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
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 void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct ethtool_ops *ops);
66 /* Backlog congestion levels */
67 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
68 #define NET_RX_DROP 1 /* packet dropped */
71 * Transmit return codes: transmit return codes originate from three different
74 * - qdisc return codes
75 * - driver transmit return codes
78 * Drivers are allowed to return any one of those in their hard_start_xmit()
79 * function. Real network devices commonly used with qdiscs should only return
80 * the driver transmit return codes though - when qdiscs are used, the actual
81 * transmission happens asynchronously, so the value is not propagated to
82 * higher layers. Virtual network devices transmit synchronously, in this case
83 * the driver transmit return codes are consumed by dev_queue_xmit(), all
84 * others are propagated to higher layers.
87 /* qdisc ->enqueue() return codes. */
88 #define NET_XMIT_SUCCESS 0x00
89 #define NET_XMIT_DROP 0x01 /* skb dropped */
90 #define NET_XMIT_CN 0x02 /* congestion notification */
91 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
92 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
94 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
95 * indicates that the device will soon be dropping packets, or already drops
96 * some packets of the same priority; prompting us to send less aggressively. */
97 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
98 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100 /* Driver transmit return codes */
101 #define NETDEV_TX_MASK 0xf0
104 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
105 NETDEV_TX_OK = 0x00, /* driver took care of packet */
106 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
107 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
109 typedef enum netdev_tx netdev_tx_t;
112 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
113 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 static inline bool dev_xmit_complete(int rc)
118 * Positive cases with an skb consumed by a driver:
119 * - successful transmission (rc == NETDEV_TX_OK)
120 * - error while transmitting (rc < 0)
121 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 if (likely(rc < NET_XMIT_MASK))
130 * Compute the worst case header length according to the protocols
134 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
135 # if defined(CONFIG_MAC80211_MESH)
136 # define LL_MAX_HEADER 128
138 # define LL_MAX_HEADER 96
141 # define LL_MAX_HEADER 32
144 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
145 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
146 #define MAX_HEADER LL_MAX_HEADER
148 #define MAX_HEADER (LL_MAX_HEADER + 48)
152 * Old network device statistics. Fields are native words
153 * (unsigned long) so they can be read and written atomically.
156 struct net_device_stats {
157 unsigned long rx_packets;
158 unsigned long tx_packets;
159 unsigned long rx_bytes;
160 unsigned long tx_bytes;
161 unsigned long rx_errors;
162 unsigned long tx_errors;
163 unsigned long rx_dropped;
164 unsigned long tx_dropped;
165 unsigned long multicast;
166 unsigned long collisions;
167 unsigned long rx_length_errors;
168 unsigned long rx_over_errors;
169 unsigned long rx_crc_errors;
170 unsigned long rx_frame_errors;
171 unsigned long rx_fifo_errors;
172 unsigned long rx_missed_errors;
173 unsigned long tx_aborted_errors;
174 unsigned long tx_carrier_errors;
175 unsigned long tx_fifo_errors;
176 unsigned long tx_heartbeat_errors;
177 unsigned long tx_window_errors;
178 unsigned long rx_compressed;
179 unsigned long tx_compressed;
183 #include <linux/cache.h>
184 #include <linux/skbuff.h>
187 #include <linux/static_key.h>
188 extern struct static_key rps_needed;
195 struct netdev_hw_addr {
196 struct list_head list;
197 unsigned char addr[MAX_ADDR_LEN];
199 #define NETDEV_HW_ADDR_T_LAN 1
200 #define NETDEV_HW_ADDR_T_SAN 2
201 #define NETDEV_HW_ADDR_T_SLAVE 3
202 #define NETDEV_HW_ADDR_T_UNICAST 4
203 #define NETDEV_HW_ADDR_T_MULTICAST 5
208 struct rcu_head rcu_head;
211 struct netdev_hw_addr_list {
212 struct list_head list;
216 #define netdev_hw_addr_list_count(l) ((l)->count)
217 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
218 #define netdev_hw_addr_list_for_each(ha, l) \
219 list_for_each_entry(ha, &(l)->list, list)
221 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
222 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
223 #define netdev_for_each_uc_addr(ha, dev) \
224 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
226 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
227 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
228 #define netdev_for_each_mc_addr(ha, dev) \
229 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
236 /* cached hardware header; allow for machine alignment needs. */
237 #define HH_DATA_MOD 16
238 #define HH_DATA_OFF(__len) \
239 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
240 #define HH_DATA_ALIGN(__len) \
241 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
242 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
245 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * dev->hard_header_len ? (dev->hard_header_len +
248 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 * We could use other alignment values, but we must maintain the
251 * relationship HH alignment <= LL alignment.
253 #define LL_RESERVED_SPACE(dev) \
254 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
255 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
256 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
259 int (*create) (struct sk_buff *skb, struct net_device *dev,
260 unsigned short type, const void *daddr,
261 const void *saddr, unsigned int len);
262 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
263 int (*rebuild)(struct sk_buff *skb);
264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
270 /* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
275 enum netdev_state_t {
277 __LINK_STATE_PRESENT,
278 __LINK_STATE_NOCARRIER,
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
285 * This structure holds at boot time configured netdevice settings. They
286 * are then used in the device probing.
288 struct netdev_boot_setup {
292 #define NETDEV_BOOT_SETUP_MAX 8
294 int __init netdev_boot_setup(char *str);
297 * Structure for NAPI scheduling similar to tasklet but with weighting
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
306 struct list_head poll_list;
310 unsigned int gro_count;
311 int (*poll)(struct napi_struct *, int);
312 #ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
316 struct net_device *dev;
317 struct sk_buff *gro_list;
319 struct list_head dev_list;
320 struct hlist_node napi_hash_node;
321 unsigned int napi_id;
325 NAPI_STATE_SCHED, /* Poll is scheduled */
326 NAPI_STATE_DISABLE, /* Disable pending */
327 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
328 NAPI_STATE_HASHED, /* In NAPI hash */
338 typedef enum gro_result gro_result_t;
341 * enum rx_handler_result - Possible return values for rx_handlers.
342 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
345 * case skb->dev was changed by rx_handler.
346 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
347 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 * rx_handlers are functions called from inside __netif_receive_skb(), to do
350 * special processing of the skb, prior to delivery to protocol handlers.
352 * Currently, a net_device can only have a single rx_handler registered. Trying
353 * to register a second rx_handler will return -EBUSY.
355 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
356 * To unregister a rx_handler on a net_device, use
357 * netdev_rx_handler_unregister().
359 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
362 * If the rx_handler consumed to skb in some way, it should return
363 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
364 * the skb to be delivered in some other ways.
366 * If the rx_handler changed skb->dev, to divert the skb to another
367 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
368 * new device will be called if it exists.
370 * If the rx_handler consider the skb should be ignored, it should return
371 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
372 * are registered on exact device (ptype->dev == skb->dev).
374 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
375 * delivered, it should return RX_HANDLER_PASS.
377 * A device without a registered rx_handler will behave as if rx_handler
378 * returned RX_HANDLER_PASS.
381 enum rx_handler_result {
387 typedef enum rx_handler_result rx_handler_result_t;
388 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
390 void __napi_schedule(struct napi_struct *n);
392 static inline bool napi_disable_pending(struct napi_struct *n)
394 return test_bit(NAPI_STATE_DISABLE, &n->state);
398 * napi_schedule_prep - check if napi can be scheduled
401 * Test if NAPI routine is already running, and if not mark
402 * it as running. This is used as a condition variable
403 * insure only one NAPI poll instance runs. We also make
404 * sure there is no pending NAPI disable.
406 static inline bool napi_schedule_prep(struct napi_struct *n)
408 return !napi_disable_pending(n) &&
409 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
413 * napi_schedule - schedule NAPI poll
416 * Schedule NAPI poll routine to be called if it is not already
419 static inline void napi_schedule(struct napi_struct *n)
421 if (napi_schedule_prep(n))
425 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
426 static inline bool napi_reschedule(struct napi_struct *napi)
428 if (napi_schedule_prep(napi)) {
429 __napi_schedule(napi);
436 * napi_complete - NAPI processing complete
439 * Mark NAPI processing as complete.
441 void __napi_complete(struct napi_struct *n);
442 void napi_complete(struct napi_struct *n);
445 * napi_by_id - lookup a NAPI by napi_id
446 * @napi_id: hashed napi_id
448 * lookup @napi_id in napi_hash table
449 * must be called under rcu_read_lock()
451 struct napi_struct *napi_by_id(unsigned int napi_id);
454 * napi_hash_add - add a NAPI to global hashtable
455 * @napi: napi context
457 * generate a new napi_id and store a @napi under it in napi_hash
459 void napi_hash_add(struct napi_struct *napi);
462 * napi_hash_del - remove a NAPI from global table
463 * @napi: napi context
465 * Warning: caller must observe rcu grace period
466 * before freeing memory containing @napi
468 void napi_hash_del(struct napi_struct *napi);
471 * napi_disable - prevent NAPI from scheduling
474 * Stop NAPI from being scheduled on this context.
475 * Waits till any outstanding processing completes.
477 static inline void napi_disable(struct napi_struct *n)
480 set_bit(NAPI_STATE_DISABLE, &n->state);
481 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
483 clear_bit(NAPI_STATE_DISABLE, &n->state);
487 * napi_enable - enable NAPI scheduling
490 * Resume NAPI from being scheduled on this context.
491 * Must be paired with napi_disable.
493 static inline void napi_enable(struct napi_struct *n)
495 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
496 smp_mb__before_clear_bit();
497 clear_bit(NAPI_STATE_SCHED, &n->state);
502 * napi_synchronize - wait until NAPI is not running
505 * Wait until NAPI is done being scheduled on this context.
506 * Waits till any outstanding processing completes but
507 * does not disable future activations.
509 static inline void napi_synchronize(const struct napi_struct *n)
511 while (test_bit(NAPI_STATE_SCHED, &n->state))
515 # define napi_synchronize(n) barrier()
518 enum netdev_queue_state_t {
519 __QUEUE_STATE_DRV_XOFF,
520 __QUEUE_STATE_STACK_XOFF,
521 __QUEUE_STATE_FROZEN,
524 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
525 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
526 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
528 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
529 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
531 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
535 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
536 * netif_tx_* functions below are used to manipulate this flag. The
537 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
538 * queue independently. The netif_xmit_*stopped functions below are called
539 * to check if the queue has been stopped by the driver or stack (either
540 * of the XOFF bits are set in the state). Drivers should not need to call
541 * netif_xmit*stopped functions, they should only be using netif_tx_*.
544 struct netdev_queue {
548 struct net_device *dev;
550 struct Qdisc *qdisc_sleeping;
554 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
560 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
563 * please use this field instead of dev->trans_start
565 unsigned long trans_start;
568 * Number of TX timeouts for this queue
569 * (/sys/class/net/DEV/Q/trans_timeout)
571 unsigned long trans_timeout;
578 } ____cacheline_aligned_in_smp;
580 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
582 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
589 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
591 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
598 * This structure holds an RPS map which can be of variable length. The
599 * map is an array of CPUs.
606 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
609 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
610 * tail pointer for that CPU's input queue at the time of last enqueue, and
611 * a hardware filter index.
613 struct rps_dev_flow {
616 unsigned int last_qtail;
618 #define RPS_NO_FILTER 0xffff
621 * The rps_dev_flow_table structure contains a table of flow mappings.
623 struct rps_dev_flow_table {
626 struct rps_dev_flow flows[0];
628 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
629 ((_num) * sizeof(struct rps_dev_flow)))
632 * The rps_sock_flow_table contains mappings of flows to the last CPU
633 * on which they were processed by the application (set in recvmsg).
635 struct rps_sock_flow_table {
639 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
640 ((_num) * sizeof(u16)))
642 #define RPS_NO_CPU 0xffff
644 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
648 unsigned int cpu, index = hash & table->mask;
650 /* We only give a hint, preemption can change cpu under us */
651 cpu = raw_smp_processor_id();
653 if (table->ents[index] != cpu)
654 table->ents[index] = cpu;
658 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
662 table->ents[hash & table->mask] = RPS_NO_CPU;
665 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
667 #ifdef CONFIG_RFS_ACCEL
668 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
671 #endif /* CONFIG_RPS */
673 /* This structure contains an instance of an RX queue. */
674 struct netdev_rx_queue {
676 struct rps_map __rcu *rps_map;
677 struct rps_dev_flow_table __rcu *rps_flow_table;
680 struct net_device *dev;
681 } ____cacheline_aligned_in_smp;
684 * RX queue sysfs structures and functions.
686 struct rx_queue_attribute {
687 struct attribute attr;
688 ssize_t (*show)(struct netdev_rx_queue *queue,
689 struct rx_queue_attribute *attr, char *buf);
690 ssize_t (*store)(struct netdev_rx_queue *queue,
691 struct rx_queue_attribute *attr, const char *buf, size_t len);
696 * This structure holds an XPS map which can be of variable length. The
697 * map is an array of queues.
701 unsigned int alloc_len;
705 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
706 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
710 * This structure holds all XPS maps for device. Maps are indexed by CPU.
712 struct xps_dev_maps {
714 struct xps_map __rcu *cpu_map[0];
716 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
717 (nr_cpu_ids * sizeof(struct xps_map *)))
718 #endif /* CONFIG_XPS */
720 #define TC_MAX_QUEUE 16
721 #define TC_BITMASK 15
722 /* HW offloaded queuing disciplines txq count and offset maps */
723 struct netdev_tc_txq {
728 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
730 * This structure is to hold information about the device
731 * configured to run FCoE protocol stack.
733 struct netdev_fcoe_hbainfo {
734 char manufacturer[64];
735 char serial_number[64];
736 char hardware_version[64];
737 char driver_version[64];
738 char optionrom_version[64];
739 char firmware_version[64];
741 char model_description[256];
745 #define MAX_PHYS_PORT_ID_LEN 32
747 /* This structure holds a unique identifier to identify the
748 * physical port used by a netdevice.
750 struct netdev_phys_port_id {
751 unsigned char id[MAX_PHYS_PORT_ID_LEN];
752 unsigned char id_len;
755 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
756 struct sk_buff *skb);
759 * This structure defines the management hooks for network devices.
760 * The following hooks can be defined; unless noted otherwise, they are
761 * optional and can be filled with a null pointer.
763 * int (*ndo_init)(struct net_device *dev);
764 * This function is called once when network device is registered.
765 * The network device can use this to any late stage initializaton
766 * or semantic validattion. It can fail with an error code which will
767 * be propogated back to register_netdev
769 * void (*ndo_uninit)(struct net_device *dev);
770 * This function is called when device is unregistered or when registration
771 * fails. It is not called if init fails.
773 * int (*ndo_open)(struct net_device *dev);
774 * This function is called when network device transistions to the up
777 * int (*ndo_stop)(struct net_device *dev);
778 * This function is called when network device transistions to the down
781 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
782 * struct net_device *dev);
783 * Called when a packet needs to be transmitted.
784 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
785 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
786 * Required can not be NULL.
788 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
789 * void *accel_priv, select_queue_fallback_t fallback);
790 * Called to decide which queue to when device supports multiple
793 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
794 * This function is called to allow device receiver to make
795 * changes to configuration when multicast or promiscious is enabled.
797 * void (*ndo_set_rx_mode)(struct net_device *dev);
798 * This function is called device changes address list filtering.
799 * If driver handles unicast address filtering, it should set
800 * IFF_UNICAST_FLT to its priv_flags.
802 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
803 * This function is called when the Media Access Control address
804 * needs to be changed. If this interface is not defined, the
805 * mac address can not be changed.
807 * int (*ndo_validate_addr)(struct net_device *dev);
808 * Test if Media Access Control address is valid for the device.
810 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
811 * Called when a user request an ioctl which can't be handled by
812 * the generic interface code. If not defined ioctl's return
813 * not supported error code.
815 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
816 * Used to set network devices bus interface parameters. This interface
817 * is retained for legacy reason, new devices should use the bus
818 * interface (PCI) for low level management.
820 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
821 * Called when a user wants to change the Maximum Transfer Unit
822 * of a device. If not defined, any request to change MTU will
823 * will return an error.
825 * void (*ndo_tx_timeout)(struct net_device *dev);
826 * Callback uses when the transmitter has not made any progress
827 * for dev->watchdog ticks.
829 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
830 * struct rtnl_link_stats64 *storage);
831 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
832 * Called when a user wants to get the network device usage
833 * statistics. Drivers must do one of the following:
834 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
835 * rtnl_link_stats64 structure passed by the caller.
836 * 2. Define @ndo_get_stats to update a net_device_stats structure
837 * (which should normally be dev->stats) and return a pointer to
838 * it. The structure may be changed asynchronously only if each
839 * field is written atomically.
840 * 3. Update dev->stats asynchronously and atomically, and define
843 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
844 * If device support VLAN filtering this function is called when a
845 * VLAN id is registered.
847 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
848 * If device support VLAN filtering this function is called when a
849 * VLAN id is unregistered.
851 * void (*ndo_poll_controller)(struct net_device *dev);
853 * SR-IOV management functions.
854 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
855 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
856 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
857 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
858 * int (*ndo_get_vf_config)(struct net_device *dev,
859 * int vf, struct ifla_vf_info *ivf);
860 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
861 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
862 * struct nlattr *port[]);
863 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
864 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
865 * Called to setup 'tc' number of traffic classes in the net device. This
866 * is always called from the stack with the rtnl lock held and netif tx
867 * queues stopped. This allows the netdevice to perform queue management
870 * Fiber Channel over Ethernet (FCoE) offload functions.
871 * int (*ndo_fcoe_enable)(struct net_device *dev);
872 * Called when the FCoE protocol stack wants to start using LLD for FCoE
873 * so the underlying device can perform whatever needed configuration or
874 * initialization to support acceleration of FCoE traffic.
876 * int (*ndo_fcoe_disable)(struct net_device *dev);
877 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
878 * so the underlying device can perform whatever needed clean-ups to
879 * stop supporting acceleration of FCoE traffic.
881 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
882 * struct scatterlist *sgl, unsigned int sgc);
883 * Called when the FCoE Initiator wants to initialize an I/O that
884 * is a possible candidate for Direct Data Placement (DDP). The LLD can
885 * perform necessary setup and returns 1 to indicate the device is set up
886 * successfully to perform DDP on this I/O, otherwise this returns 0.
888 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
889 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
890 * indicated by the FC exchange id 'xid', so the underlying device can
891 * clean up and reuse resources for later DDP requests.
893 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
894 * struct scatterlist *sgl, unsigned int sgc);
895 * Called when the FCoE Target wants to initialize an I/O that
896 * is a possible candidate for Direct Data Placement (DDP). The LLD can
897 * perform necessary setup and returns 1 to indicate the device is set up
898 * successfully to perform DDP on this I/O, otherwise this returns 0.
900 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
901 * struct netdev_fcoe_hbainfo *hbainfo);
902 * Called when the FCoE Protocol stack wants information on the underlying
903 * device. This information is utilized by the FCoE protocol stack to
904 * register attributes with Fiber Channel management service as per the
905 * FC-GS Fabric Device Management Information(FDMI) specification.
907 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
908 * Called when the underlying device wants to override default World Wide
909 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
910 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
911 * protocol stack to use.
914 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
915 * u16 rxq_index, u32 flow_id);
916 * Set hardware filter for RFS. rxq_index is the target queue index;
917 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
918 * Return the filter ID on success, or a negative error code.
920 * Slave management functions (for bridge, bonding, etc).
921 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
922 * Called to make another netdev an underling.
924 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
925 * Called to release previously enslaved netdev.
927 * Feature/offload setting functions.
928 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
929 * netdev_features_t features);
930 * Adjusts the requested feature flags according to device-specific
931 * constraints, and returns the resulting flags. Must not modify
934 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
935 * Called to update device configuration to new features. Passed
936 * feature set might be less than what was returned by ndo_fix_features()).
937 * Must return >0 or -errno if it changed dev->features itself.
939 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
940 * struct net_device *dev,
941 * const unsigned char *addr, u16 flags)
942 * Adds an FDB entry to dev for addr.
943 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
944 * struct net_device *dev,
945 * const unsigned char *addr)
946 * Deletes the FDB entry from dev coresponding to addr.
947 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
948 * struct net_device *dev, int idx)
949 * Used to add FDB entries to dump requests. Implementers should add
950 * entries to skb and update idx with the number of entries.
952 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
953 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
954 * struct net_device *dev, u32 filter_mask)
956 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
957 * Called to change device carrier. Soft-devices (like dummy, team, etc)
958 * which do not represent real hardware may define this to allow their
959 * userspace components to manage their virtual carrier state. Devices
960 * that determine carrier state from physical hardware properties (eg
961 * network cables) or protocol-dependent mechanisms (eg
962 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
964 * int (*ndo_get_phys_port_id)(struct net_device *dev,
965 * struct netdev_phys_port_id *ppid);
966 * Called to get ID of physical port of this device. If driver does
967 * not implement this, it is assumed that the hw is not able to have
968 * multiple net devices on single physical port.
970 * void (*ndo_add_vxlan_port)(struct net_device *dev,
971 * sa_family_t sa_family, __be16 port);
972 * Called by vxlan to notiy a driver about the UDP port and socket
973 * address family that vxlan is listnening to. It is called only when
974 * a new port starts listening. The operation is protected by the
975 * vxlan_net->sock_lock.
977 * void (*ndo_del_vxlan_port)(struct net_device *dev,
978 * sa_family_t sa_family, __be16 port);
979 * Called by vxlan to notify the driver about a UDP port and socket
980 * address family that vxlan is not listening to anymore. The operation
981 * is protected by the vxlan_net->sock_lock.
983 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
984 * struct net_device *dev)
985 * Called by upper layer devices to accelerate switching or other
986 * station functionality into hardware. 'pdev is the lowerdev
987 * to use for the offload and 'dev' is the net device that will
988 * back the offload. Returns a pointer to the private structure
989 * the upper layer will maintain.
990 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
991 * Called by upper layer device to delete the station created
992 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
993 * the station and priv is the structure returned by the add
995 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
996 * struct net_device *dev,
998 * Callback to use for xmit over the accelerated station. This
999 * is used in place of ndo_start_xmit on accelerated net
1002 struct net_device_ops {
1003 int (*ndo_init)(struct net_device *dev);
1004 void (*ndo_uninit)(struct net_device *dev);
1005 int (*ndo_open)(struct net_device *dev);
1006 int (*ndo_stop)(struct net_device *dev);
1007 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1008 struct net_device *dev);
1009 u16 (*ndo_select_queue)(struct net_device *dev,
1010 struct sk_buff *skb,
1012 select_queue_fallback_t fallback);
1013 void (*ndo_change_rx_flags)(struct net_device *dev,
1015 void (*ndo_set_rx_mode)(struct net_device *dev);
1016 int (*ndo_set_mac_address)(struct net_device *dev,
1018 int (*ndo_validate_addr)(struct net_device *dev);
1019 int (*ndo_do_ioctl)(struct net_device *dev,
1020 struct ifreq *ifr, int cmd);
1021 int (*ndo_set_config)(struct net_device *dev,
1023 int (*ndo_change_mtu)(struct net_device *dev,
1025 int (*ndo_neigh_setup)(struct net_device *dev,
1026 struct neigh_parms *);
1027 void (*ndo_tx_timeout) (struct net_device *dev);
1029 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1030 struct rtnl_link_stats64 *storage);
1031 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1033 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1034 __be16 proto, u16 vid);
1035 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1036 __be16 proto, u16 vid);
1037 #ifdef CONFIG_NET_POLL_CONTROLLER
1038 void (*ndo_poll_controller)(struct net_device *dev);
1039 int (*ndo_netpoll_setup)(struct net_device *dev,
1040 struct netpoll_info *info);
1041 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1043 #ifdef CONFIG_NET_RX_BUSY_POLL
1044 int (*ndo_busy_poll)(struct napi_struct *dev);
1046 int (*ndo_set_vf_mac)(struct net_device *dev,
1047 int queue, u8 *mac);
1048 int (*ndo_set_vf_vlan)(struct net_device *dev,
1049 int queue, u16 vlan, u8 qos);
1050 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
1052 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1053 int vf, bool setting);
1054 int (*ndo_get_vf_config)(struct net_device *dev,
1056 struct ifla_vf_info *ivf);
1057 int (*ndo_set_vf_link_state)(struct net_device *dev,
1058 int vf, int link_state);
1059 int (*ndo_set_vf_port)(struct net_device *dev,
1061 struct nlattr *port[]);
1062 int (*ndo_get_vf_port)(struct net_device *dev,
1063 int vf, struct sk_buff *skb);
1064 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1065 #if IS_ENABLED(CONFIG_FCOE)
1066 int (*ndo_fcoe_enable)(struct net_device *dev);
1067 int (*ndo_fcoe_disable)(struct net_device *dev);
1068 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1070 struct scatterlist *sgl,
1072 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1074 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1076 struct scatterlist *sgl,
1078 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1079 struct netdev_fcoe_hbainfo *hbainfo);
1082 #if IS_ENABLED(CONFIG_LIBFCOE)
1083 #define NETDEV_FCOE_WWNN 0
1084 #define NETDEV_FCOE_WWPN 1
1085 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1086 u64 *wwn, int type);
1089 #ifdef CONFIG_RFS_ACCEL
1090 int (*ndo_rx_flow_steer)(struct net_device *dev,
1091 const struct sk_buff *skb,
1095 int (*ndo_add_slave)(struct net_device *dev,
1096 struct net_device *slave_dev);
1097 int (*ndo_del_slave)(struct net_device *dev,
1098 struct net_device *slave_dev);
1099 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1100 netdev_features_t features);
1101 int (*ndo_set_features)(struct net_device *dev,
1102 netdev_features_t features);
1103 int (*ndo_neigh_construct)(struct neighbour *n);
1104 void (*ndo_neigh_destroy)(struct neighbour *n);
1106 int (*ndo_fdb_add)(struct ndmsg *ndm,
1107 struct nlattr *tb[],
1108 struct net_device *dev,
1109 const unsigned char *addr,
1111 int (*ndo_fdb_del)(struct ndmsg *ndm,
1112 struct nlattr *tb[],
1113 struct net_device *dev,
1114 const unsigned char *addr);
1115 int (*ndo_fdb_dump)(struct sk_buff *skb,
1116 struct netlink_callback *cb,
1117 struct net_device *dev,
1120 int (*ndo_bridge_setlink)(struct net_device *dev,
1121 struct nlmsghdr *nlh);
1122 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1124 struct net_device *dev,
1126 int (*ndo_bridge_dellink)(struct net_device *dev,
1127 struct nlmsghdr *nlh);
1128 int (*ndo_change_carrier)(struct net_device *dev,
1130 int (*ndo_get_phys_port_id)(struct net_device *dev,
1131 struct netdev_phys_port_id *ppid);
1132 void (*ndo_add_vxlan_port)(struct net_device *dev,
1133 sa_family_t sa_family,
1135 void (*ndo_del_vxlan_port)(struct net_device *dev,
1136 sa_family_t sa_family,
1139 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1140 struct net_device *dev);
1141 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1144 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1145 struct net_device *dev,
1150 * enum net_device_priv_flags - &struct net_device priv_flags
1152 * These are the &struct net_device, they are only set internally
1153 * by drivers and used in the kernel. These flags are invisible to
1154 * userspace, this means that the order of these flags can change
1155 * during any kernel release.
1157 * You should have a pretty good reason to be extending these flags.
1159 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1160 * @IFF_EBRIDGE: Ethernet bridging device
1161 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1162 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1163 * @IFF_MASTER_ALB: bonding master, balance-alb
1164 * @IFF_BONDING: bonding master or slave
1165 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1166 * @IFF_ISATAP: ISATAP interface (RFC4214)
1167 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1168 * @IFF_WAN_HDLC: WAN HDLC device
1169 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1171 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1172 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1173 * @IFF_MACVLAN_PORT: device used as macvlan port
1174 * @IFF_BRIDGE_PORT: device used as bridge port
1175 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1176 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1177 * @IFF_UNICAST_FLT: Supports unicast filtering
1178 * @IFF_TEAM_PORT: device used as team port
1179 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1180 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1181 * change when it's running
1182 * @IFF_MACVLAN: Macvlan device
1184 enum netdev_priv_flags {
1185 IFF_802_1Q_VLAN = 1<<0,
1187 IFF_SLAVE_INACTIVE = 1<<2,
1188 IFF_MASTER_8023AD = 1<<3,
1189 IFF_MASTER_ALB = 1<<4,
1191 IFF_SLAVE_NEEDARP = 1<<6,
1193 IFF_MASTER_ARPMON = 1<<8,
1194 IFF_WAN_HDLC = 1<<9,
1195 IFF_XMIT_DST_RELEASE = 1<<10,
1196 IFF_DONT_BRIDGE = 1<<11,
1197 IFF_DISABLE_NETPOLL = 1<<12,
1198 IFF_MACVLAN_PORT = 1<<13,
1199 IFF_BRIDGE_PORT = 1<<14,
1200 IFF_OVS_DATAPATH = 1<<15,
1201 IFF_TX_SKB_SHARING = 1<<16,
1202 IFF_UNICAST_FLT = 1<<17,
1203 IFF_TEAM_PORT = 1<<18,
1204 IFF_SUPP_NOFCS = 1<<19,
1205 IFF_LIVE_ADDR_CHANGE = 1<<20,
1206 IFF_MACVLAN = 1<<21,
1209 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1210 #define IFF_EBRIDGE IFF_EBRIDGE
1211 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1212 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1213 #define IFF_MASTER_ALB IFF_MASTER_ALB
1214 #define IFF_BONDING IFF_BONDING
1215 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1216 #define IFF_ISATAP IFF_ISATAP
1217 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1218 #define IFF_WAN_HDLC IFF_WAN_HDLC
1219 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1220 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1221 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1222 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1223 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1224 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1225 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1226 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1227 #define IFF_TEAM_PORT IFF_TEAM_PORT
1228 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1229 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1230 #define IFF_MACVLAN IFF_MACVLAN
1233 * The DEVICE structure.
1234 * Actually, this whole structure is a big mistake. It mixes I/O
1235 * data with strictly "high-level" data, and it has to know about
1236 * almost every data structure used in the INET module.
1238 * FIXME: cleanup struct net_device such that network protocol info
1245 * This is the first field of the "visible" part of this structure
1246 * (i.e. as seen by users in the "Space.c" file). It is the name
1249 char name[IFNAMSIZ];
1251 /* device name hash chain, please keep it close to name[] */
1252 struct hlist_node name_hlist;
1258 * I/O specific fields
1259 * FIXME: Merge these and struct ifmap into one
1261 unsigned long mem_end; /* shared mem end */
1262 unsigned long mem_start; /* shared mem start */
1263 unsigned long base_addr; /* device I/O address */
1264 int irq; /* device IRQ number */
1267 * Some hardware also needs these fields, but they are not
1268 * part of the usual set specified in Space.c.
1271 unsigned long state;
1273 struct list_head dev_list;
1274 struct list_head napi_list;
1275 struct list_head unreg_list;
1276 struct list_head close_list;
1278 /* directly linked devices, like slaves for bonding */
1280 struct list_head upper;
1281 struct list_head lower;
1284 /* all linked devices, *including* neighbours */
1286 struct list_head upper;
1287 struct list_head lower;
1291 /* currently active device features */
1292 netdev_features_t features;
1293 /* user-changeable features */
1294 netdev_features_t hw_features;
1295 /* user-requested features */
1296 netdev_features_t wanted_features;
1297 /* mask of features inheritable by VLAN devices */
1298 netdev_features_t vlan_features;
1299 /* mask of features inherited by encapsulating devices
1300 * This field indicates what encapsulation offloads
1301 * the hardware is capable of doing, and drivers will
1302 * need to set them appropriately.
1304 netdev_features_t hw_enc_features;
1305 /* mask of fetures inheritable by MPLS */
1306 netdev_features_t mpls_features;
1308 /* Interface index. Unique device identifier */
1312 struct net_device_stats stats;
1314 /* dropped packets by core network, Do not use this in drivers */
1315 atomic_long_t rx_dropped;
1316 atomic_long_t tx_dropped;
1318 /* Stats to monitor carrier on<->off transitions */
1319 atomic_t carrier_changes;
1321 #ifdef CONFIG_WIRELESS_EXT
1322 /* List of functions to handle Wireless Extensions (instead of ioctl).
1323 * See <net/iw_handler.h> for details. Jean II */
1324 const struct iw_handler_def * wireless_handlers;
1325 /* Instance data managed by the core of Wireless Extensions. */
1326 struct iw_public_data * wireless_data;
1328 /* Management operations */
1329 const struct net_device_ops *netdev_ops;
1330 const struct ethtool_ops *ethtool_ops;
1331 const struct forwarding_accel_ops *fwd_ops;
1333 /* Hardware header description */
1334 const struct header_ops *header_ops;
1336 unsigned int flags; /* interface flags (a la BSD) */
1337 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1338 * See if.h for definitions. */
1339 unsigned short gflags;
1340 unsigned short padded; /* How much padding added by alloc_netdev() */
1342 unsigned char operstate; /* RFC2863 operstate */
1343 unsigned char link_mode; /* mapping policy to operstate */
1345 unsigned char if_port; /* Selectable AUI, TP,..*/
1346 unsigned char dma; /* DMA channel */
1348 unsigned int mtu; /* interface MTU value */
1349 unsigned short type; /* interface hardware type */
1350 unsigned short hard_header_len; /* hardware hdr length */
1352 /* extra head- and tailroom the hardware may need, but not in all cases
1353 * can this be guaranteed, especially tailroom. Some cases also use
1354 * LL_MAX_HEADER instead to allocate the skb.
1356 unsigned short needed_headroom;
1357 unsigned short needed_tailroom;
1359 /* Interface address info. */
1360 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1361 unsigned char addr_assign_type; /* hw address assignment type */
1362 unsigned char addr_len; /* hardware address length */
1363 unsigned short neigh_priv_len;
1364 unsigned short dev_id; /* Used to differentiate devices
1365 * that share the same link
1368 unsigned short dev_port; /* Used to differentiate
1369 * devices that share the same
1372 spinlock_t addr_list_lock;
1373 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1374 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1375 struct netdev_hw_addr_list dev_addrs; /* list of device
1379 struct kset *queues_kset;
1383 unsigned int promiscuity;
1384 unsigned int allmulti;
1387 /* Protocol specific pointers */
1389 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1390 struct vlan_info __rcu *vlan_info; /* VLAN info */
1392 #if IS_ENABLED(CONFIG_NET_DSA)
1393 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1395 #if IS_ENABLED(CONFIG_TIPC)
1396 struct tipc_bearer __rcu *tipc_ptr; /* TIPC specific data */
1398 void *atalk_ptr; /* AppleTalk link */
1399 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1400 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1401 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1402 void *ax25_ptr; /* AX.25 specific data */
1403 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1404 assign before registering */
1407 * Cache lines mostly used on receive path (including eth_type_trans())
1409 unsigned long last_rx; /* Time of last Rx */
1411 /* Interface address info used in eth_type_trans() */
1412 unsigned char *dev_addr; /* hw address, (before bcast
1413 because most packets are
1418 struct netdev_rx_queue *_rx;
1420 /* Number of RX queues allocated at register_netdev() time */
1421 unsigned int num_rx_queues;
1423 /* Number of RX queues currently active in device */
1424 unsigned int real_num_rx_queues;
1428 rx_handler_func_t __rcu *rx_handler;
1429 void __rcu *rx_handler_data;
1431 struct netdev_queue __rcu *ingress_queue;
1432 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1436 * Cache lines mostly used on transmit path
1438 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1440 /* Number of TX queues allocated at alloc_netdev_mq() time */
1441 unsigned int num_tx_queues;
1443 /* Number of TX queues currently active in device */
1444 unsigned int real_num_tx_queues;
1446 /* root qdisc from userspace point of view */
1447 struct Qdisc *qdisc;
1449 unsigned long tx_queue_len; /* Max frames per queue allowed */
1450 spinlock_t tx_global_lock;
1453 struct xps_dev_maps __rcu *xps_maps;
1455 #ifdef CONFIG_RFS_ACCEL
1456 /* CPU reverse-mapping for RX completion interrupts, indexed
1457 * by RX queue number. Assigned by driver. This must only be
1458 * set if the ndo_rx_flow_steer operation is defined. */
1459 struct cpu_rmap *rx_cpu_rmap;
1462 /* These may be needed for future network-power-down code. */
1465 * trans_start here is expensive for high speed devices on SMP,
1466 * please use netdev_queue->trans_start instead.
1468 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1470 int watchdog_timeo; /* used by dev_watchdog() */
1471 struct timer_list watchdog_timer;
1473 /* Number of references to this device */
1474 int __percpu *pcpu_refcnt;
1476 /* delayed register/unregister */
1477 struct list_head todo_list;
1478 /* device index hash chain */
1479 struct hlist_node index_hlist;
1481 struct list_head link_watch_list;
1483 /* register/unregister state machine */
1484 enum { NETREG_UNINITIALIZED=0,
1485 NETREG_REGISTERED, /* completed register_netdevice */
1486 NETREG_UNREGISTERING, /* called unregister_netdevice */
1487 NETREG_UNREGISTERED, /* completed unregister todo */
1488 NETREG_RELEASED, /* called free_netdev */
1489 NETREG_DUMMY, /* dummy device for NAPI poll */
1492 bool dismantle; /* device is going do be freed */
1495 RTNL_LINK_INITIALIZED,
1496 RTNL_LINK_INITIALIZING,
1497 } rtnl_link_state:16;
1499 /* Called from unregister, can be used to call free_netdev */
1500 void (*destructor)(struct net_device *dev);
1502 #ifdef CONFIG_NETPOLL
1503 struct netpoll_info __rcu *npinfo;
1506 #ifdef CONFIG_NET_NS
1507 /* Network namespace this network device is inside */
1511 /* mid-layer private */
1514 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1515 struct pcpu_sw_netstats __percpu *tstats;
1516 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1517 struct pcpu_vstats __percpu *vstats; /* veth stats */
1520 struct garp_port __rcu *garp_port;
1522 struct mrp_port __rcu *mrp_port;
1524 /* class/net/name entry */
1526 /* space for optional device, statistics, and wireless sysfs groups */
1527 const struct attribute_group *sysfs_groups[4];
1528 /* space for optional per-rx queue attributes */
1529 const struct attribute_group *sysfs_rx_queue_group;
1531 /* rtnetlink link ops */
1532 const struct rtnl_link_ops *rtnl_link_ops;
1534 /* for setting kernel sock attribute on TCP connection setup */
1535 #define GSO_MAX_SIZE 65536
1536 unsigned int gso_max_size;
1537 #define GSO_MAX_SEGS 65535
1541 /* Data Center Bridging netlink ops */
1542 const struct dcbnl_rtnl_ops *dcbnl_ops;
1545 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1546 u8 prio_tc_map[TC_BITMASK + 1];
1548 #if IS_ENABLED(CONFIG_FCOE)
1549 /* max exchange id for FCoE LRO by ddp */
1550 unsigned int fcoe_ddp_xid;
1552 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1553 struct netprio_map __rcu *priomap;
1555 /* phy device may attach itself for hardware timestamping */
1556 struct phy_device *phydev;
1558 struct lock_class_key *qdisc_tx_busylock;
1560 /* group the device belongs to */
1563 struct pm_qos_request pm_qos_req;
1565 #define to_net_dev(d) container_of(d, struct net_device, dev)
1567 #define NETDEV_ALIGN 32
1570 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1572 return dev->prio_tc_map[prio & TC_BITMASK];
1576 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1578 if (tc >= dev->num_tc)
1581 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1586 void netdev_reset_tc(struct net_device *dev)
1589 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1590 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1594 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1596 if (tc >= dev->num_tc)
1599 dev->tc_to_txq[tc].count = count;
1600 dev->tc_to_txq[tc].offset = offset;
1605 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1607 if (num_tc > TC_MAX_QUEUE)
1610 dev->num_tc = num_tc;
1615 int netdev_get_num_tc(struct net_device *dev)
1621 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1624 return &dev->_tx[index];
1627 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1628 void (*f)(struct net_device *,
1629 struct netdev_queue *,
1635 for (i = 0; i < dev->num_tx_queues; i++)
1636 f(dev, &dev->_tx[i], arg);
1639 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1640 struct sk_buff *skb,
1644 * Net namespace inlines
1647 struct net *dev_net(const struct net_device *dev)
1649 return read_pnet(&dev->nd_net);
1653 void dev_net_set(struct net_device *dev, struct net *net)
1655 #ifdef CONFIG_NET_NS
1656 release_net(dev->nd_net);
1657 dev->nd_net = hold_net(net);
1661 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1663 #ifdef CONFIG_NET_DSA_TAG_DSA
1664 if (dev->dsa_ptr != NULL)
1665 return dsa_uses_dsa_tags(dev->dsa_ptr);
1671 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1673 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1674 if (dev->dsa_ptr != NULL)
1675 return dsa_uses_trailer_tags(dev->dsa_ptr);
1682 * netdev_priv - access network device private data
1683 * @dev: network device
1685 * Get network device private data
1687 static inline void *netdev_priv(const struct net_device *dev)
1689 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1692 /* Set the sysfs physical device reference for the network logical device
1693 * if set prior to registration will cause a symlink during initialization.
1695 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1697 /* Set the sysfs device type for the network logical device to allow
1698 * fine-grained identification of different network device types. For
1699 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1701 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1703 /* Default NAPI poll() weight
1704 * Device drivers are strongly advised to not use bigger value
1706 #define NAPI_POLL_WEIGHT 64
1709 * netif_napi_add - initialize a napi context
1710 * @dev: network device
1711 * @napi: napi context
1712 * @poll: polling function
1713 * @weight: default weight
1715 * netif_napi_add() must be used to initialize a napi context prior to calling
1716 * *any* of the other napi related functions.
1718 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1719 int (*poll)(struct napi_struct *, int), int weight);
1722 * netif_napi_del - remove a napi context
1723 * @napi: napi context
1725 * netif_napi_del() removes a napi context from the network device napi list
1727 void netif_napi_del(struct napi_struct *napi);
1729 struct napi_gro_cb {
1730 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1733 /* Length of frag0. */
1734 unsigned int frag0_len;
1736 /* This indicates where we are processing relative to skb->data. */
1739 /* This is non-zero if the packet cannot be merged with the new skb. */
1742 /* Save the IP ID here and check when we get to the transport layer */
1745 /* Number of segments aggregated. */
1748 /* This is non-zero if the packet may be of the same flow. */
1753 #define NAPI_GRO_FREE 1
1754 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1756 /* jiffies when first packet was created/queued */
1759 /* Used in ipv6_gro_receive() */
1762 /* Used in udp_gro_receive */
1765 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1768 /* used in skb_gro_receive() slow path */
1769 struct sk_buff *last;
1772 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1774 struct packet_type {
1775 __be16 type; /* This is really htons(ether_type). */
1776 struct net_device *dev; /* NULL is wildcarded here */
1777 int (*func) (struct sk_buff *,
1778 struct net_device *,
1779 struct packet_type *,
1780 struct net_device *);
1781 bool (*id_match)(struct packet_type *ptype,
1783 void *af_packet_priv;
1784 struct list_head list;
1787 struct offload_callbacks {
1788 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1789 netdev_features_t features);
1790 int (*gso_send_check)(struct sk_buff *skb);
1791 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1792 struct sk_buff *skb);
1793 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1796 struct packet_offload {
1797 __be16 type; /* This is really htons(ether_type). */
1798 struct offload_callbacks callbacks;
1799 struct list_head list;
1802 struct udp_offload {
1804 struct offload_callbacks callbacks;
1807 /* often modified stats are per cpu, other are shared (netdev->stats) */
1808 struct pcpu_sw_netstats {
1813 struct u64_stats_sync syncp;
1816 #define netdev_alloc_pcpu_stats(type) \
1818 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1821 for_each_possible_cpu(i) { \
1822 typeof(type) *stat; \
1823 stat = per_cpu_ptr(pcpu_stats, i); \
1824 u64_stats_init(&stat->syncp); \
1830 #include <linux/notifier.h>
1832 /* netdevice notifier chain. Please remember to update the rtnetlink
1833 * notification exclusion list in rtnetlink_event() when adding new
1836 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1837 #define NETDEV_DOWN 0x0002
1838 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1839 detected a hardware crash and restarted
1840 - we can use this eg to kick tcp sessions
1842 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1843 #define NETDEV_REGISTER 0x0005
1844 #define NETDEV_UNREGISTER 0x0006
1845 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
1846 #define NETDEV_CHANGEADDR 0x0008
1847 #define NETDEV_GOING_DOWN 0x0009
1848 #define NETDEV_CHANGENAME 0x000A
1849 #define NETDEV_FEAT_CHANGE 0x000B
1850 #define NETDEV_BONDING_FAILOVER 0x000C
1851 #define NETDEV_PRE_UP 0x000D
1852 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1853 #define NETDEV_POST_TYPE_CHANGE 0x000F
1854 #define NETDEV_POST_INIT 0x0010
1855 #define NETDEV_UNREGISTER_FINAL 0x0011
1856 #define NETDEV_RELEASE 0x0012
1857 #define NETDEV_NOTIFY_PEERS 0x0013
1858 #define NETDEV_JOIN 0x0014
1859 #define NETDEV_CHANGEUPPER 0x0015
1860 #define NETDEV_RESEND_IGMP 0x0016
1861 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
1863 int register_netdevice_notifier(struct notifier_block *nb);
1864 int unregister_netdevice_notifier(struct notifier_block *nb);
1866 struct netdev_notifier_info {
1867 struct net_device *dev;
1870 struct netdev_notifier_change_info {
1871 struct netdev_notifier_info info; /* must be first */
1872 unsigned int flags_changed;
1875 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1876 struct net_device *dev)
1881 static inline struct net_device *
1882 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1887 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1890 extern rwlock_t dev_base_lock; /* Device list lock */
1892 #define for_each_netdev(net, d) \
1893 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1894 #define for_each_netdev_reverse(net, d) \
1895 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1896 #define for_each_netdev_rcu(net, d) \
1897 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1898 #define for_each_netdev_safe(net, d, n) \
1899 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1900 #define for_each_netdev_continue(net, d) \
1901 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1902 #define for_each_netdev_continue_rcu(net, d) \
1903 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1904 #define for_each_netdev_in_bond_rcu(bond, slave) \
1905 for_each_netdev_rcu(&init_net, slave) \
1906 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1907 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1909 static inline struct net_device *next_net_device(struct net_device *dev)
1911 struct list_head *lh;
1915 lh = dev->dev_list.next;
1916 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1919 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1921 struct list_head *lh;
1925 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1926 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1929 static inline struct net_device *first_net_device(struct net *net)
1931 return list_empty(&net->dev_base_head) ? NULL :
1932 net_device_entry(net->dev_base_head.next);
1935 static inline struct net_device *first_net_device_rcu(struct net *net)
1937 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1939 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1942 int netdev_boot_setup_check(struct net_device *dev);
1943 unsigned long netdev_boot_base(const char *prefix, int unit);
1944 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1945 const char *hwaddr);
1946 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1947 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1948 void dev_add_pack(struct packet_type *pt);
1949 void dev_remove_pack(struct packet_type *pt);
1950 void __dev_remove_pack(struct packet_type *pt);
1951 void dev_add_offload(struct packet_offload *po);
1952 void dev_remove_offload(struct packet_offload *po);
1954 struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1955 unsigned short mask);
1956 struct net_device *dev_get_by_name(struct net *net, const char *name);
1957 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1958 struct net_device *__dev_get_by_name(struct net *net, const char *name);
1959 int dev_alloc_name(struct net_device *dev, const char *name);
1960 int dev_open(struct net_device *dev);
1961 int dev_close(struct net_device *dev);
1962 void dev_disable_lro(struct net_device *dev);
1963 int dev_loopback_xmit(struct sk_buff *newskb);
1964 int dev_queue_xmit(struct sk_buff *skb);
1965 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
1966 int register_netdevice(struct net_device *dev);
1967 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
1968 void unregister_netdevice_many(struct list_head *head);
1969 static inline void unregister_netdevice(struct net_device *dev)
1971 unregister_netdevice_queue(dev, NULL);
1974 int netdev_refcnt_read(const struct net_device *dev);
1975 void free_netdev(struct net_device *dev);
1976 void netdev_freemem(struct net_device *dev);
1977 void synchronize_net(void);
1978 int init_dummy_netdev(struct net_device *dev);
1980 struct net_device *dev_get_by_index(struct net *net, int ifindex);
1981 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1982 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1983 int netdev_get_name(struct net *net, char *name, int ifindex);
1984 int dev_restart(struct net_device *dev);
1985 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
1987 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1989 return NAPI_GRO_CB(skb)->data_offset;
1992 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1994 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1997 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1999 NAPI_GRO_CB(skb)->data_offset += len;
2002 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2003 unsigned int offset)
2005 return NAPI_GRO_CB(skb)->frag0 + offset;
2008 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2010 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2013 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2014 unsigned int offset)
2016 if (!pskb_may_pull(skb, hlen))
2019 NAPI_GRO_CB(skb)->frag0 = NULL;
2020 NAPI_GRO_CB(skb)->frag0_len = 0;
2021 return skb->data + offset;
2024 static inline void *skb_gro_network_header(struct sk_buff *skb)
2026 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2027 skb_network_offset(skb);
2030 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2031 const void *start, unsigned int len)
2033 if (skb->ip_summed == CHECKSUM_COMPLETE)
2034 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2035 csum_partial(start, len, 0));
2038 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2039 unsigned short type,
2040 const void *daddr, const void *saddr,
2043 if (!dev->header_ops || !dev->header_ops->create)
2046 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2049 static inline int dev_parse_header(const struct sk_buff *skb,
2050 unsigned char *haddr)
2052 const struct net_device *dev = skb->dev;
2054 if (!dev->header_ops || !dev->header_ops->parse)
2056 return dev->header_ops->parse(skb, haddr);
2059 static inline int dev_rebuild_header(struct sk_buff *skb)
2061 const struct net_device *dev = skb->dev;
2063 if (!dev->header_ops || !dev->header_ops->rebuild)
2065 return dev->header_ops->rebuild(skb);
2068 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2069 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2070 static inline int unregister_gifconf(unsigned int family)
2072 return register_gifconf(family, NULL);
2075 #ifdef CONFIG_NET_FLOW_LIMIT
2076 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2077 struct sd_flow_limit {
2079 unsigned int num_buckets;
2080 unsigned int history_head;
2081 u16 history[FLOW_LIMIT_HISTORY];
2085 extern int netdev_flow_limit_table_len;
2086 #endif /* CONFIG_NET_FLOW_LIMIT */
2089 * Incoming packets are placed on per-cpu queues
2091 struct softnet_data {
2092 struct Qdisc *output_queue;
2093 struct Qdisc **output_queue_tailp;
2094 struct list_head poll_list;
2095 struct sk_buff *completion_queue;
2096 struct sk_buff_head process_queue;
2099 unsigned int processed;
2100 unsigned int time_squeeze;
2101 unsigned int cpu_collision;
2102 unsigned int received_rps;
2105 struct softnet_data *rps_ipi_list;
2107 /* Elements below can be accessed between CPUs for RPS */
2108 struct call_single_data csd ____cacheline_aligned_in_smp;
2109 struct softnet_data *rps_ipi_next;
2111 unsigned int input_queue_head;
2112 unsigned int input_queue_tail;
2114 unsigned int dropped;
2115 struct sk_buff_head input_pkt_queue;
2116 struct napi_struct backlog;
2118 #ifdef CONFIG_NET_FLOW_LIMIT
2119 struct sd_flow_limit __rcu *flow_limit;
2123 static inline void input_queue_head_incr(struct softnet_data *sd)
2126 sd->input_queue_head++;
2130 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2131 unsigned int *qtail)
2134 *qtail = ++sd->input_queue_tail;
2138 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2140 void __netif_schedule(struct Qdisc *q);
2142 static inline void netif_schedule_queue(struct netdev_queue *txq)
2144 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
2145 __netif_schedule(txq->qdisc);
2148 static inline void netif_tx_schedule_all(struct net_device *dev)
2152 for (i = 0; i < dev->num_tx_queues; i++)
2153 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2156 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2158 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2162 * netif_start_queue - allow transmit
2163 * @dev: network device
2165 * Allow upper layers to call the device hard_start_xmit routine.
2167 static inline void netif_start_queue(struct net_device *dev)
2169 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2172 static inline void netif_tx_start_all_queues(struct net_device *dev)
2176 for (i = 0; i < dev->num_tx_queues; i++) {
2177 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2178 netif_tx_start_queue(txq);
2182 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
2184 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
2185 __netif_schedule(dev_queue->qdisc);
2189 * netif_wake_queue - restart transmit
2190 * @dev: network device
2192 * Allow upper layers to call the device hard_start_xmit routine.
2193 * Used for flow control when transmit resources are available.
2195 static inline void netif_wake_queue(struct net_device *dev)
2197 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2200 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2204 for (i = 0; i < dev->num_tx_queues; i++) {
2205 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2206 netif_tx_wake_queue(txq);
2210 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2212 if (WARN_ON(!dev_queue)) {
2213 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2216 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2220 * netif_stop_queue - stop transmitted packets
2221 * @dev: network device
2223 * Stop upper layers calling the device hard_start_xmit routine.
2224 * Used for flow control when transmit resources are unavailable.
2226 static inline void netif_stop_queue(struct net_device *dev)
2228 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2231 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2235 for (i = 0; i < dev->num_tx_queues; i++) {
2236 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2237 netif_tx_stop_queue(txq);
2241 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2243 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2247 * netif_queue_stopped - test if transmit queue is flowblocked
2248 * @dev: network device
2250 * Test if transmit queue on device is currently unable to send.
2252 static inline bool netif_queue_stopped(const struct net_device *dev)
2254 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2257 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2259 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2263 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2265 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2269 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2271 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2274 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2278 dql_queued(&dev_queue->dql, bytes);
2280 if (likely(dql_avail(&dev_queue->dql) >= 0))
2283 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2286 * The XOFF flag must be set before checking the dql_avail below,
2287 * because in netdev_tx_completed_queue we update the dql_completed
2288 * before checking the XOFF flag.
2292 /* check again in case another CPU has just made room avail */
2293 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2294 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2299 * netdev_sent_queue - report the number of bytes queued to hardware
2300 * @dev: network device
2301 * @bytes: number of bytes queued to the hardware device queue
2303 * Report the number of bytes queued for sending/completion to the network
2304 * device hardware queue. @bytes should be a good approximation and should
2305 * exactly match netdev_completed_queue() @bytes
2307 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2309 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2312 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2313 unsigned int pkts, unsigned int bytes)
2316 if (unlikely(!bytes))
2319 dql_completed(&dev_queue->dql, bytes);
2322 * Without the memory barrier there is a small possiblity that
2323 * netdev_tx_sent_queue will miss the update and cause the queue to
2324 * be stopped forever
2328 if (dql_avail(&dev_queue->dql) < 0)
2331 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2332 netif_schedule_queue(dev_queue);
2337 * netdev_completed_queue - report bytes and packets completed by device
2338 * @dev: network device
2339 * @pkts: actual number of packets sent over the medium
2340 * @bytes: actual number of bytes sent over the medium
2342 * Report the number of bytes and packets transmitted by the network device
2343 * hardware queue over the physical medium, @bytes must exactly match the
2344 * @bytes amount passed to netdev_sent_queue()
2346 static inline void netdev_completed_queue(struct net_device *dev,
2347 unsigned int pkts, unsigned int bytes)
2349 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2352 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2355 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2361 * netdev_reset_queue - reset the packets and bytes count of a network device
2362 * @dev_queue: network device
2364 * Reset the bytes and packet count of a network device and clear the
2365 * software flow control OFF bit for this network device
2367 static inline void netdev_reset_queue(struct net_device *dev_queue)
2369 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2373 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2374 * @dev: network device
2375 * @queue_index: given tx queue index
2377 * Returns 0 if given tx queue index >= number of device tx queues,
2378 * otherwise returns the originally passed tx queue index.
2380 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2382 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2383 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2384 dev->name, queue_index,
2385 dev->real_num_tx_queues);
2393 * netif_running - test if up
2394 * @dev: network device
2396 * Test if the device has been brought up.
2398 static inline bool netif_running(const struct net_device *dev)
2400 return test_bit(__LINK_STATE_START, &dev->state);
2404 * Routines to manage the subqueues on a device. We only need start
2405 * stop, and a check if it's stopped. All other device management is
2406 * done at the overall netdevice level.
2407 * Also test the device if we're multiqueue.
2411 * netif_start_subqueue - allow sending packets on subqueue
2412 * @dev: network device
2413 * @queue_index: sub queue index
2415 * Start individual transmit queue of a device with multiple transmit queues.
2417 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2419 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2421 netif_tx_start_queue(txq);
2425 * netif_stop_subqueue - stop sending packets on subqueue
2426 * @dev: network device
2427 * @queue_index: sub queue index
2429 * Stop individual transmit queue of a device with multiple transmit queues.
2431 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2433 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2434 netif_tx_stop_queue(txq);
2438 * netif_subqueue_stopped - test status of subqueue
2439 * @dev: network device
2440 * @queue_index: sub queue index
2442 * Check individual transmit queue of a device with multiple transmit queues.
2444 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2447 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2449 return netif_tx_queue_stopped(txq);
2452 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2453 struct sk_buff *skb)
2455 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2459 * netif_wake_subqueue - allow sending packets on subqueue
2460 * @dev: network device
2461 * @queue_index: sub queue index
2463 * Resume individual transmit queue of a device with multiple transmit queues.
2465 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2467 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2468 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2469 __netif_schedule(txq->qdisc);
2473 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2476 static inline int netif_set_xps_queue(struct net_device *dev,
2477 const struct cpumask *mask,
2485 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2486 * as a distribution range limit for the returned value.
2488 static inline u16 skb_tx_hash(const struct net_device *dev,
2489 const struct sk_buff *skb)
2491 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2495 * netif_is_multiqueue - test if device has multiple transmit queues
2496 * @dev: network device
2498 * Check if device has multiple transmit queues
2500 static inline bool netif_is_multiqueue(const struct net_device *dev)
2502 return dev->num_tx_queues > 1;
2505 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2508 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2510 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2517 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2518 const struct net_device *from_dev)
2522 err = netif_set_real_num_tx_queues(to_dev,
2523 from_dev->real_num_tx_queues);
2527 return netif_set_real_num_rx_queues(to_dev,
2528 from_dev->real_num_rx_queues);
2535 static inline unsigned int get_netdev_rx_queue_index(
2536 struct netdev_rx_queue *queue)
2538 struct net_device *dev = queue->dev;
2539 int index = queue - dev->_rx;
2541 BUG_ON(index >= dev->num_rx_queues);
2546 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2547 int netif_get_num_default_rss_queues(void);
2549 enum skb_free_reason {
2550 SKB_REASON_CONSUMED,
2554 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2555 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2558 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2559 * interrupt context or with hardware interrupts being disabled.
2560 * (in_irq() || irqs_disabled())
2562 * We provide four helpers that can be used in following contexts :
2564 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2565 * replacing kfree_skb(skb)
2567 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2568 * Typically used in place of consume_skb(skb) in TX completion path
2570 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2571 * replacing kfree_skb(skb)
2573 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2574 * and consumed a packet. Used in place of consume_skb(skb)
2576 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2578 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2581 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2583 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2586 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2588 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2591 static inline void dev_consume_skb_any(struct sk_buff *skb)
2593 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2596 int netif_rx(struct sk_buff *skb);
2597 int netif_rx_ni(struct sk_buff *skb);
2598 int netif_receive_skb(struct sk_buff *skb);
2599 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2600 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2601 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2602 gro_result_t napi_gro_frags(struct napi_struct *napi);
2603 struct packet_offload *gro_find_receive_by_type(__be16 type);
2604 struct packet_offload *gro_find_complete_by_type(__be16 type);
2606 static inline void napi_free_frags(struct napi_struct *napi)
2608 kfree_skb(napi->skb);
2612 int netdev_rx_handler_register(struct net_device *dev,
2613 rx_handler_func_t *rx_handler,
2614 void *rx_handler_data);
2615 void netdev_rx_handler_unregister(struct net_device *dev);
2617 bool dev_valid_name(const char *name);
2618 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2619 int dev_ethtool(struct net *net, struct ifreq *);
2620 unsigned int dev_get_flags(const struct net_device *);
2621 int __dev_change_flags(struct net_device *, unsigned int flags);
2622 int dev_change_flags(struct net_device *, unsigned int);
2623 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2624 unsigned int gchanges);
2625 int dev_change_name(struct net_device *, const char *);
2626 int dev_set_alias(struct net_device *, const char *, size_t);
2627 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2628 int dev_set_mtu(struct net_device *, int);
2629 void dev_set_group(struct net_device *, int);
2630 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2631 int dev_change_carrier(struct net_device *, bool new_carrier);
2632 int dev_get_phys_port_id(struct net_device *dev,
2633 struct netdev_phys_port_id *ppid);
2634 int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2635 struct netdev_queue *txq);
2636 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2637 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2639 extern int netdev_budget;
2641 /* Called by rtnetlink.c:rtnl_unlock() */
2642 void netdev_run_todo(void);
2645 * dev_put - release reference to device
2646 * @dev: network device
2648 * Release reference to device to allow it to be freed.
2650 static inline void dev_put(struct net_device *dev)
2652 this_cpu_dec(*dev->pcpu_refcnt);
2656 * dev_hold - get reference to device
2657 * @dev: network device
2659 * Hold reference to device to keep it from being freed.
2661 static inline void dev_hold(struct net_device *dev)
2663 this_cpu_inc(*dev->pcpu_refcnt);
2666 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2667 * and _off may be called from IRQ context, but it is caller
2668 * who is responsible for serialization of these calls.
2670 * The name carrier is inappropriate, these functions should really be
2671 * called netif_lowerlayer_*() because they represent the state of any
2672 * kind of lower layer not just hardware media.
2675 void linkwatch_init_dev(struct net_device *dev);
2676 void linkwatch_fire_event(struct net_device *dev);
2677 void linkwatch_forget_dev(struct net_device *dev);
2680 * netif_carrier_ok - test if carrier present
2681 * @dev: network device
2683 * Check if carrier is present on device
2685 static inline bool netif_carrier_ok(const struct net_device *dev)
2687 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2690 unsigned long dev_trans_start(struct net_device *dev);
2692 void __netdev_watchdog_up(struct net_device *dev);
2694 void netif_carrier_on(struct net_device *dev);
2696 void netif_carrier_off(struct net_device *dev);
2699 * netif_dormant_on - mark device as dormant.
2700 * @dev: network device
2702 * Mark device as dormant (as per RFC2863).
2704 * The dormant state indicates that the relevant interface is not
2705 * actually in a condition to pass packets (i.e., it is not 'up') but is
2706 * in a "pending" state, waiting for some external event. For "on-
2707 * demand" interfaces, this new state identifies the situation where the
2708 * interface is waiting for events to place it in the up state.
2711 static inline void netif_dormant_on(struct net_device *dev)
2713 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2714 linkwatch_fire_event(dev);
2718 * netif_dormant_off - set device as not dormant.
2719 * @dev: network device
2721 * Device is not in dormant state.
2723 static inline void netif_dormant_off(struct net_device *dev)
2725 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2726 linkwatch_fire_event(dev);
2730 * netif_dormant - test if carrier present
2731 * @dev: network device
2733 * Check if carrier is present on device
2735 static inline bool netif_dormant(const struct net_device *dev)
2737 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2742 * netif_oper_up - test if device is operational
2743 * @dev: network device
2745 * Check if carrier is operational
2747 static inline bool netif_oper_up(const struct net_device *dev)
2749 return (dev->operstate == IF_OPER_UP ||
2750 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2754 * netif_device_present - is device available or removed
2755 * @dev: network device
2757 * Check if device has not been removed from system.
2759 static inline bool netif_device_present(struct net_device *dev)
2761 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2764 void netif_device_detach(struct net_device *dev);
2766 void netif_device_attach(struct net_device *dev);
2769 * Network interface message level settings
2773 NETIF_MSG_DRV = 0x0001,
2774 NETIF_MSG_PROBE = 0x0002,
2775 NETIF_MSG_LINK = 0x0004,
2776 NETIF_MSG_TIMER = 0x0008,
2777 NETIF_MSG_IFDOWN = 0x0010,
2778 NETIF_MSG_IFUP = 0x0020,
2779 NETIF_MSG_RX_ERR = 0x0040,
2780 NETIF_MSG_TX_ERR = 0x0080,
2781 NETIF_MSG_TX_QUEUED = 0x0100,
2782 NETIF_MSG_INTR = 0x0200,
2783 NETIF_MSG_TX_DONE = 0x0400,
2784 NETIF_MSG_RX_STATUS = 0x0800,
2785 NETIF_MSG_PKTDATA = 0x1000,
2786 NETIF_MSG_HW = 0x2000,
2787 NETIF_MSG_WOL = 0x4000,
2790 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2791 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2792 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2793 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2794 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2795 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2796 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2797 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2798 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2799 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2800 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2801 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2802 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2803 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2804 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2806 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2809 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2810 return default_msg_enable_bits;
2811 if (debug_value == 0) /* no output */
2813 /* set low N bits */
2814 return (1 << debug_value) - 1;
2817 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2819 spin_lock(&txq->_xmit_lock);
2820 txq->xmit_lock_owner = cpu;
2823 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2825 spin_lock_bh(&txq->_xmit_lock);
2826 txq->xmit_lock_owner = smp_processor_id();
2829 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2831 bool ok = spin_trylock(&txq->_xmit_lock);
2833 txq->xmit_lock_owner = smp_processor_id();
2837 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2839 txq->xmit_lock_owner = -1;
2840 spin_unlock(&txq->_xmit_lock);
2843 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2845 txq->xmit_lock_owner = -1;
2846 spin_unlock_bh(&txq->_xmit_lock);
2849 static inline void txq_trans_update(struct netdev_queue *txq)
2851 if (txq->xmit_lock_owner != -1)
2852 txq->trans_start = jiffies;
2856 * netif_tx_lock - grab network device transmit lock
2857 * @dev: network device
2859 * Get network device transmit lock
2861 static inline void netif_tx_lock(struct net_device *dev)
2866 spin_lock(&dev->tx_global_lock);
2867 cpu = smp_processor_id();
2868 for (i = 0; i < dev->num_tx_queues; i++) {
2869 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2871 /* We are the only thread of execution doing a
2872 * freeze, but we have to grab the _xmit_lock in
2873 * order to synchronize with threads which are in
2874 * the ->hard_start_xmit() handler and already
2875 * checked the frozen bit.
2877 __netif_tx_lock(txq, cpu);
2878 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2879 __netif_tx_unlock(txq);
2883 static inline void netif_tx_lock_bh(struct net_device *dev)
2889 static inline void netif_tx_unlock(struct net_device *dev)
2893 for (i = 0; i < dev->num_tx_queues; i++) {
2894 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2896 /* No need to grab the _xmit_lock here. If the
2897 * queue is not stopped for another reason, we
2900 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2901 netif_schedule_queue(txq);
2903 spin_unlock(&dev->tx_global_lock);
2906 static inline void netif_tx_unlock_bh(struct net_device *dev)
2908 netif_tx_unlock(dev);
2912 #define HARD_TX_LOCK(dev, txq, cpu) { \
2913 if ((dev->features & NETIF_F_LLTX) == 0) { \
2914 __netif_tx_lock(txq, cpu); \
2918 #define HARD_TX_TRYLOCK(dev, txq) \
2919 (((dev->features & NETIF_F_LLTX) == 0) ? \
2920 __netif_tx_trylock(txq) : \
2923 #define HARD_TX_UNLOCK(dev, txq) { \
2924 if ((dev->features & NETIF_F_LLTX) == 0) { \
2925 __netif_tx_unlock(txq); \
2929 static inline void netif_tx_disable(struct net_device *dev)
2935 cpu = smp_processor_id();
2936 for (i = 0; i < dev->num_tx_queues; i++) {
2937 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2939 __netif_tx_lock(txq, cpu);
2940 netif_tx_stop_queue(txq);
2941 __netif_tx_unlock(txq);
2946 static inline void netif_addr_lock(struct net_device *dev)
2948 spin_lock(&dev->addr_list_lock);
2951 static inline void netif_addr_lock_nested(struct net_device *dev)
2953 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2956 static inline void netif_addr_lock_bh(struct net_device *dev)
2958 spin_lock_bh(&dev->addr_list_lock);
2961 static inline void netif_addr_unlock(struct net_device *dev)
2963 spin_unlock(&dev->addr_list_lock);
2966 static inline void netif_addr_unlock_bh(struct net_device *dev)
2968 spin_unlock_bh(&dev->addr_list_lock);
2972 * dev_addrs walker. Should be used only for read access. Call with
2973 * rcu_read_lock held.
2975 #define for_each_dev_addr(dev, ha) \
2976 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2978 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2980 void ether_setup(struct net_device *dev);
2982 /* Support for loadable net-drivers */
2983 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2984 void (*setup)(struct net_device *),
2985 unsigned int txqs, unsigned int rxqs);
2986 #define alloc_netdev(sizeof_priv, name, setup) \
2987 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2989 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2990 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2992 int register_netdev(struct net_device *dev);
2993 void unregister_netdev(struct net_device *dev);
2995 /* General hardware address lists handling functions */
2996 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2997 struct netdev_hw_addr_list *from_list, int addr_len);
2998 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2999 struct netdev_hw_addr_list *from_list, int addr_len);
3000 void __hw_addr_init(struct netdev_hw_addr_list *list);
3002 /* Functions used for device addresses handling */
3003 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3004 unsigned char addr_type);
3005 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3006 unsigned char addr_type);
3007 void dev_addr_flush(struct net_device *dev);
3008 int dev_addr_init(struct net_device *dev);
3010 /* Functions used for unicast addresses handling */
3011 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3012 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3013 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3014 int dev_uc_sync(struct net_device *to, struct net_device *from);
3015 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3016 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3017 void dev_uc_flush(struct net_device *dev);
3018 void dev_uc_init(struct net_device *dev);
3020 /* Functions used for multicast addresses handling */
3021 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3022 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3023 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3024 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3025 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3026 int dev_mc_sync(struct net_device *to, struct net_device *from);
3027 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3028 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3029 void dev_mc_flush(struct net_device *dev);
3030 void dev_mc_init(struct net_device *dev);
3032 /* Functions used for secondary unicast and multicast support */
3033 void dev_set_rx_mode(struct net_device *dev);
3034 void __dev_set_rx_mode(struct net_device *dev);
3035 int dev_set_promiscuity(struct net_device *dev, int inc);
3036 int dev_set_allmulti(struct net_device *dev, int inc);
3037 void netdev_state_change(struct net_device *dev);
3038 void netdev_notify_peers(struct net_device *dev);
3039 void netdev_features_change(struct net_device *dev);
3040 /* Load a device via the kmod */
3041 void dev_load(struct net *net, const char *name);
3042 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3043 struct rtnl_link_stats64 *storage);
3044 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3045 const struct net_device_stats *netdev_stats);
3047 extern int netdev_max_backlog;
3048 extern int netdev_tstamp_prequeue;
3049 extern int weight_p;
3050 extern int bpf_jit_enable;
3052 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3053 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3054 struct list_head **iter);
3056 /* iterate through upper list, must be called under RCU read lock */
3057 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3058 for (iter = &(dev)->all_adj_list.upper, \
3059 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3061 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3063 void *netdev_lower_get_next_private(struct net_device *dev,
3064 struct list_head **iter);
3065 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3066 struct list_head **iter);
3068 #define netdev_for_each_lower_private(dev, priv, iter) \
3069 for (iter = (dev)->adj_list.lower.next, \
3070 priv = netdev_lower_get_next_private(dev, &(iter)); \
3072 priv = netdev_lower_get_next_private(dev, &(iter)))
3074 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3075 for (iter = &(dev)->adj_list.lower, \
3076 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3078 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3080 void *netdev_adjacent_get_private(struct list_head *adj_list);
3081 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3082 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3083 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3084 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3085 int netdev_master_upper_dev_link(struct net_device *dev,
3086 struct net_device *upper_dev);
3087 int netdev_master_upper_dev_link_private(struct net_device *dev,
3088 struct net_device *upper_dev,
3090 void netdev_upper_dev_unlink(struct net_device *dev,
3091 struct net_device *upper_dev);
3092 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3093 void *netdev_lower_dev_get_private(struct net_device *dev,
3094 struct net_device *lower_dev);
3095 int skb_checksum_help(struct sk_buff *skb);
3096 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3097 netdev_features_t features, bool tx_path);
3098 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3099 netdev_features_t features);
3102 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3104 return __skb_gso_segment(skb, features, true);
3106 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3108 static inline bool can_checksum_protocol(netdev_features_t features,
3111 return ((features & NETIF_F_GEN_CSUM) ||
3112 ((features & NETIF_F_V4_CSUM) &&
3113 protocol == htons(ETH_P_IP)) ||
3114 ((features & NETIF_F_V6_CSUM) &&
3115 protocol == htons(ETH_P_IPV6)) ||
3116 ((features & NETIF_F_FCOE_CRC) &&
3117 protocol == htons(ETH_P_FCOE)));
3121 void netdev_rx_csum_fault(struct net_device *dev);
3123 static inline void netdev_rx_csum_fault(struct net_device *dev)
3127 /* rx skb timestamps */
3128 void net_enable_timestamp(void);
3129 void net_disable_timestamp(void);
3131 #ifdef CONFIG_PROC_FS
3132 int __init dev_proc_init(void);
3134 #define dev_proc_init() 0
3137 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3139 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3142 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3144 return netdev_class_create_file_ns(class_attr, NULL);
3147 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3149 netdev_class_remove_file_ns(class_attr, NULL);
3152 extern struct kobj_ns_type_operations net_ns_type_operations;
3154 const char *netdev_drivername(const struct net_device *dev);
3156 void linkwatch_run_queue(void);
3158 static inline netdev_features_t netdev_get_wanted_features(
3159 struct net_device *dev)
3161 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3163 netdev_features_t netdev_increment_features(netdev_features_t all,
3164 netdev_features_t one, netdev_features_t mask);
3166 /* Allow TSO being used on stacked device :
3167 * Performing the GSO segmentation before last device
3168 * is a performance improvement.
3170 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3171 netdev_features_t mask)
3173 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3176 int __netdev_update_features(struct net_device *dev);
3177 void netdev_update_features(struct net_device *dev);
3178 void netdev_change_features(struct net_device *dev);
3180 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3181 struct net_device *dev);
3183 netdev_features_t netif_skb_dev_features(struct sk_buff *skb,
3184 const struct net_device *dev);
3185 static inline netdev_features_t netif_skb_features(struct sk_buff *skb)
3187 return netif_skb_dev_features(skb, skb->dev);
3190 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3192 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3194 /* check flags correspondence */
3195 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3196 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3197 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3198 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3199 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3200 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3202 return (features & feature) == feature;
3205 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3207 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3208 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3211 static inline bool netif_needs_gso(struct sk_buff *skb,
3212 netdev_features_t features)
3214 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3215 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3216 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3219 static inline void netif_set_gso_max_size(struct net_device *dev,
3222 dev->gso_max_size = size;
3225 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3226 int pulled_hlen, u16 mac_offset,
3229 skb->protocol = protocol;
3230 skb->encapsulation = 1;
3231 skb_push(skb, pulled_hlen);
3232 skb_reset_transport_header(skb);
3233 skb->mac_header = mac_offset;
3234 skb->network_header = skb->mac_header + mac_len;
3235 skb->mac_len = mac_len;
3238 static inline bool netif_is_macvlan(struct net_device *dev)
3240 return dev->priv_flags & IFF_MACVLAN;
3243 static inline bool netif_is_bond_master(struct net_device *dev)
3245 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3248 static inline bool netif_is_bond_slave(struct net_device *dev)
3250 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3253 static inline bool netif_supports_nofcs(struct net_device *dev)
3255 return dev->priv_flags & IFF_SUPP_NOFCS;
3258 extern struct pernet_operations __net_initdata loopback_net_ops;
3260 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3262 /* netdev_printk helpers, similar to dev_printk */
3264 static inline const char *netdev_name(const struct net_device *dev)
3266 if (dev->reg_state != NETREG_REGISTERED)
3267 return "(unregistered net_device)";
3272 int netdev_printk(const char *level, const struct net_device *dev,
3273 const char *format, ...);
3275 int netdev_emerg(const struct net_device *dev, const char *format, ...);
3277 int netdev_alert(const struct net_device *dev, const char *format, ...);
3279 int netdev_crit(const struct net_device *dev, const char *format, ...);
3281 int netdev_err(const struct net_device *dev, const char *format, ...);
3283 int netdev_warn(const struct net_device *dev, const char *format, ...);
3285 int netdev_notice(const struct net_device *dev, const char *format, ...);
3287 int netdev_info(const struct net_device *dev, const char *format, ...);
3289 #define MODULE_ALIAS_NETDEV(device) \
3290 MODULE_ALIAS("netdev-" device)
3292 #if defined(CONFIG_DYNAMIC_DEBUG)
3293 #define netdev_dbg(__dev, format, args...) \
3295 dynamic_netdev_dbg(__dev, format, ##args); \
3297 #elif defined(DEBUG)
3298 #define netdev_dbg(__dev, format, args...) \
3299 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3301 #define netdev_dbg(__dev, format, args...) \
3304 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3309 #if defined(VERBOSE_DEBUG)
3310 #define netdev_vdbg netdev_dbg
3313 #define netdev_vdbg(dev, format, args...) \
3316 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3322 * netdev_WARN() acts like dev_printk(), but with the key difference
3323 * of using a WARN/WARN_ON to get the message out, including the
3324 * file/line information and a backtrace.
3326 #define netdev_WARN(dev, format, args...) \
3327 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
3329 /* netif printk helpers, similar to netdev_printk */
3331 #define netif_printk(priv, type, level, dev, fmt, args...) \
3333 if (netif_msg_##type(priv)) \
3334 netdev_printk(level, (dev), fmt, ##args); \
3337 #define netif_level(level, priv, type, dev, fmt, args...) \
3339 if (netif_msg_##type(priv)) \
3340 netdev_##level(dev, fmt, ##args); \
3343 #define netif_emerg(priv, type, dev, fmt, args...) \
3344 netif_level(emerg, priv, type, dev, fmt, ##args)
3345 #define netif_alert(priv, type, dev, fmt, args...) \
3346 netif_level(alert, priv, type, dev, fmt, ##args)
3347 #define netif_crit(priv, type, dev, fmt, args...) \
3348 netif_level(crit, priv, type, dev, fmt, ##args)
3349 #define netif_err(priv, type, dev, fmt, args...) \
3350 netif_level(err, priv, type, dev, fmt, ##args)
3351 #define netif_warn(priv, type, dev, fmt, args...) \
3352 netif_level(warn, priv, type, dev, fmt, ##args)
3353 #define netif_notice(priv, type, dev, fmt, args...) \
3354 netif_level(notice, priv, type, dev, fmt, ##args)
3355 #define netif_info(priv, type, dev, fmt, args...) \
3356 netif_level(info, priv, type, dev, fmt, ##args)
3358 #if defined(CONFIG_DYNAMIC_DEBUG)
3359 #define netif_dbg(priv, type, netdev, format, args...) \
3361 if (netif_msg_##type(priv)) \
3362 dynamic_netdev_dbg(netdev, format, ##args); \
3364 #elif defined(DEBUG)
3365 #define netif_dbg(priv, type, dev, format, args...) \
3366 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3368 #define netif_dbg(priv, type, dev, format, args...) \
3371 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3376 #if defined(VERBOSE_DEBUG)
3377 #define netif_vdbg netif_dbg
3379 #define netif_vdbg(priv, type, dev, format, args...) \
3382 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3388 * The list of packet types we will receive (as opposed to discard)
3389 * and the routines to invoke.
3391 * Why 16. Because with 16 the only overlap we get on a hash of the
3392 * low nibble of the protocol value is RARP/SNAP/X.25.
3394 * NOTE: That is no longer true with the addition of VLAN tags. Not
3395 * sure which should go first, but I bet it won't make much
3396 * difference if we are running VLANs. The good news is that
3397 * this protocol won't be in the list unless compiled in, so
3398 * the average user (w/out VLANs) will not be adversely affected.
3414 #define PTYPE_HASH_SIZE (16)
3415 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3417 #endif /* _LINUX_NETDEVICE_H */