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 <linux/prefetch.h>
34 #include <asm/cache.h>
35 #include <asm/byteorder.h>
37 #include <linux/percpu.h>
38 #include <linux/rculist.h>
39 #include <linux/dmaengine.h>
40 #include <linux/workqueue.h>
41 #include <linux/dynamic_queue_limits.h>
43 #include <linux/ethtool.h>
44 #include <net/net_namespace.h>
47 #include <net/dcbnl.h>
49 #include <net/netprio_cgroup.h>
51 #include <linux/netdev_features.h>
52 #include <linux/neighbour.h>
53 #include <uapi/linux/netdevice.h>
54 #include <uapi/linux/if_bonding.h>
61 /* 802.15.4 specific */
64 void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
67 /* Backlog congestion levels */
68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69 #define NET_RX_DROP 1 /* packet dropped */
72 * Transmit return codes: transmit return codes originate from three different
75 * - qdisc return codes
76 * - driver transmit return codes
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
88 /* qdisc ->enqueue() return codes. */
89 #define NET_XMIT_SUCCESS 0x00
90 #define NET_XMIT_DROP 0x01 /* skb dropped */
91 #define NET_XMIT_CN 0x02 /* congestion notification */
92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101 /* Driver transmit return codes */
102 #define NETDEV_TX_MASK 0xf0
105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
110 typedef enum netdev_tx netdev_tx_t;
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 static inline bool dev_xmit_complete(int rc)
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 if (likely(rc < NET_XMIT_MASK))
131 * Compute the worst case header length according to the protocols
135 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
136 # if defined(CONFIG_MAC80211_MESH)
137 # define LL_MAX_HEADER 128
139 # define LL_MAX_HEADER 96
142 # define LL_MAX_HEADER 32
145 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
147 #define MAX_HEADER LL_MAX_HEADER
149 #define MAX_HEADER (LL_MAX_HEADER + 48)
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
157 struct net_device_stats {
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
167 unsigned long collisions;
168 unsigned long rx_length_errors;
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
184 #include <linux/cache.h>
185 #include <linux/skbuff.h>
188 #include <linux/static_key.h>
189 extern struct static_key rps_needed;
196 struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
200 #define NETDEV_HW_ADDR_T_LAN 1
201 #define NETDEV_HW_ADDR_T_SAN 2
202 #define NETDEV_HW_ADDR_T_SLAVE 3
203 #define NETDEV_HW_ADDR_T_UNICAST 4
204 #define NETDEV_HW_ADDR_T_MULTICAST 5
209 struct rcu_head rcu_head;
212 struct netdev_hw_addr_list {
213 struct list_head list;
217 #define netdev_hw_addr_list_count(l) ((l)->count)
218 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219 #define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
222 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224 #define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
227 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
229 #define netdev_for_each_mc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
237 /* cached hardware header; allow for machine alignment needs. */
238 #define HH_DATA_MOD 16
239 #define HH_DATA_OFF(__len) \
240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
241 #define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
246 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
254 #define LL_RESERVED_SPACE(dev) \
255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
256 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
262 const void *saddr, unsigned int len);
263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
270 /* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
275 enum netdev_state_t {
277 __LINK_STATE_PRESENT,
278 __LINK_STATE_NOCARRIER,
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
285 * This structure holds at boot time configured netdevice settings. They
286 * are then used in the device probing.
288 struct netdev_boot_setup {
292 #define NETDEV_BOOT_SETUP_MAX 8
294 int __init netdev_boot_setup(char *str);
297 * Structure for NAPI scheduling similar to tasklet but with weighting
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
306 struct list_head poll_list;
310 unsigned int gro_count;
311 int (*poll)(struct napi_struct *, int);
312 #ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
316 struct net_device *dev;
317 struct sk_buff *gro_list;
319 struct hrtimer timer;
320 struct list_head dev_list;
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
326 NAPI_STATE_SCHED, /* Poll is scheduled */
327 NAPI_STATE_DISABLE, /* Disable pending */
328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
329 NAPI_STATE_HASHED, /* In NAPI hash */
339 typedef enum gro_result gro_result_t;
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
373 * are registered on exact device (ptype->dev == skb->dev).
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
382 enum rx_handler_result {
388 typedef enum rx_handler_result rx_handler_result_t;
389 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
391 void __napi_schedule(struct napi_struct *n);
392 void __napi_schedule_irqoff(struct napi_struct *n);
394 static inline bool napi_disable_pending(struct napi_struct *n)
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
400 * napi_schedule_prep - check if napi can be scheduled
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
408 static inline bool napi_schedule_prep(struct napi_struct *n)
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
415 * napi_schedule - schedule NAPI poll
418 * Schedule NAPI poll routine to be called if it is not already
421 static inline void napi_schedule(struct napi_struct *n)
423 if (napi_schedule_prep(n))
428 * napi_schedule_irqoff - schedule NAPI poll
431 * Variant of napi_schedule(), assuming hard irqs are masked.
433 static inline void napi_schedule_irqoff(struct napi_struct *n)
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
439 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
440 static inline bool napi_reschedule(struct napi_struct *napi)
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
449 void __napi_complete(struct napi_struct *n);
450 void napi_complete_done(struct napi_struct *n, int work_done);
452 * napi_complete - NAPI processing complete
455 * Mark NAPI processing as complete.
456 * Consider using napi_complete_done() instead.
458 static inline void napi_complete(struct napi_struct *n)
460 return napi_complete_done(n, 0);
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
470 struct napi_struct *napi_by_id(unsigned int napi_id);
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
476 * generate a new napi_id and store a @napi under it in napi_hash
478 void napi_hash_add(struct napi_struct *napi);
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
487 void napi_hash_del(struct napi_struct *napi);
490 * napi_disable - prevent NAPI from scheduling
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
496 void napi_disable(struct napi_struct *n);
499 * napi_enable - enable NAPI scheduling
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
505 static inline void napi_enable(struct napi_struct *n)
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
508 smp_mb__before_atomic();
509 clear_bit(NAPI_STATE_SCHED, &n->state);
514 * napi_synchronize - wait until NAPI is not running
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
521 static inline void napi_synchronize(const struct napi_struct *n)
523 while (test_bit(NAPI_STATE_SCHED, &n->state))
527 # define napi_synchronize(n) barrier()
530 enum netdev_queue_state_t {
531 __QUEUE_STATE_DRV_XOFF,
532 __QUEUE_STATE_STACK_XOFF,
533 __QUEUE_STATE_FROZEN,
536 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
537 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
538 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
540 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
541 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
543 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
548 * netif_tx_* functions below are used to manipulate this flag. The
549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
550 * queue independently. The netif_xmit_*stopped functions below are called
551 * to check if the queue has been stopped by the driver or stack (either
552 * of the XOFF bits are set in the state). Drivers should not need to call
553 * netif_xmit*stopped functions, they should only be using netif_tx_*.
556 struct netdev_queue {
560 struct net_device *dev;
561 struct Qdisc __rcu *qdisc;
562 struct Qdisc *qdisc_sleeping;
566 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
575 * please use this field instead of dev->trans_start
577 unsigned long trans_start;
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
583 unsigned long trans_timeout;
590 } ____cacheline_aligned_in_smp;
592 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
594 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
603 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 * This structure holds an RPS map which can be of variable length. The
611 * map is an array of CPUs.
618 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
621 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
622 * tail pointer for that CPU's input queue at the time of last enqueue, and
623 * a hardware filter index.
625 struct rps_dev_flow {
628 unsigned int last_qtail;
630 #define RPS_NO_FILTER 0xffff
633 * The rps_dev_flow_table structure contains a table of flow mappings.
635 struct rps_dev_flow_table {
638 struct rps_dev_flow flows[0];
640 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
641 ((_num) * sizeof(struct rps_dev_flow)))
644 * The rps_sock_flow_table contains mappings of flows to the last CPU
645 * on which they were processed by the application (set in recvmsg).
646 * Each entry is a 32bit value. Upper part is the high order bits
647 * of flow hash, lower part is cpu number.
648 * rps_cpu_mask is used to partition the space, depending on number of
649 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
650 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
651 * meaning we use 32-6=26 bits for the hash.
653 struct rps_sock_flow_table {
656 u32 ents[0] ____cacheline_aligned_in_smp;
658 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
660 #define RPS_NO_CPU 0xffff
662 extern u32 rps_cpu_mask;
663 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
669 unsigned int index = hash & table->mask;
670 u32 val = hash & ~rps_cpu_mask;
672 /* We only give a hint, preemption can change cpu under us */
673 val |= raw_smp_processor_id();
675 if (table->ents[index] != val)
676 table->ents[index] = val;
680 #ifdef CONFIG_RFS_ACCEL
681 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
684 #endif /* CONFIG_RPS */
686 /* This structure contains an instance of an RX queue. */
687 struct netdev_rx_queue {
689 struct rps_map __rcu *rps_map;
690 struct rps_dev_flow_table __rcu *rps_flow_table;
693 struct net_device *dev;
694 } ____cacheline_aligned_in_smp;
697 * RX queue sysfs structures and functions.
699 struct rx_queue_attribute {
700 struct attribute attr;
701 ssize_t (*show)(struct netdev_rx_queue *queue,
702 struct rx_queue_attribute *attr, char *buf);
703 ssize_t (*store)(struct netdev_rx_queue *queue,
704 struct rx_queue_attribute *attr, const char *buf, size_t len);
709 * This structure holds an XPS map which can be of variable length. The
710 * map is an array of queues.
714 unsigned int alloc_len;
718 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
719 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
723 * This structure holds all XPS maps for device. Maps are indexed by CPU.
725 struct xps_dev_maps {
727 struct xps_map __rcu *cpu_map[0];
729 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
730 (nr_cpu_ids * sizeof(struct xps_map *)))
731 #endif /* CONFIG_XPS */
733 #define TC_MAX_QUEUE 16
734 #define TC_BITMASK 15
735 /* HW offloaded queuing disciplines txq count and offset maps */
736 struct netdev_tc_txq {
741 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
743 * This structure is to hold information about the device
744 * configured to run FCoE protocol stack.
746 struct netdev_fcoe_hbainfo {
747 char manufacturer[64];
748 char serial_number[64];
749 char hardware_version[64];
750 char driver_version[64];
751 char optionrom_version[64];
752 char firmware_version[64];
754 char model_description[256];
758 #define MAX_PHYS_ITEM_ID_LEN 32
760 /* This structure holds a unique identifier to identify some
761 * physical item (port for example) used by a netdevice.
763 struct netdev_phys_item_id {
764 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
765 unsigned char id_len;
768 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
769 struct sk_buff *skb);
772 * This structure defines the management hooks for network devices.
773 * The following hooks can be defined; unless noted otherwise, they are
774 * optional and can be filled with a null pointer.
776 * int (*ndo_init)(struct net_device *dev);
777 * This function is called once when network device is registered.
778 * The network device can use this to any late stage initializaton
779 * or semantic validattion. It can fail with an error code which will
780 * be propogated back to register_netdev
782 * void (*ndo_uninit)(struct net_device *dev);
783 * This function is called when device is unregistered or when registration
784 * fails. It is not called if init fails.
786 * int (*ndo_open)(struct net_device *dev);
787 * This function is called when network device transistions to the up
790 * int (*ndo_stop)(struct net_device *dev);
791 * This function is called when network device transistions to the down
794 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
795 * struct net_device *dev);
796 * Called when a packet needs to be transmitted.
797 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
798 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
799 * Required can not be NULL.
801 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
802 * void *accel_priv, select_queue_fallback_t fallback);
803 * Called to decide which queue to when device supports multiple
806 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
807 * This function is called to allow device receiver to make
808 * changes to configuration when multicast or promiscious is enabled.
810 * void (*ndo_set_rx_mode)(struct net_device *dev);
811 * This function is called device changes address list filtering.
812 * If driver handles unicast address filtering, it should set
813 * IFF_UNICAST_FLT to its priv_flags.
815 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
816 * This function is called when the Media Access Control address
817 * needs to be changed. If this interface is not defined, the
818 * mac address can not be changed.
820 * int (*ndo_validate_addr)(struct net_device *dev);
821 * Test if Media Access Control address is valid for the device.
823 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
824 * Called when a user request an ioctl which can't be handled by
825 * the generic interface code. If not defined ioctl's return
826 * not supported error code.
828 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
829 * Used to set network devices bus interface parameters. This interface
830 * is retained for legacy reason, new devices should use the bus
831 * interface (PCI) for low level management.
833 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
834 * Called when a user wants to change the Maximum Transfer Unit
835 * of a device. If not defined, any request to change MTU will
836 * will return an error.
838 * void (*ndo_tx_timeout)(struct net_device *dev);
839 * Callback uses when the transmitter has not made any progress
840 * for dev->watchdog ticks.
842 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
843 * struct rtnl_link_stats64 *storage);
844 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
845 * Called when a user wants to get the network device usage
846 * statistics. Drivers must do one of the following:
847 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
848 * rtnl_link_stats64 structure passed by the caller.
849 * 2. Define @ndo_get_stats to update a net_device_stats structure
850 * (which should normally be dev->stats) and return a pointer to
851 * it. The structure may be changed asynchronously only if each
852 * field is written atomically.
853 * 3. Update dev->stats asynchronously and atomically, and define
856 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
857 * If device support VLAN filtering this function is called when a
858 * VLAN id is registered.
860 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
861 * If device support VLAN filtering this function is called when a
862 * VLAN id is unregistered.
864 * void (*ndo_poll_controller)(struct net_device *dev);
866 * SR-IOV management functions.
867 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
868 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
869 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
871 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
872 * int (*ndo_get_vf_config)(struct net_device *dev,
873 * int vf, struct ifla_vf_info *ivf);
874 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
875 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
876 * struct nlattr *port[]);
877 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
878 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
879 * Called to setup 'tc' number of traffic classes in the net device. This
880 * is always called from the stack with the rtnl lock held and netif tx
881 * queues stopped. This allows the netdevice to perform queue management
884 * Fiber Channel over Ethernet (FCoE) offload functions.
885 * int (*ndo_fcoe_enable)(struct net_device *dev);
886 * Called when the FCoE protocol stack wants to start using LLD for FCoE
887 * so the underlying device can perform whatever needed configuration or
888 * initialization to support acceleration of FCoE traffic.
890 * int (*ndo_fcoe_disable)(struct net_device *dev);
891 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
892 * so the underlying device can perform whatever needed clean-ups to
893 * stop supporting acceleration of FCoE traffic.
895 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
896 * struct scatterlist *sgl, unsigned int sgc);
897 * Called when the FCoE Initiator wants to initialize an I/O that
898 * is a possible candidate for Direct Data Placement (DDP). The LLD can
899 * perform necessary setup and returns 1 to indicate the device is set up
900 * successfully to perform DDP on this I/O, otherwise this returns 0.
902 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
903 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
904 * indicated by the FC exchange id 'xid', so the underlying device can
905 * clean up and reuse resources for later DDP requests.
907 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
908 * struct scatterlist *sgl, unsigned int sgc);
909 * Called when the FCoE Target wants to initialize an I/O that
910 * is a possible candidate for Direct Data Placement (DDP). The LLD can
911 * perform necessary setup and returns 1 to indicate the device is set up
912 * successfully to perform DDP on this I/O, otherwise this returns 0.
914 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
915 * struct netdev_fcoe_hbainfo *hbainfo);
916 * Called when the FCoE Protocol stack wants information on the underlying
917 * device. This information is utilized by the FCoE protocol stack to
918 * register attributes with Fiber Channel management service as per the
919 * FC-GS Fabric Device Management Information(FDMI) specification.
921 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
922 * Called when the underlying device wants to override default World Wide
923 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
924 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
925 * protocol stack to use.
928 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
929 * u16 rxq_index, u32 flow_id);
930 * Set hardware filter for RFS. rxq_index is the target queue index;
931 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
932 * Return the filter ID on success, or a negative error code.
934 * Slave management functions (for bridge, bonding, etc).
935 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
936 * Called to make another netdev an underling.
938 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
939 * Called to release previously enslaved netdev.
941 * Feature/offload setting functions.
942 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
943 * netdev_features_t features);
944 * Adjusts the requested feature flags according to device-specific
945 * constraints, and returns the resulting flags. Must not modify
948 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
949 * Called to update device configuration to new features. Passed
950 * feature set might be less than what was returned by ndo_fix_features()).
951 * Must return >0 or -errno if it changed dev->features itself.
953 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
954 * struct net_device *dev,
955 * const unsigned char *addr, u16 vid, u16 flags)
956 * Adds an FDB entry to dev for addr.
957 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
958 * struct net_device *dev,
959 * const unsigned char *addr, u16 vid)
960 * Deletes the FDB entry from dev coresponding to addr.
961 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
962 * struct net_device *dev, struct net_device *filter_dev,
964 * Used to add FDB entries to dump requests. Implementers should add
965 * entries to skb and update idx with the number of entries.
967 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
968 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
969 * struct net_device *dev, u32 filter_mask)
971 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
972 * Called to change device carrier. Soft-devices (like dummy, team, etc)
973 * which do not represent real hardware may define this to allow their
974 * userspace components to manage their virtual carrier state. Devices
975 * that determine carrier state from physical hardware properties (eg
976 * network cables) or protocol-dependent mechanisms (eg
977 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
979 * int (*ndo_get_phys_port_id)(struct net_device *dev,
980 * struct netdev_phys_item_id *ppid);
981 * Called to get ID of physical port of this device. If driver does
982 * not implement this, it is assumed that the hw is not able to have
983 * multiple net devices on single physical port.
985 * void (*ndo_add_vxlan_port)(struct net_device *dev,
986 * sa_family_t sa_family, __be16 port);
987 * Called by vxlan to notiy a driver about the UDP port and socket
988 * address family that vxlan is listnening to. It is called only when
989 * a new port starts listening. The operation is protected by the
990 * vxlan_net->sock_lock.
992 * void (*ndo_del_vxlan_port)(struct net_device *dev,
993 * sa_family_t sa_family, __be16 port);
994 * Called by vxlan to notify the driver about a UDP port and socket
995 * address family that vxlan is not listening to anymore. The operation
996 * is protected by the vxlan_net->sock_lock.
998 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
999 * struct net_device *dev)
1000 * Called by upper layer devices to accelerate switching or other
1001 * station functionality into hardware. 'pdev is the lowerdev
1002 * to use for the offload and 'dev' is the net device that will
1003 * back the offload. Returns a pointer to the private structure
1004 * the upper layer will maintain.
1005 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1006 * Called by upper layer device to delete the station created
1007 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1008 * the station and priv is the structure returned by the add
1010 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1011 * struct net_device *dev,
1013 * Callback to use for xmit over the accelerated station. This
1014 * is used in place of ndo_start_xmit on accelerated net
1016 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1017 * struct net_device *dev
1018 * netdev_features_t features);
1019 * Called by core transmit path to determine if device is capable of
1020 * performing offload operations on a given packet. This is to give
1021 * the device an opportunity to implement any restrictions that cannot
1022 * be otherwise expressed by feature flags. The check is called with
1023 * the set of features that the stack has calculated and it returns
1024 * those the driver believes to be appropriate.
1026 struct net_device_ops {
1027 int (*ndo_init)(struct net_device *dev);
1028 void (*ndo_uninit)(struct net_device *dev);
1029 int (*ndo_open)(struct net_device *dev);
1030 int (*ndo_stop)(struct net_device *dev);
1031 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1032 struct net_device *dev);
1033 u16 (*ndo_select_queue)(struct net_device *dev,
1034 struct sk_buff *skb,
1036 select_queue_fallback_t fallback);
1037 void (*ndo_change_rx_flags)(struct net_device *dev,
1039 void (*ndo_set_rx_mode)(struct net_device *dev);
1040 int (*ndo_set_mac_address)(struct net_device *dev,
1042 int (*ndo_validate_addr)(struct net_device *dev);
1043 int (*ndo_do_ioctl)(struct net_device *dev,
1044 struct ifreq *ifr, int cmd);
1045 int (*ndo_set_config)(struct net_device *dev,
1047 int (*ndo_change_mtu)(struct net_device *dev,
1049 int (*ndo_neigh_setup)(struct net_device *dev,
1050 struct neigh_parms *);
1051 void (*ndo_tx_timeout) (struct net_device *dev);
1053 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1054 struct rtnl_link_stats64 *storage);
1055 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1057 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1058 __be16 proto, u16 vid);
1059 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1060 __be16 proto, u16 vid);
1061 #ifdef CONFIG_NET_POLL_CONTROLLER
1062 void (*ndo_poll_controller)(struct net_device *dev);
1063 int (*ndo_netpoll_setup)(struct net_device *dev,
1064 struct netpoll_info *info);
1065 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1067 #ifdef CONFIG_NET_RX_BUSY_POLL
1068 int (*ndo_busy_poll)(struct napi_struct *dev);
1070 int (*ndo_set_vf_mac)(struct net_device *dev,
1071 int queue, u8 *mac);
1072 int (*ndo_set_vf_vlan)(struct net_device *dev,
1073 int queue, u16 vlan, u8 qos);
1074 int (*ndo_set_vf_rate)(struct net_device *dev,
1075 int vf, int min_tx_rate,
1077 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1078 int vf, bool setting);
1079 int (*ndo_get_vf_config)(struct net_device *dev,
1081 struct ifla_vf_info *ivf);
1082 int (*ndo_set_vf_link_state)(struct net_device *dev,
1083 int vf, int link_state);
1084 int (*ndo_set_vf_port)(struct net_device *dev,
1086 struct nlattr *port[]);
1087 int (*ndo_get_vf_port)(struct net_device *dev,
1088 int vf, struct sk_buff *skb);
1089 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1090 #if IS_ENABLED(CONFIG_FCOE)
1091 int (*ndo_fcoe_enable)(struct net_device *dev);
1092 int (*ndo_fcoe_disable)(struct net_device *dev);
1093 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1095 struct scatterlist *sgl,
1097 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1099 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1101 struct scatterlist *sgl,
1103 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1104 struct netdev_fcoe_hbainfo *hbainfo);
1107 #if IS_ENABLED(CONFIG_LIBFCOE)
1108 #define NETDEV_FCOE_WWNN 0
1109 #define NETDEV_FCOE_WWPN 1
1110 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1111 u64 *wwn, int type);
1114 #ifdef CONFIG_RFS_ACCEL
1115 int (*ndo_rx_flow_steer)(struct net_device *dev,
1116 const struct sk_buff *skb,
1120 int (*ndo_add_slave)(struct net_device *dev,
1121 struct net_device *slave_dev);
1122 int (*ndo_del_slave)(struct net_device *dev,
1123 struct net_device *slave_dev);
1124 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1125 netdev_features_t features);
1126 int (*ndo_set_features)(struct net_device *dev,
1127 netdev_features_t features);
1128 int (*ndo_neigh_construct)(struct neighbour *n);
1129 void (*ndo_neigh_destroy)(struct neighbour *n);
1131 int (*ndo_fdb_add)(struct ndmsg *ndm,
1132 struct nlattr *tb[],
1133 struct net_device *dev,
1134 const unsigned char *addr,
1137 int (*ndo_fdb_del)(struct ndmsg *ndm,
1138 struct nlattr *tb[],
1139 struct net_device *dev,
1140 const unsigned char *addr,
1142 int (*ndo_fdb_dump)(struct sk_buff *skb,
1143 struct netlink_callback *cb,
1144 struct net_device *dev,
1145 struct net_device *filter_dev,
1148 int (*ndo_bridge_setlink)(struct net_device *dev,
1149 struct nlmsghdr *nlh,
1151 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1153 struct net_device *dev,
1155 int (*ndo_bridge_dellink)(struct net_device *dev,
1156 struct nlmsghdr *nlh,
1158 int (*ndo_change_carrier)(struct net_device *dev,
1160 int (*ndo_get_phys_port_id)(struct net_device *dev,
1161 struct netdev_phys_item_id *ppid);
1162 void (*ndo_add_vxlan_port)(struct net_device *dev,
1163 sa_family_t sa_family,
1165 void (*ndo_del_vxlan_port)(struct net_device *dev,
1166 sa_family_t sa_family,
1169 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1170 struct net_device *dev);
1171 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1174 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1175 struct net_device *dev,
1177 int (*ndo_get_lock_subclass)(struct net_device *dev);
1178 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1179 struct net_device *dev,
1180 netdev_features_t features);
1184 * enum net_device_priv_flags - &struct net_device priv_flags
1186 * These are the &struct net_device, they are only set internally
1187 * by drivers and used in the kernel. These flags are invisible to
1188 * userspace, this means that the order of these flags can change
1189 * during any kernel release.
1191 * You should have a pretty good reason to be extending these flags.
1193 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1194 * @IFF_EBRIDGE: Ethernet bridging device
1195 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1196 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1197 * @IFF_MASTER_ALB: bonding master, balance-alb
1198 * @IFF_BONDING: bonding master or slave
1199 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1200 * @IFF_ISATAP: ISATAP interface (RFC4214)
1201 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1202 * @IFF_WAN_HDLC: WAN HDLC device
1203 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1205 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1206 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1207 * @IFF_MACVLAN_PORT: device used as macvlan port
1208 * @IFF_BRIDGE_PORT: device used as bridge port
1209 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1210 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1211 * @IFF_UNICAST_FLT: Supports unicast filtering
1212 * @IFF_TEAM_PORT: device used as team port
1213 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1214 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1215 * change when it's running
1216 * @IFF_MACVLAN: Macvlan device
1218 enum netdev_priv_flags {
1219 IFF_802_1Q_VLAN = 1<<0,
1221 IFF_SLAVE_INACTIVE = 1<<2,
1222 IFF_MASTER_8023AD = 1<<3,
1223 IFF_MASTER_ALB = 1<<4,
1225 IFF_SLAVE_NEEDARP = 1<<6,
1227 IFF_MASTER_ARPMON = 1<<8,
1228 IFF_WAN_HDLC = 1<<9,
1229 IFF_XMIT_DST_RELEASE = 1<<10,
1230 IFF_DONT_BRIDGE = 1<<11,
1231 IFF_DISABLE_NETPOLL = 1<<12,
1232 IFF_MACVLAN_PORT = 1<<13,
1233 IFF_BRIDGE_PORT = 1<<14,
1234 IFF_OVS_DATAPATH = 1<<15,
1235 IFF_TX_SKB_SHARING = 1<<16,
1236 IFF_UNICAST_FLT = 1<<17,
1237 IFF_TEAM_PORT = 1<<18,
1238 IFF_SUPP_NOFCS = 1<<19,
1239 IFF_LIVE_ADDR_CHANGE = 1<<20,
1240 IFF_MACVLAN = 1<<21,
1241 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1242 IFF_IPVLAN_MASTER = 1<<23,
1243 IFF_IPVLAN_SLAVE = 1<<24,
1246 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1247 #define IFF_EBRIDGE IFF_EBRIDGE
1248 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1249 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1250 #define IFF_MASTER_ALB IFF_MASTER_ALB
1251 #define IFF_BONDING IFF_BONDING
1252 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1253 #define IFF_ISATAP IFF_ISATAP
1254 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1255 #define IFF_WAN_HDLC IFF_WAN_HDLC
1256 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1257 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1258 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1259 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1260 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1261 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1262 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1263 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1264 #define IFF_TEAM_PORT IFF_TEAM_PORT
1265 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1266 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1267 #define IFF_MACVLAN IFF_MACVLAN
1268 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1269 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1270 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1273 * struct net_device - The DEVICE structure.
1274 * Actually, this whole structure is a big mistake. It mixes I/O
1275 * data with strictly "high-level" data, and it has to know about
1276 * almost every data structure used in the INET module.
1278 * @name: This is the first field of the "visible" part of this structure
1279 * (i.e. as seen by users in the "Space.c" file). It is the name
1282 * @name_hlist: Device name hash chain, please keep it close to name[]
1283 * @ifalias: SNMP alias
1284 * @mem_end: Shared memory end
1285 * @mem_start: Shared memory start
1286 * @base_addr: Device I/O address
1287 * @irq: Device IRQ number
1289 * @state: Generic network queuing layer state, see netdev_state_t
1290 * @dev_list: The global list of network devices
1291 * @napi_list: List entry, that is used for polling napi devices
1292 * @unreg_list: List entry, that is used, when we are unregistering the
1293 * device, see the function unregister_netdev
1294 * @close_list: List entry, that is used, when we are closing the device
1296 * @adj_list: Directly linked devices, like slaves for bonding
1297 * @all_adj_list: All linked devices, *including* neighbours
1298 * @features: Currently active device features
1299 * @hw_features: User-changeable features
1301 * @wanted_features: User-requested features
1302 * @vlan_features: Mask of features inheritable by VLAN devices
1304 * @hw_enc_features: Mask of features inherited by encapsulating devices
1305 * This field indicates what encapsulation
1306 * offloads the hardware is capable of doing,
1307 * and drivers will need to set them appropriately.
1309 * @mpls_features: Mask of features inheritable by MPLS
1311 * @ifindex: interface index
1312 * @iflink: unique device identifier
1314 * @stats: Statistics struct, which was left as a legacy, use
1315 * rtnl_link_stats64 instead
1317 * @rx_dropped: Dropped packets by core network,
1318 * do not use this in drivers
1319 * @tx_dropped: Dropped packets by core network,
1320 * do not use this in drivers
1322 * @carrier_changes: Stats to monitor carrier on<->off transitions
1324 * @wireless_handlers: List of functions to handle Wireless Extensions,
1326 * see <net/iw_handler.h> for details.
1327 * @wireless_data: Instance data managed by the core of wireless extensions
1329 * @netdev_ops: Includes several pointers to callbacks,
1330 * if one wants to override the ndo_*() functions
1331 * @ethtool_ops: Management operations
1332 * @fwd_ops: Management operations
1333 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1334 * of Layer 2 headers.
1336 * @flags: Interface flags (a la BSD)
1337 * @priv_flags: Like 'flags' but invisible to userspace,
1338 * see if.h for the definitions
1339 * @gflags: Global flags ( kept as legacy )
1340 * @padded: How much padding added by alloc_netdev()
1341 * @operstate: RFC2863 operstate
1342 * @link_mode: Mapping policy to operstate
1343 * @if_port: Selectable AUI, TP, ...
1345 * @mtu: Interface MTU value
1346 * @type: Interface hardware type
1347 * @hard_header_len: Hardware header length
1349 * @needed_headroom: Extra headroom the hardware may need, but not in all
1350 * cases can this be guaranteed
1351 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1352 * cases can this be guaranteed. Some cases also use
1353 * LL_MAX_HEADER instead to allocate the skb
1355 * interface address info:
1357 * @perm_addr: Permanent hw address
1358 * @addr_assign_type: Hw address assignment type
1359 * @addr_len: Hardware address length
1360 * @neigh_priv_len; Used in neigh_alloc(),
1361 * initialized only in atm/clip.c
1362 * @dev_id: Used to differentiate devices that share
1363 * the same link layer address
1364 * @dev_port: Used to differentiate devices that share
1366 * @addr_list_lock: XXX: need comments on this one
1367 * @uc: unicast mac addresses
1368 * @mc: multicast mac addresses
1369 * @dev_addrs: list of device hw addresses
1370 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1371 * @uc_promisc: Counter, that indicates, that promiscuous mode
1372 * has been enabled due to the need to listen to
1373 * additional unicast addresses in a device that
1374 * does not implement ndo_set_rx_mode()
1375 * @promiscuity: Number of times, the NIC is told to work in
1376 * Promiscuous mode, if it becomes 0 the NIC will
1377 * exit from working in Promiscuous mode
1378 * @allmulti: Counter, enables or disables allmulticast mode
1380 * @vlan_info: VLAN info
1381 * @dsa_ptr: dsa specific data
1382 * @tipc_ptr: TIPC specific data
1383 * @atalk_ptr: AppleTalk link
1384 * @ip_ptr: IPv4 specific data
1385 * @dn_ptr: DECnet specific data
1386 * @ip6_ptr: IPv6 specific data
1387 * @ax25_ptr: AX.25 specific data
1388 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1390 * @last_rx: Time of last Rx
1391 * @dev_addr: Hw address (before bcast,
1392 * because most packets are unicast)
1394 * @_rx: Array of RX queues
1395 * @num_rx_queues: Number of RX queues
1396 * allocated at register_netdev() time
1397 * @real_num_rx_queues: Number of RX queues currently active in device
1399 * @rx_handler: handler for received packets
1400 * @rx_handler_data: XXX: need comments on this one
1401 * @ingress_queue: XXX: need comments on this one
1402 * @broadcast: hw bcast address
1404 * @_tx: Array of TX queues
1405 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1406 * @real_num_tx_queues: Number of TX queues currently active in device
1407 * @qdisc: Root qdisc from userspace point of view
1408 * @tx_queue_len: Max frames per queue allowed
1409 * @tx_global_lock: XXX: need comments on this one
1411 * @xps_maps: XXX: need comments on this one
1413 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1414 * indexed by RX queue number. Assigned by driver.
1415 * This must only be set if the ndo_rx_flow_steer
1416 * operation is defined
1418 * @trans_start: Time (in jiffies) of last Tx
1419 * @watchdog_timeo: Represents the timeout that is used by
1420 * the watchdog ( see dev_watchdog() )
1421 * @watchdog_timer: List of timers
1423 * @pcpu_refcnt: Number of references to this device
1424 * @todo_list: Delayed register/unregister
1425 * @index_hlist: Device index hash chain
1426 * @link_watch_list: XXX: need comments on this one
1428 * @reg_state: Register/unregister state machine
1429 * @dismantle: Device is going to be freed
1430 * @rtnl_link_state: This enum represents the phases of creating
1433 * @destructor: Called from unregister,
1434 * can be used to call free_netdev
1435 * @npinfo: XXX: need comments on this one
1436 * @nd_net: Network namespace this network device is inside
1438 * @ml_priv: Mid-layer private
1439 * @lstats: Loopback statistics
1440 * @tstats: Tunnel statistics
1441 * @dstats: Dummy statistics
1442 * @vstats: Virtual ethernet statistics
1447 * @dev: Class/net/name entry
1448 * @sysfs_groups: Space for optional device, statistics and wireless
1451 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1452 * @rtnl_link_ops: Rtnl_link_ops
1454 * @gso_max_size: Maximum size of generic segmentation offload
1455 * @gso_max_segs: Maximum number of segments that can be passed to the
1457 * @gso_min_segs: Minimum number of segments that can be passed to the
1460 * @dcbnl_ops: Data Center Bridging netlink ops
1461 * @num_tc: Number of traffic classes in the net device
1462 * @tc_to_txq: XXX: need comments on this one
1463 * @prio_tc_map XXX: need comments on this one
1465 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1467 * @priomap: XXX: need comments on this one
1468 * @phydev: Physical device may attach itself
1469 * for hardware timestamping
1471 * @qdisc_tx_busylock: XXX: need comments on this one
1473 * @group: The group, that the device belongs to
1474 * @pm_qos_req: Power Management QoS object
1476 * FIXME: cleanup struct net_device such that network protocol info
1481 char name[IFNAMSIZ];
1482 struct hlist_node name_hlist;
1485 * I/O specific fields
1486 * FIXME: Merge these and struct ifmap into one
1488 unsigned long mem_end;
1489 unsigned long mem_start;
1490 unsigned long base_addr;
1494 * Some hardware also needs these fields (state,dev_list,
1495 * napi_list,unreg_list,close_list) but they are not
1496 * part of the usual set specified in Space.c.
1499 unsigned long state;
1501 struct list_head dev_list;
1502 struct list_head napi_list;
1503 struct list_head unreg_list;
1504 struct list_head close_list;
1505 struct list_head ptype_all;
1506 struct list_head ptype_specific;
1509 struct list_head upper;
1510 struct list_head lower;
1514 struct list_head upper;
1515 struct list_head lower;
1518 netdev_features_t features;
1519 netdev_features_t hw_features;
1520 netdev_features_t wanted_features;
1521 netdev_features_t vlan_features;
1522 netdev_features_t hw_enc_features;
1523 netdev_features_t mpls_features;
1528 struct net_device_stats stats;
1530 atomic_long_t rx_dropped;
1531 atomic_long_t tx_dropped;
1533 atomic_t carrier_changes;
1535 #ifdef CONFIG_WIRELESS_EXT
1536 const struct iw_handler_def * wireless_handlers;
1537 struct iw_public_data * wireless_data;
1539 const struct net_device_ops *netdev_ops;
1540 const struct ethtool_ops *ethtool_ops;
1541 const struct forwarding_accel_ops *fwd_ops;
1542 #ifdef CONFIG_NET_SWITCHDEV
1543 const struct swdev_ops *swdev_ops;
1546 const struct header_ops *header_ops;
1549 unsigned int priv_flags;
1551 unsigned short gflags;
1552 unsigned short padded;
1554 unsigned char operstate;
1555 unsigned char link_mode;
1557 unsigned char if_port;
1561 unsigned short type;
1562 unsigned short hard_header_len;
1564 unsigned short needed_headroom;
1565 unsigned short needed_tailroom;
1567 /* Interface address info. */
1568 unsigned char perm_addr[MAX_ADDR_LEN];
1569 unsigned char addr_assign_type;
1570 unsigned char addr_len;
1571 unsigned short neigh_priv_len;
1572 unsigned short dev_id;
1573 unsigned short dev_port;
1574 spinlock_t addr_list_lock;
1575 struct netdev_hw_addr_list uc;
1576 struct netdev_hw_addr_list mc;
1577 struct netdev_hw_addr_list dev_addrs;
1580 struct kset *queues_kset;
1583 unsigned char name_assign_type;
1586 unsigned int promiscuity;
1587 unsigned int allmulti;
1590 /* Protocol specific pointers */
1592 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1593 struct vlan_info __rcu *vlan_info;
1595 #if IS_ENABLED(CONFIG_NET_DSA)
1596 struct dsa_switch_tree *dsa_ptr;
1598 #if IS_ENABLED(CONFIG_TIPC)
1599 struct tipc_bearer __rcu *tipc_ptr;
1602 struct in_device __rcu *ip_ptr;
1603 struct dn_dev __rcu *dn_ptr;
1604 struct inet6_dev __rcu *ip6_ptr;
1606 struct wireless_dev *ieee80211_ptr;
1607 struct wpan_dev *ieee802154_ptr;
1610 * Cache lines mostly used on receive path (including eth_type_trans())
1612 unsigned long last_rx;
1614 /* Interface address info used in eth_type_trans() */
1615 unsigned char *dev_addr;
1619 struct netdev_rx_queue *_rx;
1621 unsigned int num_rx_queues;
1622 unsigned int real_num_rx_queues;
1626 unsigned long gro_flush_timeout;
1627 rx_handler_func_t __rcu *rx_handler;
1628 void __rcu *rx_handler_data;
1630 struct netdev_queue __rcu *ingress_queue;
1631 unsigned char broadcast[MAX_ADDR_LEN];
1635 * Cache lines mostly used on transmit path
1637 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1638 unsigned int num_tx_queues;
1639 unsigned int real_num_tx_queues;
1640 struct Qdisc *qdisc;
1641 unsigned long tx_queue_len;
1642 spinlock_t tx_global_lock;
1645 struct xps_dev_maps __rcu *xps_maps;
1647 #ifdef CONFIG_RFS_ACCEL
1648 struct cpu_rmap *rx_cpu_rmap;
1651 /* These may be needed for future network-power-down code. */
1654 * trans_start here is expensive for high speed devices on SMP,
1655 * please use netdev_queue->trans_start instead.
1657 unsigned long trans_start;
1660 struct timer_list watchdog_timer;
1662 int __percpu *pcpu_refcnt;
1663 struct list_head todo_list;
1665 struct hlist_node index_hlist;
1666 struct list_head link_watch_list;
1668 enum { NETREG_UNINITIALIZED=0,
1669 NETREG_REGISTERED, /* completed register_netdevice */
1670 NETREG_UNREGISTERING, /* called unregister_netdevice */
1671 NETREG_UNREGISTERED, /* completed unregister todo */
1672 NETREG_RELEASED, /* called free_netdev */
1673 NETREG_DUMMY, /* dummy device for NAPI poll */
1679 RTNL_LINK_INITIALIZED,
1680 RTNL_LINK_INITIALIZING,
1681 } rtnl_link_state:16;
1683 void (*destructor)(struct net_device *dev);
1685 #ifdef CONFIG_NETPOLL
1686 struct netpoll_info __rcu *npinfo;
1689 possible_net_t nd_net;
1691 /* mid-layer private */
1694 struct pcpu_lstats __percpu *lstats;
1695 struct pcpu_sw_netstats __percpu *tstats;
1696 struct pcpu_dstats __percpu *dstats;
1697 struct pcpu_vstats __percpu *vstats;
1700 struct garp_port __rcu *garp_port;
1701 struct mrp_port __rcu *mrp_port;
1704 const struct attribute_group *sysfs_groups[4];
1705 const struct attribute_group *sysfs_rx_queue_group;
1707 const struct rtnl_link_ops *rtnl_link_ops;
1709 /* for setting kernel sock attribute on TCP connection setup */
1710 #define GSO_MAX_SIZE 65536
1711 unsigned int gso_max_size;
1712 #define GSO_MAX_SEGS 65535
1716 const struct dcbnl_rtnl_ops *dcbnl_ops;
1719 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1720 u8 prio_tc_map[TC_BITMASK + 1];
1722 #if IS_ENABLED(CONFIG_FCOE)
1723 unsigned int fcoe_ddp_xid;
1725 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1726 struct netprio_map __rcu *priomap;
1728 struct phy_device *phydev;
1729 struct lock_class_key *qdisc_tx_busylock;
1731 struct pm_qos_request pm_qos_req;
1733 #define to_net_dev(d) container_of(d, struct net_device, dev)
1735 #define NETDEV_ALIGN 32
1738 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1740 return dev->prio_tc_map[prio & TC_BITMASK];
1744 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1746 if (tc >= dev->num_tc)
1749 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1754 void netdev_reset_tc(struct net_device *dev)
1757 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1758 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1762 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1764 if (tc >= dev->num_tc)
1767 dev->tc_to_txq[tc].count = count;
1768 dev->tc_to_txq[tc].offset = offset;
1773 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1775 if (num_tc > TC_MAX_QUEUE)
1778 dev->num_tc = num_tc;
1783 int netdev_get_num_tc(struct net_device *dev)
1789 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1792 return &dev->_tx[index];
1795 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1796 const struct sk_buff *skb)
1798 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1801 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1802 void (*f)(struct net_device *,
1803 struct netdev_queue *,
1809 for (i = 0; i < dev->num_tx_queues; i++)
1810 f(dev, &dev->_tx[i], arg);
1813 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1814 struct sk_buff *skb,
1818 * Net namespace inlines
1821 struct net *dev_net(const struct net_device *dev)
1823 return read_pnet(&dev->nd_net);
1827 void dev_net_set(struct net_device *dev, struct net *net)
1829 write_pnet(&dev->nd_net, net);
1832 static inline bool netdev_uses_dsa(struct net_device *dev)
1834 #if IS_ENABLED(CONFIG_NET_DSA)
1835 if (dev->dsa_ptr != NULL)
1836 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1842 * netdev_priv - access network device private data
1843 * @dev: network device
1845 * Get network device private data
1847 static inline void *netdev_priv(const struct net_device *dev)
1849 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1852 /* Set the sysfs physical device reference for the network logical device
1853 * if set prior to registration will cause a symlink during initialization.
1855 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1857 /* Set the sysfs device type for the network logical device to allow
1858 * fine-grained identification of different network device types. For
1859 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1861 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1863 /* Default NAPI poll() weight
1864 * Device drivers are strongly advised to not use bigger value
1866 #define NAPI_POLL_WEIGHT 64
1869 * netif_napi_add - initialize a napi context
1870 * @dev: network device
1871 * @napi: napi context
1872 * @poll: polling function
1873 * @weight: default weight
1875 * netif_napi_add() must be used to initialize a napi context prior to calling
1876 * *any* of the other napi related functions.
1878 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1879 int (*poll)(struct napi_struct *, int), int weight);
1882 * netif_napi_del - remove a napi context
1883 * @napi: napi context
1885 * netif_napi_del() removes a napi context from the network device napi list
1887 void netif_napi_del(struct napi_struct *napi);
1889 struct napi_gro_cb {
1890 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1893 /* Length of frag0. */
1894 unsigned int frag0_len;
1896 /* This indicates where we are processing relative to skb->data. */
1899 /* This is non-zero if the packet cannot be merged with the new skb. */
1902 /* Save the IP ID here and check when we get to the transport layer */
1905 /* Number of segments aggregated. */
1908 /* Start offset for remote checksum offload */
1909 u16 gro_remcsum_start;
1911 /* jiffies when first packet was created/queued */
1914 /* Used in ipv6_gro_receive() and foo-over-udp */
1917 /* This is non-zero if the packet may be of the same flow. */
1920 /* Used in udp_gro_receive */
1923 /* GRO checksum is valid */
1926 /* Number of checksums via CHECKSUM_UNNECESSARY */
1931 #define NAPI_GRO_FREE 1
1932 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1934 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1939 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1942 /* used in skb_gro_receive() slow path */
1943 struct sk_buff *last;
1946 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1948 struct packet_type {
1949 __be16 type; /* This is really htons(ether_type). */
1950 struct net_device *dev; /* NULL is wildcarded here */
1951 int (*func) (struct sk_buff *,
1952 struct net_device *,
1953 struct packet_type *,
1954 struct net_device *);
1955 bool (*id_match)(struct packet_type *ptype,
1957 void *af_packet_priv;
1958 struct list_head list;
1961 struct offload_callbacks {
1962 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1963 netdev_features_t features);
1964 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1965 struct sk_buff *skb);
1966 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1969 struct packet_offload {
1970 __be16 type; /* This is really htons(ether_type). */
1971 struct offload_callbacks callbacks;
1972 struct list_head list;
1977 struct udp_offload_callbacks {
1978 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1979 struct sk_buff *skb,
1980 struct udp_offload *uoff);
1981 int (*gro_complete)(struct sk_buff *skb,
1983 struct udp_offload *uoff);
1986 struct udp_offload {
1989 struct udp_offload_callbacks callbacks;
1992 /* often modified stats are per cpu, other are shared (netdev->stats) */
1993 struct pcpu_sw_netstats {
1998 struct u64_stats_sync syncp;
2001 #define netdev_alloc_pcpu_stats(type) \
2003 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2006 for_each_possible_cpu(i) { \
2007 typeof(type) *stat; \
2008 stat = per_cpu_ptr(pcpu_stats, i); \
2009 u64_stats_init(&stat->syncp); \
2015 #include <linux/notifier.h>
2017 /* netdevice notifier chain. Please remember to update the rtnetlink
2018 * notification exclusion list in rtnetlink_event() when adding new
2021 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2022 #define NETDEV_DOWN 0x0002
2023 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2024 detected a hardware crash and restarted
2025 - we can use this eg to kick tcp sessions
2027 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2028 #define NETDEV_REGISTER 0x0005
2029 #define NETDEV_UNREGISTER 0x0006
2030 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2031 #define NETDEV_CHANGEADDR 0x0008
2032 #define NETDEV_GOING_DOWN 0x0009
2033 #define NETDEV_CHANGENAME 0x000A
2034 #define NETDEV_FEAT_CHANGE 0x000B
2035 #define NETDEV_BONDING_FAILOVER 0x000C
2036 #define NETDEV_PRE_UP 0x000D
2037 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2038 #define NETDEV_POST_TYPE_CHANGE 0x000F
2039 #define NETDEV_POST_INIT 0x0010
2040 #define NETDEV_UNREGISTER_FINAL 0x0011
2041 #define NETDEV_RELEASE 0x0012
2042 #define NETDEV_NOTIFY_PEERS 0x0013
2043 #define NETDEV_JOIN 0x0014
2044 #define NETDEV_CHANGEUPPER 0x0015
2045 #define NETDEV_RESEND_IGMP 0x0016
2046 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2047 #define NETDEV_CHANGEINFODATA 0x0018
2048 #define NETDEV_BONDING_INFO 0x0019
2050 int register_netdevice_notifier(struct notifier_block *nb);
2051 int unregister_netdevice_notifier(struct notifier_block *nb);
2053 struct netdev_notifier_info {
2054 struct net_device *dev;
2057 struct netdev_notifier_change_info {
2058 struct netdev_notifier_info info; /* must be first */
2059 unsigned int flags_changed;
2062 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2063 struct net_device *dev)
2068 static inline struct net_device *
2069 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2074 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2077 extern rwlock_t dev_base_lock; /* Device list lock */
2079 #define for_each_netdev(net, d) \
2080 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2081 #define for_each_netdev_reverse(net, d) \
2082 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2083 #define for_each_netdev_rcu(net, d) \
2084 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2085 #define for_each_netdev_safe(net, d, n) \
2086 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2087 #define for_each_netdev_continue(net, d) \
2088 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2089 #define for_each_netdev_continue_rcu(net, d) \
2090 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2091 #define for_each_netdev_in_bond_rcu(bond, slave) \
2092 for_each_netdev_rcu(&init_net, slave) \
2093 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2094 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2096 static inline struct net_device *next_net_device(struct net_device *dev)
2098 struct list_head *lh;
2102 lh = dev->dev_list.next;
2103 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2106 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2108 struct list_head *lh;
2112 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2113 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2116 static inline struct net_device *first_net_device(struct net *net)
2118 return list_empty(&net->dev_base_head) ? NULL :
2119 net_device_entry(net->dev_base_head.next);
2122 static inline struct net_device *first_net_device_rcu(struct net *net)
2124 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2126 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2129 int netdev_boot_setup_check(struct net_device *dev);
2130 unsigned long netdev_boot_base(const char *prefix, int unit);
2131 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2132 const char *hwaddr);
2133 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2134 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2135 void dev_add_pack(struct packet_type *pt);
2136 void dev_remove_pack(struct packet_type *pt);
2137 void __dev_remove_pack(struct packet_type *pt);
2138 void dev_add_offload(struct packet_offload *po);
2139 void dev_remove_offload(struct packet_offload *po);
2141 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2142 unsigned short mask);
2143 struct net_device *dev_get_by_name(struct net *net, const char *name);
2144 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2145 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2146 int dev_alloc_name(struct net_device *dev, const char *name);
2147 int dev_open(struct net_device *dev);
2148 int dev_close(struct net_device *dev);
2149 void dev_disable_lro(struct net_device *dev);
2150 int dev_loopback_xmit(struct sk_buff *newskb);
2151 int dev_queue_xmit(struct sk_buff *skb);
2152 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2153 int register_netdevice(struct net_device *dev);
2154 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2155 void unregister_netdevice_many(struct list_head *head);
2156 static inline void unregister_netdevice(struct net_device *dev)
2158 unregister_netdevice_queue(dev, NULL);
2161 int netdev_refcnt_read(const struct net_device *dev);
2162 void free_netdev(struct net_device *dev);
2163 void netdev_freemem(struct net_device *dev);
2164 void synchronize_net(void);
2165 int init_dummy_netdev(struct net_device *dev);
2167 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2168 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2169 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2170 int netdev_get_name(struct net *net, char *name, int ifindex);
2171 int dev_restart(struct net_device *dev);
2172 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2174 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2176 return NAPI_GRO_CB(skb)->data_offset;
2179 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2181 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2184 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2186 NAPI_GRO_CB(skb)->data_offset += len;
2189 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2190 unsigned int offset)
2192 return NAPI_GRO_CB(skb)->frag0 + offset;
2195 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2197 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2200 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2201 unsigned int offset)
2203 if (!pskb_may_pull(skb, hlen))
2206 NAPI_GRO_CB(skb)->frag0 = NULL;
2207 NAPI_GRO_CB(skb)->frag0_len = 0;
2208 return skb->data + offset;
2211 static inline void *skb_gro_network_header(struct sk_buff *skb)
2213 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2214 skb_network_offset(skb);
2217 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2218 const void *start, unsigned int len)
2220 if (NAPI_GRO_CB(skb)->csum_valid)
2221 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2222 csum_partial(start, len, 0));
2225 /* GRO checksum functions. These are logical equivalents of the normal
2226 * checksum functions (in skbuff.h) except that they operate on the GRO
2227 * offsets and fields in sk_buff.
2230 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2232 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2234 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2235 skb_gro_offset(skb));
2238 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2242 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2243 skb_checksum_start_offset(skb) <
2244 skb_gro_offset(skb)) &&
2245 !skb_at_gro_remcsum_start(skb) &&
2246 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2247 (!zero_okay || check));
2250 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2253 if (NAPI_GRO_CB(skb)->csum_valid &&
2254 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2257 NAPI_GRO_CB(skb)->csum = psum;
2259 return __skb_gro_checksum_complete(skb);
2262 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2264 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2265 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2266 NAPI_GRO_CB(skb)->csum_cnt--;
2268 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2269 * verified a new top level checksum or an encapsulated one
2270 * during GRO. This saves work if we fallback to normal path.
2272 __skb_incr_checksum_unnecessary(skb);
2276 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2279 __sum16 __ret = 0; \
2280 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2281 __ret = __skb_gro_checksum_validate_complete(skb, \
2282 compute_pseudo(skb, proto)); \
2284 __skb_mark_checksum_bad(skb); \
2286 skb_gro_incr_csum_unnecessary(skb); \
2290 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2291 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2293 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2295 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2297 #define skb_gro_checksum_simple_validate(skb) \
2298 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2300 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2302 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2303 !NAPI_GRO_CB(skb)->csum_valid);
2306 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2307 __sum16 check, __wsum pseudo)
2309 NAPI_GRO_CB(skb)->csum = ~pseudo;
2310 NAPI_GRO_CB(skb)->csum_valid = 1;
2313 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2315 if (__skb_gro_checksum_convert_check(skb)) \
2316 __skb_gro_checksum_convert(skb, check, \
2317 compute_pseudo(skb, proto)); \
2320 struct gro_remcsum {
2325 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2331 static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2332 int start, int offset,
2333 struct gro_remcsum *grc,
2338 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2341 NAPI_GRO_CB(skb)->gro_remcsum_start =
2342 ((unsigned char *)ptr + start) - skb->head;
2346 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2348 /* Adjust skb->csum since we changed the packet */
2349 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2351 grc->offset = (ptr + offset) - (void *)skb->head;
2355 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2356 struct gro_remcsum *grc)
2361 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2364 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2365 unsigned short type,
2366 const void *daddr, const void *saddr,
2369 if (!dev->header_ops || !dev->header_ops->create)
2372 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2375 static inline int dev_parse_header(const struct sk_buff *skb,
2376 unsigned char *haddr)
2378 const struct net_device *dev = skb->dev;
2380 if (!dev->header_ops || !dev->header_ops->parse)
2382 return dev->header_ops->parse(skb, haddr);
2385 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2386 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2387 static inline int unregister_gifconf(unsigned int family)
2389 return register_gifconf(family, NULL);
2392 #ifdef CONFIG_NET_FLOW_LIMIT
2393 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2394 struct sd_flow_limit {
2396 unsigned int num_buckets;
2397 unsigned int history_head;
2398 u16 history[FLOW_LIMIT_HISTORY];
2402 extern int netdev_flow_limit_table_len;
2403 #endif /* CONFIG_NET_FLOW_LIMIT */
2406 * Incoming packets are placed on per-cpu queues
2408 struct softnet_data {
2409 struct list_head poll_list;
2410 struct sk_buff_head process_queue;
2413 unsigned int processed;
2414 unsigned int time_squeeze;
2415 unsigned int cpu_collision;
2416 unsigned int received_rps;
2418 struct softnet_data *rps_ipi_list;
2420 #ifdef CONFIG_NET_FLOW_LIMIT
2421 struct sd_flow_limit __rcu *flow_limit;
2423 struct Qdisc *output_queue;
2424 struct Qdisc **output_queue_tailp;
2425 struct sk_buff *completion_queue;
2428 /* Elements below can be accessed between CPUs for RPS */
2429 struct call_single_data csd ____cacheline_aligned_in_smp;
2430 struct softnet_data *rps_ipi_next;
2432 unsigned int input_queue_head;
2433 unsigned int input_queue_tail;
2435 unsigned int dropped;
2436 struct sk_buff_head input_pkt_queue;
2437 struct napi_struct backlog;
2441 static inline void input_queue_head_incr(struct softnet_data *sd)
2444 sd->input_queue_head++;
2448 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2449 unsigned int *qtail)
2452 *qtail = ++sd->input_queue_tail;
2456 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2458 void __netif_schedule(struct Qdisc *q);
2459 void netif_schedule_queue(struct netdev_queue *txq);
2461 static inline void netif_tx_schedule_all(struct net_device *dev)
2465 for (i = 0; i < dev->num_tx_queues; i++)
2466 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2469 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2471 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2475 * netif_start_queue - allow transmit
2476 * @dev: network device
2478 * Allow upper layers to call the device hard_start_xmit routine.
2480 static inline void netif_start_queue(struct net_device *dev)
2482 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2485 static inline void netif_tx_start_all_queues(struct net_device *dev)
2489 for (i = 0; i < dev->num_tx_queues; i++) {
2490 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2491 netif_tx_start_queue(txq);
2495 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2498 * netif_wake_queue - restart transmit
2499 * @dev: network device
2501 * Allow upper layers to call the device hard_start_xmit routine.
2502 * Used for flow control when transmit resources are available.
2504 static inline void netif_wake_queue(struct net_device *dev)
2506 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2509 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2513 for (i = 0; i < dev->num_tx_queues; i++) {
2514 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2515 netif_tx_wake_queue(txq);
2519 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2521 if (WARN_ON(!dev_queue)) {
2522 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2525 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2529 * netif_stop_queue - stop transmitted packets
2530 * @dev: network device
2532 * Stop upper layers calling the device hard_start_xmit routine.
2533 * Used for flow control when transmit resources are unavailable.
2535 static inline void netif_stop_queue(struct net_device *dev)
2537 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2540 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2544 for (i = 0; i < dev->num_tx_queues; i++) {
2545 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2546 netif_tx_stop_queue(txq);
2550 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2552 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2556 * netif_queue_stopped - test if transmit queue is flowblocked
2557 * @dev: network device
2559 * Test if transmit queue on device is currently unable to send.
2561 static inline bool netif_queue_stopped(const struct net_device *dev)
2563 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2566 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2568 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2572 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2574 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2578 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2580 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2584 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2585 * @dev_queue: pointer to transmit queue
2587 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2588 * to give appropriate hint to the cpu.
2590 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2593 prefetchw(&dev_queue->dql.num_queued);
2598 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2599 * @dev_queue: pointer to transmit queue
2601 * BQL enabled drivers might use this helper in their TX completion path,
2602 * to give appropriate hint to the cpu.
2604 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2607 prefetchw(&dev_queue->dql.limit);
2611 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2615 dql_queued(&dev_queue->dql, bytes);
2617 if (likely(dql_avail(&dev_queue->dql) >= 0))
2620 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2623 * The XOFF flag must be set before checking the dql_avail below,
2624 * because in netdev_tx_completed_queue we update the dql_completed
2625 * before checking the XOFF flag.
2629 /* check again in case another CPU has just made room avail */
2630 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2631 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2636 * netdev_sent_queue - report the number of bytes queued to hardware
2637 * @dev: network device
2638 * @bytes: number of bytes queued to the hardware device queue
2640 * Report the number of bytes queued for sending/completion to the network
2641 * device hardware queue. @bytes should be a good approximation and should
2642 * exactly match netdev_completed_queue() @bytes
2644 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2646 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2649 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2650 unsigned int pkts, unsigned int bytes)
2653 if (unlikely(!bytes))
2656 dql_completed(&dev_queue->dql, bytes);
2659 * Without the memory barrier there is a small possiblity that
2660 * netdev_tx_sent_queue will miss the update and cause the queue to
2661 * be stopped forever
2665 if (dql_avail(&dev_queue->dql) < 0)
2668 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2669 netif_schedule_queue(dev_queue);
2674 * netdev_completed_queue - report bytes and packets completed by device
2675 * @dev: network device
2676 * @pkts: actual number of packets sent over the medium
2677 * @bytes: actual number of bytes sent over the medium
2679 * Report the number of bytes and packets transmitted by the network device
2680 * hardware queue over the physical medium, @bytes must exactly match the
2681 * @bytes amount passed to netdev_sent_queue()
2683 static inline void netdev_completed_queue(struct net_device *dev,
2684 unsigned int pkts, unsigned int bytes)
2686 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2689 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2692 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2698 * netdev_reset_queue - reset the packets and bytes count of a network device
2699 * @dev_queue: network device
2701 * Reset the bytes and packet count of a network device and clear the
2702 * software flow control OFF bit for this network device
2704 static inline void netdev_reset_queue(struct net_device *dev_queue)
2706 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2710 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2711 * @dev: network device
2712 * @queue_index: given tx queue index
2714 * Returns 0 if given tx queue index >= number of device tx queues,
2715 * otherwise returns the originally passed tx queue index.
2717 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2719 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2720 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2721 dev->name, queue_index,
2722 dev->real_num_tx_queues);
2730 * netif_running - test if up
2731 * @dev: network device
2733 * Test if the device has been brought up.
2735 static inline bool netif_running(const struct net_device *dev)
2737 return test_bit(__LINK_STATE_START, &dev->state);
2741 * Routines to manage the subqueues on a device. We only need start
2742 * stop, and a check if it's stopped. All other device management is
2743 * done at the overall netdevice level.
2744 * Also test the device if we're multiqueue.
2748 * netif_start_subqueue - allow sending packets on subqueue
2749 * @dev: network device
2750 * @queue_index: sub queue index
2752 * Start individual transmit queue of a device with multiple transmit queues.
2754 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2756 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2758 netif_tx_start_queue(txq);
2762 * netif_stop_subqueue - stop sending packets on subqueue
2763 * @dev: network device
2764 * @queue_index: sub queue index
2766 * Stop individual transmit queue of a device with multiple transmit queues.
2768 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2770 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2771 netif_tx_stop_queue(txq);
2775 * netif_subqueue_stopped - test status of subqueue
2776 * @dev: network device
2777 * @queue_index: sub queue index
2779 * Check individual transmit queue of a device with multiple transmit queues.
2781 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2784 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2786 return netif_tx_queue_stopped(txq);
2789 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2790 struct sk_buff *skb)
2792 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2795 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2798 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2801 static inline int netif_set_xps_queue(struct net_device *dev,
2802 const struct cpumask *mask,
2810 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2811 * as a distribution range limit for the returned value.
2813 static inline u16 skb_tx_hash(const struct net_device *dev,
2814 struct sk_buff *skb)
2816 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2820 * netif_is_multiqueue - test if device has multiple transmit queues
2821 * @dev: network device
2823 * Check if device has multiple transmit queues
2825 static inline bool netif_is_multiqueue(const struct net_device *dev)
2827 return dev->num_tx_queues > 1;
2830 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2833 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2835 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2843 static inline unsigned int get_netdev_rx_queue_index(
2844 struct netdev_rx_queue *queue)
2846 struct net_device *dev = queue->dev;
2847 int index = queue - dev->_rx;
2849 BUG_ON(index >= dev->num_rx_queues);
2854 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2855 int netif_get_num_default_rss_queues(void);
2857 enum skb_free_reason {
2858 SKB_REASON_CONSUMED,
2862 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2863 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2866 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2867 * interrupt context or with hardware interrupts being disabled.
2868 * (in_irq() || irqs_disabled())
2870 * We provide four helpers that can be used in following contexts :
2872 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2873 * replacing kfree_skb(skb)
2875 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2876 * Typically used in place of consume_skb(skb) in TX completion path
2878 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2879 * replacing kfree_skb(skb)
2881 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2882 * and consumed a packet. Used in place of consume_skb(skb)
2884 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2886 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2889 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2891 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2894 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2896 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2899 static inline void dev_consume_skb_any(struct sk_buff *skb)
2901 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2904 int netif_rx(struct sk_buff *skb);
2905 int netif_rx_ni(struct sk_buff *skb);
2906 int netif_receive_skb(struct sk_buff *skb);
2907 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2908 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2909 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2910 gro_result_t napi_gro_frags(struct napi_struct *napi);
2911 struct packet_offload *gro_find_receive_by_type(__be16 type);
2912 struct packet_offload *gro_find_complete_by_type(__be16 type);
2914 static inline void napi_free_frags(struct napi_struct *napi)
2916 kfree_skb(napi->skb);
2920 int netdev_rx_handler_register(struct net_device *dev,
2921 rx_handler_func_t *rx_handler,
2922 void *rx_handler_data);
2923 void netdev_rx_handler_unregister(struct net_device *dev);
2925 bool dev_valid_name(const char *name);
2926 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2927 int dev_ethtool(struct net *net, struct ifreq *);
2928 unsigned int dev_get_flags(const struct net_device *);
2929 int __dev_change_flags(struct net_device *, unsigned int flags);
2930 int dev_change_flags(struct net_device *, unsigned int);
2931 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2932 unsigned int gchanges);
2933 int dev_change_name(struct net_device *, const char *);
2934 int dev_set_alias(struct net_device *, const char *, size_t);
2935 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2936 int dev_set_mtu(struct net_device *, int);
2937 void dev_set_group(struct net_device *, int);
2938 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2939 int dev_change_carrier(struct net_device *, bool new_carrier);
2940 int dev_get_phys_port_id(struct net_device *dev,
2941 struct netdev_phys_item_id *ppid);
2942 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2943 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2944 struct netdev_queue *txq, int *ret);
2945 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2946 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2947 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2949 extern int netdev_budget;
2951 /* Called by rtnetlink.c:rtnl_unlock() */
2952 void netdev_run_todo(void);
2955 * dev_put - release reference to device
2956 * @dev: network device
2958 * Release reference to device to allow it to be freed.
2960 static inline void dev_put(struct net_device *dev)
2962 this_cpu_dec(*dev->pcpu_refcnt);
2966 * dev_hold - get reference to device
2967 * @dev: network device
2969 * Hold reference to device to keep it from being freed.
2971 static inline void dev_hold(struct net_device *dev)
2973 this_cpu_inc(*dev->pcpu_refcnt);
2976 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2977 * and _off may be called from IRQ context, but it is caller
2978 * who is responsible for serialization of these calls.
2980 * The name carrier is inappropriate, these functions should really be
2981 * called netif_lowerlayer_*() because they represent the state of any
2982 * kind of lower layer not just hardware media.
2985 void linkwatch_init_dev(struct net_device *dev);
2986 void linkwatch_fire_event(struct net_device *dev);
2987 void linkwatch_forget_dev(struct net_device *dev);
2990 * netif_carrier_ok - test if carrier present
2991 * @dev: network device
2993 * Check if carrier is present on device
2995 static inline bool netif_carrier_ok(const struct net_device *dev)
2997 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3000 unsigned long dev_trans_start(struct net_device *dev);
3002 void __netdev_watchdog_up(struct net_device *dev);
3004 void netif_carrier_on(struct net_device *dev);
3006 void netif_carrier_off(struct net_device *dev);
3009 * netif_dormant_on - mark device as dormant.
3010 * @dev: network device
3012 * Mark device as dormant (as per RFC2863).
3014 * The dormant state indicates that the relevant interface is not
3015 * actually in a condition to pass packets (i.e., it is not 'up') but is
3016 * in a "pending" state, waiting for some external event. For "on-
3017 * demand" interfaces, this new state identifies the situation where the
3018 * interface is waiting for events to place it in the up state.
3021 static inline void netif_dormant_on(struct net_device *dev)
3023 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3024 linkwatch_fire_event(dev);
3028 * netif_dormant_off - set device as not dormant.
3029 * @dev: network device
3031 * Device is not in dormant state.
3033 static inline void netif_dormant_off(struct net_device *dev)
3035 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3036 linkwatch_fire_event(dev);
3040 * netif_dormant - test if carrier present
3041 * @dev: network device
3043 * Check if carrier is present on device
3045 static inline bool netif_dormant(const struct net_device *dev)
3047 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3052 * netif_oper_up - test if device is operational
3053 * @dev: network device
3055 * Check if carrier is operational
3057 static inline bool netif_oper_up(const struct net_device *dev)
3059 return (dev->operstate == IF_OPER_UP ||
3060 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3064 * netif_device_present - is device available or removed
3065 * @dev: network device
3067 * Check if device has not been removed from system.
3069 static inline bool netif_device_present(struct net_device *dev)
3071 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3074 void netif_device_detach(struct net_device *dev);
3076 void netif_device_attach(struct net_device *dev);
3079 * Network interface message level settings
3083 NETIF_MSG_DRV = 0x0001,
3084 NETIF_MSG_PROBE = 0x0002,
3085 NETIF_MSG_LINK = 0x0004,
3086 NETIF_MSG_TIMER = 0x0008,
3087 NETIF_MSG_IFDOWN = 0x0010,
3088 NETIF_MSG_IFUP = 0x0020,
3089 NETIF_MSG_RX_ERR = 0x0040,
3090 NETIF_MSG_TX_ERR = 0x0080,
3091 NETIF_MSG_TX_QUEUED = 0x0100,
3092 NETIF_MSG_INTR = 0x0200,
3093 NETIF_MSG_TX_DONE = 0x0400,
3094 NETIF_MSG_RX_STATUS = 0x0800,
3095 NETIF_MSG_PKTDATA = 0x1000,
3096 NETIF_MSG_HW = 0x2000,
3097 NETIF_MSG_WOL = 0x4000,
3100 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3101 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3102 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3103 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3104 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3105 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3106 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3107 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3108 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3109 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3110 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3111 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3112 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3113 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3114 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3116 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3119 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3120 return default_msg_enable_bits;
3121 if (debug_value == 0) /* no output */
3123 /* set low N bits */
3124 return (1 << debug_value) - 1;
3127 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3129 spin_lock(&txq->_xmit_lock);
3130 txq->xmit_lock_owner = cpu;
3133 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3135 spin_lock_bh(&txq->_xmit_lock);
3136 txq->xmit_lock_owner = smp_processor_id();
3139 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3141 bool ok = spin_trylock(&txq->_xmit_lock);
3143 txq->xmit_lock_owner = smp_processor_id();
3147 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3149 txq->xmit_lock_owner = -1;
3150 spin_unlock(&txq->_xmit_lock);
3153 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3155 txq->xmit_lock_owner = -1;
3156 spin_unlock_bh(&txq->_xmit_lock);
3159 static inline void txq_trans_update(struct netdev_queue *txq)
3161 if (txq->xmit_lock_owner != -1)
3162 txq->trans_start = jiffies;
3166 * netif_tx_lock - grab network device transmit lock
3167 * @dev: network device
3169 * Get network device transmit lock
3171 static inline void netif_tx_lock(struct net_device *dev)
3176 spin_lock(&dev->tx_global_lock);
3177 cpu = smp_processor_id();
3178 for (i = 0; i < dev->num_tx_queues; i++) {
3179 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3181 /* We are the only thread of execution doing a
3182 * freeze, but we have to grab the _xmit_lock in
3183 * order to synchronize with threads which are in
3184 * the ->hard_start_xmit() handler and already
3185 * checked the frozen bit.
3187 __netif_tx_lock(txq, cpu);
3188 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3189 __netif_tx_unlock(txq);
3193 static inline void netif_tx_lock_bh(struct net_device *dev)
3199 static inline void netif_tx_unlock(struct net_device *dev)
3203 for (i = 0; i < dev->num_tx_queues; i++) {
3204 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3206 /* No need to grab the _xmit_lock here. If the
3207 * queue is not stopped for another reason, we
3210 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3211 netif_schedule_queue(txq);
3213 spin_unlock(&dev->tx_global_lock);
3216 static inline void netif_tx_unlock_bh(struct net_device *dev)
3218 netif_tx_unlock(dev);
3222 #define HARD_TX_LOCK(dev, txq, cpu) { \
3223 if ((dev->features & NETIF_F_LLTX) == 0) { \
3224 __netif_tx_lock(txq, cpu); \
3228 #define HARD_TX_TRYLOCK(dev, txq) \
3229 (((dev->features & NETIF_F_LLTX) == 0) ? \
3230 __netif_tx_trylock(txq) : \
3233 #define HARD_TX_UNLOCK(dev, txq) { \
3234 if ((dev->features & NETIF_F_LLTX) == 0) { \
3235 __netif_tx_unlock(txq); \
3239 static inline void netif_tx_disable(struct net_device *dev)
3245 cpu = smp_processor_id();
3246 for (i = 0; i < dev->num_tx_queues; i++) {
3247 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3249 __netif_tx_lock(txq, cpu);
3250 netif_tx_stop_queue(txq);
3251 __netif_tx_unlock(txq);
3256 static inline void netif_addr_lock(struct net_device *dev)
3258 spin_lock(&dev->addr_list_lock);
3261 static inline void netif_addr_lock_nested(struct net_device *dev)
3263 int subclass = SINGLE_DEPTH_NESTING;
3265 if (dev->netdev_ops->ndo_get_lock_subclass)
3266 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3268 spin_lock_nested(&dev->addr_list_lock, subclass);
3271 static inline void netif_addr_lock_bh(struct net_device *dev)
3273 spin_lock_bh(&dev->addr_list_lock);
3276 static inline void netif_addr_unlock(struct net_device *dev)
3278 spin_unlock(&dev->addr_list_lock);
3281 static inline void netif_addr_unlock_bh(struct net_device *dev)
3283 spin_unlock_bh(&dev->addr_list_lock);
3287 * dev_addrs walker. Should be used only for read access. Call with
3288 * rcu_read_lock held.
3290 #define for_each_dev_addr(dev, ha) \
3291 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3293 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3295 void ether_setup(struct net_device *dev);
3297 /* Support for loadable net-drivers */
3298 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3299 unsigned char name_assign_type,
3300 void (*setup)(struct net_device *),
3301 unsigned int txqs, unsigned int rxqs);
3302 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3303 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3305 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3306 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3309 int register_netdev(struct net_device *dev);
3310 void unregister_netdev(struct net_device *dev);
3312 /* General hardware address lists handling functions */
3313 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3314 struct netdev_hw_addr_list *from_list, int addr_len);
3315 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3316 struct netdev_hw_addr_list *from_list, int addr_len);
3317 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3318 struct net_device *dev,
3319 int (*sync)(struct net_device *, const unsigned char *),
3320 int (*unsync)(struct net_device *,
3321 const unsigned char *));
3322 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3323 struct net_device *dev,
3324 int (*unsync)(struct net_device *,
3325 const unsigned char *));
3326 void __hw_addr_init(struct netdev_hw_addr_list *list);
3328 /* Functions used for device addresses handling */
3329 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3330 unsigned char addr_type);
3331 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3332 unsigned char addr_type);
3333 void dev_addr_flush(struct net_device *dev);
3334 int dev_addr_init(struct net_device *dev);
3336 /* Functions used for unicast addresses handling */
3337 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3338 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3339 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3340 int dev_uc_sync(struct net_device *to, struct net_device *from);
3341 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3342 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3343 void dev_uc_flush(struct net_device *dev);
3344 void dev_uc_init(struct net_device *dev);
3347 * __dev_uc_sync - Synchonize device's unicast list
3348 * @dev: device to sync
3349 * @sync: function to call if address should be added
3350 * @unsync: function to call if address should be removed
3352 * Add newly added addresses to the interface, and release
3353 * addresses that have been deleted.
3355 static inline int __dev_uc_sync(struct net_device *dev,
3356 int (*sync)(struct net_device *,
3357 const unsigned char *),
3358 int (*unsync)(struct net_device *,
3359 const unsigned char *))
3361 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3365 * __dev_uc_unsync - Remove synchronized addresses from device
3366 * @dev: device to sync
3367 * @unsync: function to call if address should be removed
3369 * Remove all addresses that were added to the device by dev_uc_sync().
3371 static inline void __dev_uc_unsync(struct net_device *dev,
3372 int (*unsync)(struct net_device *,
3373 const unsigned char *))
3375 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3378 /* Functions used for multicast addresses handling */
3379 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3380 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3381 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3382 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3383 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3384 int dev_mc_sync(struct net_device *to, struct net_device *from);
3385 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3386 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3387 void dev_mc_flush(struct net_device *dev);
3388 void dev_mc_init(struct net_device *dev);
3391 * __dev_mc_sync - Synchonize device's multicast list
3392 * @dev: device to sync
3393 * @sync: function to call if address should be added
3394 * @unsync: function to call if address should be removed
3396 * Add newly added addresses to the interface, and release
3397 * addresses that have been deleted.
3399 static inline int __dev_mc_sync(struct net_device *dev,
3400 int (*sync)(struct net_device *,
3401 const unsigned char *),
3402 int (*unsync)(struct net_device *,
3403 const unsigned char *))
3405 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3409 * __dev_mc_unsync - Remove synchronized addresses from device
3410 * @dev: device to sync
3411 * @unsync: function to call if address should be removed
3413 * Remove all addresses that were added to the device by dev_mc_sync().
3415 static inline void __dev_mc_unsync(struct net_device *dev,
3416 int (*unsync)(struct net_device *,
3417 const unsigned char *))
3419 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3422 /* Functions used for secondary unicast and multicast support */
3423 void dev_set_rx_mode(struct net_device *dev);
3424 void __dev_set_rx_mode(struct net_device *dev);
3425 int dev_set_promiscuity(struct net_device *dev, int inc);
3426 int dev_set_allmulti(struct net_device *dev, int inc);
3427 void netdev_state_change(struct net_device *dev);
3428 void netdev_notify_peers(struct net_device *dev);
3429 void netdev_features_change(struct net_device *dev);
3430 /* Load a device via the kmod */
3431 void dev_load(struct net *net, const char *name);
3432 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3433 struct rtnl_link_stats64 *storage);
3434 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3435 const struct net_device_stats *netdev_stats);
3437 extern int netdev_max_backlog;
3438 extern int netdev_tstamp_prequeue;
3439 extern int weight_p;
3440 extern int bpf_jit_enable;
3442 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3443 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3444 struct list_head **iter);
3445 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3446 struct list_head **iter);
3448 /* iterate through upper list, must be called under RCU read lock */
3449 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3450 for (iter = &(dev)->adj_list.upper, \
3451 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3453 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3455 /* iterate through upper list, must be called under RCU read lock */
3456 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3457 for (iter = &(dev)->all_adj_list.upper, \
3458 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3460 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3462 void *netdev_lower_get_next_private(struct net_device *dev,
3463 struct list_head **iter);
3464 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3465 struct list_head **iter);
3467 #define netdev_for_each_lower_private(dev, priv, iter) \
3468 for (iter = (dev)->adj_list.lower.next, \
3469 priv = netdev_lower_get_next_private(dev, &(iter)); \
3471 priv = netdev_lower_get_next_private(dev, &(iter)))
3473 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3474 for (iter = &(dev)->adj_list.lower, \
3475 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3477 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3479 void *netdev_lower_get_next(struct net_device *dev,
3480 struct list_head **iter);
3481 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3482 for (iter = &(dev)->adj_list.lower, \
3483 ldev = netdev_lower_get_next(dev, &(iter)); \
3485 ldev = netdev_lower_get_next(dev, &(iter)))
3487 void *netdev_adjacent_get_private(struct list_head *adj_list);
3488 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3489 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3490 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3491 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3492 int netdev_master_upper_dev_link(struct net_device *dev,
3493 struct net_device *upper_dev);
3494 int netdev_master_upper_dev_link_private(struct net_device *dev,
3495 struct net_device *upper_dev,
3497 void netdev_upper_dev_unlink(struct net_device *dev,
3498 struct net_device *upper_dev);
3499 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3500 void *netdev_lower_dev_get_private(struct net_device *dev,
3501 struct net_device *lower_dev);
3503 /* RSS keys are 40 or 52 bytes long */
3504 #define NETDEV_RSS_KEY_LEN 52
3505 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3506 void netdev_rss_key_fill(void *buffer, size_t len);
3508 int dev_get_nest_level(struct net_device *dev,
3509 bool (*type_check)(struct net_device *dev));
3510 int skb_checksum_help(struct sk_buff *skb);
3511 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3512 netdev_features_t features, bool tx_path);
3513 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3514 netdev_features_t features);
3516 struct netdev_bonding_info {
3521 struct netdev_notifier_bonding_info {
3522 struct netdev_notifier_info info; /* must be first */
3523 struct netdev_bonding_info bonding_info;
3526 void netdev_bonding_info_change(struct net_device *dev,
3527 struct netdev_bonding_info *bonding_info);
3530 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3532 return __skb_gso_segment(skb, features, true);
3534 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3536 static inline bool can_checksum_protocol(netdev_features_t features,
3539 return ((features & NETIF_F_GEN_CSUM) ||
3540 ((features & NETIF_F_V4_CSUM) &&
3541 protocol == htons(ETH_P_IP)) ||
3542 ((features & NETIF_F_V6_CSUM) &&
3543 protocol == htons(ETH_P_IPV6)) ||
3544 ((features & NETIF_F_FCOE_CRC) &&
3545 protocol == htons(ETH_P_FCOE)));
3549 void netdev_rx_csum_fault(struct net_device *dev);
3551 static inline void netdev_rx_csum_fault(struct net_device *dev)
3555 /* rx skb timestamps */
3556 void net_enable_timestamp(void);
3557 void net_disable_timestamp(void);
3559 #ifdef CONFIG_PROC_FS
3560 int __init dev_proc_init(void);
3562 #define dev_proc_init() 0
3565 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3566 struct sk_buff *skb, struct net_device *dev,
3569 skb->xmit_more = more ? 1 : 0;
3570 return ops->ndo_start_xmit(skb, dev);
3573 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3574 struct netdev_queue *txq, bool more)
3576 const struct net_device_ops *ops = dev->netdev_ops;
3579 rc = __netdev_start_xmit(ops, skb, dev, more);
3580 if (rc == NETDEV_TX_OK)
3581 txq_trans_update(txq);
3586 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3588 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3591 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3593 return netdev_class_create_file_ns(class_attr, NULL);
3596 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3598 netdev_class_remove_file_ns(class_attr, NULL);
3601 extern struct kobj_ns_type_operations net_ns_type_operations;
3603 const char *netdev_drivername(const struct net_device *dev);
3605 void linkwatch_run_queue(void);
3607 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3608 netdev_features_t f2)
3610 if (f1 & NETIF_F_GEN_CSUM)
3611 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3612 if (f2 & NETIF_F_GEN_CSUM)
3613 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3615 if (f1 & NETIF_F_GEN_CSUM)
3616 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3621 static inline netdev_features_t netdev_get_wanted_features(
3622 struct net_device *dev)
3624 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3626 netdev_features_t netdev_increment_features(netdev_features_t all,
3627 netdev_features_t one, netdev_features_t mask);
3629 /* Allow TSO being used on stacked device :
3630 * Performing the GSO segmentation before last device
3631 * is a performance improvement.
3633 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3634 netdev_features_t mask)
3636 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3639 int __netdev_update_features(struct net_device *dev);
3640 void netdev_update_features(struct net_device *dev);
3641 void netdev_change_features(struct net_device *dev);
3643 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3644 struct net_device *dev);
3646 netdev_features_t netif_skb_features(struct sk_buff *skb);
3648 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3650 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3652 /* check flags correspondence */
3653 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3654 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3655 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3656 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3657 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3658 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3659 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3660 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3661 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3662 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3663 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3664 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3665 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3667 return (features & feature) == feature;
3670 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3672 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3673 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3676 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3677 netdev_features_t features)
3679 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3680 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3681 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3684 static inline void netif_set_gso_max_size(struct net_device *dev,
3687 dev->gso_max_size = size;
3690 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3691 int pulled_hlen, u16 mac_offset,
3694 skb->protocol = protocol;
3695 skb->encapsulation = 1;
3696 skb_push(skb, pulled_hlen);
3697 skb_reset_transport_header(skb);
3698 skb->mac_header = mac_offset;
3699 skb->network_header = skb->mac_header + mac_len;
3700 skb->mac_len = mac_len;
3703 static inline bool netif_is_macvlan(struct net_device *dev)
3705 return dev->priv_flags & IFF_MACVLAN;
3708 static inline bool netif_is_macvlan_port(struct net_device *dev)
3710 return dev->priv_flags & IFF_MACVLAN_PORT;
3713 static inline bool netif_is_ipvlan(struct net_device *dev)
3715 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3718 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3720 return dev->priv_flags & IFF_IPVLAN_MASTER;
3723 static inline bool netif_is_bond_master(struct net_device *dev)
3725 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3728 static inline bool netif_is_bond_slave(struct net_device *dev)
3730 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3733 static inline bool netif_supports_nofcs(struct net_device *dev)
3735 return dev->priv_flags & IFF_SUPP_NOFCS;
3738 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3739 static inline void netif_keep_dst(struct net_device *dev)
3741 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3744 extern struct pernet_operations __net_initdata loopback_net_ops;
3746 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3748 /* netdev_printk helpers, similar to dev_printk */
3750 static inline const char *netdev_name(const struct net_device *dev)
3752 if (!dev->name[0] || strchr(dev->name, '%'))
3753 return "(unnamed net_device)";
3757 static inline const char *netdev_reg_state(const struct net_device *dev)
3759 switch (dev->reg_state) {
3760 case NETREG_UNINITIALIZED: return " (uninitialized)";
3761 case NETREG_REGISTERED: return "";
3762 case NETREG_UNREGISTERING: return " (unregistering)";
3763 case NETREG_UNREGISTERED: return " (unregistered)";
3764 case NETREG_RELEASED: return " (released)";
3765 case NETREG_DUMMY: return " (dummy)";
3768 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3769 return " (unknown)";
3773 void netdev_printk(const char *level, const struct net_device *dev,
3774 const char *format, ...);
3776 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3778 void netdev_alert(const struct net_device *dev, const char *format, ...);
3780 void netdev_crit(const struct net_device *dev, const char *format, ...);
3782 void netdev_err(const struct net_device *dev, const char *format, ...);
3784 void netdev_warn(const struct net_device *dev, const char *format, ...);
3786 void netdev_notice(const struct net_device *dev, const char *format, ...);
3788 void netdev_info(const struct net_device *dev, const char *format, ...);
3790 #define MODULE_ALIAS_NETDEV(device) \
3791 MODULE_ALIAS("netdev-" device)
3793 #if defined(CONFIG_DYNAMIC_DEBUG)
3794 #define netdev_dbg(__dev, format, args...) \
3796 dynamic_netdev_dbg(__dev, format, ##args); \
3798 #elif defined(DEBUG)
3799 #define netdev_dbg(__dev, format, args...) \
3800 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3802 #define netdev_dbg(__dev, format, args...) \
3805 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3809 #if defined(VERBOSE_DEBUG)
3810 #define netdev_vdbg netdev_dbg
3813 #define netdev_vdbg(dev, format, args...) \
3816 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3822 * netdev_WARN() acts like dev_printk(), but with the key difference
3823 * of using a WARN/WARN_ON to get the message out, including the
3824 * file/line information and a backtrace.
3826 #define netdev_WARN(dev, format, args...) \
3827 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3828 netdev_reg_state(dev), ##args)
3830 /* netif printk helpers, similar to netdev_printk */
3832 #define netif_printk(priv, type, level, dev, fmt, args...) \
3834 if (netif_msg_##type(priv)) \
3835 netdev_printk(level, (dev), fmt, ##args); \
3838 #define netif_level(level, priv, type, dev, fmt, args...) \
3840 if (netif_msg_##type(priv)) \
3841 netdev_##level(dev, fmt, ##args); \
3844 #define netif_emerg(priv, type, dev, fmt, args...) \
3845 netif_level(emerg, priv, type, dev, fmt, ##args)
3846 #define netif_alert(priv, type, dev, fmt, args...) \
3847 netif_level(alert, priv, type, dev, fmt, ##args)
3848 #define netif_crit(priv, type, dev, fmt, args...) \
3849 netif_level(crit, priv, type, dev, fmt, ##args)
3850 #define netif_err(priv, type, dev, fmt, args...) \
3851 netif_level(err, priv, type, dev, fmt, ##args)
3852 #define netif_warn(priv, type, dev, fmt, args...) \
3853 netif_level(warn, priv, type, dev, fmt, ##args)
3854 #define netif_notice(priv, type, dev, fmt, args...) \
3855 netif_level(notice, priv, type, dev, fmt, ##args)
3856 #define netif_info(priv, type, dev, fmt, args...) \
3857 netif_level(info, priv, type, dev, fmt, ##args)
3859 #if defined(CONFIG_DYNAMIC_DEBUG)
3860 #define netif_dbg(priv, type, netdev, format, args...) \
3862 if (netif_msg_##type(priv)) \
3863 dynamic_netdev_dbg(netdev, format, ##args); \
3865 #elif defined(DEBUG)
3866 #define netif_dbg(priv, type, dev, format, args...) \
3867 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3869 #define netif_dbg(priv, type, dev, format, args...) \
3872 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3877 #if defined(VERBOSE_DEBUG)
3878 #define netif_vdbg netif_dbg
3880 #define netif_vdbg(priv, type, dev, format, args...) \
3883 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3889 * The list of packet types we will receive (as opposed to discard)
3890 * and the routines to invoke.
3892 * Why 16. Because with 16 the only overlap we get on a hash of the
3893 * low nibble of the protocol value is RARP/SNAP/X.25.
3895 * NOTE: That is no longer true with the addition of VLAN tags. Not
3896 * sure which should go first, but I bet it won't make much
3897 * difference if we are running VLANs. The good news is that
3898 * this protocol won't be in the list unless compiled in, so
3899 * the average user (w/out VLANs) will not be adversely affected.
3915 #define PTYPE_HASH_SIZE (16)
3916 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3918 #endif /* _LINUX_NETDEVICE_H */