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/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
55 #include <linux/hashtable.h>
62 /* 802.15.4 specific */
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
69 void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
72 /* Backlog congestion levels */
73 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74 #define NET_RX_DROP 1 /* packet dropped */
77 * Transmit return codes: transmit return codes originate from three different
80 * - qdisc return codes
81 * - driver transmit return codes
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
90 * others are propagated to higher layers.
93 /* qdisc ->enqueue() return codes. */
94 #define NET_XMIT_SUCCESS 0x00
95 #define NET_XMIT_DROP 0x01 /* skb dropped */
96 #define NET_XMIT_CN 0x02 /* congestion notification */
97 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
99 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
102 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
103 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105 /* Driver transmit return codes */
106 #define NETDEV_TX_MASK 0xf0
109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
113 typedef enum netdev_tx netdev_tx_t;
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 static inline bool dev_xmit_complete(int rc)
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 if (likely(rc < NET_XMIT_MASK))
134 * Compute the worst-case header length according to the protocols
138 #if defined(CONFIG_HYPERV_NET)
139 # define LL_MAX_HEADER 128
140 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
141 # if defined(CONFIG_MAC80211_MESH)
142 # define LL_MAX_HEADER 128
144 # define LL_MAX_HEADER 96
147 # define LL_MAX_HEADER 32
150 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
152 #define MAX_HEADER LL_MAX_HEADER
154 #define MAX_HEADER (LL_MAX_HEADER + 48)
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
162 struct net_device_stats {
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
172 unsigned long collisions;
173 unsigned long rx_length_errors;
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
189 #include <linux/cache.h>
190 #include <linux/skbuff.h>
193 #include <linux/static_key.h>
194 extern struct static_key rps_needed;
201 struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
205 #define NETDEV_HW_ADDR_T_LAN 1
206 #define NETDEV_HW_ADDR_T_SAN 2
207 #define NETDEV_HW_ADDR_T_SLAVE 3
208 #define NETDEV_HW_ADDR_T_UNICAST 4
209 #define NETDEV_HW_ADDR_T_MULTICAST 5
214 struct rcu_head rcu_head;
217 struct netdev_hw_addr_list {
218 struct list_head list;
222 #define netdev_hw_addr_list_count(l) ((l)->count)
223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224 #define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229 #define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234 #define netdev_for_each_mc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
242 /* cached hardware header; allow for machine alignment needs. */
243 #define HH_DATA_MOD 16
244 #define HH_DATA_OFF(__len) \
245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
246 #define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
259 #define LL_RESERVED_SPACE(dev) \
260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
261 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
267 const void *saddr, unsigned int len);
268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
270 void (*cache_update)(struct hh_cache *hh,
271 const struct net_device *dev,
272 const unsigned char *haddr);
273 bool (*validate)(const char *ll_header, unsigned int len);
276 /* These flag bits are private to the generic network queueing
277 * layer; they may not be explicitly referenced by any other
281 enum netdev_state_t {
283 __LINK_STATE_PRESENT,
284 __LINK_STATE_NOCARRIER,
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
291 * This structure holds boot-time configured netdevice settings. They
292 * are then used in the device probing.
294 struct netdev_boot_setup {
298 #define NETDEV_BOOT_SETUP_MAX 8
300 int __init netdev_boot_setup(char *str);
303 * Structure for NAPI scheduling similar to tasklet but with weighting
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
309 * to the per-CPU poll_list, and whoever clears that bit
310 * can remove from the list right before clearing the bit.
312 struct list_head poll_list;
316 unsigned int gro_count;
317 int (*poll)(struct napi_struct *, int);
318 #ifdef CONFIG_NETPOLL
319 spinlock_t poll_lock;
322 struct net_device *dev;
323 struct sk_buff *gro_list;
325 struct hrtimer timer;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_DISABLE, /* Disable pending */
334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
346 typedef enum gro_result gro_result_t;
349 * enum rx_handler_result - Possible return values for rx_handlers.
350 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
352 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
353 * case skb->dev was changed by rx_handler.
354 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
355 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
357 * rx_handlers are functions called from inside __netif_receive_skb(), to do
358 * special processing of the skb, prior to delivery to protocol handlers.
360 * Currently, a net_device can only have a single rx_handler registered. Trying
361 * to register a second rx_handler will return -EBUSY.
363 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
364 * To unregister a rx_handler on a net_device, use
365 * netdev_rx_handler_unregister().
367 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
370 * If the rx_handler consumed the skb in some way, it should return
371 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
372 * the skb to be delivered in some other way.
374 * If the rx_handler changed skb->dev, to divert the skb to another
375 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
376 * new device will be called if it exists.
378 * If the rx_handler decides the skb should be ignored, it should return
379 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
380 * are registered on exact device (ptype->dev == skb->dev).
382 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
383 * delivered, it should return RX_HANDLER_PASS.
385 * A device without a registered rx_handler will behave as if rx_handler
386 * returned RX_HANDLER_PASS.
389 enum rx_handler_result {
395 typedef enum rx_handler_result rx_handler_result_t;
396 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
398 void __napi_schedule(struct napi_struct *n);
401 * When PREEMPT_RT_FULL is defined, all device interrupt handlers
402 * run as threads, and they can also be preempted (without PREEMPT_RT
403 * interrupt threads can not be preempted). Which means that calling
404 * __napi_schedule_irqoff() from an interrupt handler can be preempted
405 * and can corrupt the napi->poll_list.
407 #ifdef CONFIG_PREEMPT_RT_FULL
408 #define __napi_schedule_irqoff(n) __napi_schedule(n)
410 void __napi_schedule_irqoff(struct napi_struct *n);
413 static inline bool napi_disable_pending(struct napi_struct *n)
415 return test_bit(NAPI_STATE_DISABLE, &n->state);
419 * napi_schedule_prep - check if NAPI can be scheduled
422 * Test if NAPI routine is already running, and if not mark
423 * it as running. This is used as a condition variable to
424 * insure only one NAPI poll instance runs. We also make
425 * sure there is no pending NAPI disable.
427 static inline bool napi_schedule_prep(struct napi_struct *n)
429 return !napi_disable_pending(n) &&
430 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
434 * napi_schedule - schedule NAPI poll
437 * Schedule NAPI poll routine to be called if it is not already
440 static inline void napi_schedule(struct napi_struct *n)
442 if (napi_schedule_prep(n))
447 * napi_schedule_irqoff - schedule NAPI poll
450 * Variant of napi_schedule(), assuming hard irqs are masked.
452 static inline void napi_schedule_irqoff(struct napi_struct *n)
454 if (napi_schedule_prep(n))
455 __napi_schedule_irqoff(n);
458 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
459 static inline bool napi_reschedule(struct napi_struct *napi)
461 if (napi_schedule_prep(napi)) {
462 __napi_schedule(napi);
468 void __napi_complete(struct napi_struct *n);
469 void napi_complete_done(struct napi_struct *n, int work_done);
471 * napi_complete - NAPI processing complete
474 * Mark NAPI processing as complete.
475 * Consider using napi_complete_done() instead.
477 static inline void napi_complete(struct napi_struct *n)
479 return napi_complete_done(n, 0);
483 * napi_hash_add - add a NAPI to global hashtable
484 * @napi: NAPI context
486 * Generate a new napi_id and store a @napi under it in napi_hash.
487 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
488 * Note: This is normally automatically done from netif_napi_add(),
489 * so might disappear in a future Linux version.
491 void napi_hash_add(struct napi_struct *napi);
494 * napi_hash_del - remove a NAPI from global table
495 * @napi: NAPI context
497 * Warning: caller must observe RCU grace period
498 * before freeing memory containing @napi, if
499 * this function returns true.
500 * Note: core networking stack automatically calls it
501 * from netif_napi_del().
502 * Drivers might want to call this helper to combine all
503 * the needed RCU grace periods into a single one.
505 bool napi_hash_del(struct napi_struct *napi);
508 * napi_disable - prevent NAPI from scheduling
511 * Stop NAPI from being scheduled on this context.
512 * Waits till any outstanding processing completes.
514 void napi_disable(struct napi_struct *n);
517 * napi_enable - enable NAPI scheduling
520 * Resume NAPI from being scheduled on this context.
521 * Must be paired with napi_disable.
523 static inline void napi_enable(struct napi_struct *n)
525 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
526 smp_mb__before_atomic();
527 clear_bit(NAPI_STATE_SCHED, &n->state);
528 clear_bit(NAPI_STATE_NPSVC, &n->state);
532 * napi_synchronize - wait until NAPI is not running
535 * Wait until NAPI is done being scheduled on this context.
536 * Waits till any outstanding processing completes but
537 * does not disable future activations.
539 static inline void napi_synchronize(const struct napi_struct *n)
541 if (IS_ENABLED(CONFIG_SMP))
542 while (test_bit(NAPI_STATE_SCHED, &n->state))
548 enum netdev_queue_state_t {
549 __QUEUE_STATE_DRV_XOFF,
550 __QUEUE_STATE_STACK_XOFF,
551 __QUEUE_STATE_FROZEN,
554 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
555 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
556 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
558 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
559 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
561 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
565 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
566 * netif_tx_* functions below are used to manipulate this flag. The
567 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
568 * queue independently. The netif_xmit_*stopped functions below are called
569 * to check if the queue has been stopped by the driver or stack (either
570 * of the XOFF bits are set in the state). Drivers should not need to call
571 * netif_xmit*stopped functions, they should only be using netif_tx_*.
574 struct netdev_queue {
578 struct net_device *dev;
579 struct Qdisc __rcu *qdisc;
580 struct Qdisc *qdisc_sleeping;
584 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
587 unsigned long tx_maxrate;
589 * Number of TX timeouts for this queue
590 * (/sys/class/net/DEV/Q/trans_timeout)
592 unsigned long trans_timeout;
596 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
599 * Time (in jiffies) of last Tx
601 unsigned long trans_start;
608 } ____cacheline_aligned_in_smp;
610 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
612 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
619 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
621 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
628 * This structure holds an RPS map which can be of variable length. The
629 * map is an array of CPUs.
636 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
639 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
640 * tail pointer for that CPU's input queue at the time of last enqueue, and
641 * a hardware filter index.
643 struct rps_dev_flow {
646 unsigned int last_qtail;
648 #define RPS_NO_FILTER 0xffff
651 * The rps_dev_flow_table structure contains a table of flow mappings.
653 struct rps_dev_flow_table {
656 struct rps_dev_flow flows[0];
658 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
659 ((_num) * sizeof(struct rps_dev_flow)))
662 * The rps_sock_flow_table contains mappings of flows to the last CPU
663 * on which they were processed by the application (set in recvmsg).
664 * Each entry is a 32bit value. Upper part is the high-order bits
665 * of flow hash, lower part is CPU number.
666 * rps_cpu_mask is used to partition the space, depending on number of
667 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
668 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
669 * meaning we use 32-6=26 bits for the hash.
671 struct rps_sock_flow_table {
674 u32 ents[0] ____cacheline_aligned_in_smp;
676 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
678 #define RPS_NO_CPU 0xffff
680 extern u32 rps_cpu_mask;
681 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
683 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
687 unsigned int index = hash & table->mask;
688 u32 val = hash & ~rps_cpu_mask;
690 /* We only give a hint, preemption can change CPU under us */
691 val |= raw_smp_processor_id();
693 if (table->ents[index] != val)
694 table->ents[index] = val;
698 #ifdef CONFIG_RFS_ACCEL
699 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
702 #endif /* CONFIG_RPS */
704 /* This structure contains an instance of an RX queue. */
705 struct netdev_rx_queue {
707 struct rps_map __rcu *rps_map;
708 struct rps_dev_flow_table __rcu *rps_flow_table;
711 struct net_device *dev;
712 } ____cacheline_aligned_in_smp;
715 * RX queue sysfs structures and functions.
717 struct rx_queue_attribute {
718 struct attribute attr;
719 ssize_t (*show)(struct netdev_rx_queue *queue,
720 struct rx_queue_attribute *attr, char *buf);
721 ssize_t (*store)(struct netdev_rx_queue *queue,
722 struct rx_queue_attribute *attr, const char *buf, size_t len);
727 * This structure holds an XPS map which can be of variable length. The
728 * map is an array of queues.
732 unsigned int alloc_len;
736 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
737 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
738 - sizeof(struct xps_map)) / sizeof(u16))
741 * This structure holds all XPS maps for device. Maps are indexed by CPU.
743 struct xps_dev_maps {
745 struct xps_map __rcu *cpu_map[0];
747 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
748 (nr_cpu_ids * sizeof(struct xps_map *)))
749 #endif /* CONFIG_XPS */
751 #define TC_MAX_QUEUE 16
752 #define TC_BITMASK 15
753 /* HW offloaded queuing disciplines txq count and offset maps */
754 struct netdev_tc_txq {
759 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
761 * This structure is to hold information about the device
762 * configured to run FCoE protocol stack.
764 struct netdev_fcoe_hbainfo {
765 char manufacturer[64];
766 char serial_number[64];
767 char hardware_version[64];
768 char driver_version[64];
769 char optionrom_version[64];
770 char firmware_version[64];
772 char model_description[256];
776 #define MAX_PHYS_ITEM_ID_LEN 32
778 /* This structure holds a unique identifier to identify some
779 * physical item (port for example) used by a netdevice.
781 struct netdev_phys_item_id {
782 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
783 unsigned char id_len;
786 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
787 struct netdev_phys_item_id *b)
789 return a->id_len == b->id_len &&
790 memcmp(a->id, b->id, a->id_len) == 0;
793 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
794 struct sk_buff *skb);
796 /* These structures hold the attributes of qdisc and classifiers
797 * that are being passed to the netdevice through the setup_tc op.
807 struct tc_cls_u32_offload;
809 struct tc_to_netdev {
813 struct tc_cls_u32_offload *cls_u32;
814 struct tc_cls_flower_offload *cls_flower;
815 struct tc_cls_matchall_offload *cls_mall;
816 struct tc_cls_bpf_offload *cls_bpf;
820 /* These structures hold the attributes of xdp state that are being passed
821 * to the netdevice through the xdp op.
823 enum xdp_netdev_command {
824 /* Set or clear a bpf program used in the earliest stages of packet
825 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
826 * is responsible for calling bpf_prog_put on any old progs that are
827 * stored. In case of error, the callee need not release the new prog
828 * reference, but on success it takes ownership and must bpf_prog_put
829 * when it is no longer used.
832 /* Check if a bpf program is set on the device. The callee should
833 * return true if a program is currently attached and running.
839 enum xdp_netdev_command command;
842 struct bpf_prog *prog;
849 * This structure defines the management hooks for network devices.
850 * The following hooks can be defined; unless noted otherwise, they are
851 * optional and can be filled with a null pointer.
853 * int (*ndo_init)(struct net_device *dev);
854 * This function is called once when a network device is registered.
855 * The network device can use this for any late stage initialization
856 * or semantic validation. It can fail with an error code which will
857 * be propagated back to register_netdev.
859 * void (*ndo_uninit)(struct net_device *dev);
860 * This function is called when device is unregistered or when registration
861 * fails. It is not called if init fails.
863 * int (*ndo_open)(struct net_device *dev);
864 * This function is called when a network device transitions to the up
867 * int (*ndo_stop)(struct net_device *dev);
868 * This function is called when a network device transitions to the down
871 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
872 * struct net_device *dev);
873 * Called when a packet needs to be transmitted.
874 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
875 * the queue before that can happen; it's for obsolete devices and weird
876 * corner cases, but the stack really does a non-trivial amount
877 * of useless work if you return NETDEV_TX_BUSY.
878 * Required; cannot be NULL.
880 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
881 * netdev_features_t features);
882 * Adjusts the requested feature flags according to device-specific
883 * constraints, and returns the resulting flags. Must not modify
886 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
887 * void *accel_priv, select_queue_fallback_t fallback);
888 * Called to decide which queue to use when device supports multiple
891 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
892 * This function is called to allow device receiver to make
893 * changes to configuration when multicast or promiscuous is enabled.
895 * void (*ndo_set_rx_mode)(struct net_device *dev);
896 * This function is called device changes address list filtering.
897 * If driver handles unicast address filtering, it should set
898 * IFF_UNICAST_FLT in its priv_flags.
900 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
901 * This function is called when the Media Access Control address
902 * needs to be changed. If this interface is not defined, the
903 * MAC address can not be changed.
905 * int (*ndo_validate_addr)(struct net_device *dev);
906 * Test if Media Access Control address is valid for the device.
908 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
909 * Called when a user requests an ioctl which can't be handled by
910 * the generic interface code. If not defined ioctls return
911 * not supported error code.
913 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
914 * Used to set network devices bus interface parameters. This interface
915 * is retained for legacy reasons; new devices should use the bus
916 * interface (PCI) for low level management.
918 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
919 * Called when a user wants to change the Maximum Transfer Unit
920 * of a device. If not defined, any request to change MTU will
921 * will return an error.
923 * void (*ndo_tx_timeout)(struct net_device *dev);
924 * Callback used when the transmitter has not made any progress
925 * for dev->watchdog ticks.
927 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
928 * struct rtnl_link_stats64 *storage);
929 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
930 * Called when a user wants to get the network device usage
931 * statistics. Drivers must do one of the following:
932 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
933 * rtnl_link_stats64 structure passed by the caller.
934 * 2. Define @ndo_get_stats to update a net_device_stats structure
935 * (which should normally be dev->stats) and return a pointer to
936 * it. The structure may be changed asynchronously only if each
937 * field is written atomically.
938 * 3. Update dev->stats asynchronously and atomically, and define
941 * bool (*ndo_has_offload_stats)(int attr_id)
942 * Return true if this device supports offload stats of this attr_id.
944 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
946 * Get statistics for offload operations by attr_id. Write it into the
949 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
950 * If device supports VLAN filtering this function is called when a
951 * VLAN id is registered.
953 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
954 * If device supports VLAN filtering this function is called when a
955 * VLAN id is unregistered.
957 * void (*ndo_poll_controller)(struct net_device *dev);
959 * SR-IOV management functions.
960 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
961 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
962 * u8 qos, __be16 proto);
963 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
965 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
966 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
967 * int (*ndo_get_vf_config)(struct net_device *dev,
968 * int vf, struct ifla_vf_info *ivf);
969 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
970 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
971 * struct nlattr *port[]);
973 * Enable or disable the VF ability to query its RSS Redirection Table and
974 * Hash Key. This is needed since on some devices VF share this information
975 * with PF and querying it may introduce a theoretical security risk.
976 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
977 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
978 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
979 * Called to setup 'tc' number of traffic classes in the net device. This
980 * is always called from the stack with the rtnl lock held and netif tx
981 * queues stopped. This allows the netdevice to perform queue management
984 * Fiber Channel over Ethernet (FCoE) offload functions.
985 * int (*ndo_fcoe_enable)(struct net_device *dev);
986 * Called when the FCoE protocol stack wants to start using LLD for FCoE
987 * so the underlying device can perform whatever needed configuration or
988 * initialization to support acceleration of FCoE traffic.
990 * int (*ndo_fcoe_disable)(struct net_device *dev);
991 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
992 * so the underlying device can perform whatever needed clean-ups to
993 * stop supporting acceleration of FCoE traffic.
995 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
996 * struct scatterlist *sgl, unsigned int sgc);
997 * Called when the FCoE Initiator wants to initialize an I/O that
998 * is a possible candidate for Direct Data Placement (DDP). The LLD can
999 * perform necessary setup and returns 1 to indicate the device is set up
1000 * successfully to perform DDP on this I/O, otherwise this returns 0.
1002 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1003 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1004 * indicated by the FC exchange id 'xid', so the underlying device can
1005 * clean up and reuse resources for later DDP requests.
1007 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1008 * struct scatterlist *sgl, unsigned int sgc);
1009 * Called when the FCoE Target wants to initialize an I/O that
1010 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1011 * perform necessary setup and returns 1 to indicate the device is set up
1012 * successfully to perform DDP on this I/O, otherwise this returns 0.
1014 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1015 * struct netdev_fcoe_hbainfo *hbainfo);
1016 * Called when the FCoE Protocol stack wants information on the underlying
1017 * device. This information is utilized by the FCoE protocol stack to
1018 * register attributes with Fiber Channel management service as per the
1019 * FC-GS Fabric Device Management Information(FDMI) specification.
1021 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1022 * Called when the underlying device wants to override default World Wide
1023 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1024 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1025 * protocol stack to use.
1028 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1029 * u16 rxq_index, u32 flow_id);
1030 * Set hardware filter for RFS. rxq_index is the target queue index;
1031 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1032 * Return the filter ID on success, or a negative error code.
1034 * Slave management functions (for bridge, bonding, etc).
1035 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1036 * Called to make another netdev an underling.
1038 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1039 * Called to release previously enslaved netdev.
1041 * Feature/offload setting functions.
1042 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1043 * Called to update device configuration to new features. Passed
1044 * feature set might be less than what was returned by ndo_fix_features()).
1045 * Must return >0 or -errno if it changed dev->features itself.
1047 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1048 * struct net_device *dev,
1049 * const unsigned char *addr, u16 vid, u16 flags)
1050 * Adds an FDB entry to dev for addr.
1051 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1052 * struct net_device *dev,
1053 * const unsigned char *addr, u16 vid)
1054 * Deletes the FDB entry from dev coresponding to addr.
1055 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1056 * struct net_device *dev, struct net_device *filter_dev,
1058 * Used to add FDB entries to dump requests. Implementers should add
1059 * entries to skb and update idx with the number of entries.
1061 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1063 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1064 * struct net_device *dev, u32 filter_mask,
1066 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1069 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1070 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1071 * which do not represent real hardware may define this to allow their
1072 * userspace components to manage their virtual carrier state. Devices
1073 * that determine carrier state from physical hardware properties (eg
1074 * network cables) or protocol-dependent mechanisms (eg
1075 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1077 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1078 * struct netdev_phys_item_id *ppid);
1079 * Called to get ID of physical port of this device. If driver does
1080 * not implement this, it is assumed that the hw is not able to have
1081 * multiple net devices on single physical port.
1083 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1084 * struct udp_tunnel_info *ti);
1085 * Called by UDP tunnel to notify a driver about the UDP port and socket
1086 * address family that a UDP tunnel is listnening to. It is called only
1087 * when a new port starts listening. The operation is protected by the
1090 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1091 * struct udp_tunnel_info *ti);
1092 * Called by UDP tunnel to notify the driver about a UDP port and socket
1093 * address family that the UDP tunnel is not listening to anymore. The
1094 * operation is protected by the RTNL.
1096 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1097 * struct net_device *dev)
1098 * Called by upper layer devices to accelerate switching or other
1099 * station functionality into hardware. 'pdev is the lowerdev
1100 * to use for the offload and 'dev' is the net device that will
1101 * back the offload. Returns a pointer to the private structure
1102 * the upper layer will maintain.
1103 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1104 * Called by upper layer device to delete the station created
1105 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1106 * the station and priv is the structure returned by the add
1108 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1109 * struct net_device *dev,
1111 * Callback to use for xmit over the accelerated station. This
1112 * is used in place of ndo_start_xmit on accelerated net
1114 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1115 * struct net_device *dev
1116 * netdev_features_t features);
1117 * Called by core transmit path to determine if device is capable of
1118 * performing offload operations on a given packet. This is to give
1119 * the device an opportunity to implement any restrictions that cannot
1120 * be otherwise expressed by feature flags. The check is called with
1121 * the set of features that the stack has calculated and it returns
1122 * those the driver believes to be appropriate.
1123 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1124 * int queue_index, u32 maxrate);
1125 * Called when a user wants to set a max-rate limitation of specific
1127 * int (*ndo_get_iflink)(const struct net_device *dev);
1128 * Called to get the iflink value of this device.
1129 * void (*ndo_change_proto_down)(struct net_device *dev,
1131 * This function is used to pass protocol port error state information
1132 * to the switch driver. The switch driver can react to the proto_down
1133 * by doing a phys down on the associated switch port.
1134 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1135 * This function is used to get egress tunnel information for given skb.
1136 * This is useful for retrieving outer tunnel header parameters while
1138 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1139 * This function is used to specify the headroom that the skb must
1140 * consider when allocation skb during packet reception. Setting
1141 * appropriate rx headroom value allows avoiding skb head copy on
1142 * forward. Setting a negative value resets the rx headroom to the
1144 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1145 * This function is used to set or query state related to XDP on the
1146 * netdevice. See definition of enum xdp_netdev_command for details.
1149 struct net_device_ops {
1150 int (*ndo_init)(struct net_device *dev);
1151 void (*ndo_uninit)(struct net_device *dev);
1152 int (*ndo_open)(struct net_device *dev);
1153 int (*ndo_stop)(struct net_device *dev);
1154 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1155 struct net_device *dev);
1156 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1157 struct net_device *dev,
1158 netdev_features_t features);
1159 u16 (*ndo_select_queue)(struct net_device *dev,
1160 struct sk_buff *skb,
1162 select_queue_fallback_t fallback);
1163 void (*ndo_change_rx_flags)(struct net_device *dev,
1165 void (*ndo_set_rx_mode)(struct net_device *dev);
1166 int (*ndo_set_mac_address)(struct net_device *dev,
1168 int (*ndo_validate_addr)(struct net_device *dev);
1169 int (*ndo_do_ioctl)(struct net_device *dev,
1170 struct ifreq *ifr, int cmd);
1171 int (*ndo_set_config)(struct net_device *dev,
1173 int (*ndo_change_mtu)(struct net_device *dev,
1175 int (*ndo_neigh_setup)(struct net_device *dev,
1176 struct neigh_parms *);
1177 void (*ndo_tx_timeout) (struct net_device *dev);
1179 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1180 struct rtnl_link_stats64 *storage);
1181 bool (*ndo_has_offload_stats)(int attr_id);
1182 int (*ndo_get_offload_stats)(int attr_id,
1183 const struct net_device *dev,
1185 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1187 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1188 __be16 proto, u16 vid);
1189 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1190 __be16 proto, u16 vid);
1191 #ifdef CONFIG_NET_POLL_CONTROLLER
1192 void (*ndo_poll_controller)(struct net_device *dev);
1193 int (*ndo_netpoll_setup)(struct net_device *dev,
1194 struct netpoll_info *info);
1195 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1197 #ifdef CONFIG_NET_RX_BUSY_POLL
1198 int (*ndo_busy_poll)(struct napi_struct *dev);
1200 int (*ndo_set_vf_mac)(struct net_device *dev,
1201 int queue, u8 *mac);
1202 int (*ndo_set_vf_vlan)(struct net_device *dev,
1203 int queue, u16 vlan,
1204 u8 qos, __be16 proto);
1205 int (*ndo_set_vf_rate)(struct net_device *dev,
1206 int vf, int min_tx_rate,
1208 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1209 int vf, bool setting);
1210 int (*ndo_set_vf_trust)(struct net_device *dev,
1211 int vf, bool setting);
1212 int (*ndo_get_vf_config)(struct net_device *dev,
1214 struct ifla_vf_info *ivf);
1215 int (*ndo_set_vf_link_state)(struct net_device *dev,
1216 int vf, int link_state);
1217 int (*ndo_get_vf_stats)(struct net_device *dev,
1219 struct ifla_vf_stats
1221 int (*ndo_set_vf_port)(struct net_device *dev,
1223 struct nlattr *port[]);
1224 int (*ndo_get_vf_port)(struct net_device *dev,
1225 int vf, struct sk_buff *skb);
1226 int (*ndo_set_vf_guid)(struct net_device *dev,
1229 int (*ndo_set_vf_rss_query_en)(
1230 struct net_device *dev,
1231 int vf, bool setting);
1232 int (*ndo_setup_tc)(struct net_device *dev,
1235 struct tc_to_netdev *tc);
1236 #if IS_ENABLED(CONFIG_FCOE)
1237 int (*ndo_fcoe_enable)(struct net_device *dev);
1238 int (*ndo_fcoe_disable)(struct net_device *dev);
1239 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1241 struct scatterlist *sgl,
1243 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1245 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1247 struct scatterlist *sgl,
1249 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1250 struct netdev_fcoe_hbainfo *hbainfo);
1253 #if IS_ENABLED(CONFIG_LIBFCOE)
1254 #define NETDEV_FCOE_WWNN 0
1255 #define NETDEV_FCOE_WWPN 1
1256 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1257 u64 *wwn, int type);
1260 #ifdef CONFIG_RFS_ACCEL
1261 int (*ndo_rx_flow_steer)(struct net_device *dev,
1262 const struct sk_buff *skb,
1266 int (*ndo_add_slave)(struct net_device *dev,
1267 struct net_device *slave_dev);
1268 int (*ndo_del_slave)(struct net_device *dev,
1269 struct net_device *slave_dev);
1270 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1271 netdev_features_t features);
1272 int (*ndo_set_features)(struct net_device *dev,
1273 netdev_features_t features);
1274 int (*ndo_neigh_construct)(struct net_device *dev,
1275 struct neighbour *n);
1276 void (*ndo_neigh_destroy)(struct net_device *dev,
1277 struct neighbour *n);
1279 int (*ndo_fdb_add)(struct ndmsg *ndm,
1280 struct nlattr *tb[],
1281 struct net_device *dev,
1282 const unsigned char *addr,
1285 int (*ndo_fdb_del)(struct ndmsg *ndm,
1286 struct nlattr *tb[],
1287 struct net_device *dev,
1288 const unsigned char *addr,
1290 int (*ndo_fdb_dump)(struct sk_buff *skb,
1291 struct netlink_callback *cb,
1292 struct net_device *dev,
1293 struct net_device *filter_dev,
1296 int (*ndo_bridge_setlink)(struct net_device *dev,
1297 struct nlmsghdr *nlh,
1299 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1301 struct net_device *dev,
1304 int (*ndo_bridge_dellink)(struct net_device *dev,
1305 struct nlmsghdr *nlh,
1307 int (*ndo_change_carrier)(struct net_device *dev,
1309 int (*ndo_get_phys_port_id)(struct net_device *dev,
1310 struct netdev_phys_item_id *ppid);
1311 int (*ndo_get_phys_port_name)(struct net_device *dev,
1312 char *name, size_t len);
1313 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1314 struct udp_tunnel_info *ti);
1315 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1316 struct udp_tunnel_info *ti);
1317 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1318 struct net_device *dev);
1319 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1322 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1323 struct net_device *dev,
1325 int (*ndo_get_lock_subclass)(struct net_device *dev);
1326 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1329 int (*ndo_get_iflink)(const struct net_device *dev);
1330 int (*ndo_change_proto_down)(struct net_device *dev,
1332 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1333 struct sk_buff *skb);
1334 void (*ndo_set_rx_headroom)(struct net_device *dev,
1335 int needed_headroom);
1336 int (*ndo_xdp)(struct net_device *dev,
1337 struct netdev_xdp *xdp);
1341 * enum net_device_priv_flags - &struct net_device priv_flags
1343 * These are the &struct net_device, they are only set internally
1344 * by drivers and used in the kernel. These flags are invisible to
1345 * userspace; this means that the order of these flags can change
1346 * during any kernel release.
1348 * You should have a pretty good reason to be extending these flags.
1350 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1351 * @IFF_EBRIDGE: Ethernet bridging device
1352 * @IFF_BONDING: bonding master or slave
1353 * @IFF_ISATAP: ISATAP interface (RFC4214)
1354 * @IFF_WAN_HDLC: WAN HDLC device
1355 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1357 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1358 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1359 * @IFF_MACVLAN_PORT: device used as macvlan port
1360 * @IFF_BRIDGE_PORT: device used as bridge port
1361 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1362 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1363 * @IFF_UNICAST_FLT: Supports unicast filtering
1364 * @IFF_TEAM_PORT: device used as team port
1365 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1366 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1367 * change when it's running
1368 * @IFF_MACVLAN: Macvlan device
1369 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1370 * underlying stacked devices
1371 * @IFF_IPVLAN_MASTER: IPvlan master device
1372 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1373 * @IFF_L3MDEV_MASTER: device is an L3 master device
1374 * @IFF_NO_QUEUE: device can run without qdisc attached
1375 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1376 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1377 * @IFF_TEAM: device is a team device
1378 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1379 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1380 * entity (i.e. the master device for bridged veth)
1381 * @IFF_MACSEC: device is a MACsec device
1383 enum netdev_priv_flags {
1384 IFF_802_1Q_VLAN = 1<<0,
1388 IFF_WAN_HDLC = 1<<4,
1389 IFF_XMIT_DST_RELEASE = 1<<5,
1390 IFF_DONT_BRIDGE = 1<<6,
1391 IFF_DISABLE_NETPOLL = 1<<7,
1392 IFF_MACVLAN_PORT = 1<<8,
1393 IFF_BRIDGE_PORT = 1<<9,
1394 IFF_OVS_DATAPATH = 1<<10,
1395 IFF_TX_SKB_SHARING = 1<<11,
1396 IFF_UNICAST_FLT = 1<<12,
1397 IFF_TEAM_PORT = 1<<13,
1398 IFF_SUPP_NOFCS = 1<<14,
1399 IFF_LIVE_ADDR_CHANGE = 1<<15,
1400 IFF_MACVLAN = 1<<16,
1401 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1402 IFF_IPVLAN_MASTER = 1<<18,
1403 IFF_IPVLAN_SLAVE = 1<<19,
1404 IFF_L3MDEV_MASTER = 1<<20,
1405 IFF_NO_QUEUE = 1<<21,
1406 IFF_OPENVSWITCH = 1<<22,
1407 IFF_L3MDEV_SLAVE = 1<<23,
1409 IFF_RXFH_CONFIGURED = 1<<25,
1410 IFF_PHONY_HEADROOM = 1<<26,
1414 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1415 #define IFF_EBRIDGE IFF_EBRIDGE
1416 #define IFF_BONDING IFF_BONDING
1417 #define IFF_ISATAP IFF_ISATAP
1418 #define IFF_WAN_HDLC IFF_WAN_HDLC
1419 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1420 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1421 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1422 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1423 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1424 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1425 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1426 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1427 #define IFF_TEAM_PORT IFF_TEAM_PORT
1428 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1429 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1430 #define IFF_MACVLAN IFF_MACVLAN
1431 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1432 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1433 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1434 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1435 #define IFF_NO_QUEUE IFF_NO_QUEUE
1436 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1437 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1438 #define IFF_TEAM IFF_TEAM
1439 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1440 #define IFF_MACSEC IFF_MACSEC
1443 * struct net_device - The DEVICE structure.
1444 * Actually, this whole structure is a big mistake. It mixes I/O
1445 * data with strictly "high-level" data, and it has to know about
1446 * almost every data structure used in the INET module.
1448 * @name: This is the first field of the "visible" part of this structure
1449 * (i.e. as seen by users in the "Space.c" file). It is the name
1452 * @name_hlist: Device name hash chain, please keep it close to name[]
1453 * @ifalias: SNMP alias
1454 * @mem_end: Shared memory end
1455 * @mem_start: Shared memory start
1456 * @base_addr: Device I/O address
1457 * @irq: Device IRQ number
1459 * @carrier_changes: Stats to monitor carrier on<->off transitions
1461 * @state: Generic network queuing layer state, see netdev_state_t
1462 * @dev_list: The global list of network devices
1463 * @napi_list: List entry used for polling NAPI devices
1464 * @unreg_list: List entry when we are unregistering the
1465 * device; see the function unregister_netdev
1466 * @close_list: List entry used when we are closing the device
1467 * @ptype_all: Device-specific packet handlers for all protocols
1468 * @ptype_specific: Device-specific, protocol-specific packet handlers
1470 * @adj_list: Directly linked devices, like slaves for bonding
1471 * @all_adj_list: All linked devices, *including* neighbours
1472 * @features: Currently active device features
1473 * @hw_features: User-changeable features
1475 * @wanted_features: User-requested features
1476 * @vlan_features: Mask of features inheritable by VLAN devices
1478 * @hw_enc_features: Mask of features inherited by encapsulating devices
1479 * This field indicates what encapsulation
1480 * offloads the hardware is capable of doing,
1481 * and drivers will need to set them appropriately.
1483 * @mpls_features: Mask of features inheritable by MPLS
1485 * @ifindex: interface index
1486 * @group: The group the device belongs to
1488 * @stats: Statistics struct, which was left as a legacy, use
1489 * rtnl_link_stats64 instead
1491 * @rx_dropped: Dropped packets by core network,
1492 * do not use this in drivers
1493 * @tx_dropped: Dropped packets by core network,
1494 * do not use this in drivers
1495 * @rx_nohandler: nohandler dropped packets by core network on
1496 * inactive devices, do not use this in drivers
1498 * @wireless_handlers: List of functions to handle Wireless Extensions,
1500 * see <net/iw_handler.h> for details.
1501 * @wireless_data: Instance data managed by the core of wireless extensions
1503 * @netdev_ops: Includes several pointers to callbacks,
1504 * if one wants to override the ndo_*() functions
1505 * @ethtool_ops: Management operations
1506 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1507 * discovery handling. Necessary for e.g. 6LoWPAN.
1508 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1509 * of Layer 2 headers.
1511 * @flags: Interface flags (a la BSD)
1512 * @priv_flags: Like 'flags' but invisible to userspace,
1513 * see if.h for the definitions
1514 * @gflags: Global flags ( kept as legacy )
1515 * @padded: How much padding added by alloc_netdev()
1516 * @operstate: RFC2863 operstate
1517 * @link_mode: Mapping policy to operstate
1518 * @if_port: Selectable AUI, TP, ...
1520 * @mtu: Interface MTU value
1521 * @type: Interface hardware type
1522 * @hard_header_len: Maximum hardware header length.
1524 * @needed_headroom: Extra headroom the hardware may need, but not in all
1525 * cases can this be guaranteed
1526 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1527 * cases can this be guaranteed. Some cases also use
1528 * LL_MAX_HEADER instead to allocate the skb
1530 * interface address info:
1532 * @perm_addr: Permanent hw address
1533 * @addr_assign_type: Hw address assignment type
1534 * @addr_len: Hardware address length
1535 * @neigh_priv_len: Used in neigh_alloc()
1536 * @dev_id: Used to differentiate devices that share
1537 * the same link layer address
1538 * @dev_port: Used to differentiate devices that share
1540 * @addr_list_lock: XXX: need comments on this one
1541 * @uc_promisc: Counter that indicates promiscuous mode
1542 * has been enabled due to the need to listen to
1543 * additional unicast addresses in a device that
1544 * does not implement ndo_set_rx_mode()
1545 * @uc: unicast mac addresses
1546 * @mc: multicast mac addresses
1547 * @dev_addrs: list of device hw addresses
1548 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1549 * @promiscuity: Number of times the NIC is told to work in
1550 * promiscuous mode; if it becomes 0 the NIC will
1551 * exit promiscuous mode
1552 * @allmulti: Counter, enables or disables allmulticast mode
1554 * @vlan_info: VLAN info
1555 * @dsa_ptr: dsa specific data
1556 * @tipc_ptr: TIPC specific data
1557 * @atalk_ptr: AppleTalk link
1558 * @ip_ptr: IPv4 specific data
1559 * @dn_ptr: DECnet specific data
1560 * @ip6_ptr: IPv6 specific data
1561 * @ax25_ptr: AX.25 specific data
1562 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1564 * @last_rx: Time of last Rx
1565 * @dev_addr: Hw address (before bcast,
1566 * because most packets are unicast)
1568 * @_rx: Array of RX queues
1569 * @num_rx_queues: Number of RX queues
1570 * allocated at register_netdev() time
1571 * @real_num_rx_queues: Number of RX queues currently active in device
1573 * @rx_handler: handler for received packets
1574 * @rx_handler_data: XXX: need comments on this one
1575 * @ingress_queue: XXX: need comments on this one
1576 * @broadcast: hw bcast address
1578 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1579 * indexed by RX queue number. Assigned by driver.
1580 * This must only be set if the ndo_rx_flow_steer
1581 * operation is defined
1582 * @index_hlist: Device index hash chain
1584 * @_tx: Array of TX queues
1585 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1586 * @real_num_tx_queues: Number of TX queues currently active in device
1587 * @qdisc: Root qdisc from userspace point of view
1588 * @tx_queue_len: Max frames per queue allowed
1589 * @tx_global_lock: XXX: need comments on this one
1591 * @xps_maps: XXX: need comments on this one
1593 * @watchdog_timeo: Represents the timeout that is used by
1594 * the watchdog (see dev_watchdog())
1595 * @watchdog_timer: List of timers
1597 * @pcpu_refcnt: Number of references to this device
1598 * @todo_list: Delayed register/unregister
1599 * @link_watch_list: XXX: need comments on this one
1601 * @reg_state: Register/unregister state machine
1602 * @dismantle: Device is going to be freed
1603 * @rtnl_link_state: This enum represents the phases of creating
1606 * @destructor: Called from unregister,
1607 * can be used to call free_netdev
1608 * @npinfo: XXX: need comments on this one
1609 * @nd_net: Network namespace this network device is inside
1611 * @ml_priv: Mid-layer private
1612 * @lstats: Loopback statistics
1613 * @tstats: Tunnel statistics
1614 * @dstats: Dummy statistics
1615 * @vstats: Virtual ethernet statistics
1620 * @dev: Class/net/name entry
1621 * @sysfs_groups: Space for optional device, statistics and wireless
1624 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1625 * @rtnl_link_ops: Rtnl_link_ops
1627 * @gso_max_size: Maximum size of generic segmentation offload
1628 * @gso_max_segs: Maximum number of segments that can be passed to the
1631 * @dcbnl_ops: Data Center Bridging netlink ops
1632 * @num_tc: Number of traffic classes in the net device
1633 * @tc_to_txq: XXX: need comments on this one
1634 * @prio_tc_map: XXX: need comments on this one
1636 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1638 * @priomap: XXX: need comments on this one
1639 * @phydev: Physical device may attach itself
1640 * for hardware timestamping
1642 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1643 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1645 * @proto_down: protocol port state information can be sent to the
1646 * switch driver and used to set the phys state of the
1649 * FIXME: cleanup struct net_device such that network protocol info
1654 char name[IFNAMSIZ];
1655 struct hlist_node name_hlist;
1658 * I/O specific fields
1659 * FIXME: Merge these and struct ifmap into one
1661 unsigned long mem_end;
1662 unsigned long mem_start;
1663 unsigned long base_addr;
1666 atomic_t carrier_changes;
1669 * Some hardware also needs these fields (state,dev_list,
1670 * napi_list,unreg_list,close_list) but they are not
1671 * part of the usual set specified in Space.c.
1674 unsigned long state;
1676 struct list_head dev_list;
1677 struct list_head napi_list;
1678 struct list_head unreg_list;
1679 struct list_head close_list;
1680 struct list_head ptype_all;
1681 struct list_head ptype_specific;
1684 struct list_head upper;
1685 struct list_head lower;
1689 struct list_head upper;
1690 struct list_head lower;
1693 netdev_features_t features;
1694 netdev_features_t hw_features;
1695 netdev_features_t wanted_features;
1696 netdev_features_t vlan_features;
1697 netdev_features_t hw_enc_features;
1698 netdev_features_t mpls_features;
1699 netdev_features_t gso_partial_features;
1704 struct net_device_stats stats;
1706 atomic_long_t rx_dropped;
1707 atomic_long_t tx_dropped;
1708 atomic_long_t rx_nohandler;
1710 #ifdef CONFIG_WIRELESS_EXT
1711 const struct iw_handler_def *wireless_handlers;
1712 struct iw_public_data *wireless_data;
1714 const struct net_device_ops *netdev_ops;
1715 const struct ethtool_ops *ethtool_ops;
1716 #ifdef CONFIG_NET_SWITCHDEV
1717 const struct switchdev_ops *switchdev_ops;
1719 #ifdef CONFIG_NET_L3_MASTER_DEV
1720 const struct l3mdev_ops *l3mdev_ops;
1722 #if IS_ENABLED(CONFIG_IPV6)
1723 const struct ndisc_ops *ndisc_ops;
1726 const struct header_ops *header_ops;
1729 unsigned int priv_flags;
1731 unsigned short gflags;
1732 unsigned short padded;
1734 unsigned char operstate;
1735 unsigned char link_mode;
1737 unsigned char if_port;
1741 unsigned short type;
1742 unsigned short hard_header_len;
1744 unsigned short needed_headroom;
1745 unsigned short needed_tailroom;
1747 /* Interface address info. */
1748 unsigned char perm_addr[MAX_ADDR_LEN];
1749 unsigned char addr_assign_type;
1750 unsigned char addr_len;
1751 unsigned short neigh_priv_len;
1752 unsigned short dev_id;
1753 unsigned short dev_port;
1754 spinlock_t addr_list_lock;
1755 unsigned char name_assign_type;
1757 struct netdev_hw_addr_list uc;
1758 struct netdev_hw_addr_list mc;
1759 struct netdev_hw_addr_list dev_addrs;
1762 struct kset *queues_kset;
1764 unsigned int promiscuity;
1765 unsigned int allmulti;
1768 /* Protocol-specific pointers */
1770 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1771 struct vlan_info __rcu *vlan_info;
1773 #if IS_ENABLED(CONFIG_NET_DSA)
1774 struct dsa_switch_tree *dsa_ptr;
1776 #if IS_ENABLED(CONFIG_TIPC)
1777 struct tipc_bearer __rcu *tipc_ptr;
1780 struct in_device __rcu *ip_ptr;
1781 struct dn_dev __rcu *dn_ptr;
1782 struct inet6_dev __rcu *ip6_ptr;
1784 struct wireless_dev *ieee80211_ptr;
1785 struct wpan_dev *ieee802154_ptr;
1786 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1787 struct mpls_dev __rcu *mpls_ptr;
1791 * Cache lines mostly used on receive path (including eth_type_trans())
1793 unsigned long last_rx;
1795 /* Interface address info used in eth_type_trans() */
1796 unsigned char *dev_addr;
1799 struct netdev_rx_queue *_rx;
1801 unsigned int num_rx_queues;
1802 unsigned int real_num_rx_queues;
1805 unsigned long gro_flush_timeout;
1806 rx_handler_func_t __rcu *rx_handler;
1807 void __rcu *rx_handler_data;
1809 #ifdef CONFIG_NET_CLS_ACT
1810 struct tcf_proto __rcu *ingress_cl_list;
1812 struct netdev_queue __rcu *ingress_queue;
1813 #ifdef CONFIG_NETFILTER_INGRESS
1814 struct nf_hook_entry __rcu *nf_hooks_ingress;
1817 unsigned char broadcast[MAX_ADDR_LEN];
1818 #ifdef CONFIG_RFS_ACCEL
1819 struct cpu_rmap *rx_cpu_rmap;
1821 struct hlist_node index_hlist;
1824 * Cache lines mostly used on transmit path
1826 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1827 unsigned int num_tx_queues;
1828 unsigned int real_num_tx_queues;
1829 struct Qdisc *qdisc;
1830 #ifdef CONFIG_NET_SCHED
1831 DECLARE_HASHTABLE (qdisc_hash, 4);
1833 unsigned long tx_queue_len;
1834 spinlock_t tx_global_lock;
1838 struct xps_dev_maps __rcu *xps_maps;
1840 #ifdef CONFIG_NET_CLS_ACT
1841 struct tcf_proto __rcu *egress_cl_list;
1844 /* These may be needed for future network-power-down code. */
1845 struct timer_list watchdog_timer;
1847 int __percpu *pcpu_refcnt;
1848 struct list_head todo_list;
1850 struct list_head link_watch_list;
1852 enum { NETREG_UNINITIALIZED=0,
1853 NETREG_REGISTERED, /* completed register_netdevice */
1854 NETREG_UNREGISTERING, /* called unregister_netdevice */
1855 NETREG_UNREGISTERED, /* completed unregister todo */
1856 NETREG_RELEASED, /* called free_netdev */
1857 NETREG_DUMMY, /* dummy device for NAPI poll */
1863 RTNL_LINK_INITIALIZED,
1864 RTNL_LINK_INITIALIZING,
1865 } rtnl_link_state:16;
1867 void (*destructor)(struct net_device *dev);
1869 #ifdef CONFIG_NETPOLL
1870 struct netpoll_info __rcu *npinfo;
1873 possible_net_t nd_net;
1875 /* mid-layer private */
1878 struct pcpu_lstats __percpu *lstats;
1879 struct pcpu_sw_netstats __percpu *tstats;
1880 struct pcpu_dstats __percpu *dstats;
1881 struct pcpu_vstats __percpu *vstats;
1884 struct garp_port __rcu *garp_port;
1885 struct mrp_port __rcu *mrp_port;
1888 const struct attribute_group *sysfs_groups[4];
1889 const struct attribute_group *sysfs_rx_queue_group;
1891 const struct rtnl_link_ops *rtnl_link_ops;
1893 /* for setting kernel sock attribute on TCP connection setup */
1894 #define GSO_MAX_SIZE 65536
1895 unsigned int gso_max_size;
1896 #define GSO_MAX_SEGS 65535
1900 const struct dcbnl_rtnl_ops *dcbnl_ops;
1903 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1904 u8 prio_tc_map[TC_BITMASK + 1];
1906 #if IS_ENABLED(CONFIG_FCOE)
1907 unsigned int fcoe_ddp_xid;
1909 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1910 struct netprio_map __rcu *priomap;
1912 struct phy_device *phydev;
1913 struct lock_class_key *qdisc_tx_busylock;
1914 struct lock_class_key *qdisc_running_key;
1917 #define to_net_dev(d) container_of(d, struct net_device, dev)
1919 #define NETDEV_ALIGN 32
1922 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1924 return dev->prio_tc_map[prio & TC_BITMASK];
1928 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1930 if (tc >= dev->num_tc)
1933 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1938 void netdev_reset_tc(struct net_device *dev)
1941 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1942 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1946 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1948 if (tc >= dev->num_tc)
1951 dev->tc_to_txq[tc].count = count;
1952 dev->tc_to_txq[tc].offset = offset;
1957 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1959 if (num_tc > TC_MAX_QUEUE)
1962 dev->num_tc = num_tc;
1967 int netdev_get_num_tc(struct net_device *dev)
1973 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1976 return &dev->_tx[index];
1979 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1980 const struct sk_buff *skb)
1982 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1985 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1986 void (*f)(struct net_device *,
1987 struct netdev_queue *,
1993 for (i = 0; i < dev->num_tx_queues; i++)
1994 f(dev, &dev->_tx[i], arg);
1997 #define netdev_lockdep_set_classes(dev) \
1999 static struct lock_class_key qdisc_tx_busylock_key; \
2000 static struct lock_class_key qdisc_running_key; \
2001 static struct lock_class_key qdisc_xmit_lock_key; \
2002 static struct lock_class_key dev_addr_list_lock_key; \
2005 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2006 (dev)->qdisc_running_key = &qdisc_running_key; \
2007 lockdep_set_class(&(dev)->addr_list_lock, \
2008 &dev_addr_list_lock_key); \
2009 for (i = 0; i < (dev)->num_tx_queues; i++) \
2010 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2011 &qdisc_xmit_lock_key); \
2014 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2015 struct sk_buff *skb,
2018 /* returns the headroom that the master device needs to take in account
2019 * when forwarding to this dev
2021 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2023 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2026 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2028 if (dev->netdev_ops->ndo_set_rx_headroom)
2029 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2032 /* set the device rx headroom to the dev's default */
2033 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2035 netdev_set_rx_headroom(dev, -1);
2039 * Net namespace inlines
2042 struct net *dev_net(const struct net_device *dev)
2044 return read_pnet(&dev->nd_net);
2048 void dev_net_set(struct net_device *dev, struct net *net)
2050 write_pnet(&dev->nd_net, net);
2053 static inline bool netdev_uses_dsa(struct net_device *dev)
2055 #if IS_ENABLED(CONFIG_NET_DSA)
2056 if (dev->dsa_ptr != NULL)
2057 return dsa_uses_tagged_protocol(dev->dsa_ptr);
2063 * netdev_priv - access network device private data
2064 * @dev: network device
2066 * Get network device private data
2068 static inline void *netdev_priv(const struct net_device *dev)
2070 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2073 /* Set the sysfs physical device reference for the network logical device
2074 * if set prior to registration will cause a symlink during initialization.
2076 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2078 /* Set the sysfs device type for the network logical device to allow
2079 * fine-grained identification of different network device types. For
2080 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2082 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2084 /* Default NAPI poll() weight
2085 * Device drivers are strongly advised to not use bigger value
2087 #define NAPI_POLL_WEIGHT 64
2090 * netif_napi_add - initialize a NAPI context
2091 * @dev: network device
2092 * @napi: NAPI context
2093 * @poll: polling function
2094 * @weight: default weight
2096 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2097 * *any* of the other NAPI-related functions.
2099 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2100 int (*poll)(struct napi_struct *, int), int weight);
2103 * netif_tx_napi_add - initialize a NAPI context
2104 * @dev: network device
2105 * @napi: NAPI context
2106 * @poll: polling function
2107 * @weight: default weight
2109 * This variant of netif_napi_add() should be used from drivers using NAPI
2110 * to exclusively poll a TX queue.
2111 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2113 static inline void netif_tx_napi_add(struct net_device *dev,
2114 struct napi_struct *napi,
2115 int (*poll)(struct napi_struct *, int),
2118 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2119 netif_napi_add(dev, napi, poll, weight);
2123 * netif_napi_del - remove a NAPI context
2124 * @napi: NAPI context
2126 * netif_napi_del() removes a NAPI context from the network device NAPI list
2128 void netif_napi_del(struct napi_struct *napi);
2130 struct napi_gro_cb {
2131 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2134 /* Length of frag0. */
2135 unsigned int frag0_len;
2137 /* This indicates where we are processing relative to skb->data. */
2140 /* This is non-zero if the packet cannot be merged with the new skb. */
2143 /* Save the IP ID here and check when we get to the transport layer */
2146 /* Number of segments aggregated. */
2149 /* Start offset for remote checksum offload */
2150 u16 gro_remcsum_start;
2152 /* jiffies when first packet was created/queued */
2155 /* Used in ipv6_gro_receive() and foo-over-udp */
2158 /* This is non-zero if the packet may be of the same flow. */
2161 /* Used in tunnel GRO receive */
2164 /* GRO checksum is valid */
2167 /* Number of checksums via CHECKSUM_UNNECESSARY */
2172 #define NAPI_GRO_FREE 1
2173 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2175 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2178 /* Used in GRE, set in fou/gue_gro_receive */
2181 /* Used to determine if flush_id can be ignored */
2184 /* Number of gro_receive callbacks this packet already went through */
2185 u8 recursion_counter:4;
2189 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2192 /* used in skb_gro_receive() slow path */
2193 struct sk_buff *last;
2196 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2198 #define GRO_RECURSION_LIMIT 15
2199 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2201 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2204 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2205 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2206 struct sk_buff **head,
2207 struct sk_buff *skb)
2209 if (unlikely(gro_recursion_inc_test(skb))) {
2210 NAPI_GRO_CB(skb)->flush |= 1;
2214 return cb(head, skb);
2217 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2219 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2221 struct sk_buff **head,
2222 struct sk_buff *skb)
2224 if (unlikely(gro_recursion_inc_test(skb))) {
2225 NAPI_GRO_CB(skb)->flush |= 1;
2229 return cb(sk, head, skb);
2232 struct packet_type {
2233 __be16 type; /* This is really htons(ether_type). */
2234 struct net_device *dev; /* NULL is wildcarded here */
2235 int (*func) (struct sk_buff *,
2236 struct net_device *,
2237 struct packet_type *,
2238 struct net_device *);
2239 bool (*id_match)(struct packet_type *ptype,
2241 void *af_packet_priv;
2242 struct list_head list;
2245 struct offload_callbacks {
2246 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2247 netdev_features_t features);
2248 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2249 struct sk_buff *skb);
2250 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2253 struct packet_offload {
2254 __be16 type; /* This is really htons(ether_type). */
2256 struct offload_callbacks callbacks;
2257 struct list_head list;
2260 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2261 struct pcpu_sw_netstats {
2266 struct u64_stats_sync syncp;
2269 #define __netdev_alloc_pcpu_stats(type, gfp) \
2271 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2274 for_each_possible_cpu(__cpu) { \
2275 typeof(type) *stat; \
2276 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2277 u64_stats_init(&stat->syncp); \
2283 #define netdev_alloc_pcpu_stats(type) \
2284 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2286 enum netdev_lag_tx_type {
2287 NETDEV_LAG_TX_TYPE_UNKNOWN,
2288 NETDEV_LAG_TX_TYPE_RANDOM,
2289 NETDEV_LAG_TX_TYPE_BROADCAST,
2290 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2291 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2292 NETDEV_LAG_TX_TYPE_HASH,
2295 struct netdev_lag_upper_info {
2296 enum netdev_lag_tx_type tx_type;
2299 struct netdev_lag_lower_state_info {
2304 #include <linux/notifier.h>
2306 /* netdevice notifier chain. Please remember to update the rtnetlink
2307 * notification exclusion list in rtnetlink_event() when adding new
2310 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2311 #define NETDEV_DOWN 0x0002
2312 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2313 detected a hardware crash and restarted
2314 - we can use this eg to kick tcp sessions
2316 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2317 #define NETDEV_REGISTER 0x0005
2318 #define NETDEV_UNREGISTER 0x0006
2319 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2320 #define NETDEV_CHANGEADDR 0x0008
2321 #define NETDEV_GOING_DOWN 0x0009
2322 #define NETDEV_CHANGENAME 0x000A
2323 #define NETDEV_FEAT_CHANGE 0x000B
2324 #define NETDEV_BONDING_FAILOVER 0x000C
2325 #define NETDEV_PRE_UP 0x000D
2326 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2327 #define NETDEV_POST_TYPE_CHANGE 0x000F
2328 #define NETDEV_POST_INIT 0x0010
2329 #define NETDEV_UNREGISTER_FINAL 0x0011
2330 #define NETDEV_RELEASE 0x0012
2331 #define NETDEV_NOTIFY_PEERS 0x0013
2332 #define NETDEV_JOIN 0x0014
2333 #define NETDEV_CHANGEUPPER 0x0015
2334 #define NETDEV_RESEND_IGMP 0x0016
2335 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2336 #define NETDEV_CHANGEINFODATA 0x0018
2337 #define NETDEV_BONDING_INFO 0x0019
2338 #define NETDEV_PRECHANGEUPPER 0x001A
2339 #define NETDEV_CHANGELOWERSTATE 0x001B
2340 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2341 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2343 int register_netdevice_notifier(struct notifier_block *nb);
2344 int unregister_netdevice_notifier(struct notifier_block *nb);
2346 struct netdev_notifier_info {
2347 struct net_device *dev;
2350 struct netdev_notifier_change_info {
2351 struct netdev_notifier_info info; /* must be first */
2352 unsigned int flags_changed;
2355 struct netdev_notifier_changeupper_info {
2356 struct netdev_notifier_info info; /* must be first */
2357 struct net_device *upper_dev; /* new upper dev */
2358 bool master; /* is upper dev master */
2359 bool linking; /* is the notification for link or unlink */
2360 void *upper_info; /* upper dev info */
2363 struct netdev_notifier_changelowerstate_info {
2364 struct netdev_notifier_info info; /* must be first */
2365 void *lower_state_info; /* is lower dev state */
2368 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2369 struct net_device *dev)
2374 static inline struct net_device *
2375 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2380 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2383 extern rwlock_t dev_base_lock; /* Device list lock */
2385 #define for_each_netdev(net, d) \
2386 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2387 #define for_each_netdev_reverse(net, d) \
2388 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2389 #define for_each_netdev_rcu(net, d) \
2390 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2391 #define for_each_netdev_safe(net, d, n) \
2392 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2393 #define for_each_netdev_continue(net, d) \
2394 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2395 #define for_each_netdev_continue_rcu(net, d) \
2396 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2397 #define for_each_netdev_in_bond_rcu(bond, slave) \
2398 for_each_netdev_rcu(&init_net, slave) \
2399 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2400 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2402 static inline struct net_device *next_net_device(struct net_device *dev)
2404 struct list_head *lh;
2408 lh = dev->dev_list.next;
2409 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2412 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2414 struct list_head *lh;
2418 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2419 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2422 static inline struct net_device *first_net_device(struct net *net)
2424 return list_empty(&net->dev_base_head) ? NULL :
2425 net_device_entry(net->dev_base_head.next);
2428 static inline struct net_device *first_net_device_rcu(struct net *net)
2430 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2432 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2435 int netdev_boot_setup_check(struct net_device *dev);
2436 unsigned long netdev_boot_base(const char *prefix, int unit);
2437 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2438 const char *hwaddr);
2439 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2440 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2441 void dev_add_pack(struct packet_type *pt);
2442 void dev_remove_pack(struct packet_type *pt);
2443 void __dev_remove_pack(struct packet_type *pt);
2444 void dev_add_offload(struct packet_offload *po);
2445 void dev_remove_offload(struct packet_offload *po);
2447 int dev_get_iflink(const struct net_device *dev);
2448 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2449 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2450 unsigned short mask);
2451 struct net_device *dev_get_by_name(struct net *net, const char *name);
2452 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2453 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2454 int dev_alloc_name(struct net_device *dev, const char *name);
2455 int dev_open(struct net_device *dev);
2456 int dev_close(struct net_device *dev);
2457 int dev_close_many(struct list_head *head, bool unlink);
2458 void dev_disable_lro(struct net_device *dev);
2459 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2460 int dev_queue_xmit(struct sk_buff *skb);
2461 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2462 int register_netdevice(struct net_device *dev);
2463 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2464 void unregister_netdevice_many(struct list_head *head);
2465 static inline void unregister_netdevice(struct net_device *dev)
2467 unregister_netdevice_queue(dev, NULL);
2470 int netdev_refcnt_read(const struct net_device *dev);
2471 void free_netdev(struct net_device *dev);
2472 void netdev_freemem(struct net_device *dev);
2473 void synchronize_net(void);
2474 int init_dummy_netdev(struct net_device *dev);
2476 #define XMIT_RECURSION_LIMIT 10
2477 #ifdef CONFIG_PREEMPT_RT_FULL
2478 static inline int dev_recursion_level(void)
2480 return current->xmit_recursion;
2483 static inline int xmit_rec_read(void)
2485 return current->xmit_recursion;
2488 static inline void xmit_rec_inc(void)
2490 current->xmit_recursion++;
2493 static inline void xmit_rec_dec(void)
2495 current->xmit_recursion--;
2500 DECLARE_PER_CPU(int, xmit_recursion);
2502 static inline int dev_recursion_level(void)
2504 return this_cpu_read(xmit_recursion);
2507 static inline int xmit_rec_read(void)
2509 return __this_cpu_read(xmit_recursion);
2512 static inline void xmit_rec_inc(void)
2514 __this_cpu_inc(xmit_recursion);
2517 static inline void xmit_rec_dec(void)
2519 __this_cpu_dec(xmit_recursion);
2523 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2524 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2525 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2526 int netdev_get_name(struct net *net, char *name, int ifindex);
2527 int dev_restart(struct net_device *dev);
2528 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2530 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2532 return NAPI_GRO_CB(skb)->data_offset;
2535 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2537 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2540 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2542 NAPI_GRO_CB(skb)->data_offset += len;
2545 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2546 unsigned int offset)
2548 return NAPI_GRO_CB(skb)->frag0 + offset;
2551 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2553 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2556 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2557 unsigned int offset)
2559 if (!pskb_may_pull(skb, hlen))
2562 NAPI_GRO_CB(skb)->frag0 = NULL;
2563 NAPI_GRO_CB(skb)->frag0_len = 0;
2564 return skb->data + offset;
2567 static inline void *skb_gro_network_header(struct sk_buff *skb)
2569 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2570 skb_network_offset(skb);
2573 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2574 const void *start, unsigned int len)
2576 if (NAPI_GRO_CB(skb)->csum_valid)
2577 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2578 csum_partial(start, len, 0));
2581 /* GRO checksum functions. These are logical equivalents of the normal
2582 * checksum functions (in skbuff.h) except that they operate on the GRO
2583 * offsets and fields in sk_buff.
2586 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2588 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2590 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2593 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2597 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2598 skb_checksum_start_offset(skb) <
2599 skb_gro_offset(skb)) &&
2600 !skb_at_gro_remcsum_start(skb) &&
2601 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2602 (!zero_okay || check));
2605 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2608 if (NAPI_GRO_CB(skb)->csum_valid &&
2609 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2612 NAPI_GRO_CB(skb)->csum = psum;
2614 return __skb_gro_checksum_complete(skb);
2617 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2619 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2620 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2621 NAPI_GRO_CB(skb)->csum_cnt--;
2623 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2624 * verified a new top level checksum or an encapsulated one
2625 * during GRO. This saves work if we fallback to normal path.
2627 __skb_incr_checksum_unnecessary(skb);
2631 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2634 __sum16 __ret = 0; \
2635 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2636 __ret = __skb_gro_checksum_validate_complete(skb, \
2637 compute_pseudo(skb, proto)); \
2639 __skb_mark_checksum_bad(skb); \
2641 skb_gro_incr_csum_unnecessary(skb); \
2645 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2646 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2648 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2650 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2652 #define skb_gro_checksum_simple_validate(skb) \
2653 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2655 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2657 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2658 !NAPI_GRO_CB(skb)->csum_valid);
2661 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2662 __sum16 check, __wsum pseudo)
2664 NAPI_GRO_CB(skb)->csum = ~pseudo;
2665 NAPI_GRO_CB(skb)->csum_valid = 1;
2668 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2670 if (__skb_gro_checksum_convert_check(skb)) \
2671 __skb_gro_checksum_convert(skb, check, \
2672 compute_pseudo(skb, proto)); \
2675 struct gro_remcsum {
2680 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2686 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2687 unsigned int off, size_t hdrlen,
2688 int start, int offset,
2689 struct gro_remcsum *grc,
2693 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2695 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2698 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2702 ptr = skb_gro_header_fast(skb, off);
2703 if (skb_gro_header_hard(skb, off + plen)) {
2704 ptr = skb_gro_header_slow(skb, off + plen, off);
2709 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2712 /* Adjust skb->csum since we changed the packet */
2713 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2715 grc->offset = off + hdrlen + offset;
2721 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2722 struct gro_remcsum *grc)
2725 size_t plen = grc->offset + sizeof(u16);
2730 ptr = skb_gro_header_fast(skb, grc->offset);
2731 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2732 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2737 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2740 struct skb_csum_offl_spec {
2754 bool __skb_csum_offload_chk(struct sk_buff *skb,
2755 const struct skb_csum_offl_spec *spec,
2756 bool *csum_encapped,
2759 static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2760 const struct skb_csum_offl_spec *spec,
2761 bool *csum_encapped,
2764 if (skb->ip_summed != CHECKSUM_PARTIAL)
2767 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2770 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2771 const struct skb_csum_offl_spec *spec)
2775 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2778 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2780 static const struct skb_csum_offl_spec csum_offl_spec = {
2782 .ip_options_okay = 1,
2789 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2792 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2794 static const struct skb_csum_offl_spec csum_offl_spec = {
2796 .ip_options_okay = 1,
2802 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2805 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2806 unsigned short type,
2807 const void *daddr, const void *saddr,
2810 if (!dev->header_ops || !dev->header_ops->create)
2813 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2816 static inline int dev_parse_header(const struct sk_buff *skb,
2817 unsigned char *haddr)
2819 const struct net_device *dev = skb->dev;
2821 if (!dev->header_ops || !dev->header_ops->parse)
2823 return dev->header_ops->parse(skb, haddr);
2826 /* ll_header must have at least hard_header_len allocated */
2827 static inline bool dev_validate_header(const struct net_device *dev,
2828 char *ll_header, int len)
2830 if (likely(len >= dev->hard_header_len))
2833 if (capable(CAP_SYS_RAWIO)) {
2834 memset(ll_header + len, 0, dev->hard_header_len - len);
2838 if (dev->header_ops && dev->header_ops->validate)
2839 return dev->header_ops->validate(ll_header, len);
2844 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2845 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2846 static inline int unregister_gifconf(unsigned int family)
2848 return register_gifconf(family, NULL);
2851 #ifdef CONFIG_NET_FLOW_LIMIT
2852 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2853 struct sd_flow_limit {
2855 unsigned int num_buckets;
2856 unsigned int history_head;
2857 u16 history[FLOW_LIMIT_HISTORY];
2861 extern int netdev_flow_limit_table_len;
2862 #endif /* CONFIG_NET_FLOW_LIMIT */
2865 * Incoming packets are placed on per-CPU queues
2867 struct softnet_data {
2868 struct list_head poll_list;
2869 struct sk_buff_head process_queue;
2872 unsigned int processed;
2873 unsigned int time_squeeze;
2874 unsigned int received_rps;
2876 struct softnet_data *rps_ipi_list;
2878 #ifdef CONFIG_NET_FLOW_LIMIT
2879 struct sd_flow_limit __rcu *flow_limit;
2881 struct Qdisc *output_queue;
2882 struct Qdisc **output_queue_tailp;
2883 struct sk_buff *completion_queue;
2886 /* input_queue_head should be written by cpu owning this struct,
2887 * and only read by other cpus. Worth using a cache line.
2889 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2891 /* Elements below can be accessed between CPUs for RPS/RFS */
2892 struct call_single_data csd ____cacheline_aligned_in_smp;
2893 struct softnet_data *rps_ipi_next;
2895 unsigned int input_queue_tail;
2897 unsigned int dropped;
2898 struct sk_buff_head input_pkt_queue;
2899 struct napi_struct backlog;
2900 struct sk_buff_head tofree_queue;
2904 static inline void input_queue_head_incr(struct softnet_data *sd)
2907 sd->input_queue_head++;
2911 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2912 unsigned int *qtail)
2915 *qtail = ++sd->input_queue_tail;
2919 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2921 void __netif_schedule(struct Qdisc *q);
2922 void netif_schedule_queue(struct netdev_queue *txq);
2924 static inline void netif_tx_schedule_all(struct net_device *dev)
2928 for (i = 0; i < dev->num_tx_queues; i++)
2929 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2932 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2934 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2938 * netif_start_queue - allow transmit
2939 * @dev: network device
2941 * Allow upper layers to call the device hard_start_xmit routine.
2943 static inline void netif_start_queue(struct net_device *dev)
2945 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2948 static inline void netif_tx_start_all_queues(struct net_device *dev)
2952 for (i = 0; i < dev->num_tx_queues; i++) {
2953 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2954 netif_tx_start_queue(txq);
2958 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2961 * netif_wake_queue - restart transmit
2962 * @dev: network device
2964 * Allow upper layers to call the device hard_start_xmit routine.
2965 * Used for flow control when transmit resources are available.
2967 static inline void netif_wake_queue(struct net_device *dev)
2969 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2972 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2976 for (i = 0; i < dev->num_tx_queues; i++) {
2977 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2978 netif_tx_wake_queue(txq);
2982 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2984 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2988 * netif_stop_queue - stop transmitted packets
2989 * @dev: network device
2991 * Stop upper layers calling the device hard_start_xmit routine.
2992 * Used for flow control when transmit resources are unavailable.
2994 static inline void netif_stop_queue(struct net_device *dev)
2996 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2999 void netif_tx_stop_all_queues(struct net_device *dev);
3001 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3003 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3007 * netif_queue_stopped - test if transmit queue is flowblocked
3008 * @dev: network device
3010 * Test if transmit queue on device is currently unable to send.
3012 static inline bool netif_queue_stopped(const struct net_device *dev)
3014 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3017 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3019 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3023 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3025 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3029 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3031 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3035 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3036 * @dev_queue: pointer to transmit queue
3038 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3039 * to give appropriate hint to the CPU.
3041 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3044 prefetchw(&dev_queue->dql.num_queued);
3049 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3050 * @dev_queue: pointer to transmit queue
3052 * BQL enabled drivers might use this helper in their TX completion path,
3053 * to give appropriate hint to the CPU.
3055 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3058 prefetchw(&dev_queue->dql.limit);
3062 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3066 dql_queued(&dev_queue->dql, bytes);
3068 if (likely(dql_avail(&dev_queue->dql) >= 0))
3071 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3074 * The XOFF flag must be set before checking the dql_avail below,
3075 * because in netdev_tx_completed_queue we update the dql_completed
3076 * before checking the XOFF flag.
3080 /* check again in case another CPU has just made room avail */
3081 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3082 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3087 * netdev_sent_queue - report the number of bytes queued to hardware
3088 * @dev: network device
3089 * @bytes: number of bytes queued to the hardware device queue
3091 * Report the number of bytes queued for sending/completion to the network
3092 * device hardware queue. @bytes should be a good approximation and should
3093 * exactly match netdev_completed_queue() @bytes
3095 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3097 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3100 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3101 unsigned int pkts, unsigned int bytes)
3104 if (unlikely(!bytes))
3107 dql_completed(&dev_queue->dql, bytes);
3110 * Without the memory barrier there is a small possiblity that
3111 * netdev_tx_sent_queue will miss the update and cause the queue to
3112 * be stopped forever
3116 if (dql_avail(&dev_queue->dql) < 0)
3119 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3120 netif_schedule_queue(dev_queue);
3125 * netdev_completed_queue - report bytes and packets completed by device
3126 * @dev: network device
3127 * @pkts: actual number of packets sent over the medium
3128 * @bytes: actual number of bytes sent over the medium
3130 * Report the number of bytes and packets transmitted by the network device
3131 * hardware queue over the physical medium, @bytes must exactly match the
3132 * @bytes amount passed to netdev_sent_queue()
3134 static inline void netdev_completed_queue(struct net_device *dev,
3135 unsigned int pkts, unsigned int bytes)
3137 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3140 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3143 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3149 * netdev_reset_queue - reset the packets and bytes count of a network device
3150 * @dev_queue: network device
3152 * Reset the bytes and packet count of a network device and clear the
3153 * software flow control OFF bit for this network device
3155 static inline void netdev_reset_queue(struct net_device *dev_queue)
3157 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3161 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3162 * @dev: network device
3163 * @queue_index: given tx queue index
3165 * Returns 0 if given tx queue index >= number of device tx queues,
3166 * otherwise returns the originally passed tx queue index.
3168 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3170 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3171 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3172 dev->name, queue_index,
3173 dev->real_num_tx_queues);
3181 * netif_running - test if up
3182 * @dev: network device
3184 * Test if the device has been brought up.
3186 static inline bool netif_running(const struct net_device *dev)
3188 return test_bit(__LINK_STATE_START, &dev->state);
3192 * Routines to manage the subqueues on a device. We only need start,
3193 * stop, and a check if it's stopped. All other device management is
3194 * done at the overall netdevice level.
3195 * Also test the device if we're multiqueue.
3199 * netif_start_subqueue - allow sending packets on subqueue
3200 * @dev: network device
3201 * @queue_index: sub queue index
3203 * Start individual transmit queue of a device with multiple transmit queues.
3205 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3207 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3209 netif_tx_start_queue(txq);
3213 * netif_stop_subqueue - stop sending packets on subqueue
3214 * @dev: network device
3215 * @queue_index: sub queue index
3217 * Stop individual transmit queue of a device with multiple transmit queues.
3219 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3221 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3222 netif_tx_stop_queue(txq);
3226 * netif_subqueue_stopped - test status of subqueue
3227 * @dev: network device
3228 * @queue_index: sub queue index
3230 * Check individual transmit queue of a device with multiple transmit queues.
3232 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3235 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3237 return netif_tx_queue_stopped(txq);
3240 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3241 struct sk_buff *skb)
3243 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3246 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3249 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3252 static inline int netif_set_xps_queue(struct net_device *dev,
3253 const struct cpumask *mask,
3260 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3261 unsigned int num_tx_queues);
3264 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3265 * as a distribution range limit for the returned value.
3267 static inline u16 skb_tx_hash(const struct net_device *dev,
3268 struct sk_buff *skb)
3270 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3274 * netif_is_multiqueue - test if device has multiple transmit queues
3275 * @dev: network device
3277 * Check if device has multiple transmit queues
3279 static inline bool netif_is_multiqueue(const struct net_device *dev)
3281 return dev->num_tx_queues > 1;
3284 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3287 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3289 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3297 static inline unsigned int get_netdev_rx_queue_index(
3298 struct netdev_rx_queue *queue)
3300 struct net_device *dev = queue->dev;
3301 int index = queue - dev->_rx;
3303 BUG_ON(index >= dev->num_rx_queues);
3308 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3309 int netif_get_num_default_rss_queues(void);
3311 enum skb_free_reason {
3312 SKB_REASON_CONSUMED,
3316 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3317 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3320 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3321 * interrupt context or with hardware interrupts being disabled.
3322 * (in_irq() || irqs_disabled())
3324 * We provide four helpers that can be used in following contexts :
3326 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3327 * replacing kfree_skb(skb)
3329 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3330 * Typically used in place of consume_skb(skb) in TX completion path
3332 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3333 * replacing kfree_skb(skb)
3335 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3336 * and consumed a packet. Used in place of consume_skb(skb)
3338 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3340 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3343 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3345 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3348 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3350 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3353 static inline void dev_consume_skb_any(struct sk_buff *skb)
3355 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3358 int netif_rx(struct sk_buff *skb);
3359 int netif_rx_ni(struct sk_buff *skb);
3360 int netif_receive_skb(struct sk_buff *skb);
3361 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3362 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3363 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3364 gro_result_t napi_gro_frags(struct napi_struct *napi);
3365 struct packet_offload *gro_find_receive_by_type(__be16 type);
3366 struct packet_offload *gro_find_complete_by_type(__be16 type);
3368 static inline void napi_free_frags(struct napi_struct *napi)
3370 kfree_skb(napi->skb);
3374 bool netdev_is_rx_handler_busy(struct net_device *dev);
3375 int netdev_rx_handler_register(struct net_device *dev,
3376 rx_handler_func_t *rx_handler,
3377 void *rx_handler_data);
3378 void netdev_rx_handler_unregister(struct net_device *dev);
3380 bool dev_valid_name(const char *name);
3381 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3382 int dev_ethtool(struct net *net, struct ifreq *);
3383 unsigned int dev_get_flags(const struct net_device *);
3384 int __dev_change_flags(struct net_device *, unsigned int flags);
3385 int dev_change_flags(struct net_device *, unsigned int);
3386 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3387 unsigned int gchanges);
3388 int dev_change_name(struct net_device *, const char *);
3389 int dev_set_alias(struct net_device *, const char *, size_t);
3390 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3391 int dev_set_mtu(struct net_device *, int);
3392 void dev_set_group(struct net_device *, int);
3393 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3394 int dev_change_carrier(struct net_device *, bool new_carrier);
3395 int dev_get_phys_port_id(struct net_device *dev,
3396 struct netdev_phys_item_id *ppid);
3397 int dev_get_phys_port_name(struct net_device *dev,
3398 char *name, size_t len);
3399 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3400 int dev_change_xdp_fd(struct net_device *dev, int fd);
3401 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3402 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3403 struct netdev_queue *txq, int *ret);
3404 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3405 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3406 bool is_skb_forwardable(const struct net_device *dev,
3407 const struct sk_buff *skb);
3409 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3410 struct sk_buff *skb)
3412 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3413 unlikely(!is_skb_forwardable(dev, skb))) {
3414 atomic_long_inc(&dev->rx_dropped);
3419 skb_scrub_packet(skb, true);
3424 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3426 extern int netdev_budget;
3428 /* Called by rtnetlink.c:rtnl_unlock() */
3429 void netdev_run_todo(void);
3432 * dev_put - release reference to device
3433 * @dev: network device
3435 * Release reference to device to allow it to be freed.
3437 static inline void dev_put(struct net_device *dev)
3439 this_cpu_dec(*dev->pcpu_refcnt);
3443 * dev_hold - get reference to device
3444 * @dev: network device
3446 * Hold reference to device to keep it from being freed.
3448 static inline void dev_hold(struct net_device *dev)
3450 this_cpu_inc(*dev->pcpu_refcnt);
3453 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3454 * and _off may be called from IRQ context, but it is caller
3455 * who is responsible for serialization of these calls.
3457 * The name carrier is inappropriate, these functions should really be
3458 * called netif_lowerlayer_*() because they represent the state of any
3459 * kind of lower layer not just hardware media.
3462 void linkwatch_init_dev(struct net_device *dev);
3463 void linkwatch_fire_event(struct net_device *dev);
3464 void linkwatch_forget_dev(struct net_device *dev);
3467 * netif_carrier_ok - test if carrier present
3468 * @dev: network device
3470 * Check if carrier is present on device
3472 static inline bool netif_carrier_ok(const struct net_device *dev)
3474 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3477 unsigned long dev_trans_start(struct net_device *dev);
3479 void __netdev_watchdog_up(struct net_device *dev);
3481 void netif_carrier_on(struct net_device *dev);
3483 void netif_carrier_off(struct net_device *dev);
3486 * netif_dormant_on - mark device as dormant.
3487 * @dev: network device
3489 * Mark device as dormant (as per RFC2863).
3491 * The dormant state indicates that the relevant interface is not
3492 * actually in a condition to pass packets (i.e., it is not 'up') but is
3493 * in a "pending" state, waiting for some external event. For "on-
3494 * demand" interfaces, this new state identifies the situation where the
3495 * interface is waiting for events to place it in the up state.
3497 static inline void netif_dormant_on(struct net_device *dev)
3499 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3500 linkwatch_fire_event(dev);
3504 * netif_dormant_off - set device as not dormant.
3505 * @dev: network device
3507 * Device is not in dormant state.
3509 static inline void netif_dormant_off(struct net_device *dev)
3511 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3512 linkwatch_fire_event(dev);
3516 * netif_dormant - test if carrier present
3517 * @dev: network device
3519 * Check if carrier is present on device
3521 static inline bool netif_dormant(const struct net_device *dev)
3523 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3528 * netif_oper_up - test if device is operational
3529 * @dev: network device
3531 * Check if carrier is operational
3533 static inline bool netif_oper_up(const struct net_device *dev)
3535 return (dev->operstate == IF_OPER_UP ||
3536 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3540 * netif_device_present - is device available or removed
3541 * @dev: network device
3543 * Check if device has not been removed from system.
3545 static inline bool netif_device_present(struct net_device *dev)
3547 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3550 void netif_device_detach(struct net_device *dev);
3552 void netif_device_attach(struct net_device *dev);
3555 * Network interface message level settings
3559 NETIF_MSG_DRV = 0x0001,
3560 NETIF_MSG_PROBE = 0x0002,
3561 NETIF_MSG_LINK = 0x0004,
3562 NETIF_MSG_TIMER = 0x0008,
3563 NETIF_MSG_IFDOWN = 0x0010,
3564 NETIF_MSG_IFUP = 0x0020,
3565 NETIF_MSG_RX_ERR = 0x0040,
3566 NETIF_MSG_TX_ERR = 0x0080,
3567 NETIF_MSG_TX_QUEUED = 0x0100,
3568 NETIF_MSG_INTR = 0x0200,
3569 NETIF_MSG_TX_DONE = 0x0400,
3570 NETIF_MSG_RX_STATUS = 0x0800,
3571 NETIF_MSG_PKTDATA = 0x1000,
3572 NETIF_MSG_HW = 0x2000,
3573 NETIF_MSG_WOL = 0x4000,
3576 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3577 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3578 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3579 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3580 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3581 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3582 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3583 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3584 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3585 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3586 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3587 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3588 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3589 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3590 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3592 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3595 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3596 return default_msg_enable_bits;
3597 if (debug_value == 0) /* no output */
3599 /* set low N bits */
3600 return (1 << debug_value) - 1;
3603 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3605 spin_lock(&txq->_xmit_lock);
3606 txq->xmit_lock_owner = cpu;
3609 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3611 spin_lock_bh(&txq->_xmit_lock);
3612 txq->xmit_lock_owner = smp_processor_id();
3615 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3617 bool ok = spin_trylock(&txq->_xmit_lock);
3619 txq->xmit_lock_owner = smp_processor_id();
3623 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3625 txq->xmit_lock_owner = -1;
3626 spin_unlock(&txq->_xmit_lock);
3629 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3631 txq->xmit_lock_owner = -1;
3632 spin_unlock_bh(&txq->_xmit_lock);
3635 static inline void txq_trans_update(struct netdev_queue *txq)
3637 if (txq->xmit_lock_owner != -1)
3638 txq->trans_start = jiffies;
3641 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3642 static inline void netif_trans_update(struct net_device *dev)
3644 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3646 if (txq->trans_start != jiffies)
3647 txq->trans_start = jiffies;
3651 * netif_tx_lock - grab network device transmit lock
3652 * @dev: network device
3654 * Get network device transmit lock
3656 static inline void netif_tx_lock(struct net_device *dev)
3661 spin_lock(&dev->tx_global_lock);
3662 cpu = smp_processor_id();
3663 for (i = 0; i < dev->num_tx_queues; i++) {
3664 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3666 /* We are the only thread of execution doing a
3667 * freeze, but we have to grab the _xmit_lock in
3668 * order to synchronize with threads which are in
3669 * the ->hard_start_xmit() handler and already
3670 * checked the frozen bit.
3672 __netif_tx_lock(txq, cpu);
3673 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3674 __netif_tx_unlock(txq);
3678 static inline void netif_tx_lock_bh(struct net_device *dev)
3684 static inline void netif_tx_unlock(struct net_device *dev)
3688 for (i = 0; i < dev->num_tx_queues; i++) {
3689 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3691 /* No need to grab the _xmit_lock here. If the
3692 * queue is not stopped for another reason, we
3695 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3696 netif_schedule_queue(txq);
3698 spin_unlock(&dev->tx_global_lock);
3701 static inline void netif_tx_unlock_bh(struct net_device *dev)
3703 netif_tx_unlock(dev);
3707 #define HARD_TX_LOCK(dev, txq, cpu) { \
3708 if ((dev->features & NETIF_F_LLTX) == 0) { \
3709 __netif_tx_lock(txq, cpu); \
3713 #define HARD_TX_TRYLOCK(dev, txq) \
3714 (((dev->features & NETIF_F_LLTX) == 0) ? \
3715 __netif_tx_trylock(txq) : \
3718 #define HARD_TX_UNLOCK(dev, txq) { \
3719 if ((dev->features & NETIF_F_LLTX) == 0) { \
3720 __netif_tx_unlock(txq); \
3724 static inline void netif_tx_disable(struct net_device *dev)
3730 cpu = smp_processor_id();
3731 for (i = 0; i < dev->num_tx_queues; i++) {
3732 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3734 __netif_tx_lock(txq, cpu);
3735 netif_tx_stop_queue(txq);
3736 __netif_tx_unlock(txq);
3741 static inline void netif_addr_lock(struct net_device *dev)
3743 spin_lock(&dev->addr_list_lock);
3746 static inline void netif_addr_lock_nested(struct net_device *dev)
3748 int subclass = SINGLE_DEPTH_NESTING;
3750 if (dev->netdev_ops->ndo_get_lock_subclass)
3751 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3753 spin_lock_nested(&dev->addr_list_lock, subclass);
3756 static inline void netif_addr_lock_bh(struct net_device *dev)
3758 spin_lock_bh(&dev->addr_list_lock);
3761 static inline void netif_addr_unlock(struct net_device *dev)
3763 spin_unlock(&dev->addr_list_lock);
3766 static inline void netif_addr_unlock_bh(struct net_device *dev)
3768 spin_unlock_bh(&dev->addr_list_lock);
3772 * dev_addrs walker. Should be used only for read access. Call with
3773 * rcu_read_lock held.
3775 #define for_each_dev_addr(dev, ha) \
3776 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3778 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3780 void ether_setup(struct net_device *dev);
3782 /* Support for loadable net-drivers */
3783 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3784 unsigned char name_assign_type,
3785 void (*setup)(struct net_device *),
3786 unsigned int txqs, unsigned int rxqs);
3787 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3788 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3790 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3791 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3794 int register_netdev(struct net_device *dev);
3795 void unregister_netdev(struct net_device *dev);
3797 /* General hardware address lists handling functions */
3798 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3799 struct netdev_hw_addr_list *from_list, int addr_len);
3800 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3801 struct netdev_hw_addr_list *from_list, int addr_len);
3802 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3803 struct net_device *dev,
3804 int (*sync)(struct net_device *, const unsigned char *),
3805 int (*unsync)(struct net_device *,
3806 const unsigned char *));
3807 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3808 struct net_device *dev,
3809 int (*unsync)(struct net_device *,
3810 const unsigned char *));
3811 void __hw_addr_init(struct netdev_hw_addr_list *list);
3813 /* Functions used for device addresses handling */
3814 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3815 unsigned char addr_type);
3816 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3817 unsigned char addr_type);
3818 void dev_addr_flush(struct net_device *dev);
3819 int dev_addr_init(struct net_device *dev);
3821 /* Functions used for unicast addresses handling */
3822 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3823 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3824 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3825 int dev_uc_sync(struct net_device *to, struct net_device *from);
3826 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3827 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3828 void dev_uc_flush(struct net_device *dev);
3829 void dev_uc_init(struct net_device *dev);
3832 * __dev_uc_sync - Synchonize device's unicast list
3833 * @dev: device to sync
3834 * @sync: function to call if address should be added
3835 * @unsync: function to call if address should be removed
3837 * Add newly added addresses to the interface, and release
3838 * addresses that have been deleted.
3840 static inline int __dev_uc_sync(struct net_device *dev,
3841 int (*sync)(struct net_device *,
3842 const unsigned char *),
3843 int (*unsync)(struct net_device *,
3844 const unsigned char *))
3846 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3850 * __dev_uc_unsync - Remove synchronized addresses from device
3851 * @dev: device to sync
3852 * @unsync: function to call if address should be removed
3854 * Remove all addresses that were added to the device by dev_uc_sync().
3856 static inline void __dev_uc_unsync(struct net_device *dev,
3857 int (*unsync)(struct net_device *,
3858 const unsigned char *))
3860 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3863 /* Functions used for multicast addresses handling */
3864 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3865 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3866 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3867 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3868 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3869 int dev_mc_sync(struct net_device *to, struct net_device *from);
3870 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3871 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3872 void dev_mc_flush(struct net_device *dev);
3873 void dev_mc_init(struct net_device *dev);
3876 * __dev_mc_sync - Synchonize device's multicast list
3877 * @dev: device to sync
3878 * @sync: function to call if address should be added
3879 * @unsync: function to call if address should be removed
3881 * Add newly added addresses to the interface, and release
3882 * addresses that have been deleted.
3884 static inline int __dev_mc_sync(struct net_device *dev,
3885 int (*sync)(struct net_device *,
3886 const unsigned char *),
3887 int (*unsync)(struct net_device *,
3888 const unsigned char *))
3890 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3894 * __dev_mc_unsync - Remove synchronized addresses from device
3895 * @dev: device to sync
3896 * @unsync: function to call if address should be removed
3898 * Remove all addresses that were added to the device by dev_mc_sync().
3900 static inline void __dev_mc_unsync(struct net_device *dev,
3901 int (*unsync)(struct net_device *,
3902 const unsigned char *))
3904 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3907 /* Functions used for secondary unicast and multicast support */
3908 void dev_set_rx_mode(struct net_device *dev);
3909 void __dev_set_rx_mode(struct net_device *dev);
3910 int dev_set_promiscuity(struct net_device *dev, int inc);
3911 int dev_set_allmulti(struct net_device *dev, int inc);
3912 void netdev_state_change(struct net_device *dev);
3913 void netdev_notify_peers(struct net_device *dev);
3914 void netdev_features_change(struct net_device *dev);
3915 /* Load a device via the kmod */
3916 void dev_load(struct net *net, const char *name);
3917 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3918 struct rtnl_link_stats64 *storage);
3919 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3920 const struct net_device_stats *netdev_stats);
3922 extern int netdev_max_backlog;
3923 extern int netdev_tstamp_prequeue;
3924 extern int weight_p;
3926 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3927 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3928 struct list_head **iter);
3929 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3930 struct list_head **iter);
3932 /* iterate through upper list, must be called under RCU read lock */
3933 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3934 for (iter = &(dev)->adj_list.upper, \
3935 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3937 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3939 /* iterate through upper list, must be called under RCU read lock */
3940 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3941 for (iter = &(dev)->all_adj_list.upper, \
3942 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3944 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3946 void *netdev_lower_get_next_private(struct net_device *dev,
3947 struct list_head **iter);
3948 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3949 struct list_head **iter);
3951 #define netdev_for_each_lower_private(dev, priv, iter) \
3952 for (iter = (dev)->adj_list.lower.next, \
3953 priv = netdev_lower_get_next_private(dev, &(iter)); \
3955 priv = netdev_lower_get_next_private(dev, &(iter)))
3957 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3958 for (iter = &(dev)->adj_list.lower, \
3959 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3961 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3963 void *netdev_lower_get_next(struct net_device *dev,
3964 struct list_head **iter);
3966 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3967 for (iter = (dev)->adj_list.lower.next, \
3968 ldev = netdev_lower_get_next(dev, &(iter)); \
3970 ldev = netdev_lower_get_next(dev, &(iter)))
3972 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3973 struct list_head **iter);
3974 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3975 struct list_head **iter);
3977 #define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3978 for (iter = (dev)->all_adj_list.lower.next, \
3979 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3981 ldev = netdev_all_lower_get_next(dev, &(iter)))
3983 #define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
3984 for (iter = &(dev)->all_adj_list.lower, \
3985 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3987 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3989 void *netdev_adjacent_get_private(struct list_head *adj_list);
3990 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3991 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3992 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3993 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3994 int netdev_master_upper_dev_link(struct net_device *dev,
3995 struct net_device *upper_dev,
3996 void *upper_priv, void *upper_info);
3997 void netdev_upper_dev_unlink(struct net_device *dev,
3998 struct net_device *upper_dev);
3999 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4000 void *netdev_lower_dev_get_private(struct net_device *dev,
4001 struct net_device *lower_dev);
4002 void netdev_lower_state_changed(struct net_device *lower_dev,
4003 void *lower_state_info);
4004 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
4005 struct neighbour *n);
4006 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
4007 struct neighbour *n);
4009 /* RSS keys are 40 or 52 bytes long */
4010 #define NETDEV_RSS_KEY_LEN 52
4011 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4012 void netdev_rss_key_fill(void *buffer, size_t len);
4014 int dev_get_nest_level(struct net_device *dev);
4015 int skb_checksum_help(struct sk_buff *skb);
4016 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4017 netdev_features_t features, bool tx_path);
4018 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4019 netdev_features_t features);
4021 struct netdev_bonding_info {
4026 struct netdev_notifier_bonding_info {
4027 struct netdev_notifier_info info; /* must be first */
4028 struct netdev_bonding_info bonding_info;
4031 void netdev_bonding_info_change(struct net_device *dev,
4032 struct netdev_bonding_info *bonding_info);
4035 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4037 return __skb_gso_segment(skb, features, true);
4039 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4041 static inline bool can_checksum_protocol(netdev_features_t features,
4044 if (protocol == htons(ETH_P_FCOE))
4045 return !!(features & NETIF_F_FCOE_CRC);
4047 /* Assume this is an IP checksum (not SCTP CRC) */
4049 if (features & NETIF_F_HW_CSUM) {
4050 /* Can checksum everything */
4055 case htons(ETH_P_IP):
4056 return !!(features & NETIF_F_IP_CSUM);
4057 case htons(ETH_P_IPV6):
4058 return !!(features & NETIF_F_IPV6_CSUM);
4064 /* Map an ethertype into IP protocol if possible */
4065 static inline int eproto_to_ipproto(int eproto)
4068 case htons(ETH_P_IP):
4070 case htons(ETH_P_IPV6):
4071 return IPPROTO_IPV6;
4078 void netdev_rx_csum_fault(struct net_device *dev);
4080 static inline void netdev_rx_csum_fault(struct net_device *dev)
4084 /* rx skb timestamps */
4085 void net_enable_timestamp(void);
4086 void net_disable_timestamp(void);
4088 #ifdef CONFIG_PROC_FS
4089 int __init dev_proc_init(void);
4091 #define dev_proc_init() 0
4094 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4095 struct sk_buff *skb, struct net_device *dev,
4098 skb->xmit_more = more ? 1 : 0;
4099 return ops->ndo_start_xmit(skb, dev);
4102 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4103 struct netdev_queue *txq, bool more)
4105 const struct net_device_ops *ops = dev->netdev_ops;
4108 rc = __netdev_start_xmit(ops, skb, dev, more);
4109 if (rc == NETDEV_TX_OK)
4110 txq_trans_update(txq);
4115 int netdev_class_create_file_ns(struct class_attribute *class_attr,
4117 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4120 static inline int netdev_class_create_file(struct class_attribute *class_attr)
4122 return netdev_class_create_file_ns(class_attr, NULL);
4125 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4127 netdev_class_remove_file_ns(class_attr, NULL);
4130 extern struct kobj_ns_type_operations net_ns_type_operations;
4132 const char *netdev_drivername(const struct net_device *dev);
4134 void linkwatch_run_queue(void);
4136 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4137 netdev_features_t f2)
4139 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4140 if (f1 & NETIF_F_HW_CSUM)
4141 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4143 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4149 static inline netdev_features_t netdev_get_wanted_features(
4150 struct net_device *dev)
4152 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4154 netdev_features_t netdev_increment_features(netdev_features_t all,
4155 netdev_features_t one, netdev_features_t mask);
4157 /* Allow TSO being used on stacked device :
4158 * Performing the GSO segmentation before last device
4159 * is a performance improvement.
4161 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4162 netdev_features_t mask)
4164 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4167 int __netdev_update_features(struct net_device *dev);
4168 void netdev_update_features(struct net_device *dev);
4169 void netdev_change_features(struct net_device *dev);
4171 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4172 struct net_device *dev);
4174 netdev_features_t passthru_features_check(struct sk_buff *skb,
4175 struct net_device *dev,
4176 netdev_features_t features);
4177 netdev_features_t netif_skb_features(struct sk_buff *skb);
4179 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4181 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4183 /* check flags correspondence */
4184 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4185 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4186 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4187 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4188 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4189 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4190 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4191 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4192 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4193 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4194 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4195 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4196 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4197 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4198 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4199 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4201 return (features & feature) == feature;
4204 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4206 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4207 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4210 static inline bool netif_needs_gso(struct sk_buff *skb,
4211 netdev_features_t features)
4213 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4214 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4215 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4218 static inline void netif_set_gso_max_size(struct net_device *dev,
4221 dev->gso_max_size = size;
4224 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4225 int pulled_hlen, u16 mac_offset,
4228 skb->protocol = protocol;
4229 skb->encapsulation = 1;
4230 skb_push(skb, pulled_hlen);
4231 skb_reset_transport_header(skb);
4232 skb->mac_header = mac_offset;
4233 skb->network_header = skb->mac_header + mac_len;
4234 skb->mac_len = mac_len;
4237 static inline bool netif_is_macsec(const struct net_device *dev)
4239 return dev->priv_flags & IFF_MACSEC;
4242 static inline bool netif_is_macvlan(const struct net_device *dev)
4244 return dev->priv_flags & IFF_MACVLAN;
4247 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4249 return dev->priv_flags & IFF_MACVLAN_PORT;
4252 static inline bool netif_is_ipvlan(const struct net_device *dev)
4254 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4257 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4259 return dev->priv_flags & IFF_IPVLAN_MASTER;
4262 static inline bool netif_is_bond_master(const struct net_device *dev)
4264 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4267 static inline bool netif_is_bond_slave(const struct net_device *dev)
4269 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4272 static inline bool netif_supports_nofcs(struct net_device *dev)
4274 return dev->priv_flags & IFF_SUPP_NOFCS;
4277 static inline bool netif_is_l3_master(const struct net_device *dev)
4279 return dev->priv_flags & IFF_L3MDEV_MASTER;
4282 static inline bool netif_is_l3_slave(const struct net_device *dev)
4284 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4287 static inline bool netif_is_bridge_master(const struct net_device *dev)
4289 return dev->priv_flags & IFF_EBRIDGE;
4292 static inline bool netif_is_bridge_port(const struct net_device *dev)
4294 return dev->priv_flags & IFF_BRIDGE_PORT;
4297 static inline bool netif_is_ovs_master(const struct net_device *dev)
4299 return dev->priv_flags & IFF_OPENVSWITCH;
4302 static inline bool netif_is_team_master(const struct net_device *dev)
4304 return dev->priv_flags & IFF_TEAM;
4307 static inline bool netif_is_team_port(const struct net_device *dev)
4309 return dev->priv_flags & IFF_TEAM_PORT;
4312 static inline bool netif_is_lag_master(const struct net_device *dev)
4314 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4317 static inline bool netif_is_lag_port(const struct net_device *dev)
4319 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4322 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4324 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4327 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4328 static inline void netif_keep_dst(struct net_device *dev)
4330 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4333 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4334 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4336 /* TODO: reserve and use an additional IFF bit, if we get more users */
4337 return dev->priv_flags & IFF_MACSEC;
4340 extern struct pernet_operations __net_initdata loopback_net_ops;
4342 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4344 /* netdev_printk helpers, similar to dev_printk */
4346 static inline const char *netdev_name(const struct net_device *dev)
4348 if (!dev->name[0] || strchr(dev->name, '%'))
4349 return "(unnamed net_device)";
4353 static inline const char *netdev_reg_state(const struct net_device *dev)
4355 switch (dev->reg_state) {
4356 case NETREG_UNINITIALIZED: return " (uninitialized)";
4357 case NETREG_REGISTERED: return "";
4358 case NETREG_UNREGISTERING: return " (unregistering)";
4359 case NETREG_UNREGISTERED: return " (unregistered)";
4360 case NETREG_RELEASED: return " (released)";
4361 case NETREG_DUMMY: return " (dummy)";
4364 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4365 return " (unknown)";
4369 void netdev_printk(const char *level, const struct net_device *dev,
4370 const char *format, ...);
4372 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4374 void netdev_alert(const struct net_device *dev, const char *format, ...);
4376 void netdev_crit(const struct net_device *dev, const char *format, ...);
4378 void netdev_err(const struct net_device *dev, const char *format, ...);
4380 void netdev_warn(const struct net_device *dev, const char *format, ...);
4382 void netdev_notice(const struct net_device *dev, const char *format, ...);
4384 void netdev_info(const struct net_device *dev, const char *format, ...);
4386 #define MODULE_ALIAS_NETDEV(device) \
4387 MODULE_ALIAS("netdev-" device)
4389 #if defined(CONFIG_DYNAMIC_DEBUG)
4390 #define netdev_dbg(__dev, format, args...) \
4392 dynamic_netdev_dbg(__dev, format, ##args); \
4394 #elif defined(DEBUG)
4395 #define netdev_dbg(__dev, format, args...) \
4396 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4398 #define netdev_dbg(__dev, format, args...) \
4401 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4405 #if defined(VERBOSE_DEBUG)
4406 #define netdev_vdbg netdev_dbg
4409 #define netdev_vdbg(dev, format, args...) \
4412 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4418 * netdev_WARN() acts like dev_printk(), but with the key difference
4419 * of using a WARN/WARN_ON to get the message out, including the
4420 * file/line information and a backtrace.
4422 #define netdev_WARN(dev, format, args...) \
4423 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4424 netdev_reg_state(dev), ##args)
4426 /* netif printk helpers, similar to netdev_printk */
4428 #define netif_printk(priv, type, level, dev, fmt, args...) \
4430 if (netif_msg_##type(priv)) \
4431 netdev_printk(level, (dev), fmt, ##args); \
4434 #define netif_level(level, priv, type, dev, fmt, args...) \
4436 if (netif_msg_##type(priv)) \
4437 netdev_##level(dev, fmt, ##args); \
4440 #define netif_emerg(priv, type, dev, fmt, args...) \
4441 netif_level(emerg, priv, type, dev, fmt, ##args)
4442 #define netif_alert(priv, type, dev, fmt, args...) \
4443 netif_level(alert, priv, type, dev, fmt, ##args)
4444 #define netif_crit(priv, type, dev, fmt, args...) \
4445 netif_level(crit, priv, type, dev, fmt, ##args)
4446 #define netif_err(priv, type, dev, fmt, args...) \
4447 netif_level(err, priv, type, dev, fmt, ##args)
4448 #define netif_warn(priv, type, dev, fmt, args...) \
4449 netif_level(warn, priv, type, dev, fmt, ##args)
4450 #define netif_notice(priv, type, dev, fmt, args...) \
4451 netif_level(notice, priv, type, dev, fmt, ##args)
4452 #define netif_info(priv, type, dev, fmt, args...) \
4453 netif_level(info, priv, type, dev, fmt, ##args)
4455 #if defined(CONFIG_DYNAMIC_DEBUG)
4456 #define netif_dbg(priv, type, netdev, format, args...) \
4458 if (netif_msg_##type(priv)) \
4459 dynamic_netdev_dbg(netdev, format, ##args); \
4461 #elif defined(DEBUG)
4462 #define netif_dbg(priv, type, dev, format, args...) \
4463 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4465 #define netif_dbg(priv, type, dev, format, args...) \
4468 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4473 #if defined(VERBOSE_DEBUG)
4474 #define netif_vdbg netif_dbg
4476 #define netif_vdbg(priv, type, dev, format, args...) \
4479 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4485 * The list of packet types we will receive (as opposed to discard)
4486 * and the routines to invoke.
4488 * Why 16. Because with 16 the only overlap we get on a hash of the
4489 * low nibble of the protocol value is RARP/SNAP/X.25.
4491 * NOTE: That is no longer true with the addition of VLAN tags. Not
4492 * sure which should go first, but I bet it won't make much
4493 * difference if we are running VLANs. The good news is that
4494 * this protocol won't be in the list unless compiled in, so
4495 * the average user (w/out VLANs) will not be adversely affected.
4511 #define PTYPE_HASH_SIZE (16)
4512 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4514 #endif /* _LINUX_NETDEVICE_H */