2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
59 /* source back-compat hooks */
60 #define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
63 extern void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct 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))
135 * Compute the worst case header length according to the protocols
139 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
140 # if defined(CONFIG_MAC80211_MESH)
141 # define LL_MAX_HEADER 128
143 # define LL_MAX_HEADER 96
145 #elif IS_ENABLED(CONFIG_TR)
146 # define LL_MAX_HEADER 48
148 # define LL_MAX_HEADER 32
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
163 struct net_device_stats {
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
173 unsigned long collisions;
174 unsigned long rx_length_errors;
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed;
202 struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
206 #define NETDEV_HW_ADDR_T_LAN 1
207 #define NETDEV_HW_ADDR_T_SAN 2
208 #define NETDEV_HW_ADDR_T_SLAVE 3
209 #define NETDEV_HW_ADDR_T_UNICAST 4
210 #define NETDEV_HW_ADDR_T_MULTICAST 5
214 struct rcu_head rcu_head;
217 struct netdev_hw_addr_list {
218 struct list_head list;
222 #define netdev_hw_addr_list_count(l) ((l)->count)
223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224 #define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229 #define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234 #define netdev_for_each_mc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
242 /* cached hardware header; allow for machine alignment needs. */
243 #define HH_DATA_MOD 16
244 #define HH_DATA_OFF(__len) \
245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
246 #define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
259 #define LL_RESERVED_SPACE(dev) \
260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
261 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
267 const void *saddr, unsigned int len);
268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
269 int (*rebuild)(struct sk_buff *skb);
270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
276 /* These flag bits are private to the generic network queueing
277 * layer, they may not be explicitly referenced by any other
281 enum netdev_state_t {
283 __LINK_STATE_PRESENT,
284 __LINK_STATE_NOCARRIER,
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
291 * This structure holds at boot time configured netdevice settings. They
292 * are then used in the device probing.
294 struct netdev_boot_setup {
298 #define NETDEV_BOOT_SETUP_MAX 8
300 extern int __init netdev_boot_setup(char *str);
303 * Structure for NAPI scheduling similar to tasklet but with weighting
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
309 * to the per-cpu poll_list, and whoever clears that bit
310 * can remove from the list right before clearing the bit.
312 struct list_head poll_list;
316 unsigned int gro_count;
317 int (*poll)(struct napi_struct *, int);
318 #ifdef CONFIG_NETPOLL
319 spinlock_t poll_lock;
322 struct net_device *dev;
323 struct sk_buff *gro_list;
325 struct list_head dev_list;
329 NAPI_STATE_SCHED, /* Poll is scheduled */
330 NAPI_STATE_DISABLE, /* Disable pending */
331 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
341 typedef enum gro_result gro_result_t;
344 * enum rx_handler_result - Possible return values for rx_handlers.
345 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
347 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
348 * case skb->dev was changed by rx_handler.
349 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
350 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
352 * rx_handlers are functions called from inside __netif_receive_skb(), to do
353 * special processing of the skb, prior to delivery to protocol handlers.
355 * Currently, a net_device can only have a single rx_handler registered. Trying
356 * to register a second rx_handler will return -EBUSY.
358 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
359 * To unregister a rx_handler on a net_device, use
360 * netdev_rx_handler_unregister().
362 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
365 * If the rx_handler consumed to skb in some way, it should return
366 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
367 * the skb to be delivered in some other ways.
369 * If the rx_handler changed skb->dev, to divert the skb to another
370 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
371 * new device will be called if it exists.
373 * If the rx_handler consider the skb should be ignored, it should return
374 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
375 * are registered on exact device (ptype->dev == skb->dev).
377 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
378 * delivered, it should return RX_HANDLER_PASS.
380 * A device without a registered rx_handler will behave as if rx_handler
381 * returned RX_HANDLER_PASS.
384 enum rx_handler_result {
390 typedef enum rx_handler_result rx_handler_result_t;
391 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
393 extern void __napi_schedule(struct napi_struct *n);
395 static inline bool napi_disable_pending(struct napi_struct *n)
397 return test_bit(NAPI_STATE_DISABLE, &n->state);
401 * napi_schedule_prep - check if napi can be scheduled
404 * Test if NAPI routine is already running, and if not mark
405 * it as running. This is used as a condition variable
406 * insure only one NAPI poll instance runs. We also make
407 * sure there is no pending NAPI disable.
409 static inline bool napi_schedule_prep(struct napi_struct *n)
411 return !napi_disable_pending(n) &&
412 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
416 * napi_schedule - schedule NAPI poll
419 * Schedule NAPI poll routine to be called if it is not already
422 static inline void napi_schedule(struct napi_struct *n)
424 if (napi_schedule_prep(n))
428 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
429 static inline bool napi_reschedule(struct napi_struct *napi)
431 if (napi_schedule_prep(napi)) {
432 __napi_schedule(napi);
439 * napi_complete - NAPI processing complete
442 * Mark NAPI processing as complete.
444 extern void __napi_complete(struct napi_struct *n);
445 extern void napi_complete(struct napi_struct *n);
448 * napi_disable - prevent NAPI from scheduling
451 * Stop NAPI from being scheduled on this context.
452 * Waits till any outstanding processing completes.
454 static inline void napi_disable(struct napi_struct *n)
456 set_bit(NAPI_STATE_DISABLE, &n->state);
457 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
459 clear_bit(NAPI_STATE_DISABLE, &n->state);
463 * napi_enable - enable NAPI scheduling
466 * Resume NAPI from being scheduled on this context.
467 * Must be paired with napi_disable.
469 static inline void napi_enable(struct napi_struct *n)
471 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
472 smp_mb__before_clear_bit();
473 clear_bit(NAPI_STATE_SCHED, &n->state);
478 * napi_synchronize - wait until NAPI is not running
481 * Wait until NAPI is done being scheduled on this context.
482 * Waits till any outstanding processing completes but
483 * does not disable future activations.
485 static inline void napi_synchronize(const struct napi_struct *n)
487 while (test_bit(NAPI_STATE_SCHED, &n->state))
491 # define napi_synchronize(n) barrier()
494 enum netdev_queue_state_t {
495 __QUEUE_STATE_DRV_XOFF,
496 __QUEUE_STATE_STACK_XOFF,
497 __QUEUE_STATE_FROZEN,
498 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
499 (1 << __QUEUE_STATE_STACK_XOFF))
500 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
501 (1 << __QUEUE_STATE_FROZEN))
504 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
505 * netif_tx_* functions below are used to manipulate this flag. The
506 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
507 * queue independently. The netif_xmit_*stopped functions below are called
508 * to check if the queue has been stopped by the driver or stack (either
509 * of the XOFF bits are set in the state). Drivers should not need to call
510 * netif_xmit*stopped functions, they should only be using netif_tx_*.
513 struct netdev_queue {
517 struct net_device *dev;
519 struct Qdisc *qdisc_sleeping;
523 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
529 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
532 * please use this field instead of dev->trans_start
534 unsigned long trans_start;
537 * Number of TX timeouts for this queue
538 * (/sys/class/net/DEV/Q/trans_timeout)
540 unsigned long trans_timeout;
547 } ____cacheline_aligned_in_smp;
549 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
551 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
558 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
560 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
567 * This structure holds an RPS map which can be of variable length. The
568 * map is an array of CPUs.
575 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
578 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
579 * tail pointer for that CPU's input queue at the time of last enqueue, and
580 * a hardware filter index.
582 struct rps_dev_flow {
585 unsigned int last_qtail;
587 #define RPS_NO_FILTER 0xffff
590 * The rps_dev_flow_table structure contains a table of flow mappings.
592 struct rps_dev_flow_table {
595 struct work_struct free_work;
596 struct rps_dev_flow flows[0];
598 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
599 ((_num) * sizeof(struct rps_dev_flow)))
602 * The rps_sock_flow_table contains mappings of flows to the last CPU
603 * on which they were processed by the application (set in recvmsg).
605 struct rps_sock_flow_table {
609 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
610 ((_num) * sizeof(u16)))
612 #define RPS_NO_CPU 0xffff
614 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
618 unsigned int cpu, index = hash & table->mask;
620 /* We only give a hint, preemption can change cpu under us */
621 cpu = raw_smp_processor_id();
623 if (table->ents[index] != cpu)
624 table->ents[index] = cpu;
628 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
632 table->ents[hash & table->mask] = RPS_NO_CPU;
635 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
637 #ifdef CONFIG_RFS_ACCEL
638 extern bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
639 u32 flow_id, u16 filter_id);
642 /* This structure contains an instance of an RX queue. */
643 struct netdev_rx_queue {
644 struct rps_map __rcu *rps_map;
645 struct rps_dev_flow_table __rcu *rps_flow_table;
647 struct net_device *dev;
648 } ____cacheline_aligned_in_smp;
649 #endif /* CONFIG_RPS */
653 * This structure holds an XPS map which can be of variable length. The
654 * map is an array of queues.
658 unsigned int alloc_len;
662 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
663 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
667 * This structure holds all XPS maps for device. Maps are indexed by CPU.
669 struct xps_dev_maps {
671 struct xps_map __rcu *cpu_map[0];
673 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
674 (nr_cpu_ids * sizeof(struct xps_map *)))
675 #endif /* CONFIG_XPS */
677 #define TC_MAX_QUEUE 16
678 #define TC_BITMASK 15
679 /* HW offloaded queuing disciplines txq count and offset maps */
680 struct netdev_tc_txq {
685 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
687 * This structure is to hold information about the device
688 * configured to run FCoE protocol stack.
690 struct netdev_fcoe_hbainfo {
691 char manufacturer[64];
692 char serial_number[64];
693 char hardware_version[64];
694 char driver_version[64];
695 char optionrom_version[64];
696 char firmware_version[64];
698 char model_description[256];
703 * This structure defines the management hooks for network devices.
704 * The following hooks can be defined; unless noted otherwise, they are
705 * optional and can be filled with a null pointer.
707 * int (*ndo_init)(struct net_device *dev);
708 * This function is called once when network device is registered.
709 * The network device can use this to any late stage initializaton
710 * or semantic validattion. It can fail with an error code which will
711 * be propogated back to register_netdev
713 * void (*ndo_uninit)(struct net_device *dev);
714 * This function is called when device is unregistered or when registration
715 * fails. It is not called if init fails.
717 * int (*ndo_open)(struct net_device *dev);
718 * This function is called when network device transistions to the up
721 * int (*ndo_stop)(struct net_device *dev);
722 * This function is called when network device transistions to the down
725 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
726 * struct net_device *dev);
727 * Called when a packet needs to be transmitted.
728 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
729 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
730 * Required can not be NULL.
732 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
733 * Called to decide which queue to when device supports multiple
736 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
737 * This function is called to allow device receiver to make
738 * changes to configuration when multicast or promiscious is enabled.
740 * void (*ndo_set_rx_mode)(struct net_device *dev);
741 * This function is called device changes address list filtering.
742 * If driver handles unicast address filtering, it should set
743 * IFF_UNICAST_FLT to its priv_flags.
745 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
746 * This function is called when the Media Access Control address
747 * needs to be changed. If this interface is not defined, the
748 * mac address can not be changed.
750 * int (*ndo_validate_addr)(struct net_device *dev);
751 * Test if Media Access Control address is valid for the device.
753 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
754 * Called when a user request an ioctl which can't be handled by
755 * the generic interface code. If not defined ioctl's return
756 * not supported error code.
758 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
759 * Used to set network devices bus interface parameters. This interface
760 * is retained for legacy reason, new devices should use the bus
761 * interface (PCI) for low level management.
763 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
764 * Called when a user wants to change the Maximum Transfer Unit
765 * of a device. If not defined, any request to change MTU will
766 * will return an error.
768 * void (*ndo_tx_timeout)(struct net_device *dev);
769 * Callback uses when the transmitter has not made any progress
770 * for dev->watchdog ticks.
772 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
773 * struct rtnl_link_stats64 *storage);
774 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
775 * Called when a user wants to get the network device usage
776 * statistics. Drivers must do one of the following:
777 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
778 * rtnl_link_stats64 structure passed by the caller.
779 * 2. Define @ndo_get_stats to update a net_device_stats structure
780 * (which should normally be dev->stats) and return a pointer to
781 * it. The structure may be changed asynchronously only if each
782 * field is written atomically.
783 * 3. Update dev->stats asynchronously and atomically, and define
786 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, unsigned short vid);
787 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER)
788 * this function is called when a VLAN id is registered.
790 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
791 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER)
792 * this function is called when a VLAN id is unregistered.
794 * void (*ndo_poll_controller)(struct net_device *dev);
796 * SR-IOV management functions.
797 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
798 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
799 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
800 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
801 * int (*ndo_get_vf_config)(struct net_device *dev,
802 * int vf, struct ifla_vf_info *ivf);
803 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
804 * struct nlattr *port[]);
805 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
806 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
807 * Called to setup 'tc' number of traffic classes in the net device. This
808 * is always called from the stack with the rtnl lock held and netif tx
809 * queues stopped. This allows the netdevice to perform queue management
812 * Fiber Channel over Ethernet (FCoE) offload functions.
813 * int (*ndo_fcoe_enable)(struct net_device *dev);
814 * Called when the FCoE protocol stack wants to start using LLD for FCoE
815 * so the underlying device can perform whatever needed configuration or
816 * initialization to support acceleration of FCoE traffic.
818 * int (*ndo_fcoe_disable)(struct net_device *dev);
819 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
820 * so the underlying device can perform whatever needed clean-ups to
821 * stop supporting acceleration of FCoE traffic.
823 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
824 * struct scatterlist *sgl, unsigned int sgc);
825 * Called when the FCoE Initiator wants to initialize an I/O that
826 * is a possible candidate for Direct Data Placement (DDP). The LLD can
827 * perform necessary setup and returns 1 to indicate the device is set up
828 * successfully to perform DDP on this I/O, otherwise this returns 0.
830 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
831 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
832 * indicated by the FC exchange id 'xid', so the underlying device can
833 * clean up and reuse resources for later DDP requests.
835 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
836 * struct scatterlist *sgl, unsigned int sgc);
837 * Called when the FCoE Target wants to initialize an I/O that
838 * is a possible candidate for Direct Data Placement (DDP). The LLD can
839 * perform necessary setup and returns 1 to indicate the device is set up
840 * successfully to perform DDP on this I/O, otherwise this returns 0.
842 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
843 * struct netdev_fcoe_hbainfo *hbainfo);
844 * Called when the FCoE Protocol stack wants information on the underlying
845 * device. This information is utilized by the FCoE protocol stack to
846 * register attributes with Fiber Channel management service as per the
847 * FC-GS Fabric Device Management Information(FDMI) specification.
849 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
850 * Called when the underlying device wants to override default World Wide
851 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
852 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
853 * protocol stack to use.
856 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
857 * u16 rxq_index, u32 flow_id);
858 * Set hardware filter for RFS. rxq_index is the target queue index;
859 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
860 * Return the filter ID on success, or a negative error code.
862 * Slave management functions (for bridge, bonding, etc). User should
863 * call netdev_set_master() to set dev->master properly.
864 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
865 * Called to make another netdev an underling.
867 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
868 * Called to release previously enslaved netdev.
870 * Feature/offload setting functions.
871 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
872 * netdev_features_t features);
873 * Adjusts the requested feature flags according to device-specific
874 * constraints, and returns the resulting flags. Must not modify
877 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
878 * Called to update device configuration to new features. Passed
879 * feature set might be less than what was returned by ndo_fix_features()).
880 * Must return >0 or -errno if it changed dev->features itself.
882 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
883 * struct net_device *dev,
884 * const unsigned char *addr, u16 flags)
885 * Adds an FDB entry to dev for addr.
886 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct net_device *dev,
887 * const unsigned char *addr)
888 * Deletes the FDB entry from dev coresponding to addr.
889 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
890 * struct net_device *dev, int idx)
891 * Used to add FDB entries to dump requests. Implementers should add
892 * entries to skb and update idx with the number of entries.
894 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
895 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
896 * struct net_device *dev)
898 struct net_device_ops {
899 int (*ndo_init)(struct net_device *dev);
900 void (*ndo_uninit)(struct net_device *dev);
901 int (*ndo_open)(struct net_device *dev);
902 int (*ndo_stop)(struct net_device *dev);
903 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
904 struct net_device *dev);
905 u16 (*ndo_select_queue)(struct net_device *dev,
906 struct sk_buff *skb);
907 void (*ndo_change_rx_flags)(struct net_device *dev,
909 void (*ndo_set_rx_mode)(struct net_device *dev);
910 int (*ndo_set_mac_address)(struct net_device *dev,
912 int (*ndo_validate_addr)(struct net_device *dev);
913 int (*ndo_do_ioctl)(struct net_device *dev,
914 struct ifreq *ifr, int cmd);
915 int (*ndo_set_config)(struct net_device *dev,
917 int (*ndo_change_mtu)(struct net_device *dev,
919 int (*ndo_neigh_setup)(struct net_device *dev,
920 struct neigh_parms *);
921 void (*ndo_tx_timeout) (struct net_device *dev);
923 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
924 struct rtnl_link_stats64 *storage);
925 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
927 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
929 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
931 #ifdef CONFIG_NET_POLL_CONTROLLER
932 void (*ndo_poll_controller)(struct net_device *dev);
933 int (*ndo_netpoll_setup)(struct net_device *dev,
934 struct netpoll_info *info,
936 void (*ndo_netpoll_cleanup)(struct net_device *dev);
938 int (*ndo_set_vf_mac)(struct net_device *dev,
940 int (*ndo_set_vf_vlan)(struct net_device *dev,
941 int queue, u16 vlan, u8 qos);
942 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
944 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
945 int vf, bool setting);
946 int (*ndo_get_vf_config)(struct net_device *dev,
948 struct ifla_vf_info *ivf);
949 int (*ndo_set_vf_port)(struct net_device *dev,
951 struct nlattr *port[]);
952 int (*ndo_get_vf_port)(struct net_device *dev,
953 int vf, struct sk_buff *skb);
954 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
955 #if IS_ENABLED(CONFIG_FCOE)
956 int (*ndo_fcoe_enable)(struct net_device *dev);
957 int (*ndo_fcoe_disable)(struct net_device *dev);
958 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
960 struct scatterlist *sgl,
962 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
964 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
966 struct scatterlist *sgl,
968 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
969 struct netdev_fcoe_hbainfo *hbainfo);
972 #if IS_ENABLED(CONFIG_LIBFCOE)
973 #define NETDEV_FCOE_WWNN 0
974 #define NETDEV_FCOE_WWPN 1
975 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
979 #ifdef CONFIG_RFS_ACCEL
980 int (*ndo_rx_flow_steer)(struct net_device *dev,
981 const struct sk_buff *skb,
985 int (*ndo_add_slave)(struct net_device *dev,
986 struct net_device *slave_dev);
987 int (*ndo_del_slave)(struct net_device *dev,
988 struct net_device *slave_dev);
989 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
990 netdev_features_t features);
991 int (*ndo_set_features)(struct net_device *dev,
992 netdev_features_t features);
993 int (*ndo_neigh_construct)(struct neighbour *n);
994 void (*ndo_neigh_destroy)(struct neighbour *n);
996 int (*ndo_fdb_add)(struct ndmsg *ndm,
998 struct net_device *dev,
999 const unsigned char *addr,
1001 int (*ndo_fdb_del)(struct ndmsg *ndm,
1002 struct net_device *dev,
1003 const unsigned char *addr);
1004 int (*ndo_fdb_dump)(struct sk_buff *skb,
1005 struct netlink_callback *cb,
1006 struct net_device *dev,
1009 int (*ndo_bridge_setlink)(struct net_device *dev,
1010 struct nlmsghdr *nlh);
1011 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1013 struct net_device *dev);
1017 * The DEVICE structure.
1018 * Actually, this whole structure is a big mistake. It mixes I/O
1019 * data with strictly "high-level" data, and it has to know about
1020 * almost every data structure used in the INET module.
1022 * FIXME: cleanup struct net_device such that network protocol info
1029 * This is the first field of the "visible" part of this structure
1030 * (i.e. as seen by users in the "Space.c" file). It is the name
1033 char name[IFNAMSIZ];
1035 /* device name hash chain, please keep it close to name[] */
1036 struct hlist_node name_hlist;
1042 * I/O specific fields
1043 * FIXME: Merge these and struct ifmap into one
1045 unsigned long mem_end; /* shared mem end */
1046 unsigned long mem_start; /* shared mem start */
1047 unsigned long base_addr; /* device I/O address */
1048 unsigned int irq; /* device IRQ number */
1051 * Some hardware also needs these fields, but they are not
1052 * part of the usual set specified in Space.c.
1055 unsigned long state;
1057 struct list_head dev_list;
1058 struct list_head napi_list;
1059 struct list_head unreg_list;
1061 /* currently active device features */
1062 netdev_features_t features;
1063 /* user-changeable features */
1064 netdev_features_t hw_features;
1065 /* user-requested features */
1066 netdev_features_t wanted_features;
1067 /* mask of features inheritable by VLAN devices */
1068 netdev_features_t vlan_features;
1069 /* mask of features inherited by encapsulating devices
1070 * This field indicates what encapsulation offloads
1071 * the hardware is capable of doing, and drivers will
1072 * need to set them appropriately.
1074 netdev_features_t hw_enc_features;
1076 /* Interface index. Unique device identifier */
1080 struct net_device_stats stats;
1081 atomic_long_t rx_dropped; /* dropped packets by core network
1082 * Do not use this in drivers.
1085 #ifdef CONFIG_WIRELESS_EXT
1086 /* List of functions to handle Wireless Extensions (instead of ioctl).
1087 * See <net/iw_handler.h> for details. Jean II */
1088 const struct iw_handler_def * wireless_handlers;
1089 /* Instance data managed by the core of Wireless Extensions. */
1090 struct iw_public_data * wireless_data;
1092 /* Management operations */
1093 const struct net_device_ops *netdev_ops;
1094 const struct ethtool_ops *ethtool_ops;
1096 /* Hardware header description */
1097 const struct header_ops *header_ops;
1099 unsigned int flags; /* interface flags (a la BSD) */
1100 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1101 * See if.h for definitions. */
1102 unsigned short gflags;
1103 unsigned short padded; /* How much padding added by alloc_netdev() */
1105 unsigned char operstate; /* RFC2863 operstate */
1106 unsigned char link_mode; /* mapping policy to operstate */
1108 unsigned char if_port; /* Selectable AUI, TP,..*/
1109 unsigned char dma; /* DMA channel */
1111 unsigned int mtu; /* interface MTU value */
1112 unsigned short type; /* interface hardware type */
1113 unsigned short hard_header_len; /* hardware hdr length */
1115 /* extra head- and tailroom the hardware may need, but not in all cases
1116 * can this be guaranteed, especially tailroom. Some cases also use
1117 * LL_MAX_HEADER instead to allocate the skb.
1119 unsigned short needed_headroom;
1120 unsigned short needed_tailroom;
1122 /* Interface address info. */
1123 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1124 unsigned char addr_assign_type; /* hw address assignment type */
1125 unsigned char addr_len; /* hardware address length */
1126 unsigned char neigh_priv_len;
1127 unsigned short dev_id; /* for shared network cards */
1129 spinlock_t addr_list_lock;
1130 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1131 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1133 unsigned int promiscuity;
1134 unsigned int allmulti;
1137 /* Protocol specific pointers */
1139 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1140 struct vlan_info __rcu *vlan_info; /* VLAN info */
1142 #if IS_ENABLED(CONFIG_NET_DSA)
1143 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1145 void *atalk_ptr; /* AppleTalk link */
1146 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1147 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1148 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1149 void *ax25_ptr; /* AX.25 specific data */
1150 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1151 assign before registering */
1154 * Cache lines mostly used on receive path (including eth_type_trans())
1156 unsigned long last_rx; /* Time of last Rx
1157 * This should not be set in
1158 * drivers, unless really needed,
1159 * because network stack (bonding)
1160 * use it if/when necessary, to
1161 * avoid dirtying this cache line.
1164 struct net_device *master; /* Pointer to master device of a group,
1165 * which this device is member of.
1168 /* Interface address info used in eth_type_trans() */
1169 unsigned char *dev_addr; /* hw address, (before bcast
1170 because most packets are
1173 struct netdev_hw_addr_list dev_addrs; /* list of device
1176 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1179 struct kset *queues_kset;
1183 struct netdev_rx_queue *_rx;
1185 /* Number of RX queues allocated at register_netdev() time */
1186 unsigned int num_rx_queues;
1188 /* Number of RX queues currently active in device */
1189 unsigned int real_num_rx_queues;
1191 #ifdef CONFIG_RFS_ACCEL
1192 /* CPU reverse-mapping for RX completion interrupts, indexed
1193 * by RX queue number. Assigned by driver. This must only be
1194 * set if the ndo_rx_flow_steer operation is defined. */
1195 struct cpu_rmap *rx_cpu_rmap;
1199 rx_handler_func_t __rcu *rx_handler;
1200 void __rcu *rx_handler_data;
1202 struct netdev_queue __rcu *ingress_queue;
1205 * Cache lines mostly used on transmit path
1207 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1209 /* Number of TX queues allocated at alloc_netdev_mq() time */
1210 unsigned int num_tx_queues;
1212 /* Number of TX queues currently active in device */
1213 unsigned int real_num_tx_queues;
1215 /* root qdisc from userspace point of view */
1216 struct Qdisc *qdisc;
1218 unsigned long tx_queue_len; /* Max frames per queue allowed */
1219 spinlock_t tx_global_lock;
1222 struct xps_dev_maps __rcu *xps_maps;
1225 /* These may be needed for future network-power-down code. */
1228 * trans_start here is expensive for high speed devices on SMP,
1229 * please use netdev_queue->trans_start instead.
1231 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1233 int watchdog_timeo; /* used by dev_watchdog() */
1234 struct timer_list watchdog_timer;
1236 /* Number of references to this device */
1237 int __percpu *pcpu_refcnt;
1239 /* delayed register/unregister */
1240 struct list_head todo_list;
1241 /* device index hash chain */
1242 struct hlist_node index_hlist;
1244 struct list_head link_watch_list;
1246 /* register/unregister state machine */
1247 enum { NETREG_UNINITIALIZED=0,
1248 NETREG_REGISTERED, /* completed register_netdevice */
1249 NETREG_UNREGISTERING, /* called unregister_netdevice */
1250 NETREG_UNREGISTERED, /* completed unregister todo */
1251 NETREG_RELEASED, /* called free_netdev */
1252 NETREG_DUMMY, /* dummy device for NAPI poll */
1255 bool dismantle; /* device is going do be freed */
1258 RTNL_LINK_INITIALIZED,
1259 RTNL_LINK_INITIALIZING,
1260 } rtnl_link_state:16;
1262 /* Called from unregister, can be used to call free_netdev */
1263 void (*destructor)(struct net_device *dev);
1265 #ifdef CONFIG_NETPOLL
1266 struct netpoll_info *npinfo;
1269 #ifdef CONFIG_NET_NS
1270 /* Network namespace this network device is inside */
1274 /* mid-layer private */
1277 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1278 struct pcpu_tstats __percpu *tstats; /* tunnel stats */
1279 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1282 struct garp_port __rcu *garp_port;
1284 /* class/net/name entry */
1286 /* space for optional device, statistics, and wireless sysfs groups */
1287 const struct attribute_group *sysfs_groups[4];
1289 /* rtnetlink link ops */
1290 const struct rtnl_link_ops *rtnl_link_ops;
1292 /* for setting kernel sock attribute on TCP connection setup */
1293 #define GSO_MAX_SIZE 65536
1294 unsigned int gso_max_size;
1295 #define GSO_MAX_SEGS 65535
1299 /* Data Center Bridging netlink ops */
1300 const struct dcbnl_rtnl_ops *dcbnl_ops;
1303 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1304 u8 prio_tc_map[TC_BITMASK + 1];
1306 #if IS_ENABLED(CONFIG_FCOE)
1307 /* max exchange id for FCoE LRO by ddp */
1308 unsigned int fcoe_ddp_xid;
1310 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
1311 struct netprio_map __rcu *priomap;
1313 /* phy device may attach itself for hardware timestamping */
1314 struct phy_device *phydev;
1316 struct lock_class_key *qdisc_tx_busylock;
1318 /* group the device belongs to */
1321 struct pm_qos_request pm_qos_req;
1323 #define to_net_dev(d) container_of(d, struct net_device, dev)
1325 #define NETDEV_ALIGN 32
1328 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1330 return dev->prio_tc_map[prio & TC_BITMASK];
1334 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1336 if (tc >= dev->num_tc)
1339 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1344 void netdev_reset_tc(struct net_device *dev)
1347 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1348 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1352 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1354 if (tc >= dev->num_tc)
1357 dev->tc_to_txq[tc].count = count;
1358 dev->tc_to_txq[tc].offset = offset;
1363 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1365 if (num_tc > TC_MAX_QUEUE)
1368 dev->num_tc = num_tc;
1373 int netdev_get_num_tc(struct net_device *dev)
1379 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1382 return &dev->_tx[index];
1385 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1386 void (*f)(struct net_device *,
1387 struct netdev_queue *,
1393 for (i = 0; i < dev->num_tx_queues; i++)
1394 f(dev, &dev->_tx[i], arg);
1397 extern struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1398 struct sk_buff *skb);
1401 * Net namespace inlines
1404 struct net *dev_net(const struct net_device *dev)
1406 return read_pnet(&dev->nd_net);
1410 void dev_net_set(struct net_device *dev, struct net *net)
1412 #ifdef CONFIG_NET_NS
1413 release_net(dev->nd_net);
1414 dev->nd_net = hold_net(net);
1418 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1420 #ifdef CONFIG_NET_DSA_TAG_DSA
1421 if (dev->dsa_ptr != NULL)
1422 return dsa_uses_dsa_tags(dev->dsa_ptr);
1428 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1430 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1431 if (dev->dsa_ptr != NULL)
1432 return dsa_uses_trailer_tags(dev->dsa_ptr);
1439 * netdev_priv - access network device private data
1440 * @dev: network device
1442 * Get network device private data
1444 static inline void *netdev_priv(const struct net_device *dev)
1446 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1449 /* Set the sysfs physical device reference for the network logical device
1450 * if set prior to registration will cause a symlink during initialization.
1452 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1454 /* Set the sysfs device type for the network logical device to allow
1455 * fin grained indentification of different network device types. For
1456 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1458 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1461 * netif_napi_add - initialize a napi context
1462 * @dev: network device
1463 * @napi: napi context
1464 * @poll: polling function
1465 * @weight: default weight
1467 * netif_napi_add() must be used to initialize a napi context prior to calling
1468 * *any* of the other napi related functions.
1470 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1471 int (*poll)(struct napi_struct *, int), int weight);
1474 * netif_napi_del - remove a napi context
1475 * @napi: napi context
1477 * netif_napi_del() removes a napi context from the network device napi list
1479 void netif_napi_del(struct napi_struct *napi);
1481 struct napi_gro_cb {
1482 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1485 /* Length of frag0. */
1486 unsigned int frag0_len;
1488 /* This indicates where we are processing relative to skb->data. */
1491 /* This is non-zero if the packet cannot be merged with the new skb. */
1494 /* Number of segments aggregated. */
1497 /* This is non-zero if the packet may be of the same flow. */
1502 #define NAPI_GRO_FREE 1
1503 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1505 /* jiffies when first packet was created/queued */
1508 /* Used in ipv6_gro_receive() */
1511 /* used in skb_gro_receive() slow path */
1512 struct sk_buff *last;
1515 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1517 struct packet_type {
1518 __be16 type; /* This is really htons(ether_type). */
1519 struct net_device *dev; /* NULL is wildcarded here */
1520 int (*func) (struct sk_buff *,
1521 struct net_device *,
1522 struct packet_type *,
1523 struct net_device *);
1524 bool (*id_match)(struct packet_type *ptype,
1526 void *af_packet_priv;
1527 struct list_head list;
1530 struct offload_callbacks {
1531 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1532 netdev_features_t features);
1533 int (*gso_send_check)(struct sk_buff *skb);
1534 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1535 struct sk_buff *skb);
1536 int (*gro_complete)(struct sk_buff *skb);
1539 struct packet_offload {
1540 __be16 type; /* This is really htons(ether_type). */
1541 struct offload_callbacks callbacks;
1542 struct list_head list;
1545 #include <linux/notifier.h>
1547 /* netdevice notifier chain. Please remember to update the rtnetlink
1548 * notification exclusion list in rtnetlink_event() when adding new
1551 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1552 #define NETDEV_DOWN 0x0002
1553 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1554 detected a hardware crash and restarted
1555 - we can use this eg to kick tcp sessions
1557 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1558 #define NETDEV_REGISTER 0x0005
1559 #define NETDEV_UNREGISTER 0x0006
1560 #define NETDEV_CHANGEMTU 0x0007
1561 #define NETDEV_CHANGEADDR 0x0008
1562 #define NETDEV_GOING_DOWN 0x0009
1563 #define NETDEV_CHANGENAME 0x000A
1564 #define NETDEV_FEAT_CHANGE 0x000B
1565 #define NETDEV_BONDING_FAILOVER 0x000C
1566 #define NETDEV_PRE_UP 0x000D
1567 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1568 #define NETDEV_POST_TYPE_CHANGE 0x000F
1569 #define NETDEV_POST_INIT 0x0010
1570 #define NETDEV_UNREGISTER_FINAL 0x0011
1571 #define NETDEV_RELEASE 0x0012
1572 #define NETDEV_NOTIFY_PEERS 0x0013
1573 #define NETDEV_JOIN 0x0014
1575 extern int register_netdevice_notifier(struct notifier_block *nb);
1576 extern int unregister_netdevice_notifier(struct notifier_block *nb);
1577 extern int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1580 extern rwlock_t dev_base_lock; /* Device list lock */
1582 extern seqcount_t devnet_rename_seq; /* Device rename seq */
1585 #define for_each_netdev(net, d) \
1586 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1587 #define for_each_netdev_reverse(net, d) \
1588 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1589 #define for_each_netdev_rcu(net, d) \
1590 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1591 #define for_each_netdev_safe(net, d, n) \
1592 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1593 #define for_each_netdev_continue(net, d) \
1594 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1595 #define for_each_netdev_continue_rcu(net, d) \
1596 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1597 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1599 static inline struct net_device *next_net_device(struct net_device *dev)
1601 struct list_head *lh;
1605 lh = dev->dev_list.next;
1606 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1609 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1611 struct list_head *lh;
1615 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1616 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1619 static inline struct net_device *first_net_device(struct net *net)
1621 return list_empty(&net->dev_base_head) ? NULL :
1622 net_device_entry(net->dev_base_head.next);
1625 static inline struct net_device *first_net_device_rcu(struct net *net)
1627 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1629 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1632 extern int netdev_boot_setup_check(struct net_device *dev);
1633 extern unsigned long netdev_boot_base(const char *prefix, int unit);
1634 extern struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1635 const char *hwaddr);
1636 extern struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1637 extern struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1638 extern void dev_add_pack(struct packet_type *pt);
1639 extern void dev_remove_pack(struct packet_type *pt);
1640 extern void __dev_remove_pack(struct packet_type *pt);
1641 extern void dev_add_offload(struct packet_offload *po);
1642 extern void dev_remove_offload(struct packet_offload *po);
1643 extern void __dev_remove_offload(struct packet_offload *po);
1645 extern struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1646 unsigned short mask);
1647 extern struct net_device *dev_get_by_name(struct net *net, const char *name);
1648 extern struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1649 extern struct net_device *__dev_get_by_name(struct net *net, const char *name);
1650 extern int dev_alloc_name(struct net_device *dev, const char *name);
1651 extern int dev_open(struct net_device *dev);
1652 extern int dev_close(struct net_device *dev);
1653 extern void dev_disable_lro(struct net_device *dev);
1654 extern int dev_loopback_xmit(struct sk_buff *newskb);
1655 extern int dev_queue_xmit(struct sk_buff *skb);
1656 extern int register_netdevice(struct net_device *dev);
1657 extern void unregister_netdevice_queue(struct net_device *dev,
1658 struct list_head *head);
1659 extern void unregister_netdevice_many(struct list_head *head);
1660 static inline void unregister_netdevice(struct net_device *dev)
1662 unregister_netdevice_queue(dev, NULL);
1665 extern int netdev_refcnt_read(const struct net_device *dev);
1666 extern void free_netdev(struct net_device *dev);
1667 extern void synchronize_net(void);
1668 extern int init_dummy_netdev(struct net_device *dev);
1669 extern void netdev_resync_ops(struct net_device *dev);
1671 extern struct net_device *dev_get_by_index(struct net *net, int ifindex);
1672 extern struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1673 extern struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1674 extern int dev_restart(struct net_device *dev);
1675 #ifdef CONFIG_NETPOLL_TRAP
1676 extern int netpoll_trap(void);
1678 extern int skb_gro_receive(struct sk_buff **head,
1679 struct sk_buff *skb);
1681 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1683 return NAPI_GRO_CB(skb)->data_offset;
1686 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1688 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1691 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1693 NAPI_GRO_CB(skb)->data_offset += len;
1696 static inline void *skb_gro_header_fast(struct sk_buff *skb,
1697 unsigned int offset)
1699 return NAPI_GRO_CB(skb)->frag0 + offset;
1702 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1704 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1707 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1708 unsigned int offset)
1710 if (!pskb_may_pull(skb, hlen))
1713 NAPI_GRO_CB(skb)->frag0 = NULL;
1714 NAPI_GRO_CB(skb)->frag0_len = 0;
1715 return skb->data + offset;
1718 static inline void *skb_gro_mac_header(struct sk_buff *skb)
1720 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
1723 static inline void *skb_gro_network_header(struct sk_buff *skb)
1725 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1726 skb_network_offset(skb);
1729 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1730 unsigned short type,
1731 const void *daddr, const void *saddr,
1734 if (!dev->header_ops || !dev->header_ops->create)
1737 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
1740 static inline int dev_parse_header(const struct sk_buff *skb,
1741 unsigned char *haddr)
1743 const struct net_device *dev = skb->dev;
1745 if (!dev->header_ops || !dev->header_ops->parse)
1747 return dev->header_ops->parse(skb, haddr);
1750 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
1751 extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf);
1752 static inline int unregister_gifconf(unsigned int family)
1754 return register_gifconf(family, NULL);
1758 * Incoming packets are placed on per-cpu queues
1760 struct softnet_data {
1761 struct Qdisc *output_queue;
1762 struct Qdisc **output_queue_tailp;
1763 struct list_head poll_list;
1764 struct sk_buff *completion_queue;
1765 struct sk_buff_head process_queue;
1768 unsigned int processed;
1769 unsigned int time_squeeze;
1770 unsigned int cpu_collision;
1771 unsigned int received_rps;
1774 struct softnet_data *rps_ipi_list;
1776 /* Elements below can be accessed between CPUs for RPS */
1777 struct call_single_data csd ____cacheline_aligned_in_smp;
1778 struct softnet_data *rps_ipi_next;
1780 unsigned int input_queue_head;
1781 unsigned int input_queue_tail;
1783 unsigned int dropped;
1784 struct sk_buff_head input_pkt_queue;
1785 struct napi_struct backlog;
1788 static inline void input_queue_head_incr(struct softnet_data *sd)
1791 sd->input_queue_head++;
1795 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1796 unsigned int *qtail)
1799 *qtail = ++sd->input_queue_tail;
1803 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1805 extern void __netif_schedule(struct Qdisc *q);
1807 static inline void netif_schedule_queue(struct netdev_queue *txq)
1809 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
1810 __netif_schedule(txq->qdisc);
1813 static inline void netif_tx_schedule_all(struct net_device *dev)
1817 for (i = 0; i < dev->num_tx_queues; i++)
1818 netif_schedule_queue(netdev_get_tx_queue(dev, i));
1821 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
1823 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1827 * netif_start_queue - allow transmit
1828 * @dev: network device
1830 * Allow upper layers to call the device hard_start_xmit routine.
1832 static inline void netif_start_queue(struct net_device *dev)
1834 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1837 static inline void netif_tx_start_all_queues(struct net_device *dev)
1841 for (i = 0; i < dev->num_tx_queues; i++) {
1842 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1843 netif_tx_start_queue(txq);
1847 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1849 #ifdef CONFIG_NETPOLL_TRAP
1850 if (netpoll_trap()) {
1851 netif_tx_start_queue(dev_queue);
1855 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
1856 __netif_schedule(dev_queue->qdisc);
1860 * netif_wake_queue - restart transmit
1861 * @dev: network device
1863 * Allow upper layers to call the device hard_start_xmit routine.
1864 * Used for flow control when transmit resources are available.
1866 static inline void netif_wake_queue(struct net_device *dev)
1868 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1871 static inline void netif_tx_wake_all_queues(struct net_device *dev)
1875 for (i = 0; i < dev->num_tx_queues; i++) {
1876 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1877 netif_tx_wake_queue(txq);
1881 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
1883 if (WARN_ON(!dev_queue)) {
1884 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
1887 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1891 * netif_stop_queue - stop transmitted packets
1892 * @dev: network device
1894 * Stop upper layers calling the device hard_start_xmit routine.
1895 * Used for flow control when transmit resources are unavailable.
1897 static inline void netif_stop_queue(struct net_device *dev)
1899 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1902 static inline void netif_tx_stop_all_queues(struct net_device *dev)
1906 for (i = 0; i < dev->num_tx_queues; i++) {
1907 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1908 netif_tx_stop_queue(txq);
1912 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
1914 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1918 * netif_queue_stopped - test if transmit queue is flowblocked
1919 * @dev: network device
1921 * Test if transmit queue on device is currently unable to send.
1923 static inline bool netif_queue_stopped(const struct net_device *dev)
1925 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1928 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
1930 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
1933 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
1935 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
1938 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
1942 dql_queued(&dev_queue->dql, bytes);
1944 if (likely(dql_avail(&dev_queue->dql) >= 0))
1947 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
1950 * The XOFF flag must be set before checking the dql_avail below,
1951 * because in netdev_tx_completed_queue we update the dql_completed
1952 * before checking the XOFF flag.
1956 /* check again in case another CPU has just made room avail */
1957 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
1958 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
1962 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
1964 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
1967 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
1968 unsigned int pkts, unsigned int bytes)
1971 if (unlikely(!bytes))
1974 dql_completed(&dev_queue->dql, bytes);
1977 * Without the memory barrier there is a small possiblity that
1978 * netdev_tx_sent_queue will miss the update and cause the queue to
1979 * be stopped forever
1983 if (dql_avail(&dev_queue->dql) < 0)
1986 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
1987 netif_schedule_queue(dev_queue);
1991 static inline void netdev_completed_queue(struct net_device *dev,
1992 unsigned int pkts, unsigned int bytes)
1994 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
1997 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2000 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2005 static inline void netdev_reset_queue(struct net_device *dev_queue)
2007 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2011 * netif_running - test if up
2012 * @dev: network device
2014 * Test if the device has been brought up.
2016 static inline bool netif_running(const struct net_device *dev)
2018 return test_bit(__LINK_STATE_START, &dev->state);
2022 * Routines to manage the subqueues on a device. We only need start
2023 * stop, and a check if it's stopped. All other device management is
2024 * done at the overall netdevice level.
2025 * Also test the device if we're multiqueue.
2029 * netif_start_subqueue - allow sending packets on subqueue
2030 * @dev: network device
2031 * @queue_index: sub queue index
2033 * Start individual transmit queue of a device with multiple transmit queues.
2035 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2037 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2039 netif_tx_start_queue(txq);
2043 * netif_stop_subqueue - stop sending packets on subqueue
2044 * @dev: network device
2045 * @queue_index: sub queue index
2047 * Stop individual transmit queue of a device with multiple transmit queues.
2049 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2051 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2052 #ifdef CONFIG_NETPOLL_TRAP
2056 netif_tx_stop_queue(txq);
2060 * netif_subqueue_stopped - test status of subqueue
2061 * @dev: network device
2062 * @queue_index: sub queue index
2064 * Check individual transmit queue of a device with multiple transmit queues.
2066 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2069 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2071 return netif_tx_queue_stopped(txq);
2074 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2075 struct sk_buff *skb)
2077 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2081 * netif_wake_subqueue - allow sending packets on subqueue
2082 * @dev: network device
2083 * @queue_index: sub queue index
2085 * Resume individual transmit queue of a device with multiple transmit queues.
2087 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2089 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2090 #ifdef CONFIG_NETPOLL_TRAP
2094 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2095 __netif_schedule(txq->qdisc);
2099 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2100 * as a distribution range limit for the returned value.
2102 static inline u16 skb_tx_hash(const struct net_device *dev,
2103 const struct sk_buff *skb)
2105 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2109 * netif_is_multiqueue - test if device has multiple transmit queues
2110 * @dev: network device
2112 * Check if device has multiple transmit queues
2114 static inline bool netif_is_multiqueue(const struct net_device *dev)
2116 return dev->num_tx_queues > 1;
2119 extern int netif_set_real_num_tx_queues(struct net_device *dev,
2123 extern int netif_set_real_num_rx_queues(struct net_device *dev,
2126 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2133 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2134 const struct net_device *from_dev)
2138 err = netif_set_real_num_tx_queues(to_dev,
2139 from_dev->real_num_tx_queues);
2143 return netif_set_real_num_rx_queues(to_dev,
2144 from_dev->real_num_rx_queues);
2150 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2151 extern int netif_get_num_default_rss_queues(void);
2153 /* Use this variant when it is known for sure that it
2154 * is executing from hardware interrupt context or with hardware interrupts
2157 extern void dev_kfree_skb_irq(struct sk_buff *skb);
2159 /* Use this variant in places where it could be invoked
2160 * from either hardware interrupt or other context, with hardware interrupts
2161 * either disabled or enabled.
2163 extern void dev_kfree_skb_any(struct sk_buff *skb);
2165 extern int netif_rx(struct sk_buff *skb);
2166 extern int netif_rx_ni(struct sk_buff *skb);
2167 extern int netif_receive_skb(struct sk_buff *skb);
2168 extern gro_result_t napi_gro_receive(struct napi_struct *napi,
2169 struct sk_buff *skb);
2170 extern void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2171 extern struct sk_buff * napi_get_frags(struct napi_struct *napi);
2172 extern gro_result_t napi_gro_frags(struct napi_struct *napi);
2174 static inline void napi_free_frags(struct napi_struct *napi)
2176 kfree_skb(napi->skb);
2180 extern int netdev_rx_handler_register(struct net_device *dev,
2181 rx_handler_func_t *rx_handler,
2182 void *rx_handler_data);
2183 extern void netdev_rx_handler_unregister(struct net_device *dev);
2185 extern bool dev_valid_name(const char *name);
2186 extern int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2187 extern int dev_ethtool(struct net *net, struct ifreq *);
2188 extern unsigned int dev_get_flags(const struct net_device *);
2189 extern int __dev_change_flags(struct net_device *, unsigned int flags);
2190 extern int dev_change_flags(struct net_device *, unsigned int);
2191 extern void __dev_notify_flags(struct net_device *, unsigned int old_flags);
2192 extern int dev_change_name(struct net_device *, const char *);
2193 extern int dev_set_alias(struct net_device *, const char *, size_t);
2194 extern int dev_change_net_namespace(struct net_device *,
2195 struct net *, const char *);
2196 extern int dev_set_mtu(struct net_device *, int);
2197 extern void dev_set_group(struct net_device *, int);
2198 extern int dev_set_mac_address(struct net_device *,
2200 extern int dev_hard_start_xmit(struct sk_buff *skb,
2201 struct net_device *dev,
2202 struct netdev_queue *txq);
2203 extern int dev_forward_skb(struct net_device *dev,
2204 struct sk_buff *skb);
2206 extern int netdev_budget;
2208 /* Called by rtnetlink.c:rtnl_unlock() */
2209 extern void netdev_run_todo(void);
2212 * dev_put - release reference to device
2213 * @dev: network device
2215 * Release reference to device to allow it to be freed.
2217 static inline void dev_put(struct net_device *dev)
2219 this_cpu_dec(*dev->pcpu_refcnt);
2223 * dev_hold - get reference to device
2224 * @dev: network device
2226 * Hold reference to device to keep it from being freed.
2228 static inline void dev_hold(struct net_device *dev)
2230 this_cpu_inc(*dev->pcpu_refcnt);
2233 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2234 * and _off may be called from IRQ context, but it is caller
2235 * who is responsible for serialization of these calls.
2237 * The name carrier is inappropriate, these functions should really be
2238 * called netif_lowerlayer_*() because they represent the state of any
2239 * kind of lower layer not just hardware media.
2242 extern void linkwatch_init_dev(struct net_device *dev);
2243 extern void linkwatch_fire_event(struct net_device *dev);
2244 extern void linkwatch_forget_dev(struct net_device *dev);
2247 * netif_carrier_ok - test if carrier present
2248 * @dev: network device
2250 * Check if carrier is present on device
2252 static inline bool netif_carrier_ok(const struct net_device *dev)
2254 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2257 extern unsigned long dev_trans_start(struct net_device *dev);
2259 extern void __netdev_watchdog_up(struct net_device *dev);
2261 extern void netif_carrier_on(struct net_device *dev);
2263 extern void netif_carrier_off(struct net_device *dev);
2266 * netif_dormant_on - mark device as dormant.
2267 * @dev: network device
2269 * Mark device as dormant (as per RFC2863).
2271 * The dormant state indicates that the relevant interface is not
2272 * actually in a condition to pass packets (i.e., it is not 'up') but is
2273 * in a "pending" state, waiting for some external event. For "on-
2274 * demand" interfaces, this new state identifies the situation where the
2275 * interface is waiting for events to place it in the up state.
2278 static inline void netif_dormant_on(struct net_device *dev)
2280 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2281 linkwatch_fire_event(dev);
2285 * netif_dormant_off - set device as not dormant.
2286 * @dev: network device
2288 * Device is not in dormant state.
2290 static inline void netif_dormant_off(struct net_device *dev)
2292 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2293 linkwatch_fire_event(dev);
2297 * netif_dormant - test if carrier present
2298 * @dev: network device
2300 * Check if carrier is present on device
2302 static inline bool netif_dormant(const struct net_device *dev)
2304 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2309 * netif_oper_up - test if device is operational
2310 * @dev: network device
2312 * Check if carrier is operational
2314 static inline bool netif_oper_up(const struct net_device *dev)
2316 return (dev->operstate == IF_OPER_UP ||
2317 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2321 * netif_device_present - is device available or removed
2322 * @dev: network device
2324 * Check if device has not been removed from system.
2326 static inline bool netif_device_present(struct net_device *dev)
2328 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2331 extern void netif_device_detach(struct net_device *dev);
2333 extern void netif_device_attach(struct net_device *dev);
2336 * Network interface message level settings
2340 NETIF_MSG_DRV = 0x0001,
2341 NETIF_MSG_PROBE = 0x0002,
2342 NETIF_MSG_LINK = 0x0004,
2343 NETIF_MSG_TIMER = 0x0008,
2344 NETIF_MSG_IFDOWN = 0x0010,
2345 NETIF_MSG_IFUP = 0x0020,
2346 NETIF_MSG_RX_ERR = 0x0040,
2347 NETIF_MSG_TX_ERR = 0x0080,
2348 NETIF_MSG_TX_QUEUED = 0x0100,
2349 NETIF_MSG_INTR = 0x0200,
2350 NETIF_MSG_TX_DONE = 0x0400,
2351 NETIF_MSG_RX_STATUS = 0x0800,
2352 NETIF_MSG_PKTDATA = 0x1000,
2353 NETIF_MSG_HW = 0x2000,
2354 NETIF_MSG_WOL = 0x4000,
2357 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2358 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2359 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2360 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2361 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2362 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2363 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2364 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2365 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2366 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2367 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2368 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2369 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2370 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2371 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2373 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2376 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2377 return default_msg_enable_bits;
2378 if (debug_value == 0) /* no output */
2380 /* set low N bits */
2381 return (1 << debug_value) - 1;
2384 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2386 spin_lock(&txq->_xmit_lock);
2387 txq->xmit_lock_owner = cpu;
2390 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2392 spin_lock_bh(&txq->_xmit_lock);
2393 txq->xmit_lock_owner = smp_processor_id();
2396 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2398 bool ok = spin_trylock(&txq->_xmit_lock);
2400 txq->xmit_lock_owner = smp_processor_id();
2404 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2406 txq->xmit_lock_owner = -1;
2407 spin_unlock(&txq->_xmit_lock);
2410 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2412 txq->xmit_lock_owner = -1;
2413 spin_unlock_bh(&txq->_xmit_lock);
2416 static inline void txq_trans_update(struct netdev_queue *txq)
2418 if (txq->xmit_lock_owner != -1)
2419 txq->trans_start = jiffies;
2423 * netif_tx_lock - grab network device transmit lock
2424 * @dev: network device
2426 * Get network device transmit lock
2428 static inline void netif_tx_lock(struct net_device *dev)
2433 spin_lock(&dev->tx_global_lock);
2434 cpu = smp_processor_id();
2435 for (i = 0; i < dev->num_tx_queues; i++) {
2436 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2438 /* We are the only thread of execution doing a
2439 * freeze, but we have to grab the _xmit_lock in
2440 * order to synchronize with threads which are in
2441 * the ->hard_start_xmit() handler and already
2442 * checked the frozen bit.
2444 __netif_tx_lock(txq, cpu);
2445 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2446 __netif_tx_unlock(txq);
2450 static inline void netif_tx_lock_bh(struct net_device *dev)
2456 static inline void netif_tx_unlock(struct net_device *dev)
2460 for (i = 0; i < dev->num_tx_queues; i++) {
2461 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2463 /* No need to grab the _xmit_lock here. If the
2464 * queue is not stopped for another reason, we
2467 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2468 netif_schedule_queue(txq);
2470 spin_unlock(&dev->tx_global_lock);
2473 static inline void netif_tx_unlock_bh(struct net_device *dev)
2475 netif_tx_unlock(dev);
2479 #define HARD_TX_LOCK(dev, txq, cpu) { \
2480 if ((dev->features & NETIF_F_LLTX) == 0) { \
2481 __netif_tx_lock(txq, cpu); \
2485 #define HARD_TX_UNLOCK(dev, txq) { \
2486 if ((dev->features & NETIF_F_LLTX) == 0) { \
2487 __netif_tx_unlock(txq); \
2491 static inline void netif_tx_disable(struct net_device *dev)
2497 cpu = smp_processor_id();
2498 for (i = 0; i < dev->num_tx_queues; i++) {
2499 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2501 __netif_tx_lock(txq, cpu);
2502 netif_tx_stop_queue(txq);
2503 __netif_tx_unlock(txq);
2508 static inline void netif_addr_lock(struct net_device *dev)
2510 spin_lock(&dev->addr_list_lock);
2513 static inline void netif_addr_lock_nested(struct net_device *dev)
2515 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2518 static inline void netif_addr_lock_bh(struct net_device *dev)
2520 spin_lock_bh(&dev->addr_list_lock);
2523 static inline void netif_addr_unlock(struct net_device *dev)
2525 spin_unlock(&dev->addr_list_lock);
2528 static inline void netif_addr_unlock_bh(struct net_device *dev)
2530 spin_unlock_bh(&dev->addr_list_lock);
2534 * dev_addrs walker. Should be used only for read access. Call with
2535 * rcu_read_lock held.
2537 #define for_each_dev_addr(dev, ha) \
2538 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2540 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2542 extern void ether_setup(struct net_device *dev);
2544 /* Support for loadable net-drivers */
2545 extern struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2546 void (*setup)(struct net_device *),
2547 unsigned int txqs, unsigned int rxqs);
2548 #define alloc_netdev(sizeof_priv, name, setup) \
2549 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2551 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2552 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2554 extern int register_netdev(struct net_device *dev);
2555 extern void unregister_netdev(struct net_device *dev);
2557 /* General hardware address lists handling functions */
2558 extern int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2559 struct netdev_hw_addr_list *from_list,
2560 int addr_len, unsigned char addr_type);
2561 extern void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2562 struct netdev_hw_addr_list *from_list,
2563 int addr_len, unsigned char addr_type);
2564 extern int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2565 struct netdev_hw_addr_list *from_list,
2567 extern void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2568 struct netdev_hw_addr_list *from_list,
2570 extern void __hw_addr_flush(struct netdev_hw_addr_list *list);
2571 extern void __hw_addr_init(struct netdev_hw_addr_list *list);
2573 /* Functions used for device addresses handling */
2574 extern int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2575 unsigned char addr_type);
2576 extern int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2577 unsigned char addr_type);
2578 extern int dev_addr_add_multiple(struct net_device *to_dev,
2579 struct net_device *from_dev,
2580 unsigned char addr_type);
2581 extern int dev_addr_del_multiple(struct net_device *to_dev,
2582 struct net_device *from_dev,
2583 unsigned char addr_type);
2584 extern void dev_addr_flush(struct net_device *dev);
2585 extern int dev_addr_init(struct net_device *dev);
2587 /* Functions used for unicast addresses handling */
2588 extern int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2589 extern int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2590 extern int dev_uc_del(struct net_device *dev, const unsigned char *addr);
2591 extern int dev_uc_sync(struct net_device *to, struct net_device *from);
2592 extern void dev_uc_unsync(struct net_device *to, struct net_device *from);
2593 extern void dev_uc_flush(struct net_device *dev);
2594 extern void dev_uc_init(struct net_device *dev);
2596 /* Functions used for multicast addresses handling */
2597 extern int dev_mc_add(struct net_device *dev, const unsigned char *addr);
2598 extern int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
2599 extern int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
2600 extern int dev_mc_del(struct net_device *dev, const unsigned char *addr);
2601 extern int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
2602 extern int dev_mc_sync(struct net_device *to, struct net_device *from);
2603 extern void dev_mc_unsync(struct net_device *to, struct net_device *from);
2604 extern void dev_mc_flush(struct net_device *dev);
2605 extern void dev_mc_init(struct net_device *dev);
2607 /* Functions used for secondary unicast and multicast support */
2608 extern void dev_set_rx_mode(struct net_device *dev);
2609 extern void __dev_set_rx_mode(struct net_device *dev);
2610 extern int dev_set_promiscuity(struct net_device *dev, int inc);
2611 extern int dev_set_allmulti(struct net_device *dev, int inc);
2612 extern void netdev_state_change(struct net_device *dev);
2613 extern void netdev_notify_peers(struct net_device *dev);
2614 extern void netdev_features_change(struct net_device *dev);
2615 /* Load a device via the kmod */
2616 extern void dev_load(struct net *net, const char *name);
2617 extern void dev_mcast_init(void);
2618 extern struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2619 struct rtnl_link_stats64 *storage);
2620 extern void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
2621 const struct net_device_stats *netdev_stats);
2623 extern int netdev_max_backlog;
2624 extern int netdev_tstamp_prequeue;
2625 extern int weight_p;
2626 extern int bpf_jit_enable;
2627 extern int netdev_set_master(struct net_device *dev, struct net_device *master);
2628 extern int netdev_set_bond_master(struct net_device *dev,
2629 struct net_device *master);
2630 extern int skb_checksum_help(struct sk_buff *skb);
2631 extern struct sk_buff *skb_gso_segment(struct sk_buff *skb,
2632 netdev_features_t features);
2634 extern void netdev_rx_csum_fault(struct net_device *dev);
2636 static inline void netdev_rx_csum_fault(struct net_device *dev)
2640 /* rx skb timestamps */
2641 extern void net_enable_timestamp(void);
2642 extern void net_disable_timestamp(void);
2644 #ifdef CONFIG_PROC_FS
2645 extern void *dev_seq_start(struct seq_file *seq, loff_t *pos);
2646 extern void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos);
2647 extern void dev_seq_stop(struct seq_file *seq, void *v);
2650 extern int netdev_class_create_file(struct class_attribute *class_attr);
2651 extern void netdev_class_remove_file(struct class_attribute *class_attr);
2653 extern struct kobj_ns_type_operations net_ns_type_operations;
2655 extern const char *netdev_drivername(const struct net_device *dev);
2657 extern void linkwatch_run_queue(void);
2659 static inline netdev_features_t netdev_get_wanted_features(
2660 struct net_device *dev)
2662 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2664 netdev_features_t netdev_increment_features(netdev_features_t all,
2665 netdev_features_t one, netdev_features_t mask);
2666 int __netdev_update_features(struct net_device *dev);
2667 void netdev_update_features(struct net_device *dev);
2668 void netdev_change_features(struct net_device *dev);
2670 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
2671 struct net_device *dev);
2673 netdev_features_t netif_skb_features(struct sk_buff *skb);
2675 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
2677 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
2679 /* check flags correspondence */
2680 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
2681 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
2682 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
2683 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
2684 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
2685 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
2687 return (features & feature) == feature;
2690 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
2692 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
2693 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
2696 static inline bool netif_needs_gso(struct sk_buff *skb,
2697 netdev_features_t features)
2699 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
2700 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
2701 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
2704 static inline void netif_set_gso_max_size(struct net_device *dev,
2707 dev->gso_max_size = size;
2710 static inline bool netif_is_bond_slave(struct net_device *dev)
2712 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
2715 static inline bool netif_supports_nofcs(struct net_device *dev)
2717 return dev->priv_flags & IFF_SUPP_NOFCS;
2720 extern struct pernet_operations __net_initdata loopback_net_ops;
2722 /* Logging, debugging and troubleshooting/diagnostic helpers. */
2724 /* netdev_printk helpers, similar to dev_printk */
2726 static inline const char *netdev_name(const struct net_device *dev)
2728 if (dev->reg_state != NETREG_REGISTERED)
2729 return "(unregistered net_device)";
2733 extern __printf(3, 4)
2734 int netdev_printk(const char *level, const struct net_device *dev,
2735 const char *format, ...);
2736 extern __printf(2, 3)
2737 int netdev_emerg(const struct net_device *dev, const char *format, ...);
2738 extern __printf(2, 3)
2739 int netdev_alert(const struct net_device *dev, const char *format, ...);
2740 extern __printf(2, 3)
2741 int netdev_crit(const struct net_device *dev, const char *format, ...);
2742 extern __printf(2, 3)
2743 int netdev_err(const struct net_device *dev, const char *format, ...);
2744 extern __printf(2, 3)
2745 int netdev_warn(const struct net_device *dev, const char *format, ...);
2746 extern __printf(2, 3)
2747 int netdev_notice(const struct net_device *dev, const char *format, ...);
2748 extern __printf(2, 3)
2749 int netdev_info(const struct net_device *dev, const char *format, ...);
2751 #define MODULE_ALIAS_NETDEV(device) \
2752 MODULE_ALIAS("netdev-" device)
2754 #if defined(CONFIG_DYNAMIC_DEBUG)
2755 #define netdev_dbg(__dev, format, args...) \
2757 dynamic_netdev_dbg(__dev, format, ##args); \
2759 #elif defined(DEBUG)
2760 #define netdev_dbg(__dev, format, args...) \
2761 netdev_printk(KERN_DEBUG, __dev, format, ##args)
2763 #define netdev_dbg(__dev, format, args...) \
2766 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
2771 #if defined(VERBOSE_DEBUG)
2772 #define netdev_vdbg netdev_dbg
2775 #define netdev_vdbg(dev, format, args...) \
2778 netdev_printk(KERN_DEBUG, dev, format, ##args); \
2784 * netdev_WARN() acts like dev_printk(), but with the key difference
2785 * of using a WARN/WARN_ON to get the message out, including the
2786 * file/line information and a backtrace.
2788 #define netdev_WARN(dev, format, args...) \
2789 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args);
2791 /* netif printk helpers, similar to netdev_printk */
2793 #define netif_printk(priv, type, level, dev, fmt, args...) \
2795 if (netif_msg_##type(priv)) \
2796 netdev_printk(level, (dev), fmt, ##args); \
2799 #define netif_level(level, priv, type, dev, fmt, args...) \
2801 if (netif_msg_##type(priv)) \
2802 netdev_##level(dev, fmt, ##args); \
2805 #define netif_emerg(priv, type, dev, fmt, args...) \
2806 netif_level(emerg, priv, type, dev, fmt, ##args)
2807 #define netif_alert(priv, type, dev, fmt, args...) \
2808 netif_level(alert, priv, type, dev, fmt, ##args)
2809 #define netif_crit(priv, type, dev, fmt, args...) \
2810 netif_level(crit, priv, type, dev, fmt, ##args)
2811 #define netif_err(priv, type, dev, fmt, args...) \
2812 netif_level(err, priv, type, dev, fmt, ##args)
2813 #define netif_warn(priv, type, dev, fmt, args...) \
2814 netif_level(warn, priv, type, dev, fmt, ##args)
2815 #define netif_notice(priv, type, dev, fmt, args...) \
2816 netif_level(notice, priv, type, dev, fmt, ##args)
2817 #define netif_info(priv, type, dev, fmt, args...) \
2818 netif_level(info, priv, type, dev, fmt, ##args)
2820 #if defined(CONFIG_DYNAMIC_DEBUG)
2821 #define netif_dbg(priv, type, netdev, format, args...) \
2823 if (netif_msg_##type(priv)) \
2824 dynamic_netdev_dbg(netdev, format, ##args); \
2826 #elif defined(DEBUG)
2827 #define netif_dbg(priv, type, dev, format, args...) \
2828 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
2830 #define netif_dbg(priv, type, dev, format, args...) \
2833 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2838 #if defined(VERBOSE_DEBUG)
2839 #define netif_vdbg netif_dbg
2841 #define netif_vdbg(priv, type, dev, format, args...) \
2844 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2849 #endif /* _LINUX_NETDEVICE_H */