2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
59 /* source back-compat hooks */
60 #define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
63 void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct ethtool_ops *ops);
66 /* hardware address assignment types */
67 #define NET_ADDR_PERM 0 /* address is permanent (default) */
68 #define NET_ADDR_RANDOM 1 /* address is generated randomly */
69 #define NET_ADDR_STOLEN 2 /* address is stolen from other device */
70 #define NET_ADDR_SET 3 /* address is set using
71 * dev_set_mac_address() */
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
78 * Transmit return codes: transmit return codes originate from three different
81 * - qdisc return codes
82 * - driver transmit return codes
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously, in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), all
91 * others are propagated to higher layers.
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK 0xf0
111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
116 typedef enum netdev_tx netdev_tx_t;
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
122 static inline bool dev_xmit_complete(int rc)
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
130 if (likely(rc < NET_XMIT_MASK))
137 * Compute the worst case header length according to the protocols
141 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
145 # define LL_MAX_HEADER 96
148 # define LL_MAX_HEADER 32
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
163 struct net_device_stats {
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
173 unsigned long collisions;
174 unsigned long rx_length_errors;
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed;
202 struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
206 #define NETDEV_HW_ADDR_T_LAN 1
207 #define NETDEV_HW_ADDR_T_SAN 2
208 #define NETDEV_HW_ADDR_T_SLAVE 3
209 #define NETDEV_HW_ADDR_T_UNICAST 4
210 #define NETDEV_HW_ADDR_T_MULTICAST 5
215 struct rcu_head rcu_head;
218 struct netdev_hw_addr_list {
219 struct list_head list;
223 #define netdev_hw_addr_list_count(l) ((l)->count)
224 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225 #define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
228 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230 #define netdev_for_each_uc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
233 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
235 #define netdev_for_each_mc_addr(ha, dev) \
236 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
252 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
268 const void *saddr, unsigned int len);
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*rebuild)(struct sk_buff *skb);
271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
277 /* These flag bits are private to the generic network queueing
278 * layer, they may not be explicitly referenced by any other
282 enum netdev_state_t {
284 __LINK_STATE_PRESENT,
285 __LINK_STATE_NOCARRIER,
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
292 * This structure holds at boot time configured netdevice settings. They
293 * are then used in the device probing.
295 struct netdev_boot_setup {
299 #define NETDEV_BOOT_SETUP_MAX 8
301 int __init netdev_boot_setup(char *str);
304 * Structure for NAPI scheduling similar to tasklet but with weighting
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-cpu poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
313 struct list_head poll_list;
317 unsigned int gro_count;
318 int (*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
320 spinlock_t poll_lock;
323 struct net_device *dev;
324 struct sk_buff *gro_list;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_DISABLE, /* Disable pending */
334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED, /* In NAPI hash */
345 typedef enum gro_result gro_result_t;
348 * enum rx_handler_result - Possible return values for rx_handlers.
349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
352 * case skb->dev was changed by rx_handler.
353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
354 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
356 * rx_handlers are functions called from inside __netif_receive_skb(), to do
357 * special processing of the skb, prior to delivery to protocol handlers.
359 * Currently, a net_device can only have a single rx_handler registered. Trying
360 * to register a second rx_handler will return -EBUSY.
362 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
363 * To unregister a rx_handler on a net_device, use
364 * netdev_rx_handler_unregister().
366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
369 * If the rx_handler consumed to skb in some way, it should return
370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
371 * the skb to be delivered in some other ways.
373 * If the rx_handler changed skb->dev, to divert the skb to another
374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
375 * new device will be called if it exists.
377 * If the rx_handler consider the skb should be ignored, it should return
378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
379 * are registered on exact device (ptype->dev == skb->dev).
381 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
382 * delivered, it should return RX_HANDLER_PASS.
384 * A device without a registered rx_handler will behave as if rx_handler
385 * returned RX_HANDLER_PASS.
388 enum rx_handler_result {
394 typedef enum rx_handler_result rx_handler_result_t;
395 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
397 void __napi_schedule(struct napi_struct *n);
399 static inline bool napi_disable_pending(struct napi_struct *n)
401 return test_bit(NAPI_STATE_DISABLE, &n->state);
405 * napi_schedule_prep - check if napi can be scheduled
408 * Test if NAPI routine is already running, and if not mark
409 * it as running. This is used as a condition variable
410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
413 static inline bool napi_schedule_prep(struct napi_struct *n)
415 return !napi_disable_pending(n) &&
416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
420 * napi_schedule - schedule NAPI poll
423 * Schedule NAPI poll routine to be called if it is not already
426 static inline void napi_schedule(struct napi_struct *n)
428 if (napi_schedule_prep(n))
432 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
433 static inline bool napi_reschedule(struct napi_struct *napi)
435 if (napi_schedule_prep(napi)) {
436 __napi_schedule(napi);
443 * napi_complete - NAPI processing complete
446 * Mark NAPI processing as complete.
448 void __napi_complete(struct napi_struct *n);
449 void napi_complete(struct napi_struct *n);
452 * napi_by_id - lookup a NAPI by napi_id
453 * @napi_id: hashed napi_id
455 * lookup @napi_id in napi_hash table
456 * must be called under rcu_read_lock()
458 struct napi_struct *napi_by_id(unsigned int napi_id);
461 * napi_hash_add - add a NAPI to global hashtable
462 * @napi: napi context
464 * generate a new napi_id and store a @napi under it in napi_hash
466 void napi_hash_add(struct napi_struct *napi);
469 * napi_hash_del - remove a NAPI from global table
470 * @napi: napi context
472 * Warning: caller must observe rcu grace period
473 * before freeing memory containing @napi
475 void napi_hash_del(struct napi_struct *napi);
478 * napi_disable - prevent NAPI from scheduling
481 * Stop NAPI from being scheduled on this context.
482 * Waits till any outstanding processing completes.
484 static inline void napi_disable(struct napi_struct *n)
487 set_bit(NAPI_STATE_DISABLE, &n->state);
488 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
490 clear_bit(NAPI_STATE_DISABLE, &n->state);
494 * napi_enable - enable NAPI scheduling
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
500 static inline void napi_enable(struct napi_struct *n)
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
503 smp_mb__before_clear_bit();
504 clear_bit(NAPI_STATE_SCHED, &n->state);
509 * napi_synchronize - wait until NAPI is not running
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
516 static inline void napi_synchronize(const struct napi_struct *n)
518 while (test_bit(NAPI_STATE_SCHED, &n->state))
522 # define napi_synchronize(n) barrier()
525 enum netdev_queue_state_t {
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
528 __QUEUE_STATE_FROZEN,
529 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
530 (1 << __QUEUE_STATE_STACK_XOFF))
531 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
532 (1 << __QUEUE_STATE_FROZEN))
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,
672 /* This structure contains an instance of an RX queue. */
673 struct netdev_rx_queue {
674 struct rps_map __rcu *rps_map;
675 struct rps_dev_flow_table __rcu *rps_flow_table;
677 struct net_device *dev;
678 } ____cacheline_aligned_in_smp;
679 #endif /* CONFIG_RPS */
683 * This structure holds an XPS map which can be of variable length. The
684 * map is an array of queues.
688 unsigned int alloc_len;
692 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
693 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
697 * This structure holds all XPS maps for device. Maps are indexed by CPU.
699 struct xps_dev_maps {
701 struct xps_map __rcu *cpu_map[0];
703 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
704 (nr_cpu_ids * sizeof(struct xps_map *)))
705 #endif /* CONFIG_XPS */
707 #define TC_MAX_QUEUE 16
708 #define TC_BITMASK 15
709 /* HW offloaded queuing disciplines txq count and offset maps */
710 struct netdev_tc_txq {
715 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
717 * This structure is to hold information about the device
718 * configured to run FCoE protocol stack.
720 struct netdev_fcoe_hbainfo {
721 char manufacturer[64];
722 char serial_number[64];
723 char hardware_version[64];
724 char driver_version[64];
725 char optionrom_version[64];
726 char firmware_version[64];
728 char model_description[256];
732 #define MAX_PHYS_PORT_ID_LEN 32
734 /* This structure holds a unique identifier to identify the
735 * physical port used by a netdevice.
737 struct netdev_phys_port_id {
738 unsigned char id[MAX_PHYS_PORT_ID_LEN];
739 unsigned char id_len;
743 * This structure defines the management hooks for network devices.
744 * The following hooks can be defined; unless noted otherwise, they are
745 * optional and can be filled with a null pointer.
747 * int (*ndo_init)(struct net_device *dev);
748 * This function is called once when network device is registered.
749 * The network device can use this to any late stage initializaton
750 * or semantic validattion. It can fail with an error code which will
751 * be propogated back to register_netdev
753 * void (*ndo_uninit)(struct net_device *dev);
754 * This function is called when device is unregistered or when registration
755 * fails. It is not called if init fails.
757 * int (*ndo_open)(struct net_device *dev);
758 * This function is called when network device transistions to the up
761 * int (*ndo_stop)(struct net_device *dev);
762 * This function is called when network device transistions to the down
765 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
766 * struct net_device *dev);
767 * Called when a packet needs to be transmitted.
768 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
769 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
770 * Required can not be NULL.
772 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
774 * Called to decide which queue to when device supports multiple
777 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
778 * This function is called to allow device receiver to make
779 * changes to configuration when multicast or promiscious is enabled.
781 * void (*ndo_set_rx_mode)(struct net_device *dev);
782 * This function is called device changes address list filtering.
783 * If driver handles unicast address filtering, it should set
784 * IFF_UNICAST_FLT to its priv_flags.
786 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
787 * This function is called when the Media Access Control address
788 * needs to be changed. If this interface is not defined, the
789 * mac address can not be changed.
791 * int (*ndo_validate_addr)(struct net_device *dev);
792 * Test if Media Access Control address is valid for the device.
794 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
795 * Called when a user request an ioctl which can't be handled by
796 * the generic interface code. If not defined ioctl's return
797 * not supported error code.
799 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
800 * Used to set network devices bus interface parameters. This interface
801 * is retained for legacy reason, new devices should use the bus
802 * interface (PCI) for low level management.
804 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
805 * Called when a user wants to change the Maximum Transfer Unit
806 * of a device. If not defined, any request to change MTU will
807 * will return an error.
809 * void (*ndo_tx_timeout)(struct net_device *dev);
810 * Callback uses when the transmitter has not made any progress
811 * for dev->watchdog ticks.
813 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
814 * struct rtnl_link_stats64 *storage);
815 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
816 * Called when a user wants to get the network device usage
817 * statistics. Drivers must do one of the following:
818 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
819 * rtnl_link_stats64 structure passed by the caller.
820 * 2. Define @ndo_get_stats to update a net_device_stats structure
821 * (which should normally be dev->stats) and return a pointer to
822 * it. The structure may be changed asynchronously only if each
823 * field is written atomically.
824 * 3. Update dev->stats asynchronously and atomically, and define
827 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
828 * If device support VLAN filtering this function is called when a
829 * VLAN id is registered.
831 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
832 * If device support VLAN filtering this function is called when a
833 * VLAN id is unregistered.
835 * void (*ndo_poll_controller)(struct net_device *dev);
837 * SR-IOV management functions.
838 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
839 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
840 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
841 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
842 * int (*ndo_get_vf_config)(struct net_device *dev,
843 * int vf, struct ifla_vf_info *ivf);
844 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
845 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
846 * struct nlattr *port[]);
847 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
848 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
849 * Called to setup 'tc' number of traffic classes in the net device. This
850 * is always called from the stack with the rtnl lock held and netif tx
851 * queues stopped. This allows the netdevice to perform queue management
854 * Fiber Channel over Ethernet (FCoE) offload functions.
855 * int (*ndo_fcoe_enable)(struct net_device *dev);
856 * Called when the FCoE protocol stack wants to start using LLD for FCoE
857 * so the underlying device can perform whatever needed configuration or
858 * initialization to support acceleration of FCoE traffic.
860 * int (*ndo_fcoe_disable)(struct net_device *dev);
861 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
862 * so the underlying device can perform whatever needed clean-ups to
863 * stop supporting acceleration of FCoE traffic.
865 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
866 * struct scatterlist *sgl, unsigned int sgc);
867 * Called when the FCoE Initiator wants to initialize an I/O that
868 * is a possible candidate for Direct Data Placement (DDP). The LLD can
869 * perform necessary setup and returns 1 to indicate the device is set up
870 * successfully to perform DDP on this I/O, otherwise this returns 0.
872 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
873 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
874 * indicated by the FC exchange id 'xid', so the underlying device can
875 * clean up and reuse resources for later DDP requests.
877 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
878 * struct scatterlist *sgl, unsigned int sgc);
879 * Called when the FCoE Target wants to initialize an I/O that
880 * is a possible candidate for Direct Data Placement (DDP). The LLD can
881 * perform necessary setup and returns 1 to indicate the device is set up
882 * successfully to perform DDP on this I/O, otherwise this returns 0.
884 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
885 * struct netdev_fcoe_hbainfo *hbainfo);
886 * Called when the FCoE Protocol stack wants information on the underlying
887 * device. This information is utilized by the FCoE protocol stack to
888 * register attributes with Fiber Channel management service as per the
889 * FC-GS Fabric Device Management Information(FDMI) specification.
891 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
892 * Called when the underlying device wants to override default World Wide
893 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
894 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
895 * protocol stack to use.
898 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
899 * u16 rxq_index, u32 flow_id);
900 * Set hardware filter for RFS. rxq_index is the target queue index;
901 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
902 * Return the filter ID on success, or a negative error code.
904 * Slave management functions (for bridge, bonding, etc).
905 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
906 * Called to make another netdev an underling.
908 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
909 * Called to release previously enslaved netdev.
911 * Feature/offload setting functions.
912 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
913 * netdev_features_t features);
914 * Adjusts the requested feature flags according to device-specific
915 * constraints, and returns the resulting flags. Must not modify
918 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
919 * Called to update device configuration to new features. Passed
920 * feature set might be less than what was returned by ndo_fix_features()).
921 * Must return >0 or -errno if it changed dev->features itself.
923 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
924 * struct net_device *dev,
925 * const unsigned char *addr, u16 flags)
926 * Adds an FDB entry to dev for addr.
927 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
928 * struct net_device *dev,
929 * const unsigned char *addr)
930 * Deletes the FDB entry from dev coresponding to addr.
931 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
932 * struct net_device *dev, int idx)
933 * Used to add FDB entries to dump requests. Implementers should add
934 * entries to skb and update idx with the number of entries.
936 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
937 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
938 * struct net_device *dev, u32 filter_mask)
940 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
941 * Called to change device carrier. Soft-devices (like dummy, team, etc)
942 * which do not represent real hardware may define this to allow their
943 * userspace components to manage their virtual carrier state. Devices
944 * that determine carrier state from physical hardware properties (eg
945 * network cables) or protocol-dependent mechanisms (eg
946 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
948 * int (*ndo_get_phys_port_id)(struct net_device *dev,
949 * struct netdev_phys_port_id *ppid);
950 * Called to get ID of physical port of this device. If driver does
951 * not implement this, it is assumed that the hw is not able to have
952 * multiple net devices on single physical port.
954 * void (*ndo_add_vxlan_port)(struct net_device *dev,
955 * sa_family_t sa_family, __be16 port);
956 * Called by vxlan to notiy a driver about the UDP port and socket
957 * address family that vxlan is listnening to. It is called only when
958 * a new port starts listening. The operation is protected by the
959 * vxlan_net->sock_lock.
961 * void (*ndo_del_vxlan_port)(struct net_device *dev,
962 * sa_family_t sa_family, __be16 port);
963 * Called by vxlan to notify the driver about a UDP port and socket
964 * address family that vxlan is not listening to anymore. The operation
965 * is protected by the vxlan_net->sock_lock.
967 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
968 * struct net_device *dev)
969 * Called by upper layer devices to accelerate switching or other
970 * station functionality into hardware. 'pdev is the lowerdev
971 * to use for the offload and 'dev' is the net device that will
972 * back the offload. Returns a pointer to the private structure
973 * the upper layer will maintain.
974 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
975 * Called by upper layer device to delete the station created
976 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
977 * the station and priv is the structure returned by the add
979 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
980 * struct net_device *dev,
982 * Callback to use for xmit over the accelerated station. This
983 * is used in place of ndo_start_xmit on accelerated net
986 struct net_device_ops {
987 int (*ndo_init)(struct net_device *dev);
988 void (*ndo_uninit)(struct net_device *dev);
989 int (*ndo_open)(struct net_device *dev);
990 int (*ndo_stop)(struct net_device *dev);
991 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
992 struct net_device *dev);
993 u16 (*ndo_select_queue)(struct net_device *dev,
996 void (*ndo_change_rx_flags)(struct net_device *dev,
998 void (*ndo_set_rx_mode)(struct net_device *dev);
999 int (*ndo_set_mac_address)(struct net_device *dev,
1001 int (*ndo_validate_addr)(struct net_device *dev);
1002 int (*ndo_do_ioctl)(struct net_device *dev,
1003 struct ifreq *ifr, int cmd);
1004 int (*ndo_set_config)(struct net_device *dev,
1006 int (*ndo_change_mtu)(struct net_device *dev,
1008 int (*ndo_neigh_setup)(struct net_device *dev,
1009 struct neigh_parms *);
1010 void (*ndo_tx_timeout) (struct net_device *dev);
1012 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1013 struct rtnl_link_stats64 *storage);
1014 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1016 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1017 __be16 proto, u16 vid);
1018 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1019 __be16 proto, u16 vid);
1020 #ifdef CONFIG_NET_POLL_CONTROLLER
1021 void (*ndo_poll_controller)(struct net_device *dev);
1022 int (*ndo_netpoll_setup)(struct net_device *dev,
1023 struct netpoll_info *info,
1025 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1027 #ifdef CONFIG_NET_RX_BUSY_POLL
1028 int (*ndo_busy_poll)(struct napi_struct *dev);
1030 int (*ndo_set_vf_mac)(struct net_device *dev,
1031 int queue, u8 *mac);
1032 int (*ndo_set_vf_vlan)(struct net_device *dev,
1033 int queue, u16 vlan, u8 qos);
1034 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
1036 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1037 int vf, bool setting);
1038 int (*ndo_get_vf_config)(struct net_device *dev,
1040 struct ifla_vf_info *ivf);
1041 int (*ndo_set_vf_link_state)(struct net_device *dev,
1042 int vf, int link_state);
1043 int (*ndo_set_vf_port)(struct net_device *dev,
1045 struct nlattr *port[]);
1046 int (*ndo_get_vf_port)(struct net_device *dev,
1047 int vf, struct sk_buff *skb);
1048 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1049 #if IS_ENABLED(CONFIG_FCOE)
1050 int (*ndo_fcoe_enable)(struct net_device *dev);
1051 int (*ndo_fcoe_disable)(struct net_device *dev);
1052 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1054 struct scatterlist *sgl,
1056 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1058 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1060 struct scatterlist *sgl,
1062 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1063 struct netdev_fcoe_hbainfo *hbainfo);
1066 #if IS_ENABLED(CONFIG_LIBFCOE)
1067 #define NETDEV_FCOE_WWNN 0
1068 #define NETDEV_FCOE_WWPN 1
1069 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1070 u64 *wwn, int type);
1073 #ifdef CONFIG_RFS_ACCEL
1074 int (*ndo_rx_flow_steer)(struct net_device *dev,
1075 const struct sk_buff *skb,
1079 int (*ndo_add_slave)(struct net_device *dev,
1080 struct net_device *slave_dev);
1081 int (*ndo_del_slave)(struct net_device *dev,
1082 struct net_device *slave_dev);
1083 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1084 netdev_features_t features);
1085 int (*ndo_set_features)(struct net_device *dev,
1086 netdev_features_t features);
1087 int (*ndo_neigh_construct)(struct neighbour *n);
1088 void (*ndo_neigh_destroy)(struct neighbour *n);
1090 int (*ndo_fdb_add)(struct ndmsg *ndm,
1091 struct nlattr *tb[],
1092 struct net_device *dev,
1093 const unsigned char *addr,
1095 int (*ndo_fdb_del)(struct ndmsg *ndm,
1096 struct nlattr *tb[],
1097 struct net_device *dev,
1098 const unsigned char *addr);
1099 int (*ndo_fdb_dump)(struct sk_buff *skb,
1100 struct netlink_callback *cb,
1101 struct net_device *dev,
1104 int (*ndo_bridge_setlink)(struct net_device *dev,
1105 struct nlmsghdr *nlh);
1106 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1108 struct net_device *dev,
1110 int (*ndo_bridge_dellink)(struct net_device *dev,
1111 struct nlmsghdr *nlh);
1112 int (*ndo_change_carrier)(struct net_device *dev,
1114 int (*ndo_get_phys_port_id)(struct net_device *dev,
1115 struct netdev_phys_port_id *ppid);
1116 void (*ndo_add_vxlan_port)(struct net_device *dev,
1117 sa_family_t sa_family,
1119 void (*ndo_del_vxlan_port)(struct net_device *dev,
1120 sa_family_t sa_family,
1123 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1124 struct net_device *dev);
1125 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1128 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1129 struct net_device *dev,
1134 * The DEVICE structure.
1135 * Actually, this whole structure is a big mistake. It mixes I/O
1136 * data with strictly "high-level" data, and it has to know about
1137 * almost every data structure used in the INET module.
1139 * FIXME: cleanup struct net_device such that network protocol info
1146 * This is the first field of the "visible" part of this structure
1147 * (i.e. as seen by users in the "Space.c" file). It is the name
1150 char name[IFNAMSIZ];
1152 /* device name hash chain, please keep it close to name[] */
1153 struct hlist_node name_hlist;
1159 * I/O specific fields
1160 * FIXME: Merge these and struct ifmap into one
1162 unsigned long mem_end; /* shared mem end */
1163 unsigned long mem_start; /* shared mem start */
1164 unsigned long base_addr; /* device I/O address */
1165 int irq; /* device IRQ number */
1168 * Some hardware also needs these fields, but they are not
1169 * part of the usual set specified in Space.c.
1172 unsigned long state;
1174 struct list_head dev_list;
1175 struct list_head napi_list;
1176 struct list_head unreg_list;
1177 struct list_head close_list;
1179 /* directly linked devices, like slaves for bonding */
1181 struct list_head upper;
1182 struct list_head lower;
1185 /* all linked devices, *including* neighbours */
1187 struct list_head upper;
1188 struct list_head lower;
1192 /* currently active device features */
1193 netdev_features_t features;
1194 /* user-changeable features */
1195 netdev_features_t hw_features;
1196 /* user-requested features */
1197 netdev_features_t wanted_features;
1198 /* mask of features inheritable by VLAN devices */
1199 netdev_features_t vlan_features;
1200 /* mask of features inherited by encapsulating devices
1201 * This field indicates what encapsulation offloads
1202 * the hardware is capable of doing, and drivers will
1203 * need to set them appropriately.
1205 netdev_features_t hw_enc_features;
1206 /* mask of fetures inheritable by MPLS */
1207 netdev_features_t mpls_features;
1209 /* Interface index. Unique device identifier */
1213 struct net_device_stats stats;
1214 atomic_long_t rx_dropped; /* dropped packets by core network
1215 * Do not use this in drivers.
1218 #ifdef CONFIG_WIRELESS_EXT
1219 /* List of functions to handle Wireless Extensions (instead of ioctl).
1220 * See <net/iw_handler.h> for details. Jean II */
1221 const struct iw_handler_def * wireless_handlers;
1222 /* Instance data managed by the core of Wireless Extensions. */
1223 struct iw_public_data * wireless_data;
1225 /* Management operations */
1226 const struct net_device_ops *netdev_ops;
1227 const struct ethtool_ops *ethtool_ops;
1228 const struct forwarding_accel_ops *fwd_ops;
1230 /* Hardware header description */
1231 const struct header_ops *header_ops;
1233 unsigned int flags; /* interface flags (a la BSD) */
1234 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1235 * See if.h for definitions. */
1236 unsigned short gflags;
1237 unsigned short padded; /* How much padding added by alloc_netdev() */
1239 unsigned char operstate; /* RFC2863 operstate */
1240 unsigned char link_mode; /* mapping policy to operstate */
1242 unsigned char if_port; /* Selectable AUI, TP,..*/
1243 unsigned char dma; /* DMA channel */
1245 unsigned int mtu; /* interface MTU value */
1246 unsigned short type; /* interface hardware type */
1247 unsigned short hard_header_len; /* hardware hdr length */
1249 /* extra head- and tailroom the hardware may need, but not in all cases
1250 * can this be guaranteed, especially tailroom. Some cases also use
1251 * LL_MAX_HEADER instead to allocate the skb.
1253 unsigned short needed_headroom;
1254 unsigned short needed_tailroom;
1256 /* Interface address info. */
1257 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1258 unsigned char addr_assign_type; /* hw address assignment type */
1259 unsigned char addr_len; /* hardware address length */
1260 unsigned short neigh_priv_len;
1261 unsigned short dev_id; /* Used to differentiate devices
1262 * that share the same link
1265 spinlock_t addr_list_lock;
1266 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1267 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1268 struct netdev_hw_addr_list dev_addrs; /* list of device
1272 struct kset *queues_kset;
1276 unsigned int promiscuity;
1277 unsigned int allmulti;
1280 /* Protocol specific pointers */
1282 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1283 struct vlan_info __rcu *vlan_info; /* VLAN info */
1285 #if IS_ENABLED(CONFIG_NET_DSA)
1286 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1288 #if IS_ENABLED(CONFIG_TIPC)
1289 struct tipc_bearer __rcu *tipc_ptr; /* TIPC specific data */
1291 void *atalk_ptr; /* AppleTalk link */
1292 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1293 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1294 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1295 void *ax25_ptr; /* AX.25 specific data */
1296 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1297 assign before registering */
1300 * Cache lines mostly used on receive path (including eth_type_trans())
1302 unsigned long last_rx; /* Time of last Rx
1303 * This should not be set in
1304 * drivers, unless really needed,
1305 * because network stack (bonding)
1306 * use it if/when necessary, to
1307 * avoid dirtying this cache line.
1310 /* Interface address info used in eth_type_trans() */
1311 unsigned char *dev_addr; /* hw address, (before bcast
1312 because most packets are
1317 struct netdev_rx_queue *_rx;
1319 /* Number of RX queues allocated at register_netdev() time */
1320 unsigned int num_rx_queues;
1322 /* Number of RX queues currently active in device */
1323 unsigned int real_num_rx_queues;
1327 rx_handler_func_t __rcu *rx_handler;
1328 void __rcu *rx_handler_data;
1330 struct netdev_queue __rcu *ingress_queue;
1331 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1335 * Cache lines mostly used on transmit path
1337 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1339 /* Number of TX queues allocated at alloc_netdev_mq() time */
1340 unsigned int num_tx_queues;
1342 /* Number of TX queues currently active in device */
1343 unsigned int real_num_tx_queues;
1345 /* root qdisc from userspace point of view */
1346 struct Qdisc *qdisc;
1348 unsigned long tx_queue_len; /* Max frames per queue allowed */
1349 spinlock_t tx_global_lock;
1352 struct xps_dev_maps __rcu *xps_maps;
1354 #ifdef CONFIG_RFS_ACCEL
1355 /* CPU reverse-mapping for RX completion interrupts, indexed
1356 * by RX queue number. Assigned by driver. This must only be
1357 * set if the ndo_rx_flow_steer operation is defined. */
1358 struct cpu_rmap *rx_cpu_rmap;
1361 /* These may be needed for future network-power-down code. */
1364 * trans_start here is expensive for high speed devices on SMP,
1365 * please use netdev_queue->trans_start instead.
1367 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1369 int watchdog_timeo; /* used by dev_watchdog() */
1370 struct timer_list watchdog_timer;
1372 /* Number of references to this device */
1373 int __percpu *pcpu_refcnt;
1375 /* delayed register/unregister */
1376 struct list_head todo_list;
1377 /* device index hash chain */
1378 struct hlist_node index_hlist;
1380 struct list_head link_watch_list;
1382 /* register/unregister state machine */
1383 enum { NETREG_UNINITIALIZED=0,
1384 NETREG_REGISTERED, /* completed register_netdevice */
1385 NETREG_UNREGISTERING, /* called unregister_netdevice */
1386 NETREG_UNREGISTERED, /* completed unregister todo */
1387 NETREG_RELEASED, /* called free_netdev */
1388 NETREG_DUMMY, /* dummy device for NAPI poll */
1391 bool dismantle; /* device is going do be freed */
1394 RTNL_LINK_INITIALIZED,
1395 RTNL_LINK_INITIALIZING,
1396 } rtnl_link_state:16;
1398 /* Called from unregister, can be used to call free_netdev */
1399 void (*destructor)(struct net_device *dev);
1401 #ifdef CONFIG_NETPOLL
1402 struct netpoll_info __rcu *npinfo;
1405 #ifdef CONFIG_NET_NS
1406 /* Network namespace this network device is inside */
1410 /* mid-layer private */
1413 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1414 struct pcpu_sw_netstats __percpu *tstats;
1415 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1416 struct pcpu_vstats __percpu *vstats; /* veth stats */
1419 struct garp_port __rcu *garp_port;
1421 struct mrp_port __rcu *mrp_port;
1423 /* class/net/name entry */
1425 /* space for optional device, statistics, and wireless sysfs groups */
1426 const struct attribute_group *sysfs_groups[4];
1428 /* rtnetlink link ops */
1429 const struct rtnl_link_ops *rtnl_link_ops;
1431 /* for setting kernel sock attribute on TCP connection setup */
1432 #define GSO_MAX_SIZE 65536
1433 unsigned int gso_max_size;
1434 #define GSO_MAX_SEGS 65535
1438 /* Data Center Bridging netlink ops */
1439 const struct dcbnl_rtnl_ops *dcbnl_ops;
1442 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1443 u8 prio_tc_map[TC_BITMASK + 1];
1445 #if IS_ENABLED(CONFIG_FCOE)
1446 /* max exchange id for FCoE LRO by ddp */
1447 unsigned int fcoe_ddp_xid;
1449 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1450 struct netprio_map __rcu *priomap;
1452 /* phy device may attach itself for hardware timestamping */
1453 struct phy_device *phydev;
1455 struct lock_class_key *qdisc_tx_busylock;
1457 /* group the device belongs to */
1460 struct pm_qos_request pm_qos_req;
1462 #define to_net_dev(d) container_of(d, struct net_device, dev)
1464 #define NETDEV_ALIGN 32
1467 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1469 return dev->prio_tc_map[prio & TC_BITMASK];
1473 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1475 if (tc >= dev->num_tc)
1478 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1483 void netdev_reset_tc(struct net_device *dev)
1486 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1487 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1491 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1493 if (tc >= dev->num_tc)
1496 dev->tc_to_txq[tc].count = count;
1497 dev->tc_to_txq[tc].offset = offset;
1502 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1504 if (num_tc > TC_MAX_QUEUE)
1507 dev->num_tc = num_tc;
1512 int netdev_get_num_tc(struct net_device *dev)
1518 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1521 return &dev->_tx[index];
1524 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1525 void (*f)(struct net_device *,
1526 struct netdev_queue *,
1532 for (i = 0; i < dev->num_tx_queues; i++)
1533 f(dev, &dev->_tx[i], arg);
1536 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1537 struct sk_buff *skb,
1539 u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb);
1542 * Net namespace inlines
1545 struct net *dev_net(const struct net_device *dev)
1547 return read_pnet(&dev->nd_net);
1551 void dev_net_set(struct net_device *dev, struct net *net)
1553 #ifdef CONFIG_NET_NS
1554 release_net(dev->nd_net);
1555 dev->nd_net = hold_net(net);
1559 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1561 #ifdef CONFIG_NET_DSA_TAG_DSA
1562 if (dev->dsa_ptr != NULL)
1563 return dsa_uses_dsa_tags(dev->dsa_ptr);
1569 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1571 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1572 if (dev->dsa_ptr != NULL)
1573 return dsa_uses_trailer_tags(dev->dsa_ptr);
1580 * netdev_priv - access network device private data
1581 * @dev: network device
1583 * Get network device private data
1585 static inline void *netdev_priv(const struct net_device *dev)
1587 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1590 /* Set the sysfs physical device reference for the network logical device
1591 * if set prior to registration will cause a symlink during initialization.
1593 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1595 /* Set the sysfs device type for the network logical device to allow
1596 * fine-grained identification of different network device types. For
1597 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1599 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1601 /* Default NAPI poll() weight
1602 * Device drivers are strongly advised to not use bigger value
1604 #define NAPI_POLL_WEIGHT 64
1607 * netif_napi_add - initialize a napi context
1608 * @dev: network device
1609 * @napi: napi context
1610 * @poll: polling function
1611 * @weight: default weight
1613 * netif_napi_add() must be used to initialize a napi context prior to calling
1614 * *any* of the other napi related functions.
1616 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1617 int (*poll)(struct napi_struct *, int), int weight);
1620 * netif_napi_del - remove a napi context
1621 * @napi: napi context
1623 * netif_napi_del() removes a napi context from the network device napi list
1625 void netif_napi_del(struct napi_struct *napi);
1627 struct napi_gro_cb {
1628 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1631 /* Length of frag0. */
1632 unsigned int frag0_len;
1634 /* This indicates where we are processing relative to skb->data. */
1637 /* This is non-zero if the packet cannot be merged with the new skb. */
1640 /* Save the IP ID here and check when we get to the transport layer */
1643 /* Number of segments aggregated. */
1646 /* This is non-zero if the packet may be of the same flow. */
1651 #define NAPI_GRO_FREE 1
1652 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1654 /* jiffies when first packet was created/queued */
1657 /* Used in ipv6_gro_receive() */
1660 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1663 /* used in skb_gro_receive() slow path */
1664 struct sk_buff *last;
1667 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1669 struct packet_type {
1670 __be16 type; /* This is really htons(ether_type). */
1671 struct net_device *dev; /* NULL is wildcarded here */
1672 int (*func) (struct sk_buff *,
1673 struct net_device *,
1674 struct packet_type *,
1675 struct net_device *);
1676 bool (*id_match)(struct packet_type *ptype,
1678 void *af_packet_priv;
1679 struct list_head list;
1682 struct offload_callbacks {
1683 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1684 netdev_features_t features);
1685 int (*gso_send_check)(struct sk_buff *skb);
1686 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1687 struct sk_buff *skb);
1688 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1691 struct packet_offload {
1692 __be16 type; /* This is really htons(ether_type). */
1693 struct offload_callbacks callbacks;
1694 struct list_head list;
1697 /* often modified stats are per cpu, other are shared (netdev->stats) */
1698 struct pcpu_sw_netstats {
1703 struct u64_stats_sync syncp;
1706 #include <linux/notifier.h>
1708 /* netdevice notifier chain. Please remember to update the rtnetlink
1709 * notification exclusion list in rtnetlink_event() when adding new
1712 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1713 #define NETDEV_DOWN 0x0002
1714 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1715 detected a hardware crash and restarted
1716 - we can use this eg to kick tcp sessions
1718 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1719 #define NETDEV_REGISTER 0x0005
1720 #define NETDEV_UNREGISTER 0x0006
1721 #define NETDEV_CHANGEMTU 0x0007
1722 #define NETDEV_CHANGEADDR 0x0008
1723 #define NETDEV_GOING_DOWN 0x0009
1724 #define NETDEV_CHANGENAME 0x000A
1725 #define NETDEV_FEAT_CHANGE 0x000B
1726 #define NETDEV_BONDING_FAILOVER 0x000C
1727 #define NETDEV_PRE_UP 0x000D
1728 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1729 #define NETDEV_POST_TYPE_CHANGE 0x000F
1730 #define NETDEV_POST_INIT 0x0010
1731 #define NETDEV_UNREGISTER_FINAL 0x0011
1732 #define NETDEV_RELEASE 0x0012
1733 #define NETDEV_NOTIFY_PEERS 0x0013
1734 #define NETDEV_JOIN 0x0014
1735 #define NETDEV_CHANGEUPPER 0x0015
1736 #define NETDEV_RESEND_IGMP 0x0016
1738 int register_netdevice_notifier(struct notifier_block *nb);
1739 int unregister_netdevice_notifier(struct notifier_block *nb);
1741 struct netdev_notifier_info {
1742 struct net_device *dev;
1745 struct netdev_notifier_change_info {
1746 struct netdev_notifier_info info; /* must be first */
1747 unsigned int flags_changed;
1750 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1751 struct net_device *dev)
1756 static inline struct net_device *
1757 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1762 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1765 extern rwlock_t dev_base_lock; /* Device list lock */
1767 #define for_each_netdev(net, d) \
1768 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1769 #define for_each_netdev_reverse(net, d) \
1770 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1771 #define for_each_netdev_rcu(net, d) \
1772 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1773 #define for_each_netdev_safe(net, d, n) \
1774 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1775 #define for_each_netdev_continue(net, d) \
1776 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1777 #define for_each_netdev_continue_rcu(net, d) \
1778 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1779 #define for_each_netdev_in_bond_rcu(bond, slave) \
1780 for_each_netdev_rcu(&init_net, slave) \
1781 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1782 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1784 static inline struct net_device *next_net_device(struct net_device *dev)
1786 struct list_head *lh;
1790 lh = dev->dev_list.next;
1791 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1794 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1796 struct list_head *lh;
1800 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1801 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1804 static inline struct net_device *first_net_device(struct net *net)
1806 return list_empty(&net->dev_base_head) ? NULL :
1807 net_device_entry(net->dev_base_head.next);
1810 static inline struct net_device *first_net_device_rcu(struct net *net)
1812 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1814 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1817 int netdev_boot_setup_check(struct net_device *dev);
1818 unsigned long netdev_boot_base(const char *prefix, int unit);
1819 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1820 const char *hwaddr);
1821 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1822 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1823 void dev_add_pack(struct packet_type *pt);
1824 void dev_remove_pack(struct packet_type *pt);
1825 void __dev_remove_pack(struct packet_type *pt);
1826 void dev_add_offload(struct packet_offload *po);
1827 void dev_remove_offload(struct packet_offload *po);
1829 struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1830 unsigned short mask);
1831 struct net_device *dev_get_by_name(struct net *net, const char *name);
1832 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1833 struct net_device *__dev_get_by_name(struct net *net, const char *name);
1834 int dev_alloc_name(struct net_device *dev, const char *name);
1835 int dev_open(struct net_device *dev);
1836 int dev_close(struct net_device *dev);
1837 void dev_disable_lro(struct net_device *dev);
1838 int dev_loopback_xmit(struct sk_buff *newskb);
1839 int dev_queue_xmit(struct sk_buff *skb);
1840 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
1841 int register_netdevice(struct net_device *dev);
1842 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
1843 void unregister_netdevice_many(struct list_head *head);
1844 static inline void unregister_netdevice(struct net_device *dev)
1846 unregister_netdevice_queue(dev, NULL);
1849 int netdev_refcnt_read(const struct net_device *dev);
1850 void free_netdev(struct net_device *dev);
1851 void netdev_freemem(struct net_device *dev);
1852 void synchronize_net(void);
1853 int init_dummy_netdev(struct net_device *dev);
1855 struct net_device *dev_get_by_index(struct net *net, int ifindex);
1856 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1857 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1858 int netdev_get_name(struct net *net, char *name, int ifindex);
1859 int dev_restart(struct net_device *dev);
1860 #ifdef CONFIG_NETPOLL_TRAP
1861 int netpoll_trap(void);
1863 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
1865 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1867 return NAPI_GRO_CB(skb)->data_offset;
1870 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1872 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1875 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1877 NAPI_GRO_CB(skb)->data_offset += len;
1880 static inline void *skb_gro_header_fast(struct sk_buff *skb,
1881 unsigned int offset)
1883 return NAPI_GRO_CB(skb)->frag0 + offset;
1886 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1888 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1891 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1892 unsigned int offset)
1894 if (!pskb_may_pull(skb, hlen))
1897 NAPI_GRO_CB(skb)->frag0 = NULL;
1898 NAPI_GRO_CB(skb)->frag0_len = 0;
1899 return skb->data + offset;
1902 static inline void *skb_gro_mac_header(struct sk_buff *skb)
1904 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
1907 static inline void *skb_gro_network_header(struct sk_buff *skb)
1909 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1910 skb_network_offset(skb);
1913 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
1914 const void *start, unsigned int len)
1916 if (skb->ip_summed == CHECKSUM_COMPLETE)
1917 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
1918 csum_partial(start, len, 0));
1921 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1922 unsigned short type,
1923 const void *daddr, const void *saddr,
1926 if (!dev->header_ops || !dev->header_ops->create)
1929 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
1932 static inline int dev_parse_header(const struct sk_buff *skb,
1933 unsigned char *haddr)
1935 const struct net_device *dev = skb->dev;
1937 if (!dev->header_ops || !dev->header_ops->parse)
1939 return dev->header_ops->parse(skb, haddr);
1942 static inline int dev_rebuild_header(struct sk_buff *skb)
1944 const struct net_device *dev = skb->dev;
1946 if (!dev->header_ops || !dev->header_ops->rebuild)
1948 return dev->header_ops->rebuild(skb);
1951 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
1952 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1953 static inline int unregister_gifconf(unsigned int family)
1955 return register_gifconf(family, NULL);
1958 #ifdef CONFIG_NET_FLOW_LIMIT
1959 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
1960 struct sd_flow_limit {
1962 unsigned int num_buckets;
1963 unsigned int history_head;
1964 u16 history[FLOW_LIMIT_HISTORY];
1968 extern int netdev_flow_limit_table_len;
1969 #endif /* CONFIG_NET_FLOW_LIMIT */
1972 * Incoming packets are placed on per-cpu queues
1974 struct softnet_data {
1975 struct Qdisc *output_queue;
1976 struct Qdisc **output_queue_tailp;
1977 struct list_head poll_list;
1978 struct sk_buff *completion_queue;
1979 struct sk_buff_head process_queue;
1982 unsigned int processed;
1983 unsigned int time_squeeze;
1984 unsigned int cpu_collision;
1985 unsigned int received_rps;
1988 struct softnet_data *rps_ipi_list;
1990 /* Elements below can be accessed between CPUs for RPS */
1991 struct call_single_data csd ____cacheline_aligned_in_smp;
1992 struct softnet_data *rps_ipi_next;
1994 unsigned int input_queue_head;
1995 unsigned int input_queue_tail;
1997 unsigned int dropped;
1998 struct sk_buff_head input_pkt_queue;
1999 struct napi_struct backlog;
2001 #ifdef CONFIG_NET_FLOW_LIMIT
2002 struct sd_flow_limit __rcu *flow_limit;
2006 static inline void input_queue_head_incr(struct softnet_data *sd)
2009 sd->input_queue_head++;
2013 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2014 unsigned int *qtail)
2017 *qtail = ++sd->input_queue_tail;
2021 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2023 void __netif_schedule(struct Qdisc *q);
2025 static inline void netif_schedule_queue(struct netdev_queue *txq)
2027 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
2028 __netif_schedule(txq->qdisc);
2031 static inline void netif_tx_schedule_all(struct net_device *dev)
2035 for (i = 0; i < dev->num_tx_queues; i++)
2036 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2039 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2041 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2045 * netif_start_queue - allow transmit
2046 * @dev: network device
2048 * Allow upper layers to call the device hard_start_xmit routine.
2050 static inline void netif_start_queue(struct net_device *dev)
2052 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2055 static inline void netif_tx_start_all_queues(struct net_device *dev)
2059 for (i = 0; i < dev->num_tx_queues; i++) {
2060 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2061 netif_tx_start_queue(txq);
2065 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
2067 #ifdef CONFIG_NETPOLL_TRAP
2068 if (netpoll_trap()) {
2069 netif_tx_start_queue(dev_queue);
2073 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
2074 __netif_schedule(dev_queue->qdisc);
2078 * netif_wake_queue - restart transmit
2079 * @dev: network device
2081 * Allow upper layers to call the device hard_start_xmit routine.
2082 * Used for flow control when transmit resources are available.
2084 static inline void netif_wake_queue(struct net_device *dev)
2086 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2089 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2093 for (i = 0; i < dev->num_tx_queues; i++) {
2094 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2095 netif_tx_wake_queue(txq);
2099 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2101 if (WARN_ON(!dev_queue)) {
2102 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2105 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2109 * netif_stop_queue - stop transmitted packets
2110 * @dev: network device
2112 * Stop upper layers calling the device hard_start_xmit routine.
2113 * Used for flow control when transmit resources are unavailable.
2115 static inline void netif_stop_queue(struct net_device *dev)
2117 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2120 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2124 for (i = 0; i < dev->num_tx_queues; i++) {
2125 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2126 netif_tx_stop_queue(txq);
2130 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2132 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2136 * netif_queue_stopped - test if transmit queue is flowblocked
2137 * @dev: network device
2139 * Test if transmit queue on device is currently unable to send.
2141 static inline bool netif_queue_stopped(const struct net_device *dev)
2143 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2146 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2148 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2151 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2153 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2156 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2160 dql_queued(&dev_queue->dql, bytes);
2162 if (likely(dql_avail(&dev_queue->dql) >= 0))
2165 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2168 * The XOFF flag must be set before checking the dql_avail below,
2169 * because in netdev_tx_completed_queue we update the dql_completed
2170 * before checking the XOFF flag.
2174 /* check again in case another CPU has just made room avail */
2175 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2176 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2181 * netdev_sent_queue - report the number of bytes queued to hardware
2182 * @dev: network device
2183 * @bytes: number of bytes queued to the hardware device queue
2185 * Report the number of bytes queued for sending/completion to the network
2186 * device hardware queue. @bytes should be a good approximation and should
2187 * exactly match netdev_completed_queue() @bytes
2189 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2191 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2194 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2195 unsigned int pkts, unsigned int bytes)
2198 if (unlikely(!bytes))
2201 dql_completed(&dev_queue->dql, bytes);
2204 * Without the memory barrier there is a small possiblity that
2205 * netdev_tx_sent_queue will miss the update and cause the queue to
2206 * be stopped forever
2210 if (dql_avail(&dev_queue->dql) < 0)
2213 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2214 netif_schedule_queue(dev_queue);
2219 * netdev_completed_queue - report bytes and packets completed by device
2220 * @dev: network device
2221 * @pkts: actual number of packets sent over the medium
2222 * @bytes: actual number of bytes sent over the medium
2224 * Report the number of bytes and packets transmitted by the network device
2225 * hardware queue over the physical medium, @bytes must exactly match the
2226 * @bytes amount passed to netdev_sent_queue()
2228 static inline void netdev_completed_queue(struct net_device *dev,
2229 unsigned int pkts, unsigned int bytes)
2231 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2234 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2237 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2243 * netdev_reset_queue - reset the packets and bytes count of a network device
2244 * @dev_queue: network device
2246 * Reset the bytes and packet count of a network device and clear the
2247 * software flow control OFF bit for this network device
2249 static inline void netdev_reset_queue(struct net_device *dev_queue)
2251 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2255 * netif_running - test if up
2256 * @dev: network device
2258 * Test if the device has been brought up.
2260 static inline bool netif_running(const struct net_device *dev)
2262 return test_bit(__LINK_STATE_START, &dev->state);
2266 * Routines to manage the subqueues on a device. We only need start
2267 * stop, and a check if it's stopped. All other device management is
2268 * done at the overall netdevice level.
2269 * Also test the device if we're multiqueue.
2273 * netif_start_subqueue - allow sending packets on subqueue
2274 * @dev: network device
2275 * @queue_index: sub queue index
2277 * Start individual transmit queue of a device with multiple transmit queues.
2279 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2281 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2283 netif_tx_start_queue(txq);
2287 * netif_stop_subqueue - stop sending packets on subqueue
2288 * @dev: network device
2289 * @queue_index: sub queue index
2291 * Stop individual transmit queue of a device with multiple transmit queues.
2293 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2295 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2296 #ifdef CONFIG_NETPOLL_TRAP
2300 netif_tx_stop_queue(txq);
2304 * netif_subqueue_stopped - test status of subqueue
2305 * @dev: network device
2306 * @queue_index: sub queue index
2308 * Check individual transmit queue of a device with multiple transmit queues.
2310 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2313 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2315 return netif_tx_queue_stopped(txq);
2318 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2319 struct sk_buff *skb)
2321 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2325 * netif_wake_subqueue - allow sending packets on subqueue
2326 * @dev: network device
2327 * @queue_index: sub queue index
2329 * Resume individual transmit queue of a device with multiple transmit queues.
2331 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2333 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2334 #ifdef CONFIG_NETPOLL_TRAP
2338 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2339 __netif_schedule(txq->qdisc);
2343 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2346 static inline int netif_set_xps_queue(struct net_device *dev,
2347 const struct cpumask *mask,
2355 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2356 * as a distribution range limit for the returned value.
2358 static inline u16 skb_tx_hash(const struct net_device *dev,
2359 const struct sk_buff *skb)
2361 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2365 * netif_is_multiqueue - test if device has multiple transmit queues
2366 * @dev: network device
2368 * Check if device has multiple transmit queues
2370 static inline bool netif_is_multiqueue(const struct net_device *dev)
2372 return dev->num_tx_queues > 1;
2375 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2378 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2380 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2387 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2388 const struct net_device *from_dev)
2392 err = netif_set_real_num_tx_queues(to_dev,
2393 from_dev->real_num_tx_queues);
2397 return netif_set_real_num_rx_queues(to_dev,
2398 from_dev->real_num_rx_queues);
2404 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2405 int netif_get_num_default_rss_queues(void);
2407 enum skb_free_reason {
2408 SKB_REASON_CONSUMED,
2412 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2413 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2416 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2417 * interrupt context or with hardware interrupts being disabled.
2418 * (in_irq() || irqs_disabled())
2420 * We provide four helpers that can be used in following contexts :
2422 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2423 * replacing kfree_skb(skb)
2425 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2426 * Typically used in place of consume_skb(skb) in TX completion path
2428 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2429 * replacing kfree_skb(skb)
2431 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2432 * and consumed a packet. Used in place of consume_skb(skb)
2434 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2436 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2439 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2441 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2444 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2446 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2449 static inline void dev_consume_skb_any(struct sk_buff *skb)
2451 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2454 int netif_rx(struct sk_buff *skb);
2455 int netif_rx_ni(struct sk_buff *skb);
2456 int netif_receive_skb(struct sk_buff *skb);
2457 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2458 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2459 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2460 gro_result_t napi_gro_frags(struct napi_struct *napi);
2461 struct packet_offload *gro_find_receive_by_type(__be16 type);
2462 struct packet_offload *gro_find_complete_by_type(__be16 type);
2464 static inline void napi_free_frags(struct napi_struct *napi)
2466 kfree_skb(napi->skb);
2470 int netdev_rx_handler_register(struct net_device *dev,
2471 rx_handler_func_t *rx_handler,
2472 void *rx_handler_data);
2473 void netdev_rx_handler_unregister(struct net_device *dev);
2475 bool dev_valid_name(const char *name);
2476 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2477 int dev_ethtool(struct net *net, struct ifreq *);
2478 unsigned int dev_get_flags(const struct net_device *);
2479 int __dev_change_flags(struct net_device *, unsigned int flags);
2480 int dev_change_flags(struct net_device *, unsigned int);
2481 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2482 unsigned int gchanges);
2483 int dev_change_name(struct net_device *, const char *);
2484 int dev_set_alias(struct net_device *, const char *, size_t);
2485 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2486 int dev_set_mtu(struct net_device *, int);
2487 void dev_set_group(struct net_device *, int);
2488 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2489 int dev_change_carrier(struct net_device *, bool new_carrier);
2490 int dev_get_phys_port_id(struct net_device *dev,
2491 struct netdev_phys_port_id *ppid);
2492 int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2493 struct netdev_queue *txq);
2494 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2496 extern int netdev_budget;
2498 /* Called by rtnetlink.c:rtnl_unlock() */
2499 void netdev_run_todo(void);
2502 * dev_put - release reference to device
2503 * @dev: network device
2505 * Release reference to device to allow it to be freed.
2507 static inline void dev_put(struct net_device *dev)
2509 this_cpu_dec(*dev->pcpu_refcnt);
2513 * dev_hold - get reference to device
2514 * @dev: network device
2516 * Hold reference to device to keep it from being freed.
2518 static inline void dev_hold(struct net_device *dev)
2520 this_cpu_inc(*dev->pcpu_refcnt);
2523 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2524 * and _off may be called from IRQ context, but it is caller
2525 * who is responsible for serialization of these calls.
2527 * The name carrier is inappropriate, these functions should really be
2528 * called netif_lowerlayer_*() because they represent the state of any
2529 * kind of lower layer not just hardware media.
2532 void linkwatch_init_dev(struct net_device *dev);
2533 void linkwatch_fire_event(struct net_device *dev);
2534 void linkwatch_forget_dev(struct net_device *dev);
2537 * netif_carrier_ok - test if carrier present
2538 * @dev: network device
2540 * Check if carrier is present on device
2542 static inline bool netif_carrier_ok(const struct net_device *dev)
2544 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2547 unsigned long dev_trans_start(struct net_device *dev);
2549 void __netdev_watchdog_up(struct net_device *dev);
2551 void netif_carrier_on(struct net_device *dev);
2553 void netif_carrier_off(struct net_device *dev);
2556 * netif_dormant_on - mark device as dormant.
2557 * @dev: network device
2559 * Mark device as dormant (as per RFC2863).
2561 * The dormant state indicates that the relevant interface is not
2562 * actually in a condition to pass packets (i.e., it is not 'up') but is
2563 * in a "pending" state, waiting for some external event. For "on-
2564 * demand" interfaces, this new state identifies the situation where the
2565 * interface is waiting for events to place it in the up state.
2568 static inline void netif_dormant_on(struct net_device *dev)
2570 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2571 linkwatch_fire_event(dev);
2575 * netif_dormant_off - set device as not dormant.
2576 * @dev: network device
2578 * Device is not in dormant state.
2580 static inline void netif_dormant_off(struct net_device *dev)
2582 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2583 linkwatch_fire_event(dev);
2587 * netif_dormant - test if carrier present
2588 * @dev: network device
2590 * Check if carrier is present on device
2592 static inline bool netif_dormant(const struct net_device *dev)
2594 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2599 * netif_oper_up - test if device is operational
2600 * @dev: network device
2602 * Check if carrier is operational
2604 static inline bool netif_oper_up(const struct net_device *dev)
2606 return (dev->operstate == IF_OPER_UP ||
2607 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2611 * netif_device_present - is device available or removed
2612 * @dev: network device
2614 * Check if device has not been removed from system.
2616 static inline bool netif_device_present(struct net_device *dev)
2618 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2621 void netif_device_detach(struct net_device *dev);
2623 void netif_device_attach(struct net_device *dev);
2626 * Network interface message level settings
2630 NETIF_MSG_DRV = 0x0001,
2631 NETIF_MSG_PROBE = 0x0002,
2632 NETIF_MSG_LINK = 0x0004,
2633 NETIF_MSG_TIMER = 0x0008,
2634 NETIF_MSG_IFDOWN = 0x0010,
2635 NETIF_MSG_IFUP = 0x0020,
2636 NETIF_MSG_RX_ERR = 0x0040,
2637 NETIF_MSG_TX_ERR = 0x0080,
2638 NETIF_MSG_TX_QUEUED = 0x0100,
2639 NETIF_MSG_INTR = 0x0200,
2640 NETIF_MSG_TX_DONE = 0x0400,
2641 NETIF_MSG_RX_STATUS = 0x0800,
2642 NETIF_MSG_PKTDATA = 0x1000,
2643 NETIF_MSG_HW = 0x2000,
2644 NETIF_MSG_WOL = 0x4000,
2647 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2648 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2649 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2650 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2651 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2652 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2653 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2654 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2655 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2656 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2657 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2658 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2659 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2660 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2661 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2663 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2666 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2667 return default_msg_enable_bits;
2668 if (debug_value == 0) /* no output */
2670 /* set low N bits */
2671 return (1 << debug_value) - 1;
2674 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2676 spin_lock(&txq->_xmit_lock);
2677 txq->xmit_lock_owner = cpu;
2680 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2682 spin_lock_bh(&txq->_xmit_lock);
2683 txq->xmit_lock_owner = smp_processor_id();
2686 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2688 bool ok = spin_trylock(&txq->_xmit_lock);
2690 txq->xmit_lock_owner = smp_processor_id();
2694 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2696 txq->xmit_lock_owner = -1;
2697 spin_unlock(&txq->_xmit_lock);
2700 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2702 txq->xmit_lock_owner = -1;
2703 spin_unlock_bh(&txq->_xmit_lock);
2706 static inline void txq_trans_update(struct netdev_queue *txq)
2708 if (txq->xmit_lock_owner != -1)
2709 txq->trans_start = jiffies;
2713 * netif_tx_lock - grab network device transmit lock
2714 * @dev: network device
2716 * Get network device transmit lock
2718 static inline void netif_tx_lock(struct net_device *dev)
2723 spin_lock(&dev->tx_global_lock);
2724 cpu = smp_processor_id();
2725 for (i = 0; i < dev->num_tx_queues; i++) {
2726 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2728 /* We are the only thread of execution doing a
2729 * freeze, but we have to grab the _xmit_lock in
2730 * order to synchronize with threads which are in
2731 * the ->hard_start_xmit() handler and already
2732 * checked the frozen bit.
2734 __netif_tx_lock(txq, cpu);
2735 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2736 __netif_tx_unlock(txq);
2740 static inline void netif_tx_lock_bh(struct net_device *dev)
2746 static inline void netif_tx_unlock(struct net_device *dev)
2750 for (i = 0; i < dev->num_tx_queues; i++) {
2751 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2753 /* No need to grab the _xmit_lock here. If the
2754 * queue is not stopped for another reason, we
2757 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2758 netif_schedule_queue(txq);
2760 spin_unlock(&dev->tx_global_lock);
2763 static inline void netif_tx_unlock_bh(struct net_device *dev)
2765 netif_tx_unlock(dev);
2769 #define HARD_TX_LOCK(dev, txq, cpu) { \
2770 if ((dev->features & NETIF_F_LLTX) == 0) { \
2771 __netif_tx_lock(txq, cpu); \
2775 #define HARD_TX_UNLOCK(dev, txq) { \
2776 if ((dev->features & NETIF_F_LLTX) == 0) { \
2777 __netif_tx_unlock(txq); \
2781 static inline void netif_tx_disable(struct net_device *dev)
2787 cpu = smp_processor_id();
2788 for (i = 0; i < dev->num_tx_queues; i++) {
2789 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2791 __netif_tx_lock(txq, cpu);
2792 netif_tx_stop_queue(txq);
2793 __netif_tx_unlock(txq);
2798 static inline void netif_addr_lock(struct net_device *dev)
2800 spin_lock(&dev->addr_list_lock);
2803 static inline void netif_addr_lock_nested(struct net_device *dev)
2805 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2808 static inline void netif_addr_lock_bh(struct net_device *dev)
2810 spin_lock_bh(&dev->addr_list_lock);
2813 static inline void netif_addr_unlock(struct net_device *dev)
2815 spin_unlock(&dev->addr_list_lock);
2818 static inline void netif_addr_unlock_bh(struct net_device *dev)
2820 spin_unlock_bh(&dev->addr_list_lock);
2824 * dev_addrs walker. Should be used only for read access. Call with
2825 * rcu_read_lock held.
2827 #define for_each_dev_addr(dev, ha) \
2828 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2830 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2832 void ether_setup(struct net_device *dev);
2834 /* Support for loadable net-drivers */
2835 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2836 void (*setup)(struct net_device *),
2837 unsigned int txqs, unsigned int rxqs);
2838 #define alloc_netdev(sizeof_priv, name, setup) \
2839 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2841 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2842 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2844 int register_netdev(struct net_device *dev);
2845 void unregister_netdev(struct net_device *dev);
2847 /* General hardware address lists handling functions */
2848 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2849 struct netdev_hw_addr_list *from_list, int addr_len);
2850 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2851 struct netdev_hw_addr_list *from_list, int addr_len);
2852 void __hw_addr_init(struct netdev_hw_addr_list *list);
2854 /* Functions used for device addresses handling */
2855 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2856 unsigned char addr_type);
2857 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2858 unsigned char addr_type);
2859 void dev_addr_flush(struct net_device *dev);
2860 int dev_addr_init(struct net_device *dev);
2862 /* Functions used for unicast addresses handling */
2863 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2864 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2865 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
2866 int dev_uc_sync(struct net_device *to, struct net_device *from);
2867 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
2868 void dev_uc_unsync(struct net_device *to, struct net_device *from);
2869 void dev_uc_flush(struct net_device *dev);
2870 void dev_uc_init(struct net_device *dev);
2872 /* Functions used for multicast addresses handling */
2873 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
2874 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
2875 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
2876 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
2877 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
2878 int dev_mc_sync(struct net_device *to, struct net_device *from);
2879 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
2880 void dev_mc_unsync(struct net_device *to, struct net_device *from);
2881 void dev_mc_flush(struct net_device *dev);
2882 void dev_mc_init(struct net_device *dev);
2884 /* Functions used for secondary unicast and multicast support */
2885 void dev_set_rx_mode(struct net_device *dev);
2886 void __dev_set_rx_mode(struct net_device *dev);
2887 int dev_set_promiscuity(struct net_device *dev, int inc);
2888 int dev_set_allmulti(struct net_device *dev, int inc);
2889 void netdev_state_change(struct net_device *dev);
2890 void netdev_notify_peers(struct net_device *dev);
2891 void netdev_features_change(struct net_device *dev);
2892 /* Load a device via the kmod */
2893 void dev_load(struct net *net, const char *name);
2894 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2895 struct rtnl_link_stats64 *storage);
2896 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
2897 const struct net_device_stats *netdev_stats);
2899 extern int netdev_max_backlog;
2900 extern int netdev_tstamp_prequeue;
2901 extern int weight_p;
2902 extern int bpf_jit_enable;
2904 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
2905 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
2906 struct list_head **iter);
2908 /* iterate through upper list, must be called under RCU read lock */
2909 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
2910 for (iter = &(dev)->all_adj_list.upper, \
2911 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
2913 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
2915 void *netdev_lower_get_next_private(struct net_device *dev,
2916 struct list_head **iter);
2917 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
2918 struct list_head **iter);
2920 #define netdev_for_each_lower_private(dev, priv, iter) \
2921 for (iter = (dev)->adj_list.lower.next, \
2922 priv = netdev_lower_get_next_private(dev, &(iter)); \
2924 priv = netdev_lower_get_next_private(dev, &(iter)))
2926 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
2927 for (iter = &(dev)->adj_list.lower, \
2928 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
2930 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
2932 void *netdev_adjacent_get_private(struct list_head *adj_list);
2933 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
2934 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
2935 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
2936 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
2937 int netdev_master_upper_dev_link(struct net_device *dev,
2938 struct net_device *upper_dev);
2939 int netdev_master_upper_dev_link_private(struct net_device *dev,
2940 struct net_device *upper_dev,
2942 void netdev_upper_dev_unlink(struct net_device *dev,
2943 struct net_device *upper_dev);
2944 void *netdev_lower_dev_get_private(struct net_device *dev,
2945 struct net_device *lower_dev);
2946 int skb_checksum_help(struct sk_buff *skb);
2947 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
2948 netdev_features_t features, bool tx_path);
2949 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
2950 netdev_features_t features);
2953 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
2955 return __skb_gso_segment(skb, features, true);
2957 __be16 skb_network_protocol(struct sk_buff *skb);
2959 static inline bool can_checksum_protocol(netdev_features_t features,
2962 return ((features & NETIF_F_GEN_CSUM) ||
2963 ((features & NETIF_F_V4_CSUM) &&
2964 protocol == htons(ETH_P_IP)) ||
2965 ((features & NETIF_F_V6_CSUM) &&
2966 protocol == htons(ETH_P_IPV6)) ||
2967 ((features & NETIF_F_FCOE_CRC) &&
2968 protocol == htons(ETH_P_FCOE)));
2972 void netdev_rx_csum_fault(struct net_device *dev);
2974 static inline void netdev_rx_csum_fault(struct net_device *dev)
2978 /* rx skb timestamps */
2979 void net_enable_timestamp(void);
2980 void net_disable_timestamp(void);
2982 #ifdef CONFIG_PROC_FS
2983 int __init dev_proc_init(void);
2985 #define dev_proc_init() 0
2988 int netdev_class_create_file_ns(struct class_attribute *class_attr,
2990 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
2993 static inline int netdev_class_create_file(struct class_attribute *class_attr)
2995 return netdev_class_create_file_ns(class_attr, NULL);
2998 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3000 netdev_class_remove_file_ns(class_attr, NULL);
3003 extern struct kobj_ns_type_operations net_ns_type_operations;
3005 const char *netdev_drivername(const struct net_device *dev);
3007 void linkwatch_run_queue(void);
3009 static inline netdev_features_t netdev_get_wanted_features(
3010 struct net_device *dev)
3012 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3014 netdev_features_t netdev_increment_features(netdev_features_t all,
3015 netdev_features_t one, netdev_features_t mask);
3017 /* Allow TSO being used on stacked device :
3018 * Performing the GSO segmentation before last device
3019 * is a performance improvement.
3021 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3022 netdev_features_t mask)
3024 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3027 int __netdev_update_features(struct net_device *dev);
3028 void netdev_update_features(struct net_device *dev);
3029 void netdev_change_features(struct net_device *dev);
3031 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3032 struct net_device *dev);
3034 netdev_features_t netif_skb_features(struct sk_buff *skb);
3036 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3038 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3040 /* check flags correspondence */
3041 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3042 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3043 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3044 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3045 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3046 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3048 return (features & feature) == feature;
3051 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3053 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3054 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3057 static inline bool netif_needs_gso(struct sk_buff *skb,
3058 netdev_features_t features)
3060 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3061 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3062 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3065 static inline void netif_set_gso_max_size(struct net_device *dev,
3068 dev->gso_max_size = size;
3071 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3072 int pulled_hlen, u16 mac_offset,
3075 skb->protocol = protocol;
3076 skb->encapsulation = 1;
3077 skb_push(skb, pulled_hlen);
3078 skb_reset_transport_header(skb);
3079 skb->mac_header = mac_offset;
3080 skb->network_header = skb->mac_header + mac_len;
3081 skb->mac_len = mac_len;
3084 static inline bool netif_is_macvlan(struct net_device *dev)
3086 return dev->priv_flags & IFF_MACVLAN;
3089 static inline bool netif_is_bond_master(struct net_device *dev)
3091 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3094 static inline bool netif_is_bond_slave(struct net_device *dev)
3096 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3099 static inline bool netif_supports_nofcs(struct net_device *dev)
3101 return dev->priv_flags & IFF_SUPP_NOFCS;
3104 extern struct pernet_operations __net_initdata loopback_net_ops;
3106 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3108 /* netdev_printk helpers, similar to dev_printk */
3110 static inline const char *netdev_name(const struct net_device *dev)
3112 if (dev->reg_state != NETREG_REGISTERED)
3113 return "(unregistered net_device)";
3118 int netdev_printk(const char *level, const struct net_device *dev,
3119 const char *format, ...);
3121 int netdev_emerg(const struct net_device *dev, const char *format, ...);
3123 int netdev_alert(const struct net_device *dev, const char *format, ...);
3125 int netdev_crit(const struct net_device *dev, const char *format, ...);
3127 int netdev_err(const struct net_device *dev, const char *format, ...);
3129 int netdev_warn(const struct net_device *dev, const char *format, ...);
3131 int netdev_notice(const struct net_device *dev, const char *format, ...);
3133 int netdev_info(const struct net_device *dev, const char *format, ...);
3135 #define MODULE_ALIAS_NETDEV(device) \
3136 MODULE_ALIAS("netdev-" device)
3138 #if defined(CONFIG_DYNAMIC_DEBUG)
3139 #define netdev_dbg(__dev, format, args...) \
3141 dynamic_netdev_dbg(__dev, format, ##args); \
3143 #elif defined(DEBUG)
3144 #define netdev_dbg(__dev, format, args...) \
3145 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3147 #define netdev_dbg(__dev, format, args...) \
3150 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3155 #if defined(VERBOSE_DEBUG)
3156 #define netdev_vdbg netdev_dbg
3159 #define netdev_vdbg(dev, format, args...) \
3162 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3168 * netdev_WARN() acts like dev_printk(), but with the key difference
3169 * of using a WARN/WARN_ON to get the message out, including the
3170 * file/line information and a backtrace.
3172 #define netdev_WARN(dev, format, args...) \
3173 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
3175 /* netif printk helpers, similar to netdev_printk */
3177 #define netif_printk(priv, type, level, dev, fmt, args...) \
3179 if (netif_msg_##type(priv)) \
3180 netdev_printk(level, (dev), fmt, ##args); \
3183 #define netif_level(level, priv, type, dev, fmt, args...) \
3185 if (netif_msg_##type(priv)) \
3186 netdev_##level(dev, fmt, ##args); \
3189 #define netif_emerg(priv, type, dev, fmt, args...) \
3190 netif_level(emerg, priv, type, dev, fmt, ##args)
3191 #define netif_alert(priv, type, dev, fmt, args...) \
3192 netif_level(alert, priv, type, dev, fmt, ##args)
3193 #define netif_crit(priv, type, dev, fmt, args...) \
3194 netif_level(crit, priv, type, dev, fmt, ##args)
3195 #define netif_err(priv, type, dev, fmt, args...) \
3196 netif_level(err, priv, type, dev, fmt, ##args)
3197 #define netif_warn(priv, type, dev, fmt, args...) \
3198 netif_level(warn, priv, type, dev, fmt, ##args)
3199 #define netif_notice(priv, type, dev, fmt, args...) \
3200 netif_level(notice, priv, type, dev, fmt, ##args)
3201 #define netif_info(priv, type, dev, fmt, args...) \
3202 netif_level(info, priv, type, dev, fmt, ##args)
3204 #if defined(CONFIG_DYNAMIC_DEBUG)
3205 #define netif_dbg(priv, type, netdev, format, args...) \
3207 if (netif_msg_##type(priv)) \
3208 dynamic_netdev_dbg(netdev, format, ##args); \
3210 #elif defined(DEBUG)
3211 #define netif_dbg(priv, type, dev, format, args...) \
3212 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3214 #define netif_dbg(priv, type, dev, format, args...) \
3217 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3222 #if defined(VERBOSE_DEBUG)
3223 #define netif_vdbg netif_dbg
3225 #define netif_vdbg(priv, type, dev, format, args...) \
3228 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3234 * The list of packet types we will receive (as opposed to discard)
3235 * and the routines to invoke.
3237 * Why 16. Because with 16 the only overlap we get on a hash of the
3238 * low nibble of the protocol value is RARP/SNAP/X.25.
3240 * NOTE: That is no longer true with the addition of VLAN tags. Not
3241 * sure which should go first, but I bet it won't make much
3242 * difference if we are running VLANs. The good news is that
3243 * this protocol won't be in the list unless compiled in, so
3244 * the average user (w/out VLANs) will not be adversely affected.
3260 #define PTYPE_HASH_SIZE (16)
3261 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3263 #endif /* _LINUX_NETDEVICE_H */