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>
63 /* source back-compat hooks */
64 #define SET_ETHTOOL_OPS(netdev,ops) \
65 ( (netdev)->ethtool_ops = (ops) )
67 /* hardware address assignment types */
68 #define NET_ADDR_PERM 0 /* address is permanent (default) */
69 #define NET_ADDR_RANDOM 1 /* address is generated randomly */
70 #define NET_ADDR_STOLEN 2 /* address is stolen from other device */
72 /* Backlog congestion levels */
73 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74 #define NET_RX_DROP 1 /* packet dropped */
77 * Transmit return codes: transmit return codes originate from three different
80 * - qdisc return codes
81 * - driver transmit return codes
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
88 * higher layers. Virtual network devices transmit synchronously, in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), all
90 * others are propagated to higher layers.
93 /* qdisc ->enqueue() return codes. */
94 #define NET_XMIT_SUCCESS 0x00
95 #define NET_XMIT_DROP 0x01 /* skb dropped */
96 #define NET_XMIT_CN 0x02 /* congestion notification */
97 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
98 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
103 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
104 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106 /* Driver transmit return codes */
107 #define NETDEV_TX_MASK 0xf0
110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
113 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
115 typedef enum netdev_tx netdev_tx_t;
118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121 static inline bool dev_xmit_complete(int rc)
124 * Positive cases with an skb consumed by a driver:
125 * - successful transmission (rc == NETDEV_TX_OK)
126 * - error while transmitting (rc < 0)
127 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 if (likely(rc < NET_XMIT_MASK))
137 #define MAX_ADDR_LEN 32 /* Largest hardware address length */
139 /* Initial net device group. All devices belong to group 0 by default. */
140 #define INIT_NETDEV_GROUP 0
144 * Compute the worst case header length according to the protocols
148 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
149 # if defined(CONFIG_MAC80211_MESH)
150 # define LL_MAX_HEADER 128
152 # define LL_MAX_HEADER 96
154 #elif IS_ENABLED(CONFIG_TR)
155 # define LL_MAX_HEADER 48
157 # define LL_MAX_HEADER 32
160 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
161 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
162 #define MAX_HEADER LL_MAX_HEADER
164 #define MAX_HEADER (LL_MAX_HEADER + 48)
168 * Old network device statistics. Fields are native words
169 * (unsigned long) so they can be read and written atomically.
172 struct net_device_stats {
173 unsigned long rx_packets;
174 unsigned long tx_packets;
175 unsigned long rx_bytes;
176 unsigned long tx_bytes;
177 unsigned long rx_errors;
178 unsigned long tx_errors;
179 unsigned long rx_dropped;
180 unsigned long tx_dropped;
181 unsigned long multicast;
182 unsigned long collisions;
183 unsigned long rx_length_errors;
184 unsigned long rx_over_errors;
185 unsigned long rx_crc_errors;
186 unsigned long rx_frame_errors;
187 unsigned long rx_fifo_errors;
188 unsigned long rx_missed_errors;
189 unsigned long tx_aborted_errors;
190 unsigned long tx_carrier_errors;
191 unsigned long tx_fifo_errors;
192 unsigned long tx_heartbeat_errors;
193 unsigned long tx_window_errors;
194 unsigned long rx_compressed;
195 unsigned long tx_compressed;
198 #endif /* __KERNEL__ */
201 /* Media selection options. */
214 #include <linux/cache.h>
215 #include <linux/skbuff.h>
218 #include <linux/jump_label.h>
219 extern struct jump_label_key rps_needed;
226 struct netdev_hw_addr {
227 struct list_head list;
228 unsigned char addr[MAX_ADDR_LEN];
230 #define NETDEV_HW_ADDR_T_LAN 1
231 #define NETDEV_HW_ADDR_T_SAN 2
232 #define NETDEV_HW_ADDR_T_SLAVE 3
233 #define NETDEV_HW_ADDR_T_UNICAST 4
234 #define NETDEV_HW_ADDR_T_MULTICAST 5
238 struct rcu_head rcu_head;
241 struct netdev_hw_addr_list {
242 struct list_head list;
246 #define netdev_hw_addr_list_count(l) ((l)->count)
247 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
248 #define netdev_hw_addr_list_for_each(ha, l) \
249 list_for_each_entry(ha, &(l)->list, list)
251 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
252 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
253 #define netdev_for_each_uc_addr(ha, dev) \
254 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
256 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
257 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
258 #define netdev_for_each_mc_addr(ha, dev) \
259 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
266 /* cached hardware header; allow for machine alignment needs. */
267 #define HH_DATA_MOD 16
268 #define HH_DATA_OFF(__len) \
269 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
270 #define HH_DATA_ALIGN(__len) \
271 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
272 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
275 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
277 * dev->hard_header_len ? (dev->hard_header_len +
278 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
280 * We could use other alignment values, but we must maintain the
281 * relationship HH alignment <= LL alignment.
283 #define LL_RESERVED_SPACE(dev) \
284 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
285 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
286 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
289 int (*create) (struct sk_buff *skb, struct net_device *dev,
290 unsigned short type, const void *daddr,
291 const void *saddr, unsigned len);
292 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
293 int (*rebuild)(struct sk_buff *skb);
294 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
295 void (*cache_update)(struct hh_cache *hh,
296 const struct net_device *dev,
297 const unsigned char *haddr);
300 /* These flag bits are private to the generic network queueing
301 * layer, they may not be explicitly referenced by any other
305 enum netdev_state_t {
307 __LINK_STATE_PRESENT,
308 __LINK_STATE_NOCARRIER,
309 __LINK_STATE_LINKWATCH_PENDING,
310 __LINK_STATE_DORMANT,
315 * This structure holds at boot time configured netdevice settings. They
316 * are then used in the device probing.
318 struct netdev_boot_setup {
322 #define NETDEV_BOOT_SETUP_MAX 8
324 extern int __init netdev_boot_setup(char *str);
327 * Structure for NAPI scheduling similar to tasklet but with weighting
330 /* The poll_list must only be managed by the entity which
331 * changes the state of the NAPI_STATE_SCHED bit. This means
332 * whoever atomically sets that bit can add this napi_struct
333 * to the per-cpu poll_list, and whoever clears that bit
334 * can remove from the list right before clearing the bit.
336 struct list_head poll_list;
340 int (*poll)(struct napi_struct *, int);
341 #ifdef CONFIG_NETPOLL
342 spinlock_t poll_lock;
346 unsigned int gro_count;
348 struct net_device *dev;
349 struct list_head dev_list;
350 struct sk_buff *gro_list;
355 NAPI_STATE_SCHED, /* Poll is scheduled */
356 NAPI_STATE_DISABLE, /* Disable pending */
357 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
367 typedef enum gro_result gro_result_t;
370 * enum rx_handler_result - Possible return values for rx_handlers.
371 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
373 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
374 * case skb->dev was changed by rx_handler.
375 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
376 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
378 * rx_handlers are functions called from inside __netif_receive_skb(), to do
379 * special processing of the skb, prior to delivery to protocol handlers.
381 * Currently, a net_device can only have a single rx_handler registered. Trying
382 * to register a second rx_handler will return -EBUSY.
384 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
385 * To unregister a rx_handler on a net_device, use
386 * netdev_rx_handler_unregister().
388 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
391 * If the rx_handler consumed to skb in some way, it should return
392 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
393 * the skb to be delivered in some other ways.
395 * If the rx_handler changed skb->dev, to divert the skb to another
396 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
397 * new device will be called if it exists.
399 * If the rx_handler consider the skb should be ignored, it should return
400 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
401 * are registred on exact device (ptype->dev == skb->dev).
403 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
404 * delivered, it should return RX_HANDLER_PASS.
406 * A device without a registered rx_handler will behave as if rx_handler
407 * returned RX_HANDLER_PASS.
410 enum rx_handler_result {
416 typedef enum rx_handler_result rx_handler_result_t;
417 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
419 extern void __napi_schedule(struct napi_struct *n);
421 static inline int napi_disable_pending(struct napi_struct *n)
423 return test_bit(NAPI_STATE_DISABLE, &n->state);
427 * napi_schedule_prep - check if napi can be scheduled
430 * Test if NAPI routine is already running, and if not mark
431 * it as running. This is used as a condition variable
432 * insure only one NAPI poll instance runs. We also make
433 * sure there is no pending NAPI disable.
435 static inline int napi_schedule_prep(struct napi_struct *n)
437 return !napi_disable_pending(n) &&
438 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
442 * napi_schedule - schedule NAPI poll
445 * Schedule NAPI poll routine to be called if it is not already
448 static inline void napi_schedule(struct napi_struct *n)
450 if (napi_schedule_prep(n))
454 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
455 static inline int napi_reschedule(struct napi_struct *napi)
457 if (napi_schedule_prep(napi)) {
458 __napi_schedule(napi);
465 * napi_complete - NAPI processing complete
468 * Mark NAPI processing as complete.
470 extern void __napi_complete(struct napi_struct *n);
471 extern void napi_complete(struct napi_struct *n);
474 * napi_disable - prevent NAPI from scheduling
477 * Stop NAPI from being scheduled on this context.
478 * Waits till any outstanding processing completes.
480 static inline void napi_disable(struct napi_struct *n)
482 set_bit(NAPI_STATE_DISABLE, &n->state);
483 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
485 clear_bit(NAPI_STATE_DISABLE, &n->state);
489 * napi_enable - enable NAPI scheduling
492 * Resume NAPI from being scheduled on this context.
493 * Must be paired with napi_disable.
495 static inline void napi_enable(struct napi_struct *n)
497 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
498 smp_mb__before_clear_bit();
499 clear_bit(NAPI_STATE_SCHED, &n->state);
504 * napi_synchronize - wait until NAPI is not running
507 * Wait until NAPI is done being scheduled on this context.
508 * Waits till any outstanding processing completes but
509 * does not disable future activations.
511 static inline void napi_synchronize(const struct napi_struct *n)
513 while (test_bit(NAPI_STATE_SCHED, &n->state))
517 # define napi_synchronize(n) barrier()
520 enum netdev_queue_state_t {
521 __QUEUE_STATE_DRV_XOFF,
522 __QUEUE_STATE_STACK_XOFF,
523 __QUEUE_STATE_FROZEN,
524 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
525 (1 << __QUEUE_STATE_STACK_XOFF))
526 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
527 (1 << __QUEUE_STATE_FROZEN))
530 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
531 * netif_tx_* functions below are used to manipulate this flag. The
532 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
533 * queue independently. The netif_xmit_*stopped functions below are called
534 * to check if the queue has been stopped by the driver or stack (either
535 * of the XOFF bits are set in the state). Drivers should not need to call
536 * netif_xmit*stopped functions, they should only be using netif_tx_*.
539 struct netdev_queue {
543 struct net_device *dev;
545 struct Qdisc *qdisc_sleeping;
549 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
555 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
558 * please use this field instead of dev->trans_start
560 unsigned long trans_start;
563 * Number of TX timeouts for this queue
564 * (/sys/class/net/DEV/Q/trans_timeout)
566 unsigned long trans_timeout;
573 } ____cacheline_aligned_in_smp;
575 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
577 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
584 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
586 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
593 * This structure holds an RPS map which can be of variable length. The
594 * map is an array of CPUs.
601 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + (_num * sizeof(u16)))
604 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
605 * tail pointer for that CPU's input queue at the time of last enqueue, and
606 * a hardware filter index.
608 struct rps_dev_flow {
611 unsigned int last_qtail;
613 #define RPS_NO_FILTER 0xffff
616 * The rps_dev_flow_table structure contains a table of flow mappings.
618 struct rps_dev_flow_table {
621 struct work_struct free_work;
622 struct rps_dev_flow flows[0];
624 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
625 (_num * sizeof(struct rps_dev_flow)))
628 * The rps_sock_flow_table contains mappings of flows to the last CPU
629 * on which they were processed by the application (set in recvmsg).
631 struct rps_sock_flow_table {
635 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
636 (_num * sizeof(u16)))
638 #define RPS_NO_CPU 0xffff
640 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
644 unsigned int cpu, index = hash & table->mask;
646 /* We only give a hint, preemption can change cpu under us */
647 cpu = raw_smp_processor_id();
649 if (table->ents[index] != cpu)
650 table->ents[index] = cpu;
654 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
658 table->ents[hash & table->mask] = RPS_NO_CPU;
661 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
663 #ifdef CONFIG_RFS_ACCEL
664 extern bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
665 u32 flow_id, u16 filter_id);
668 /* This structure contains an instance of an RX queue. */
669 struct netdev_rx_queue {
670 struct rps_map __rcu *rps_map;
671 struct rps_dev_flow_table __rcu *rps_flow_table;
673 struct net_device *dev;
674 } ____cacheline_aligned_in_smp;
675 #endif /* CONFIG_RPS */
679 * This structure holds an XPS map which can be of variable length. The
680 * map is an array of queues.
684 unsigned int alloc_len;
688 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + (_num * sizeof(u16)))
689 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
693 * This structure holds all XPS maps for device. Maps are indexed by CPU.
695 struct xps_dev_maps {
697 struct xps_map __rcu *cpu_map[0];
699 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
700 (nr_cpu_ids * sizeof(struct xps_map *)))
701 #endif /* CONFIG_XPS */
703 #define TC_MAX_QUEUE 16
704 #define TC_BITMASK 15
705 /* HW offloaded queuing disciplines txq count and offset maps */
706 struct netdev_tc_txq {
712 * This structure defines the management hooks for network devices.
713 * The following hooks can be defined; unless noted otherwise, they are
714 * optional and can be filled with a null pointer.
716 * int (*ndo_init)(struct net_device *dev);
717 * This function is called once when network device is registered.
718 * The network device can use this to any late stage initializaton
719 * or semantic validattion. It can fail with an error code which will
720 * be propogated back to register_netdev
722 * void (*ndo_uninit)(struct net_device *dev);
723 * This function is called when device is unregistered or when registration
724 * fails. It is not called if init fails.
726 * int (*ndo_open)(struct net_device *dev);
727 * This function is called when network device transistions to the up
730 * int (*ndo_stop)(struct net_device *dev);
731 * This function is called when network device transistions to the down
734 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
735 * struct net_device *dev);
736 * Called when a packet needs to be transmitted.
737 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
738 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
739 * Required can not be NULL.
741 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
742 * Called to decide which queue to when device supports multiple
745 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
746 * This function is called to allow device receiver to make
747 * changes to configuration when multicast or promiscious is enabled.
749 * void (*ndo_set_rx_mode)(struct net_device *dev);
750 * This function is called device changes address list filtering.
751 * If driver handles unicast address filtering, it should set
752 * IFF_UNICAST_FLT to its priv_flags.
754 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
755 * This function is called when the Media Access Control address
756 * needs to be changed. If this interface is not defined, the
757 * mac address can not be changed.
759 * int (*ndo_validate_addr)(struct net_device *dev);
760 * Test if Media Access Control address is valid for the device.
762 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
763 * Called when a user request an ioctl which can't be handled by
764 * the generic interface code. If not defined ioctl's return
765 * not supported error code.
767 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
768 * Used to set network devices bus interface parameters. This interface
769 * is retained for legacy reason, new devices should use the bus
770 * interface (PCI) for low level management.
772 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
773 * Called when a user wants to change the Maximum Transfer Unit
774 * of a device. If not defined, any request to change MTU will
775 * will return an error.
777 * void (*ndo_tx_timeout)(struct net_device *dev);
778 * Callback uses when the transmitter has not made any progress
779 * for dev->watchdog ticks.
781 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
782 * struct rtnl_link_stats64 *storage);
783 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
784 * Called when a user wants to get the network device usage
785 * statistics. Drivers must do one of the following:
786 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
787 * rtnl_link_stats64 structure passed by the caller.
788 * 2. Define @ndo_get_stats to update a net_device_stats structure
789 * (which should normally be dev->stats) and return a pointer to
790 * it. The structure may be changed asynchronously only if each
791 * field is written atomically.
792 * 3. Update dev->stats asynchronously and atomically, and define
795 * void (*ndo_vlan_rx_add_vid)(struct net_device *dev, unsigned short vid);
796 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER)
797 * this function is called when a VLAN id is registered.
799 * void (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
800 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER)
801 * this function is called when a VLAN id is unregistered.
803 * void (*ndo_poll_controller)(struct net_device *dev);
805 * SR-IOV management functions.
806 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
807 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
808 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
809 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
810 * int (*ndo_get_vf_config)(struct net_device *dev,
811 * int vf, struct ifla_vf_info *ivf);
812 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
813 * struct nlattr *port[]);
814 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
815 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
816 * Called to setup 'tc' number of traffic classes in the net device. This
817 * is always called from the stack with the rtnl lock held and netif tx
818 * queues stopped. This allows the netdevice to perform queue management
821 * Fiber Channel over Ethernet (FCoE) offload functions.
822 * int (*ndo_fcoe_enable)(struct net_device *dev);
823 * Called when the FCoE protocol stack wants to start using LLD for FCoE
824 * so the underlying device can perform whatever needed configuration or
825 * initialization to support acceleration of FCoE traffic.
827 * int (*ndo_fcoe_disable)(struct net_device *dev);
828 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
829 * so the underlying device can perform whatever needed clean-ups to
830 * stop supporting acceleration of FCoE traffic.
832 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
833 * struct scatterlist *sgl, unsigned int sgc);
834 * Called when the FCoE Initiator wants to initialize an I/O that
835 * is a possible candidate for Direct Data Placement (DDP). The LLD can
836 * perform necessary setup and returns 1 to indicate the device is set up
837 * successfully to perform DDP on this I/O, otherwise this returns 0.
839 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
840 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
841 * indicated by the FC exchange id 'xid', so the underlying device can
842 * clean up and reuse resources for later DDP requests.
844 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
845 * struct scatterlist *sgl, unsigned int sgc);
846 * Called when the FCoE Target wants to initialize an I/O that
847 * is a possible candidate for Direct Data Placement (DDP). The LLD can
848 * perform necessary setup and returns 1 to indicate the device is set up
849 * successfully to perform DDP on this I/O, otherwise this returns 0.
851 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
852 * Called when the underlying device wants to override default World Wide
853 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
854 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
855 * protocol stack to use.
858 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
859 * u16 rxq_index, u32 flow_id);
860 * Set hardware filter for RFS. rxq_index is the target queue index;
861 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
862 * Return the filter ID on success, or a negative error code.
864 * Slave management functions (for bridge, bonding, etc). User should
865 * call netdev_set_master() to set dev->master properly.
866 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
867 * Called to make another netdev an underling.
869 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
870 * Called to release previously enslaved netdev.
872 * Feature/offload setting functions.
873 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
874 * netdev_features_t features);
875 * Adjusts the requested feature flags according to device-specific
876 * constraints, and returns the resulting flags. Must not modify
879 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
880 * Called to update device configuration to new features. Passed
881 * feature set might be less than what was returned by ndo_fix_features()).
882 * Must return >0 or -errno if it changed dev->features itself.
885 struct net_device_ops {
886 int (*ndo_init)(struct net_device *dev);
887 void (*ndo_uninit)(struct net_device *dev);
888 int (*ndo_open)(struct net_device *dev);
889 int (*ndo_stop)(struct net_device *dev);
890 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
891 struct net_device *dev);
892 u16 (*ndo_select_queue)(struct net_device *dev,
893 struct sk_buff *skb);
894 void (*ndo_change_rx_flags)(struct net_device *dev,
896 void (*ndo_set_rx_mode)(struct net_device *dev);
897 int (*ndo_set_mac_address)(struct net_device *dev,
899 int (*ndo_validate_addr)(struct net_device *dev);
900 int (*ndo_do_ioctl)(struct net_device *dev,
901 struct ifreq *ifr, int cmd);
902 int (*ndo_set_config)(struct net_device *dev,
904 int (*ndo_change_mtu)(struct net_device *dev,
906 int (*ndo_neigh_setup)(struct net_device *dev,
907 struct neigh_parms *);
908 void (*ndo_tx_timeout) (struct net_device *dev);
910 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
911 struct rtnl_link_stats64 *storage);
912 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
914 void (*ndo_vlan_rx_add_vid)(struct net_device *dev,
916 void (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
918 #ifdef CONFIG_NET_POLL_CONTROLLER
919 void (*ndo_poll_controller)(struct net_device *dev);
920 int (*ndo_netpoll_setup)(struct net_device *dev,
921 struct netpoll_info *info);
922 void (*ndo_netpoll_cleanup)(struct net_device *dev);
924 int (*ndo_set_vf_mac)(struct net_device *dev,
926 int (*ndo_set_vf_vlan)(struct net_device *dev,
927 int queue, u16 vlan, u8 qos);
928 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
930 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
931 int vf, bool setting);
932 int (*ndo_get_vf_config)(struct net_device *dev,
934 struct ifla_vf_info *ivf);
935 int (*ndo_set_vf_port)(struct net_device *dev,
937 struct nlattr *port[]);
938 int (*ndo_get_vf_port)(struct net_device *dev,
939 int vf, struct sk_buff *skb);
940 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
941 #if IS_ENABLED(CONFIG_FCOE)
942 int (*ndo_fcoe_enable)(struct net_device *dev);
943 int (*ndo_fcoe_disable)(struct net_device *dev);
944 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
946 struct scatterlist *sgl,
948 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
950 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
952 struct scatterlist *sgl,
956 #if IS_ENABLED(CONFIG_LIBFCOE)
957 #define NETDEV_FCOE_WWNN 0
958 #define NETDEV_FCOE_WWPN 1
959 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
963 #ifdef CONFIG_RFS_ACCEL
964 int (*ndo_rx_flow_steer)(struct net_device *dev,
965 const struct sk_buff *skb,
969 int (*ndo_add_slave)(struct net_device *dev,
970 struct net_device *slave_dev);
971 int (*ndo_del_slave)(struct net_device *dev,
972 struct net_device *slave_dev);
973 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
974 netdev_features_t features);
975 int (*ndo_set_features)(struct net_device *dev,
976 netdev_features_t features);
980 * The DEVICE structure.
981 * Actually, this whole structure is a big mistake. It mixes I/O
982 * data with strictly "high-level" data, and it has to know about
983 * almost every data structure used in the INET module.
985 * FIXME: cleanup struct net_device such that network protocol info
992 * This is the first field of the "visible" part of this structure
993 * (i.e. as seen by users in the "Space.c" file). It is the name
998 struct pm_qos_request pm_qos_req;
1000 /* device name hash chain */
1001 struct hlist_node name_hlist;
1006 * I/O specific fields
1007 * FIXME: Merge these and struct ifmap into one
1009 unsigned long mem_end; /* shared mem end */
1010 unsigned long mem_start; /* shared mem start */
1011 unsigned long base_addr; /* device I/O address */
1012 unsigned int irq; /* device IRQ number */
1015 * Some hardware also needs these fields, but they are not
1016 * part of the usual set specified in Space.c.
1019 unsigned long state;
1021 struct list_head dev_list;
1022 struct list_head napi_list;
1023 struct list_head unreg_list;
1025 /* currently active device features */
1026 netdev_features_t features;
1027 /* user-changeable features */
1028 netdev_features_t hw_features;
1029 /* user-requested features */
1030 netdev_features_t wanted_features;
1031 /* mask of features inheritable by VLAN devices */
1032 netdev_features_t vlan_features;
1034 /* Interface index. Unique device identifier */
1038 struct net_device_stats stats;
1039 atomic_long_t rx_dropped; /* dropped packets by core network
1040 * Do not use this in drivers.
1043 #ifdef CONFIG_WIRELESS_EXT
1044 /* List of functions to handle Wireless Extensions (instead of ioctl).
1045 * See <net/iw_handler.h> for details. Jean II */
1046 const struct iw_handler_def * wireless_handlers;
1047 /* Instance data managed by the core of Wireless Extensions. */
1048 struct iw_public_data * wireless_data;
1050 /* Management operations */
1051 const struct net_device_ops *netdev_ops;
1052 const struct ethtool_ops *ethtool_ops;
1054 /* Hardware header description */
1055 const struct header_ops *header_ops;
1057 unsigned int flags; /* interface flags (a la BSD) */
1058 unsigned int priv_flags; /* Like 'flags' but invisible to userspace. */
1059 unsigned short gflags;
1060 unsigned short padded; /* How much padding added by alloc_netdev() */
1062 unsigned char operstate; /* RFC2863 operstate */
1063 unsigned char link_mode; /* mapping policy to operstate */
1065 unsigned char if_port; /* Selectable AUI, TP,..*/
1066 unsigned char dma; /* DMA channel */
1068 unsigned int mtu; /* interface MTU value */
1069 unsigned short type; /* interface hardware type */
1070 unsigned short hard_header_len; /* hardware hdr length */
1072 /* extra head- and tailroom the hardware may need, but not in all cases
1073 * can this be guaranteed, especially tailroom. Some cases also use
1074 * LL_MAX_HEADER instead to allocate the skb.
1076 unsigned short needed_headroom;
1077 unsigned short needed_tailroom;
1079 /* Interface address info. */
1080 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1081 unsigned char addr_assign_type; /* hw address assignment type */
1082 unsigned char addr_len; /* hardware address length */
1083 unsigned char neigh_priv_len;
1084 unsigned short dev_id; /* for shared network cards */
1086 spinlock_t addr_list_lock;
1087 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1088 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1090 unsigned int promiscuity;
1091 unsigned int allmulti;
1094 /* Protocol specific pointers */
1096 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1097 struct vlan_group __rcu *vlgrp; /* VLAN group */
1099 #if IS_ENABLED(CONFIG_NET_DSA)
1100 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1102 void *atalk_ptr; /* AppleTalk link */
1103 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1104 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1105 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1106 void *ec_ptr; /* Econet specific data */
1107 void *ax25_ptr; /* AX.25 specific data */
1108 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1109 assign before registering */
1112 * Cache lines mostly used on receive path (including eth_type_trans())
1114 unsigned long last_rx; /* Time of last Rx
1115 * This should not be set in
1116 * drivers, unless really needed,
1117 * because network stack (bonding)
1118 * use it if/when necessary, to
1119 * avoid dirtying this cache line.
1122 struct net_device *master; /* Pointer to master device of a group,
1123 * which this device is member of.
1126 /* Interface address info used in eth_type_trans() */
1127 unsigned char *dev_addr; /* hw address, (before bcast
1128 because most packets are
1131 struct netdev_hw_addr_list dev_addrs; /* list of device
1134 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1137 struct kset *queues_kset;
1141 struct netdev_rx_queue *_rx;
1143 /* Number of RX queues allocated at register_netdev() time */
1144 unsigned int num_rx_queues;
1146 /* Number of RX queues currently active in device */
1147 unsigned int real_num_rx_queues;
1149 #ifdef CONFIG_RFS_ACCEL
1150 /* CPU reverse-mapping for RX completion interrupts, indexed
1151 * by RX queue number. Assigned by driver. This must only be
1152 * set if the ndo_rx_flow_steer operation is defined. */
1153 struct cpu_rmap *rx_cpu_rmap;
1157 rx_handler_func_t __rcu *rx_handler;
1158 void __rcu *rx_handler_data;
1160 struct netdev_queue __rcu *ingress_queue;
1163 * Cache lines mostly used on transmit path
1165 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1167 /* Number of TX queues allocated at alloc_netdev_mq() time */
1168 unsigned int num_tx_queues;
1170 /* Number of TX queues currently active in device */
1171 unsigned int real_num_tx_queues;
1173 /* root qdisc from userspace point of view */
1174 struct Qdisc *qdisc;
1176 unsigned long tx_queue_len; /* Max frames per queue allowed */
1177 spinlock_t tx_global_lock;
1180 struct xps_dev_maps __rcu *xps_maps;
1183 /* These may be needed for future network-power-down code. */
1186 * trans_start here is expensive for high speed devices on SMP,
1187 * please use netdev_queue->trans_start instead.
1189 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1191 int watchdog_timeo; /* used by dev_watchdog() */
1192 struct timer_list watchdog_timer;
1194 /* Number of references to this device */
1195 int __percpu *pcpu_refcnt;
1197 /* delayed register/unregister */
1198 struct list_head todo_list;
1199 /* device index hash chain */
1200 struct hlist_node index_hlist;
1202 struct list_head link_watch_list;
1204 /* register/unregister state machine */
1205 enum { NETREG_UNINITIALIZED=0,
1206 NETREG_REGISTERED, /* completed register_netdevice */
1207 NETREG_UNREGISTERING, /* called unregister_netdevice */
1208 NETREG_UNREGISTERED, /* completed unregister todo */
1209 NETREG_RELEASED, /* called free_netdev */
1210 NETREG_DUMMY, /* dummy device for NAPI poll */
1213 bool dismantle; /* device is going do be freed */
1216 RTNL_LINK_INITIALIZED,
1217 RTNL_LINK_INITIALIZING,
1218 } rtnl_link_state:16;
1220 /* Called from unregister, can be used to call free_netdev */
1221 void (*destructor)(struct net_device *dev);
1223 #ifdef CONFIG_NETPOLL
1224 struct netpoll_info *npinfo;
1227 #ifdef CONFIG_NET_NS
1228 /* Network namespace this network device is inside */
1232 /* mid-layer private */
1235 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1236 struct pcpu_tstats __percpu *tstats; /* tunnel stats */
1237 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1240 struct garp_port __rcu *garp_port;
1242 /* class/net/name entry */
1244 /* space for optional device, statistics, and wireless sysfs groups */
1245 const struct attribute_group *sysfs_groups[4];
1247 /* rtnetlink link ops */
1248 const struct rtnl_link_ops *rtnl_link_ops;
1250 /* for setting kernel sock attribute on TCP connection setup */
1251 #define GSO_MAX_SIZE 65536
1252 unsigned int gso_max_size;
1255 /* Data Center Bridging netlink ops */
1256 const struct dcbnl_rtnl_ops *dcbnl_ops;
1259 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1260 u8 prio_tc_map[TC_BITMASK + 1];
1262 #if IS_ENABLED(CONFIG_FCOE)
1263 /* max exchange id for FCoE LRO by ddp */
1264 unsigned int fcoe_ddp_xid;
1266 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
1267 struct netprio_map __rcu *priomap;
1269 /* phy device may attach itself for hardware timestamping */
1270 struct phy_device *phydev;
1272 /* group the device belongs to */
1275 #define to_net_dev(d) container_of(d, struct net_device, dev)
1277 #define NETDEV_ALIGN 32
1280 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1282 return dev->prio_tc_map[prio & TC_BITMASK];
1286 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1288 if (tc >= dev->num_tc)
1291 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1296 void netdev_reset_tc(struct net_device *dev)
1299 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1300 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1304 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1306 if (tc >= dev->num_tc)
1309 dev->tc_to_txq[tc].count = count;
1310 dev->tc_to_txq[tc].offset = offset;
1315 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1317 if (num_tc > TC_MAX_QUEUE)
1320 dev->num_tc = num_tc;
1325 int netdev_get_num_tc(struct net_device *dev)
1331 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1334 return &dev->_tx[index];
1337 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1338 void (*f)(struct net_device *,
1339 struct netdev_queue *,
1345 for (i = 0; i < dev->num_tx_queues; i++)
1346 f(dev, &dev->_tx[i], arg);
1350 * Net namespace inlines
1353 struct net *dev_net(const struct net_device *dev)
1355 return read_pnet(&dev->nd_net);
1359 void dev_net_set(struct net_device *dev, struct net *net)
1361 #ifdef CONFIG_NET_NS
1362 release_net(dev->nd_net);
1363 dev->nd_net = hold_net(net);
1367 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1369 #ifdef CONFIG_NET_DSA_TAG_DSA
1370 if (dev->dsa_ptr != NULL)
1371 return dsa_uses_dsa_tags(dev->dsa_ptr);
1377 #ifndef CONFIG_NET_NS
1378 static inline void skb_set_dev(struct sk_buff *skb, struct net_device *dev)
1382 #else /* CONFIG_NET_NS */
1383 void skb_set_dev(struct sk_buff *skb, struct net_device *dev);
1386 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1388 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1389 if (dev->dsa_ptr != NULL)
1390 return dsa_uses_trailer_tags(dev->dsa_ptr);
1397 * netdev_priv - access network device private data
1398 * @dev: network device
1400 * Get network device private data
1402 static inline void *netdev_priv(const struct net_device *dev)
1404 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1407 /* Set the sysfs physical device reference for the network logical device
1408 * if set prior to registration will cause a symlink during initialization.
1410 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1412 /* Set the sysfs device type for the network logical device to allow
1413 * fin grained indentification of different network device types. For
1414 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1416 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1419 * netif_napi_add - initialize a napi context
1420 * @dev: network device
1421 * @napi: napi context
1422 * @poll: polling function
1423 * @weight: default weight
1425 * netif_napi_add() must be used to initialize a napi context prior to calling
1426 * *any* of the other napi related functions.
1428 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1429 int (*poll)(struct napi_struct *, int), int weight);
1432 * netif_napi_del - remove a napi context
1433 * @napi: napi context
1435 * netif_napi_del() removes a napi context from the network device napi list
1437 void netif_napi_del(struct napi_struct *napi);
1439 struct napi_gro_cb {
1440 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1443 /* Length of frag0. */
1444 unsigned int frag0_len;
1446 /* This indicates where we are processing relative to skb->data. */
1449 /* This is non-zero if the packet may be of the same flow. */
1452 /* This is non-zero if the packet cannot be merged with the new skb. */
1455 /* Number of segments aggregated. */
1462 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1464 struct packet_type {
1465 __be16 type; /* This is really htons(ether_type). */
1466 struct net_device *dev; /* NULL is wildcarded here */
1467 int (*func) (struct sk_buff *,
1468 struct net_device *,
1469 struct packet_type *,
1470 struct net_device *);
1471 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1472 netdev_features_t features);
1473 int (*gso_send_check)(struct sk_buff *skb);
1474 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1475 struct sk_buff *skb);
1476 int (*gro_complete)(struct sk_buff *skb);
1477 void *af_packet_priv;
1478 struct list_head list;
1481 #include <linux/notifier.h>
1483 /* netdevice notifier chain. Please remember to update the rtnetlink
1484 * notification exclusion list in rtnetlink_event() when adding new
1487 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1488 #define NETDEV_DOWN 0x0002
1489 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1490 detected a hardware crash and restarted
1491 - we can use this eg to kick tcp sessions
1493 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1494 #define NETDEV_REGISTER 0x0005
1495 #define NETDEV_UNREGISTER 0x0006
1496 #define NETDEV_CHANGEMTU 0x0007
1497 #define NETDEV_CHANGEADDR 0x0008
1498 #define NETDEV_GOING_DOWN 0x0009
1499 #define NETDEV_CHANGENAME 0x000A
1500 #define NETDEV_FEAT_CHANGE 0x000B
1501 #define NETDEV_BONDING_FAILOVER 0x000C
1502 #define NETDEV_PRE_UP 0x000D
1503 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1504 #define NETDEV_POST_TYPE_CHANGE 0x000F
1505 #define NETDEV_POST_INIT 0x0010
1506 #define NETDEV_UNREGISTER_BATCH 0x0011
1507 #define NETDEV_RELEASE 0x0012
1508 #define NETDEV_NOTIFY_PEERS 0x0013
1509 #define NETDEV_JOIN 0x0014
1511 extern int register_netdevice_notifier(struct notifier_block *nb);
1512 extern int unregister_netdevice_notifier(struct notifier_block *nb);
1513 extern int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1516 extern rwlock_t dev_base_lock; /* Device list lock */
1519 #define for_each_netdev(net, d) \
1520 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1521 #define for_each_netdev_reverse(net, d) \
1522 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1523 #define for_each_netdev_rcu(net, d) \
1524 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1525 #define for_each_netdev_safe(net, d, n) \
1526 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1527 #define for_each_netdev_continue(net, d) \
1528 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1529 #define for_each_netdev_continue_rcu(net, d) \
1530 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1531 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1533 static inline struct net_device *next_net_device(struct net_device *dev)
1535 struct list_head *lh;
1539 lh = dev->dev_list.next;
1540 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1543 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1545 struct list_head *lh;
1549 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1550 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1553 static inline struct net_device *first_net_device(struct net *net)
1555 return list_empty(&net->dev_base_head) ? NULL :
1556 net_device_entry(net->dev_base_head.next);
1559 static inline struct net_device *first_net_device_rcu(struct net *net)
1561 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1563 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1566 extern int netdev_boot_setup_check(struct net_device *dev);
1567 extern unsigned long netdev_boot_base(const char *prefix, int unit);
1568 extern struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1569 const char *hwaddr);
1570 extern struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1571 extern struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1572 extern void dev_add_pack(struct packet_type *pt);
1573 extern void dev_remove_pack(struct packet_type *pt);
1574 extern void __dev_remove_pack(struct packet_type *pt);
1576 extern struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1577 unsigned short mask);
1578 extern struct net_device *dev_get_by_name(struct net *net, const char *name);
1579 extern struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1580 extern struct net_device *__dev_get_by_name(struct net *net, const char *name);
1581 extern int dev_alloc_name(struct net_device *dev, const char *name);
1582 extern int dev_open(struct net_device *dev);
1583 extern int dev_close(struct net_device *dev);
1584 extern void dev_disable_lro(struct net_device *dev);
1585 extern int dev_queue_xmit(struct sk_buff *skb);
1586 extern int register_netdevice(struct net_device *dev);
1587 extern void unregister_netdevice_queue(struct net_device *dev,
1588 struct list_head *head);
1589 extern void unregister_netdevice_many(struct list_head *head);
1590 static inline void unregister_netdevice(struct net_device *dev)
1592 unregister_netdevice_queue(dev, NULL);
1595 extern int netdev_refcnt_read(const struct net_device *dev);
1596 extern void free_netdev(struct net_device *dev);
1597 extern void synchronize_net(void);
1598 extern int init_dummy_netdev(struct net_device *dev);
1599 extern void netdev_resync_ops(struct net_device *dev);
1601 extern struct net_device *dev_get_by_index(struct net *net, int ifindex);
1602 extern struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1603 extern struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1604 extern int dev_restart(struct net_device *dev);
1605 #ifdef CONFIG_NETPOLL_TRAP
1606 extern int netpoll_trap(void);
1608 extern int skb_gro_receive(struct sk_buff **head,
1609 struct sk_buff *skb);
1610 extern void skb_gro_reset_offset(struct sk_buff *skb);
1612 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1614 return NAPI_GRO_CB(skb)->data_offset;
1617 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1619 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1622 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1624 NAPI_GRO_CB(skb)->data_offset += len;
1627 static inline void *skb_gro_header_fast(struct sk_buff *skb,
1628 unsigned int offset)
1630 return NAPI_GRO_CB(skb)->frag0 + offset;
1633 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1635 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1638 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1639 unsigned int offset)
1641 if (!pskb_may_pull(skb, hlen))
1644 NAPI_GRO_CB(skb)->frag0 = NULL;
1645 NAPI_GRO_CB(skb)->frag0_len = 0;
1646 return skb->data + offset;
1649 static inline void *skb_gro_mac_header(struct sk_buff *skb)
1651 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
1654 static inline void *skb_gro_network_header(struct sk_buff *skb)
1656 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1657 skb_network_offset(skb);
1660 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1661 unsigned short type,
1662 const void *daddr, const void *saddr,
1665 if (!dev->header_ops || !dev->header_ops->create)
1668 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
1671 static inline int dev_parse_header(const struct sk_buff *skb,
1672 unsigned char *haddr)
1674 const struct net_device *dev = skb->dev;
1676 if (!dev->header_ops || !dev->header_ops->parse)
1678 return dev->header_ops->parse(skb, haddr);
1681 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
1682 extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf);
1683 static inline int unregister_gifconf(unsigned int family)
1685 return register_gifconf(family, NULL);
1689 * Incoming packets are placed on per-cpu queues
1691 struct softnet_data {
1692 struct Qdisc *output_queue;
1693 struct Qdisc **output_queue_tailp;
1694 struct list_head poll_list;
1695 struct sk_buff *completion_queue;
1696 struct sk_buff_head process_queue;
1699 unsigned int processed;
1700 unsigned int time_squeeze;
1701 unsigned int cpu_collision;
1702 unsigned int received_rps;
1705 struct softnet_data *rps_ipi_list;
1707 /* Elements below can be accessed between CPUs for RPS */
1708 struct call_single_data csd ____cacheline_aligned_in_smp;
1709 struct softnet_data *rps_ipi_next;
1711 unsigned int input_queue_head;
1712 unsigned int input_queue_tail;
1715 struct sk_buff_head input_pkt_queue;
1716 struct napi_struct backlog;
1719 static inline void input_queue_head_incr(struct softnet_data *sd)
1722 sd->input_queue_head++;
1726 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1727 unsigned int *qtail)
1730 *qtail = ++sd->input_queue_tail;
1734 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1736 extern void __netif_schedule(struct Qdisc *q);
1738 static inline void netif_schedule_queue(struct netdev_queue *txq)
1740 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
1741 __netif_schedule(txq->qdisc);
1744 static inline void netif_tx_schedule_all(struct net_device *dev)
1748 for (i = 0; i < dev->num_tx_queues; i++)
1749 netif_schedule_queue(netdev_get_tx_queue(dev, i));
1752 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
1754 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1758 * netif_start_queue - allow transmit
1759 * @dev: network device
1761 * Allow upper layers to call the device hard_start_xmit routine.
1763 static inline void netif_start_queue(struct net_device *dev)
1765 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1768 static inline void netif_tx_start_all_queues(struct net_device *dev)
1772 for (i = 0; i < dev->num_tx_queues; i++) {
1773 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1774 netif_tx_start_queue(txq);
1778 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1780 #ifdef CONFIG_NETPOLL_TRAP
1781 if (netpoll_trap()) {
1782 netif_tx_start_queue(dev_queue);
1786 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
1787 __netif_schedule(dev_queue->qdisc);
1791 * netif_wake_queue - restart transmit
1792 * @dev: network device
1794 * Allow upper layers to call the device hard_start_xmit routine.
1795 * Used for flow control when transmit resources are available.
1797 static inline void netif_wake_queue(struct net_device *dev)
1799 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1802 static inline void netif_tx_wake_all_queues(struct net_device *dev)
1806 for (i = 0; i < dev->num_tx_queues; i++) {
1807 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1808 netif_tx_wake_queue(txq);
1812 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
1814 if (WARN_ON(!dev_queue)) {
1815 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
1818 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1822 * netif_stop_queue - stop transmitted packets
1823 * @dev: network device
1825 * Stop upper layers calling the device hard_start_xmit routine.
1826 * Used for flow control when transmit resources are unavailable.
1828 static inline void netif_stop_queue(struct net_device *dev)
1830 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1833 static inline void netif_tx_stop_all_queues(struct net_device *dev)
1837 for (i = 0; i < dev->num_tx_queues; i++) {
1838 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1839 netif_tx_stop_queue(txq);
1843 static inline int netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
1845 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1849 * netif_queue_stopped - test if transmit queue is flowblocked
1850 * @dev: network device
1852 * Test if transmit queue on device is currently unable to send.
1854 static inline int netif_queue_stopped(const struct net_device *dev)
1856 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1859 static inline int netif_xmit_stopped(const struct netdev_queue *dev_queue)
1861 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
1864 static inline int netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
1866 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
1869 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
1873 dql_queued(&dev_queue->dql, bytes);
1874 if (unlikely(dql_avail(&dev_queue->dql) < 0)) {
1875 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
1876 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
1877 clear_bit(__QUEUE_STATE_STACK_XOFF,
1883 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
1885 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
1888 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
1889 unsigned pkts, unsigned bytes)
1892 if (likely(bytes)) {
1893 dql_completed(&dev_queue->dql, bytes);
1894 if (unlikely(test_bit(__QUEUE_STATE_STACK_XOFF,
1895 &dev_queue->state) &&
1896 dql_avail(&dev_queue->dql) >= 0)) {
1897 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF,
1899 netif_schedule_queue(dev_queue);
1905 static inline void netdev_completed_queue(struct net_device *dev,
1906 unsigned pkts, unsigned bytes)
1908 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
1911 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
1918 static inline void netdev_reset_queue(struct net_device *dev_queue)
1920 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
1924 * netif_running - test if up
1925 * @dev: network device
1927 * Test if the device has been brought up.
1929 static inline int netif_running(const struct net_device *dev)
1931 return test_bit(__LINK_STATE_START, &dev->state);
1935 * Routines to manage the subqueues on a device. We only need start
1936 * stop, and a check if it's stopped. All other device management is
1937 * done at the overall netdevice level.
1938 * Also test the device if we're multiqueue.
1942 * netif_start_subqueue - allow sending packets on subqueue
1943 * @dev: network device
1944 * @queue_index: sub queue index
1946 * Start individual transmit queue of a device with multiple transmit queues.
1948 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
1950 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
1952 netif_tx_start_queue(txq);
1956 * netif_stop_subqueue - stop sending packets on subqueue
1957 * @dev: network device
1958 * @queue_index: sub queue index
1960 * Stop individual transmit queue of a device with multiple transmit queues.
1962 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
1964 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
1965 #ifdef CONFIG_NETPOLL_TRAP
1969 netif_tx_stop_queue(txq);
1973 * netif_subqueue_stopped - test status of subqueue
1974 * @dev: network device
1975 * @queue_index: sub queue index
1977 * Check individual transmit queue of a device with multiple transmit queues.
1979 static inline int __netif_subqueue_stopped(const struct net_device *dev,
1982 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
1984 return netif_tx_queue_stopped(txq);
1987 static inline int netif_subqueue_stopped(const struct net_device *dev,
1988 struct sk_buff *skb)
1990 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
1994 * netif_wake_subqueue - allow sending packets on subqueue
1995 * @dev: network device
1996 * @queue_index: sub queue index
1998 * Resume individual transmit queue of a device with multiple transmit queues.
2000 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2002 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2003 #ifdef CONFIG_NETPOLL_TRAP
2007 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2008 __netif_schedule(txq->qdisc);
2012 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2013 * as a distribution range limit for the returned value.
2015 static inline u16 skb_tx_hash(const struct net_device *dev,
2016 const struct sk_buff *skb)
2018 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2022 * netif_is_multiqueue - test if device has multiple transmit queues
2023 * @dev: network device
2025 * Check if device has multiple transmit queues
2027 static inline int netif_is_multiqueue(const struct net_device *dev)
2029 return dev->num_tx_queues > 1;
2032 extern int netif_set_real_num_tx_queues(struct net_device *dev,
2036 extern int netif_set_real_num_rx_queues(struct net_device *dev,
2039 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2046 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2047 const struct net_device *from_dev)
2049 netif_set_real_num_tx_queues(to_dev, from_dev->real_num_tx_queues);
2051 return netif_set_real_num_rx_queues(to_dev,
2052 from_dev->real_num_rx_queues);
2058 /* Use this variant when it is known for sure that it
2059 * is executing from hardware interrupt context or with hardware interrupts
2062 extern void dev_kfree_skb_irq(struct sk_buff *skb);
2064 /* Use this variant in places where it could be invoked
2065 * from either hardware interrupt or other context, with hardware interrupts
2066 * either disabled or enabled.
2068 extern void dev_kfree_skb_any(struct sk_buff *skb);
2070 extern int netif_rx(struct sk_buff *skb);
2071 extern int netif_rx_ni(struct sk_buff *skb);
2072 extern int netif_receive_skb(struct sk_buff *skb);
2073 extern gro_result_t dev_gro_receive(struct napi_struct *napi,
2074 struct sk_buff *skb);
2075 extern gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb);
2076 extern gro_result_t napi_gro_receive(struct napi_struct *napi,
2077 struct sk_buff *skb);
2078 extern void napi_gro_flush(struct napi_struct *napi);
2079 extern struct sk_buff * napi_get_frags(struct napi_struct *napi);
2080 extern gro_result_t napi_frags_finish(struct napi_struct *napi,
2081 struct sk_buff *skb,
2083 extern struct sk_buff * napi_frags_skb(struct napi_struct *napi);
2084 extern gro_result_t napi_gro_frags(struct napi_struct *napi);
2086 static inline void napi_free_frags(struct napi_struct *napi)
2088 kfree_skb(napi->skb);
2092 extern int netdev_rx_handler_register(struct net_device *dev,
2093 rx_handler_func_t *rx_handler,
2094 void *rx_handler_data);
2095 extern void netdev_rx_handler_unregister(struct net_device *dev);
2097 extern int dev_valid_name(const char *name);
2098 extern int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2099 extern int dev_ethtool(struct net *net, struct ifreq *);
2100 extern unsigned dev_get_flags(const struct net_device *);
2101 extern int __dev_change_flags(struct net_device *, unsigned int flags);
2102 extern int dev_change_flags(struct net_device *, unsigned);
2103 extern void __dev_notify_flags(struct net_device *, unsigned int old_flags);
2104 extern int dev_change_name(struct net_device *, const char *);
2105 extern int dev_set_alias(struct net_device *, const char *, size_t);
2106 extern int dev_change_net_namespace(struct net_device *,
2107 struct net *, const char *);
2108 extern int dev_set_mtu(struct net_device *, int);
2109 extern void dev_set_group(struct net_device *, int);
2110 extern int dev_set_mac_address(struct net_device *,
2112 extern int dev_hard_start_xmit(struct sk_buff *skb,
2113 struct net_device *dev,
2114 struct netdev_queue *txq);
2115 extern int dev_forward_skb(struct net_device *dev,
2116 struct sk_buff *skb);
2118 extern int netdev_budget;
2120 /* Called by rtnetlink.c:rtnl_unlock() */
2121 extern void netdev_run_todo(void);
2124 * dev_put - release reference to device
2125 * @dev: network device
2127 * Release reference to device to allow it to be freed.
2129 static inline void dev_put(struct net_device *dev)
2131 irqsafe_cpu_dec(*dev->pcpu_refcnt);
2135 * dev_hold - get reference to device
2136 * @dev: network device
2138 * Hold reference to device to keep it from being freed.
2140 static inline void dev_hold(struct net_device *dev)
2142 irqsafe_cpu_inc(*dev->pcpu_refcnt);
2145 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2146 * and _off may be called from IRQ context, but it is caller
2147 * who is responsible for serialization of these calls.
2149 * The name carrier is inappropriate, these functions should really be
2150 * called netif_lowerlayer_*() because they represent the state of any
2151 * kind of lower layer not just hardware media.
2154 extern void linkwatch_fire_event(struct net_device *dev);
2155 extern void linkwatch_forget_dev(struct net_device *dev);
2158 * netif_carrier_ok - test if carrier present
2159 * @dev: network device
2161 * Check if carrier is present on device
2163 static inline int netif_carrier_ok(const struct net_device *dev)
2165 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2168 extern unsigned long dev_trans_start(struct net_device *dev);
2170 extern void __netdev_watchdog_up(struct net_device *dev);
2172 extern void netif_carrier_on(struct net_device *dev);
2174 extern void netif_carrier_off(struct net_device *dev);
2176 extern void netif_notify_peers(struct net_device *dev);
2179 * netif_dormant_on - mark device as dormant.
2180 * @dev: network device
2182 * Mark device as dormant (as per RFC2863).
2184 * The dormant state indicates that the relevant interface is not
2185 * actually in a condition to pass packets (i.e., it is not 'up') but is
2186 * in a "pending" state, waiting for some external event. For "on-
2187 * demand" interfaces, this new state identifies the situation where the
2188 * interface is waiting for events to place it in the up state.
2191 static inline void netif_dormant_on(struct net_device *dev)
2193 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2194 linkwatch_fire_event(dev);
2198 * netif_dormant_off - set device as not dormant.
2199 * @dev: network device
2201 * Device is not in dormant state.
2203 static inline void netif_dormant_off(struct net_device *dev)
2205 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2206 linkwatch_fire_event(dev);
2210 * netif_dormant - test if carrier present
2211 * @dev: network device
2213 * Check if carrier is present on device
2215 static inline int netif_dormant(const struct net_device *dev)
2217 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2222 * netif_oper_up - test if device is operational
2223 * @dev: network device
2225 * Check if carrier is operational
2227 static inline int netif_oper_up(const struct net_device *dev)
2229 return (dev->operstate == IF_OPER_UP ||
2230 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2234 * netif_device_present - is device available or removed
2235 * @dev: network device
2237 * Check if device has not been removed from system.
2239 static inline int netif_device_present(struct net_device *dev)
2241 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2244 extern void netif_device_detach(struct net_device *dev);
2246 extern void netif_device_attach(struct net_device *dev);
2249 * Network interface message level settings
2253 NETIF_MSG_DRV = 0x0001,
2254 NETIF_MSG_PROBE = 0x0002,
2255 NETIF_MSG_LINK = 0x0004,
2256 NETIF_MSG_TIMER = 0x0008,
2257 NETIF_MSG_IFDOWN = 0x0010,
2258 NETIF_MSG_IFUP = 0x0020,
2259 NETIF_MSG_RX_ERR = 0x0040,
2260 NETIF_MSG_TX_ERR = 0x0080,
2261 NETIF_MSG_TX_QUEUED = 0x0100,
2262 NETIF_MSG_INTR = 0x0200,
2263 NETIF_MSG_TX_DONE = 0x0400,
2264 NETIF_MSG_RX_STATUS = 0x0800,
2265 NETIF_MSG_PKTDATA = 0x1000,
2266 NETIF_MSG_HW = 0x2000,
2267 NETIF_MSG_WOL = 0x4000,
2270 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2271 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2272 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2273 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2274 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2275 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2276 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2277 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2278 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2279 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2280 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2281 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2282 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2283 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2284 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2286 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2289 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2290 return default_msg_enable_bits;
2291 if (debug_value == 0) /* no output */
2293 /* set low N bits */
2294 return (1 << debug_value) - 1;
2297 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2299 spin_lock(&txq->_xmit_lock);
2300 txq->xmit_lock_owner = cpu;
2303 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2305 spin_lock_bh(&txq->_xmit_lock);
2306 txq->xmit_lock_owner = smp_processor_id();
2309 static inline int __netif_tx_trylock(struct netdev_queue *txq)
2311 int ok = spin_trylock(&txq->_xmit_lock);
2313 txq->xmit_lock_owner = smp_processor_id();
2317 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2319 txq->xmit_lock_owner = -1;
2320 spin_unlock(&txq->_xmit_lock);
2323 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2325 txq->xmit_lock_owner = -1;
2326 spin_unlock_bh(&txq->_xmit_lock);
2329 static inline void txq_trans_update(struct netdev_queue *txq)
2331 if (txq->xmit_lock_owner != -1)
2332 txq->trans_start = jiffies;
2336 * netif_tx_lock - grab network device transmit lock
2337 * @dev: network device
2339 * Get network device transmit lock
2341 static inline void netif_tx_lock(struct net_device *dev)
2346 spin_lock(&dev->tx_global_lock);
2347 cpu = smp_processor_id();
2348 for (i = 0; i < dev->num_tx_queues; i++) {
2349 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2351 /* We are the only thread of execution doing a
2352 * freeze, but we have to grab the _xmit_lock in
2353 * order to synchronize with threads which are in
2354 * the ->hard_start_xmit() handler and already
2355 * checked the frozen bit.
2357 __netif_tx_lock(txq, cpu);
2358 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2359 __netif_tx_unlock(txq);
2363 static inline void netif_tx_lock_bh(struct net_device *dev)
2369 static inline void netif_tx_unlock(struct net_device *dev)
2373 for (i = 0; i < dev->num_tx_queues; i++) {
2374 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2376 /* No need to grab the _xmit_lock here. If the
2377 * queue is not stopped for another reason, we
2380 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2381 netif_schedule_queue(txq);
2383 spin_unlock(&dev->tx_global_lock);
2386 static inline void netif_tx_unlock_bh(struct net_device *dev)
2388 netif_tx_unlock(dev);
2392 #define HARD_TX_LOCK(dev, txq, cpu) { \
2393 if ((dev->features & NETIF_F_LLTX) == 0) { \
2394 __netif_tx_lock(txq, cpu); \
2398 #define HARD_TX_UNLOCK(dev, txq) { \
2399 if ((dev->features & NETIF_F_LLTX) == 0) { \
2400 __netif_tx_unlock(txq); \
2404 static inline void netif_tx_disable(struct net_device *dev)
2410 cpu = smp_processor_id();
2411 for (i = 0; i < dev->num_tx_queues; i++) {
2412 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2414 __netif_tx_lock(txq, cpu);
2415 netif_tx_stop_queue(txq);
2416 __netif_tx_unlock(txq);
2421 static inline void netif_addr_lock(struct net_device *dev)
2423 spin_lock(&dev->addr_list_lock);
2426 static inline void netif_addr_lock_bh(struct net_device *dev)
2428 spin_lock_bh(&dev->addr_list_lock);
2431 static inline void netif_addr_unlock(struct net_device *dev)
2433 spin_unlock(&dev->addr_list_lock);
2436 static inline void netif_addr_unlock_bh(struct net_device *dev)
2438 spin_unlock_bh(&dev->addr_list_lock);
2442 * dev_addrs walker. Should be used only for read access. Call with
2443 * rcu_read_lock held.
2445 #define for_each_dev_addr(dev, ha) \
2446 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2448 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2450 extern void ether_setup(struct net_device *dev);
2452 /* Support for loadable net-drivers */
2453 extern struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2454 void (*setup)(struct net_device *),
2455 unsigned int txqs, unsigned int rxqs);
2456 #define alloc_netdev(sizeof_priv, name, setup) \
2457 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2459 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2460 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2462 extern int register_netdev(struct net_device *dev);
2463 extern void unregister_netdev(struct net_device *dev);
2465 /* General hardware address lists handling functions */
2466 extern int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2467 struct netdev_hw_addr_list *from_list,
2468 int addr_len, unsigned char addr_type);
2469 extern void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2470 struct netdev_hw_addr_list *from_list,
2471 int addr_len, unsigned char addr_type);
2472 extern int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2473 struct netdev_hw_addr_list *from_list,
2475 extern void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2476 struct netdev_hw_addr_list *from_list,
2478 extern void __hw_addr_flush(struct netdev_hw_addr_list *list);
2479 extern void __hw_addr_init(struct netdev_hw_addr_list *list);
2481 /* Functions used for device addresses handling */
2482 extern int dev_addr_add(struct net_device *dev, unsigned char *addr,
2483 unsigned char addr_type);
2484 extern int dev_addr_del(struct net_device *dev, unsigned char *addr,
2485 unsigned char addr_type);
2486 extern int dev_addr_add_multiple(struct net_device *to_dev,
2487 struct net_device *from_dev,
2488 unsigned char addr_type);
2489 extern int dev_addr_del_multiple(struct net_device *to_dev,
2490 struct net_device *from_dev,
2491 unsigned char addr_type);
2492 extern void dev_addr_flush(struct net_device *dev);
2493 extern int dev_addr_init(struct net_device *dev);
2495 /* Functions used for unicast addresses handling */
2496 extern int dev_uc_add(struct net_device *dev, unsigned char *addr);
2497 extern int dev_uc_del(struct net_device *dev, unsigned char *addr);
2498 extern int dev_uc_sync(struct net_device *to, struct net_device *from);
2499 extern void dev_uc_unsync(struct net_device *to, struct net_device *from);
2500 extern void dev_uc_flush(struct net_device *dev);
2501 extern void dev_uc_init(struct net_device *dev);
2503 /* Functions used for multicast addresses handling */
2504 extern int dev_mc_add(struct net_device *dev, unsigned char *addr);
2505 extern int dev_mc_add_global(struct net_device *dev, unsigned char *addr);
2506 extern int dev_mc_del(struct net_device *dev, unsigned char *addr);
2507 extern int dev_mc_del_global(struct net_device *dev, unsigned char *addr);
2508 extern int dev_mc_sync(struct net_device *to, struct net_device *from);
2509 extern void dev_mc_unsync(struct net_device *to, struct net_device *from);
2510 extern void dev_mc_flush(struct net_device *dev);
2511 extern void dev_mc_init(struct net_device *dev);
2513 /* Functions used for secondary unicast and multicast support */
2514 extern void dev_set_rx_mode(struct net_device *dev);
2515 extern void __dev_set_rx_mode(struct net_device *dev);
2516 extern int dev_set_promiscuity(struct net_device *dev, int inc);
2517 extern int dev_set_allmulti(struct net_device *dev, int inc);
2518 extern void netdev_state_change(struct net_device *dev);
2519 extern int netdev_bonding_change(struct net_device *dev,
2520 unsigned long event);
2521 extern void netdev_features_change(struct net_device *dev);
2522 /* Load a device via the kmod */
2523 extern void dev_load(struct net *net, const char *name);
2524 extern void dev_mcast_init(void);
2525 extern struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2526 struct rtnl_link_stats64 *storage);
2528 extern int netdev_max_backlog;
2529 extern int netdev_tstamp_prequeue;
2530 extern int weight_p;
2531 extern int bpf_jit_enable;
2532 extern int netdev_set_master(struct net_device *dev, struct net_device *master);
2533 extern int netdev_set_bond_master(struct net_device *dev,
2534 struct net_device *master);
2535 extern int skb_checksum_help(struct sk_buff *skb);
2536 extern struct sk_buff *skb_gso_segment(struct sk_buff *skb,
2537 netdev_features_t features);
2539 extern void netdev_rx_csum_fault(struct net_device *dev);
2541 static inline void netdev_rx_csum_fault(struct net_device *dev)
2545 /* rx skb timestamps */
2546 extern void net_enable_timestamp(void);
2547 extern void net_disable_timestamp(void);
2549 #ifdef CONFIG_PROC_FS
2550 extern void *dev_seq_start(struct seq_file *seq, loff_t *pos);
2551 extern void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos);
2552 extern void dev_seq_stop(struct seq_file *seq, void *v);
2555 extern int netdev_class_create_file(struct class_attribute *class_attr);
2556 extern void netdev_class_remove_file(struct class_attribute *class_attr);
2558 extern struct kobj_ns_type_operations net_ns_type_operations;
2560 extern const char *netdev_drivername(const struct net_device *dev);
2562 extern void linkwatch_run_queue(void);
2564 static inline netdev_features_t netdev_get_wanted_features(
2565 struct net_device *dev)
2567 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2569 netdev_features_t netdev_increment_features(netdev_features_t all,
2570 netdev_features_t one, netdev_features_t mask);
2571 int __netdev_update_features(struct net_device *dev);
2572 void netdev_update_features(struct net_device *dev);
2573 void netdev_change_features(struct net_device *dev);
2575 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
2576 struct net_device *dev);
2578 netdev_features_t netif_skb_features(struct sk_buff *skb);
2580 static inline int net_gso_ok(netdev_features_t features, int gso_type)
2582 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
2584 /* check flags correspondence */
2585 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
2586 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
2587 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
2588 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
2589 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
2590 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
2592 return (features & feature) == feature;
2595 static inline int skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
2597 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
2598 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
2601 static inline int netif_needs_gso(struct sk_buff *skb,
2602 netdev_features_t features)
2604 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
2605 unlikely(skb->ip_summed != CHECKSUM_PARTIAL));
2608 static inline void netif_set_gso_max_size(struct net_device *dev,
2611 dev->gso_max_size = size;
2614 static inline int netif_is_bond_slave(struct net_device *dev)
2616 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
2619 extern struct pernet_operations __net_initdata loopback_net_ops;
2621 /* Logging, debugging and troubleshooting/diagnostic helpers. */
2623 /* netdev_printk helpers, similar to dev_printk */
2625 static inline const char *netdev_name(const struct net_device *dev)
2627 if (dev->reg_state != NETREG_REGISTERED)
2628 return "(unregistered net_device)";
2632 extern int __netdev_printk(const char *level, const struct net_device *dev,
2633 struct va_format *vaf);
2635 extern __printf(3, 4)
2636 int netdev_printk(const char *level, const struct net_device *dev,
2637 const char *format, ...);
2638 extern __printf(2, 3)
2639 int netdev_emerg(const struct net_device *dev, const char *format, ...);
2640 extern __printf(2, 3)
2641 int netdev_alert(const struct net_device *dev, const char *format, ...);
2642 extern __printf(2, 3)
2643 int netdev_crit(const struct net_device *dev, const char *format, ...);
2644 extern __printf(2, 3)
2645 int netdev_err(const struct net_device *dev, const char *format, ...);
2646 extern __printf(2, 3)
2647 int netdev_warn(const struct net_device *dev, const char *format, ...);
2648 extern __printf(2, 3)
2649 int netdev_notice(const struct net_device *dev, const char *format, ...);
2650 extern __printf(2, 3)
2651 int netdev_info(const struct net_device *dev, const char *format, ...);
2653 #define MODULE_ALIAS_NETDEV(device) \
2654 MODULE_ALIAS("netdev-" device)
2657 #define netdev_dbg(__dev, format, args...) \
2658 netdev_printk(KERN_DEBUG, __dev, format, ##args)
2659 #elif defined(CONFIG_DYNAMIC_DEBUG)
2660 #define netdev_dbg(__dev, format, args...) \
2662 dynamic_netdev_dbg(__dev, format, ##args); \
2665 #define netdev_dbg(__dev, format, args...) \
2668 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
2673 #if defined(VERBOSE_DEBUG)
2674 #define netdev_vdbg netdev_dbg
2677 #define netdev_vdbg(dev, format, args...) \
2680 netdev_printk(KERN_DEBUG, dev, format, ##args); \
2686 * netdev_WARN() acts like dev_printk(), but with the key difference
2687 * of using a WARN/WARN_ON to get the message out, including the
2688 * file/line information and a backtrace.
2690 #define netdev_WARN(dev, format, args...) \
2691 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args);
2693 /* netif printk helpers, similar to netdev_printk */
2695 #define netif_printk(priv, type, level, dev, fmt, args...) \
2697 if (netif_msg_##type(priv)) \
2698 netdev_printk(level, (dev), fmt, ##args); \
2701 #define netif_level(level, priv, type, dev, fmt, args...) \
2703 if (netif_msg_##type(priv)) \
2704 netdev_##level(dev, fmt, ##args); \
2707 #define netif_emerg(priv, type, dev, fmt, args...) \
2708 netif_level(emerg, priv, type, dev, fmt, ##args)
2709 #define netif_alert(priv, type, dev, fmt, args...) \
2710 netif_level(alert, priv, type, dev, fmt, ##args)
2711 #define netif_crit(priv, type, dev, fmt, args...) \
2712 netif_level(crit, priv, type, dev, fmt, ##args)
2713 #define netif_err(priv, type, dev, fmt, args...) \
2714 netif_level(err, priv, type, dev, fmt, ##args)
2715 #define netif_warn(priv, type, dev, fmt, args...) \
2716 netif_level(warn, priv, type, dev, fmt, ##args)
2717 #define netif_notice(priv, type, dev, fmt, args...) \
2718 netif_level(notice, priv, type, dev, fmt, ##args)
2719 #define netif_info(priv, type, dev, fmt, args...) \
2720 netif_level(info, priv, type, dev, fmt, ##args)
2723 #define netif_dbg(priv, type, dev, format, args...) \
2724 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
2725 #elif defined(CONFIG_DYNAMIC_DEBUG)
2726 #define netif_dbg(priv, type, netdev, format, args...) \
2728 if (netif_msg_##type(priv)) \
2729 dynamic_netdev_dbg(netdev, format, ##args); \
2732 #define netif_dbg(priv, type, dev, format, args...) \
2735 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2740 #if defined(VERBOSE_DEBUG)
2741 #define netif_vdbg netif_dbg
2743 #define netif_vdbg(priv, type, dev, format, args...) \
2746 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2751 #endif /* __KERNEL__ */
2753 #endif /* _LINUX_NETDEVICE_H */