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
29 #include <linux/if_ether.h>
30 #include <linux/if_packet.h>
31 #include <linux/if_link.h>
34 #include <linux/pm_qos.h>
35 #include <linux/timer.h>
36 #include <linux/delay.h>
37 #include <linux/atomic.h>
38 #include <asm/cache.h>
39 #include <asm/byteorder.h>
41 #include <linux/device.h>
42 #include <linux/percpu.h>
43 #include <linux/rculist.h>
44 #include <linux/dmaengine.h>
45 #include <linux/workqueue.h>
46 #include <linux/dynamic_queue_limits.h>
48 #include <linux/ethtool.h>
49 #include <net/net_namespace.h>
52 #include <net/dcbnl.h>
54 #include <net/netprio_cgroup.h>
56 #include <linux/netdev_features.h>
62 /* source back-compat hooks */
63 #define SET_ETHTOOL_OPS(netdev,ops) \
64 ( (netdev)->ethtool_ops = (ops) )
66 /* hardware address assignment types */
67 #define NET_ADDR_PERM 0 /* address is permanent (default) */
68 #define NET_ADDR_RANDOM 1 /* address is generated randomly */
69 #define NET_ADDR_STOLEN 2 /* address is stolen from other device */
71 /* Backlog congestion levels */
72 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
73 #define NET_RX_DROP 1 /* packet dropped */
76 * Transmit return codes: transmit return codes originate from three different
79 * - qdisc return codes
80 * - driver transmit return codes
83 * Drivers are allowed to return any one of those in their hard_start_xmit()
84 * function. Real network devices commonly used with qdiscs should only return
85 * the driver transmit return codes though - when qdiscs are used, the actual
86 * transmission happens asynchronously, so the value is not propagated to
87 * higher layers. Virtual network devices transmit synchronously, in this case
88 * the driver transmit return codes are consumed by dev_queue_xmit(), all
89 * others are propagated to higher layers.
92 /* qdisc ->enqueue() return codes. */
93 #define NET_XMIT_SUCCESS 0x00
94 #define NET_XMIT_DROP 0x01 /* skb dropped */
95 #define NET_XMIT_CN 0x02 /* congestion notification */
96 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
97 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
99 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
102 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
103 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105 /* Driver transmit return codes */
106 #define NETDEV_TX_MASK 0xf0
109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
112 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
114 typedef enum netdev_tx netdev_tx_t;
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 static inline bool dev_xmit_complete(int rc)
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 if (likely(rc < NET_XMIT_MASK))
136 #define MAX_ADDR_LEN 32 /* Largest hardware address length */
138 /* Initial net device group. All devices belong to group 0 by default. */
139 #define INIT_NETDEV_GROUP 0
143 * Compute the worst case header length according to the protocols
147 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
148 # if defined(CONFIG_MAC80211_MESH)
149 # define LL_MAX_HEADER 128
151 # define LL_MAX_HEADER 96
153 #elif IS_ENABLED(CONFIG_TR)
154 # define LL_MAX_HEADER 48
156 # define LL_MAX_HEADER 32
159 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
160 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
161 #define MAX_HEADER LL_MAX_HEADER
163 #define MAX_HEADER (LL_MAX_HEADER + 48)
167 * Old network device statistics. Fields are native words
168 * (unsigned long) so they can be read and written atomically.
171 struct net_device_stats {
172 unsigned long rx_packets;
173 unsigned long tx_packets;
174 unsigned long rx_bytes;
175 unsigned long tx_bytes;
176 unsigned long rx_errors;
177 unsigned long tx_errors;
178 unsigned long rx_dropped;
179 unsigned long tx_dropped;
180 unsigned long multicast;
181 unsigned long collisions;
182 unsigned long rx_length_errors;
183 unsigned long rx_over_errors;
184 unsigned long rx_crc_errors;
185 unsigned long rx_frame_errors;
186 unsigned long rx_fifo_errors;
187 unsigned long rx_missed_errors;
188 unsigned long tx_aborted_errors;
189 unsigned long tx_carrier_errors;
190 unsigned long tx_fifo_errors;
191 unsigned long tx_heartbeat_errors;
192 unsigned long tx_window_errors;
193 unsigned long rx_compressed;
194 unsigned long tx_compressed;
197 #endif /* __KERNEL__ */
200 /* Media selection options. */
213 #include <linux/cache.h>
214 #include <linux/skbuff.h>
217 #include <linux/jump_label.h>
218 extern struct jump_label_key rps_needed;
225 struct netdev_hw_addr {
226 struct list_head list;
227 unsigned char addr[MAX_ADDR_LEN];
229 #define NETDEV_HW_ADDR_T_LAN 1
230 #define NETDEV_HW_ADDR_T_SAN 2
231 #define NETDEV_HW_ADDR_T_SLAVE 3
232 #define NETDEV_HW_ADDR_T_UNICAST 4
233 #define NETDEV_HW_ADDR_T_MULTICAST 5
237 struct rcu_head rcu_head;
240 struct netdev_hw_addr_list {
241 struct list_head list;
245 #define netdev_hw_addr_list_count(l) ((l)->count)
246 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
247 #define netdev_hw_addr_list_for_each(ha, l) \
248 list_for_each_entry(ha, &(l)->list, list)
250 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
251 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
252 #define netdev_for_each_uc_addr(ha, dev) \
253 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
255 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
256 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
257 #define netdev_for_each_mc_addr(ha, dev) \
258 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
265 /* cached hardware header; allow for machine alignment needs. */
266 #define HH_DATA_MOD 16
267 #define HH_DATA_OFF(__len) \
268 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
269 #define HH_DATA_ALIGN(__len) \
270 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
271 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
274 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
276 * dev->hard_header_len ? (dev->hard_header_len +
277 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
279 * We could use other alignment values, but we must maintain the
280 * relationship HH alignment <= LL alignment.
282 #define LL_RESERVED_SPACE(dev) \
283 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
284 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
285 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
288 int (*create) (struct sk_buff *skb, struct net_device *dev,
289 unsigned short type, const void *daddr,
290 const void *saddr, unsigned len);
291 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
292 int (*rebuild)(struct sk_buff *skb);
293 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
294 void (*cache_update)(struct hh_cache *hh,
295 const struct net_device *dev,
296 const unsigned char *haddr);
299 /* These flag bits are private to the generic network queueing
300 * layer, they may not be explicitly referenced by any other
304 enum netdev_state_t {
306 __LINK_STATE_PRESENT,
307 __LINK_STATE_NOCARRIER,
308 __LINK_STATE_LINKWATCH_PENDING,
309 __LINK_STATE_DORMANT,
314 * This structure holds at boot time configured netdevice settings. They
315 * are then used in the device probing.
317 struct netdev_boot_setup {
321 #define NETDEV_BOOT_SETUP_MAX 8
323 extern int __init netdev_boot_setup(char *str);
326 * Structure for NAPI scheduling similar to tasklet but with weighting
329 /* The poll_list must only be managed by the entity which
330 * changes the state of the NAPI_STATE_SCHED bit. This means
331 * whoever atomically sets that bit can add this napi_struct
332 * to the per-cpu poll_list, and whoever clears that bit
333 * can remove from the list right before clearing the bit.
335 struct list_head poll_list;
339 int (*poll)(struct napi_struct *, int);
340 #ifdef CONFIG_NETPOLL
341 spinlock_t poll_lock;
345 unsigned int gro_count;
347 struct net_device *dev;
348 struct list_head dev_list;
349 struct sk_buff *gro_list;
354 NAPI_STATE_SCHED, /* Poll is scheduled */
355 NAPI_STATE_DISABLE, /* Disable pending */
356 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
366 typedef enum gro_result gro_result_t;
369 * enum rx_handler_result - Possible return values for rx_handlers.
370 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
372 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
373 * case skb->dev was changed by rx_handler.
374 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
375 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
377 * rx_handlers are functions called from inside __netif_receive_skb(), to do
378 * special processing of the skb, prior to delivery to protocol handlers.
380 * Currently, a net_device can only have a single rx_handler registered. Trying
381 * to register a second rx_handler will return -EBUSY.
383 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
384 * To unregister a rx_handler on a net_device, use
385 * netdev_rx_handler_unregister().
387 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
390 * If the rx_handler consumed to skb in some way, it should return
391 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
392 * the skb to be delivered in some other ways.
394 * If the rx_handler changed skb->dev, to divert the skb to another
395 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
396 * new device will be called if it exists.
398 * If the rx_handler consider the skb should be ignored, it should return
399 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
400 * are registred on exact device (ptype->dev == skb->dev).
402 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
403 * delivered, it should return RX_HANDLER_PASS.
405 * A device without a registered rx_handler will behave as if rx_handler
406 * returned RX_HANDLER_PASS.
409 enum rx_handler_result {
415 typedef enum rx_handler_result rx_handler_result_t;
416 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
418 extern void __napi_schedule(struct napi_struct *n);
420 static inline int napi_disable_pending(struct napi_struct *n)
422 return test_bit(NAPI_STATE_DISABLE, &n->state);
426 * napi_schedule_prep - check if napi can be scheduled
429 * Test if NAPI routine is already running, and if not mark
430 * it as running. This is used as a condition variable
431 * insure only one NAPI poll instance runs. We also make
432 * sure there is no pending NAPI disable.
434 static inline int napi_schedule_prep(struct napi_struct *n)
436 return !napi_disable_pending(n) &&
437 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
441 * napi_schedule - schedule NAPI poll
444 * Schedule NAPI poll routine to be called if it is not already
447 static inline void napi_schedule(struct napi_struct *n)
449 if (napi_schedule_prep(n))
453 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
454 static inline int napi_reschedule(struct napi_struct *napi)
456 if (napi_schedule_prep(napi)) {
457 __napi_schedule(napi);
464 * napi_complete - NAPI processing complete
467 * Mark NAPI processing as complete.
469 extern void __napi_complete(struct napi_struct *n);
470 extern void napi_complete(struct napi_struct *n);
473 * napi_disable - prevent NAPI from scheduling
476 * Stop NAPI from being scheduled on this context.
477 * Waits till any outstanding processing completes.
479 static inline void napi_disable(struct napi_struct *n)
481 set_bit(NAPI_STATE_DISABLE, &n->state);
482 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
484 clear_bit(NAPI_STATE_DISABLE, &n->state);
488 * napi_enable - enable NAPI scheduling
491 * Resume NAPI from being scheduled on this context.
492 * Must be paired with napi_disable.
494 static inline void napi_enable(struct napi_struct *n)
496 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
497 smp_mb__before_clear_bit();
498 clear_bit(NAPI_STATE_SCHED, &n->state);
503 * napi_synchronize - wait until NAPI is not running
506 * Wait until NAPI is done being scheduled on this context.
507 * Waits till any outstanding processing completes but
508 * does not disable future activations.
510 static inline void napi_synchronize(const struct napi_struct *n)
512 while (test_bit(NAPI_STATE_SCHED, &n->state))
516 # define napi_synchronize(n) barrier()
519 enum netdev_queue_state_t {
520 __QUEUE_STATE_DRV_XOFF,
521 __QUEUE_STATE_STACK_XOFF,
522 __QUEUE_STATE_FROZEN,
523 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
524 (1 << __QUEUE_STATE_STACK_XOFF))
525 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
526 (1 << __QUEUE_STATE_FROZEN))
529 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
530 * netif_tx_* functions below are used to manipulate this flag. The
531 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
532 * queue independently. The netif_xmit_*stopped functions below are called
533 * to check if the queue has been stopped by the driver or stack (either
534 * of the XOFF bits are set in the state). Drivers should not need to call
535 * netif_xmit*stopped functions, they should only be using netif_tx_*.
538 struct netdev_queue {
542 struct net_device *dev;
544 struct Qdisc *qdisc_sleeping;
548 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
554 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
557 * please use this field instead of dev->trans_start
559 unsigned long trans_start;
562 * Number of TX timeouts for this queue
563 * (/sys/class/net/DEV/Q/trans_timeout)
565 unsigned long trans_timeout;
572 } ____cacheline_aligned_in_smp;
574 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
576 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
583 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
585 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
592 * This structure holds an RPS map which can be of variable length. The
593 * map is an array of CPUs.
600 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
603 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
604 * tail pointer for that CPU's input queue at the time of last enqueue, and
605 * a hardware filter index.
607 struct rps_dev_flow {
610 unsigned int last_qtail;
612 #define RPS_NO_FILTER 0xffff
615 * The rps_dev_flow_table structure contains a table of flow mappings.
617 struct rps_dev_flow_table {
620 struct work_struct free_work;
621 struct rps_dev_flow flows[0];
623 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
624 ((_num) * sizeof(struct rps_dev_flow)))
627 * The rps_sock_flow_table contains mappings of flows to the last CPU
628 * on which they were processed by the application (set in recvmsg).
630 struct rps_sock_flow_table {
634 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
635 ((_num) * sizeof(u16)))
637 #define RPS_NO_CPU 0xffff
639 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
643 unsigned int cpu, index = hash & table->mask;
645 /* We only give a hint, preemption can change cpu under us */
646 cpu = raw_smp_processor_id();
648 if (table->ents[index] != cpu)
649 table->ents[index] = cpu;
653 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
657 table->ents[hash & table->mask] = RPS_NO_CPU;
660 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
662 #ifdef CONFIG_RFS_ACCEL
663 extern bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
664 u32 flow_id, u16 filter_id);
667 /* This structure contains an instance of an RX queue. */
668 struct netdev_rx_queue {
669 struct rps_map __rcu *rps_map;
670 struct rps_dev_flow_table __rcu *rps_flow_table;
672 struct net_device *dev;
673 } ____cacheline_aligned_in_smp;
674 #endif /* CONFIG_RPS */
678 * This structure holds an XPS map which can be of variable length. The
679 * map is an array of queues.
683 unsigned int alloc_len;
687 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
688 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
692 * This structure holds all XPS maps for device. Maps are indexed by CPU.
694 struct xps_dev_maps {
696 struct xps_map __rcu *cpu_map[0];
698 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
699 (nr_cpu_ids * sizeof(struct xps_map *)))
700 #endif /* CONFIG_XPS */
702 #define TC_MAX_QUEUE 16
703 #define TC_BITMASK 15
704 /* HW offloaded queuing disciplines txq count and offset maps */
705 struct netdev_tc_txq {
711 * This structure defines the management hooks for network devices.
712 * The following hooks can be defined; unless noted otherwise, they are
713 * optional and can be filled with a null pointer.
715 * int (*ndo_init)(struct net_device *dev);
716 * This function is called once when network device is registered.
717 * The network device can use this to any late stage initializaton
718 * or semantic validattion. It can fail with an error code which will
719 * be propogated back to register_netdev
721 * void (*ndo_uninit)(struct net_device *dev);
722 * This function is called when device is unregistered or when registration
723 * fails. It is not called if init fails.
725 * int (*ndo_open)(struct net_device *dev);
726 * This function is called when network device transistions to the up
729 * int (*ndo_stop)(struct net_device *dev);
730 * This function is called when network device transistions to the down
733 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
734 * struct net_device *dev);
735 * Called when a packet needs to be transmitted.
736 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
737 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
738 * Required can not be NULL.
740 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
741 * Called to decide which queue to when device supports multiple
744 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
745 * This function is called to allow device receiver to make
746 * changes to configuration when multicast or promiscious is enabled.
748 * void (*ndo_set_rx_mode)(struct net_device *dev);
749 * This function is called device changes address list filtering.
750 * If driver handles unicast address filtering, it should set
751 * IFF_UNICAST_FLT to its priv_flags.
753 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
754 * This function is called when the Media Access Control address
755 * needs to be changed. If this interface is not defined, the
756 * mac address can not be changed.
758 * int (*ndo_validate_addr)(struct net_device *dev);
759 * Test if Media Access Control address is valid for the device.
761 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
762 * Called when a user request an ioctl which can't be handled by
763 * the generic interface code. If not defined ioctl's return
764 * not supported error code.
766 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
767 * Used to set network devices bus interface parameters. This interface
768 * is retained for legacy reason, new devices should use the bus
769 * interface (PCI) for low level management.
771 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
772 * Called when a user wants to change the Maximum Transfer Unit
773 * of a device. If not defined, any request to change MTU will
774 * will return an error.
776 * void (*ndo_tx_timeout)(struct net_device *dev);
777 * Callback uses when the transmitter has not made any progress
778 * for dev->watchdog ticks.
780 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
781 * struct rtnl_link_stats64 *storage);
782 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
783 * Called when a user wants to get the network device usage
784 * statistics. Drivers must do one of the following:
785 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
786 * rtnl_link_stats64 structure passed by the caller.
787 * 2. Define @ndo_get_stats to update a net_device_stats structure
788 * (which should normally be dev->stats) and return a pointer to
789 * it. The structure may be changed asynchronously only if each
790 * field is written atomically.
791 * 3. Update dev->stats asynchronously and atomically, and define
794 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, unsigned short vid);
795 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER)
796 * this function is called when a VLAN id is registered.
798 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
799 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER)
800 * this function is called when a VLAN id is unregistered.
802 * void (*ndo_poll_controller)(struct net_device *dev);
804 * SR-IOV management functions.
805 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
806 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
807 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
808 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
809 * int (*ndo_get_vf_config)(struct net_device *dev,
810 * int vf, struct ifla_vf_info *ivf);
811 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
812 * struct nlattr *port[]);
813 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
814 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
815 * Called to setup 'tc' number of traffic classes in the net device. This
816 * is always called from the stack with the rtnl lock held and netif tx
817 * queues stopped. This allows the netdevice to perform queue management
820 * Fiber Channel over Ethernet (FCoE) offload functions.
821 * int (*ndo_fcoe_enable)(struct net_device *dev);
822 * Called when the FCoE protocol stack wants to start using LLD for FCoE
823 * so the underlying device can perform whatever needed configuration or
824 * initialization to support acceleration of FCoE traffic.
826 * int (*ndo_fcoe_disable)(struct net_device *dev);
827 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
828 * so the underlying device can perform whatever needed clean-ups to
829 * stop supporting acceleration of FCoE traffic.
831 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
832 * struct scatterlist *sgl, unsigned int sgc);
833 * Called when the FCoE Initiator wants to initialize an I/O that
834 * is a possible candidate for Direct Data Placement (DDP). The LLD can
835 * perform necessary setup and returns 1 to indicate the device is set up
836 * successfully to perform DDP on this I/O, otherwise this returns 0.
838 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
839 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
840 * indicated by the FC exchange id 'xid', so the underlying device can
841 * clean up and reuse resources for later DDP requests.
843 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
844 * struct scatterlist *sgl, unsigned int sgc);
845 * Called when the FCoE Target wants to initialize an I/O that
846 * is a possible candidate for Direct Data Placement (DDP). The LLD can
847 * perform necessary setup and returns 1 to indicate the device is set up
848 * successfully to perform DDP on this I/O, otherwise this returns 0.
850 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
851 * Called when the underlying device wants to override default World Wide
852 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
853 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
854 * protocol stack to use.
857 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
858 * u16 rxq_index, u32 flow_id);
859 * Set hardware filter for RFS. rxq_index is the target queue index;
860 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
861 * Return the filter ID on success, or a negative error code.
863 * Slave management functions (for bridge, bonding, etc). User should
864 * call netdev_set_master() to set dev->master properly.
865 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
866 * Called to make another netdev an underling.
868 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
869 * Called to release previously enslaved netdev.
871 * Feature/offload setting functions.
872 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
873 * netdev_features_t features);
874 * Adjusts the requested feature flags according to device-specific
875 * constraints, and returns the resulting flags. Must not modify
878 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
879 * Called to update device configuration to new features. Passed
880 * feature set might be less than what was returned by ndo_fix_features()).
881 * Must return >0 or -errno if it changed dev->features itself.
884 struct net_device_ops {
885 int (*ndo_init)(struct net_device *dev);
886 void (*ndo_uninit)(struct net_device *dev);
887 int (*ndo_open)(struct net_device *dev);
888 int (*ndo_stop)(struct net_device *dev);
889 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
890 struct net_device *dev);
891 u16 (*ndo_select_queue)(struct net_device *dev,
892 struct sk_buff *skb);
893 void (*ndo_change_rx_flags)(struct net_device *dev,
895 void (*ndo_set_rx_mode)(struct net_device *dev);
896 int (*ndo_set_mac_address)(struct net_device *dev,
898 int (*ndo_validate_addr)(struct net_device *dev);
899 int (*ndo_do_ioctl)(struct net_device *dev,
900 struct ifreq *ifr, int cmd);
901 int (*ndo_set_config)(struct net_device *dev,
903 int (*ndo_change_mtu)(struct net_device *dev,
905 int (*ndo_neigh_setup)(struct net_device *dev,
906 struct neigh_parms *);
907 void (*ndo_tx_timeout) (struct net_device *dev);
909 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
910 struct rtnl_link_stats64 *storage);
911 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
913 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
915 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
917 #ifdef CONFIG_NET_POLL_CONTROLLER
918 void (*ndo_poll_controller)(struct net_device *dev);
919 int (*ndo_netpoll_setup)(struct net_device *dev,
920 struct netpoll_info *info);
921 void (*ndo_netpoll_cleanup)(struct net_device *dev);
923 int (*ndo_set_vf_mac)(struct net_device *dev,
925 int (*ndo_set_vf_vlan)(struct net_device *dev,
926 int queue, u16 vlan, u8 qos);
927 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
929 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
930 int vf, bool setting);
931 int (*ndo_get_vf_config)(struct net_device *dev,
933 struct ifla_vf_info *ivf);
934 int (*ndo_set_vf_port)(struct net_device *dev,
936 struct nlattr *port[]);
937 int (*ndo_get_vf_port)(struct net_device *dev,
938 int vf, struct sk_buff *skb);
939 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
940 #if IS_ENABLED(CONFIG_FCOE)
941 int (*ndo_fcoe_enable)(struct net_device *dev);
942 int (*ndo_fcoe_disable)(struct net_device *dev);
943 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
945 struct scatterlist *sgl,
947 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
949 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
951 struct scatterlist *sgl,
955 #if IS_ENABLED(CONFIG_LIBFCOE)
956 #define NETDEV_FCOE_WWNN 0
957 #define NETDEV_FCOE_WWPN 1
958 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
962 #ifdef CONFIG_RFS_ACCEL
963 int (*ndo_rx_flow_steer)(struct net_device *dev,
964 const struct sk_buff *skb,
968 int (*ndo_add_slave)(struct net_device *dev,
969 struct net_device *slave_dev);
970 int (*ndo_del_slave)(struct net_device *dev,
971 struct net_device *slave_dev);
972 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
973 netdev_features_t features);
974 int (*ndo_set_features)(struct net_device *dev,
975 netdev_features_t features);
976 int (*ndo_neigh_construct)(struct neighbour *n);
977 void (*ndo_neigh_destroy)(struct neighbour *n);
981 * The DEVICE structure.
982 * Actually, this whole structure is a big mistake. It mixes I/O
983 * data with strictly "high-level" data, and it has to know about
984 * almost every data structure used in the INET module.
986 * FIXME: cleanup struct net_device such that network protocol info
993 * This is the first field of the "visible" part of this structure
994 * (i.e. as seen by users in the "Space.c" file). It is the name
999 struct pm_qos_request pm_qos_req;
1001 /* device name hash chain */
1002 struct hlist_node name_hlist;
1007 * I/O specific fields
1008 * FIXME: Merge these and struct ifmap into one
1010 unsigned long mem_end; /* shared mem end */
1011 unsigned long mem_start; /* shared mem start */
1012 unsigned long base_addr; /* device I/O address */
1013 unsigned int irq; /* device IRQ number */
1016 * Some hardware also needs these fields, but they are not
1017 * part of the usual set specified in Space.c.
1020 unsigned long state;
1022 struct list_head dev_list;
1023 struct list_head napi_list;
1024 struct list_head unreg_list;
1026 /* currently active device features */
1027 netdev_features_t features;
1028 /* user-changeable features */
1029 netdev_features_t hw_features;
1030 /* user-requested features */
1031 netdev_features_t wanted_features;
1032 /* mask of features inheritable by VLAN devices */
1033 netdev_features_t vlan_features;
1035 /* Interface index. Unique device identifier */
1039 struct net_device_stats stats;
1040 atomic_long_t rx_dropped; /* dropped packets by core network
1041 * Do not use this in drivers.
1044 #ifdef CONFIG_WIRELESS_EXT
1045 /* List of functions to handle Wireless Extensions (instead of ioctl).
1046 * See <net/iw_handler.h> for details. Jean II */
1047 const struct iw_handler_def * wireless_handlers;
1048 /* Instance data managed by the core of Wireless Extensions. */
1049 struct iw_public_data * wireless_data;
1051 /* Management operations */
1052 const struct net_device_ops *netdev_ops;
1053 const struct ethtool_ops *ethtool_ops;
1055 /* Hardware header description */
1056 const struct header_ops *header_ops;
1058 unsigned int flags; /* interface flags (a la BSD) */
1059 unsigned int priv_flags; /* Like 'flags' but invisible to userspace. */
1060 unsigned short gflags;
1061 unsigned short padded; /* How much padding added by alloc_netdev() */
1063 unsigned char operstate; /* RFC2863 operstate */
1064 unsigned char link_mode; /* mapping policy to operstate */
1066 unsigned char if_port; /* Selectable AUI, TP,..*/
1067 unsigned char dma; /* DMA channel */
1069 unsigned int mtu; /* interface MTU value */
1070 unsigned short type; /* interface hardware type */
1071 unsigned short hard_header_len; /* hardware hdr length */
1073 /* extra head- and tailroom the hardware may need, but not in all cases
1074 * can this be guaranteed, especially tailroom. Some cases also use
1075 * LL_MAX_HEADER instead to allocate the skb.
1077 unsigned short needed_headroom;
1078 unsigned short needed_tailroom;
1080 /* Interface address info. */
1081 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1082 unsigned char addr_assign_type; /* hw address assignment type */
1083 unsigned char addr_len; /* hardware address length */
1084 unsigned char neigh_priv_len;
1085 unsigned short dev_id; /* for shared network cards */
1087 spinlock_t addr_list_lock;
1088 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1089 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1091 unsigned int promiscuity;
1092 unsigned int allmulti;
1095 /* Protocol specific pointers */
1097 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1098 struct vlan_info __rcu *vlan_info; /* VLAN info */
1100 #if IS_ENABLED(CONFIG_NET_DSA)
1101 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1103 void *atalk_ptr; /* AppleTalk link */
1104 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1105 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1106 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1107 void *ec_ptr; /* Econet specific data */
1108 void *ax25_ptr; /* AX.25 specific data */
1109 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1110 assign before registering */
1113 * Cache lines mostly used on receive path (including eth_type_trans())
1115 unsigned long last_rx; /* Time of last Rx
1116 * This should not be set in
1117 * drivers, unless really needed,
1118 * because network stack (bonding)
1119 * use it if/when necessary, to
1120 * avoid dirtying this cache line.
1123 struct net_device *master; /* Pointer to master device of a group,
1124 * which this device is member of.
1127 /* Interface address info used in eth_type_trans() */
1128 unsigned char *dev_addr; /* hw address, (before bcast
1129 because most packets are
1132 struct netdev_hw_addr_list dev_addrs; /* list of device
1135 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1138 struct kset *queues_kset;
1142 struct netdev_rx_queue *_rx;
1144 /* Number of RX queues allocated at register_netdev() time */
1145 unsigned int num_rx_queues;
1147 /* Number of RX queues currently active in device */
1148 unsigned int real_num_rx_queues;
1150 #ifdef CONFIG_RFS_ACCEL
1151 /* CPU reverse-mapping for RX completion interrupts, indexed
1152 * by RX queue number. Assigned by driver. This must only be
1153 * set if the ndo_rx_flow_steer operation is defined. */
1154 struct cpu_rmap *rx_cpu_rmap;
1158 rx_handler_func_t __rcu *rx_handler;
1159 void __rcu *rx_handler_data;
1161 struct netdev_queue __rcu *ingress_queue;
1164 * Cache lines mostly used on transmit path
1166 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1168 /* Number of TX queues allocated at alloc_netdev_mq() time */
1169 unsigned int num_tx_queues;
1171 /* Number of TX queues currently active in device */
1172 unsigned int real_num_tx_queues;
1174 /* root qdisc from userspace point of view */
1175 struct Qdisc *qdisc;
1177 unsigned long tx_queue_len; /* Max frames per queue allowed */
1178 spinlock_t tx_global_lock;
1181 struct xps_dev_maps __rcu *xps_maps;
1184 /* These may be needed for future network-power-down code. */
1187 * trans_start here is expensive for high speed devices on SMP,
1188 * please use netdev_queue->trans_start instead.
1190 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1192 int watchdog_timeo; /* used by dev_watchdog() */
1193 struct timer_list watchdog_timer;
1195 /* Number of references to this device */
1196 int __percpu *pcpu_refcnt;
1198 /* delayed register/unregister */
1199 struct list_head todo_list;
1200 /* device index hash chain */
1201 struct hlist_node index_hlist;
1203 struct list_head link_watch_list;
1205 /* register/unregister state machine */
1206 enum { NETREG_UNINITIALIZED=0,
1207 NETREG_REGISTERED, /* completed register_netdevice */
1208 NETREG_UNREGISTERING, /* called unregister_netdevice */
1209 NETREG_UNREGISTERED, /* completed unregister todo */
1210 NETREG_RELEASED, /* called free_netdev */
1211 NETREG_DUMMY, /* dummy device for NAPI poll */
1214 bool dismantle; /* device is going do be freed */
1217 RTNL_LINK_INITIALIZED,
1218 RTNL_LINK_INITIALIZING,
1219 } rtnl_link_state:16;
1221 /* Called from unregister, can be used to call free_netdev */
1222 void (*destructor)(struct net_device *dev);
1224 #ifdef CONFIG_NETPOLL
1225 struct netpoll_info *npinfo;
1228 #ifdef CONFIG_NET_NS
1229 /* Network namespace this network device is inside */
1233 /* mid-layer private */
1236 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1237 struct pcpu_tstats __percpu *tstats; /* tunnel stats */
1238 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1241 struct garp_port __rcu *garp_port;
1243 /* class/net/name entry */
1245 /* space for optional device, statistics, and wireless sysfs groups */
1246 const struct attribute_group *sysfs_groups[4];
1248 /* rtnetlink link ops */
1249 const struct rtnl_link_ops *rtnl_link_ops;
1251 /* for setting kernel sock attribute on TCP connection setup */
1252 #define GSO_MAX_SIZE 65536
1253 unsigned int gso_max_size;
1256 /* Data Center Bridging netlink ops */
1257 const struct dcbnl_rtnl_ops *dcbnl_ops;
1260 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1261 u8 prio_tc_map[TC_BITMASK + 1];
1263 #if IS_ENABLED(CONFIG_FCOE)
1264 /* max exchange id for FCoE LRO by ddp */
1265 unsigned int fcoe_ddp_xid;
1267 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
1268 struct netprio_map __rcu *priomap;
1270 /* phy device may attach itself for hardware timestamping */
1271 struct phy_device *phydev;
1273 /* group the device belongs to */
1276 #define to_net_dev(d) container_of(d, struct net_device, dev)
1278 #define NETDEV_ALIGN 32
1281 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1283 return dev->prio_tc_map[prio & TC_BITMASK];
1287 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1289 if (tc >= dev->num_tc)
1292 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1297 void netdev_reset_tc(struct net_device *dev)
1300 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1301 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1305 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1307 if (tc >= dev->num_tc)
1310 dev->tc_to_txq[tc].count = count;
1311 dev->tc_to_txq[tc].offset = offset;
1316 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1318 if (num_tc > TC_MAX_QUEUE)
1321 dev->num_tc = num_tc;
1326 int netdev_get_num_tc(struct net_device *dev)
1332 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1335 return &dev->_tx[index];
1338 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1339 void (*f)(struct net_device *,
1340 struct netdev_queue *,
1346 for (i = 0; i < dev->num_tx_queues; i++)
1347 f(dev, &dev->_tx[i], arg);
1351 * Net namespace inlines
1354 struct net *dev_net(const struct net_device *dev)
1356 return read_pnet(&dev->nd_net);
1360 void dev_net_set(struct net_device *dev, struct net *net)
1362 #ifdef CONFIG_NET_NS
1363 release_net(dev->nd_net);
1364 dev->nd_net = hold_net(net);
1368 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1370 #ifdef CONFIG_NET_DSA_TAG_DSA
1371 if (dev->dsa_ptr != NULL)
1372 return dsa_uses_dsa_tags(dev->dsa_ptr);
1378 #ifndef CONFIG_NET_NS
1379 static inline void skb_set_dev(struct sk_buff *skb, struct net_device *dev)
1383 #else /* CONFIG_NET_NS */
1384 void skb_set_dev(struct sk_buff *skb, struct net_device *dev);
1387 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1389 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1390 if (dev->dsa_ptr != NULL)
1391 return dsa_uses_trailer_tags(dev->dsa_ptr);
1398 * netdev_priv - access network device private data
1399 * @dev: network device
1401 * Get network device private data
1403 static inline void *netdev_priv(const struct net_device *dev)
1405 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1408 /* Set the sysfs physical device reference for the network logical device
1409 * if set prior to registration will cause a symlink during initialization.
1411 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1413 /* Set the sysfs device type for the network logical device to allow
1414 * fin grained indentification of different network device types. For
1415 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1417 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1420 * netif_napi_add - initialize a napi context
1421 * @dev: network device
1422 * @napi: napi context
1423 * @poll: polling function
1424 * @weight: default weight
1426 * netif_napi_add() must be used to initialize a napi context prior to calling
1427 * *any* of the other napi related functions.
1429 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1430 int (*poll)(struct napi_struct *, int), int weight);
1433 * netif_napi_del - remove a napi context
1434 * @napi: napi context
1436 * netif_napi_del() removes a napi context from the network device napi list
1438 void netif_napi_del(struct napi_struct *napi);
1440 struct napi_gro_cb {
1441 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1444 /* Length of frag0. */
1445 unsigned int frag0_len;
1447 /* This indicates where we are processing relative to skb->data. */
1450 /* This is non-zero if the packet may be of the same flow. */
1453 /* This is non-zero if the packet cannot be merged with the new skb. */
1456 /* Number of segments aggregated. */
1463 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1465 struct packet_type {
1466 __be16 type; /* This is really htons(ether_type). */
1467 struct net_device *dev; /* NULL is wildcarded here */
1468 int (*func) (struct sk_buff *,
1469 struct net_device *,
1470 struct packet_type *,
1471 struct net_device *);
1472 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1473 netdev_features_t features);
1474 int (*gso_send_check)(struct sk_buff *skb);
1475 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1476 struct sk_buff *skb);
1477 int (*gro_complete)(struct sk_buff *skb);
1478 void *af_packet_priv;
1479 struct list_head list;
1482 #include <linux/notifier.h>
1484 /* netdevice notifier chain. Please remember to update the rtnetlink
1485 * notification exclusion list in rtnetlink_event() when adding new
1488 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1489 #define NETDEV_DOWN 0x0002
1490 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1491 detected a hardware crash and restarted
1492 - we can use this eg to kick tcp sessions
1494 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1495 #define NETDEV_REGISTER 0x0005
1496 #define NETDEV_UNREGISTER 0x0006
1497 #define NETDEV_CHANGEMTU 0x0007
1498 #define NETDEV_CHANGEADDR 0x0008
1499 #define NETDEV_GOING_DOWN 0x0009
1500 #define NETDEV_CHANGENAME 0x000A
1501 #define NETDEV_FEAT_CHANGE 0x000B
1502 #define NETDEV_BONDING_FAILOVER 0x000C
1503 #define NETDEV_PRE_UP 0x000D
1504 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1505 #define NETDEV_POST_TYPE_CHANGE 0x000F
1506 #define NETDEV_POST_INIT 0x0010
1507 #define NETDEV_UNREGISTER_BATCH 0x0011
1508 #define NETDEV_RELEASE 0x0012
1509 #define NETDEV_NOTIFY_PEERS 0x0013
1510 #define NETDEV_JOIN 0x0014
1512 extern int register_netdevice_notifier(struct notifier_block *nb);
1513 extern int unregister_netdevice_notifier(struct notifier_block *nb);
1514 extern int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1517 extern rwlock_t dev_base_lock; /* Device list lock */
1520 #define for_each_netdev(net, d) \
1521 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1522 #define for_each_netdev_reverse(net, d) \
1523 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1524 #define for_each_netdev_rcu(net, d) \
1525 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1526 #define for_each_netdev_safe(net, d, n) \
1527 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1528 #define for_each_netdev_continue(net, d) \
1529 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1530 #define for_each_netdev_continue_rcu(net, d) \
1531 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1532 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1534 static inline struct net_device *next_net_device(struct net_device *dev)
1536 struct list_head *lh;
1540 lh = dev->dev_list.next;
1541 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1544 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1546 struct list_head *lh;
1550 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1551 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1554 static inline struct net_device *first_net_device(struct net *net)
1556 return list_empty(&net->dev_base_head) ? NULL :
1557 net_device_entry(net->dev_base_head.next);
1560 static inline struct net_device *first_net_device_rcu(struct net *net)
1562 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1564 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1567 extern int netdev_boot_setup_check(struct net_device *dev);
1568 extern unsigned long netdev_boot_base(const char *prefix, int unit);
1569 extern struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1570 const char *hwaddr);
1571 extern struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1572 extern struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1573 extern void dev_add_pack(struct packet_type *pt);
1574 extern void dev_remove_pack(struct packet_type *pt);
1575 extern void __dev_remove_pack(struct packet_type *pt);
1577 extern struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1578 unsigned short mask);
1579 extern struct net_device *dev_get_by_name(struct net *net, const char *name);
1580 extern struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1581 extern struct net_device *__dev_get_by_name(struct net *net, const char *name);
1582 extern int dev_alloc_name(struct net_device *dev, const char *name);
1583 extern int dev_open(struct net_device *dev);
1584 extern int dev_close(struct net_device *dev);
1585 extern void dev_disable_lro(struct net_device *dev);
1586 extern int dev_queue_xmit(struct sk_buff *skb);
1587 extern int register_netdevice(struct net_device *dev);
1588 extern void unregister_netdevice_queue(struct net_device *dev,
1589 struct list_head *head);
1590 extern void unregister_netdevice_many(struct list_head *head);
1591 static inline void unregister_netdevice(struct net_device *dev)
1593 unregister_netdevice_queue(dev, NULL);
1596 extern int netdev_refcnt_read(const struct net_device *dev);
1597 extern void free_netdev(struct net_device *dev);
1598 extern void synchronize_net(void);
1599 extern int init_dummy_netdev(struct net_device *dev);
1600 extern void netdev_resync_ops(struct net_device *dev);
1602 extern struct net_device *dev_get_by_index(struct net *net, int ifindex);
1603 extern struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1604 extern struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1605 extern int dev_restart(struct net_device *dev);
1606 #ifdef CONFIG_NETPOLL_TRAP
1607 extern int netpoll_trap(void);
1609 extern int skb_gro_receive(struct sk_buff **head,
1610 struct sk_buff *skb);
1611 extern void skb_gro_reset_offset(struct sk_buff *skb);
1613 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1615 return NAPI_GRO_CB(skb)->data_offset;
1618 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1620 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1623 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1625 NAPI_GRO_CB(skb)->data_offset += len;
1628 static inline void *skb_gro_header_fast(struct sk_buff *skb,
1629 unsigned int offset)
1631 return NAPI_GRO_CB(skb)->frag0 + offset;
1634 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1636 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1639 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1640 unsigned int offset)
1642 if (!pskb_may_pull(skb, hlen))
1645 NAPI_GRO_CB(skb)->frag0 = NULL;
1646 NAPI_GRO_CB(skb)->frag0_len = 0;
1647 return skb->data + offset;
1650 static inline void *skb_gro_mac_header(struct sk_buff *skb)
1652 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
1655 static inline void *skb_gro_network_header(struct sk_buff *skb)
1657 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1658 skb_network_offset(skb);
1661 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1662 unsigned short type,
1663 const void *daddr, const void *saddr,
1666 if (!dev->header_ops || !dev->header_ops->create)
1669 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
1672 static inline int dev_parse_header(const struct sk_buff *skb,
1673 unsigned char *haddr)
1675 const struct net_device *dev = skb->dev;
1677 if (!dev->header_ops || !dev->header_ops->parse)
1679 return dev->header_ops->parse(skb, haddr);
1682 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
1683 extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf);
1684 static inline int unregister_gifconf(unsigned int family)
1686 return register_gifconf(family, NULL);
1690 * Incoming packets are placed on per-cpu queues
1692 struct softnet_data {
1693 struct Qdisc *output_queue;
1694 struct Qdisc **output_queue_tailp;
1695 struct list_head poll_list;
1696 struct sk_buff *completion_queue;
1697 struct sk_buff_head process_queue;
1700 unsigned int processed;
1701 unsigned int time_squeeze;
1702 unsigned int cpu_collision;
1703 unsigned int received_rps;
1706 struct softnet_data *rps_ipi_list;
1708 /* Elements below can be accessed between CPUs for RPS */
1709 struct call_single_data csd ____cacheline_aligned_in_smp;
1710 struct softnet_data *rps_ipi_next;
1712 unsigned int input_queue_head;
1713 unsigned int input_queue_tail;
1716 struct sk_buff_head input_pkt_queue;
1717 struct napi_struct backlog;
1720 static inline void input_queue_head_incr(struct softnet_data *sd)
1723 sd->input_queue_head++;
1727 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1728 unsigned int *qtail)
1731 *qtail = ++sd->input_queue_tail;
1735 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1737 extern void __netif_schedule(struct Qdisc *q);
1739 static inline void netif_schedule_queue(struct netdev_queue *txq)
1741 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
1742 __netif_schedule(txq->qdisc);
1745 static inline void netif_tx_schedule_all(struct net_device *dev)
1749 for (i = 0; i < dev->num_tx_queues; i++)
1750 netif_schedule_queue(netdev_get_tx_queue(dev, i));
1753 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
1755 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1759 * netif_start_queue - allow transmit
1760 * @dev: network device
1762 * Allow upper layers to call the device hard_start_xmit routine.
1764 static inline void netif_start_queue(struct net_device *dev)
1766 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1769 static inline void netif_tx_start_all_queues(struct net_device *dev)
1773 for (i = 0; i < dev->num_tx_queues; i++) {
1774 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1775 netif_tx_start_queue(txq);
1779 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1781 #ifdef CONFIG_NETPOLL_TRAP
1782 if (netpoll_trap()) {
1783 netif_tx_start_queue(dev_queue);
1787 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
1788 __netif_schedule(dev_queue->qdisc);
1792 * netif_wake_queue - restart transmit
1793 * @dev: network device
1795 * Allow upper layers to call the device hard_start_xmit routine.
1796 * Used for flow control when transmit resources are available.
1798 static inline void netif_wake_queue(struct net_device *dev)
1800 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1803 static inline void netif_tx_wake_all_queues(struct net_device *dev)
1807 for (i = 0; i < dev->num_tx_queues; i++) {
1808 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1809 netif_tx_wake_queue(txq);
1813 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
1815 if (WARN_ON(!dev_queue)) {
1816 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
1819 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1823 * netif_stop_queue - stop transmitted packets
1824 * @dev: network device
1826 * Stop upper layers calling the device hard_start_xmit routine.
1827 * Used for flow control when transmit resources are unavailable.
1829 static inline void netif_stop_queue(struct net_device *dev)
1831 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1834 static inline void netif_tx_stop_all_queues(struct net_device *dev)
1838 for (i = 0; i < dev->num_tx_queues; i++) {
1839 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1840 netif_tx_stop_queue(txq);
1844 static inline int netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
1846 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1850 * netif_queue_stopped - test if transmit queue is flowblocked
1851 * @dev: network device
1853 * Test if transmit queue on device is currently unable to send.
1855 static inline int netif_queue_stopped(const struct net_device *dev)
1857 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1860 static inline int netif_xmit_stopped(const struct netdev_queue *dev_queue)
1862 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
1865 static inline int netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
1867 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
1870 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
1874 dql_queued(&dev_queue->dql, bytes);
1875 if (unlikely(dql_avail(&dev_queue->dql) < 0)) {
1876 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
1877 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
1878 clear_bit(__QUEUE_STATE_STACK_XOFF,
1884 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
1886 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
1889 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
1890 unsigned pkts, unsigned bytes)
1893 if (likely(bytes)) {
1894 dql_completed(&dev_queue->dql, bytes);
1895 if (unlikely(test_bit(__QUEUE_STATE_STACK_XOFF,
1896 &dev_queue->state) &&
1897 dql_avail(&dev_queue->dql) >= 0)) {
1898 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF,
1900 netif_schedule_queue(dev_queue);
1906 static inline void netdev_completed_queue(struct net_device *dev,
1907 unsigned pkts, unsigned bytes)
1909 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
1912 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
1919 static inline void netdev_reset_queue(struct net_device *dev_queue)
1921 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
1925 * netif_running - test if up
1926 * @dev: network device
1928 * Test if the device has been brought up.
1930 static inline int netif_running(const struct net_device *dev)
1932 return test_bit(__LINK_STATE_START, &dev->state);
1936 * Routines to manage the subqueues on a device. We only need start
1937 * stop, and a check if it's stopped. All other device management is
1938 * done at the overall netdevice level.
1939 * Also test the device if we're multiqueue.
1943 * netif_start_subqueue - allow sending packets on subqueue
1944 * @dev: network device
1945 * @queue_index: sub queue index
1947 * Start individual transmit queue of a device with multiple transmit queues.
1949 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
1951 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
1953 netif_tx_start_queue(txq);
1957 * netif_stop_subqueue - stop sending packets on subqueue
1958 * @dev: network device
1959 * @queue_index: sub queue index
1961 * Stop individual transmit queue of a device with multiple transmit queues.
1963 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
1965 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
1966 #ifdef CONFIG_NETPOLL_TRAP
1970 netif_tx_stop_queue(txq);
1974 * netif_subqueue_stopped - test status of subqueue
1975 * @dev: network device
1976 * @queue_index: sub queue index
1978 * Check individual transmit queue of a device with multiple transmit queues.
1980 static inline int __netif_subqueue_stopped(const struct net_device *dev,
1983 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
1985 return netif_tx_queue_stopped(txq);
1988 static inline int netif_subqueue_stopped(const struct net_device *dev,
1989 struct sk_buff *skb)
1991 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
1995 * netif_wake_subqueue - allow sending packets on subqueue
1996 * @dev: network device
1997 * @queue_index: sub queue index
1999 * Resume individual transmit queue of a device with multiple transmit queues.
2001 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2003 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2004 #ifdef CONFIG_NETPOLL_TRAP
2008 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2009 __netif_schedule(txq->qdisc);
2013 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2014 * as a distribution range limit for the returned value.
2016 static inline u16 skb_tx_hash(const struct net_device *dev,
2017 const struct sk_buff *skb)
2019 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2023 * netif_is_multiqueue - test if device has multiple transmit queues
2024 * @dev: network device
2026 * Check if device has multiple transmit queues
2028 static inline int netif_is_multiqueue(const struct net_device *dev)
2030 return dev->num_tx_queues > 1;
2033 extern int netif_set_real_num_tx_queues(struct net_device *dev,
2037 extern int netif_set_real_num_rx_queues(struct net_device *dev,
2040 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2047 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2048 const struct net_device *from_dev)
2050 netif_set_real_num_tx_queues(to_dev, from_dev->real_num_tx_queues);
2052 return netif_set_real_num_rx_queues(to_dev,
2053 from_dev->real_num_rx_queues);
2059 /* Use this variant when it is known for sure that it
2060 * is executing from hardware interrupt context or with hardware interrupts
2063 extern void dev_kfree_skb_irq(struct sk_buff *skb);
2065 /* Use this variant in places where it could be invoked
2066 * from either hardware interrupt or other context, with hardware interrupts
2067 * either disabled or enabled.
2069 extern void dev_kfree_skb_any(struct sk_buff *skb);
2071 extern int netif_rx(struct sk_buff *skb);
2072 extern int netif_rx_ni(struct sk_buff *skb);
2073 extern int netif_receive_skb(struct sk_buff *skb);
2074 extern gro_result_t dev_gro_receive(struct napi_struct *napi,
2075 struct sk_buff *skb);
2076 extern gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb);
2077 extern gro_result_t napi_gro_receive(struct napi_struct *napi,
2078 struct sk_buff *skb);
2079 extern void napi_gro_flush(struct napi_struct *napi);
2080 extern struct sk_buff * napi_get_frags(struct napi_struct *napi);
2081 extern gro_result_t napi_frags_finish(struct napi_struct *napi,
2082 struct sk_buff *skb,
2084 extern struct sk_buff * napi_frags_skb(struct napi_struct *napi);
2085 extern gro_result_t napi_gro_frags(struct napi_struct *napi);
2087 static inline void napi_free_frags(struct napi_struct *napi)
2089 kfree_skb(napi->skb);
2093 extern int netdev_rx_handler_register(struct net_device *dev,
2094 rx_handler_func_t *rx_handler,
2095 void *rx_handler_data);
2096 extern void netdev_rx_handler_unregister(struct net_device *dev);
2098 extern int dev_valid_name(const char *name);
2099 extern int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2100 extern int dev_ethtool(struct net *net, struct ifreq *);
2101 extern unsigned dev_get_flags(const struct net_device *);
2102 extern int __dev_change_flags(struct net_device *, unsigned int flags);
2103 extern int dev_change_flags(struct net_device *, unsigned);
2104 extern void __dev_notify_flags(struct net_device *, unsigned int old_flags);
2105 extern int dev_change_name(struct net_device *, const char *);
2106 extern int dev_set_alias(struct net_device *, const char *, size_t);
2107 extern int dev_change_net_namespace(struct net_device *,
2108 struct net *, const char *);
2109 extern int dev_set_mtu(struct net_device *, int);
2110 extern void dev_set_group(struct net_device *, int);
2111 extern int dev_set_mac_address(struct net_device *,
2113 extern int dev_hard_start_xmit(struct sk_buff *skb,
2114 struct net_device *dev,
2115 struct netdev_queue *txq);
2116 extern int dev_forward_skb(struct net_device *dev,
2117 struct sk_buff *skb);
2119 extern int netdev_budget;
2121 /* Called by rtnetlink.c:rtnl_unlock() */
2122 extern void netdev_run_todo(void);
2125 * dev_put - release reference to device
2126 * @dev: network device
2128 * Release reference to device to allow it to be freed.
2130 static inline void dev_put(struct net_device *dev)
2132 irqsafe_cpu_dec(*dev->pcpu_refcnt);
2136 * dev_hold - get reference to device
2137 * @dev: network device
2139 * Hold reference to device to keep it from being freed.
2141 static inline void dev_hold(struct net_device *dev)
2143 irqsafe_cpu_inc(*dev->pcpu_refcnt);
2146 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2147 * and _off may be called from IRQ context, but it is caller
2148 * who is responsible for serialization of these calls.
2150 * The name carrier is inappropriate, these functions should really be
2151 * called netif_lowerlayer_*() because they represent the state of any
2152 * kind of lower layer not just hardware media.
2155 extern void linkwatch_fire_event(struct net_device *dev);
2156 extern void linkwatch_forget_dev(struct net_device *dev);
2159 * netif_carrier_ok - test if carrier present
2160 * @dev: network device
2162 * Check if carrier is present on device
2164 static inline int netif_carrier_ok(const struct net_device *dev)
2166 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2169 extern unsigned long dev_trans_start(struct net_device *dev);
2171 extern void __netdev_watchdog_up(struct net_device *dev);
2173 extern void netif_carrier_on(struct net_device *dev);
2175 extern void netif_carrier_off(struct net_device *dev);
2177 extern void netif_notify_peers(struct net_device *dev);
2180 * netif_dormant_on - mark device as dormant.
2181 * @dev: network device
2183 * Mark device as dormant (as per RFC2863).
2185 * The dormant state indicates that the relevant interface is not
2186 * actually in a condition to pass packets (i.e., it is not 'up') but is
2187 * in a "pending" state, waiting for some external event. For "on-
2188 * demand" interfaces, this new state identifies the situation where the
2189 * interface is waiting for events to place it in the up state.
2192 static inline void netif_dormant_on(struct net_device *dev)
2194 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2195 linkwatch_fire_event(dev);
2199 * netif_dormant_off - set device as not dormant.
2200 * @dev: network device
2202 * Device is not in dormant state.
2204 static inline void netif_dormant_off(struct net_device *dev)
2206 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2207 linkwatch_fire_event(dev);
2211 * netif_dormant - test if carrier present
2212 * @dev: network device
2214 * Check if carrier is present on device
2216 static inline int netif_dormant(const struct net_device *dev)
2218 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2223 * netif_oper_up - test if device is operational
2224 * @dev: network device
2226 * Check if carrier is operational
2228 static inline int netif_oper_up(const struct net_device *dev)
2230 return (dev->operstate == IF_OPER_UP ||
2231 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2235 * netif_device_present - is device available or removed
2236 * @dev: network device
2238 * Check if device has not been removed from system.
2240 static inline int netif_device_present(struct net_device *dev)
2242 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2245 extern void netif_device_detach(struct net_device *dev);
2247 extern void netif_device_attach(struct net_device *dev);
2250 * Network interface message level settings
2254 NETIF_MSG_DRV = 0x0001,
2255 NETIF_MSG_PROBE = 0x0002,
2256 NETIF_MSG_LINK = 0x0004,
2257 NETIF_MSG_TIMER = 0x0008,
2258 NETIF_MSG_IFDOWN = 0x0010,
2259 NETIF_MSG_IFUP = 0x0020,
2260 NETIF_MSG_RX_ERR = 0x0040,
2261 NETIF_MSG_TX_ERR = 0x0080,
2262 NETIF_MSG_TX_QUEUED = 0x0100,
2263 NETIF_MSG_INTR = 0x0200,
2264 NETIF_MSG_TX_DONE = 0x0400,
2265 NETIF_MSG_RX_STATUS = 0x0800,
2266 NETIF_MSG_PKTDATA = 0x1000,
2267 NETIF_MSG_HW = 0x2000,
2268 NETIF_MSG_WOL = 0x4000,
2271 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2272 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2273 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2274 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2275 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2276 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2277 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2278 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2279 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2280 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2281 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2282 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2283 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2284 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2285 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2287 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2290 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2291 return default_msg_enable_bits;
2292 if (debug_value == 0) /* no output */
2294 /* set low N bits */
2295 return (1 << debug_value) - 1;
2298 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2300 spin_lock(&txq->_xmit_lock);
2301 txq->xmit_lock_owner = cpu;
2304 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2306 spin_lock_bh(&txq->_xmit_lock);
2307 txq->xmit_lock_owner = smp_processor_id();
2310 static inline int __netif_tx_trylock(struct netdev_queue *txq)
2312 int ok = spin_trylock(&txq->_xmit_lock);
2314 txq->xmit_lock_owner = smp_processor_id();
2318 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2320 txq->xmit_lock_owner = -1;
2321 spin_unlock(&txq->_xmit_lock);
2324 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2326 txq->xmit_lock_owner = -1;
2327 spin_unlock_bh(&txq->_xmit_lock);
2330 static inline void txq_trans_update(struct netdev_queue *txq)
2332 if (txq->xmit_lock_owner != -1)
2333 txq->trans_start = jiffies;
2337 * netif_tx_lock - grab network device transmit lock
2338 * @dev: network device
2340 * Get network device transmit lock
2342 static inline void netif_tx_lock(struct net_device *dev)
2347 spin_lock(&dev->tx_global_lock);
2348 cpu = smp_processor_id();
2349 for (i = 0; i < dev->num_tx_queues; i++) {
2350 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2352 /* We are the only thread of execution doing a
2353 * freeze, but we have to grab the _xmit_lock in
2354 * order to synchronize with threads which are in
2355 * the ->hard_start_xmit() handler and already
2356 * checked the frozen bit.
2358 __netif_tx_lock(txq, cpu);
2359 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2360 __netif_tx_unlock(txq);
2364 static inline void netif_tx_lock_bh(struct net_device *dev)
2370 static inline void netif_tx_unlock(struct net_device *dev)
2374 for (i = 0; i < dev->num_tx_queues; i++) {
2375 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2377 /* No need to grab the _xmit_lock here. If the
2378 * queue is not stopped for another reason, we
2381 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2382 netif_schedule_queue(txq);
2384 spin_unlock(&dev->tx_global_lock);
2387 static inline void netif_tx_unlock_bh(struct net_device *dev)
2389 netif_tx_unlock(dev);
2393 #define HARD_TX_LOCK(dev, txq, cpu) { \
2394 if ((dev->features & NETIF_F_LLTX) == 0) { \
2395 __netif_tx_lock(txq, cpu); \
2399 #define HARD_TX_UNLOCK(dev, txq) { \
2400 if ((dev->features & NETIF_F_LLTX) == 0) { \
2401 __netif_tx_unlock(txq); \
2405 static inline void netif_tx_disable(struct net_device *dev)
2411 cpu = smp_processor_id();
2412 for (i = 0; i < dev->num_tx_queues; i++) {
2413 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2415 __netif_tx_lock(txq, cpu);
2416 netif_tx_stop_queue(txq);
2417 __netif_tx_unlock(txq);
2422 static inline void netif_addr_lock(struct net_device *dev)
2424 spin_lock(&dev->addr_list_lock);
2427 static inline void netif_addr_lock_bh(struct net_device *dev)
2429 spin_lock_bh(&dev->addr_list_lock);
2432 static inline void netif_addr_unlock(struct net_device *dev)
2434 spin_unlock(&dev->addr_list_lock);
2437 static inline void netif_addr_unlock_bh(struct net_device *dev)
2439 spin_unlock_bh(&dev->addr_list_lock);
2443 * dev_addrs walker. Should be used only for read access. Call with
2444 * rcu_read_lock held.
2446 #define for_each_dev_addr(dev, ha) \
2447 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2449 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2451 extern void ether_setup(struct net_device *dev);
2453 /* Support for loadable net-drivers */
2454 extern struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2455 void (*setup)(struct net_device *),
2456 unsigned int txqs, unsigned int rxqs);
2457 #define alloc_netdev(sizeof_priv, name, setup) \
2458 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2460 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2461 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2463 extern int register_netdev(struct net_device *dev);
2464 extern void unregister_netdev(struct net_device *dev);
2466 /* General hardware address lists handling functions */
2467 extern int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2468 struct netdev_hw_addr_list *from_list,
2469 int addr_len, unsigned char addr_type);
2470 extern void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2471 struct netdev_hw_addr_list *from_list,
2472 int addr_len, unsigned char addr_type);
2473 extern int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2474 struct netdev_hw_addr_list *from_list,
2476 extern void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2477 struct netdev_hw_addr_list *from_list,
2479 extern void __hw_addr_flush(struct netdev_hw_addr_list *list);
2480 extern void __hw_addr_init(struct netdev_hw_addr_list *list);
2482 /* Functions used for device addresses handling */
2483 extern int dev_addr_add(struct net_device *dev, unsigned char *addr,
2484 unsigned char addr_type);
2485 extern int dev_addr_del(struct net_device *dev, unsigned char *addr,
2486 unsigned char addr_type);
2487 extern int dev_addr_add_multiple(struct net_device *to_dev,
2488 struct net_device *from_dev,
2489 unsigned char addr_type);
2490 extern int dev_addr_del_multiple(struct net_device *to_dev,
2491 struct net_device *from_dev,
2492 unsigned char addr_type);
2493 extern void dev_addr_flush(struct net_device *dev);
2494 extern int dev_addr_init(struct net_device *dev);
2496 /* Functions used for unicast addresses handling */
2497 extern int dev_uc_add(struct net_device *dev, unsigned char *addr);
2498 extern int dev_uc_del(struct net_device *dev, unsigned char *addr);
2499 extern int dev_uc_sync(struct net_device *to, struct net_device *from);
2500 extern void dev_uc_unsync(struct net_device *to, struct net_device *from);
2501 extern void dev_uc_flush(struct net_device *dev);
2502 extern void dev_uc_init(struct net_device *dev);
2504 /* Functions used for multicast addresses handling */
2505 extern int dev_mc_add(struct net_device *dev, unsigned char *addr);
2506 extern int dev_mc_add_global(struct net_device *dev, unsigned char *addr);
2507 extern int dev_mc_del(struct net_device *dev, unsigned char *addr);
2508 extern int dev_mc_del_global(struct net_device *dev, unsigned char *addr);
2509 extern int dev_mc_sync(struct net_device *to, struct net_device *from);
2510 extern void dev_mc_unsync(struct net_device *to, struct net_device *from);
2511 extern void dev_mc_flush(struct net_device *dev);
2512 extern void dev_mc_init(struct net_device *dev);
2514 /* Functions used for secondary unicast and multicast support */
2515 extern void dev_set_rx_mode(struct net_device *dev);
2516 extern void __dev_set_rx_mode(struct net_device *dev);
2517 extern int dev_set_promiscuity(struct net_device *dev, int inc);
2518 extern int dev_set_allmulti(struct net_device *dev, int inc);
2519 extern void netdev_state_change(struct net_device *dev);
2520 extern int netdev_bonding_change(struct net_device *dev,
2521 unsigned long event);
2522 extern void netdev_features_change(struct net_device *dev);
2523 /* Load a device via the kmod */
2524 extern void dev_load(struct net *net, const char *name);
2525 extern void dev_mcast_init(void);
2526 extern struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2527 struct rtnl_link_stats64 *storage);
2529 extern int netdev_max_backlog;
2530 extern int netdev_tstamp_prequeue;
2531 extern int weight_p;
2532 extern int bpf_jit_enable;
2533 extern int netdev_set_master(struct net_device *dev, struct net_device *master);
2534 extern int netdev_set_bond_master(struct net_device *dev,
2535 struct net_device *master);
2536 extern int skb_checksum_help(struct sk_buff *skb);
2537 extern struct sk_buff *skb_gso_segment(struct sk_buff *skb,
2538 netdev_features_t features);
2540 extern void netdev_rx_csum_fault(struct net_device *dev);
2542 static inline void netdev_rx_csum_fault(struct net_device *dev)
2546 /* rx skb timestamps */
2547 extern void net_enable_timestamp(void);
2548 extern void net_disable_timestamp(void);
2550 #ifdef CONFIG_PROC_FS
2551 extern void *dev_seq_start(struct seq_file *seq, loff_t *pos);
2552 extern void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos);
2553 extern void dev_seq_stop(struct seq_file *seq, void *v);
2554 extern int dev_seq_open_ops(struct inode *inode, struct file *file,
2555 const struct seq_operations *ops);
2558 extern int netdev_class_create_file(struct class_attribute *class_attr);
2559 extern void netdev_class_remove_file(struct class_attribute *class_attr);
2561 extern struct kobj_ns_type_operations net_ns_type_operations;
2563 extern const char *netdev_drivername(const struct net_device *dev);
2565 extern void linkwatch_run_queue(void);
2567 static inline netdev_features_t netdev_get_wanted_features(
2568 struct net_device *dev)
2570 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2572 netdev_features_t netdev_increment_features(netdev_features_t all,
2573 netdev_features_t one, netdev_features_t mask);
2574 int __netdev_update_features(struct net_device *dev);
2575 void netdev_update_features(struct net_device *dev);
2576 void netdev_change_features(struct net_device *dev);
2578 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
2579 struct net_device *dev);
2581 netdev_features_t netif_skb_features(struct sk_buff *skb);
2583 static inline int net_gso_ok(netdev_features_t features, int gso_type)
2585 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
2587 /* check flags correspondence */
2588 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
2589 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
2590 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
2591 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
2592 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
2593 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
2595 return (features & feature) == feature;
2598 static inline int skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
2600 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
2601 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
2604 static inline int netif_needs_gso(struct sk_buff *skb,
2605 netdev_features_t features)
2607 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
2608 unlikely(skb->ip_summed != CHECKSUM_PARTIAL));
2611 static inline void netif_set_gso_max_size(struct net_device *dev,
2614 dev->gso_max_size = size;
2617 static inline int netif_is_bond_slave(struct net_device *dev)
2619 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
2622 extern struct pernet_operations __net_initdata loopback_net_ops;
2624 /* Logging, debugging and troubleshooting/diagnostic helpers. */
2626 /* netdev_printk helpers, similar to dev_printk */
2628 static inline const char *netdev_name(const struct net_device *dev)
2630 if (dev->reg_state != NETREG_REGISTERED)
2631 return "(unregistered net_device)";
2635 extern int __netdev_printk(const char *level, const struct net_device *dev,
2636 struct va_format *vaf);
2638 extern __printf(3, 4)
2639 int netdev_printk(const char *level, const struct net_device *dev,
2640 const char *format, ...);
2641 extern __printf(2, 3)
2642 int netdev_emerg(const struct net_device *dev, const char *format, ...);
2643 extern __printf(2, 3)
2644 int netdev_alert(const struct net_device *dev, const char *format, ...);
2645 extern __printf(2, 3)
2646 int netdev_crit(const struct net_device *dev, const char *format, ...);
2647 extern __printf(2, 3)
2648 int netdev_err(const struct net_device *dev, const char *format, ...);
2649 extern __printf(2, 3)
2650 int netdev_warn(const struct net_device *dev, const char *format, ...);
2651 extern __printf(2, 3)
2652 int netdev_notice(const struct net_device *dev, const char *format, ...);
2653 extern __printf(2, 3)
2654 int netdev_info(const struct net_device *dev, const char *format, ...);
2656 #define MODULE_ALIAS_NETDEV(device) \
2657 MODULE_ALIAS("netdev-" device)
2660 #define netdev_dbg(__dev, format, args...) \
2661 netdev_printk(KERN_DEBUG, __dev, format, ##args)
2662 #elif defined(CONFIG_DYNAMIC_DEBUG)
2663 #define netdev_dbg(__dev, format, args...) \
2665 dynamic_netdev_dbg(__dev, format, ##args); \
2668 #define netdev_dbg(__dev, format, args...) \
2671 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
2676 #if defined(VERBOSE_DEBUG)
2677 #define netdev_vdbg netdev_dbg
2680 #define netdev_vdbg(dev, format, args...) \
2683 netdev_printk(KERN_DEBUG, dev, format, ##args); \
2689 * netdev_WARN() acts like dev_printk(), but with the key difference
2690 * of using a WARN/WARN_ON to get the message out, including the
2691 * file/line information and a backtrace.
2693 #define netdev_WARN(dev, format, args...) \
2694 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args);
2696 /* netif printk helpers, similar to netdev_printk */
2698 #define netif_printk(priv, type, level, dev, fmt, args...) \
2700 if (netif_msg_##type(priv)) \
2701 netdev_printk(level, (dev), fmt, ##args); \
2704 #define netif_level(level, priv, type, dev, fmt, args...) \
2706 if (netif_msg_##type(priv)) \
2707 netdev_##level(dev, fmt, ##args); \
2710 #define netif_emerg(priv, type, dev, fmt, args...) \
2711 netif_level(emerg, priv, type, dev, fmt, ##args)
2712 #define netif_alert(priv, type, dev, fmt, args...) \
2713 netif_level(alert, priv, type, dev, fmt, ##args)
2714 #define netif_crit(priv, type, dev, fmt, args...) \
2715 netif_level(crit, priv, type, dev, fmt, ##args)
2716 #define netif_err(priv, type, dev, fmt, args...) \
2717 netif_level(err, priv, type, dev, fmt, ##args)
2718 #define netif_warn(priv, type, dev, fmt, args...) \
2719 netif_level(warn, priv, type, dev, fmt, ##args)
2720 #define netif_notice(priv, type, dev, fmt, args...) \
2721 netif_level(notice, priv, type, dev, fmt, ##args)
2722 #define netif_info(priv, type, dev, fmt, args...) \
2723 netif_level(info, priv, type, dev, fmt, ##args)
2726 #define netif_dbg(priv, type, dev, format, args...) \
2727 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
2728 #elif defined(CONFIG_DYNAMIC_DEBUG)
2729 #define netif_dbg(priv, type, netdev, format, args...) \
2731 if (netif_msg_##type(priv)) \
2732 dynamic_netdev_dbg(netdev, format, ##args); \
2735 #define netif_dbg(priv, type, dev, format, args...) \
2738 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2743 #if defined(VERBOSE_DEBUG)
2744 #define netif_vdbg netif_dbg
2746 #define netif_vdbg(priv, type, dev, format, args...) \
2749 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2754 #endif /* __KERNEL__ */
2756 #endif /* _LINUX_NETDEVICE_H */