that a UDP tunnel is listnening to. It is called only
1075 * when a new port starts listening. The operation is protected by the
1078 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1079 * struct udp_tunnel_info *ti);
1080 * Called by UDP tunnel to notify the driver about a UDP port and socket
1081 * address family that the UDP tunnel is not listening to anymore. The
1082 * operation is protected by the RTNL.
1084 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1085 * struct net_device *dev)
1086 * Called by upper layer devices to accelerate switching or other
1087 * station functionality into hardware. 'pdev is the lowerdev
1088 * to use for the offload and 'dev' is the net device that will
1089 * back the offload. Returns a pointer to the private structure
1090 * the upper layer will maintain.
1091 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1092 * Called by upper layer device to delete the station created
1093 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1094 * the station and priv is the structure returned by the add
1096 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1097 * struct net_device *dev,
1099 * Callback to use for xmit over the accelerated station. This
1100 * is used in place of ndo_start_xmit on accelerated net
1102 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1103 * struct net_device *dev
1104 * netdev_features_t features);
1105 * Called by core transmit path to determine if device is capable of
1106 * performing offload operations on a given packet. This is to give
1107 * the device an opportunity to implement any restrictions that cannot
1108 * be otherwise expressed by feature flags. The check is called with
1109 * the set of features that the stack has calculated and it returns
1110 * those the driver believes to be appropriate.
1111 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1112 * int queue_index, u32 maxrate);
1113 * Called when a user wants to set a max-rate limitation of specific
1115 * int (*ndo_get_iflink)(const struct net_device *dev);
1116 * Called to get the iflink value of this device.
1117 * void (*ndo_change_proto_down)(struct net_device *dev,
1119 * This function is used to pass protocol port error state information
1120 * to the switch driver. The switch driver can react to the proto_down
1121 * by doing a phys down on the associated switch port.
1122 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1123 * This function is used to get egress tunnel information for given skb.
1124 * This is useful for retrieving outer tunnel header parameters while
1126 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1127 * This function is used to specify the headroom that the skb must
1128 * consider when allocation skb during packet reception. Setting
1129 * appropriate rx headroom value allows avoiding skb head copy on
1130 * forward. Setting a negative value resets the rx headroom to the
1132 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1133 * This function is used to set or query state related to XDP on the
1134 * netdevice. See definition of enum xdp_netdev_command for details.
1137 struct net_device_ops {
1138 int (*ndo_init)(struct net_device *dev);
1139 void (*ndo_uninit)(struct net_device *dev);
1140 int (*ndo_open)(struct net_device *dev);
1141 int (*ndo_stop)(struct net_device *dev);
1142 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1143 struct net_device *dev);
1144 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1145 struct net_device *dev,
1146 netdev_features_t features);
1147 u16 (*ndo_select_queue)(struct net_device *dev,
1148 struct sk_buff *skb,
1150 select_queue_fallback_t fallback);
1151 void (*ndo_change_rx_flags)(struct net_device *dev,
1153 void (*ndo_set_rx_mode)(struct net_device *dev);
1154 int (*ndo_set_mac_address)(struct net_device *dev,
1156 int (*ndo_validate_addr)(struct net_device *dev);
1157 int (*ndo_do_ioctl)(struct net_device *dev,
1158 struct ifreq *ifr, int cmd);
1159 int (*ndo_set_config)(struct net_device *dev,
1161 int (*ndo_change_mtu)(struct net_device *dev,
1163 int (*ndo_neigh_setup)(struct net_device *dev,
1164 struct neigh_parms *);
1165 void (*ndo_tx_timeout) (struct net_device *dev);
1167 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1168 struct rtnl_link_stats64 *storage);
1169 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1170 int (*ndo_get_offload_stats)(int attr_id,
1171 const struct net_device *dev,
1173 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1175 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1176 __be16 proto, u16 vid);
1177 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1178 __be16 proto, u16 vid);
1179 #ifdef CONFIG_NET_POLL_CONTROLLER
1180 void (*ndo_poll_controller)(struct net_device *dev);
1181 int (*ndo_netpoll_setup)(struct net_device *dev,
1182 struct netpoll_info *info);
1183 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1185 #ifdef CONFIG_NET_RX_BUSY_POLL
1186 int (*ndo_busy_poll)(struct napi_struct *dev);
1188 int (*ndo_set_vf_mac)(struct net_device *dev,
1189 int queue, u8 *mac);
1190 int (*ndo_set_vf_vlan)(struct net_device *dev,
1191 int queue, u16 vlan,
1192 u8 qos, __be16 proto);
1193 int (*ndo_set_vf_rate)(struct net_device *dev,
1194 int vf, int min_tx_rate,
1196 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1197 int vf, bool setting);
1198 int (*ndo_set_vf_trust)(struct net_device *dev,
1199 int vf, bool setting);
1200 int (*ndo_get_vf_config)(struct net_device *dev,
1202 struct ifla_vf_info *ivf);
1203 int (*ndo_set_vf_link_state)(struct net_device *dev,
1204 int vf, int link_state);
1205 int (*ndo_get_vf_stats)(struct net_device *dev,
1207 struct ifla_vf_stats
1209 int (*ndo_set_vf_port)(struct net_device *dev,
1211 struct nlattr *port[]);
1212 int (*ndo_get_vf_port)(struct net_device *dev,
1213 int vf, struct sk_buff *skb);
1214 int (*ndo_set_vf_guid)(struct net_device *dev,
1217 int (*ndo_set_vf_rss_query_en)(
1218 struct net_device *dev,
1219 int vf, bool setting);
1220 int (*ndo_setup_tc)(struct net_device *dev,
1223 struct tc_to_netdev *tc);
1224 #if IS_ENABLED(CONFIG_FCOE)
1225 int (*ndo_fcoe_enable)(struct net_device *dev);
1226 int (*ndo_fcoe_disable)(struct net_device *dev);
1227 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1229 struct scatterlist *sgl,
1231 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1233 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1235 struct scatterlist *sgl,
1237 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1238 struct netdev_fcoe_hbainfo *hbainfo);
1241 #if IS_ENABLED(CONFIG_LIBFCOE)
1242 #define NETDEV_FCOE_WWNN 0
1243 #define NETDEV_FCOE_WWPN 1
1244 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1245 u64 *wwn, int type);
1248 #ifdef CONFIG_RFS_ACCEL
1249 int (*ndo_rx_flow_steer)(struct net_device *dev,
1250 const struct sk_buff *skb,
1254 int (*ndo_add_slave)(struct net_device *dev,
1255 struct net_device *slave_dev);
1256 int (*ndo_del_slave)(struct net_device *dev,
1257 struct net_device *slave_dev);
1258 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1259 netdev_features_t features);
1260 int (*ndo_set_features)(struct net_device *dev,
1261 netdev_features_t features);
1262 int (*ndo_neigh_construct)(struct net_device *dev,
1263 struct neighbour *n);
1264 void (*ndo_neigh_destroy)(struct net_device *dev,
1265 struct neighbour *n);
1267 int (*ndo_fdb_add)(struct ndmsg *ndm,
1268 struct nlattr *tb[],
1269 struct net_device *dev,
1270 const unsigned char *addr,
1273 int (*ndo_fdb_del)(struct ndmsg *ndm,
1274 struct nlattr *tb[],
1275 struct net_device *dev,
1276 const unsigned char *addr,
1278 int (*ndo_fdb_dump)(struct sk_buff *skb,
1279 struct netlink_callback *cb,
1280 struct net_device *dev,
1281 struct net_device *filter_dev,
1284 int (*ndo_bridge_setlink)(struct net_device *dev,
1285 struct nlmsghdr *nlh,
1287 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1289 struct net_device *dev,
1292 int (*ndo_bridge_dellink)(struct net_device *dev,
1293 struct nlmsghdr *nlh,
1295 int (*ndo_change_carrier)(struct net_device *dev,
1297 int (*ndo_get_phys_port_id)(struct net_device *dev,
1298 struct netdev_phys_item_id *ppid);
1299 int (*ndo_get_phys_port_name)(struct net_device *dev,
1300 char *name, size_t len);
1301 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1302 struct udp_tunnel_info *ti);
1303 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1304 struct udp_tunnel_info *ti);
1305 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1306 struct net_device *dev);
1307 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1310 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1311 struct net_device *dev,
1313 int (*ndo_get_lock_subclass)(struct net_device *dev);
1314 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1317 int (*ndo_get_iflink)(const struct net_device *dev);
1318 int (*ndo_change_proto_down)(struct net_device *dev,
1320 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1321 struct sk_buff *skb);
1322 void (*ndo_set_rx_headroom)(struct net_device *dev,
1323 int needed_headroom);
1324 int (*ndo_xdp)(struct net_device *dev,
1325 struct netdev_xdp *xdp);
1329 * enum net_device_priv_flags - &struct net_device priv_flags
1331 * These are the &struct net_device, they are only set internally
1332 * by drivers and used in the kernel. These flags are invisible to
1333 * userspace; this means that the order of these flags can change
1334 * during any kernel release.
1336 * You should have a pretty good reason to be extending these flags.
1338 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1339 * @IFF_EBRIDGE: Ethernet bridging device
1340 * @IFF_BONDING: bonding master or slave
1341 * @IFF_ISATAP: ISATAP interface (RFC4214)
1342 * @IFF_WAN_HDLC: WAN HDLC device
1343 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1345 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1346 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1347 * @IFF_MACVLAN_PORT: device used as macvlan port
1348 * @IFF_BRIDGE_PORT: device used as bridge port
1349 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1350 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1351 * @IFF_UNICAST_FLT: Supports unicast filtering
1352 * @IFF_TEAM_PORT: device used as team port
1353 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1354 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1355 * change when it's running
1356 * @IFF_MACVLAN: Macvlan device
1357 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1358 * underlying stacked devices
1359 * @IFF_IPVLAN_MASTER: IPvlan master device
1360 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1361 * @IFF_L3MDEV_MASTER: device is an L3 master device
1362 * @IFF_NO_QUEUE: device can run without qdisc attached
1363 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1364 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1365 * @IFF_TEAM: device is a team device
1366 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1367 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1368 * entity (i.e. the master device for bridged veth)
1369 * @IFF_MACSEC: device is a MACsec device
1371 enum netdev_priv_flags {
1372 IFF_802_1Q_VLAN = 1<<0,
1376 IFF_WAN_HDLC = 1<<4,
1377 IFF_XMIT_DST_RELEASE = 1<<5,
1378 IFF_DONT_BRIDGE = 1<<6,
1379 IFF_DISABLE_NETPOLL = 1<<7,
1380 IFF_MACVLAN_PORT = 1<<8,
1381 IFF_BRIDGE_PORT = 1<<9,
1382 IFF_OVS_DATAPATH = 1<<10,
1383 IFF_TX_SKB_SHARING = 1<<11,
1384 IFF_UNICAST_FLT = 1<<12,
1385 IFF_TEAM_PORT = 1<<13,
1386 IFF_SUPP_NOFCS = 1<<14,
1387 IFF_LIVE_ADDR_CHANGE = 1<<15,
1388 IFF_MACVLAN = 1<<16,
1389 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1390 IFF_IPVLAN_MASTER = 1<<18,
1391 IFF_IPVLAN_SLAVE = 1<<19,
1392 IFF_L3MDEV_MASTER = 1<<20,
1393 IFF_NO_QUEUE = 1<<21,
1394 IFF_OPENVSWITCH = 1<<22,
1395 IFF_L3MDEV_SLAVE = 1<<23,
1397 IFF_RXFH_CONFIGURED = 1<<25,
1398 IFF_PHONY_HEADROOM = 1<<26,
1402 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1403 #define IFF_EBRIDGE IFF_EBRIDGE
1404 #define IFF_BONDING IFF_BONDING
1405 #define IFF_ISATAP IFF_ISATAP
1406 #define IFF_WAN_HDLC IFF_WAN_HDLC
1407 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1408 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1409 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1410 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1411 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1412 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1413 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1414 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1415 #define IFF_TEAM_PORT IFF_TEAM_PORT
1416 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1417 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1418 #define IFF_MACVLAN IFF_MACVLAN
1419 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1420 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1421 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1422 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1423 #define IFF_NO_QUEUE IFF_NO_QUEUE
1424 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1425 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1426 #define IFF_TEAM IFF_TEAM
1427 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1428 #define IFF_MACSEC IFF_MACSEC
1431 * struct net_device - The DEVICE structure.
1432 * Actually, this whole structure is a big mistake. It mixes I/O
1433 * data with strictly "high-level" data, and it has to know about
1434 * almost every data structure used in the INET module.
1436 * @name: This is the first field of the "visible" part of this structure
1437 * (i.e. as seen by users in the "Space.c" file). It is the name
1440 * @name_hlist: Device name hash chain, please keep it close to name[]
1441 * @ifalias: SNMP alias
1442 * @mem_end: Shared memory end
1443 * @mem_start: Shared memory start
1444 * @base_addr: Device I/O address
1445 * @irq: Device IRQ number
1447 * @carrier_changes: Stats to monitor carrier on<->off transitions
1449 * @state: Generic network queuing layer state, see netdev_state_t
1450 * @dev_list: The global list of network devices
1451 * @napi_list: List entry used for polling NAPI devices
1452 * @unreg_list: List entry when we are unregistering the
1453 * device; see the function unregister_netdev
1454 * @close_list: List entry used when we are closing the device
1455 * @ptype_all: Device-specific packet handlers for all protocols
1456 * @ptype_specific: Device-specific, protocol-specific packet handlers
1458 * @adj_list: Directly linked devices, like slaves for bonding
1459 * @features: Currently active device features
1460 * @hw_features: User-changeable features
1462 * @wanted_features: User-requested features
1463 * @vlan_features: Mask of features inheritable by VLAN devices
1465 * @hw_enc_features: Mask of features inherited by encapsulating devices
1466 * This field indicates what encapsulation
1467 * offloads the hardware is capable of doing,
1468 * and drivers will need to set them appropriately.
1470 * @mpls_features: Mask of features inheritable by MPLS
1472 * @ifindex: interface index
1473 * @group: The group the device belongs to
1475 * @stats: Statistics struct, which was left as a legacy, use
1476 * rtnl_link_stats64 instead
1478 * @rx_dropped: Dropped packets by core network,
1479 * do not use this in drivers
1480 * @tx_dropped: Dropped packets by core network,
1481 * do not use this in drivers
1482 * @rx_nohandler: nohandler dropped packets by core network on
1483 * inactive devices, do not use this in drivers
1485 * @wireless_handlers: List of functions to handle Wireless Extensions,
1487 * see <net/iw_handler.h> for details.
1488 * @wireless_data: Instance data managed by the core of wireless extensions
1490 * @netdev_ops: Includes several pointers to callbacks,
1491 * if one wants to override the ndo_*() functions
1492 * @ethtool_ops: Management operations
1493 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1494 * discovery handling. Necessary for e.g. 6LoWPAN.
1495 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1496 * of Layer 2 headers.
1498 * @flags: Interface flags (a la BSD)
1499 * @priv_flags: Like 'flags' but invisible to userspace,
1500 * see if.h for the definitions
1501 * @gflags: Global flags ( kept as legacy )
1502 * @padded: How much padding added by alloc_netdev()
1503 * @operstate: RFC2863 operstate
1504 * @link_mode: Mapping policy to operstate
1505 * @if_port: Selectable AUI, TP, ...
1507 * @mtu: Interface MTU value
1508 * @min_mtu: Interface Minimum MTU value
1509 * @max_mtu: Interface Maximum MTU value
1510 * @type: Interface hardware type
1511 * @hard_header_len: Maximum hardware header length.
1513 * @needed_headroom: Extra headroom the hardware may need, but not in all
1514 * cases can this be guaranteed
1515 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1516 * cases can this be guaranteed. Some cases also use
1517 * LL_MAX_HEADER instead to allocate the skb
1519 * interface address info:
1521 * @perm_addr: Permanent hw address
1522 * @addr_assign_type: Hw address assignment type
1523 * @addr_len: Hardware address length
1524 * @neigh_priv_len: Used in neigh_alloc()
1525 * @dev_id: Used to differentiate devices that share
1526 * the same link layer address
1527 * @dev_port: Used to differentiate devices that share
1529 * @addr_list_lock: XXX: need comments on this one
1530 * @uc_promisc: Counter that indicates promiscuous mode
1531 * has been enabled due to the need to listen to
1532 * additional unicast addresses in a device that
1533 * does not implement ndo_set_rx_mode()
1534 * @uc: unicast mac addresses
1535 * @mc: multicast mac addresses
1536 * @dev_addrs: list of device hw addresses
1537 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1538 * @promiscuity: Number of times the NIC is told to work in
1539 * promiscuous mode; if it becomes 0 the NIC will
1540 * exit promiscuous mode
1541 * @allmulti: Counter, enables or disables allmulticast mode
1543 * @vlan_info: VLAN info
1544 * @dsa_ptr: dsa specific data
1545 * @tipc_ptr: TIPC specific data
1546 * @atalk_ptr: AppleTalk link
1547 * @ip_ptr: IPv4 specific data
1548 * @dn_ptr: DECnet specific data
1549 * @ip6_ptr: IPv6 specific data
1550 * @ax25_ptr: AX.25 specific data
1551 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1553 * @last_rx: Time of last Rx
1554 * @dev_addr: Hw address (before bcast,
1555 * because most packets are unicast)
1557 * @_rx: Array of RX queues
1558 * @num_rx_queues: Number of RX queues
1559 * allocated at register_netdev() time
1560 * @real_num_rx_queues: Number of RX queues currently active in device
1562 * @rx_handler: handler for received packets
1563 * @rx_handler_data: XXX: need comments on this one
1564 * @ingress_queue: XXX: need comments on this one
1565 * @broadcast: hw bcast address
1567 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1568 * indexed by RX queue number. Assigned by driver.
1569 * This must only be set if the ndo_rx_flow_steer
1570 * operation is defined
1571 * @index_hlist: Device index hash chain
1573 * @_tx: Array of TX queues
1574 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1575 * @real_num_tx_queues: Number of TX queues currently active in device
1576 * @qdisc: Root qdisc from userspace point of view
1577 * @tx_queue_len: Max frames per queue allowed
1578 * @tx_global_lock: XXX: need comments on this one
1580 * @xps_maps: XXX: need comments on this one
1582 * @watchdog_timeo: Represents the timeout that is used by
1583 * the watchdog (see dev_watchdog())
1584 * @watchdog_timer: List of timers
1586 * @pcpu_refcnt: Number of references to this device
1587 * @todo_list: Delayed register/unregister
1588 * @link_watch_list: XXX: need comments on this one
1590 * @reg_state: Register/unregister state machine
1591 * @dismantle: Device is going to be freed
1592 * @rtnl_link_state: This enum represents the phases of creating
1595 * @destructor: Called from unregister,
1596 * can be used to call free_netdev
1597 * @npinfo: XXX: need comments on this one
1598 * @nd_net: Network namespace this network device is inside
1600 * @ml_priv: Mid-layer private
1601 * @lstats: Loopback statistics
1602 * @tstats: Tunnel statistics
1603 * @dstats: Dummy statistics
1604 * @vstats: Virtual ethernet statistics
1609 * @dev: Class/net/name entry
1610 * @sysfs_groups: Space for optional device, statistics and wireless
1613 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1614 * @rtnl_link_ops: Rtnl_link_ops
1616 * @gso_max_size: Maximum size of generic segmentation offload
1617 * @gso_max_segs: Maximum number of segments that can be passed to the
1620 * @dcbnl_ops: Data Center Bridging netlink ops
1621 * @num_tc: Number of traffic classes in the net device
1622 * @tc_to_txq: XXX: need comments on this one
1623 * @prio_tc_map: XXX: need comments on this one
1625 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1627 * @priomap: XXX: need comments on this one
1628 * @phydev: Physical device may attach itself
1629 * for hardware timestamping
1631 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1632 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1634 * @proto_down: protocol port state information can be sent to the
1635 * switch driver and used to set the phys state of the
1638 * FIXME: cleanup struct net_device such that network protocol info
1643 char name[IFNAMSIZ];
1644 struct hlist_node name_hlist;
1647 * I/O specific fields
1648 * FIXME: Merge these and struct ifmap into one
1650 unsigned long mem_end;
1651 unsigned long mem_start;
1652 unsigned long base_addr;
1655 atomic_t carrier_changes;
1658 * Some hardware also needs these fields (state,dev_list,
1659 * napi_list,unreg_list,close_list) but they are not
1660 * part of the usual set specified in Space.c.
1663 unsigned long state;
1665 struct list_head dev_list;
1666 struct list_head napi_list;
1667 struct list_head unreg_list;
1668 struct list_head close_list;
1669 struct list_head ptype_all;
1670 struct list_head ptype_specific;
1673 struct list_head upper;
1674 struct list_head lower;
1677 netdev_features_t features;
1678 netdev_features_t hw_features;
1679 netdev_features_t wanted_features;
1680 netdev_features_t vlan_features;
1681 netdev_features_t hw_enc_features;
1682 netdev_features_t mpls_features;
1683 netdev_features_t gso_partial_features;
1688 struct net_device_stats stats;
1690 atomic_long_t rx_dropped;
1691 atomic_long_t tx_dropped;
1692 atomic_long_t rx_nohandler;
1694 #ifdef CONFIG_WIRELESS_EXT
1695 const struct iw_handler_def *wireless_handlers;
1696 struct iw_public_data *wireless_data;
1698 const struct net_device_ops *netdev_ops;
1699 const struct ethtool_ops *ethtool_ops;
1700 #ifdef CONFIG_NET_SWITCHDEV
1701 const struct switchdev_ops *switchdev_ops;
1703 #ifdef CONFIG_NET_L3_MASTER_DEV
1704 const struct l3mdev_ops *l3mdev_ops;
1706 #if IS_ENABLED(CONFIG_IPV6)
1707 const struct ndisc_ops *ndisc_ops;
1710 const struct header_ops *header_ops;
1713 unsigned int priv_flags;
1715 unsigned short gflags;
1716 unsigned short padded;
1718 unsigned char operstate;
1719 unsigned char link_mode;
1721 unsigned char if_port;
1725 unsigned int min_mtu;
1726 unsigned int max_mtu;
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 nf_hook_entry __rcu *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;
1923 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1924 void netdev_reset_tc(struct net_device *dev);
1925 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1926 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1929 int netdev_get_num_tc(struct net_device *dev)
1935 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1938 return &dev->_tx[index];
1941 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1942 const struct sk_buff *skb)
1944 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1947 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1948 void (*f)(struct net_device *,
1949 struct netdev_queue *,
1955 for (i = 0; i < dev->num_tx_queues; i++)
1956 f(dev, &dev->_tx[i], arg);
1959 #define netdev_lockdep_set_classes(dev) \
1961 static struct lock_class_key qdisc_tx_busylock_key; \
1962 static struct lock_class_key qdisc_running_key; \
1963 static struct lock_class_key qdisc_xmit_lock_key; \
1964 static struct lock_class_key dev_addr_list_lock_key; \
1967 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1968 (dev)->qdisc_running_key = &qdisc_running_key; \
1969 lockdep_set_class(&(dev)->addr_list_lock, \
1970 &dev_addr_list_lock_key); \
1971 for (i = 0; i < (dev)->num_tx_queues; i++) \
1972 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1973 &qdisc_xmit_lock_key); \
1976 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1977 struct sk_buff *skb,
1980 /* returns the headroom that the master device needs to take in account
1981 * when forwarding to this dev
1983 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1985 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1988 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1990 if (dev->netdev_ops->ndo_set_rx_headroom)
1991 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1994 /* set the device rx headroom to the dev's default */
1995 static inline void netdev_reset_rx_headroom(struct net_device *dev)
1997 netdev_set_rx_headroom(dev, -1);
2001 * Net namespace inlines
2004 struct net *dev_net(const struct net_device *dev)
2006 return read_pnet(&dev->nd_net);
2010 void dev_net_set(struct net_device *dev, struct net *net)
2012 write_pnet(&dev->nd_net, net);
2015 static inline bool netdev_uses_dsa(struct net_device *dev)
2017 #if IS_ENABLED(CONFIG_NET_DSA)
2018 if (dev->dsa_ptr != NULL)
2019 return dsa_uses_tagged_protocol(dev->dsa_ptr);
2025 * netdev_priv - access network device private data
2026 * @dev: network device
2028 * Get network device private data
2030 static inline void *netdev_priv(const struct net_device *dev)
2032 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2035 /* Set the sysfs physical device reference for the network logical device
2036 * if set prior to registration will cause a symlink during initialization.
2038 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2040 /* Set the sysfs device type for the network logical device to allow
2041 * fine-grained identification of different network device types. For
2042 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2044 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2046 /* Default NAPI poll() weight
2047 * Device drivers are strongly advised to not use bigger value
2049 #define NAPI_POLL_WEIGHT 64
2052 * netif_napi_add - initialize a NAPI context
2053 * @dev: network device
2054 * @napi: NAPI context
2055 * @poll: polling function
2056 * @weight: default weight
2058 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2059 * *any* of the other NAPI-related functions.
2061 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2062 int (*poll)(struct napi_struct *, int), int weight);
2065 * netif_tx_napi_add - initialize a NAPI context
2066 * @dev: network device
2067 * @napi: NAPI context
2068 * @poll: polling function
2069 * @weight: default weight
2071 * This variant of netif_napi_add() should be used from drivers using NAPI
2072 * to exclusively poll a TX queue.
2073 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2075 static inline void netif_tx_napi_add(struct net_device *dev,
2076 struct napi_struct *napi,
2077 int (*poll)(struct napi_struct *, int),
2080 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2081 netif_napi_add(dev, napi, poll, weight);
2085 * netif_napi_del - remove a NAPI context
2086 * @napi: NAPI context
2088 * netif_napi_del() removes a NAPI context from the network device NAPI list
2090 void netif_napi_del(struct napi_struct *napi);
2092 struct napi_gro_cb {
2093 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2096 /* Length of frag0. */
2097 unsigned int frag0_len;
2099 /* This indicates where we are processing relative to skb->data. */
2102 /* This is non-zero if the packet cannot be merged with the new skb. */
2105 /* Save the IP ID here and check when we get to the transport layer */
2108 /* Number of segments aggregated. */
2111 /* Start offset for remote checksum offload */
2112 u16 gro_remcsum_start;
2114 /* jiffies when first packet was created/queued */
2117 /* Used in ipv6_gro_receive() and foo-over-udp */
2120 /* This is non-zero if the packet may be of the same flow. */
2123 /* Used in tunnel GRO receive */
2126 /* GRO checksum is valid */
2129 /* Number of checksums via CHECKSUM_UNNECESSARY */
2134 #define NAPI_GRO_FREE 1
2135 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2137 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2140 /* Used in GRE, set in fou/gue_gro_receive */
2143 /* Used to determine if flush_id can be ignored */
2146 /* Number of gro_receive callbacks this packet already went through */
2147 u8 recursion_counter:4;
2151 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2154 /* used in skb_gro_receive() slow path */
2155 struct sk_buff *last;
2158 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2160 #define GRO_RECURSION_LIMIT 15
2161 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2163 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2166 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2167 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2168 struct sk_buff **head,
2169 struct sk_buff *skb)
2171 if (unlikely(gro_recursion_inc_test(skb))) {
2172 NAPI_GRO_CB(skb)->flush |= 1;
2176 return cb(head, skb);
2179 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2181 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2183 struct sk_buff **head,
2184 struct sk_buff *skb)
2186 if (unlikely(gro_recursion_inc_test(skb))) {
2187 NAPI_GRO_CB(skb)->flush |= 1;
2191 return cb(sk, head, skb);
2194 struct packet_type {
2195 __be16 type; /* This is really htons(ether_type). */
2196 struct net_device *dev; /* NULL is wildcarded here */
2197 int (*func) (struct sk_buff *,
2198 struct net_device *,
2199 struct packet_type *,
2200 struct net_device *);
2201 bool (*id_match)(struct packet_type *ptype,
2203 void *af_packet_priv;
2204 struct list_head list;
2207 struct offload_callbacks {
2208 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2209 netdev_features_t features);
2210 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2211 struct sk_buff *skb);
2212 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2215 struct packet_offload {
2216 __be16 type; /* This is really htons(ether_type). */
2218 struct offload_callbacks callbacks;
2219 struct list_head list;
2222 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2223 struct pcpu_sw_netstats {
2228 struct u64_stats_sync syncp;
2231 #define __netdev_alloc_pcpu_stats(type, gfp) \
2233 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2236 for_each_possible_cpu(__cpu) { \
2237 typeof(type) *stat; \
2238 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2239 u64_stats_init(&stat->syncp); \
2245 #define netdev_alloc_pcpu_stats(type) \
2246 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2248 enum netdev_lag_tx_type {
2249 NETDEV_LAG_TX_TYPE_UNKNOWN,
2250 NETDEV_LAG_TX_TYPE_RANDOM,
2251 NETDEV_LAG_TX_TYPE_BROADCAST,
2252 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2253 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2254 NETDEV_LAG_TX_TYPE_HASH,
2257 struct netdev_lag_upper_info {
2258 enum netdev_lag_tx_type tx_type;
2261 struct netdev_lag_lower_state_info {
2266 #include <linux/notifier.h>
2268 /* netdevice notifier chain. Please remember to update the rtnetlink
2269 * notification exclusion list in rtnetlink_event() when adding new
2272 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2273 #define NETDEV_DOWN 0x0002
2274 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2275 detected a hardware crash and restarted
2276 - we can use this eg to kick tcp sessions
2278 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2279 #define NETDEV_REGISTER 0x0005
2280 #define NETDEV_UNREGISTER 0x0006
2281 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2282 #define NETDEV_CHANGEADDR 0x0008
2283 #define NETDEV_GOING_DOWN 0x0009
2284 #define NETDEV_CHANGENAME 0x000A
2285 #define NETDEV_FEAT_CHANGE 0x000B
2286 #define NETDEV_BONDING_FAILOVER 0x000C
2287 #define NETDEV_PRE_UP 0x000D
2288 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2289 #define NETDEV_POST_TYPE_CHANGE 0x000F
2290 #define NETDEV_POST_INIT 0x0010
2291 #define NETDEV_UNREGISTER_FINAL 0x0011
2292 #define NETDEV_RELEASE 0x0012
2293 #define NETDEV_NOTIFY_PEERS 0x0013
2294 #define NETDEV_JOIN 0x0014
2295 #define NETDEV_CHANGEUPPER 0x0015
2296 #define NETDEV_RESEND_IGMP 0x0016
2297 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2298 #define NETDEV_CHANGEINFODATA 0x0018
2299 #define NETDEV_BONDING_INFO 0x0019
2300 #define NETDEV_PRECHANGEUPPER 0x001A
2301 #define NETDEV_CHANGELOWERSTATE 0x001B
2302 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2303 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2305 int register_netdevice_notifier(struct notifier_block *nb);
2306 int unregister_netdevice_notifier(struct notifier_block *nb);
2308 struct netdev_notifier_info {
2309 struct net_device *dev;
2312 struct netdev_notifier_change_info {
2313 struct netdev_notifier_info info; /* must be first */
2314 unsigned int flags_changed;
2317 struct netdev_notifier_changeupper_info {
2318 struct netdev_notifier_info info; /* must be first */
2319 struct net_device *upper_dev; /* new upper dev */
2320 bool master; /* is upper dev master */
2321 bool linking; /* is the notification for link or unlink */
2322 void *upper_info; /* upper dev info */
2325 struct netdev_notifier_changelowerstate_info {
2326 struct netdev_notifier_info info; /* must be first */
2327 void *lower_state_info; /* is lower dev state */
2330 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2331 struct net_device *dev)
2336 static inline struct net_device *
2337 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2342 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2345 extern rwlock_t dev_base_lock; /* Device list lock */
2347 #define for_each_netdev(net, d) \
2348 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2349 #define for_each_netdev_reverse(net, d) \
2350 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2351 #define for_each_netdev_rcu(net, d) \
2352 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2353 #define for_each_netdev_safe(net, d, n) \
2354 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2355 #define for_each_netdev_continue(net, d) \
2356 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2357 #define for_each_netdev_continue_rcu(net, d) \
2358 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2359 #define for_each_netdev_in_bond_rcu(bond, slave) \
2360 for_each_netdev_rcu(&init_net, slave) \
2361 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2362 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2364 static inline struct net_device *next_net_device(struct net_device *dev)
2366 struct list_head *lh;
2370 lh = dev->dev_list.next;
2371 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2374 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2376 struct list_head *lh;
2380 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2381 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2384 static inline struct net_device *first_net_device(struct net *net)
2386 return list_empty(&net->dev_base_head) ? NULL :
2387 net_device_entry(net->dev_base_head.next);
2390 static inline struct net_device *first_net_device_rcu(struct net *net)
2392 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2394 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2397 int netdev_boot_setup_check(struct net_device *dev);
2398 unsigned long netdev_boot_base(const char *prefix, int unit);
2399 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2400 const char *hwaddr);
2401 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2402 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2403 void dev_add_pack(struct packet_type *pt);
2404 void dev_remove_pack(struct packet_type *pt);
2405 void __dev_remove_pack(struct packet_type *pt);
2406 void dev_add_offload(struct packet_offload *po);
2407 void dev_remove_offload(struct packet_offload *po);
2409 int dev_get_iflink(const struct net_device *dev);
2410 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2411 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2412 unsigned short mask);
2413 struct net_device *dev_get_by_name(struct net *net, const char *name);
2414 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2415 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2416 int dev_alloc_name(struct net_device *dev, const char *name);
2417 int dev_open(struct net_device *dev);
2418 int dev_close(struct net_device *dev);
2419 int dev_close_many(struct list_head *head, bool unlink);
2420 void dev_disable_lro(struct net_device *dev);
2421 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2422 int dev_queue_xmit(struct sk_buff *skb);
2423 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2424 int register_netdevice(struct net_device *dev);
2425 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2426 void unregister_netdevice_many(struct list_head *head);
2427 static inline void unregister_netdevice(struct net_device *dev)
2429 unregister_netdevice_queue(dev, NULL);
2432 int netdev_refcnt_read(const struct net_device *dev);
2433 void free_netdev(struct net_device *dev);
2434 void netdev_freemem(struct net_device *dev);
2435 void synchronize_net(void);
2436 int init_dummy_netdev(struct net_device *dev);
2438 DECLARE_PER_CPU(int, xmit_recursion);
2439 #define XMIT_RECURSION_LIMIT 10
2441 static inline int dev_recursion_level(void)
2443 return this_cpu_read(xmit_recursion);
2446 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2447 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2448 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2449 int netdev_get_name(struct net *net, char *name, int ifindex);
2450 int dev_restart(struct net_device *dev);
2451 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2453 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2455 return NAPI_GRO_CB(skb)->data_offset;
2458 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2460 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2463 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2465 NAPI_GRO_CB(skb)->data_offset += len;
2468 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2469 unsigned int offset)
2471 return NAPI_GRO_CB(skb)->frag0 + offset;
2474 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2476 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2479 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2480 unsigned int offset)
2482 if (!pskb_may_pull(skb, hlen))
2485 NAPI_GRO_CB(skb)->frag0 = NULL;
2486 NAPI_GRO_CB(skb)->frag0_len = 0;
2487 return skb->data + offset;
2490 static inline void *skb_gro_network_header(struct sk_buff *skb)
2492 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2493 skb_network_offset(skb);
2496 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2497 const void *start, unsigned int len)
2499 if (NAPI_GRO_CB(skb)->csum_valid)
2500 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2501 csum_partial(start, len, 0));
2504 /* GRO checksum functions. These are logical equivalents of the normal
2505 * checksum functions (in skbuff.h) except that they operate on the GRO
2506 * offsets and fields in sk_buff.
2509 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2511 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2513 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2516 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2520 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2521 skb_checksum_start_offset(skb) <
2522 skb_gro_offset(skb)) &&
2523 !skb_at_gro_remcsum_start(skb) &&
2524 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2525 (!zero_okay || check));
2528 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2531 if (NAPI_GRO_CB(skb)->csum_valid &&
2532 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2535 NAPI_GRO_CB(skb)->csum = psum;
2537 return __skb_gro_checksum_complete(skb);
2540 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2542 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2543 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2544 NAPI_GRO_CB(skb)->csum_cnt--;
2546 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2547 * verified a new top level checksum or an encapsulated one
2548 * during GRO. This saves work if we fallback to normal path.
2550 __skb_incr_checksum_unnecessary(skb);
2554 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2557 __sum16 __ret = 0; \
2558 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2559 __ret = __skb_gro_checksum_validate_complete(skb, \
2560 compute_pseudo(skb, proto)); \
2562 __skb_mark_checksum_bad(skb); \
2564 skb_gro_incr_csum_unnecessary(skb); \
2568 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2569 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2571 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2573 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2575 #define skb_gro_checksum_simple_validate(skb) \
2576 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2578 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2580 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2581 !NAPI_GRO_CB(skb)->csum_valid);
2584 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2585 __sum16 check, __wsum pseudo)
2587 NAPI_GRO_CB(skb)->csum = ~pseudo;
2588 NAPI_GRO_CB(skb)->csum_valid = 1;
2591 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2593 if (__skb_gro_checksum_convert_check(skb)) \
2594 __skb_gro_checksum_convert(skb, check, \
2595 compute_pseudo(skb, proto)); \
2598 struct gro_remcsum {
2603 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2609 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2610 unsigned int off, size_t hdrlen,
2611 int start, int offset,
2612 struct gro_remcsum *grc,
2616 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2618 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2621 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2625 ptr = skb_gro_header_fast(skb, off);
2626 if (skb_gro_header_hard(skb, off + plen)) {
2627 ptr = skb_gro_header_slow(skb, off + plen, off);
2632 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2635 /* Adjust skb->csum since we changed the packet */
2636 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2638 grc->offset = off + hdrlen + offset;
2644 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2645 struct gro_remcsum *grc)
2648 size_t plen = grc->offset + sizeof(u16);
2653 ptr = skb_gro_header_fast(skb, grc->offset);
2654 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2655 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2660 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2663 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2664 unsigned short type,
2665 const void *daddr, const void *saddr,
2668 if (!dev->header_ops || !dev->header_ops->create)
2671 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2674 static inline int dev_parse_header(const struct sk_buff *skb,
2675 unsigned char *haddr)
2677 const struct net_device *dev = skb->dev;
2679 if (!dev->header_ops || !dev->header_ops->parse)
2681 return dev->header_ops->parse(skb, haddr);
2684 /* ll_header must have at least hard_header_len allocated */
2685 static inline bool dev_validate_header(const struct net_device *dev,
2686 char *ll_header, int len)
2688 if (likely(len >= dev->hard_header_len))
2691 if (capable(CAP_SYS_RAWIO)) {
2692 memset(ll_header + len, 0, dev->hard_header_len - len);
2696 if (dev->header_ops && dev->header_ops->validate)
2697 return dev->header_ops->validate(ll_header, len);
2702 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2703 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2704 static inline int unregister_gifconf(unsigned int family)
2706 return register_gifconf(family, NULL);
2709 #ifdef CONFIG_NET_FLOW_LIMIT
2710 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2711 struct sd_flow_limit {
2713 unsigned int num_buckets;
2714 unsigned int history_head;
2715 u16 history[FLOW_LIMIT_HISTORY];
2719 extern int netdev_flow_limit_table_len;
2720 #endif /* CONFIG_NET_FLOW_LIMIT */
2723 * Incoming packets are placed on per-CPU queues
2725 struct softnet_data {
2726 struct list_head poll_list;
2727 struct sk_buff_head process_queue;
2730 unsigned int processed;
2731 unsigned int time_squeeze;
2732 unsigned int received_rps;
2734 struct softnet_data *rps_ipi_list;
2736 #ifdef CONFIG_NET_FLOW_LIMIT
2737 struct sd_flow_limit __rcu *flow_limit;
2739 struct Qdisc *output_queue;
2740 struct Qdisc **output_queue_tailp;
2741 struct sk_buff *completion_queue;
2744 /* input_queue_head should be written by cpu owning this struct,
2745 * and only read by other cpus. Worth using a cache line.
2747 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2749 /* Elements below can be accessed between CPUs for RPS/RFS */
2750 struct call_single_data csd ____cacheline_aligned_in_smp;
2751 struct softnet_data *rps_ipi_next;
2753 unsigned int input_queue_tail;
2755 unsigned int dropped;
2756 struct sk_buff_head input_pkt_queue;
2757 struct napi_struct backlog;
2761 static inline void input_queue_head_incr(struct softnet_data *sd)
2764 sd->input_queue_head++;
2768 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2769 unsigned int *qtail)
2772 *qtail = ++sd->input_queue_tail;
2776 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2778 void __netif_schedule(struct Qdisc *q);
2779 void netif_schedule_queue(struct netdev_queue *txq);
2781 static inline void netif_tx_schedule_all(struct net_device *dev)
2785 for (i = 0; i < dev->num_tx_queues; i++)
2786 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2789 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2791 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2795 * netif_start_queue - allow transmit
2796 * @dev: network device
2798 * Allow upper layers to call the device hard_start_xmit routine.
2800 static inline void netif_start_queue(struct net_device *dev)
2802 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2805 static inline void netif_tx_start_all_queues(struct net_device *dev)
2809 for (i = 0; i < dev->num_tx_queues; i++) {
2810 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2811 netif_tx_start_queue(txq);
2815 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2818 * netif_wake_queue - restart transmit
2819 * @dev: network device
2821 * Allow upper layers to call the device hard_start_xmit routine.
2822 * Used for flow control when transmit resources are available.
2824 static inline void netif_wake_queue(struct net_device *dev)
2826 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2829 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2833 for (i = 0; i < dev->num_tx_queues; i++) {
2834 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2835 netif_tx_wake_queue(txq);
2839 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2841 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2845 * netif_stop_queue - stop transmitted packets
2846 * @dev: network device
2848 * Stop upper layers calling the device hard_start_xmit routine.
2849 * Used for flow control when transmit resources are unavailable.
2851 static inline void netif_stop_queue(struct net_device *dev)
2853 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2856 void netif_tx_stop_all_queues(struct net_device *dev);
2858 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2860 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2864 * netif_queue_stopped - test if transmit queue is flowblocked
2865 * @dev: network device
2867 * Test if transmit queue on device is currently unable to send.
2869 static inline bool netif_queue_stopped(const struct net_device *dev)
2871 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2874 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2876 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2880 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2882 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2886 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2888 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2892 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2893 * @dev_queue: pointer to transmit queue
2895 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2896 * to give appropriate hint to the CPU.
2898 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2901 prefetchw(&dev_queue->dql.num_queued);
2906 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2907 * @dev_queue: pointer to transmit queue
2909 * BQL enabled drivers might use this helper in their TX completion path,
2910 * to give appropriate hint to the CPU.
2912 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2915 prefetchw(&dev_queue->dql.limit);
2919 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2923 dql_queued(&dev_queue->dql, bytes);
2925 if (likely(dql_avail(&dev_queue->dql) >= 0))
2928 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2931 * The XOFF flag must be set before checking the dql_avail below,
2932 * because in netdev_tx_completed_queue we update the dql_completed
2933 * before checking the XOFF flag.
2937 /* check again in case another CPU has just made room avail */
2938 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2939 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2944 * netdev_sent_queue - report the number of bytes queued to hardware
2945 * @dev: network device
2946 * @bytes: number of bytes queued to the hardware device queue
2948 * Report the number of bytes queued for sending/completion to the network
2949 * device hardware queue. @bytes should be a good approximation and should
2950 * exactly match netdev_completed_queue() @bytes
2952 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2954 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2957 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2958 unsigned int pkts, unsigned int bytes)
2961 if (unlikely(!bytes))
2964 dql_completed(&dev_queue->dql, bytes);
2967 * Without the memory barrier there is a small possiblity that
2968 * netdev_tx_sent_queue will miss the update and cause the queue to
2969 * be stopped forever
2973 if (dql_avail(&dev_queue->dql) < 0)
2976 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2977 netif_schedule_queue(dev_queue);
2982 * netdev_completed_queue - report bytes and packets completed by device
2983 * @dev: network device
2984 * @pkts: actual number of packets sent over the medium
2985 * @bytes: actual number of bytes sent over the medium
2987 * Report the number of bytes and packets transmitted by the network device
2988 * hardware queue over the physical medium, @bytes must exactly match the
2989 * @bytes amount passed to netdev_sent_queue()
2991 static inline void netdev_completed_queue(struct net_device *dev,
2992 unsigned int pkts, unsigned int bytes)
2994 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2997 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3000 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3006 * netdev_reset_queue - reset the packets and bytes count of a network device
3007 * @dev_queue: network device
3009 * Reset the bytes and packet count of a network device and clear the
3010 * software flow control OFF bit for this network device
3012 static inline void netdev_reset_queue(struct net_device *dev_queue)
3014 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3018 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3019 * @dev: network device
3020 * @queue_index: given tx queue index
3022 * Returns 0 if given tx queue index >= number of device tx queues,
3023 * otherwise returns the originally passed tx queue index.
3025 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3027 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3028 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3029 dev->name, queue_index,
3030 dev->real_num_tx_queues);
3038 * netif_running - test if up
3039 * @dev: network device
3041 * Test if the device has been brought up.
3043 static inline bool netif_running(const struct net_device *dev)
3045 return test_bit(__LINK_STATE_START, &dev->state);
3049 * Routines to manage the subqueues on a device. We only need start,
3050 * stop, and a check if it's stopped. All other device management is
3051 * done at the overall netdevice level.
3052 * Also test the device if we're multiqueue.
3056 * netif_start_subqueue - allow sending packets on subqueue
3057 * @dev: network device
3058 * @queue_index: sub queue index
3060 * Start individual transmit queue of a device with multiple transmit queues.
3062 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3064 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3066 netif_tx_start_queue(txq);
3070 * netif_stop_subqueue - stop sending packets on subqueue
3071 * @dev: network device
3072 * @queue_index: sub queue index
3074 * Stop individual transmit queue of a device with multiple transmit queues.
3076 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3078 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3079 netif_tx_stop_queue(txq);
3083 * netif_subqueue_stopped - test status of subqueue
3084 * @dev: network device
3085 * @queue_index: sub queue index
3087 * Check individual transmit queue of a device with multiple transmit queues.
3089 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3092 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3094 return netif_tx_queue_stopped(txq);
3097 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3098 struct sk_buff *skb)
3100 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3103 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3106 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3109 static inline int netif_set_xps_queue(struct net_device *dev,
3110 const struct cpumask *mask,
3117 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3118 unsigned int num_tx_queues);
3121 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3122 * as a distribution range limit for the returned value.
3124 static inline u16 skb_tx_hash(const struct net_device *dev,
3125 struct sk_buff *skb)
3127 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3131 * netif_is_multiqueue - test if device has multiple transmit queues
3132 * @dev: network device
3134 * Check if device has multiple transmit queues
3136 static inline bool netif_is_multiqueue(const struct net_device *dev)
3138 return dev->num_tx_queues > 1;
3141 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3144 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3146 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3154 static inline unsigned int get_netdev_rx_queue_index(
3155 struct netdev_rx_queue *queue)
3157 struct net_device *dev = queue->dev;
3158 int index = queue - dev->_rx;
3160 BUG_ON(index >= dev->num_rx_queues);
3165 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3166 int netif_get_num_default_rss_queues(void);
3168 enum skb_free_reason {
3169 SKB_REASON_CONSUMED,
3173 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3174 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3177 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3178 * interrupt context or with hardware interrupts being disabled.
3179 * (in_irq() || irqs_disabled())
3181 * We provide four helpers that can be used in following contexts :
3183 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3184 * replacing kfree_skb(skb)
3186 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3187 * Typically used in place of consume_skb(skb) in TX completion path
3189 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3190 * replacing kfree_skb(skb)
3192 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3193 * and consumed a packet. Used in place of consume_skb(skb)
3195 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3197 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3200 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3202 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3205 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3207 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3210 static inline void dev_consume_skb_any(struct sk_buff *skb)
3212 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3215 int netif_rx(struct sk_buff *skb);
3216 int netif_rx_ni(struct sk_buff *skb);
3217 int netif_receive_skb(struct sk_buff *skb);
3218 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3219 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3220 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3221 gro_result_t napi_gro_frags(struct napi_struct *napi);
3222 struct packet_offload *gro_find_receive_by_type(__be16 type);
3223 struct packet_offload *gro_find_complete_by_type(__be16 type);
3225 static inline void napi_free_frags(struct napi_struct *napi)
3227 kfree_skb(napi->skb);
3231 bool netdev_is_rx_handler_busy(struct net_device *dev);
3232 int netdev_rx_handler_register(struct net_device *dev,
3233 rx_handler_func_t *rx_handler,
3234 void *rx_handler_data);
3235 void netdev_rx_handler_unregister(struct net_device *dev);
3237 bool dev_valid_name(const char *name);
3238 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3239 int dev_ethtool(struct net *net, struct ifreq *);
3240 unsigned int dev_get_flags(const struct net_device *);
3241 int __dev_change_flags(struct net_device *, unsigned int flags);
3242 int dev_change_flags(struct net_device *, unsigned int);
3243 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3244 unsigned int gchanges);
3245 int dev_change_name(struct net_device *, const char *);
3246 int dev_set_alias(struct net_device *, const char *, size_t);
3247 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3248 int dev_set_mtu(struct net_device *, int);
3249 void dev_set_group(struct net_device *, int);
3250 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3251 int dev_change_carrier(struct net_device *, bool new_carrier);
3252 int dev_get_phys_port_id(struct net_device *dev,
3253 struct netdev_phys_item_id *ppid);
3254 int dev_get_phys_port_name(struct net_device *dev,
3255 char *name, size_t len);
3256 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3257 int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
3258 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3259 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3260 struct netdev_queue *txq, int *ret);
3261 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3262 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3263 bool is_skb_forwardable(const struct net_device *dev,
3264 const struct sk_buff *skb);
3266 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3267 struct sk_buff *skb)
3269 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3270 unlikely(!is_skb_forwardable(dev, skb))) {
3271 atomic_long_inc(&dev->rx_dropped);
3276 skb_scrub_packet(skb, true);
3281 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3283 extern int netdev_budget;
3285 /* Called by rtnetlink.c:rtnl_unlock() */
3286 void netdev_run_todo(void);
3289 * dev_put - release reference to device
3290 * @dev: network device
3292 * Release reference to device to allow it to be freed.
3294 static inline void dev_put(struct net_device *dev)
3296 this_cpu_dec(*dev->pcpu_refcnt);
3300 * dev_hold - get reference to device
3301 * @dev: network device
3303 * Hold reference to device to keep it from being freed.
3305 static inline void dev_hold(struct net_device *dev)
3307 this_cpu_inc(*dev->pcpu_refcnt);
3310 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3311 * and _off may be called from IRQ context, but it is caller
3312 * who is responsible for serialization of these calls.
3314 * The name carrier is inappropriate, these functions should really be
3315 * called netif_lowerlayer_*() because they represent the state of any
3316 * kind of lower layer not just hardware media.
3319 void linkwatch_init_dev(struct net_device *dev);
3320 void linkwatch_fire_event(struct net_device *dev);
3321 void linkwatch_forget_dev(struct net_device *dev);
3324 * netif_carrier_ok - test if carrier present
3325 * @dev: network device
3327 * Check if carrier is present on device
3329 static inline bool netif_carrier_ok(const struct net_device *dev)
3331 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3334 unsigned long dev_trans_start(struct net_device *dev);
3336 void __netdev_watchdog_up(struct net_device *dev);
3338 void netif_carrier_on(struct net_device *dev);
3340 void netif_carrier_off(struct net_device *dev);
3343 * netif_dormant_on - mark device as dormant.
3344 * @dev: network device
3346 * Mark device as dormant (as per RFC2863).
3348 * The dormant state indicates that the relevant interface is not
3349 * actually in a condition to pass packets (i.e., it is not 'up') but is
3350 * in a "pending" state, waiting for some external event. For "on-
3351 * demand" interfaces, this new state identifies the situation where the
3352 * interface is waiting for events to place it in the up state.
3354 static inline void netif_dormant_on(struct net_device *dev)
3356 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3357 linkwatch_fire_event(dev);
3361 * netif_dormant_off - set device as not dormant.
3362 * @dev: network device
3364 * Device is not in dormant state.
3366 static inline void netif_dormant_off(struct net_device *dev)
3368 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3369 linkwatch_fire_event(dev);
3373 * netif_dormant - test if carrier present
3374 * @dev: network device
3376 * Check if carrier is present on device
3378 static inline bool netif_dormant(const struct net_device *dev)
3380 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3385 * netif_oper_up - test if device is operational
3386 * @dev: network device
3388 * Check if carrier is operational
3390 static inline bool netif_oper_up(const struct net_device *dev)
3392 return (dev->operstate == IF_OPER_UP ||
3393 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3397 * netif_device_present - is device available or removed
3398 * @dev: network device
3400 * Check if device has not been removed from system.
3402 static inline bool netif_device_present(struct net_device *dev)
3404 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3407 void netif_device_detach(struct net_device *dev);
3409 void netif_device_attach(struct net_device *dev);
3412 * Network interface message level settings
3416 NETIF_MSG_DRV = 0x0001,
3417 NETIF_MSG_PROBE = 0x0002,
3418 NETIF_MSG_LINK = 0x0004,
3419 NETIF_MSG_TIMER = 0x0008,
3420 NETIF_MSG_IFDOWN = 0x0010,
3421 NETIF_MSG_IFUP = 0x0020,
3422 NETIF_MSG_RX_ERR = 0x0040,
3423 NETIF_MSG_TX_ERR = 0x0080,
3424 NETIF_MSG_TX_QUEUED = 0x0100,
3425 NETIF_MSG_INTR = 0x0200,
3426 NETIF_MSG_TX_DONE = 0x0400,
3427 NETIF_MSG_RX_STATUS = 0x0800,
3428 NETIF_MSG_PKTDATA = 0x1000,
3429 NETIF_MSG_HW = 0x2000,
3430 NETIF_MSG_WOL = 0x4000,
3433 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3434 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3435 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3436 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3437 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3438 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3439 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3440 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3441 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3442 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3443 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3444 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3445 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3446 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3447 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3449 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3452 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3453 return default_msg_enable_bits;
3454 if (debug_value == 0) /* no output */
3456 /* set low N bits */
3457 return (1 << debug_value) - 1;
3460 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3462 spin_lock(&txq->_xmit_lock);
3463 txq->xmit_lock_owner = cpu;
3466 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3468 __acquire(&txq->_xmit_lock);
3472 static inline void __netif_tx_release(struct netdev_queue *txq)
3474 __release(&txq->_xmit_lock);
3477 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3479 spin_lock_bh(&txq->_xmit_lock);
3480 txq->xmit_lock_owner = smp_processor_id();
3483 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3485 bool ok = spin_trylock(&txq->_xmit_lock);
3487 txq->xmit_lock_owner = smp_processor_id();
3491 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3493 txq->xmit_lock_owner = -1;
3494 spin_unlock(&txq->_xmit_lock);
3497 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3499 txq->xmit_lock_owner = -1;
3500 spin_unlock_bh(&txq->_xmit_lock);
3503 static inline void txq_trans_update(struct netdev_queue *txq)
3505 if (txq->xmit_lock_owner != -1)
3506 txq->trans_start = jiffies;
3509 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3510 static inline void netif_trans_update(struct net_device *dev)
3512 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3514 if (txq->trans_start != jiffies)
3515 txq->trans_start = jiffies;
3519 * netif_tx_lock - grab network device transmit lock
3520 * @dev: network device
3522 * Get network device transmit lock
3524 static inline void netif_tx_lock(struct net_device *dev)
3529 spin_lock(&dev->tx_global_lock);
3530 cpu = smp_processor_id();
3531 for (i = 0; i < dev->num_tx_queues; i++) {
3532 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3534 /* We are the only thread of execution doing a
3535 * freeze, but we have to grab the _xmit_lock in
3536 * order to synchronize with threads which are in
3537 * the ->hard_start_xmit() handler and already
3538 * checked the frozen bit.
3540 __netif_tx_lock(txq, cpu);
3541 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3542 __netif_tx_unlock(txq);
3546 static inline void netif_tx_lock_bh(struct net_device *dev)
3552 static inline void netif_tx_unlock(struct net_device *dev)
3556 for (i = 0; i < dev->num_tx_queues; i++) {
3557 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3559 /* No need to grab the _xmit_lock here. If the
3560 * queue is not stopped for another reason, we
3563 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3564 netif_schedule_queue(txq);
3566 spin_unlock(&dev->tx_global_lock);
3569 static inline void netif_tx_unlock_bh(struct net_device *dev)
3571 netif_tx_unlock(dev);
3575 #define HARD_TX_LOCK(dev, txq, cpu) { \
3576 if ((dev->features & NETIF_F_LLTX) == 0) { \
3577 __netif_tx_lock(txq, cpu); \
3579 __netif_tx_acquire(txq); \
3583 #define HARD_TX_TRYLOCK(dev, txq) \
3584 (((dev->features & NETIF_F_LLTX) == 0) ? \
3585 __netif_tx_trylock(txq) : \
3586 __netif_tx_acquire(txq))
3588 #define HARD_TX_UNLOCK(dev, txq) { \
3589 if ((dev->features & NETIF_F_LLTX) == 0) { \
3590 __netif_tx_unlock(txq); \
3592 __netif_tx_release(txq); \
3596 static inline void netif_tx_disable(struct net_device *dev)
3602 cpu = smp_processor_id();
3603 for (i = 0; i < dev->num_tx_queues; i++) {
3604 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3606 __netif_tx_lock(txq, cpu);
3607 netif_tx_stop_queue(txq);
3608 __netif_tx_unlock(txq);
3613 static inline void netif_addr_lock(struct net_device *dev)
3615 spin_lock(&dev->addr_list_lock);
3618 static inline void netif_addr_lock_nested(struct net_device *dev)
3620 int subclass = SINGLE_DEPTH_NESTING;
3622 if (dev->netdev_ops->ndo_get_lock_subclass)
3623 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3625 spin_lock_nested(&dev->addr_list_lock, subclass);
3628 static inline void netif_addr_lock_bh(struct net_device *dev)
3630 spin_lock_bh(&dev->addr_list_lock);
3633 static inline void netif_addr_unlock(struct net_device *dev)
3635 spin_unlock(&dev->addr_list_lock);
3638 static inline void netif_addr_unlock_bh(struct net_device *dev)
3640 spin_unlock_bh(&dev->addr_list_lock);
3644 * dev_addrs walker. Should be used only for read access. Call with
3645 * rcu_read_lock held.
3647 #define for_each_dev_addr(dev, ha) \
3648 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3650 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3652 void ether_setup(struct net_device *dev);
3654 /* Support for loadable net-drivers */
3655 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3656 unsigned char name_assign_type,
3657 void (*setup)(struct net_device *),
3658 unsigned int txqs, unsigned int rxqs);
3659 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3660 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3662 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3663 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3666 int register_netdev(struct net_device *dev);
3667 void unregister_netdev(struct net_device *dev);
3669 /* General hardware address lists handling functions */
3670 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3671 struct netdev_hw_addr_list *from_list, int addr_len);
3672 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3673 struct netdev_hw_addr_list *from_list, int addr_len);
3674 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3675 struct net_device *dev,
3676 int (*sync)(struct net_device *, const unsigned char *),
3677 int (*unsync)(struct net_device *,
3678 const unsigned char *));
3679 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3680 struct net_device *dev,
3681 int (*unsync)(struct net_device *,
3682 const unsigned char *));
3683 void __hw_addr_init(struct netdev_hw_addr_list *list);
3685 /* Functions used for device addresses handling */
3686 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3687 unsigned char addr_type);
3688 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3689 unsigned char addr_type);
3690 void dev_addr_flush(struct net_device *dev);
3691 int dev_addr_init(struct net_device *dev);
3693 /* Functions used for unicast addresses handling */
3694 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3695 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3696 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3697 int dev_uc_sync(struct net_device *to, struct net_device *from);
3698 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3699 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3700 void dev_uc_flush(struct net_device *dev);
3701 void dev_uc_init(struct net_device *dev);
3704 * __dev_uc_sync - Synchonize device's unicast list
3705 * @dev: device to sync
3706 * @sync: function to call if address should be added
3707 * @unsync: function to call if address should be removed
3709 * Add newly added addresses to the interface, and release
3710 * addresses that have been deleted.
3712 static inline int __dev_uc_sync(struct net_device *dev,
3713 int (*sync)(struct net_device *,
3714 const unsigned char *),
3715 int (*unsync)(struct net_device *,
3716 const unsigned char *))
3718 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3722 * __dev_uc_unsync - Remove synchronized addresses from device
3723 * @dev: device to sync
3724 * @unsync: function to call if address should be removed
3726 * Remove all addresses that were added to the device by dev_uc_sync().
3728 static inline void __dev_uc_unsync(struct net_device *dev,
3729 int (*unsync)(struct net_device *,
3730 const unsigned char *))
3732 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3735 /* Functions used for multicast addresses handling */
3736 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3737 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3738 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3739 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3740 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3741 int dev_mc_sync(struct net_device *to, struct net_device *from);
3742 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3743 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3744 void dev_mc_flush(struct net_device *dev);
3745 void dev_mc_init(struct net_device *dev);
3748 * __dev_mc_sync - Synchonize device's multicast list
3749 * @dev: device to sync
3750 * @sync: function to call if address should be added
3751 * @unsync: function to call if address should be removed
3753 * Add newly added addresses to the interface, and release
3754 * addresses that have been deleted.
3756 static inline int __dev_mc_sync(struct net_device *dev,
3757 int (*sync)(struct net_device *,
3758 const unsigned char *),
3759 int (*unsync)(struct net_device *,
3760 const unsigned char *))
3762 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3766 * __dev_mc_unsync - Remove synchronized addresses from device
3767 * @dev: device to sync
3768 * @unsync: function to call if address should be removed
3770 * Remove all addresses that were added to the device by dev_mc_sync().
3772 static inline void __dev_mc_unsync(struct net_device *dev,
3773 int (*unsync)(struct net_device *,
3774 const unsigned char *))
3776 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3779 /* Functions used for secondary unicast and multicast support */
3780 void dev_set_rx_mode(struct net_device *dev);
3781 void __dev_set_rx_mode(struct net_device *dev);
3782 int dev_set_promiscuity(struct net_device *dev, int inc);
3783 int dev_set_allmulti(struct net_device *dev, int inc);
3784 void netdev_state_change(struct net_device *dev);
3785 void netdev_notify_peers(struct net_device *dev);
3786 void netdev_features_change(struct net_device *dev);
3787 /* Load a device via the kmod */
3788 void dev_load(struct net *net, const char *name);
3789 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3790 struct rtnl_link_stats64 *storage);
3791 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3792 const struct net_device_stats *netdev_stats);
3794 extern int netdev_max_backlog;
3795 extern int netdev_tstamp_prequeue;
3796 extern int weight_p;
3798 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3799 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3800 struct list_head **iter);
3801 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3802 struct list_head **iter);
3804 /* iterate through upper list, must be called under RCU read lock */
3805 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3806 for (iter = &(dev)->adj_list.upper, \
3807 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3809 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3811 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3812 int (*fn)(struct net_device *upper_dev,
3816 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3817 struct net_device *upper_dev);
3819 void *netdev_lower_get_next_private(struct net_device *dev,
3820 struct list_head **iter);
3821 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3822 struct list_head **iter);
3824 #define netdev_for_each_lower_private(dev, priv, iter) \
3825 for (iter = (dev)->adj_list.lower.next, \
3826 priv = netdev_lower_get_next_private(dev, &(iter)); \
3828 priv = netdev_lower_get_next_private(dev, &(iter)))
3830 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3831 for (iter = &(dev)->adj_list.lower, \
3832 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3834 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3836 void *netdev_lower_get_next(struct net_device *dev,
3837 struct list_head **iter);
3839 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3840 for (iter = (dev)->adj_list.lower.next, \
3841 ldev = netdev_lower_get_next(dev, &(iter)); \
3843 ldev = netdev_lower_get_next(dev, &(iter)))
3845 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3846 struct list_head **iter);
3847 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3848 struct list_head **iter);
3850 int netdev_walk_all_lower_dev(struct net_device *dev,
3851 int (*fn)(struct net_device *lower_dev,
3854 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3855 int (*fn)(struct net_device *lower_dev,
3859 void *netdev_adjacent_get_private(struct list_head *adj_list);
3860 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3861 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3862 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3863 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3864 int netdev_master_upper_dev_link(struct net_device *dev,
3865 struct net_device *upper_dev,
3866 void *upper_priv, void *upper_info);
3867 void netdev_upper_dev_unlink(struct net_device *dev,
3868 struct net_device *upper_dev);
3869 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3870 void *netdev_lower_dev_get_private(struct net_device *dev,
3871 struct net_device *lower_dev);
3872 void netdev_lower_state_changed(struct net_device *lower_dev,
3873 void *lower_state_info);
3874 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3875 struct neighbour *n);
3876 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3877 struct neighbour *n);
3879 /* RSS keys are 40 or 52 bytes long */
3880 #define NETDEV_RSS_KEY_LEN 52
3881 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3882 void netdev_rss_key_fill(void *buffer, size_t len);
3884 int dev_get_nest_level(struct net_device *dev);
3885 int skb_checksum_help(struct sk_buff *skb);
3886 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3887 netdev_features_t features, bool tx_path);
3888 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3889 netdev_features_t features);
3891 struct netdev_bonding_info {
3896 struct netdev_notifier_bonding_info {
3897 struct netdev_notifier_info info; /* must be first */
3898 struct netdev_bonding_info bonding_info;
3901 void netdev_bonding_info_change(struct net_device *dev,
3902 struct netdev_bonding_info *bonding_info);
3905 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3907 return __skb_gso_segment(skb, features, true);
3909 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3911 static inline bool can_checksum_protocol(netdev_features_t features,
3914 if (protocol == htons(ETH_P_FCOE))
3915 return !!(features & NETIF_F_FCOE_CRC);
3917 /* Assume this is an IP checksum (not SCTP CRC) */
3919 if (features & NETIF_F_HW_CSUM) {
3920 /* Can checksum everything */
3925 case htons(ETH_P_IP):
3926 return !!(features & NETIF_F_IP_CSUM);
3927 case htons(ETH_P_IPV6):
3928 return !!(features & NETIF_F_IPV6_CSUM);
3935 void netdev_rx_csum_fault(struct net_device *dev);
3937 static inline void netdev_rx_csum_fault(struct net_device *dev)
3941 /* rx skb timestamps */
3942 void net_enable_timestamp(void);
3943 void net_disable_timestamp(void);
3945 #ifdef CONFIG_PROC_FS
3946 int __init dev_proc_init(void);
3948 #define dev_proc_init() 0
3951 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3952 struct sk_buff *skb, struct net_device *dev,
3955 skb->xmit_more = more ? 1 : 0;
3956 return ops->ndo_start_xmit(skb, dev);
3959 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3960 struct netdev_queue *txq, bool more)
3962 const struct net_device_ops *ops = dev->netdev_ops;
3965 rc = __netdev_start_xmit(ops, skb, dev, more);
3966 if (rc == NETDEV_TX_OK)
3967 txq_trans_update(txq);
3972 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3974 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3977 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3979 return netdev_class_create_file_ns(class_attr, NULL);
3982 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3984 netdev_class_remove_file_ns(class_attr, NULL);
3987 extern struct kobj_ns_type_operations net_ns_type_operations;
3989 const char *netdev_drivername(const struct net_device *dev);
3991 void linkwatch_run_queue(void);
3993 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3994 netdev_features_t f2)
3996 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3997 if (f1 & NETIF_F_HW_CSUM)
3998 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4000 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4006 static inline netdev_features_t netdev_get_wanted_features(
4007 struct net_device *dev)
4009 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4011 netdev_features_t netdev_increment_features(netdev_features_t all,
4012 netdev_features_t one, netdev_features_t mask);
4014 /* Allow TSO being used on stacked device :
4015 * Performing the GSO segmentation before last device
4016 * is a performance improvement.
4018 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4019 netdev_features_t mask)
4021 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4024 int __netdev_update_features(struct net_device *dev);
4025 void netdev_update_features(struct net_device *dev);
4026 void netdev_change_features(struct net_device *dev);
4028 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4029 struct net_device *dev);
4031 netdev_features_t passthru_features_check(struct sk_buff *skb,
4032 struct net_device *dev,
4033 netdev_features_t features);
4034 netdev_features_t netif_skb_features(struct sk_buff *skb);
4036 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4038 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4040 /* check flags correspondence */
4041 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4042 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4043 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4044 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4045 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4046 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4047 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4048 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4049 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4050 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4051 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4052 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4053 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4054 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4055 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4056 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4058 return (features & feature) == feature;
4061 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4063 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4064 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4067 static inline bool netif_needs_gso(struct sk_buff *skb,
4068 netdev_features_t features)
4070 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4071 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4072 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4075 static inline void netif_set_gso_max_size(struct net_device *dev,
4078 dev->gso_max_size = size;
4081 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4082 int pulled_hlen, u16 mac_offset,
4085 skb->protocol = protocol;
4086 skb->encapsulation = 1;
4087 skb_push(skb, pulled_hlen);
4088 skb_reset_transport_header(skb);
4089 skb->mac_header = mac_offset;
4090 skb->network_header = skb->mac_header + mac_len;
4091 skb->mac_len = mac_len;
4094 static inline bool netif_is_macsec(const struct net_device *dev)
4096 return dev->priv_flags & IFF_MACSEC;
4099 static inline bool netif_is_macvlan(const struct net_device *dev)
4101 return dev->priv_flags & IFF_MACVLAN;
4104 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4106 return dev->priv_flags & IFF_MACVLAN_PORT;
4109 static inline bool netif_is_ipvlan(const struct net_device *dev)
4111 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4114 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4116 return dev->priv_flags & IFF_IPVLAN_MASTER;
4119 static inline bool netif_is_bond_master(const struct net_device *dev)
4121 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4124 static inline bool netif_is_bond_slave(const struct net_device *dev)
4126 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4129 static inline bool netif_supports_nofcs(struct net_device *dev)
4131 return dev->priv_flags & IFF_SUPP_NOFCS;
4134 static inline bool netif_is_l3_master(const struct net_device *dev)
4136 return dev->priv_flags & IFF_L3MDEV_MASTER;
4139 static inline bool netif_is_l3_slave(const struct net_device *dev)
4141 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4144 static inline bool netif_is_bridge_master(const struct net_device *dev)
4146 return dev->priv_flags & IFF_EBRIDGE;
4149 static inline bool netif_is_bridge_port(const struct net_device *dev)
4151 return dev->priv_flags & IFF_BRIDGE_PORT;
4154 static inline bool netif_is_ovs_master(const struct net_device *dev)
4156 return dev->priv_flags & IFF_OPENVSWITCH;
4159 static inline bool netif_is_team_master(const struct net_device *dev)
4161 return dev->priv_flags & IFF_TEAM;
4164 static inline bool netif_is_team_port(const struct net_device *dev)
4166 return dev->priv_flags & IFF_TEAM_PORT;
4169 static inline bool netif_is_lag_master(const struct net_device *dev)
4171 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4174 static inline bool netif_is_lag_port(const struct net_device *dev)
4176 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4179 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4181 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4184 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4185 static inline void netif_keep_dst(struct net_device *dev)
4187 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4190 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4191 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4193 /* TODO: reserve and use an additional IFF bit, if we get more users */
4194 return dev->priv_flags & IFF_MACSEC;
4197 extern struct pernet_operations __net_initdata loopback_net_ops;
4199 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4201 /* netdev_printk helpers, similar to dev_printk */
4203 static inline const char *netdev_name(const struct net_device *dev)
4205 if (!dev->name[0] || strchr(dev->name, '%'))
4206 return "(unnamed net_device)";
4210 static inline const char *netdev_reg_state(const struct net_device *dev)
4212 switch (dev->reg_state) {
4213 case NETREG_UNINITIALIZED: return " (uninitialized)";
4214 case NETREG_REGISTERED: return "";
4215 case NETREG_UNREGISTERING: return " (unregistering)";
4216 case NETREG_UNREGISTERED: return " (unregistered)";
4217 case NETREG_RELEASED: return " (released)";
4218 case NETREG_DUMMY: return " (dummy)";
4221 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4222 return " (unknown)";
4226 void netdev_printk(const char *level, const struct net_device *dev,
4227 const char *format, ...);
4229 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4231 void netdev_alert(const struct net_device *dev, const char *format, ...);
4233 void netdev_crit(const struct net_device *dev, const char *format, ...);
4235 void netdev_err(const struct net_device *dev, const char *format, ...);
4237 void netdev_warn(const struct net_device *dev, const char *format, ...);
4239 void netdev_notice(const struct net_device *dev, const char *format, ...);
4241 void netdev_info(const struct net_device *dev, const char *format, ...);
4243 #define MODULE_ALIAS_NETDEV(device) \
4244 MODULE_ALIAS("netdev-" device)
4246 #if defined(CONFIG_DYNAMIC_DEBUG)
4247 #define netdev_dbg(__dev, format, args...) \
4249 dynamic_netdev_dbg(__dev, format, ##args); \
4251 #elif defined(DEBUG)
4252 #define netdev_dbg(__dev, format, args...) \
4253 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4255 #define netdev_dbg(__dev, format, args...) \
4258 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4262 #if defined(VERBOSE_DEBUG)
4263 #define netdev_vdbg netdev_dbg
4266 #define netdev_vdbg(dev, format, args...) \
4269 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4275 * netdev_WARN() acts like dev_printk(), but with the key difference
4276 * of using a WARN/WARN_ON to get the message out, including the
4277 * file/line information and a backtrace.
4279 #define netdev_WARN(dev, format, args...) \
4280 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4281 netdev_reg_state(dev), ##args)
4283 /* netif printk helpers, similar to netdev_printk */
4285 #define netif_printk(priv, type, level, dev, fmt, args...) \
4287 if (netif_msg_##type(priv)) \
4288 netdev_printk(level, (dev), fmt, ##args); \
4291 #define netif_level(level, priv, type, dev, fmt, args...) \
4293 if (netif_msg_##type(priv)) \
4294 netdev_##level(dev, fmt, ##args); \
4297 #define netif_emerg(priv, type, dev, fmt, args...) \
4298 netif_level(emerg, priv, type, dev, fmt, ##args)
4299 #define netif_alert(priv, type, dev, fmt, args...) \
4300 netif_level(alert, priv, type, dev, fmt, ##args)
4301 #define netif_crit(priv, type, dev, fmt, args...) \
4302 netif_level(crit, priv, type, dev, fmt, ##args)
4303 #define netif_err(priv, type, dev, fmt, args...) \
4304 netif_level(err, priv, type, dev, fmt, ##args)
4305 #define netif_warn(priv, type, dev, fmt, args...) \
4306 netif_level(warn, priv, type, dev, fmt, ##args)
4307 #define netif_notice(priv, type, dev, fmt, args...) \
4308 netif_level(notice, priv, type, dev, fmt, ##args)
4309 #define netif_info(priv, type, dev, fmt, args...) \
4310 netif_level(info, priv, type, dev, fmt, ##args)
4312 #if defined(CONFIG_DYNAMIC_DEBUG)
4313 #define netif_dbg(priv, type, netdev, format, args...) \
4315 if (netif_msg_##type(priv)) \
4316 dynamic_netdev_dbg(netdev, format, ##args); \
4318 #elif defined(DEBUG)
4319 #define netif_dbg(priv, type, dev, format, args...) \
4320 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4322 #define netif_dbg(priv, type, dev, format, args...) \
4325 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4330 #if defined(VERBOSE_DEBUG)
4331 #define netif_vdbg netif_dbg
4333 #define netif_vdbg(priv, type, dev, format, args...) \
4336 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4342 * The list of packet types we will receive (as opposed to discard)
4343 * and the routines to invoke.
4345 * Why 16. Because with 16 the only overlap we get on a hash of the
4346 * low nibble of the protocol value is RARP/SNAP/X.25.
4348 * NOTE: That is no longer true with the addition of VLAN tags. Not
4349 * sure which should go first, but I bet it won't make much
4350 * difference if we are running VLANs. The good news is that
4351 * this protocol won't be in the list unless compiled in, so
4352 * the average user (w/out VLANs) will not be adversely affected.
4368 #define PTYPE_HASH_SIZE (16)
4369 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4371 #endif /* _LINUX_NETDEVICE_H */