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/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <linux/prefetch.h>
34 #include <asm/cache.h>
35 #include <asm/byteorder.h>
37 #include <linux/percpu.h>
38 #include <linux/rculist.h>
39 #include <linux/dmaengine.h>
40 #include <linux/workqueue.h>
41 #include <linux/dynamic_queue_limits.h>
43 #include <linux/ethtool.h>
44 #include <net/net_namespace.h>
47 #include <net/dcbnl.h>
49 #include <net/netprio_cgroup.h>
51 #include <linux/netdev_features.h>
52 #include <linux/neighbour.h>
53 #include <uapi/linux/netdevice.h>
54 #include <uapi/linux/if_bonding.h>
61 /* 802.15.4 specific */
65 void netdev_set_default_ethtool_ops(struct net_device *dev,
66 const struct ethtool_ops *ops);
68 /* Backlog congestion levels */
69 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
70 #define NET_RX_DROP 1 /* packet dropped */
73 * Transmit return codes: transmit return codes originate from three different
76 * - qdisc return codes
77 * - driver transmit return codes
80 * Drivers are allowed to return any one of those in their hard_start_xmit()
81 * function. Real network devices commonly used with qdiscs should only return
82 * the driver transmit return codes though - when qdiscs are used, the actual
83 * transmission happens asynchronously, so the value is not propagated to
84 * higher layers. Virtual network devices transmit synchronously, in this case
85 * the driver transmit return codes are consumed by dev_queue_xmit(), all
86 * others are propagated to higher layers.
89 /* qdisc ->enqueue() return codes. */
90 #define NET_XMIT_SUCCESS 0x00
91 #define NET_XMIT_DROP 0x01 /* skb dropped */
92 #define NET_XMIT_CN 0x02 /* congestion notification */
93 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
94 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
96 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
97 * indicates that the device will soon be dropping packets, or already drops
98 * some packets of the same priority; prompting us to send less aggressively. */
99 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
100 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
102 /* Driver transmit return codes */
103 #define NETDEV_TX_MASK 0xf0
106 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
107 NETDEV_TX_OK = 0x00, /* driver took care of packet */
108 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
109 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
111 typedef enum netdev_tx netdev_tx_t;
114 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
115 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
117 static inline bool dev_xmit_complete(int rc)
120 * Positive cases with an skb consumed by a driver:
121 * - successful transmission (rc == NETDEV_TX_OK)
122 * - error while transmitting (rc < 0)
123 * - error while queueing to a different device (rc & NET_XMIT_MASK)
125 if (likely(rc < NET_XMIT_MASK))
132 * Compute the worst case header length according to the protocols
136 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
137 # if defined(CONFIG_MAC80211_MESH)
138 # define LL_MAX_HEADER 128
140 # define LL_MAX_HEADER 96
143 # define LL_MAX_HEADER 32
146 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
147 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
148 #define MAX_HEADER LL_MAX_HEADER
150 #define MAX_HEADER (LL_MAX_HEADER + 48)
154 * Old network device statistics. Fields are native words
155 * (unsigned long) so they can be read and written atomically.
158 struct net_device_stats {
159 unsigned long rx_packets;
160 unsigned long tx_packets;
161 unsigned long rx_bytes;
162 unsigned long tx_bytes;
163 unsigned long rx_errors;
164 unsigned long tx_errors;
165 unsigned long rx_dropped;
166 unsigned long tx_dropped;
167 unsigned long multicast;
168 unsigned long collisions;
169 unsigned long rx_length_errors;
170 unsigned long rx_over_errors;
171 unsigned long rx_crc_errors;
172 unsigned long rx_frame_errors;
173 unsigned long rx_fifo_errors;
174 unsigned long rx_missed_errors;
175 unsigned long tx_aborted_errors;
176 unsigned long tx_carrier_errors;
177 unsigned long tx_fifo_errors;
178 unsigned long tx_heartbeat_errors;
179 unsigned long tx_window_errors;
180 unsigned long rx_compressed;
181 unsigned long tx_compressed;
185 #include <linux/cache.h>
186 #include <linux/skbuff.h>
189 #include <linux/static_key.h>
190 extern struct static_key rps_needed;
197 struct netdev_hw_addr {
198 struct list_head list;
199 unsigned char addr[MAX_ADDR_LEN];
201 #define NETDEV_HW_ADDR_T_LAN 1
202 #define NETDEV_HW_ADDR_T_SAN 2
203 #define NETDEV_HW_ADDR_T_SLAVE 3
204 #define NETDEV_HW_ADDR_T_UNICAST 4
205 #define NETDEV_HW_ADDR_T_MULTICAST 5
210 struct rcu_head rcu_head;
213 struct netdev_hw_addr_list {
214 struct list_head list;
218 #define netdev_hw_addr_list_count(l) ((l)->count)
219 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
220 #define netdev_hw_addr_list_for_each(ha, l) \
221 list_for_each_entry(ha, &(l)->list, list)
223 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
224 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
225 #define netdev_for_each_uc_addr(ha, dev) \
226 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
228 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
229 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
230 #define netdev_for_each_mc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
238 /* cached hardware header; allow for machine alignment needs. */
239 #define HH_DATA_MOD 16
240 #define HH_DATA_OFF(__len) \
241 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
242 #define HH_DATA_ALIGN(__len) \
243 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
244 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
247 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
249 * dev->hard_header_len ? (dev->hard_header_len +
250 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
252 * We could use other alignment values, but we must maintain the
253 * relationship HH alignment <= LL alignment.
255 #define LL_RESERVED_SPACE(dev) \
256 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
257 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
258 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
261 int (*create) (struct sk_buff *skb, struct net_device *dev,
262 unsigned short type, const void *daddr,
263 const void *saddr, unsigned int len);
264 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
265 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
266 void (*cache_update)(struct hh_cache *hh,
267 const struct net_device *dev,
268 const unsigned char *haddr);
271 /* These flag bits are private to the generic network queueing
272 * layer, they may not be explicitly referenced by any other
276 enum netdev_state_t {
278 __LINK_STATE_PRESENT,
279 __LINK_STATE_NOCARRIER,
280 __LINK_STATE_LINKWATCH_PENDING,
281 __LINK_STATE_DORMANT,
286 * This structure holds at boot time configured netdevice settings. They
287 * are then used in the device probing.
289 struct netdev_boot_setup {
293 #define NETDEV_BOOT_SETUP_MAX 8
295 int __init netdev_boot_setup(char *str);
298 * Structure for NAPI scheduling similar to tasklet but with weighting
301 /* The poll_list must only be managed by the entity which
302 * changes the state of the NAPI_STATE_SCHED bit. This means
303 * whoever atomically sets that bit can add this napi_struct
304 * to the per-cpu poll_list, and whoever clears that bit
305 * can remove from the list right before clearing the bit.
307 struct list_head poll_list;
311 unsigned int gro_count;
312 int (*poll)(struct napi_struct *, int);
313 #ifdef CONFIG_NETPOLL
314 spinlock_t poll_lock;
317 struct net_device *dev;
318 struct sk_buff *gro_list;
320 struct hrtimer timer;
321 struct list_head dev_list;
322 struct hlist_node napi_hash_node;
323 unsigned int napi_id;
327 NAPI_STATE_SCHED, /* Poll is scheduled */
328 NAPI_STATE_DISABLE, /* Disable pending */
329 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
330 NAPI_STATE_HASHED, /* In NAPI hash */
340 typedef enum gro_result gro_result_t;
343 * enum rx_handler_result - Possible return values for rx_handlers.
344 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
346 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
347 * case skb->dev was changed by rx_handler.
348 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
349 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
351 * rx_handlers are functions called from inside __netif_receive_skb(), to do
352 * special processing of the skb, prior to delivery to protocol handlers.
354 * Currently, a net_device can only have a single rx_handler registered. Trying
355 * to register a second rx_handler will return -EBUSY.
357 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
358 * To unregister a rx_handler on a net_device, use
359 * netdev_rx_handler_unregister().
361 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
364 * If the rx_handler consumed to skb in some way, it should return
365 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
366 * the skb to be delivered in some other ways.
368 * If the rx_handler changed skb->dev, to divert the skb to another
369 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
370 * new device will be called if it exists.
372 * If the rx_handler consider the skb should be ignored, it should return
373 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
374 * are registered on exact device (ptype->dev == skb->dev).
376 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
377 * delivered, it should return RX_HANDLER_PASS.
379 * A device without a registered rx_handler will behave as if rx_handler
380 * returned RX_HANDLER_PASS.
383 enum rx_handler_result {
389 typedef enum rx_handler_result rx_handler_result_t;
390 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
392 void __napi_schedule(struct napi_struct *n);
393 void __napi_schedule_irqoff(struct napi_struct *n);
395 static inline bool napi_disable_pending(struct napi_struct *n)
397 return test_bit(NAPI_STATE_DISABLE, &n->state);
401 * napi_schedule_prep - check if napi can be scheduled
404 * Test if NAPI routine is already running, and if not mark
405 * it as running. This is used as a condition variable
406 * insure only one NAPI poll instance runs. We also make
407 * sure there is no pending NAPI disable.
409 static inline bool napi_schedule_prep(struct napi_struct *n)
411 return !napi_disable_pending(n) &&
412 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
416 * napi_schedule - schedule NAPI poll
419 * Schedule NAPI poll routine to be called if it is not already
422 static inline void napi_schedule(struct napi_struct *n)
424 if (napi_schedule_prep(n))
429 * napi_schedule_irqoff - schedule NAPI poll
432 * Variant of napi_schedule(), assuming hard irqs are masked.
434 static inline void napi_schedule_irqoff(struct napi_struct *n)
436 if (napi_schedule_prep(n))
437 __napi_schedule_irqoff(n);
440 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
441 static inline bool napi_reschedule(struct napi_struct *napi)
443 if (napi_schedule_prep(napi)) {
444 __napi_schedule(napi);
450 void __napi_complete(struct napi_struct *n);
451 void napi_complete_done(struct napi_struct *n, int work_done);
453 * napi_complete - NAPI processing complete
456 * Mark NAPI processing as complete.
457 * Consider using napi_complete_done() instead.
459 static inline void napi_complete(struct napi_struct *n)
461 return napi_complete_done(n, 0);
465 * napi_by_id - lookup a NAPI by napi_id
466 * @napi_id: hashed napi_id
468 * lookup @napi_id in napi_hash table
469 * must be called under rcu_read_lock()
471 struct napi_struct *napi_by_id(unsigned int napi_id);
474 * napi_hash_add - add a NAPI to global hashtable
475 * @napi: napi context
477 * generate a new napi_id and store a @napi under it in napi_hash
479 void napi_hash_add(struct napi_struct *napi);
482 * napi_hash_del - remove a NAPI from global table
483 * @napi: napi context
485 * Warning: caller must observe rcu grace period
486 * before freeing memory containing @napi
488 void napi_hash_del(struct napi_struct *napi);
491 * napi_disable - prevent NAPI from scheduling
494 * Stop NAPI from being scheduled on this context.
495 * Waits till any outstanding processing completes.
497 void napi_disable(struct napi_struct *n);
500 * napi_enable - enable NAPI scheduling
503 * Resume NAPI from being scheduled on this context.
504 * Must be paired with napi_disable.
506 static inline void napi_enable(struct napi_struct *n)
508 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
509 smp_mb__before_atomic();
510 clear_bit(NAPI_STATE_SCHED, &n->state);
515 * napi_synchronize - wait until NAPI is not running
518 * Wait until NAPI is done being scheduled on this context.
519 * Waits till any outstanding processing completes but
520 * does not disable future activations.
522 static inline void napi_synchronize(const struct napi_struct *n)
524 while (test_bit(NAPI_STATE_SCHED, &n->state))
528 # define napi_synchronize(n) barrier()
531 enum netdev_queue_state_t {
532 __QUEUE_STATE_DRV_XOFF,
533 __QUEUE_STATE_STACK_XOFF,
534 __QUEUE_STATE_FROZEN,
537 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
538 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
539 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
541 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
542 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
544 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
548 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
549 * netif_tx_* functions below are used to manipulate this flag. The
550 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
551 * queue independently. The netif_xmit_*stopped functions below are called
552 * to check if the queue has been stopped by the driver or stack (either
553 * of the XOFF bits are set in the state). Drivers should not need to call
554 * netif_xmit*stopped functions, they should only be using netif_tx_*.
557 struct netdev_queue {
561 struct net_device *dev;
562 struct Qdisc __rcu *qdisc;
563 struct Qdisc *qdisc_sleeping;
567 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
576 * please use this field instead of dev->trans_start
578 unsigned long trans_start;
581 * Number of TX timeouts for this queue
582 * (/sys/class/net/DEV/Q/trans_timeout)
584 unsigned long trans_timeout;
591 unsigned long tx_maxrate;
592 } ____cacheline_aligned_in_smp;
594 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
596 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
603 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
605 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
612 * This structure holds an RPS map which can be of variable length. The
613 * map is an array of CPUs.
620 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
623 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
624 * tail pointer for that CPU's input queue at the time of last enqueue, and
625 * a hardware filter index.
627 struct rps_dev_flow {
630 unsigned int last_qtail;
632 #define RPS_NO_FILTER 0xffff
635 * The rps_dev_flow_table structure contains a table of flow mappings.
637 struct rps_dev_flow_table {
640 struct rps_dev_flow flows[0];
642 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
643 ((_num) * sizeof(struct rps_dev_flow)))
646 * The rps_sock_flow_table contains mappings of flows to the last CPU
647 * on which they were processed by the application (set in recvmsg).
648 * Each entry is a 32bit value. Upper part is the high order bits
649 * of flow hash, lower part is cpu number.
650 * rps_cpu_mask is used to partition the space, depending on number of
651 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
652 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
653 * meaning we use 32-6=26 bits for the hash.
655 struct rps_sock_flow_table {
658 u32 ents[0] ____cacheline_aligned_in_smp;
660 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
662 #define RPS_NO_CPU 0xffff
664 extern u32 rps_cpu_mask;
665 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
667 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
671 unsigned int index = hash & table->mask;
672 u32 val = hash & ~rps_cpu_mask;
674 /* We only give a hint, preemption can change cpu under us */
675 val |= raw_smp_processor_id();
677 if (table->ents[index] != val)
678 table->ents[index] = val;
682 #ifdef CONFIG_RFS_ACCEL
683 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
686 #endif /* CONFIG_RPS */
688 /* This structure contains an instance of an RX queue. */
689 struct netdev_rx_queue {
691 struct rps_map __rcu *rps_map;
692 struct rps_dev_flow_table __rcu *rps_flow_table;
695 struct net_device *dev;
696 } ____cacheline_aligned_in_smp;
699 * RX queue sysfs structures and functions.
701 struct rx_queue_attribute {
702 struct attribute attr;
703 ssize_t (*show)(struct netdev_rx_queue *queue,
704 struct rx_queue_attribute *attr, char *buf);
705 ssize_t (*store)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, const char *buf, size_t len);
711 * This structure holds an XPS map which can be of variable length. The
712 * map is an array of queues.
716 unsigned int alloc_len;
720 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
721 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
725 * This structure holds all XPS maps for device. Maps are indexed by CPU.
727 struct xps_dev_maps {
729 struct xps_map __rcu *cpu_map[0];
731 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
732 (nr_cpu_ids * sizeof(struct xps_map *)))
733 #endif /* CONFIG_XPS */
735 #define TC_MAX_QUEUE 16
736 #define TC_BITMASK 15
737 /* HW offloaded queuing disciplines txq count and offset maps */
738 struct netdev_tc_txq {
743 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
745 * This structure is to hold information about the device
746 * configured to run FCoE protocol stack.
748 struct netdev_fcoe_hbainfo {
749 char manufacturer[64];
750 char serial_number[64];
751 char hardware_version[64];
752 char driver_version[64];
753 char optionrom_version[64];
754 char firmware_version[64];
756 char model_description[256];
760 #define MAX_PHYS_ITEM_ID_LEN 32
762 /* This structure holds a unique identifier to identify some
763 * physical item (port for example) used by a netdevice.
765 struct netdev_phys_item_id {
766 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
767 unsigned char id_len;
770 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
771 struct sk_buff *skb);
774 * This structure defines the management hooks for network devices.
775 * The following hooks can be defined; unless noted otherwise, they are
776 * optional and can be filled with a null pointer.
778 * int (*ndo_init)(struct net_device *dev);
779 * This function is called once when network device is registered.
780 * The network device can use this to any late stage initializaton
781 * or semantic validattion. It can fail with an error code which will
782 * be propogated back to register_netdev
784 * void (*ndo_uninit)(struct net_device *dev);
785 * This function is called when device is unregistered or when registration
786 * fails. It is not called if init fails.
788 * int (*ndo_open)(struct net_device *dev);
789 * This function is called when network device transistions to the up
792 * int (*ndo_stop)(struct net_device *dev);
793 * This function is called when network device transistions to the down
796 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
797 * struct net_device *dev);
798 * Called when a packet needs to be transmitted.
799 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
800 * the queue before that can happen; it's for obsolete devices and weird
801 * corner cases, but the stack really does a non-trivial amount
802 * of useless work if you return NETDEV_TX_BUSY.
803 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
804 * Required can not be NULL.
806 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
807 * void *accel_priv, select_queue_fallback_t fallback);
808 * Called to decide which queue to when device supports multiple
811 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
812 * This function is called to allow device receiver to make
813 * changes to configuration when multicast or promiscious is enabled.
815 * void (*ndo_set_rx_mode)(struct net_device *dev);
816 * This function is called device changes address list filtering.
817 * If driver handles unicast address filtering, it should set
818 * IFF_UNICAST_FLT to its priv_flags.
820 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
821 * This function is called when the Media Access Control address
822 * needs to be changed. If this interface is not defined, the
823 * mac address can not be changed.
825 * int (*ndo_validate_addr)(struct net_device *dev);
826 * Test if Media Access Control address is valid for the device.
828 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
829 * Called when a user request an ioctl which can't be handled by
830 * the generic interface code. If not defined ioctl's return
831 * not supported error code.
833 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
834 * Used to set network devices bus interface parameters. This interface
835 * is retained for legacy reason, new devices should use the bus
836 * interface (PCI) for low level management.
838 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
839 * Called when a user wants to change the Maximum Transfer Unit
840 * of a device. If not defined, any request to change MTU will
841 * will return an error.
843 * void (*ndo_tx_timeout)(struct net_device *dev);
844 * Callback uses when the transmitter has not made any progress
845 * for dev->watchdog ticks.
847 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
848 * struct rtnl_link_stats64 *storage);
849 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
850 * Called when a user wants to get the network device usage
851 * statistics. Drivers must do one of the following:
852 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
853 * rtnl_link_stats64 structure passed by the caller.
854 * 2. Define @ndo_get_stats to update a net_device_stats structure
855 * (which should normally be dev->stats) and return a pointer to
856 * it. The structure may be changed asynchronously only if each
857 * field is written atomically.
858 * 3. Update dev->stats asynchronously and atomically, and define
861 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
862 * If device support VLAN filtering this function is called when a
863 * VLAN id is registered.
865 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
866 * If device support VLAN filtering this function is called when a
867 * VLAN id is unregistered.
869 * void (*ndo_poll_controller)(struct net_device *dev);
871 * SR-IOV management functions.
872 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
873 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
874 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
876 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
877 * int (*ndo_get_vf_config)(struct net_device *dev,
878 * int vf, struct ifla_vf_info *ivf);
879 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
880 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
881 * struct nlattr *port[]);
883 * Enable or disable the VF ability to query its RSS Redirection Table and
884 * Hash Key. This is needed since on some devices VF share this information
885 * with PF and querying it may adduce a theoretical security risk.
886 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
887 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
888 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
889 * Called to setup 'tc' number of traffic classes in the net device. This
890 * is always called from the stack with the rtnl lock held and netif tx
891 * queues stopped. This allows the netdevice to perform queue management
894 * Fiber Channel over Ethernet (FCoE) offload functions.
895 * int (*ndo_fcoe_enable)(struct net_device *dev);
896 * Called when the FCoE protocol stack wants to start using LLD for FCoE
897 * so the underlying device can perform whatever needed configuration or
898 * initialization to support acceleration of FCoE traffic.
900 * int (*ndo_fcoe_disable)(struct net_device *dev);
901 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
902 * so the underlying device can perform whatever needed clean-ups to
903 * stop supporting acceleration of FCoE traffic.
905 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
906 * struct scatterlist *sgl, unsigned int sgc);
907 * Called when the FCoE Initiator wants to initialize an I/O that
908 * is a possible candidate for Direct Data Placement (DDP). The LLD can
909 * perform necessary setup and returns 1 to indicate the device is set up
910 * successfully to perform DDP on this I/O, otherwise this returns 0.
912 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
913 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
914 * indicated by the FC exchange id 'xid', so the underlying device can
915 * clean up and reuse resources for later DDP requests.
917 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
918 * struct scatterlist *sgl, unsigned int sgc);
919 * Called when the FCoE Target wants to initialize an I/O that
920 * is a possible candidate for Direct Data Placement (DDP). The LLD can
921 * perform necessary setup and returns 1 to indicate the device is set up
922 * successfully to perform DDP on this I/O, otherwise this returns 0.
924 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
925 * struct netdev_fcoe_hbainfo *hbainfo);
926 * Called when the FCoE Protocol stack wants information on the underlying
927 * device. This information is utilized by the FCoE protocol stack to
928 * register attributes with Fiber Channel management service as per the
929 * FC-GS Fabric Device Management Information(FDMI) specification.
931 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
932 * Called when the underlying device wants to override default World Wide
933 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
934 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
935 * protocol stack to use.
938 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
939 * u16 rxq_index, u32 flow_id);
940 * Set hardware filter for RFS. rxq_index is the target queue index;
941 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
942 * Return the filter ID on success, or a negative error code.
944 * Slave management functions (for bridge, bonding, etc).
945 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
946 * Called to make another netdev an underling.
948 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
949 * Called to release previously enslaved netdev.
951 * Feature/offload setting functions.
952 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
953 * netdev_features_t features);
954 * Adjusts the requested feature flags according to device-specific
955 * constraints, and returns the resulting flags. Must not modify
958 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
959 * Called to update device configuration to new features. Passed
960 * feature set might be less than what was returned by ndo_fix_features()).
961 * Must return >0 or -errno if it changed dev->features itself.
963 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
964 * struct net_device *dev,
965 * const unsigned char *addr, u16 vid, u16 flags)
966 * Adds an FDB entry to dev for addr.
967 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
968 * struct net_device *dev,
969 * const unsigned char *addr, u16 vid)
970 * Deletes the FDB entry from dev coresponding to addr.
971 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
972 * struct net_device *dev, struct net_device *filter_dev,
974 * Used to add FDB entries to dump requests. Implementers should add
975 * entries to skb and update idx with the number of entries.
977 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
979 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
980 * struct net_device *dev, u32 filter_mask)
981 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
984 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
985 * Called to change device carrier. Soft-devices (like dummy, team, etc)
986 * which do not represent real hardware may define this to allow their
987 * userspace components to manage their virtual carrier state. Devices
988 * that determine carrier state from physical hardware properties (eg
989 * network cables) or protocol-dependent mechanisms (eg
990 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
992 * int (*ndo_get_phys_port_id)(struct net_device *dev,
993 * struct netdev_phys_item_id *ppid);
994 * Called to get ID of physical port of this device. If driver does
995 * not implement this, it is assumed that the hw is not able to have
996 * multiple net devices on single physical port.
998 * void (*ndo_add_vxlan_port)(struct net_device *dev,
999 * sa_family_t sa_family, __be16 port);
1000 * Called by vxlan to notiy a driver about the UDP port and socket
1001 * address family that vxlan is listnening to. It is called only when
1002 * a new port starts listening. The operation is protected by the
1003 * vxlan_net->sock_lock.
1005 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1006 * sa_family_t sa_family, __be16 port);
1007 * Called by vxlan to notify the driver about a UDP port and socket
1008 * address family that vxlan is not listening to anymore. The operation
1009 * is protected by the vxlan_net->sock_lock.
1011 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1012 * struct net_device *dev)
1013 * Called by upper layer devices to accelerate switching or other
1014 * station functionality into hardware. 'pdev is the lowerdev
1015 * to use for the offload and 'dev' is the net device that will
1016 * back the offload. Returns a pointer to the private structure
1017 * the upper layer will maintain.
1018 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1019 * Called by upper layer device to delete the station created
1020 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1021 * the station and priv is the structure returned by the add
1023 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1024 * struct net_device *dev,
1026 * Callback to use for xmit over the accelerated station. This
1027 * is used in place of ndo_start_xmit on accelerated net
1029 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1030 * struct net_device *dev
1031 * netdev_features_t features);
1032 * Called by core transmit path to determine if device is capable of
1033 * performing offload operations on a given packet. This is to give
1034 * the device an opportunity to implement any restrictions that cannot
1035 * be otherwise expressed by feature flags. The check is called with
1036 * the set of features that the stack has calculated and it returns
1037 * those the driver believes to be appropriate.
1038 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1039 * int queue_index, u32 maxrate);
1040 * Called when a user wants to set a max-rate limitation of specific
1042 * int (*ndo_get_iflink)(const struct net_device *dev);
1043 * Called to get the iflink value of this device.
1045 struct net_device_ops {
1046 int (*ndo_init)(struct net_device *dev);
1047 void (*ndo_uninit)(struct net_device *dev);
1048 int (*ndo_open)(struct net_device *dev);
1049 int (*ndo_stop)(struct net_device *dev);
1050 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1051 struct net_device *dev);
1052 u16 (*ndo_select_queue)(struct net_device *dev,
1053 struct sk_buff *skb,
1055 select_queue_fallback_t fallback);
1056 void (*ndo_change_rx_flags)(struct net_device *dev,
1058 void (*ndo_set_rx_mode)(struct net_device *dev);
1059 int (*ndo_set_mac_address)(struct net_device *dev,
1061 int (*ndo_validate_addr)(struct net_device *dev);
1062 int (*ndo_do_ioctl)(struct net_device *dev,
1063 struct ifreq *ifr, int cmd);
1064 int (*ndo_set_config)(struct net_device *dev,
1066 int (*ndo_change_mtu)(struct net_device *dev,
1068 int (*ndo_neigh_setup)(struct net_device *dev,
1069 struct neigh_parms *);
1070 void (*ndo_tx_timeout) (struct net_device *dev);
1072 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1073 struct rtnl_link_stats64 *storage);
1074 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1076 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1077 __be16 proto, u16 vid);
1078 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1079 __be16 proto, u16 vid);
1080 #ifdef CONFIG_NET_POLL_CONTROLLER
1081 void (*ndo_poll_controller)(struct net_device *dev);
1082 int (*ndo_netpoll_setup)(struct net_device *dev,
1083 struct netpoll_info *info);
1084 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1086 #ifdef CONFIG_NET_RX_BUSY_POLL
1087 int (*ndo_busy_poll)(struct napi_struct *dev);
1089 int (*ndo_set_vf_mac)(struct net_device *dev,
1090 int queue, u8 *mac);
1091 int (*ndo_set_vf_vlan)(struct net_device *dev,
1092 int queue, u16 vlan, u8 qos);
1093 int (*ndo_set_vf_rate)(struct net_device *dev,
1094 int vf, int min_tx_rate,
1096 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1097 int vf, bool setting);
1098 int (*ndo_get_vf_config)(struct net_device *dev,
1100 struct ifla_vf_info *ivf);
1101 int (*ndo_set_vf_link_state)(struct net_device *dev,
1102 int vf, int link_state);
1103 int (*ndo_set_vf_port)(struct net_device *dev,
1105 struct nlattr *port[]);
1106 int (*ndo_get_vf_port)(struct net_device *dev,
1107 int vf, struct sk_buff *skb);
1108 int (*ndo_set_vf_rss_query_en)(
1109 struct net_device *dev,
1110 int vf, bool setting);
1111 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1112 #if IS_ENABLED(CONFIG_FCOE)
1113 int (*ndo_fcoe_enable)(struct net_device *dev);
1114 int (*ndo_fcoe_disable)(struct net_device *dev);
1115 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1117 struct scatterlist *sgl,
1119 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1121 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1123 struct scatterlist *sgl,
1125 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1126 struct netdev_fcoe_hbainfo *hbainfo);
1129 #if IS_ENABLED(CONFIG_LIBFCOE)
1130 #define NETDEV_FCOE_WWNN 0
1131 #define NETDEV_FCOE_WWPN 1
1132 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1133 u64 *wwn, int type);
1136 #ifdef CONFIG_RFS_ACCEL
1137 int (*ndo_rx_flow_steer)(struct net_device *dev,
1138 const struct sk_buff *skb,
1142 int (*ndo_add_slave)(struct net_device *dev,
1143 struct net_device *slave_dev);
1144 int (*ndo_del_slave)(struct net_device *dev,
1145 struct net_device *slave_dev);
1146 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1147 netdev_features_t features);
1148 int (*ndo_set_features)(struct net_device *dev,
1149 netdev_features_t features);
1150 int (*ndo_neigh_construct)(struct neighbour *n);
1151 void (*ndo_neigh_destroy)(struct neighbour *n);
1153 int (*ndo_fdb_add)(struct ndmsg *ndm,
1154 struct nlattr *tb[],
1155 struct net_device *dev,
1156 const unsigned char *addr,
1159 int (*ndo_fdb_del)(struct ndmsg *ndm,
1160 struct nlattr *tb[],
1161 struct net_device *dev,
1162 const unsigned char *addr,
1164 int (*ndo_fdb_dump)(struct sk_buff *skb,
1165 struct netlink_callback *cb,
1166 struct net_device *dev,
1167 struct net_device *filter_dev,
1170 int (*ndo_bridge_setlink)(struct net_device *dev,
1171 struct nlmsghdr *nlh,
1173 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1175 struct net_device *dev,
1177 int (*ndo_bridge_dellink)(struct net_device *dev,
1178 struct nlmsghdr *nlh,
1180 int (*ndo_change_carrier)(struct net_device *dev,
1182 int (*ndo_get_phys_port_id)(struct net_device *dev,
1183 struct netdev_phys_item_id *ppid);
1184 int (*ndo_get_phys_port_name)(struct net_device *dev,
1185 char *name, size_t len);
1186 void (*ndo_add_vxlan_port)(struct net_device *dev,
1187 sa_family_t sa_family,
1189 void (*ndo_del_vxlan_port)(struct net_device *dev,
1190 sa_family_t sa_family,
1193 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1194 struct net_device *dev);
1195 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1198 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1199 struct net_device *dev,
1201 int (*ndo_get_lock_subclass)(struct net_device *dev);
1202 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1203 struct net_device *dev,
1204 netdev_features_t features);
1205 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1208 int (*ndo_get_iflink)(const struct net_device *dev);
1212 * enum net_device_priv_flags - &struct net_device priv_flags
1214 * These are the &struct net_device, they are only set internally
1215 * by drivers and used in the kernel. These flags are invisible to
1216 * userspace, this means that the order of these flags can change
1217 * during any kernel release.
1219 * You should have a pretty good reason to be extending these flags.
1221 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1222 * @IFF_EBRIDGE: Ethernet bridging device
1223 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1224 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1225 * @IFF_MASTER_ALB: bonding master, balance-alb
1226 * @IFF_BONDING: bonding master or slave
1227 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1228 * @IFF_ISATAP: ISATAP interface (RFC4214)
1229 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1230 * @IFF_WAN_HDLC: WAN HDLC device
1231 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1233 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1234 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1235 * @IFF_MACVLAN_PORT: device used as macvlan port
1236 * @IFF_BRIDGE_PORT: device used as bridge port
1237 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1238 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1239 * @IFF_UNICAST_FLT: Supports unicast filtering
1240 * @IFF_TEAM_PORT: device used as team port
1241 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1242 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1243 * change when it's running
1244 * @IFF_MACVLAN: Macvlan device
1246 enum netdev_priv_flags {
1247 IFF_802_1Q_VLAN = 1<<0,
1249 IFF_SLAVE_INACTIVE = 1<<2,
1250 IFF_MASTER_8023AD = 1<<3,
1251 IFF_MASTER_ALB = 1<<4,
1253 IFF_SLAVE_NEEDARP = 1<<6,
1255 IFF_MASTER_ARPMON = 1<<8,
1256 IFF_WAN_HDLC = 1<<9,
1257 IFF_XMIT_DST_RELEASE = 1<<10,
1258 IFF_DONT_BRIDGE = 1<<11,
1259 IFF_DISABLE_NETPOLL = 1<<12,
1260 IFF_MACVLAN_PORT = 1<<13,
1261 IFF_BRIDGE_PORT = 1<<14,
1262 IFF_OVS_DATAPATH = 1<<15,
1263 IFF_TX_SKB_SHARING = 1<<16,
1264 IFF_UNICAST_FLT = 1<<17,
1265 IFF_TEAM_PORT = 1<<18,
1266 IFF_SUPP_NOFCS = 1<<19,
1267 IFF_LIVE_ADDR_CHANGE = 1<<20,
1268 IFF_MACVLAN = 1<<21,
1269 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1270 IFF_IPVLAN_MASTER = 1<<23,
1271 IFF_IPVLAN_SLAVE = 1<<24,
1274 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1275 #define IFF_EBRIDGE IFF_EBRIDGE
1276 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1277 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1278 #define IFF_MASTER_ALB IFF_MASTER_ALB
1279 #define IFF_BONDING IFF_BONDING
1280 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1281 #define IFF_ISATAP IFF_ISATAP
1282 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1283 #define IFF_WAN_HDLC IFF_WAN_HDLC
1284 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1285 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1286 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1287 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1288 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1289 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1290 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1291 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1292 #define IFF_TEAM_PORT IFF_TEAM_PORT
1293 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1294 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1295 #define IFF_MACVLAN IFF_MACVLAN
1296 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1297 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1298 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1301 * struct net_device - The DEVICE structure.
1302 * Actually, this whole structure is a big mistake. It mixes I/O
1303 * data with strictly "high-level" data, and it has to know about
1304 * almost every data structure used in the INET module.
1306 * @name: This is the first field of the "visible" part of this structure
1307 * (i.e. as seen by users in the "Space.c" file). It is the name
1310 * @name_hlist: Device name hash chain, please keep it close to name[]
1311 * @ifalias: SNMP alias
1312 * @mem_end: Shared memory end
1313 * @mem_start: Shared memory start
1314 * @base_addr: Device I/O address
1315 * @irq: Device IRQ number
1317 * @carrier_changes: Stats to monitor carrier on<->off transitions
1319 * @state: Generic network queuing layer state, see netdev_state_t
1320 * @dev_list: The global list of network devices
1321 * @napi_list: List entry, that is used for polling napi devices
1322 * @unreg_list: List entry, that is used, when we are unregistering the
1323 * device, see the function unregister_netdev
1324 * @close_list: List entry, that is used, when we are closing the device
1326 * @adj_list: Directly linked devices, like slaves for bonding
1327 * @all_adj_list: All linked devices, *including* neighbours
1328 * @features: Currently active device features
1329 * @hw_features: User-changeable features
1331 * @wanted_features: User-requested features
1332 * @vlan_features: Mask of features inheritable by VLAN devices
1334 * @hw_enc_features: Mask of features inherited by encapsulating devices
1335 * This field indicates what encapsulation
1336 * offloads the hardware is capable of doing,
1337 * and drivers will need to set them appropriately.
1339 * @mpls_features: Mask of features inheritable by MPLS
1341 * @ifindex: interface index
1342 * @group: The group, that the device belongs to
1344 * @stats: Statistics struct, which was left as a legacy, use
1345 * rtnl_link_stats64 instead
1347 * @rx_dropped: Dropped packets by core network,
1348 * do not use this in drivers
1349 * @tx_dropped: Dropped packets by core network,
1350 * do not use this in drivers
1352 * @wireless_handlers: List of functions to handle Wireless Extensions,
1354 * see <net/iw_handler.h> for details.
1355 * @wireless_data: Instance data managed by the core of wireless extensions
1357 * @netdev_ops: Includes several pointers to callbacks,
1358 * if one wants to override the ndo_*() functions
1359 * @ethtool_ops: Management operations
1360 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1361 * of Layer 2 headers.
1363 * @flags: Interface flags (a la BSD)
1364 * @priv_flags: Like 'flags' but invisible to userspace,
1365 * see if.h for the definitions
1366 * @gflags: Global flags ( kept as legacy )
1367 * @padded: How much padding added by alloc_netdev()
1368 * @operstate: RFC2863 operstate
1369 * @link_mode: Mapping policy to operstate
1370 * @if_port: Selectable AUI, TP, ...
1372 * @mtu: Interface MTU value
1373 * @type: Interface hardware type
1374 * @hard_header_len: Hardware header length
1376 * @needed_headroom: Extra headroom the hardware may need, but not in all
1377 * cases can this be guaranteed
1378 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1379 * cases can this be guaranteed. Some cases also use
1380 * LL_MAX_HEADER instead to allocate the skb
1382 * interface address info:
1384 * @perm_addr: Permanent hw address
1385 * @addr_assign_type: Hw address assignment type
1386 * @addr_len: Hardware address length
1387 * @neigh_priv_len; Used in neigh_alloc(),
1388 * initialized only in atm/clip.c
1389 * @dev_id: Used to differentiate devices that share
1390 * the same link layer address
1391 * @dev_port: Used to differentiate devices that share
1393 * @addr_list_lock: XXX: need comments on this one
1394 * @uc_promisc: Counter, that indicates, that promiscuous mode
1395 * has been enabled due to the need to listen to
1396 * additional unicast addresses in a device that
1397 * does not implement ndo_set_rx_mode()
1398 * @uc: unicast mac addresses
1399 * @mc: multicast mac addresses
1400 * @dev_addrs: list of device hw addresses
1401 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1402 * @promiscuity: Number of times, the NIC is told to work in
1403 * Promiscuous mode, if it becomes 0 the NIC will
1404 * exit from working in Promiscuous mode
1405 * @allmulti: Counter, enables or disables allmulticast mode
1407 * @vlan_info: VLAN info
1408 * @dsa_ptr: dsa specific data
1409 * @tipc_ptr: TIPC specific data
1410 * @atalk_ptr: AppleTalk link
1411 * @ip_ptr: IPv4 specific data
1412 * @dn_ptr: DECnet specific data
1413 * @ip6_ptr: IPv6 specific data
1414 * @ax25_ptr: AX.25 specific data
1415 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1417 * @last_rx: Time of last Rx
1418 * @dev_addr: Hw address (before bcast,
1419 * because most packets are unicast)
1421 * @_rx: Array of RX queues
1422 * @num_rx_queues: Number of RX queues
1423 * allocated at register_netdev() time
1424 * @real_num_rx_queues: Number of RX queues currently active in device
1426 * @rx_handler: handler for received packets
1427 * @rx_handler_data: XXX: need comments on this one
1428 * @ingress_queue: XXX: need comments on this one
1429 * @broadcast: hw bcast address
1431 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1432 * indexed by RX queue number. Assigned by driver.
1433 * This must only be set if the ndo_rx_flow_steer
1434 * operation is defined
1435 * @index_hlist: Device index hash chain
1437 * @_tx: Array of TX queues
1438 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1439 * @real_num_tx_queues: Number of TX queues currently active in device
1440 * @qdisc: Root qdisc from userspace point of view
1441 * @tx_queue_len: Max frames per queue allowed
1442 * @tx_global_lock: XXX: need comments on this one
1444 * @xps_maps: XXX: need comments on this one
1446 * @trans_start: Time (in jiffies) of last Tx
1447 * @watchdog_timeo: Represents the timeout that is used by
1448 * the watchdog ( see dev_watchdog() )
1449 * @watchdog_timer: List of timers
1451 * @pcpu_refcnt: Number of references to this device
1452 * @todo_list: Delayed register/unregister
1453 * @link_watch_list: XXX: need comments on this one
1455 * @reg_state: Register/unregister state machine
1456 * @dismantle: Device is going to be freed
1457 * @rtnl_link_state: This enum represents the phases of creating
1460 * @destructor: Called from unregister,
1461 * can be used to call free_netdev
1462 * @npinfo: XXX: need comments on this one
1463 * @nd_net: Network namespace this network device is inside
1465 * @ml_priv: Mid-layer private
1466 * @lstats: Loopback statistics
1467 * @tstats: Tunnel statistics
1468 * @dstats: Dummy statistics
1469 * @vstats: Virtual ethernet statistics
1474 * @dev: Class/net/name entry
1475 * @sysfs_groups: Space for optional device, statistics and wireless
1478 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1479 * @rtnl_link_ops: Rtnl_link_ops
1481 * @gso_max_size: Maximum size of generic segmentation offload
1482 * @gso_max_segs: Maximum number of segments that can be passed to the
1484 * @gso_min_segs: Minimum number of segments that can be passed to the
1487 * @dcbnl_ops: Data Center Bridging netlink ops
1488 * @num_tc: Number of traffic classes in the net device
1489 * @tc_to_txq: XXX: need comments on this one
1490 * @prio_tc_map XXX: need comments on this one
1492 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1494 * @priomap: XXX: need comments on this one
1495 * @phydev: Physical device may attach itself
1496 * for hardware timestamping
1498 * @qdisc_tx_busylock: XXX: need comments on this one
1500 * @pm_qos_req: Power Management QoS object
1502 * FIXME: cleanup struct net_device such that network protocol info
1507 char name[IFNAMSIZ];
1508 struct hlist_node name_hlist;
1511 * I/O specific fields
1512 * FIXME: Merge these and struct ifmap into one
1514 unsigned long mem_end;
1515 unsigned long mem_start;
1516 unsigned long base_addr;
1519 atomic_t carrier_changes;
1522 * Some hardware also needs these fields (state,dev_list,
1523 * napi_list,unreg_list,close_list) but they are not
1524 * part of the usual set specified in Space.c.
1527 unsigned long state;
1529 struct list_head dev_list;
1530 struct list_head napi_list;
1531 struct list_head unreg_list;
1532 struct list_head close_list;
1533 struct list_head ptype_all;
1534 struct list_head ptype_specific;
1537 struct list_head upper;
1538 struct list_head lower;
1542 struct list_head upper;
1543 struct list_head lower;
1546 netdev_features_t features;
1547 netdev_features_t hw_features;
1548 netdev_features_t wanted_features;
1549 netdev_features_t vlan_features;
1550 netdev_features_t hw_enc_features;
1551 netdev_features_t mpls_features;
1556 struct net_device_stats stats;
1558 atomic_long_t rx_dropped;
1559 atomic_long_t tx_dropped;
1561 #ifdef CONFIG_WIRELESS_EXT
1562 const struct iw_handler_def * wireless_handlers;
1563 struct iw_public_data * wireless_data;
1565 const struct net_device_ops *netdev_ops;
1566 const struct ethtool_ops *ethtool_ops;
1567 #ifdef CONFIG_NET_SWITCHDEV
1568 const struct swdev_ops *swdev_ops;
1571 const struct header_ops *header_ops;
1574 unsigned int priv_flags;
1576 unsigned short gflags;
1577 unsigned short padded;
1579 unsigned char operstate;
1580 unsigned char link_mode;
1582 unsigned char if_port;
1586 unsigned short type;
1587 unsigned short hard_header_len;
1589 unsigned short needed_headroom;
1590 unsigned short needed_tailroom;
1592 /* Interface address info. */
1593 unsigned char perm_addr[MAX_ADDR_LEN];
1594 unsigned char addr_assign_type;
1595 unsigned char addr_len;
1596 unsigned short neigh_priv_len;
1597 unsigned short dev_id;
1598 unsigned short dev_port;
1599 spinlock_t addr_list_lock;
1600 unsigned char name_assign_type;
1602 struct netdev_hw_addr_list uc;
1603 struct netdev_hw_addr_list mc;
1604 struct netdev_hw_addr_list dev_addrs;
1607 struct kset *queues_kset;
1609 unsigned int promiscuity;
1610 unsigned int allmulti;
1613 /* Protocol specific pointers */
1615 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1616 struct vlan_info __rcu *vlan_info;
1618 #if IS_ENABLED(CONFIG_NET_DSA)
1619 struct dsa_switch_tree *dsa_ptr;
1621 #if IS_ENABLED(CONFIG_TIPC)
1622 struct tipc_bearer __rcu *tipc_ptr;
1625 struct in_device __rcu *ip_ptr;
1626 struct dn_dev __rcu *dn_ptr;
1627 struct inet6_dev __rcu *ip6_ptr;
1629 struct wireless_dev *ieee80211_ptr;
1630 struct wpan_dev *ieee802154_ptr;
1631 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1632 struct mpls_dev __rcu *mpls_ptr;
1636 * Cache lines mostly used on receive path (including eth_type_trans())
1638 unsigned long last_rx;
1640 /* Interface address info used in eth_type_trans() */
1641 unsigned char *dev_addr;
1645 struct netdev_rx_queue *_rx;
1647 unsigned int num_rx_queues;
1648 unsigned int real_num_rx_queues;
1652 unsigned long gro_flush_timeout;
1653 rx_handler_func_t __rcu *rx_handler;
1654 void __rcu *rx_handler_data;
1656 struct netdev_queue __rcu *ingress_queue;
1657 unsigned char broadcast[MAX_ADDR_LEN];
1658 #ifdef CONFIG_RFS_ACCEL
1659 struct cpu_rmap *rx_cpu_rmap;
1661 struct hlist_node index_hlist;
1664 * Cache lines mostly used on transmit path
1666 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1667 unsigned int num_tx_queues;
1668 unsigned int real_num_tx_queues;
1669 struct Qdisc *qdisc;
1670 unsigned long tx_queue_len;
1671 spinlock_t tx_global_lock;
1675 struct xps_dev_maps __rcu *xps_maps;
1678 /* These may be needed for future network-power-down code. */
1681 * trans_start here is expensive for high speed devices on SMP,
1682 * please use netdev_queue->trans_start instead.
1684 unsigned long trans_start;
1686 struct timer_list watchdog_timer;
1688 int __percpu *pcpu_refcnt;
1689 struct list_head todo_list;
1691 struct list_head link_watch_list;
1693 enum { NETREG_UNINITIALIZED=0,
1694 NETREG_REGISTERED, /* completed register_netdevice */
1695 NETREG_UNREGISTERING, /* called unregister_netdevice */
1696 NETREG_UNREGISTERED, /* completed unregister todo */
1697 NETREG_RELEASED, /* called free_netdev */
1698 NETREG_DUMMY, /* dummy device for NAPI poll */
1704 RTNL_LINK_INITIALIZED,
1705 RTNL_LINK_INITIALIZING,
1706 } rtnl_link_state:16;
1708 void (*destructor)(struct net_device *dev);
1710 #ifdef CONFIG_NETPOLL
1711 struct netpoll_info __rcu *npinfo;
1714 possible_net_t nd_net;
1716 /* mid-layer private */
1719 struct pcpu_lstats __percpu *lstats;
1720 struct pcpu_sw_netstats __percpu *tstats;
1721 struct pcpu_dstats __percpu *dstats;
1722 struct pcpu_vstats __percpu *vstats;
1725 struct garp_port __rcu *garp_port;
1726 struct mrp_port __rcu *mrp_port;
1729 const struct attribute_group *sysfs_groups[4];
1730 const struct attribute_group *sysfs_rx_queue_group;
1732 const struct rtnl_link_ops *rtnl_link_ops;
1734 /* for setting kernel sock attribute on TCP connection setup */
1735 #define GSO_MAX_SIZE 65536
1736 unsigned int gso_max_size;
1737 #define GSO_MAX_SEGS 65535
1741 const struct dcbnl_rtnl_ops *dcbnl_ops;
1744 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1745 u8 prio_tc_map[TC_BITMASK + 1];
1747 #if IS_ENABLED(CONFIG_FCOE)
1748 unsigned int fcoe_ddp_xid;
1750 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1751 struct netprio_map __rcu *priomap;
1753 struct phy_device *phydev;
1754 struct lock_class_key *qdisc_tx_busylock;
1756 #define to_net_dev(d) container_of(d, struct net_device, dev)
1758 #define NETDEV_ALIGN 32
1761 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1763 return dev->prio_tc_map[prio & TC_BITMASK];
1767 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1769 if (tc >= dev->num_tc)
1772 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1777 void netdev_reset_tc(struct net_device *dev)
1780 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1781 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1785 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1787 if (tc >= dev->num_tc)
1790 dev->tc_to_txq[tc].count = count;
1791 dev->tc_to_txq[tc].offset = offset;
1796 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1798 if (num_tc > TC_MAX_QUEUE)
1801 dev->num_tc = num_tc;
1806 int netdev_get_num_tc(struct net_device *dev)
1812 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1815 return &dev->_tx[index];
1818 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1819 const struct sk_buff *skb)
1821 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1824 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1825 void (*f)(struct net_device *,
1826 struct netdev_queue *,
1832 for (i = 0; i < dev->num_tx_queues; i++)
1833 f(dev, &dev->_tx[i], arg);
1836 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1837 struct sk_buff *skb,
1841 * Net namespace inlines
1844 struct net *dev_net(const struct net_device *dev)
1846 return read_pnet(&dev->nd_net);
1850 void dev_net_set(struct net_device *dev, struct net *net)
1852 write_pnet(&dev->nd_net, net);
1855 static inline bool netdev_uses_dsa(struct net_device *dev)
1857 #if IS_ENABLED(CONFIG_NET_DSA)
1858 if (dev->dsa_ptr != NULL)
1859 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1865 * netdev_priv - access network device private data
1866 * @dev: network device
1868 * Get network device private data
1870 static inline void *netdev_priv(const struct net_device *dev)
1872 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1875 /* Set the sysfs physical device reference for the network logical device
1876 * if set prior to registration will cause a symlink during initialization.
1878 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1880 /* Set the sysfs device type for the network logical device to allow
1881 * fine-grained identification of different network device types. For
1882 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1884 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1886 /* Default NAPI poll() weight
1887 * Device drivers are strongly advised to not use bigger value
1889 #define NAPI_POLL_WEIGHT 64
1892 * netif_napi_add - initialize a napi context
1893 * @dev: network device
1894 * @napi: napi context
1895 * @poll: polling function
1896 * @weight: default weight
1898 * netif_napi_add() must be used to initialize a napi context prior to calling
1899 * *any* of the other napi related functions.
1901 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1902 int (*poll)(struct napi_struct *, int), int weight);
1905 * netif_napi_del - remove a napi context
1906 * @napi: napi context
1908 * netif_napi_del() removes a napi context from the network device napi list
1910 void netif_napi_del(struct napi_struct *napi);
1912 struct napi_gro_cb {
1913 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1916 /* Length of frag0. */
1917 unsigned int frag0_len;
1919 /* This indicates where we are processing relative to skb->data. */
1922 /* This is non-zero if the packet cannot be merged with the new skb. */
1925 /* Save the IP ID here and check when we get to the transport layer */
1928 /* Number of segments aggregated. */
1931 /* Start offset for remote checksum offload */
1932 u16 gro_remcsum_start;
1934 /* jiffies when first packet was created/queued */
1937 /* Used in ipv6_gro_receive() and foo-over-udp */
1940 /* This is non-zero if the packet may be of the same flow. */
1943 /* Used in udp_gro_receive */
1946 /* GRO checksum is valid */
1949 /* Number of checksums via CHECKSUM_UNNECESSARY */
1954 #define NAPI_GRO_FREE 1
1955 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1957 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1962 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1965 /* used in skb_gro_receive() slow path */
1966 struct sk_buff *last;
1969 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1971 struct packet_type {
1972 __be16 type; /* This is really htons(ether_type). */
1973 struct net_device *dev; /* NULL is wildcarded here */
1974 int (*func) (struct sk_buff *,
1975 struct net_device *,
1976 struct packet_type *,
1977 struct net_device *);
1978 bool (*id_match)(struct packet_type *ptype,
1980 void *af_packet_priv;
1981 struct list_head list;
1984 struct offload_callbacks {
1985 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1986 netdev_features_t features);
1987 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1988 struct sk_buff *skb);
1989 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1992 struct packet_offload {
1993 __be16 type; /* This is really htons(ether_type). */
1994 struct offload_callbacks callbacks;
1995 struct list_head list;
2000 struct udp_offload_callbacks {
2001 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2002 struct sk_buff *skb,
2003 struct udp_offload *uoff);
2004 int (*gro_complete)(struct sk_buff *skb,
2006 struct udp_offload *uoff);
2009 struct udp_offload {
2012 struct udp_offload_callbacks callbacks;
2015 /* often modified stats are per cpu, other are shared (netdev->stats) */
2016 struct pcpu_sw_netstats {
2021 struct u64_stats_sync syncp;
2024 #define netdev_alloc_pcpu_stats(type) \
2026 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2029 for_each_possible_cpu(__cpu) { \
2030 typeof(type) *stat; \
2031 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2032 u64_stats_init(&stat->syncp); \
2038 #include <linux/notifier.h>
2040 /* netdevice notifier chain. Please remember to update the rtnetlink
2041 * notification exclusion list in rtnetlink_event() when adding new
2044 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2045 #define NETDEV_DOWN 0x0002
2046 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2047 detected a hardware crash and restarted
2048 - we can use this eg to kick tcp sessions
2050 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2051 #define NETDEV_REGISTER 0x0005
2052 #define NETDEV_UNREGISTER 0x0006
2053 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2054 #define NETDEV_CHANGEADDR 0x0008
2055 #define NETDEV_GOING_DOWN 0x0009
2056 #define NETDEV_CHANGENAME 0x000A
2057 #define NETDEV_FEAT_CHANGE 0x000B
2058 #define NETDEV_BONDING_FAILOVER 0x000C
2059 #define NETDEV_PRE_UP 0x000D
2060 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2061 #define NETDEV_POST_TYPE_CHANGE 0x000F
2062 #define NETDEV_POST_INIT 0x0010
2063 #define NETDEV_UNREGISTER_FINAL 0x0011
2064 #define NETDEV_RELEASE 0x0012
2065 #define NETDEV_NOTIFY_PEERS 0x0013
2066 #define NETDEV_JOIN 0x0014
2067 #define NETDEV_CHANGEUPPER 0x0015
2068 #define NETDEV_RESEND_IGMP 0x0016
2069 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2070 #define NETDEV_CHANGEINFODATA 0x0018
2071 #define NETDEV_BONDING_INFO 0x0019
2073 int register_netdevice_notifier(struct notifier_block *nb);
2074 int unregister_netdevice_notifier(struct notifier_block *nb);
2076 struct netdev_notifier_info {
2077 struct net_device *dev;
2080 struct netdev_notifier_change_info {
2081 struct netdev_notifier_info info; /* must be first */
2082 unsigned int flags_changed;
2085 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2086 struct net_device *dev)
2091 static inline struct net_device *
2092 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2097 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2100 extern rwlock_t dev_base_lock; /* Device list lock */
2102 #define for_each_netdev(net, d) \
2103 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2104 #define for_each_netdev_reverse(net, d) \
2105 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2106 #define for_each_netdev_rcu(net, d) \
2107 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2108 #define for_each_netdev_safe(net, d, n) \
2109 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2110 #define for_each_netdev_continue(net, d) \
2111 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2112 #define for_each_netdev_continue_rcu(net, d) \
2113 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2114 #define for_each_netdev_in_bond_rcu(bond, slave) \
2115 for_each_netdev_rcu(&init_net, slave) \
2116 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2117 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2119 static inline struct net_device *next_net_device(struct net_device *dev)
2121 struct list_head *lh;
2125 lh = dev->dev_list.next;
2126 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2129 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2131 struct list_head *lh;
2135 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2136 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2139 static inline struct net_device *first_net_device(struct net *net)
2141 return list_empty(&net->dev_base_head) ? NULL :
2142 net_device_entry(net->dev_base_head.next);
2145 static inline struct net_device *first_net_device_rcu(struct net *net)
2147 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2149 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2152 int netdev_boot_setup_check(struct net_device *dev);
2153 unsigned long netdev_boot_base(const char *prefix, int unit);
2154 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2155 const char *hwaddr);
2156 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2157 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2158 void dev_add_pack(struct packet_type *pt);
2159 void dev_remove_pack(struct packet_type *pt);
2160 void __dev_remove_pack(struct packet_type *pt);
2161 void dev_add_offload(struct packet_offload *po);
2162 void dev_remove_offload(struct packet_offload *po);
2164 int dev_get_iflink(const struct net_device *dev);
2165 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2166 unsigned short mask);
2167 struct net_device *dev_get_by_name(struct net *net, const char *name);
2168 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2169 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2170 int dev_alloc_name(struct net_device *dev, const char *name);
2171 int dev_open(struct net_device *dev);
2172 int dev_close(struct net_device *dev);
2173 int dev_close_many(struct list_head *head, bool unlink);
2174 void dev_disable_lro(struct net_device *dev);
2175 int dev_loopback_xmit(struct sock *sk, struct sk_buff *newskb);
2176 int dev_queue_xmit_sk(struct sock *sk, struct sk_buff *skb);
2177 static inline int dev_queue_xmit(struct sk_buff *skb)
2179 return dev_queue_xmit_sk(skb->sk, skb);
2181 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2182 int register_netdevice(struct net_device *dev);
2183 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2184 void unregister_netdevice_many(struct list_head *head);
2185 static inline void unregister_netdevice(struct net_device *dev)
2187 unregister_netdevice_queue(dev, NULL);
2190 int netdev_refcnt_read(const struct net_device *dev);
2191 void free_netdev(struct net_device *dev);
2192 void netdev_freemem(struct net_device *dev);
2193 void synchronize_net(void);
2194 int init_dummy_netdev(struct net_device *dev);
2196 DECLARE_PER_CPU(int, xmit_recursion);
2197 static inline int dev_recursion_level(void)
2199 return this_cpu_read(xmit_recursion);
2202 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2203 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2204 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2205 int netdev_get_name(struct net *net, char *name, int ifindex);
2206 int dev_restart(struct net_device *dev);
2207 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2209 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2211 return NAPI_GRO_CB(skb)->data_offset;
2214 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2216 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2219 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2221 NAPI_GRO_CB(skb)->data_offset += len;
2224 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2225 unsigned int offset)
2227 return NAPI_GRO_CB(skb)->frag0 + offset;
2230 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2232 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2235 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2236 unsigned int offset)
2238 if (!pskb_may_pull(skb, hlen))
2241 NAPI_GRO_CB(skb)->frag0 = NULL;
2242 NAPI_GRO_CB(skb)->frag0_len = 0;
2243 return skb->data + offset;
2246 static inline void *skb_gro_network_header(struct sk_buff *skb)
2248 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2249 skb_network_offset(skb);
2252 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2253 const void *start, unsigned int len)
2255 if (NAPI_GRO_CB(skb)->csum_valid)
2256 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2257 csum_partial(start, len, 0));
2260 /* GRO checksum functions. These are logical equivalents of the normal
2261 * checksum functions (in skbuff.h) except that they operate on the GRO
2262 * offsets and fields in sk_buff.
2265 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2267 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2269 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2270 skb_gro_offset(skb));
2273 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2277 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2278 skb_checksum_start_offset(skb) <
2279 skb_gro_offset(skb)) &&
2280 !skb_at_gro_remcsum_start(skb) &&
2281 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2282 (!zero_okay || check));
2285 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2288 if (NAPI_GRO_CB(skb)->csum_valid &&
2289 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2292 NAPI_GRO_CB(skb)->csum = psum;
2294 return __skb_gro_checksum_complete(skb);
2297 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2299 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2300 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2301 NAPI_GRO_CB(skb)->csum_cnt--;
2303 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2304 * verified a new top level checksum or an encapsulated one
2305 * during GRO. This saves work if we fallback to normal path.
2307 __skb_incr_checksum_unnecessary(skb);
2311 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2314 __sum16 __ret = 0; \
2315 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2316 __ret = __skb_gro_checksum_validate_complete(skb, \
2317 compute_pseudo(skb, proto)); \
2319 __skb_mark_checksum_bad(skb); \
2321 skb_gro_incr_csum_unnecessary(skb); \
2325 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2326 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2328 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2330 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2332 #define skb_gro_checksum_simple_validate(skb) \
2333 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2335 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2337 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2338 !NAPI_GRO_CB(skb)->csum_valid);
2341 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2342 __sum16 check, __wsum pseudo)
2344 NAPI_GRO_CB(skb)->csum = ~pseudo;
2345 NAPI_GRO_CB(skb)->csum_valid = 1;
2348 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2350 if (__skb_gro_checksum_convert_check(skb)) \
2351 __skb_gro_checksum_convert(skb, check, \
2352 compute_pseudo(skb, proto)); \
2355 struct gro_remcsum {
2360 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2366 static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2367 int start, int offset,
2368 struct gro_remcsum *grc,
2373 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2376 NAPI_GRO_CB(skb)->gro_remcsum_start =
2377 ((unsigned char *)ptr + start) - skb->head;
2381 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2383 /* Adjust skb->csum since we changed the packet */
2384 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2386 grc->offset = (ptr + offset) - (void *)skb->head;
2390 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2391 struct gro_remcsum *grc)
2396 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2399 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2400 unsigned short type,
2401 const void *daddr, const void *saddr,
2404 if (!dev->header_ops || !dev->header_ops->create)
2407 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2410 static inline int dev_parse_header(const struct sk_buff *skb,
2411 unsigned char *haddr)
2413 const struct net_device *dev = skb->dev;
2415 if (!dev->header_ops || !dev->header_ops->parse)
2417 return dev->header_ops->parse(skb, haddr);
2420 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2421 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2422 static inline int unregister_gifconf(unsigned int family)
2424 return register_gifconf(family, NULL);
2427 #ifdef CONFIG_NET_FLOW_LIMIT
2428 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2429 struct sd_flow_limit {
2431 unsigned int num_buckets;
2432 unsigned int history_head;
2433 u16 history[FLOW_LIMIT_HISTORY];
2437 extern int netdev_flow_limit_table_len;
2438 #endif /* CONFIG_NET_FLOW_LIMIT */
2441 * Incoming packets are placed on per-cpu queues
2443 struct softnet_data {
2444 struct list_head poll_list;
2445 struct sk_buff_head process_queue;
2448 unsigned int processed;
2449 unsigned int time_squeeze;
2450 unsigned int cpu_collision;
2451 unsigned int received_rps;
2453 struct softnet_data *rps_ipi_list;
2455 #ifdef CONFIG_NET_FLOW_LIMIT
2456 struct sd_flow_limit __rcu *flow_limit;
2458 struct Qdisc *output_queue;
2459 struct Qdisc **output_queue_tailp;
2460 struct sk_buff *completion_queue;
2463 /* Elements below can be accessed between CPUs for RPS */
2464 struct call_single_data csd ____cacheline_aligned_in_smp;
2465 struct softnet_data *rps_ipi_next;
2467 unsigned int input_queue_head;
2468 unsigned int input_queue_tail;
2470 unsigned int dropped;
2471 struct sk_buff_head input_pkt_queue;
2472 struct napi_struct backlog;
2476 static inline void input_queue_head_incr(struct softnet_data *sd)
2479 sd->input_queue_head++;
2483 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2484 unsigned int *qtail)
2487 *qtail = ++sd->input_queue_tail;
2491 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2493 void __netif_schedule(struct Qdisc *q);
2494 void netif_schedule_queue(struct netdev_queue *txq);
2496 static inline void netif_tx_schedule_all(struct net_device *dev)
2500 for (i = 0; i < dev->num_tx_queues; i++)
2501 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2504 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2506 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2510 * netif_start_queue - allow transmit
2511 * @dev: network device
2513 * Allow upper layers to call the device hard_start_xmit routine.
2515 static inline void netif_start_queue(struct net_device *dev)
2517 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2520 static inline void netif_tx_start_all_queues(struct net_device *dev)
2524 for (i = 0; i < dev->num_tx_queues; i++) {
2525 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2526 netif_tx_start_queue(txq);
2530 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2533 * netif_wake_queue - restart transmit
2534 * @dev: network device
2536 * Allow upper layers to call the device hard_start_xmit routine.
2537 * Used for flow control when transmit resources are available.
2539 static inline void netif_wake_queue(struct net_device *dev)
2541 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2544 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2548 for (i = 0; i < dev->num_tx_queues; i++) {
2549 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2550 netif_tx_wake_queue(txq);
2554 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2556 if (WARN_ON(!dev_queue)) {
2557 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2560 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2564 * netif_stop_queue - stop transmitted packets
2565 * @dev: network device
2567 * Stop upper layers calling the device hard_start_xmit routine.
2568 * Used for flow control when transmit resources are unavailable.
2570 static inline void netif_stop_queue(struct net_device *dev)
2572 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2575 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2579 for (i = 0; i < dev->num_tx_queues; i++) {
2580 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2581 netif_tx_stop_queue(txq);
2585 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2587 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2591 * netif_queue_stopped - test if transmit queue is flowblocked
2592 * @dev: network device
2594 * Test if transmit queue on device is currently unable to send.
2596 static inline bool netif_queue_stopped(const struct net_device *dev)
2598 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2601 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2603 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2607 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2609 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2613 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2615 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2619 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2620 * @dev_queue: pointer to transmit queue
2622 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2623 * to give appropriate hint to the cpu.
2625 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2628 prefetchw(&dev_queue->dql.num_queued);
2633 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2634 * @dev_queue: pointer to transmit queue
2636 * BQL enabled drivers might use this helper in their TX completion path,
2637 * to give appropriate hint to the cpu.
2639 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2642 prefetchw(&dev_queue->dql.limit);
2646 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2650 dql_queued(&dev_queue->dql, bytes);
2652 if (likely(dql_avail(&dev_queue->dql) >= 0))
2655 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2658 * The XOFF flag must be set before checking the dql_avail below,
2659 * because in netdev_tx_completed_queue we update the dql_completed
2660 * before checking the XOFF flag.
2664 /* check again in case another CPU has just made room avail */
2665 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2666 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2671 * netdev_sent_queue - report the number of bytes queued to hardware
2672 * @dev: network device
2673 * @bytes: number of bytes queued to the hardware device queue
2675 * Report the number of bytes queued for sending/completion to the network
2676 * device hardware queue. @bytes should be a good approximation and should
2677 * exactly match netdev_completed_queue() @bytes
2679 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2681 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2684 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2685 unsigned int pkts, unsigned int bytes)
2688 if (unlikely(!bytes))
2691 dql_completed(&dev_queue->dql, bytes);
2694 * Without the memory barrier there is a small possiblity that
2695 * netdev_tx_sent_queue will miss the update and cause the queue to
2696 * be stopped forever
2700 if (dql_avail(&dev_queue->dql) < 0)
2703 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2704 netif_schedule_queue(dev_queue);
2709 * netdev_completed_queue - report bytes and packets completed by device
2710 * @dev: network device
2711 * @pkts: actual number of packets sent over the medium
2712 * @bytes: actual number of bytes sent over the medium
2714 * Report the number of bytes and packets transmitted by the network device
2715 * hardware queue over the physical medium, @bytes must exactly match the
2716 * @bytes amount passed to netdev_sent_queue()
2718 static inline void netdev_completed_queue(struct net_device *dev,
2719 unsigned int pkts, unsigned int bytes)
2721 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2724 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2727 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2733 * netdev_reset_queue - reset the packets and bytes count of a network device
2734 * @dev_queue: network device
2736 * Reset the bytes and packet count of a network device and clear the
2737 * software flow control OFF bit for this network device
2739 static inline void netdev_reset_queue(struct net_device *dev_queue)
2741 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2745 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2746 * @dev: network device
2747 * @queue_index: given tx queue index
2749 * Returns 0 if given tx queue index >= number of device tx queues,
2750 * otherwise returns the originally passed tx queue index.
2752 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2754 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2755 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2756 dev->name, queue_index,
2757 dev->real_num_tx_queues);
2765 * netif_running - test if up
2766 * @dev: network device
2768 * Test if the device has been brought up.
2770 static inline bool netif_running(const struct net_device *dev)
2772 return test_bit(__LINK_STATE_START, &dev->state);
2776 * Routines to manage the subqueues on a device. We only need start
2777 * stop, and a check if it's stopped. All other device management is
2778 * done at the overall netdevice level.
2779 * Also test the device if we're multiqueue.
2783 * netif_start_subqueue - allow sending packets on subqueue
2784 * @dev: network device
2785 * @queue_index: sub queue index
2787 * Start individual transmit queue of a device with multiple transmit queues.
2789 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2791 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2793 netif_tx_start_queue(txq);
2797 * netif_stop_subqueue - stop sending packets on subqueue
2798 * @dev: network device
2799 * @queue_index: sub queue index
2801 * Stop individual transmit queue of a device with multiple transmit queues.
2803 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2805 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2806 netif_tx_stop_queue(txq);
2810 * netif_subqueue_stopped - test status of subqueue
2811 * @dev: network device
2812 * @queue_index: sub queue index
2814 * Check individual transmit queue of a device with multiple transmit queues.
2816 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2819 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2821 return netif_tx_queue_stopped(txq);
2824 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2825 struct sk_buff *skb)
2827 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2830 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2833 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2836 static inline int netif_set_xps_queue(struct net_device *dev,
2837 const struct cpumask *mask,
2845 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2846 * as a distribution range limit for the returned value.
2848 static inline u16 skb_tx_hash(const struct net_device *dev,
2849 struct sk_buff *skb)
2851 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2855 * netif_is_multiqueue - test if device has multiple transmit queues
2856 * @dev: network device
2858 * Check if device has multiple transmit queues
2860 static inline bool netif_is_multiqueue(const struct net_device *dev)
2862 return dev->num_tx_queues > 1;
2865 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2868 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2870 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2878 static inline unsigned int get_netdev_rx_queue_index(
2879 struct netdev_rx_queue *queue)
2881 struct net_device *dev = queue->dev;
2882 int index = queue - dev->_rx;
2884 BUG_ON(index >= dev->num_rx_queues);
2889 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2890 int netif_get_num_default_rss_queues(void);
2892 enum skb_free_reason {
2893 SKB_REASON_CONSUMED,
2897 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2898 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2901 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2902 * interrupt context or with hardware interrupts being disabled.
2903 * (in_irq() || irqs_disabled())
2905 * We provide four helpers that can be used in following contexts :
2907 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2908 * replacing kfree_skb(skb)
2910 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2911 * Typically used in place of consume_skb(skb) in TX completion path
2913 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2914 * replacing kfree_skb(skb)
2916 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2917 * and consumed a packet. Used in place of consume_skb(skb)
2919 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2921 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2924 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2926 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2929 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2931 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2934 static inline void dev_consume_skb_any(struct sk_buff *skb)
2936 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2939 int netif_rx(struct sk_buff *skb);
2940 int netif_rx_ni(struct sk_buff *skb);
2941 int netif_receive_skb_sk(struct sock *sk, struct sk_buff *skb);
2942 static inline int netif_receive_skb(struct sk_buff *skb)
2944 return netif_receive_skb_sk(skb->sk, skb);
2946 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2947 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2948 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2949 gro_result_t napi_gro_frags(struct napi_struct *napi);
2950 struct packet_offload *gro_find_receive_by_type(__be16 type);
2951 struct packet_offload *gro_find_complete_by_type(__be16 type);
2953 static inline void napi_free_frags(struct napi_struct *napi)
2955 kfree_skb(napi->skb);
2959 int netdev_rx_handler_register(struct net_device *dev,
2960 rx_handler_func_t *rx_handler,
2961 void *rx_handler_data);
2962 void netdev_rx_handler_unregister(struct net_device *dev);
2964 bool dev_valid_name(const char *name);
2965 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2966 int dev_ethtool(struct net *net, struct ifreq *);
2967 unsigned int dev_get_flags(const struct net_device *);
2968 int __dev_change_flags(struct net_device *, unsigned int flags);
2969 int dev_change_flags(struct net_device *, unsigned int);
2970 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2971 unsigned int gchanges);
2972 int dev_change_name(struct net_device *, const char *);
2973 int dev_set_alias(struct net_device *, const char *, size_t);
2974 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2975 int dev_set_mtu(struct net_device *, int);
2976 void dev_set_group(struct net_device *, int);
2977 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2978 int dev_change_carrier(struct net_device *, bool new_carrier);
2979 int dev_get_phys_port_id(struct net_device *dev,
2980 struct netdev_phys_item_id *ppid);
2981 int dev_get_phys_port_name(struct net_device *dev,
2982 char *name, size_t len);
2983 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2984 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2985 struct netdev_queue *txq, int *ret);
2986 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2987 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2988 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2990 extern int netdev_budget;
2992 /* Called by rtnetlink.c:rtnl_unlock() */
2993 void netdev_run_todo(void);
2996 * dev_put - release reference to device
2997 * @dev: network device
2999 * Release reference to device to allow it to be freed.
3001 static inline void dev_put(struct net_device *dev)
3003 this_cpu_dec(*dev->pcpu_refcnt);
3007 * dev_hold - get reference to device
3008 * @dev: network device
3010 * Hold reference to device to keep it from being freed.
3012 static inline void dev_hold(struct net_device *dev)
3014 this_cpu_inc(*dev->pcpu_refcnt);
3017 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3018 * and _off may be called from IRQ context, but it is caller
3019 * who is responsible for serialization of these calls.
3021 * The name carrier is inappropriate, these functions should really be
3022 * called netif_lowerlayer_*() because they represent the state of any
3023 * kind of lower layer not just hardware media.
3026 void linkwatch_init_dev(struct net_device *dev);
3027 void linkwatch_fire_event(struct net_device *dev);
3028 void linkwatch_forget_dev(struct net_device *dev);
3031 * netif_carrier_ok - test if carrier present
3032 * @dev: network device
3034 * Check if carrier is present on device
3036 static inline bool netif_carrier_ok(const struct net_device *dev)
3038 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3041 unsigned long dev_trans_start(struct net_device *dev);
3043 void __netdev_watchdog_up(struct net_device *dev);
3045 void netif_carrier_on(struct net_device *dev);
3047 void netif_carrier_off(struct net_device *dev);
3050 * netif_dormant_on - mark device as dormant.
3051 * @dev: network device
3053 * Mark device as dormant (as per RFC2863).
3055 * The dormant state indicates that the relevant interface is not
3056 * actually in a condition to pass packets (i.e., it is not 'up') but is
3057 * in a "pending" state, waiting for some external event. For "on-
3058 * demand" interfaces, this new state identifies the situation where the
3059 * interface is waiting for events to place it in the up state.
3062 static inline void netif_dormant_on(struct net_device *dev)
3064 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3065 linkwatch_fire_event(dev);
3069 * netif_dormant_off - set device as not dormant.
3070 * @dev: network device
3072 * Device is not in dormant state.
3074 static inline void netif_dormant_off(struct net_device *dev)
3076 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3077 linkwatch_fire_event(dev);
3081 * netif_dormant - test if carrier present
3082 * @dev: network device
3084 * Check if carrier is present on device
3086 static inline bool netif_dormant(const struct net_device *dev)
3088 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3093 * netif_oper_up - test if device is operational
3094 * @dev: network device
3096 * Check if carrier is operational
3098 static inline bool netif_oper_up(const struct net_device *dev)
3100 return (dev->operstate == IF_OPER_UP ||
3101 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3105 * netif_device_present - is device available or removed
3106 * @dev: network device
3108 * Check if device has not been removed from system.
3110 static inline bool netif_device_present(struct net_device *dev)
3112 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3115 void netif_device_detach(struct net_device *dev);
3117 void netif_device_attach(struct net_device *dev);
3120 * Network interface message level settings
3124 NETIF_MSG_DRV = 0x0001,
3125 NETIF_MSG_PROBE = 0x0002,
3126 NETIF_MSG_LINK = 0x0004,
3127 NETIF_MSG_TIMER = 0x0008,
3128 NETIF_MSG_IFDOWN = 0x0010,
3129 NETIF_MSG_IFUP = 0x0020,
3130 NETIF_MSG_RX_ERR = 0x0040,
3131 NETIF_MSG_TX_ERR = 0x0080,
3132 NETIF_MSG_TX_QUEUED = 0x0100,
3133 NETIF_MSG_INTR = 0x0200,
3134 NETIF_MSG_TX_DONE = 0x0400,
3135 NETIF_MSG_RX_STATUS = 0x0800,
3136 NETIF_MSG_PKTDATA = 0x1000,
3137 NETIF_MSG_HW = 0x2000,
3138 NETIF_MSG_WOL = 0x4000,
3141 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3142 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3143 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3144 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3145 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3146 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3147 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3148 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3149 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3150 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3151 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3152 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3153 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3154 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3155 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3157 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3160 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3161 return default_msg_enable_bits;
3162 if (debug_value == 0) /* no output */
3164 /* set low N bits */
3165 return (1 << debug_value) - 1;
3168 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3170 spin_lock(&txq->_xmit_lock);
3171 txq->xmit_lock_owner = cpu;
3174 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3176 spin_lock_bh(&txq->_xmit_lock);
3177 txq->xmit_lock_owner = smp_processor_id();
3180 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3182 bool ok = spin_trylock(&txq->_xmit_lock);
3184 txq->xmit_lock_owner = smp_processor_id();
3188 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3190 txq->xmit_lock_owner = -1;
3191 spin_unlock(&txq->_xmit_lock);
3194 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3196 txq->xmit_lock_owner = -1;
3197 spin_unlock_bh(&txq->_xmit_lock);
3200 static inline void txq_trans_update(struct netdev_queue *txq)
3202 if (txq->xmit_lock_owner != -1)
3203 txq->trans_start = jiffies;
3207 * netif_tx_lock - grab network device transmit lock
3208 * @dev: network device
3210 * Get network device transmit lock
3212 static inline void netif_tx_lock(struct net_device *dev)
3217 spin_lock(&dev->tx_global_lock);
3218 cpu = smp_processor_id();
3219 for (i = 0; i < dev->num_tx_queues; i++) {
3220 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3222 /* We are the only thread of execution doing a
3223 * freeze, but we have to grab the _xmit_lock in
3224 * order to synchronize with threads which are in
3225 * the ->hard_start_xmit() handler and already
3226 * checked the frozen bit.
3228 __netif_tx_lock(txq, cpu);
3229 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3230 __netif_tx_unlock(txq);
3234 static inline void netif_tx_lock_bh(struct net_device *dev)
3240 static inline void netif_tx_unlock(struct net_device *dev)
3244 for (i = 0; i < dev->num_tx_queues; i++) {
3245 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3247 /* No need to grab the _xmit_lock here. If the
3248 * queue is not stopped for another reason, we
3251 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3252 netif_schedule_queue(txq);
3254 spin_unlock(&dev->tx_global_lock);
3257 static inline void netif_tx_unlock_bh(struct net_device *dev)
3259 netif_tx_unlock(dev);
3263 #define HARD_TX_LOCK(dev, txq, cpu) { \
3264 if ((dev->features & NETIF_F_LLTX) == 0) { \
3265 __netif_tx_lock(txq, cpu); \
3269 #define HARD_TX_TRYLOCK(dev, txq) \
3270 (((dev->features & NETIF_F_LLTX) == 0) ? \
3271 __netif_tx_trylock(txq) : \
3274 #define HARD_TX_UNLOCK(dev, txq) { \
3275 if ((dev->features & NETIF_F_LLTX) == 0) { \
3276 __netif_tx_unlock(txq); \
3280 static inline void netif_tx_disable(struct net_device *dev)
3286 cpu = smp_processor_id();
3287 for (i = 0; i < dev->num_tx_queues; i++) {
3288 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3290 __netif_tx_lock(txq, cpu);
3291 netif_tx_stop_queue(txq);
3292 __netif_tx_unlock(txq);
3297 static inline void netif_addr_lock(struct net_device *dev)
3299 spin_lock(&dev->addr_list_lock);
3302 static inline void netif_addr_lock_nested(struct net_device *dev)
3304 int subclass = SINGLE_DEPTH_NESTING;
3306 if (dev->netdev_ops->ndo_get_lock_subclass)
3307 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3309 spin_lock_nested(&dev->addr_list_lock, subclass);
3312 static inline void netif_addr_lock_bh(struct net_device *dev)
3314 spin_lock_bh(&dev->addr_list_lock);
3317 static inline void netif_addr_unlock(struct net_device *dev)
3319 spin_unlock(&dev->addr_list_lock);
3322 static inline void netif_addr_unlock_bh(struct net_device *dev)
3324 spin_unlock_bh(&dev->addr_list_lock);
3328 * dev_addrs walker. Should be used only for read access. Call with
3329 * rcu_read_lock held.
3331 #define for_each_dev_addr(dev, ha) \
3332 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3334 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3336 void ether_setup(struct net_device *dev);
3338 /* Support for loadable net-drivers */
3339 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3340 unsigned char name_assign_type,
3341 void (*setup)(struct net_device *),
3342 unsigned int txqs, unsigned int rxqs);
3343 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3344 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3346 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3347 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3350 int register_netdev(struct net_device *dev);
3351 void unregister_netdev(struct net_device *dev);
3353 /* General hardware address lists handling functions */
3354 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3355 struct netdev_hw_addr_list *from_list, int addr_len);
3356 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3357 struct netdev_hw_addr_list *from_list, int addr_len);
3358 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3359 struct net_device *dev,
3360 int (*sync)(struct net_device *, const unsigned char *),
3361 int (*unsync)(struct net_device *,
3362 const unsigned char *));
3363 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3364 struct net_device *dev,
3365 int (*unsync)(struct net_device *,
3366 const unsigned char *));
3367 void __hw_addr_init(struct netdev_hw_addr_list *list);
3369 /* Functions used for device addresses handling */
3370 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3371 unsigned char addr_type);
3372 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3373 unsigned char addr_type);
3374 void dev_addr_flush(struct net_device *dev);
3375 int dev_addr_init(struct net_device *dev);
3377 /* Functions used for unicast addresses handling */
3378 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3379 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3380 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3381 int dev_uc_sync(struct net_device *to, struct net_device *from);
3382 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3383 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3384 void dev_uc_flush(struct net_device *dev);
3385 void dev_uc_init(struct net_device *dev);
3388 * __dev_uc_sync - Synchonize device's unicast list
3389 * @dev: device to sync
3390 * @sync: function to call if address should be added
3391 * @unsync: function to call if address should be removed
3393 * Add newly added addresses to the interface, and release
3394 * addresses that have been deleted.
3396 static inline int __dev_uc_sync(struct net_device *dev,
3397 int (*sync)(struct net_device *,
3398 const unsigned char *),
3399 int (*unsync)(struct net_device *,
3400 const unsigned char *))
3402 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3406 * __dev_uc_unsync - Remove synchronized addresses from device
3407 * @dev: device to sync
3408 * @unsync: function to call if address should be removed
3410 * Remove all addresses that were added to the device by dev_uc_sync().
3412 static inline void __dev_uc_unsync(struct net_device *dev,
3413 int (*unsync)(struct net_device *,
3414 const unsigned char *))
3416 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3419 /* Functions used for multicast addresses handling */
3420 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3421 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3422 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3423 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3424 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3425 int dev_mc_sync(struct net_device *to, struct net_device *from);
3426 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3427 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3428 void dev_mc_flush(struct net_device *dev);
3429 void dev_mc_init(struct net_device *dev);
3432 * __dev_mc_sync - Synchonize device's multicast list
3433 * @dev: device to sync
3434 * @sync: function to call if address should be added
3435 * @unsync: function to call if address should be removed
3437 * Add newly added addresses to the interface, and release
3438 * addresses that have been deleted.
3440 static inline int __dev_mc_sync(struct net_device *dev,
3441 int (*sync)(struct net_device *,
3442 const unsigned char *),
3443 int (*unsync)(struct net_device *,
3444 const unsigned char *))
3446 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3450 * __dev_mc_unsync - Remove synchronized addresses from device
3451 * @dev: device to sync
3452 * @unsync: function to call if address should be removed
3454 * Remove all addresses that were added to the device by dev_mc_sync().
3456 static inline void __dev_mc_unsync(struct net_device *dev,
3457 int (*unsync)(struct net_device *,
3458 const unsigned char *))
3460 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3463 /* Functions used for secondary unicast and multicast support */
3464 void dev_set_rx_mode(struct net_device *dev);
3465 void __dev_set_rx_mode(struct net_device *dev);
3466 int dev_set_promiscuity(struct net_device *dev, int inc);
3467 int dev_set_allmulti(struct net_device *dev, int inc);
3468 void netdev_state_change(struct net_device *dev);
3469 void netdev_notify_peers(struct net_device *dev);
3470 void netdev_features_change(struct net_device *dev);
3471 /* Load a device via the kmod */
3472 void dev_load(struct net *net, const char *name);
3473 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3474 struct rtnl_link_stats64 *storage);
3475 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3476 const struct net_device_stats *netdev_stats);
3478 extern int netdev_max_backlog;
3479 extern int netdev_tstamp_prequeue;
3480 extern int weight_p;
3481 extern int bpf_jit_enable;
3483 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3484 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3485 struct list_head **iter);
3486 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3487 struct list_head **iter);
3489 /* iterate through upper list, must be called under RCU read lock */
3490 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3491 for (iter = &(dev)->adj_list.upper, \
3492 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3494 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3496 /* iterate through upper list, must be called under RCU read lock */
3497 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3498 for (iter = &(dev)->all_adj_list.upper, \
3499 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3501 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3503 void *netdev_lower_get_next_private(struct net_device *dev,
3504 struct list_head **iter);
3505 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3506 struct list_head **iter);
3508 #define netdev_for_each_lower_private(dev, priv, iter) \
3509 for (iter = (dev)->adj_list.lower.next, \
3510 priv = netdev_lower_get_next_private(dev, &(iter)); \
3512 priv = netdev_lower_get_next_private(dev, &(iter)))
3514 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3515 for (iter = &(dev)->adj_list.lower, \
3516 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3518 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3520 void *netdev_lower_get_next(struct net_device *dev,
3521 struct list_head **iter);
3522 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3523 for (iter = &(dev)->adj_list.lower, \
3524 ldev = netdev_lower_get_next(dev, &(iter)); \
3526 ldev = netdev_lower_get_next(dev, &(iter)))
3528 void *netdev_adjacent_get_private(struct list_head *adj_list);
3529 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3530 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3531 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3532 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3533 int netdev_master_upper_dev_link(struct net_device *dev,
3534 struct net_device *upper_dev);
3535 int netdev_master_upper_dev_link_private(struct net_device *dev,
3536 struct net_device *upper_dev,
3538 void netdev_upper_dev_unlink(struct net_device *dev,
3539 struct net_device *upper_dev);
3540 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3541 void *netdev_lower_dev_get_private(struct net_device *dev,
3542 struct net_device *lower_dev);
3544 /* RSS keys are 40 or 52 bytes long */
3545 #define NETDEV_RSS_KEY_LEN 52
3546 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3547 void netdev_rss_key_fill(void *buffer, size_t len);
3549 int dev_get_nest_level(struct net_device *dev,
3550 bool (*type_check)(struct net_device *dev));
3551 int skb_checksum_help(struct sk_buff *skb);
3552 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3553 netdev_features_t features, bool tx_path);
3554 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3555 netdev_features_t features);
3557 struct netdev_bonding_info {
3562 struct netdev_notifier_bonding_info {
3563 struct netdev_notifier_info info; /* must be first */
3564 struct netdev_bonding_info bonding_info;
3567 void netdev_bonding_info_change(struct net_device *dev,
3568 struct netdev_bonding_info *bonding_info);
3571 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3573 return __skb_gso_segment(skb, features, true);
3575 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3577 static inline bool can_checksum_protocol(netdev_features_t features,
3580 return ((features & NETIF_F_GEN_CSUM) ||
3581 ((features & NETIF_F_V4_CSUM) &&
3582 protocol == htons(ETH_P_IP)) ||
3583 ((features & NETIF_F_V6_CSUM) &&
3584 protocol == htons(ETH_P_IPV6)) ||
3585 ((features & NETIF_F_FCOE_CRC) &&
3586 protocol == htons(ETH_P_FCOE)));
3590 void netdev_rx_csum_fault(struct net_device *dev);
3592 static inline void netdev_rx_csum_fault(struct net_device *dev)
3596 /* rx skb timestamps */
3597 void net_enable_timestamp(void);
3598 void net_disable_timestamp(void);
3600 #ifdef CONFIG_PROC_FS
3601 int __init dev_proc_init(void);
3603 #define dev_proc_init() 0
3606 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3607 struct sk_buff *skb, struct net_device *dev,
3610 skb->xmit_more = more ? 1 : 0;
3611 return ops->ndo_start_xmit(skb, dev);
3614 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3615 struct netdev_queue *txq, bool more)
3617 const struct net_device_ops *ops = dev->netdev_ops;
3620 rc = __netdev_start_xmit(ops, skb, dev, more);
3621 if (rc == NETDEV_TX_OK)
3622 txq_trans_update(txq);
3627 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3629 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3632 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3634 return netdev_class_create_file_ns(class_attr, NULL);
3637 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3639 netdev_class_remove_file_ns(class_attr, NULL);
3642 extern struct kobj_ns_type_operations net_ns_type_operations;
3644 const char *netdev_drivername(const struct net_device *dev);
3646 void linkwatch_run_queue(void);
3648 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3649 netdev_features_t f2)
3651 if (f1 & NETIF_F_GEN_CSUM)
3652 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3653 if (f2 & NETIF_F_GEN_CSUM)
3654 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3656 if (f1 & NETIF_F_GEN_CSUM)
3657 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3662 static inline netdev_features_t netdev_get_wanted_features(
3663 struct net_device *dev)
3665 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3667 netdev_features_t netdev_increment_features(netdev_features_t all,
3668 netdev_features_t one, netdev_features_t mask);
3670 /* Allow TSO being used on stacked device :
3671 * Performing the GSO segmentation before last device
3672 * is a performance improvement.
3674 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3675 netdev_features_t mask)
3677 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3680 int __netdev_update_features(struct net_device *dev);
3681 void netdev_update_features(struct net_device *dev);
3682 void netdev_change_features(struct net_device *dev);
3684 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3685 struct net_device *dev);
3687 netdev_features_t passthru_features_check(struct sk_buff *skb,
3688 struct net_device *dev,
3689 netdev_features_t features);
3690 netdev_features_t netif_skb_features(struct sk_buff *skb);
3692 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3694 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3696 /* check flags correspondence */
3697 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3698 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3699 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3700 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3701 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3702 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3703 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3704 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3705 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3706 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3707 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3708 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3709 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3711 return (features & feature) == feature;
3714 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3716 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3717 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3720 static inline bool netif_needs_gso(struct sk_buff *skb,
3721 netdev_features_t features)
3723 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3724 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3725 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3728 static inline void netif_set_gso_max_size(struct net_device *dev,
3731 dev->gso_max_size = size;
3734 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3735 int pulled_hlen, u16 mac_offset,
3738 skb->protocol = protocol;
3739 skb->encapsulation = 1;
3740 skb_push(skb, pulled_hlen);
3741 skb_reset_transport_header(skb);
3742 skb->mac_header = mac_offset;
3743 skb->network_header = skb->mac_header + mac_len;
3744 skb->mac_len = mac_len;
3747 static inline bool netif_is_macvlan(struct net_device *dev)
3749 return dev->priv_flags & IFF_MACVLAN;
3752 static inline bool netif_is_macvlan_port(struct net_device *dev)
3754 return dev->priv_flags & IFF_MACVLAN_PORT;
3757 static inline bool netif_is_ipvlan(struct net_device *dev)
3759 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3762 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3764 return dev->priv_flags & IFF_IPVLAN_MASTER;
3767 static inline bool netif_is_bond_master(struct net_device *dev)
3769 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3772 static inline bool netif_is_bond_slave(struct net_device *dev)
3774 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3777 static inline bool netif_supports_nofcs(struct net_device *dev)
3779 return dev->priv_flags & IFF_SUPP_NOFCS;
3782 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3783 static inline void netif_keep_dst(struct net_device *dev)
3785 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3788 extern struct pernet_operations __net_initdata loopback_net_ops;
3790 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3792 /* netdev_printk helpers, similar to dev_printk */
3794 static inline const char *netdev_name(const struct net_device *dev)
3796 if (!dev->name[0] || strchr(dev->name, '%'))
3797 return "(unnamed net_device)";
3801 static inline const char *netdev_reg_state(const struct net_device *dev)
3803 switch (dev->reg_state) {
3804 case NETREG_UNINITIALIZED: return " (uninitialized)";
3805 case NETREG_REGISTERED: return "";
3806 case NETREG_UNREGISTERING: return " (unregistering)";
3807 case NETREG_UNREGISTERED: return " (unregistered)";
3808 case NETREG_RELEASED: return " (released)";
3809 case NETREG_DUMMY: return " (dummy)";
3812 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3813 return " (unknown)";
3817 void netdev_printk(const char *level, const struct net_device *dev,
3818 const char *format, ...);
3820 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3822 void netdev_alert(const struct net_device *dev, const char *format, ...);
3824 void netdev_crit(const struct net_device *dev, const char *format, ...);
3826 void netdev_err(const struct net_device *dev, const char *format, ...);
3828 void netdev_warn(const struct net_device *dev, const char *format, ...);
3830 void netdev_notice(const struct net_device *dev, const char *format, ...);
3832 void netdev_info(const struct net_device *dev, const char *format, ...);
3834 #define MODULE_ALIAS_NETDEV(device) \
3835 MODULE_ALIAS("netdev-" device)
3837 #if defined(CONFIG_DYNAMIC_DEBUG)
3838 #define netdev_dbg(__dev, format, args...) \
3840 dynamic_netdev_dbg(__dev, format, ##args); \
3842 #elif defined(DEBUG)
3843 #define netdev_dbg(__dev, format, args...) \
3844 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3846 #define netdev_dbg(__dev, format, args...) \
3849 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3853 #if defined(VERBOSE_DEBUG)
3854 #define netdev_vdbg netdev_dbg
3857 #define netdev_vdbg(dev, format, args...) \
3860 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3866 * netdev_WARN() acts like dev_printk(), but with the key difference
3867 * of using a WARN/WARN_ON to get the message out, including the
3868 * file/line information and a backtrace.
3870 #define netdev_WARN(dev, format, args...) \
3871 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3872 netdev_reg_state(dev), ##args)
3874 /* netif printk helpers, similar to netdev_printk */
3876 #define netif_printk(priv, type, level, dev, fmt, args...) \
3878 if (netif_msg_##type(priv)) \
3879 netdev_printk(level, (dev), fmt, ##args); \
3882 #define netif_level(level, priv, type, dev, fmt, args...) \
3884 if (netif_msg_##type(priv)) \
3885 netdev_##level(dev, fmt, ##args); \
3888 #define netif_emerg(priv, type, dev, fmt, args...) \
3889 netif_level(emerg, priv, type, dev, fmt, ##args)
3890 #define netif_alert(priv, type, dev, fmt, args...) \
3891 netif_level(alert, priv, type, dev, fmt, ##args)
3892 #define netif_crit(priv, type, dev, fmt, args...) \
3893 netif_level(crit, priv, type, dev, fmt, ##args)
3894 #define netif_err(priv, type, dev, fmt, args...) \
3895 netif_level(err, priv, type, dev, fmt, ##args)
3896 #define netif_warn(priv, type, dev, fmt, args...) \
3897 netif_level(warn, priv, type, dev, fmt, ##args)
3898 #define netif_notice(priv, type, dev, fmt, args...) \
3899 netif_level(notice, priv, type, dev, fmt, ##args)
3900 #define netif_info(priv, type, dev, fmt, args...) \
3901 netif_level(info, priv, type, dev, fmt, ##args)
3903 #if defined(CONFIG_DYNAMIC_DEBUG)
3904 #define netif_dbg(priv, type, netdev, format, args...) \
3906 if (netif_msg_##type(priv)) \
3907 dynamic_netdev_dbg(netdev, format, ##args); \
3909 #elif defined(DEBUG)
3910 #define netif_dbg(priv, type, dev, format, args...) \
3911 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3913 #define netif_dbg(priv, type, dev, format, args...) \
3916 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3921 #if defined(VERBOSE_DEBUG)
3922 #define netif_vdbg netif_dbg
3924 #define netif_vdbg(priv, type, dev, format, args...) \
3927 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3933 * The list of packet types we will receive (as opposed to discard)
3934 * and the routines to invoke.
3936 * Why 16. Because with 16 the only overlap we get on a hash of the
3937 * low nibble of the protocol value is RARP/SNAP/X.25.
3939 * NOTE: That is no longer true with the addition of VLAN tags. Not
3940 * sure which should go first, but I bet it won't make much
3941 * difference if we are running VLANs. The good news is that
3942 * this protocol won't be in the list unless compiled in, so
3943 * the average user (w/out VLANs) will not be adversely affected.
3959 #define PTYPE_HASH_SIZE (16)
3960 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3962 #endif /* _LINUX_NETDEVICE_H */