blob: ce784d5018e07dab38807313fa4666e6b7cc4223 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * 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>
Alan Cox113aa832008-10-13 19:01:08 -070014 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
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.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
Jean Pihete8db0be2011-08-25 15:35:03 +020028#include <linux/pm_qos.h>
Al Virod7fe0f22006-12-03 23:15:30 -050029#include <linux/timer.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050030#include <linux/bug.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070031#include <linux/delay.h>
Arun Sharma600634972011-07-26 16:09:06 -070032#include <linux/atomic.h>
Eric Dumazet53511452014-10-08 08:19:27 -070033#include <linux/prefetch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <asm/cache.h>
35#include <asm/byteorder.h>
36
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#include <linux/percpu.h>
David S. Miller4d5b78c2009-05-06 16:52:51 -070038#include <linux/rculist.h>
Chris Leechdb217332006-06-17 21:24:58 -070039#include <linux/dmaengine.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070040#include <linux/workqueue.h>
Tom Herbert114cf582011-11-28 16:33:09 +000041#include <linux/dynamic_queue_limits.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
Patrick McHardyb1b67dd2009-04-20 04:49:28 +000043#include <linux/ethtool.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020044#include <net/net_namespace.h>
Lennert Buytenhekcf85d082008-10-07 13:45:02 +000045#include <net/dsa.h>
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -080046#ifdef CONFIG_DCB
Alexander Duyck2f90b862008-11-20 20:52:10 -080047#include <net/dcbnl.h>
48#endif
Neil Horman5bc14212011-11-22 05:10:51 +000049#include <net/netprio_cgroup.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020050
Michał Mirosława59e2ec2011-11-15 15:29:55 +000051#include <linux/netdev_features.h>
John Fastabend77162022012-04-15 06:43:56 +000052#include <linux/neighbour.h>
David Howells607ca462012-10-13 10:46:48 +010053#include <uapi/linux/netdevice.h>
Moni Shoua61bd3852015-02-03 16:48:29 +020054#include <uapi/linux/if_bonding.h>
Michał Mirosława59e2ec2011-11-15 15:29:55 +000055
Jeff Moyer115c1d62005-06-22 22:05:31 -070056struct netpoll_info;
Paul Gortmaker313162d2012-01-30 11:46:54 -050057struct device;
Richard Cochranc1f19b52010-07-17 08:49:36 +000058struct phy_device;
Johannes Berg704232c2007-04-23 12:20:05 -070059/* 802.11 specific */
60struct wireless_dev;
Alexander Aring98a18b62014-11-02 06:44:54 +010061/* 802.15.4 specific */
62struct wpan_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
Joe Perchesf629d202013-09-26 14:48:15 -070064void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000066
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000067/* Backlog congestion levels */
68#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69#define NET_RX_DROP 1 /* packet dropped */
70
Patrick McHardy572a9d72009-11-10 06:14:14 +000071/*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88/* qdisc ->enqueue() return codes. */
89#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000090#define NET_XMIT_DROP 0x01 /* skb dropped */
91#define NET_XMIT_CN 0x02 /* congestion notification */
92#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -020095/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +000098#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -070099#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000101/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000102#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000103
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000104enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000109};
110typedef enum netdev_tx netdev_tx_t;
111
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000112/*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116static inline bool dev_xmit_complete(int rc)
117{
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128}
129
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130/*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800134
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000135#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700136# if defined(CONFIG_MAC80211_MESH)
137# define LL_MAX_HEADER 128
138# else
139# define LL_MAX_HEADER 96
140# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700142# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143#endif
144
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000145#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147#define MAX_HEADER LL_MAX_HEADER
148#else
149#define MAX_HEADER (LL_MAX_HEADER + 48)
150#endif
151
152/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800156
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800157struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
181};
182
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183
184#include <linux/cache.h>
185#include <linux/skbuff.h>
186
Eric Dumazetadc93002011-11-17 03:13:26 +0000187#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100188#include <linux/static_key.h>
189extern struct static_key rps_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000190#endif
191
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192struct neighbour;
193struct neigh_parms;
194struct sk_buff;
195
Jiri Pirkof001fde2009-05-05 02:48:28 +0000196struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
199 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000200#define NETDEV_HW_ADDR_T_LAN 1
201#define NETDEV_HW_ADDR_T_SAN 2
202#define NETDEV_HW_ADDR_T_SLAVE 3
203#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000204#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000205 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000206 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700207 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400208 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000209 struct rcu_head rcu_head;
210};
211
Jiri Pirko31278e72009-06-17 01:12:19 +0000212struct netdev_hw_addr_list {
213 struct list_head list;
214 int count;
215};
216
Jiri Pirko22bedad32010-04-01 21:22:57 +0000217#define netdev_hw_addr_list_count(l) ((l)->count)
218#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219#define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
221
222#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800224#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800226
Jiri Pirko22bedad32010-04-01 21:22:57 +0000227#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700229#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece32010-02-04 10:22:25 -0800231
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800232struct hh_cache {
David S. Millerf6b72b622011-07-14 07:53:20 -0700233 u16 hh_len;
David S. Miller5c25f682011-07-13 00:51:10 -0700234 u16 __pad;
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800235 seqlock_t hh_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236
237 /* cached hardware header; allow for machine alignment needs. */
238#define HH_DATA_MOD 16
239#define HH_DATA_OFF(__len) \
Jiri Benc5ba0eac2005-06-02 16:48:05 -0700240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241#define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
244};
245
246/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * Alternative is:
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 *
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
253 */
254#define LL_RESERVED_SPACE(dev) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700259struct header_ops {
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
Eric Dumazet95c96172012-04-15 05:58:06 +0000262 const void *saddr, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
264 int (*rebuild)(struct sk_buff *skb);
David S. Millere69dd332011-07-12 23:28:12 -0700265 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700266 void (*cache_update)(struct hh_cache *hh,
267 const struct net_device *dev,
268 const unsigned char *haddr);
269};
270
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271/* These flag bits are private to the generic network queueing
272 * layer, they may not be explicitly referenced by any other
273 * code.
274 */
275
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800276enum netdev_state_t {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 __LINK_STATE_START,
278 __LINK_STATE_PRESENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 __LINK_STATE_NOCARRIER,
Stefan Rompfb00055a2006-03-20 17:09:11 -0800280 __LINK_STATE_LINKWATCH_PENDING,
281 __LINK_STATE_DORMANT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282};
283
284
285/*
286 * This structure holds at boot time configured netdevice settings. They
Graf Yangfe2918b2009-02-05 21:26:19 -0800287 * are then used in the device probing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 */
289struct netdev_boot_setup {
290 char name[IFNAMSIZ];
291 struct ifmap map;
292};
293#define NETDEV_BOOT_SETUP_MAX 8
294
Joe Perchesf629d202013-09-26 14:48:15 -0700295int __init netdev_boot_setup(char *str);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296
297/*
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700298 * Structure for NAPI scheduling similar to tasklet but with weighting
299 */
300struct napi_struct {
301 /* The poll_list must only be managed by the entity which
302 * changes the state of the NAPI_STATE_SCHED bit. This means
303 * whoever atomically sets that bit can add this napi_struct
304 * to the per-cpu poll_list, and whoever clears that bit
305 * can remove from the list right before clearing the bit.
306 */
307 struct list_head poll_list;
308
309 unsigned long state;
310 int weight;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000311 unsigned int gro_count;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700312 int (*poll)(struct napi_struct *, int);
313#ifdef CONFIG_NETPOLL
314 spinlock_t poll_lock;
315 int poll_owner;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700316#endif
Herbert Xu5d38a072009-01-04 16:13:40 -0800317 struct net_device *dev;
Herbert Xud565b0a2008-12-15 23:38:52 -0800318 struct sk_buff *gro_list;
Herbert Xu5d38a072009-01-04 16:13:40 -0800319 struct sk_buff *skb;
Eric Dumazet3b47d302014-11-06 21:09:44 -0800320 struct hrtimer timer;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000321 struct list_head dev_list;
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300322 struct hlist_node napi_hash_node;
323 unsigned int napi_id;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700324};
325
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800326enum {
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700327 NAPI_STATE_SCHED, /* Poll is scheduled */
David S. Millera0a46192008-01-07 20:35:07 -0800328 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800329 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300330 NAPI_STATE_HASHED, /* In NAPI hash */
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700331};
332
Ben Hutchings5b252f02009-10-29 07:17:09 +0000333enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700334 GRO_MERGED,
335 GRO_MERGED_FREE,
336 GRO_HELD,
337 GRO_NORMAL,
338 GRO_DROP,
339};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000340typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700341
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000342/*
343 * enum rx_handler_result - Possible return values for rx_handlers.
344 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
345 * further.
346 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
347 * case skb->dev was changed by rx_handler.
348 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
349 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
350 *
351 * rx_handlers are functions called from inside __netif_receive_skb(), to do
352 * special processing of the skb, prior to delivery to protocol handlers.
353 *
354 * Currently, a net_device can only have a single rx_handler registered. Trying
355 * to register a second rx_handler will return -EBUSY.
356 *
357 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
358 * To unregister a rx_handler on a net_device, use
359 * netdev_rx_handler_unregister().
360 *
361 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
362 * do with the skb.
363 *
364 * If the rx_handler consumed to skb in some way, it should return
365 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
366 * the skb to be delivered in some other ways.
367 *
368 * If the rx_handler changed skb->dev, to divert the skb to another
369 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
370 * new device will be called if it exists.
371 *
372 * If the rx_handler consider the skb should be ignored, it should return
373 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400374 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000375 *
376 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
377 * delivered, it should return RX_HANDLER_PASS.
378 *
379 * A device without a registered rx_handler will behave as if rx_handler
380 * returned RX_HANDLER_PASS.
381 */
382
383enum rx_handler_result {
384 RX_HANDLER_CONSUMED,
385 RX_HANDLER_ANOTHER,
386 RX_HANDLER_EXACT,
387 RX_HANDLER_PASS,
388};
389typedef enum rx_handler_result rx_handler_result_t;
390typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000391
Joe Perchesf629d202013-09-26 14:48:15 -0700392void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700393void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700394
David S. Miller4d295152012-03-07 21:02:35 -0500395static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800396{
397 return test_bit(NAPI_STATE_DISABLE, &n->state);
398}
399
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700400/**
401 * napi_schedule_prep - check if napi can be scheduled
402 * @n: napi context
403 *
404 * Test if NAPI routine is already running, and if not mark
405 * it as running. This is used as a condition variable
David S. Millera0a46192008-01-07 20:35:07 -0800406 * insure only one NAPI poll instance runs. We also make
407 * sure there is no pending NAPI disable.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700408 */
David S. Miller4d295152012-03-07 21:02:35 -0500409static inline bool napi_schedule_prep(struct napi_struct *n)
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700410{
David S. Millera0a46192008-01-07 20:35:07 -0800411 return !napi_disable_pending(n) &&
412 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700413}
414
415/**
416 * napi_schedule - schedule NAPI poll
417 * @n: napi context
418 *
419 * Schedule NAPI poll routine to be called if it is not already
420 * running.
421 */
422static inline void napi_schedule(struct napi_struct *n)
423{
424 if (napi_schedule_prep(n))
425 __napi_schedule(n);
426}
427
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700428/**
429 * napi_schedule_irqoff - schedule NAPI poll
430 * @n: napi context
431 *
432 * Variant of napi_schedule(), assuming hard irqs are masked.
433 */
434static inline void napi_schedule_irqoff(struct napi_struct *n)
435{
436 if (napi_schedule_prep(n))
437 __napi_schedule_irqoff(n);
438}
439
Roland Dreierbfe13f52007-10-09 15:47:37 -0700440/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
David S. Miller4d295152012-03-07 21:02:35 -0500441static inline bool napi_reschedule(struct napi_struct *napi)
Roland Dreierbfe13f52007-10-09 15:47:37 -0700442{
443 if (napi_schedule_prep(napi)) {
444 __napi_schedule(napi);
David S. Miller4d295152012-03-07 21:02:35 -0500445 return true;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700446 }
David S. Miller4d295152012-03-07 21:02:35 -0500447 return false;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700448}
449
Eric Dumazet3b47d302014-11-06 21:09:44 -0800450void __napi_complete(struct napi_struct *n);
451void napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700452/**
453 * napi_complete - NAPI processing complete
454 * @n: napi context
455 *
456 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800457 * Consider using napi_complete_done() instead.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700458 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800459static inline void napi_complete(struct napi_struct *n)
460{
461 return napi_complete_done(n, 0);
462}
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700463
464/**
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300465 * napi_by_id - lookup a NAPI by napi_id
466 * @napi_id: hashed napi_id
467 *
468 * lookup @napi_id in napi_hash table
469 * must be called under rcu_read_lock()
470 */
Joe Perchesf629d202013-09-26 14:48:15 -0700471struct napi_struct *napi_by_id(unsigned int napi_id);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300472
473/**
474 * napi_hash_add - add a NAPI to global hashtable
475 * @napi: napi context
476 *
477 * generate a new napi_id and store a @napi under it in napi_hash
478 */
Joe Perchesf629d202013-09-26 14:48:15 -0700479void napi_hash_add(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300480
481/**
482 * napi_hash_del - remove a NAPI from global table
483 * @napi: napi context
484 *
485 * Warning: caller must observe rcu grace period
486 * before freeing memory containing @napi
487 */
Joe Perchesf629d202013-09-26 14:48:15 -0700488void napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300489
490/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700491 * napi_disable - prevent NAPI from scheduling
492 * @n: napi context
493 *
494 * Stop NAPI from being scheduled on this context.
495 * Waits till any outstanding processing completes.
496 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800497void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700498
499/**
500 * napi_enable - enable NAPI scheduling
501 * @n: napi context
502 *
503 * Resume NAPI from being scheduled on this context.
504 * Must be paired with napi_disable.
505 */
506static inline void napi_enable(struct napi_struct *n)
507{
508 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100509 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700510 clear_bit(NAPI_STATE_SCHED, &n->state);
511}
512
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700513#ifdef CONFIG_SMP
514/**
515 * napi_synchronize - wait until NAPI is not running
516 * @n: napi context
517 *
518 * Wait until NAPI is done being scheduled on this context.
519 * Waits till any outstanding processing completes but
520 * does not disable future activations.
521 */
522static inline void napi_synchronize(const struct napi_struct *n)
523{
524 while (test_bit(NAPI_STATE_SCHED, &n->state))
525 msleep(1);
526}
527#else
528# define napi_synchronize(n) barrier()
529#endif
530
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800531enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000532 __QUEUE_STATE_DRV_XOFF,
533 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700534 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700535};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200536
537#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
538#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
539#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
540
541#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
542#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
543 QUEUE_STATE_FROZEN)
544#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
545 QUEUE_STATE_FROZEN)
546
Tom Herbert734664982011-11-28 16:32:44 +0000547/*
548 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
549 * netif_tx_* functions below are used to manipulate this flag. The
550 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
551 * queue independently. The netif_xmit_*stopped functions below are called
552 * to check if the queue has been stopped by the driver or stack (either
553 * of the XOFF bits are set in the state). Drivers should not need to call
554 * netif_xmit*stopped functions, they should only be using netif_tx_*.
555 */
David S. Miller79d16382008-07-08 23:14:46 -0700556
David S. Millerbb949fb2008-07-08 16:55:56 -0700557struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700558/*
559 * read mostly part
560 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700561 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700562 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700563 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000564#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000565 struct kobject kobj;
566#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000567#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
568 int numa_node;
569#endif
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700570/*
571 * write mostly part
572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700575 /*
576 * please use this field instead of dev->trans_start
577 */
578 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000579
580 /*
581 * Number of TX timeouts for this queue
582 * (/sys/class/net/DEV/Q/trans_timeout)
583 */
584 unsigned long trans_timeout;
Tom Herbert114cf582011-11-28 16:33:09 +0000585
586 unsigned long state;
587
588#ifdef CONFIG_BQL
589 struct dql dql;
590#endif
David S. Millere8a04642008-07-17 00:34:19 -0700591} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700592
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000593static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
594{
595#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 return q->numa_node;
597#else
Changli Gaob236da62010-12-14 03:09:15 +0000598 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000599#endif
600}
601
602static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
603{
604#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 q->numa_node = node;
606#endif
607}
608
Eric Dumazetdf334542010-03-24 19:13:54 +0000609#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000610/*
611 * This structure holds an RPS map which can be of variable length. The
612 * map is an array of CPUs.
613 */
614struct rps_map {
615 unsigned int len;
616 struct rcu_head rcu;
617 u16 cpus[0];
618};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000619#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000620
Tom Herbertfec5e652010-04-16 16:01:27 -0700621/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000622 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
623 * tail pointer for that CPU's input queue at the time of last enqueue, and
624 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700625 */
626struct rps_dev_flow {
627 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000628 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700629 unsigned int last_qtail;
630};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000631#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700632
633/*
634 * The rps_dev_flow_table structure contains a table of flow mappings.
635 */
636struct rps_dev_flow_table {
637 unsigned int mask;
638 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700639 struct rps_dev_flow flows[0];
640};
641#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000642 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700643
644/*
645 * The rps_sock_flow_table contains mappings of flows to the last CPU
646 * on which they were processed by the application (set in recvmsg).
647 */
648struct rps_sock_flow_table {
649 unsigned int mask;
650 u16 ents[0];
651};
652#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000653 ((_num) * sizeof(u16)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700654
655#define RPS_NO_CPU 0xffff
656
657static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
658 u32 hash)
659{
660 if (table && hash) {
661 unsigned int cpu, index = hash & table->mask;
662
663 /* We only give a hint, preemption can change cpu under us */
664 cpu = raw_smp_processor_id();
665
666 if (table->ents[index] != cpu)
667 table->ents[index] = cpu;
668 }
669}
670
671static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
672 u32 hash)
673{
674 if (table && hash)
675 table->ents[hash & table->mask] = RPS_NO_CPU;
676}
677
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000678extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
Tom Herbertfec5e652010-04-16 16:01:27 -0700679
Ben Hutchingsc4454772011-01-19 11:03:53 +0000680#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700681bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
682 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000683#endif
Michael Daltona953be52014-01-16 22:23:28 -0800684#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000685
Tom Herbert0a9627f2010-03-16 08:03:29 +0000686/* This structure contains an instance of an RX queue. */
687struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800688#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000689 struct rps_map __rcu *rps_map;
690 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800691#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000692 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000693 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000694} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800695
696/*
697 * RX queue sysfs structures and functions.
698 */
699struct rx_queue_attribute {
700 struct attribute attr;
701 ssize_t (*show)(struct netdev_rx_queue *queue,
702 struct rx_queue_attribute *attr, char *buf);
703 ssize_t (*store)(struct netdev_rx_queue *queue,
704 struct rx_queue_attribute *attr, const char *buf, size_t len);
705};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800706
Tom Herbertbf264142010-11-26 08:36:09 +0000707#ifdef CONFIG_XPS
708/*
709 * This structure holds an XPS map which can be of variable length. The
710 * map is an array of queues.
711 */
712struct xps_map {
713 unsigned int len;
714 unsigned int alloc_len;
715 struct rcu_head rcu;
716 u16 queues[0];
717};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000718#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Tom Herbertbf264142010-11-26 08:36:09 +0000719#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
720 / sizeof(u16))
721
722/*
723 * This structure holds all XPS maps for device. Maps are indexed by CPU.
724 */
725struct xps_dev_maps {
726 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000727 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000728};
729#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
730 (nr_cpu_ids * sizeof(struct xps_map *)))
731#endif /* CONFIG_XPS */
732
John Fastabend4f57c082011-01-17 08:06:04 +0000733#define TC_MAX_QUEUE 16
734#define TC_BITMASK 15
735/* HW offloaded queuing disciplines txq count and offset maps */
736struct netdev_tc_txq {
737 u16 count;
738 u16 offset;
739};
740
Neerav Parikh68bad942012-01-04 20:23:39 +0000741#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
742/*
743 * This structure is to hold information about the device
744 * configured to run FCoE protocol stack.
745 */
746struct netdev_fcoe_hbainfo {
747 char manufacturer[64];
748 char serial_number[64];
749 char hardware_version[64];
750 char driver_version[64];
751 char optionrom_version[64];
752 char firmware_version[64];
753 char model[256];
754 char model_description[256];
755};
756#endif
757
Jiri Pirko02637fc2014-11-28 14:34:16 +0100758#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200759
Jiri Pirko02637fc2014-11-28 14:34:16 +0100760/* This structure holds a unique identifier to identify some
761 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200762 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100763struct netdev_phys_item_id {
764 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200765 unsigned char id_len;
766};
767
Daniel Borkmann99932d42014-02-16 15:55:20 +0100768typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
769 struct sk_buff *skb);
770
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800771/*
772 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800773 * The following hooks can be defined; unless noted otherwise, they are
774 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800775 *
776 * int (*ndo_init)(struct net_device *dev);
777 * This function is called once when network device is registered.
778 * The network device can use this to any late stage initializaton
779 * or semantic validattion. It can fail with an error code which will
780 * be propogated back to register_netdev
781 *
782 * void (*ndo_uninit)(struct net_device *dev);
783 * This function is called when device is unregistered or when registration
784 * fails. It is not called if init fails.
785 *
786 * int (*ndo_open)(struct net_device *dev);
787 * This function is called when network device transistions to the up
788 * state.
789 *
790 * int (*ndo_stop)(struct net_device *dev);
791 * This function is called when network device transistions to the down
792 * state.
793 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000794 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
795 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800796 * Called when a packet needs to be transmitted.
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000797 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
798 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
Stephen Hemminger00829822008-11-20 20:14:53 -0800799 * Required can not be NULL.
800 *
Jason Wangf663dd92014-01-10 16:18:26 +0800801 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100802 * void *accel_priv, select_queue_fallback_t fallback);
Stephen Hemminger00829822008-11-20 20:14:53 -0800803 * Called to decide which queue to when device supports multiple
804 * transmit queues.
805 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800806 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
807 * This function is called to allow device receiver to make
808 * changes to configuration when multicast or promiscious is enabled.
809 *
810 * void (*ndo_set_rx_mode)(struct net_device *dev);
811 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000812 * If driver handles unicast address filtering, it should set
813 * IFF_UNICAST_FLT to its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800814 *
815 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
816 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700817 * needs to be changed. If this interface is not defined, the
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800818 * mac address can not be changed.
819 *
820 * int (*ndo_validate_addr)(struct net_device *dev);
821 * Test if Media Access Control address is valid for the device.
822 *
823 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
824 * Called when a user request an ioctl which can't be handled by
825 * the generic interface code. If not defined ioctl's return
826 * not supported error code.
827 *
828 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
829 * Used to set network devices bus interface parameters. This interface
830 * is retained for legacy reason, new devices should use the bus
831 * interface (PCI) for low level management.
832 *
833 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
834 * Called when a user wants to change the Maximum Transfer Unit
835 * of a device. If not defined, any request to change MTU will
836 * will return an error.
837 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800838 * void (*ndo_tx_timeout)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800839 * Callback uses when the transmitter has not made any progress
840 * for dev->watchdog ticks.
841 *
Ben Hutchings3cfde792010-07-09 09:11:52 +0000842 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
Eric Dumazet28172732010-07-07 14:58:56 -0700843 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700844 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800845 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000846 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000847 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
848 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700849 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000850 * (which should normally be dev->stats) and return a pointer to
851 * it. The structure may be changed asynchronously only if each
852 * field is written atomically.
853 * 3. Update dev->stats asynchronously and atomically, and define
854 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800855 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530856 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Patrick McHardy80d5c362013-04-19 02:04:28 +0000857 * If device support VLAN filtering this function is called when a
858 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800859 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530860 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Patrick McHardy80d5c362013-04-19 02:04:28 +0000861 * If device support VLAN filtering this function is called when a
862 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800863 *
864 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000865 *
866 * SR-IOV management functions.
867 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
868 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400869 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
870 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000871 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000872 * int (*ndo_get_vf_config)(struct net_device *dev,
873 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300874 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700875 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
876 * struct nlattr *port[]);
877 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +0000878 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
879 * Called to setup 'tc' number of traffic classes in the net device. This
880 * is always called from the stack with the rtnl lock held and netif tx
881 * queues stopped. This allows the netdevice to perform queue management
882 * safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000883 *
Yi Zoue9bce842011-03-09 08:48:03 +0000884 * Fiber Channel over Ethernet (FCoE) offload functions.
885 * int (*ndo_fcoe_enable)(struct net_device *dev);
886 * Called when the FCoE protocol stack wants to start using LLD for FCoE
887 * so the underlying device can perform whatever needed configuration or
888 * initialization to support acceleration of FCoE traffic.
889 *
890 * int (*ndo_fcoe_disable)(struct net_device *dev);
891 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
892 * so the underlying device can perform whatever needed clean-ups to
893 * stop supporting acceleration of FCoE traffic.
894 *
895 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
896 * struct scatterlist *sgl, unsigned int sgc);
897 * Called when the FCoE Initiator wants to initialize an I/O that
898 * is a possible candidate for Direct Data Placement (DDP). The LLD can
899 * perform necessary setup and returns 1 to indicate the device is set up
900 * successfully to perform DDP on this I/O, otherwise this returns 0.
901 *
902 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
903 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
904 * indicated by the FC exchange id 'xid', so the underlying device can
905 * clean up and reuse resources for later DDP requests.
906 *
907 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
908 * struct scatterlist *sgl, unsigned int sgc);
909 * Called when the FCoE Target wants to initialize an I/O that
910 * is a possible candidate for Direct Data Placement (DDP). The LLD can
911 * perform necessary setup and returns 1 to indicate the device is set up
912 * successfully to perform DDP on this I/O, otherwise this returns 0.
913 *
Neerav Parikh68bad942012-01-04 20:23:39 +0000914 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
915 * struct netdev_fcoe_hbainfo *hbainfo);
916 * Called when the FCoE Protocol stack wants information on the underlying
917 * device. This information is utilized by the FCoE protocol stack to
918 * register attributes with Fiber Channel management service as per the
919 * FC-GS Fabric Device Management Information(FDMI) specification.
920 *
Yi Zoue9bce842011-03-09 08:48:03 +0000921 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
922 * Called when the underlying device wants to override default World Wide
923 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
924 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
925 * protocol stack to use.
926 *
Ben Hutchingsc4454772011-01-19 11:03:53 +0000927 * RFS acceleration.
928 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
929 * u16 rxq_index, u32 flow_id);
930 * Set hardware filter for RFS. rxq_index is the target queue index;
931 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
932 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +0000933 *
Jiri Pirko8b98a702013-01-03 22:49:02 +0000934 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +0000935 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
936 * Called to make another netdev an underling.
937 *
938 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
939 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +0000940 *
941 * Feature/offload setting functions.
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000942 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
943 * netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +0000944 * Adjusts the requested feature flags according to device-specific
945 * constraints, and returns the resulting flags. Must not modify
946 * the device state.
947 *
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000948 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +0000949 * Called to update device configuration to new features. Passed
950 * feature set might be less than what was returned by ndo_fix_features()).
951 * Must return >0 or -errno if it changed dev->features itself.
952 *
stephen hemmingeredc7d572012-10-01 12:32:33 +0000953 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
954 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100955 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +0000956 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +0000957 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
958 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100959 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +0000960 * Deletes the FDB entry from dev coresponding to addr.
961 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -0400962 * struct net_device *dev, struct net_device *filter_dev,
963 * int idx)
John Fastabend77162022012-04-15 06:43:56 +0000964 * Used to add FDB entries to dump requests. Implementers should add
965 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +0000966 *
967 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
968 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Dmitry Kravkov24f11a52013-03-27 06:54:00 +0000969 * struct net_device *dev, u32 filter_mask)
Jiri Pirko4bf84c32012-12-27 23:49:37 +0000970 *
971 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
972 * Called to change device carrier. Soft-devices (like dummy, team, etc)
973 * which do not represent real hardware may define this to allow their
974 * userspace components to manage their virtual carrier state. Devices
975 * that determine carrier state from physical hardware properties (eg
976 * network cables) or protocol-dependent mechanisms (eg
977 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200978 *
979 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +0100980 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +0200981 * Called to get ID of physical port of this device. If driver does
982 * not implement this, it is assumed that the hw is not able to have
983 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700984 *
985 * void (*ndo_add_vxlan_port)(struct net_device *dev,
Joseph Gasparakis35e42372013-09-13 07:34:13 -0700986 * sa_family_t sa_family, __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700987 * Called by vxlan to notiy a driver about the UDP port and socket
988 * address family that vxlan is listnening to. It is called only when
989 * a new port starts listening. The operation is protected by the
990 * vxlan_net->sock_lock.
991 *
992 * void (*ndo_del_vxlan_port)(struct net_device *dev,
Joseph Gasparakis35e42372013-09-13 07:34:13 -0700993 * sa_family_t sa_family, __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700994 * Called by vxlan to notify the driver about a UDP port and socket
995 * address family that vxlan is not listening to anymore. The operation
996 * is protected by the vxlan_net->sock_lock.
John Fastabenda6cc0cf2013-11-06 09:54:46 -0800997 *
998 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
999 * struct net_device *dev)
1000 * Called by upper layer devices to accelerate switching or other
1001 * station functionality into hardware. 'pdev is the lowerdev
1002 * to use for the offload and 'dev' is the net device that will
1003 * back the offload. Returns a pointer to the private structure
1004 * the upper layer will maintain.
1005 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1006 * Called by upper layer device to delete the station created
1007 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1008 * the station and priv is the structure returned by the add
1009 * operation.
1010 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1011 * struct net_device *dev,
1012 * void *priv);
1013 * Callback to use for xmit over the accelerated station. This
1014 * is used in place of ndo_start_xmit on accelerated net
1015 * devices.
Jesse Gross5f352272014-12-23 22:37:26 -08001016 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1017 * struct net_device *dev
1018 * netdev_features_t features);
Tom Herbert04ffcb22014-10-14 15:19:06 -07001019 * Called by core transmit path to determine if device is capable of
Jesse Gross5f352272014-12-23 22:37:26 -08001020 * performing offload operations on a given packet. This is to give
1021 * the device an opportunity to implement any restrictions that cannot
1022 * be otherwise expressed by feature flags. The check is called with
1023 * the set of features that the stack has calculated and it returns
1024 * those the driver believes to be appropriate.
Jiri Pirko007f7902014-11-28 14:34:17 +01001025 *
1026 * int (*ndo_switch_parent_id_get)(struct net_device *dev,
1027 * struct netdev_phys_item_id *psid);
1028 * Called to get an ID of the switch chip this port is part of.
1029 * If driver implements this, it indicates that it represents a port
1030 * of a switch chip.
Scott Feldman38dcf352014-11-28 14:34:20 +01001031 * int (*ndo_switch_port_stp_update)(struct net_device *dev, u8 state);
1032 * Called to notify switch device port of bridge port STP
1033 * state change.
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001034 */
1035struct net_device_ops {
1036 int (*ndo_init)(struct net_device *dev);
1037 void (*ndo_uninit)(struct net_device *dev);
1038 int (*ndo_open)(struct net_device *dev);
1039 int (*ndo_stop)(struct net_device *dev);
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +00001040 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
Stephen Hemminger00829822008-11-20 20:14:53 -08001041 struct net_device *dev);
1042 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001043 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001044 void *accel_priv,
1045 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001046 void (*ndo_change_rx_flags)(struct net_device *dev,
1047 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001048 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001049 int (*ndo_set_mac_address)(struct net_device *dev,
1050 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001051 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001052 int (*ndo_do_ioctl)(struct net_device *dev,
1053 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001054 int (*ndo_set_config)(struct net_device *dev,
1055 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001056 int (*ndo_change_mtu)(struct net_device *dev,
1057 int new_mtu);
1058 int (*ndo_neigh_setup)(struct net_device *dev,
1059 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001060 void (*ndo_tx_timeout) (struct net_device *dev);
1061
Eric Dumazet28172732010-07-07 14:58:56 -07001062 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1063 struct rtnl_link_stats64 *storage);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001064 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1065
Jiri Pirko8e586132011-12-08 19:52:37 -05001066 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001067 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001068 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001069 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001070#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001071 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001072 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001073 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001074 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001075#endif
Cong Wange0d10952013-08-01 11:10:25 +08001076#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir8b80cda2013-07-10 17:13:26 +03001077 int (*ndo_busy_poll)(struct napi_struct *dev);
Eliezer Tamir06021292013-06-10 11:39:50 +03001078#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001079 int (*ndo_set_vf_mac)(struct net_device *dev,
1080 int queue, u8 *mac);
1081 int (*ndo_set_vf_vlan)(struct net_device *dev,
1082 int queue, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001083 int (*ndo_set_vf_rate)(struct net_device *dev,
1084 int vf, int min_tx_rate,
1085 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001086 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1087 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001088 int (*ndo_get_vf_config)(struct net_device *dev,
1089 int vf,
1090 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001091 int (*ndo_set_vf_link_state)(struct net_device *dev,
1092 int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -07001093 int (*ndo_set_vf_port)(struct net_device *dev,
1094 int vf,
1095 struct nlattr *port[]);
1096 int (*ndo_get_vf_port)(struct net_device *dev,
1097 int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +00001098 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001099#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001100 int (*ndo_fcoe_enable)(struct net_device *dev);
1101 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001102 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1103 u16 xid,
1104 struct scatterlist *sgl,
1105 unsigned int sgc);
1106 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1107 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001108 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1109 u16 xid,
1110 struct scatterlist *sgl,
1111 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001112 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1113 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001114#endif
1115
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001116#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001117#define NETDEV_FCOE_WWNN 0
1118#define NETDEV_FCOE_WWPN 1
1119 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1120 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001121#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001122
Ben Hutchingsc4454772011-01-19 11:03:53 +00001123#ifdef CONFIG_RFS_ACCEL
1124 int (*ndo_rx_flow_steer)(struct net_device *dev,
1125 const struct sk_buff *skb,
1126 u16 rxq_index,
1127 u32 flow_id);
1128#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001129 int (*ndo_add_slave)(struct net_device *dev,
1130 struct net_device *slave_dev);
1131 int (*ndo_del_slave)(struct net_device *dev,
1132 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001133 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1134 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001135 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001136 netdev_features_t features);
David Millerda6a8fa2011-07-25 00:01:38 +00001137 int (*ndo_neigh_construct)(struct neighbour *n);
David S. Miller447f2192011-12-19 15:04:41 -05001138 void (*ndo_neigh_destroy)(struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001139
1140 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001141 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001142 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001143 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001144 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001145 u16 flags);
1146 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001147 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001148 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001149 const unsigned char *addr,
1150 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001151 int (*ndo_fdb_dump)(struct sk_buff *skb,
1152 struct netlink_callback *cb,
1153 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001154 struct net_device *filter_dev,
John Fastabend77162022012-04-15 06:43:56 +00001155 int idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001156
1157 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001158 struct nlmsghdr *nlh,
1159 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001160 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1161 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001162 struct net_device *dev,
1163 u32 filter_mask);
Vlad Yasevich407af322013-02-13 12:00:12 +00001164 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001165 struct nlmsghdr *nlh,
1166 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001167 int (*ndo_change_carrier)(struct net_device *dev,
1168 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001169 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001170 struct netdev_phys_item_id *ppid);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001171 void (*ndo_add_vxlan_port)(struct net_device *dev,
1172 sa_family_t sa_family,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001173 __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001174 void (*ndo_del_vxlan_port)(struct net_device *dev,
1175 sa_family_t sa_family,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001176 __be16 port);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001177
1178 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1179 struct net_device *dev);
1180 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1181 void *priv);
1182
1183 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1184 struct net_device *dev,
1185 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001186 int (*ndo_get_lock_subclass)(struct net_device *dev);
Jesse Gross5f352272014-12-23 22:37:26 -08001187 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1188 struct net_device *dev,
1189 netdev_features_t features);
Jiri Pirko007f7902014-11-28 14:34:17 +01001190#ifdef CONFIG_NET_SWITCHDEV
1191 int (*ndo_switch_parent_id_get)(struct net_device *dev,
1192 struct netdev_phys_item_id *psid);
Scott Feldman38dcf352014-11-28 14:34:20 +01001193 int (*ndo_switch_port_stp_update)(struct net_device *dev,
1194 u8 state);
Jiri Pirko007f7902014-11-28 14:34:17 +01001195#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001196};
1197
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001198/**
1199 * enum net_device_priv_flags - &struct net_device priv_flags
1200 *
1201 * These are the &struct net_device, they are only set internally
1202 * by drivers and used in the kernel. These flags are invisible to
1203 * userspace, this means that the order of these flags can change
1204 * during any kernel release.
1205 *
1206 * You should have a pretty good reason to be extending these flags.
1207 *
1208 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1209 * @IFF_EBRIDGE: Ethernet bridging device
1210 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1211 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1212 * @IFF_MASTER_ALB: bonding master, balance-alb
1213 * @IFF_BONDING: bonding master or slave
1214 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1215 * @IFF_ISATAP: ISATAP interface (RFC4214)
1216 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1217 * @IFF_WAN_HDLC: WAN HDLC device
1218 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1219 * release skb->dst
1220 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1221 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1222 * @IFF_MACVLAN_PORT: device used as macvlan port
1223 * @IFF_BRIDGE_PORT: device used as bridge port
1224 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1225 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1226 * @IFF_UNICAST_FLT: Supports unicast filtering
1227 * @IFF_TEAM_PORT: device used as team port
1228 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1229 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1230 * change when it's running
1231 * @IFF_MACVLAN: Macvlan device
1232 */
1233enum netdev_priv_flags {
1234 IFF_802_1Q_VLAN = 1<<0,
1235 IFF_EBRIDGE = 1<<1,
1236 IFF_SLAVE_INACTIVE = 1<<2,
1237 IFF_MASTER_8023AD = 1<<3,
1238 IFF_MASTER_ALB = 1<<4,
1239 IFF_BONDING = 1<<5,
1240 IFF_SLAVE_NEEDARP = 1<<6,
1241 IFF_ISATAP = 1<<7,
1242 IFF_MASTER_ARPMON = 1<<8,
1243 IFF_WAN_HDLC = 1<<9,
1244 IFF_XMIT_DST_RELEASE = 1<<10,
1245 IFF_DONT_BRIDGE = 1<<11,
1246 IFF_DISABLE_NETPOLL = 1<<12,
1247 IFF_MACVLAN_PORT = 1<<13,
1248 IFF_BRIDGE_PORT = 1<<14,
1249 IFF_OVS_DATAPATH = 1<<15,
1250 IFF_TX_SKB_SHARING = 1<<16,
1251 IFF_UNICAST_FLT = 1<<17,
1252 IFF_TEAM_PORT = 1<<18,
1253 IFF_SUPP_NOFCS = 1<<19,
1254 IFF_LIVE_ADDR_CHANGE = 1<<20,
1255 IFF_MACVLAN = 1<<21,
Eric Dumazet02875872014-10-05 18:38:35 -07001256 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001257 IFF_IPVLAN_MASTER = 1<<23,
1258 IFF_IPVLAN_SLAVE = 1<<24,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001259};
1260
1261#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1262#define IFF_EBRIDGE IFF_EBRIDGE
1263#define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1264#define IFF_MASTER_8023AD IFF_MASTER_8023AD
1265#define IFF_MASTER_ALB IFF_MASTER_ALB
1266#define IFF_BONDING IFF_BONDING
1267#define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1268#define IFF_ISATAP IFF_ISATAP
1269#define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1270#define IFF_WAN_HDLC IFF_WAN_HDLC
1271#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1272#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1273#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1274#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1275#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1276#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1277#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1278#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1279#define IFF_TEAM_PORT IFF_TEAM_PORT
1280#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1281#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1282#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001283#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001284#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1285#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001286
Karoly Kemeny536721b2014-07-30 20:27:36 +02001287/**
1288 * struct net_device - The DEVICE structure.
1289 * Actually, this whole structure is a big mistake. It mixes I/O
1290 * data with strictly "high-level" data, and it has to know about
1291 * almost every data structure used in the INET module.
1292 *
1293 * @name: This is the first field of the "visible" part of this structure
1294 * (i.e. as seen by users in the "Space.c" file). It is the name
1295 * of the interface.
1296 *
1297 * @name_hlist: Device name hash chain, please keep it close to name[]
1298 * @ifalias: SNMP alias
1299 * @mem_end: Shared memory end
1300 * @mem_start: Shared memory start
1301 * @base_addr: Device I/O address
1302 * @irq: Device IRQ number
1303 *
1304 * @state: Generic network queuing layer state, see netdev_state_t
1305 * @dev_list: The global list of network devices
1306 * @napi_list: List entry, that is used for polling napi devices
1307 * @unreg_list: List entry, that is used, when we are unregistering the
1308 * device, see the function unregister_netdev
1309 * @close_list: List entry, that is used, when we are closing the device
1310 *
1311 * @adj_list: Directly linked devices, like slaves for bonding
1312 * @all_adj_list: All linked devices, *including* neighbours
1313 * @features: Currently active device features
1314 * @hw_features: User-changeable features
1315 *
1316 * @wanted_features: User-requested features
1317 * @vlan_features: Mask of features inheritable by VLAN devices
1318 *
1319 * @hw_enc_features: Mask of features inherited by encapsulating devices
1320 * This field indicates what encapsulation
1321 * offloads the hardware is capable of doing,
1322 * and drivers will need to set them appropriately.
1323 *
1324 * @mpls_features: Mask of features inheritable by MPLS
1325 *
1326 * @ifindex: interface index
1327 * @iflink: unique device identifier
1328 *
1329 * @stats: Statistics struct, which was left as a legacy, use
1330 * rtnl_link_stats64 instead
1331 *
1332 * @rx_dropped: Dropped packets by core network,
1333 * do not use this in drivers
1334 * @tx_dropped: Dropped packets by core network,
1335 * do not use this in drivers
1336 *
1337 * @carrier_changes: Stats to monitor carrier on<->off transitions
1338 *
1339 * @wireless_handlers: List of functions to handle Wireless Extensions,
1340 * instead of ioctl,
1341 * see <net/iw_handler.h> for details.
1342 * @wireless_data: Instance data managed by the core of wireless extensions
1343 *
1344 * @netdev_ops: Includes several pointers to callbacks,
1345 * if one wants to override the ndo_*() functions
1346 * @ethtool_ops: Management operations
1347 * @fwd_ops: Management operations
1348 * @header_ops: Includes callbacks for creating,parsing,rebuilding,etc
1349 * of Layer 2 headers.
1350 *
1351 * @flags: Interface flags (a la BSD)
1352 * @priv_flags: Like 'flags' but invisible to userspace,
1353 * see if.h for the definitions
1354 * @gflags: Global flags ( kept as legacy )
1355 * @padded: How much padding added by alloc_netdev()
1356 * @operstate: RFC2863 operstate
1357 * @link_mode: Mapping policy to operstate
1358 * @if_port: Selectable AUI, TP, ...
1359 * @dma: DMA channel
1360 * @mtu: Interface MTU value
1361 * @type: Interface hardware type
1362 * @hard_header_len: Hardware header length
1363 *
1364 * @needed_headroom: Extra headroom the hardware may need, but not in all
1365 * cases can this be guaranteed
1366 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1367 * cases can this be guaranteed. Some cases also use
1368 * LL_MAX_HEADER instead to allocate the skb
1369 *
1370 * interface address info:
1371 *
1372 * @perm_addr: Permanent hw address
1373 * @addr_assign_type: Hw address assignment type
1374 * @addr_len: Hardware address length
1375 * @neigh_priv_len; Used in neigh_alloc(),
1376 * initialized only in atm/clip.c
1377 * @dev_id: Used to differentiate devices that share
1378 * the same link layer address
1379 * @dev_port: Used to differentiate devices that share
1380 * the same function
1381 * @addr_list_lock: XXX: need comments on this one
1382 * @uc: unicast mac addresses
1383 * @mc: multicast mac addresses
1384 * @dev_addrs: list of device hw addresses
1385 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1386 * @uc_promisc: Counter, that indicates, that promiscuous mode
1387 * has been enabled due to the need to listen to
1388 * additional unicast addresses in a device that
1389 * does not implement ndo_set_rx_mode()
1390 * @promiscuity: Number of times, the NIC is told to work in
1391 * Promiscuous mode, if it becomes 0 the NIC will
1392 * exit from working in Promiscuous mode
1393 * @allmulti: Counter, enables or disables allmulticast mode
1394 *
1395 * @vlan_info: VLAN info
1396 * @dsa_ptr: dsa specific data
1397 * @tipc_ptr: TIPC specific data
1398 * @atalk_ptr: AppleTalk link
1399 * @ip_ptr: IPv4 specific data
1400 * @dn_ptr: DECnet specific data
1401 * @ip6_ptr: IPv6 specific data
1402 * @ax25_ptr: AX.25 specific data
1403 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1404 *
1405 * @last_rx: Time of last Rx
1406 * @dev_addr: Hw address (before bcast,
1407 * because most packets are unicast)
1408 *
1409 * @_rx: Array of RX queues
1410 * @num_rx_queues: Number of RX queues
1411 * allocated at register_netdev() time
1412 * @real_num_rx_queues: Number of RX queues currently active in device
1413 *
1414 * @rx_handler: handler for received packets
1415 * @rx_handler_data: XXX: need comments on this one
1416 * @ingress_queue: XXX: need comments on this one
1417 * @broadcast: hw bcast address
1418 *
1419 * @_tx: Array of TX queues
1420 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1421 * @real_num_tx_queues: Number of TX queues currently active in device
1422 * @qdisc: Root qdisc from userspace point of view
1423 * @tx_queue_len: Max frames per queue allowed
1424 * @tx_global_lock: XXX: need comments on this one
1425 *
1426 * @xps_maps: XXX: need comments on this one
1427 *
1428 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1429 * indexed by RX queue number. Assigned by driver.
1430 * This must only be set if the ndo_rx_flow_steer
1431 * operation is defined
1432 *
1433 * @trans_start: Time (in jiffies) of last Tx
1434 * @watchdog_timeo: Represents the timeout that is used by
1435 * the watchdog ( see dev_watchdog() )
1436 * @watchdog_timer: List of timers
1437 *
1438 * @pcpu_refcnt: Number of references to this device
1439 * @todo_list: Delayed register/unregister
1440 * @index_hlist: Device index hash chain
1441 * @link_watch_list: XXX: need comments on this one
1442 *
1443 * @reg_state: Register/unregister state machine
1444 * @dismantle: Device is going to be freed
1445 * @rtnl_link_state: This enum represents the phases of creating
1446 * a new link
1447 *
1448 * @destructor: Called from unregister,
1449 * can be used to call free_netdev
1450 * @npinfo: XXX: need comments on this one
1451 * @nd_net: Network namespace this network device is inside
1452 *
1453 * @ml_priv: Mid-layer private
1454 * @lstats: Loopback statistics
1455 * @tstats: Tunnel statistics
1456 * @dstats: Dummy statistics
1457 * @vstats: Virtual ethernet statistics
1458 *
1459 * @garp_port: GARP
1460 * @mrp_port: MRP
1461 *
1462 * @dev: Class/net/name entry
1463 * @sysfs_groups: Space for optional device, statistics and wireless
1464 * sysfs groups
1465 *
1466 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1467 * @rtnl_link_ops: Rtnl_link_ops
1468 *
1469 * @gso_max_size: Maximum size of generic segmentation offload
1470 * @gso_max_segs: Maximum number of segments that can be passed to the
1471 * NIC for GSO
Eric Dumazetfcbeb972014-10-05 10:11:27 -07001472 * @gso_min_segs: Minimum number of segments that can be passed to the
1473 * NIC for GSO
Karoly Kemeny536721b2014-07-30 20:27:36 +02001474 *
1475 * @dcbnl_ops: Data Center Bridging netlink ops
1476 * @num_tc: Number of traffic classes in the net device
1477 * @tc_to_txq: XXX: need comments on this one
1478 * @prio_tc_map XXX: need comments on this one
1479 *
1480 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1481 *
1482 * @priomap: XXX: need comments on this one
1483 * @phydev: Physical device may attach itself
1484 * for hardware timestamping
1485 *
1486 * @qdisc_tx_busylock: XXX: need comments on this one
1487 *
1488 * @group: The group, that the device belongs to
1489 * @pm_qos_req: Power Management QoS object
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 *
1491 * FIXME: cleanup struct net_device such that network protocol info
1492 * moves out.
1493 */
1494
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001495struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001497 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001498 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 /*
1500 * I/O specific fields
1501 * FIXME: Merge these and struct ifmap into one
1502 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001503 unsigned long mem_end;
1504 unsigned long mem_start;
1505 unsigned long base_addr;
1506 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507
1508 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001509 * Some hardware also needs these fields (state,dev_list,
1510 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 * part of the usual set specified in Space.c.
1512 */
1513
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 unsigned long state;
1515
Pavel Emelianov7562f872007-05-03 15:13:45 -07001516 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001517 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001518 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001519 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001520 struct list_head ptype_all;
1521 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001522
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001523 struct {
1524 struct list_head upper;
1525 struct list_head lower;
1526 } adj_list;
1527
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001528 struct {
1529 struct list_head upper;
1530 struct list_head lower;
1531 } all_adj_list;
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001532
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001533 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001534 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001535 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001536 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001537 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001538 netdev_features_t mpls_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001539
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 int ifindex;
1541 int iflink;
1542
Rusty Russellc45d2862007-03-28 14:29:08 -07001543 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001544
Eric Dumazet015f0682014-03-27 08:45:56 -07001545 atomic_long_t rx_dropped;
1546 atomic_long_t tx_dropped;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547
david decotigny2d3b4792014-03-29 09:48:35 -07001548 atomic_t carrier_changes;
1549
Johannes Bergb86e0282007-04-26 20:48:23 -07001550#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 const struct iw_handler_def * wireless_handlers;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 struct iw_public_data * wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001553#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001554 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001555 const struct ethtool_ops *ethtool_ops;
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001556 const struct forwarding_accel_ops *fwd_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001558 const struct header_ops *header_ops;
1559
Karoly Kemeny536721b2014-07-30 20:27:36 +02001560 unsigned int flags;
1561 unsigned int priv_flags;
1562
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001564 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565
Karoly Kemeny536721b2014-07-30 20:27:36 +02001566 unsigned char operstate;
1567 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001568
Karoly Kemeny536721b2014-07-30 20:27:36 +02001569 unsigned char if_port;
1570 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001571
Karoly Kemeny536721b2014-07-30 20:27:36 +02001572 unsigned int mtu;
1573 unsigned short type;
1574 unsigned short hard_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575
Johannes Bergf5184d22008-05-12 20:48:31 -07001576 unsigned short needed_headroom;
1577 unsigned short needed_tailroom;
1578
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001580 unsigned char perm_addr[MAX_ADDR_LEN];
1581 unsigned char addr_assign_type;
1582 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001583 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001584 unsigned short dev_id;
1585 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001586 spinlock_t addr_list_lock;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001587 struct netdev_hw_addr_list uc;
1588 struct netdev_hw_addr_list mc;
1589 struct netdev_hw_addr_list dev_addrs;
1590
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001591#ifdef CONFIG_SYSFS
1592 struct kset *queues_kset;
1593#endif
1594
Tom Gundersen685343f2014-07-14 16:37:22 +02001595 unsigned char name_assign_type;
1596
Joe Perches2d348d12011-07-25 16:17:35 -07001597 bool uc_promisc;
Wang Chen9d45abe2008-06-17 21:12:48 -07001598 unsigned int promiscuity;
1599 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601
1602 /* Protocol specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001603
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001604#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001605 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001606#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001607#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001608 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001609#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001610#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001611 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001612#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001613 void *atalk_ptr;
1614 struct in_device __rcu *ip_ptr;
1615 struct dn_dev __rcu *dn_ptr;
1616 struct inet6_dev __rcu *ip6_ptr;
1617 void *ax25_ptr;
1618 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001619 struct wpan_dev *ieee802154_ptr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001621/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001622 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001623 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001624 unsigned long last_rx;
Eric Dumazet4dc89132010-08-31 07:40:16 +00001625
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001626 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001627 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001628
Tom Herbert0a9627f2010-03-16 08:03:29 +00001629
Michael Daltona953be52014-01-16 22:23:28 -08001630#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001631 struct netdev_rx_queue *_rx;
1632
Tom Herbert0a9627f2010-03-16 08:03:29 +00001633 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001634 unsigned int real_num_rx_queues;
Ben Hutchingsc4454772011-01-19 11:03:53 +00001635
Eric Dumazetdf334542010-03-24 19:13:54 +00001636#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001637
Eric Dumazet3b47d302014-11-06 21:09:44 -08001638 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001639 rx_handler_func_t __rcu *rx_handler;
1640 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001641
Eric Dumazet24824a02010-10-02 06:11:55 +00001642 struct netdev_queue __rcu *ingress_queue;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001643 unsigned char broadcast[MAX_ADDR_LEN];
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001644
Eric Dumazetcd135392010-09-16 02:58:13 +00001645
1646/*
1647 * Cache lines mostly used on transmit path
1648 */
David S. Millere8a04642008-07-17 00:34:19 -07001649 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1650 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001651 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001652 struct Qdisc *qdisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001653 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001654 spinlock_t tx_global_lock;
Eric Dumazetcd135392010-09-16 02:58:13 +00001655
Tom Herbertbf264142010-11-26 08:36:09 +00001656#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001657 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001658#endif
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001659#ifdef CONFIG_RFS_ACCEL
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001660 struct cpu_rmap *rx_cpu_rmap;
1661#endif
Tom Herbert1d24eb42010-11-21 13:17:27 +00001662
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001663 /* These may be needed for future network-power-down code. */
Eric Dumazet9d214932009-05-17 20:55:16 -07001664
1665 /*
1666 * trans_start here is expensive for high speed devices on SMP,
1667 * please use netdev_queue->trans_start instead.
1668 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001669 unsigned long trans_start;
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001670
Karoly Kemeny536721b2014-07-30 20:27:36 +02001671 int watchdog_timeo;
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001672 struct timer_list watchdog_timer;
1673
Eric Dumazet29b44332010-10-11 10:22:12 +00001674 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676
Karoly Kemeny536721b2014-07-30 20:27:36 +02001677 struct hlist_node index_hlist;
Eric Dumazete014deb2009-11-17 05:59:21 +00001678 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001679
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001681 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682 NETREG_UNREGISTERING, /* called unregister_netdevice */
1683 NETREG_UNREGISTERED, /* completed unregister todo */
1684 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001685 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001686 } reg_state:8;
1687
Karoly Kemeny536721b2014-07-30 20:27:36 +02001688 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001689
1690 enum {
1691 RTNL_LINK_INITIALIZED,
1692 RTNL_LINK_INITIALIZING,
1693 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001695 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001698 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699#endif
David S. Millereae792b2008-07-15 03:03:33 -07001700
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001701#ifdef CONFIG_NET_NS
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001702 struct net *nd_net;
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001703#endif
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001704
David S. Miller49517042008-05-12 03:29:11 -07001705 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001706 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001707 void *ml_priv;
1708 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001709 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001710 struct pcpu_dstats __percpu *dstats;
1711 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001712 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001713
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001714 struct garp_port __rcu *garp_port;
David Wardfebf0182013-02-08 17:17:06 +00001715 struct mrp_port __rcu *mrp_port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716
Karoly Kemeny536721b2014-07-30 20:27:36 +02001717 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001718 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001719 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001720
Patrick McHardy38f7b872007-06-13 12:03:51 -07001721 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001722
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001723 /* for setting kernel sock attribute on TCP connection setup */
1724#define GSO_MAX_SIZE 65536
1725 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001726#define GSO_MAX_SEGS 65535
1727 u16 gso_max_segs;
Eric Dumazetfcbeb972014-10-05 10:11:27 -07001728 u16 gso_min_segs;
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001729#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001730 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001731#endif
John Fastabend4f57c082011-01-17 08:06:04 +00001732 u8 num_tc;
1733 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1734 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001735
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001736#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001737 unsigned int fcoe_ddp_xid;
1738#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001739#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001740 struct netprio_map __rcu *priomap;
1741#endif
Richard Cochranc1f19b52010-07-17 08:49:36 +00001742 struct phy_device *phydev;
Eric Dumazet23d3b8b2012-09-05 01:02:56 +00001743 struct lock_class_key *qdisc_tx_busylock;
Vlad Dogarucbda10f2011-01-13 23:38:30 +00001744 int group;
Eric Dumazet91364612012-06-11 06:36:13 +00001745 struct pm_qos_request pm_qos_req;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001747#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748
1749#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750
David S. Millere8a04642008-07-17 00:34:19 -07001751static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001752int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1753{
1754 return dev->prio_tc_map[prio & TC_BITMASK];
1755}
1756
1757static inline
1758int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1759{
1760 if (tc >= dev->num_tc)
1761 return -EINVAL;
1762
1763 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1764 return 0;
1765}
1766
1767static inline
1768void netdev_reset_tc(struct net_device *dev)
1769{
1770 dev->num_tc = 0;
1771 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1772 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1773}
1774
1775static inline
1776int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1777{
1778 if (tc >= dev->num_tc)
1779 return -EINVAL;
1780
1781 dev->tc_to_txq[tc].count = count;
1782 dev->tc_to_txq[tc].offset = offset;
1783 return 0;
1784}
1785
1786static inline
1787int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1788{
1789 if (num_tc > TC_MAX_QUEUE)
1790 return -EINVAL;
1791
1792 dev->num_tc = num_tc;
1793 return 0;
1794}
1795
1796static inline
1797int netdev_get_num_tc(struct net_device *dev)
1798{
1799 return dev->num_tc;
1800}
1801
1802static inline
David S. Millere8a04642008-07-17 00:34:19 -07001803struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1804 unsigned int index)
1805{
1806 return &dev->_tx[index];
1807}
1808
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001809static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1810 const struct sk_buff *skb)
1811{
1812 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1813}
1814
David S. Millere8a04642008-07-17 00:34:19 -07001815static inline void netdev_for_each_tx_queue(struct net_device *dev,
1816 void (*f)(struct net_device *,
1817 struct netdev_queue *,
1818 void *),
1819 void *arg)
1820{
1821 unsigned int i;
1822
1823 for (i = 0; i < dev->num_tx_queues; i++)
1824 f(dev, &dev->_tx[i], arg);
1825}
1826
Joe Perchesf629d202013-09-26 14:48:15 -07001827struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001828 struct sk_buff *skb,
1829 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001830
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001831/*
1832 * Net namespace inlines
1833 */
1834static inline
1835struct net *dev_net(const struct net_device *dev)
1836{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001837 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001838}
1839
1840static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07001841void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001842{
1843#ifdef CONFIG_NET_NS
Denis V. Lunevf3005d72008-04-16 02:02:18 -07001844 release_net(dev->nd_net);
1845 dev->nd_net = hold_net(net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001846#endif
1847}
1848
Florian Fainelli3e8a72d2014-08-27 17:04:46 -07001849static inline bool netdev_uses_dsa(struct net_device *dev)
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00001850{
Alexander Duyck3fc88672014-09-12 23:12:46 -07001851#if IS_ENABLED(CONFIG_NET_DSA)
Florian Fainelli5aed85c2014-08-27 17:04:52 -07001852 if (dev->dsa_ptr != NULL)
1853 return dsa_uses_tagged_protocol(dev->dsa_ptr);
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00001854#endif
Florian Fainelli5aed85c2014-08-27 17:04:52 -07001855 return false;
Lennert Buytenhek396138f2008-10-07 13:46:07 +00001856}
1857
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001858/**
1859 * netdev_priv - access network device private data
1860 * @dev: network device
1861 *
1862 * Get network device private data
1863 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07001864static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00001866 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867}
1868
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869/* Set the sysfs physical device reference for the network logical device
1870 * if set prior to registration will cause a symlink during initialization.
1871 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001872#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873
Marcel Holtmann384912e2009-08-31 21:08:19 +00001874/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02001875 * fine-grained identification of different network device types. For
Marcel Holtmann384912e2009-08-31 21:08:19 +00001876 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1877 */
1878#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1879
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00001880/* Default NAPI poll() weight
1881 * Device drivers are strongly advised to not use bigger value
1882 */
1883#define NAPI_POLL_WEIGHT 64
1884
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07001885/**
1886 * netif_napi_add - initialize a napi context
1887 * @dev: network device
1888 * @napi: napi context
1889 * @poll: polling function
1890 * @weight: default weight
1891 *
1892 * netif_napi_add() must be used to initialize a napi context prior to calling
1893 * *any* of the other napi related functions.
1894 */
Herbert Xud565b0a2008-12-15 23:38:52 -08001895void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1896 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001897
Alexander Duyckd8156532008-07-08 15:13:05 -07001898/**
1899 * netif_napi_del - remove a napi context
1900 * @napi: napi context
1901 *
1902 * netif_napi_del() removes a napi context from the network device napi list
1903 */
Herbert Xud565b0a2008-12-15 23:38:52 -08001904void netif_napi_del(struct napi_struct *napi);
1905
1906struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00001907 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1908 void *frag0;
1909
Herbert Xu74895942009-05-26 18:50:27 +00001910 /* Length of frag0. */
1911 unsigned int frag0_len;
1912
Herbert Xu86911732009-01-29 14:19:50 +00001913 /* This indicates where we are processing relative to skb->data. */
1914 int data_offset;
1915
Herbert Xud565b0a2008-12-15 23:38:52 -08001916 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08001917 u16 flush;
1918
1919 /* Save the IP ID here and check when we get to the transport layer */
1920 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08001921
1922 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00001923 u16 count;
1924
1925 /* This is non-zero if the packet may be of the same flow. */
1926 u8 same_flow;
Herbert Xu5d38a072009-01-04 16:13:40 -08001927
1928 /* Free the skb? */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00001929 u8 free;
Eric Dumazetd7e88832012-04-30 08:10:34 +00001930#define NAPI_GRO_FREE 1
1931#define NAPI_GRO_FREE_STOLEN_HEAD 2
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00001932
1933 /* jiffies when first packet was created/queued */
1934 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02001935
Tom Herbertafe93322014-09-17 12:25:57 -07001936 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02001937 u16 proto;
1938
1939 /* Used in udp_gro_receive */
Tom Herbert573e8fc2014-08-22 13:33:47 -07001940 u8 udp_mark:1;
1941
1942 /* GRO checksum is valid */
1943 u8 csum_valid:1;
1944
Tom Herbert662880f2014-08-27 21:26:56 -07001945 /* Number of checksums via CHECKSUM_UNNECESSARY */
1946 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00001947
Tom Herbertefc98d02014-10-03 15:48:08 -07001948 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1949 u8 is_ipv6:1;
1950
Jerry Chubf5a7552014-01-07 10:23:19 -08001951 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1952 __wsum csum;
1953
Eric Dumazetc3c7c252012-12-06 13:54:59 +00001954 /* used in skb_gro_receive() slow path */
1955 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08001956};
1957
1958#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07001959
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07001961 __be16 type; /* This is really htons(ether_type). */
1962 struct net_device *dev; /* NULL is wildcarded here */
1963 int (*func) (struct sk_buff *,
1964 struct net_device *,
1965 struct packet_type *,
1966 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00001967 bool (*id_match)(struct packet_type *ptype,
1968 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969 void *af_packet_priv;
1970 struct list_head list;
1971};
1972
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001973struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1975 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08001977 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08001978 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001979};
1980
1981struct packet_offload {
1982 __be16 type; /* This is really htons(ether_type). */
1983 struct offload_callbacks callbacks;
1984 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985};
1986
Tom Herberta2b12f32015-01-12 17:00:37 -08001987struct udp_offload;
1988
1989struct udp_offload_callbacks {
1990 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1991 struct sk_buff *skb,
1992 struct udp_offload *uoff);
1993 int (*gro_complete)(struct sk_buff *skb,
1994 int nhoff,
1995 struct udp_offload *uoff);
1996};
1997
Or Gerlitzb582ef02014-01-20 13:59:19 +02001998struct udp_offload {
1999 __be16 port;
Tom Herbertafe93322014-09-17 12:25:57 -07002000 u8 ipproto;
Tom Herberta2b12f32015-01-12 17:00:37 -08002001 struct udp_offload_callbacks callbacks;
Or Gerlitzb582ef02014-01-20 13:59:19 +02002002};
2003
Li RongQing8f849852014-01-04 13:57:59 +08002004/* often modified stats are per cpu, other are shared (netdev->stats) */
2005struct pcpu_sw_netstats {
2006 u64 rx_packets;
2007 u64 rx_bytes;
2008 u64 tx_packets;
2009 u64 tx_bytes;
2010 struct u64_stats_sync syncp;
2011};
2012
WANG Cong1c213bd2014-02-13 11:46:28 -08002013#define netdev_alloc_pcpu_stats(type) \
2014({ \
stephen hemminger693350c2014-03-10 09:41:46 -07002015 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
WANG Cong1c213bd2014-02-13 11:46:28 -08002016 if (pcpu_stats) { \
2017 int i; \
2018 for_each_possible_cpu(i) { \
2019 typeof(type) *stat; \
2020 stat = per_cpu_ptr(pcpu_stats, i); \
2021 u64_stats_init(&stat->syncp); \
2022 } \
2023 } \
2024 pcpu_stats; \
2025})
2026
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027#include <linux/notifier.h>
2028
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002029/* netdevice notifier chain. Please remember to update the rtnetlink
2030 * notification exclusion list in rtnetlink_event() when adding new
2031 * types.
2032 */
2033#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2034#define NETDEV_DOWN 0x0002
2035#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2036 detected a hardware crash and restarted
2037 - we can use this eg to kick tcp sessions
2038 once done */
2039#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2040#define NETDEV_REGISTER 0x0005
2041#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002042#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002043#define NETDEV_CHANGEADDR 0x0008
2044#define NETDEV_GOING_DOWN 0x0009
2045#define NETDEV_CHANGENAME 0x000A
2046#define NETDEV_FEAT_CHANGE 0x000B
2047#define NETDEV_BONDING_FAILOVER 0x000C
2048#define NETDEV_PRE_UP 0x000D
2049#define NETDEV_PRE_TYPE_CHANGE 0x000E
2050#define NETDEV_POST_TYPE_CHANGE 0x000F
2051#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002052#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002053#define NETDEV_RELEASE 0x0012
2054#define NETDEV_NOTIFY_PEERS 0x0013
2055#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002056#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002057#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002058#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002059#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002060#define NETDEV_BONDING_INFO 0x0019
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002061
Joe Perchesf629d202013-09-26 14:48:15 -07002062int register_netdevice_notifier(struct notifier_block *nb);
2063int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002064
2065struct netdev_notifier_info {
2066 struct net_device *dev;
2067};
2068
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002069struct netdev_notifier_change_info {
2070 struct netdev_notifier_info info; /* must be first */
2071 unsigned int flags_changed;
2072};
2073
Cong Wang75538c22013-05-29 11:30:50 +08002074static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2075 struct net_device *dev)
2076{
2077 info->dev = dev;
2078}
2079
Jiri Pirko351638e2013-05-28 01:30:21 +00002080static inline struct net_device *
2081netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2082{
2083 return info->dev;
2084}
2085
Joe Perchesf629d202013-09-26 14:48:15 -07002086int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002087
2088
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089extern rwlock_t dev_base_lock; /* Device list lock */
2090
Eric W. Biederman881d9662007-09-17 11:56:21 -07002091#define for_each_netdev(net, d) \
2092 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002093#define for_each_netdev_reverse(net, d) \
2094 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002095#define for_each_netdev_rcu(net, d) \
2096 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002097#define for_each_netdev_safe(net, d, n) \
2098 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2099#define for_each_netdev_continue(net, d) \
2100 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002101#define for_each_netdev_continue_rcu(net, d) \
2102 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002103#define for_each_netdev_in_bond_rcu(bond, slave) \
2104 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002105 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002106#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2107
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002108static inline struct net_device *next_net_device(struct net_device *dev)
2109{
2110 struct list_head *lh;
2111 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002112
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002113 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002114 lh = dev->dev_list.next;
2115 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2116}
2117
Eric Dumazetce81b762009-11-11 17:34:30 +00002118static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2119{
2120 struct list_head *lh;
2121 struct net *net;
2122
2123 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002124 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002125 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2126}
2127
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002128static inline struct net_device *first_net_device(struct net *net)
2129{
2130 return list_empty(&net->dev_base_head) ? NULL :
2131 net_device_entry(net->dev_base_head.next);
2132}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002133
Eric Dumazetccf43432011-01-26 18:08:02 +00002134static inline struct net_device *first_net_device_rcu(struct net *net)
2135{
2136 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2137
2138 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2139}
2140
Joe Perchesf629d202013-09-26 14:48:15 -07002141int netdev_boot_setup_check(struct net_device *dev);
2142unsigned long netdev_boot_base(const char *prefix, int unit);
2143struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2144 const char *hwaddr);
2145struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2146struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2147void dev_add_pack(struct packet_type *pt);
2148void dev_remove_pack(struct packet_type *pt);
2149void __dev_remove_pack(struct packet_type *pt);
2150void dev_add_offload(struct packet_offload *po);
2151void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152
WANG Cong6c555492014-09-11 15:35:09 -07002153struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2154 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002155struct net_device *dev_get_by_name(struct net *net, const char *name);
2156struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2157struct net_device *__dev_get_by_name(struct net *net, const char *name);
2158int dev_alloc_name(struct net_device *dev, const char *name);
2159int dev_open(struct net_device *dev);
2160int dev_close(struct net_device *dev);
2161void dev_disable_lro(struct net_device *dev);
2162int dev_loopback_xmit(struct sk_buff *newskb);
2163int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002164int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002165int register_netdevice(struct net_device *dev);
2166void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2167void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002168static inline void unregister_netdevice(struct net_device *dev)
2169{
2170 unregister_netdevice_queue(dev, NULL);
2171}
2172
Joe Perchesf629d202013-09-26 14:48:15 -07002173int netdev_refcnt_read(const struct net_device *dev);
2174void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002175void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002176void synchronize_net(void);
2177int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002178
Joe Perchesf629d202013-09-26 14:48:15 -07002179struct net_device *dev_get_by_index(struct net *net, int ifindex);
2180struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2181struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2182int netdev_get_name(struct net *net, char *name, int ifindex);
2183int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002184int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002185
2186static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2187{
2188 return NAPI_GRO_CB(skb)->data_offset;
2189}
2190
2191static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2192{
2193 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2194}
2195
2196static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2197{
2198 NAPI_GRO_CB(skb)->data_offset += len;
2199}
2200
Herbert Xua5b1cf22009-05-26 18:50:28 +00002201static inline void *skb_gro_header_fast(struct sk_buff *skb,
2202 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002203{
Herbert Xu78a478d2009-05-26 18:50:21 +00002204 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002205}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206
Herbert Xua5b1cf22009-05-26 18:50:28 +00002207static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2208{
2209 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2210}
2211
2212static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2213 unsigned int offset)
2214{
Herbert Xu17dd7592011-07-27 06:16:28 -07002215 if (!pskb_may_pull(skb, hlen))
2216 return NULL;
2217
Herbert Xua5b1cf22009-05-26 18:50:28 +00002218 NAPI_GRO_CB(skb)->frag0 = NULL;
2219 NAPI_GRO_CB(skb)->frag0_len = 0;
Herbert Xu17dd7592011-07-27 06:16:28 -07002220 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002221}
2222
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002223static inline void *skb_gro_network_header(struct sk_buff *skb)
2224{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002225 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2226 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002227}
2228
Jerry Chubf5a7552014-01-07 10:23:19 -08002229static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2230 const void *start, unsigned int len)
2231{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002232 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002233 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2234 csum_partial(start, len, 0));
2235}
2236
Tom Herbert573e8fc2014-08-22 13:33:47 -07002237/* GRO checksum functions. These are logical equivalents of the normal
2238 * checksum functions (in skbuff.h) except that they operate on the GRO
2239 * offsets and fields in sk_buff.
2240 */
2241
2242__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2243
2244static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2245 bool zero_okay,
2246 __sum16 check)
2247{
2248 return (skb->ip_summed != CHECKSUM_PARTIAL &&
Tom Herbert662880f2014-08-27 21:26:56 -07002249 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002250 (!zero_okay || check));
2251}
2252
2253static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2254 __wsum psum)
2255{
2256 if (NAPI_GRO_CB(skb)->csum_valid &&
2257 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2258 return 0;
2259
2260 NAPI_GRO_CB(skb)->csum = psum;
2261
2262 return __skb_gro_checksum_complete(skb);
2263}
2264
Tom Herbert573e8fc2014-08-22 13:33:47 -07002265static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2266{
Tom Herbert662880f2014-08-27 21:26:56 -07002267 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2268 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2269 NAPI_GRO_CB(skb)->csum_cnt--;
2270 } else {
2271 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2272 * verified a new top level checksum or an encapsulated one
2273 * during GRO. This saves work if we fallback to normal path.
2274 */
2275 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002276 }
2277}
2278
2279#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2280 compute_pseudo) \
2281({ \
2282 __sum16 __ret = 0; \
2283 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2284 __ret = __skb_gro_checksum_validate_complete(skb, \
2285 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002286 if (__ret) \
2287 __skb_mark_checksum_bad(skb); \
2288 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002289 skb_gro_incr_csum_unnecessary(skb); \
2290 __ret; \
2291})
2292
2293#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2294 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2295
2296#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2297 compute_pseudo) \
2298 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2299
2300#define skb_gro_checksum_simple_validate(skb) \
2301 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2302
Tom Herbertd96535a2014-08-31 15:12:42 -07002303static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2304{
2305 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2306 !NAPI_GRO_CB(skb)->csum_valid);
2307}
2308
2309static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2310 __sum16 check, __wsum pseudo)
2311{
2312 NAPI_GRO_CB(skb)->csum = ~pseudo;
2313 NAPI_GRO_CB(skb)->csum_valid = 1;
2314}
2315
2316#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2317do { \
2318 if (__skb_gro_checksum_convert_check(skb)) \
2319 __skb_gro_checksum_convert(skb, check, \
2320 compute_pseudo(skb, proto)); \
2321} while (0)
2322
Tom Herbertdcdc8992015-02-02 16:07:34 -08002323static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2324 int start, int offset)
2325{
2326 __wsum delta;
2327
2328 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2329
2330 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2331
2332 /* Adjust skb->csum since we changed the packet */
2333 skb->csum = csum_add(skb->csum, delta);
2334 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2335}
2336
2337
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002338static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2339 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002340 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002341 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002342{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002343 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002344 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002345
2346 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002347}
2348
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002349static inline int dev_parse_header(const struct sk_buff *skb,
2350 unsigned char *haddr)
2351{
2352 const struct net_device *dev = skb->dev;
2353
Patrick McHardy1b833362007-10-18 05:09:28 -07002354 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002355 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002356 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002357}
2358
David S. Miller22053692013-12-31 16:23:35 -05002359static inline int dev_rebuild_header(struct sk_buff *skb)
2360{
2361 const struct net_device *dev = skb->dev;
2362
2363 if (!dev->header_ops || !dev->header_ops->rebuild)
2364 return 0;
2365 return dev->header_ops->rebuild(skb);
2366}
2367
Linus Torvalds1da177e2005-04-16 15:20:36 -07002368typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002369int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002370static inline int unregister_gifconf(unsigned int family)
2371{
2372 return register_gifconf(family, NULL);
2373}
2374
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002375#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002376#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002377struct sd_flow_limit {
2378 u64 count;
2379 unsigned int num_buckets;
2380 unsigned int history_head;
2381 u16 history[FLOW_LIMIT_HISTORY];
2382 u8 buckets[];
2383};
2384
2385extern int netdev_flow_limit_table_len;
2386#endif /* CONFIG_NET_FLOW_LIMIT */
2387
Linus Torvalds1da177e2005-04-16 15:20:36 -07002388/*
Eric Dumazet88751272010-04-19 05:07:33 +00002389 * Incoming packets are placed on per-cpu queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002391struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002392 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002393 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002394
Changli Gaodee42872010-05-02 05:42:16 +00002395 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002396 unsigned int processed;
2397 unsigned int time_squeeze;
2398 unsigned int cpu_collision;
2399 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002400#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002401 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002402#endif
2403#ifdef CONFIG_NET_FLOW_LIMIT
2404 struct sd_flow_limit __rcu *flow_limit;
2405#endif
2406 struct Qdisc *output_queue;
2407 struct Qdisc **output_queue_tailp;
2408 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002409
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002410#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002411 /* Elements below can be accessed between CPUs for RPS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002412 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002413 struct softnet_data *rps_ipi_next;
2414 unsigned int cpu;
Tom Herbertfec5e652010-04-16 16:01:27 -07002415 unsigned int input_queue_head;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002416 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002417#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002418 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002419 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002420 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002421
Linus Torvalds1da177e2005-04-16 15:20:36 -07002422};
2423
Tom Herbert76cc8b12010-05-20 18:37:59 +00002424static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002425{
2426#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002427 sd->input_queue_head++;
2428#endif
2429}
2430
2431static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2432 unsigned int *qtail)
2433{
2434#ifdef CONFIG_RPS
2435 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002436#endif
2437}
2438
Tom Herbert0a9627f2010-03-16 08:03:29 +00002439DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440
Joe Perchesf629d202013-09-26 14:48:15 -07002441void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002442void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002444static inline void netif_tx_schedule_all(struct net_device *dev)
2445{
2446 unsigned int i;
2447
2448 for (i = 0; i < dev->num_tx_queues; i++)
2449 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2450}
2451
Dave Jonesd29f7492008-07-22 14:09:06 -07002452static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2453{
Tom Herbert734664982011-11-28 16:32:44 +00002454 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002455}
2456
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002457/**
2458 * netif_start_queue - allow transmit
2459 * @dev: network device
2460 *
2461 * Allow upper layers to call the device hard_start_xmit routine.
2462 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002463static inline void netif_start_queue(struct net_device *dev)
2464{
David S. Millere8a04642008-07-17 00:34:19 -07002465 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002466}
2467
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002468static inline void netif_tx_start_all_queues(struct net_device *dev)
2469{
2470 unsigned int i;
2471
2472 for (i = 0; i < dev->num_tx_queues; i++) {
2473 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2474 netif_tx_start_queue(txq);
2475 }
2476}
2477
John Fastabend46e5da40a2014-09-12 20:04:52 -07002478void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002479
Dave Jonesd29f7492008-07-22 14:09:06 -07002480/**
2481 * netif_wake_queue - restart transmit
2482 * @dev: network device
2483 *
2484 * Allow upper layers to call the device hard_start_xmit routine.
2485 * Used for flow control when transmit resources are available.
2486 */
David S. Miller79d16382008-07-08 23:14:46 -07002487static inline void netif_wake_queue(struct net_device *dev)
2488{
David S. Millere8a04642008-07-17 00:34:19 -07002489 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002490}
2491
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002492static inline void netif_tx_wake_all_queues(struct net_device *dev)
2493{
2494 unsigned int i;
2495
2496 for (i = 0; i < dev->num_tx_queues; i++) {
2497 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2498 netif_tx_wake_queue(txq);
2499 }
2500}
2501
Dave Jonesd29f7492008-07-22 14:09:06 -07002502static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2503{
Guillaume Chazarain18543a62010-11-06 06:39:32 +00002504 if (WARN_ON(!dev_queue)) {
Joe Perches256ee432011-03-01 07:06:12 +00002505 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
Guillaume Chazarain18543a62010-11-06 06:39:32 +00002506 return;
2507 }
Tom Herbert734664982011-11-28 16:32:44 +00002508 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002509}
2510
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002511/**
2512 * netif_stop_queue - stop transmitted packets
2513 * @dev: network device
2514 *
2515 * Stop upper layers calling the device hard_start_xmit routine.
2516 * Used for flow control when transmit resources are unavailable.
2517 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002518static inline void netif_stop_queue(struct net_device *dev)
2519{
David S. Millere8a04642008-07-17 00:34:19 -07002520 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002521}
2522
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002523static inline void netif_tx_stop_all_queues(struct net_device *dev)
2524{
2525 unsigned int i;
2526
2527 for (i = 0; i < dev->num_tx_queues; i++) {
2528 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2529 netif_tx_stop_queue(txq);
2530 }
2531}
2532
David S. Miller4d295152012-03-07 21:02:35 -05002533static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002534{
Tom Herbert734664982011-11-28 16:32:44 +00002535 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002536}
2537
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002538/**
2539 * netif_queue_stopped - test if transmit queue is flowblocked
2540 * @dev: network device
2541 *
2542 * Test if transmit queue on device is currently unable to send.
2543 */
David S. Miller4d295152012-03-07 21:02:35 -05002544static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002545{
David S. Millere8a04642008-07-17 00:34:19 -07002546 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002547}
2548
David S. Miller4d295152012-03-07 21:02:35 -05002549static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002550{
Tom Herbert734664982011-11-28 16:32:44 +00002551 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2552}
2553
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002554static inline bool
2555netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002556{
2557 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2558}
2559
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002560static inline bool
2561netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2562{
2563 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2564}
2565
Eric Dumazet53511452014-10-08 08:19:27 -07002566/**
2567 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2568 * @dev_queue: pointer to transmit queue
2569 *
2570 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2571 * to give appropriate hint to the cpu.
2572 */
2573static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2574{
2575#ifdef CONFIG_BQL
2576 prefetchw(&dev_queue->dql.num_queued);
2577#endif
2578}
2579
2580/**
2581 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2582 * @dev_queue: pointer to transmit queue
2583 *
2584 * BQL enabled drivers might use this helper in their TX completion path,
2585 * to give appropriate hint to the cpu.
2586 */
2587static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2588{
2589#ifdef CONFIG_BQL
2590 prefetchw(&dev_queue->dql.limit);
2591#endif
2592}
2593
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002594static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2595 unsigned int bytes)
2596{
Tom Herbert114cf582011-11-28 16:33:09 +00002597#ifdef CONFIG_BQL
2598 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002599
2600 if (likely(dql_avail(&dev_queue->dql) >= 0))
2601 return;
2602
2603 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2604
2605 /*
2606 * The XOFF flag must be set before checking the dql_avail below,
2607 * because in netdev_tx_completed_queue we update the dql_completed
2608 * before checking the XOFF flag.
2609 */
2610 smp_mb();
2611
2612 /* check again in case another CPU has just made room avail */
2613 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2614 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002615#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002616}
2617
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002618/**
2619 * netdev_sent_queue - report the number of bytes queued to hardware
2620 * @dev: network device
2621 * @bytes: number of bytes queued to the hardware device queue
2622 *
2623 * Report the number of bytes queued for sending/completion to the network
2624 * device hardware queue. @bytes should be a good approximation and should
2625 * exactly match netdev_completed_queue() @bytes
2626 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002627static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2628{
2629 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2630}
2631
2632static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002633 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002634{
Tom Herbert114cf582011-11-28 16:33:09 +00002635#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002636 if (unlikely(!bytes))
2637 return;
2638
2639 dql_completed(&dev_queue->dql, bytes);
2640
2641 /*
2642 * Without the memory barrier there is a small possiblity that
2643 * netdev_tx_sent_queue will miss the update and cause the queue to
2644 * be stopped forever
2645 */
2646 smp_mb();
2647
2648 if (dql_avail(&dev_queue->dql) < 0)
2649 return;
2650
2651 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2652 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00002653#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002654}
2655
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002656/**
2657 * netdev_completed_queue - report bytes and packets completed by device
2658 * @dev: network device
2659 * @pkts: actual number of packets sent over the medium
2660 * @bytes: actual number of bytes sent over the medium
2661 *
2662 * Report the number of bytes and packets transmitted by the network device
2663 * hardware queue over the physical medium, @bytes must exactly match the
2664 * @bytes amount passed to netdev_sent_queue()
2665 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002666static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00002667 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002668{
2669 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2670}
2671
2672static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2673{
Tom Herbert114cf582011-11-28 16:33:09 +00002674#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00002675 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002676 dql_reset(&q->dql);
2677#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002678}
2679
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002680/**
2681 * netdev_reset_queue - reset the packets and bytes count of a network device
2682 * @dev_queue: network device
2683 *
2684 * Reset the bytes and packet count of a network device and clear the
2685 * software flow control OFF bit for this network device
2686 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002687static inline void netdev_reset_queue(struct net_device *dev_queue)
2688{
2689 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07002690}
2691
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002692/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01002693 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2694 * @dev: network device
2695 * @queue_index: given tx queue index
2696 *
2697 * Returns 0 if given tx queue index >= number of device tx queues,
2698 * otherwise returns the originally passed tx queue index.
2699 */
2700static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2701{
2702 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2703 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2704 dev->name, queue_index,
2705 dev->real_num_tx_queues);
2706 return 0;
2707 }
2708
2709 return queue_index;
2710}
2711
2712/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002713 * netif_running - test if up
2714 * @dev: network device
2715 *
2716 * Test if the device has been brought up.
2717 */
David S. Miller4d295152012-03-07 21:02:35 -05002718static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002719{
2720 return test_bit(__LINK_STATE_START, &dev->state);
2721}
2722
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002723/*
2724 * Routines to manage the subqueues on a device. We only need start
2725 * stop, and a check if it's stopped. All other device management is
2726 * done at the overall netdevice level.
2727 * Also test the device if we're multiqueue.
2728 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002729
2730/**
2731 * netif_start_subqueue - allow sending packets on subqueue
2732 * @dev: network device
2733 * @queue_index: sub queue index
2734 *
2735 * Start individual transmit queue of a device with multiple transmit queues.
2736 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002737static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2738{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002739 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00002740
2741 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002742}
2743
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002744/**
2745 * netif_stop_subqueue - stop sending packets on subqueue
2746 * @dev: network device
2747 * @queue_index: sub queue index
2748 *
2749 * Stop individual transmit queue of a device with multiple transmit queues.
2750 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002751static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2752{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002753 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00002754 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002755}
2756
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002757/**
2758 * netif_subqueue_stopped - test status of subqueue
2759 * @dev: network device
2760 * @queue_index: sub queue index
2761 *
2762 * Check individual transmit queue of a device with multiple transmit queues.
2763 */
David S. Miller4d295152012-03-07 21:02:35 -05002764static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2765 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002766{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002767 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00002768
2769 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002770}
2771
David S. Miller4d295152012-03-07 21:02:35 -05002772static inline bool netif_subqueue_stopped(const struct net_device *dev,
2773 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07002774{
2775 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2776}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002777
John Fastabend46e5da40a2014-09-12 20:04:52 -07002778void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002779
Alexander Duyck537c00d2013-01-10 08:57:02 +00002780#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04002781int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07002782 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00002783#else
2784static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03002785 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00002786 u16 index)
2787{
2788 return 0;
2789}
2790#endif
2791
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002792/*
2793 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2794 * as a distribution range limit for the returned value.
2795 */
2796static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07002797 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002798{
2799 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2800}
2801
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002802/**
2803 * netif_is_multiqueue - test if device has multiple transmit queues
2804 * @dev: network device
2805 *
2806 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002807 */
David S. Miller4d295152012-03-07 21:02:35 -05002808static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002809{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002810 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002811}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002812
Joe Perchesf629d202013-09-26 14:48:15 -07002813int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00002814
Michael Daltona953be52014-01-16 22:23:28 -08002815#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07002816int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00002817#else
2818static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2819 unsigned int rxq)
2820{
2821 return 0;
2822}
2823#endif
2824
Michael Daltona953be52014-01-16 22:23:28 -08002825#ifdef CONFIG_SYSFS
2826static inline unsigned int get_netdev_rx_queue_index(
2827 struct netdev_rx_queue *queue)
2828{
2829 struct net_device *dev = queue->dev;
2830 int index = queue - dev->_rx;
2831
2832 BUG_ON(index >= dev->num_rx_queues);
2833 return index;
2834}
2835#endif
2836
Yuval Mintz16917b82012-07-01 03:18:50 +00002837#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07002838int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00002839
Eric Dumazete6247022013-12-05 04:45:08 -08002840enum skb_free_reason {
2841 SKB_REASON_CONSUMED,
2842 SKB_REASON_DROPPED,
2843};
Linus Torvalds1da177e2005-04-16 15:20:36 -07002844
Eric Dumazete6247022013-12-05 04:45:08 -08002845void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2846void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2847
2848/*
2849 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2850 * interrupt context or with hardware interrupts being disabled.
2851 * (in_irq() || irqs_disabled())
2852 *
2853 * We provide four helpers that can be used in following contexts :
2854 *
2855 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2856 * replacing kfree_skb(skb)
2857 *
2858 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2859 * Typically used in place of consume_skb(skb) in TX completion path
2860 *
2861 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2862 * replacing kfree_skb(skb)
2863 *
2864 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2865 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002866 */
Eric Dumazete6247022013-12-05 04:45:08 -08002867static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2868{
2869 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2870}
2871
2872static inline void dev_consume_skb_irq(struct sk_buff *skb)
2873{
2874 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2875}
2876
2877static inline void dev_kfree_skb_any(struct sk_buff *skb)
2878{
2879 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2880}
2881
2882static inline void dev_consume_skb_any(struct sk_buff *skb)
2883{
2884 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2885}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002886
Joe Perchesf629d202013-09-26 14:48:15 -07002887int netif_rx(struct sk_buff *skb);
2888int netif_rx_ni(struct sk_buff *skb);
2889int netif_receive_skb(struct sk_buff *skb);
2890gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2891void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2892struct sk_buff *napi_get_frags(struct napi_struct *napi);
2893gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08002894struct packet_offload *gro_find_receive_by_type(__be16 type);
2895struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07002896
2897static inline void napi_free_frags(struct napi_struct *napi)
2898{
2899 kfree_skb(napi->skb);
2900 napi->skb = NULL;
2901}
2902
Joe Perchesf629d202013-09-26 14:48:15 -07002903int netdev_rx_handler_register(struct net_device *dev,
2904 rx_handler_func_t *rx_handler,
2905 void *rx_handler_data);
2906void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00002907
Joe Perchesf629d202013-09-26 14:48:15 -07002908bool dev_valid_name(const char *name);
2909int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2910int dev_ethtool(struct net *net, struct ifreq *);
2911unsigned int dev_get_flags(const struct net_device *);
2912int __dev_change_flags(struct net_device *, unsigned int flags);
2913int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04002914void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2915 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07002916int dev_change_name(struct net_device *, const char *);
2917int dev_set_alias(struct net_device *, const char *, size_t);
2918int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2919int dev_set_mtu(struct net_device *, int);
2920void dev_set_group(struct net_device *, int);
2921int dev_set_mac_address(struct net_device *, struct sockaddr *);
2922int dev_change_carrier(struct net_device *, bool new_carrier);
2923int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01002924 struct netdev_phys_item_id *ppid);
Eric Dumazet55a93b32014-10-03 15:31:07 -07002925struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07002926struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2927 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08002928int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07002929int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Vlad Yasevich1ee481f2014-03-27 17:32:29 -04002930bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002931
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002932extern int netdev_budget;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002933
2934/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07002935void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002936
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002937/**
2938 * dev_put - release reference to device
2939 * @dev: network device
2940 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07002941 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002942 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002943static inline void dev_put(struct net_device *dev)
2944{
Christoph Lameter933393f2011-12-22 11:58:51 -06002945 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002946}
2947
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002948/**
2949 * dev_hold - get reference to device
2950 * @dev: network device
2951 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07002952 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002953 */
Stephen Hemminger15333062006-03-20 22:32:28 -08002954static inline void dev_hold(struct net_device *dev)
2955{
Christoph Lameter933393f2011-12-22 11:58:51 -06002956 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08002957}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002958
2959/* Carrier loss detection, dial on demand. The functions netif_carrier_on
2960 * and _off may be called from IRQ context, but it is caller
2961 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08002962 *
2963 * The name carrier is inappropriate, these functions should really be
2964 * called netif_lowerlayer_*() because they represent the state of any
2965 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002966 */
2967
Joe Perchesf629d202013-09-26 14:48:15 -07002968void linkwatch_init_dev(struct net_device *dev);
2969void linkwatch_fire_event(struct net_device *dev);
2970void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002971
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002972/**
2973 * netif_carrier_ok - test if carrier present
2974 * @dev: network device
2975 *
2976 * Check if carrier is present on device
2977 */
David S. Miller4d295152012-03-07 21:02:35 -05002978static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002979{
2980 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2981}
2982
Joe Perchesf629d202013-09-26 14:48:15 -07002983unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07002984
Joe Perchesf629d202013-09-26 14:48:15 -07002985void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002986
Joe Perchesf629d202013-09-26 14:48:15 -07002987void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002988
Joe Perchesf629d202013-09-26 14:48:15 -07002989void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002990
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002991/**
2992 * netif_dormant_on - mark device as dormant.
2993 * @dev: network device
2994 *
2995 * Mark device as dormant (as per RFC2863).
2996 *
2997 * The dormant state indicates that the relevant interface is not
2998 * actually in a condition to pass packets (i.e., it is not 'up') but is
2999 * in a "pending" state, waiting for some external event. For "on-
3000 * demand" interfaces, this new state identifies the situation where the
3001 * interface is waiting for events to place it in the up state.
3002 *
3003 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003004static inline void netif_dormant_on(struct net_device *dev)
3005{
3006 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3007 linkwatch_fire_event(dev);
3008}
3009
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003010/**
3011 * netif_dormant_off - set device as not dormant.
3012 * @dev: network device
3013 *
3014 * Device is not in dormant state.
3015 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003016static inline void netif_dormant_off(struct net_device *dev)
3017{
3018 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3019 linkwatch_fire_event(dev);
3020}
3021
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003022/**
3023 * netif_dormant - test if carrier present
3024 * @dev: network device
3025 *
3026 * Check if carrier is present on device
3027 */
David S. Miller4d295152012-03-07 21:02:35 -05003028static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003029{
3030 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3031}
3032
3033
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003034/**
3035 * netif_oper_up - test if device is operational
3036 * @dev: network device
3037 *
3038 * Check if carrier is operational
3039 */
David S. Miller4d295152012-03-07 21:02:35 -05003040static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003041{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003042 return (dev->operstate == IF_OPER_UP ||
3043 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3044}
3045
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003046/**
3047 * netif_device_present - is device available or removed
3048 * @dev: network device
3049 *
3050 * Check if device has not been removed from system.
3051 */
David S. Miller4d295152012-03-07 21:02:35 -05003052static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003053{
3054 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3055}
3056
Joe Perchesf629d202013-09-26 14:48:15 -07003057void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003058
Joe Perchesf629d202013-09-26 14:48:15 -07003059void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003060
3061/*
3062 * Network interface message level settings
3063 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003064
3065enum {
3066 NETIF_MSG_DRV = 0x0001,
3067 NETIF_MSG_PROBE = 0x0002,
3068 NETIF_MSG_LINK = 0x0004,
3069 NETIF_MSG_TIMER = 0x0008,
3070 NETIF_MSG_IFDOWN = 0x0010,
3071 NETIF_MSG_IFUP = 0x0020,
3072 NETIF_MSG_RX_ERR = 0x0040,
3073 NETIF_MSG_TX_ERR = 0x0080,
3074 NETIF_MSG_TX_QUEUED = 0x0100,
3075 NETIF_MSG_INTR = 0x0200,
3076 NETIF_MSG_TX_DONE = 0x0400,
3077 NETIF_MSG_RX_STATUS = 0x0800,
3078 NETIF_MSG_PKTDATA = 0x1000,
3079 NETIF_MSG_HW = 0x2000,
3080 NETIF_MSG_WOL = 0x4000,
3081};
3082
3083#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3084#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3085#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3086#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3087#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3088#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3089#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3090#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3091#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3092#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3093#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3094#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3095#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3096#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3097#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3098
3099static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3100{
3101 /* use default */
3102 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3103 return default_msg_enable_bits;
3104 if (debug_value == 0) /* no output */
3105 return 0;
3106 /* set low N bits */
3107 return (1 << debug_value) - 1;
3108}
3109
David S. Millerc773e842008-07-08 23:13:53 -07003110static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003111{
David S. Millerc773e842008-07-08 23:13:53 -07003112 spin_lock(&txq->_xmit_lock);
3113 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003114}
3115
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003116static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3117{
3118 spin_lock_bh(&txq->_xmit_lock);
3119 txq->xmit_lock_owner = smp_processor_id();
3120}
3121
David S. Miller4d295152012-03-07 21:02:35 -05003122static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003123{
David S. Miller4d295152012-03-07 21:02:35 -05003124 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003125 if (likely(ok))
3126 txq->xmit_lock_owner = smp_processor_id();
3127 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003128}
3129
David S. Millerc773e842008-07-08 23:13:53 -07003130static inline void __netif_tx_unlock(struct netdev_queue *txq)
3131{
3132 txq->xmit_lock_owner = -1;
3133 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003134}
3135
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003136static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3137{
3138 txq->xmit_lock_owner = -1;
3139 spin_unlock_bh(&txq->_xmit_lock);
3140}
3141
Eric Dumazet08baf562009-05-25 22:58:01 -07003142static inline void txq_trans_update(struct netdev_queue *txq)
3143{
3144 if (txq->xmit_lock_owner != -1)
3145 txq->trans_start = jiffies;
3146}
3147
David S. Millerc3f26a22008-07-31 16:58:50 -07003148/**
3149 * netif_tx_lock - grab network device transmit lock
3150 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003151 *
3152 * Get network device transmit lock
3153 */
3154static inline void netif_tx_lock(struct net_device *dev)
3155{
3156 unsigned int i;
3157 int cpu;
3158
3159 spin_lock(&dev->tx_global_lock);
3160 cpu = smp_processor_id();
3161 for (i = 0; i < dev->num_tx_queues; i++) {
3162 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3163
3164 /* We are the only thread of execution doing a
3165 * freeze, but we have to grab the _xmit_lock in
3166 * order to synchronize with threads which are in
3167 * the ->hard_start_xmit() handler and already
3168 * checked the frozen bit.
3169 */
3170 __netif_tx_lock(txq, cpu);
3171 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3172 __netif_tx_unlock(txq);
3173 }
3174}
3175
3176static inline void netif_tx_lock_bh(struct net_device *dev)
3177{
3178 local_bh_disable();
3179 netif_tx_lock(dev);
3180}
3181
Herbert Xu932ff272006-06-09 12:20:56 -07003182static inline void netif_tx_unlock(struct net_device *dev)
3183{
David S. Millere8a04642008-07-17 00:34:19 -07003184 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003185
David S. Millere8a04642008-07-17 00:34:19 -07003186 for (i = 0; i < dev->num_tx_queues; i++) {
3187 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003188
David S. Millerc3f26a22008-07-31 16:58:50 -07003189 /* No need to grab the _xmit_lock here. If the
3190 * queue is not stopped for another reason, we
3191 * force a schedule.
3192 */
3193 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003194 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003195 }
3196 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003197}
3198
3199static inline void netif_tx_unlock_bh(struct net_device *dev)
3200{
David S. Millere8a04642008-07-17 00:34:19 -07003201 netif_tx_unlock(dev);
3202 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003203}
3204
David S. Millerc773e842008-07-08 23:13:53 -07003205#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003206 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003207 __netif_tx_lock(txq, cpu); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003208 } \
3209}
3210
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003211#define HARD_TX_TRYLOCK(dev, txq) \
3212 (((dev->features & NETIF_F_LLTX) == 0) ? \
3213 __netif_tx_trylock(txq) : \
3214 true )
3215
David S. Millerc773e842008-07-08 23:13:53 -07003216#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003217 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003218 __netif_tx_unlock(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003219 } \
3220}
3221
Linus Torvalds1da177e2005-04-16 15:20:36 -07003222static inline void netif_tx_disable(struct net_device *dev)
3223{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003224 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003225 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003226
David S. Millerc3f26a22008-07-31 16:58:50 -07003227 local_bh_disable();
3228 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003229 for (i = 0; i < dev->num_tx_queues; i++) {
3230 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003231
3232 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003233 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003234 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003235 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003236 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003237}
3238
David S. Millere308a5d2008-07-15 00:13:44 -07003239static inline void netif_addr_lock(struct net_device *dev)
3240{
3241 spin_lock(&dev->addr_list_lock);
3242}
3243
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003244static inline void netif_addr_lock_nested(struct net_device *dev)
3245{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003246 int subclass = SINGLE_DEPTH_NESTING;
3247
3248 if (dev->netdev_ops->ndo_get_lock_subclass)
3249 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3250
3251 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003252}
3253
David S. Millere308a5d2008-07-15 00:13:44 -07003254static inline void netif_addr_lock_bh(struct net_device *dev)
3255{
3256 spin_lock_bh(&dev->addr_list_lock);
3257}
3258
3259static inline void netif_addr_unlock(struct net_device *dev)
3260{
3261 spin_unlock(&dev->addr_list_lock);
3262}
3263
3264static inline void netif_addr_unlock_bh(struct net_device *dev)
3265{
3266 spin_unlock_bh(&dev->addr_list_lock);
3267}
3268
Jiri Pirkof001fde2009-05-05 02:48:28 +00003269/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003270 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003271 * rcu_read_lock held.
3272 */
3273#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003274 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003275
Linus Torvalds1da177e2005-04-16 15:20:36 -07003276/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3277
Joe Perchesf629d202013-09-26 14:48:15 -07003278void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003279
3280/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003281struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003282 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003283 void (*setup)(struct net_device *),
3284 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003285#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3286 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003287
Tom Gundersenc835a672014-07-14 16:37:24 +02003288#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3289 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3290 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003291
Joe Perchesf629d202013-09-26 14:48:15 -07003292int register_netdev(struct net_device *dev);
3293void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003294
Jiri Pirko22bedad32010-04-01 21:22:57 +00003295/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003296int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3297 struct netdev_hw_addr_list *from_list, int addr_len);
3298void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3299 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003300int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3301 struct net_device *dev,
3302 int (*sync)(struct net_device *, const unsigned char *),
3303 int (*unsync)(struct net_device *,
3304 const unsigned char *));
3305void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3306 struct net_device *dev,
3307 int (*unsync)(struct net_device *,
3308 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003309void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003310
Jiri Pirkof001fde2009-05-05 02:48:28 +00003311/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003312int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3313 unsigned char addr_type);
3314int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3315 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003316void dev_addr_flush(struct net_device *dev);
3317int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003318
3319/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003320int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3321int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3322int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3323int dev_uc_sync(struct net_device *to, struct net_device *from);
3324int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3325void dev_uc_unsync(struct net_device *to, struct net_device *from);
3326void dev_uc_flush(struct net_device *dev);
3327void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003328
Alexander Duyck670e5b82014-05-28 18:44:46 -07003329/**
3330 * __dev_uc_sync - Synchonize device's unicast list
3331 * @dev: device to sync
3332 * @sync: function to call if address should be added
3333 * @unsync: function to call if address should be removed
3334 *
3335 * Add newly added addresses to the interface, and release
3336 * addresses that have been deleted.
3337 **/
3338static inline int __dev_uc_sync(struct net_device *dev,
3339 int (*sync)(struct net_device *,
3340 const unsigned char *),
3341 int (*unsync)(struct net_device *,
3342 const unsigned char *))
3343{
3344 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3345}
3346
3347/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003348 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003349 * @dev: device to sync
3350 * @unsync: function to call if address should be removed
3351 *
3352 * Remove all addresses that were added to the device by dev_uc_sync().
3353 **/
3354static inline void __dev_uc_unsync(struct net_device *dev,
3355 int (*unsync)(struct net_device *,
3356 const unsigned char *))
3357{
3358 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3359}
3360
Jiri Pirko22bedad32010-04-01 21:22:57 +00003361/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003362int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3363int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3364int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3365int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3366int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3367int dev_mc_sync(struct net_device *to, struct net_device *from);
3368int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3369void dev_mc_unsync(struct net_device *to, struct net_device *from);
3370void dev_mc_flush(struct net_device *dev);
3371void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003372
Alexander Duyck670e5b82014-05-28 18:44:46 -07003373/**
3374 * __dev_mc_sync - Synchonize device's multicast list
3375 * @dev: device to sync
3376 * @sync: function to call if address should be added
3377 * @unsync: function to call if address should be removed
3378 *
3379 * Add newly added addresses to the interface, and release
3380 * addresses that have been deleted.
3381 **/
3382static inline int __dev_mc_sync(struct net_device *dev,
3383 int (*sync)(struct net_device *,
3384 const unsigned char *),
3385 int (*unsync)(struct net_device *,
3386 const unsigned char *))
3387{
3388 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3389}
3390
3391/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003392 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003393 * @dev: device to sync
3394 * @unsync: function to call if address should be removed
3395 *
3396 * Remove all addresses that were added to the device by dev_mc_sync().
3397 **/
3398static inline void __dev_mc_unsync(struct net_device *dev,
3399 int (*unsync)(struct net_device *,
3400 const unsigned char *))
3401{
3402 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3403}
3404
Linus Torvalds1da177e2005-04-16 15:20:36 -07003405/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003406void dev_set_rx_mode(struct net_device *dev);
3407void __dev_set_rx_mode(struct net_device *dev);
3408int dev_set_promiscuity(struct net_device *dev, int inc);
3409int dev_set_allmulti(struct net_device *dev, int inc);
3410void netdev_state_change(struct net_device *dev);
3411void netdev_notify_peers(struct net_device *dev);
3412void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003413/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003414void dev_load(struct net *net, const char *name);
3415struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3416 struct rtnl_link_stats64 *storage);
3417void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3418 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003419
3420extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003421extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003422extern int weight_p;
Eric Dumazet0a148422011-04-20 09:27:32 +00003423extern int bpf_jit_enable;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003424
Joe Perchesf629d202013-09-26 14:48:15 -07003425bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003426struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3427 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003428struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3429 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003430
3431/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003432#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3433 for (iter = &(dev)->adj_list.upper, \
3434 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3435 updev; \
3436 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3437
3438/* iterate through upper list, must be called under RCU read lock */
Veaceslav Falico2f268f12013-09-25 09:20:07 +02003439#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3440 for (iter = &(dev)->all_adj_list.upper, \
3441 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3442 updev; \
3443 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003444
Joe Perchesf629d202013-09-26 14:48:15 -07003445void *netdev_lower_get_next_private(struct net_device *dev,
3446 struct list_head **iter);
3447void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3448 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003449
3450#define netdev_for_each_lower_private(dev, priv, iter) \
3451 for (iter = (dev)->adj_list.lower.next, \
3452 priv = netdev_lower_get_next_private(dev, &(iter)); \
3453 priv; \
3454 priv = netdev_lower_get_next_private(dev, &(iter)))
3455
3456#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3457 for (iter = &(dev)->adj_list.lower, \
3458 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3459 priv; \
3460 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3461
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003462void *netdev_lower_get_next(struct net_device *dev,
3463 struct list_head **iter);
3464#define netdev_for_each_lower_dev(dev, ldev, iter) \
3465 for (iter = &(dev)->adj_list.lower, \
3466 ldev = netdev_lower_get_next(dev, &(iter)); \
3467 ldev; \
3468 ldev = netdev_lower_get_next(dev, &(iter)))
3469
Joe Perchesf629d202013-09-26 14:48:15 -07003470void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003471void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003472struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3473struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3474int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3475int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003476 struct net_device *upper_dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003477int netdev_master_upper_dev_link_private(struct net_device *dev,
3478 struct net_device *upper_dev,
3479 void *private);
3480void netdev_upper_dev_unlink(struct net_device *dev,
3481 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003482void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003483void *netdev_lower_dev_get_private(struct net_device *dev,
3484 struct net_device *lower_dev);
Eric Dumazet960fb622014-11-16 06:23:05 -08003485
3486/* RSS keys are 40 or 52 bytes long */
3487#define NETDEV_RSS_KEY_LEN 52
3488extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3489void netdev_rss_key_fill(void *buffer, size_t len);
3490
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003491int dev_get_nest_level(struct net_device *dev,
3492 bool (*type_check)(struct net_device *dev));
Joe Perchesf629d202013-09-26 14:48:15 -07003493int skb_checksum_help(struct sk_buff *skb);
3494struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3495 netdev_features_t features, bool tx_path);
3496struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3497 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003498
Moni Shoua61bd3852015-02-03 16:48:29 +02003499struct netdev_bonding_info {
3500 ifslave slave;
3501 ifbond master;
3502};
3503
3504struct netdev_notifier_bonding_info {
3505 struct netdev_notifier_info info; /* must be first */
3506 struct netdev_bonding_info bonding_info;
3507};
3508
3509void netdev_bonding_info_change(struct net_device *dev,
3510 struct netdev_bonding_info *bonding_info);
3511
Cong Wang12b00042013-02-05 16:36:38 +00003512static inline
3513struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3514{
3515 return __skb_gso_segment(skb, features, true);
3516}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003517__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003518
3519static inline bool can_checksum_protocol(netdev_features_t features,
3520 __be16 protocol)
3521{
3522 return ((features & NETIF_F_GEN_CSUM) ||
3523 ((features & NETIF_F_V4_CSUM) &&
3524 protocol == htons(ETH_P_IP)) ||
3525 ((features & NETIF_F_V6_CSUM) &&
3526 protocol == htons(ETH_P_IPV6)) ||
3527 ((features & NETIF_F_FCOE_CRC) &&
3528 protocol == htons(ETH_P_FCOE)));
3529}
Cong Wang12b00042013-02-05 16:36:38 +00003530
Herbert Xufb286bb2005-11-10 13:01:24 -08003531#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003532void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003533#else
3534static inline void netdev_rx_csum_fault(struct net_device *dev)
3535{
3536}
3537#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003538/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003539void net_enable_timestamp(void);
3540void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003541
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003542#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003543int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003544#else
3545#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003546#endif
3547
David S. Miller47982482014-08-22 16:21:53 -07003548static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc22014-08-29 21:55:22 -07003549 struct sk_buff *skb, struct net_device *dev,
3550 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003551{
David S. Millerfa2dbdc22014-08-29 21:55:22 -07003552 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003553 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003554}
3555
David S. Miller10b3ad82014-08-29 21:07:24 -07003556static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc22014-08-29 21:55:22 -07003557 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07003558{
3559 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07003560 int rc;
David S. Miller47982482014-08-22 16:21:53 -07003561
David S. Millerfa2dbdc22014-08-29 21:55:22 -07003562 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07003563 if (rc == NETDEV_TX_OK)
3564 txq_trans_update(txq);
3565
3566 return rc;
David S. Miller47982482014-08-22 16:21:53 -07003567}
3568
Linus Torvalds42a2d922013-11-13 17:40:34 +09003569int netdev_class_create_file_ns(struct class_attribute *class_attr,
3570 const void *ns);
3571void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3572 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04003573
3574static inline int netdev_class_create_file(struct class_attribute *class_attr)
3575{
3576 return netdev_class_create_file_ns(class_attr, NULL);
3577}
3578
3579static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3580{
3581 netdev_class_remove_file_ns(class_attr, NULL);
3582}
Jay Vosburghb8a97872008-06-13 18:12:04 -07003583
Johannes Berg04600792010-08-05 17:45:15 +02003584extern struct kobj_ns_type_operations net_ns_type_operations;
3585
Joe Perchesf629d202013-09-26 14:48:15 -07003586const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07003587
Joe Perchesf629d202013-09-26 14:48:15 -07003588void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003589
Michal Kubečekda081432014-05-20 08:29:25 +02003590static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3591 netdev_features_t f2)
3592{
3593 if (f1 & NETIF_F_GEN_CSUM)
3594 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3595 if (f2 & NETIF_F_GEN_CSUM)
3596 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3597 f1 &= f2;
3598 if (f1 & NETIF_F_GEN_CSUM)
3599 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3600
3601 return f1;
3602}
3603
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003604static inline netdev_features_t netdev_get_wanted_features(
3605 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00003606{
3607 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3608}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003609netdev_features_t netdev_increment_features(netdev_features_t all,
3610 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00003611
3612/* Allow TSO being used on stacked device :
3613 * Performing the GSO segmentation before last device
3614 * is a performance improvement.
3615 */
3616static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3617 netdev_features_t mask)
3618{
3619 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3620}
3621
Michał Mirosław6cb6a272011-04-02 22:48:47 -07003622int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00003623void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00003624void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07003625
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08003626void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3627 struct net_device *dev);
3628
Florian Westphalc1e756b2014-05-05 15:00:44 +02003629netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00003630
David S. Miller4d295152012-03-07 21:02:35 -05003631static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07003632{
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003633 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00003634
3635 /* check flags correspondence */
3636 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3637 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3638 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3639 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3640 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3641 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07003642 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3643 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3644 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3645 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3646 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3647 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08003648 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00003649
Herbert Xubcd76112006-06-30 13:36:35 -07003650 return (features & feature) == feature;
3651}
3652
David S. Miller4d295152012-03-07 21:02:35 -05003653static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07003654{
Herbert Xu278b2512009-06-03 21:20:51 -07003655 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07003656 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07003657}
3658
Tom Herbert04ffcb22014-10-14 15:19:06 -07003659static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05003660 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07003661{
Jesse Grossfc741212011-01-09 06:23:32 +00003662 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00003663 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3664 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07003665}
3666
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07003667static inline void netif_set_gso_max_size(struct net_device *dev,
3668 unsigned int size)
3669{
3670 dev->gso_max_size = size;
3671}
3672
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08003673static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3674 int pulled_hlen, u16 mac_offset,
3675 int mac_len)
3676{
3677 skb->protocol = protocol;
3678 skb->encapsulation = 1;
3679 skb_push(skb, pulled_hlen);
3680 skb_reset_transport_header(skb);
3681 skb->mac_header = mac_offset;
3682 skb->network_header = skb->mac_header + mac_len;
3683 skb->mac_len = mac_len;
3684}
3685
John Fastabenda6cc0cf2013-11-06 09:54:46 -08003686static inline bool netif_is_macvlan(struct net_device *dev)
3687{
3688 return dev->priv_flags & IFF_MACVLAN;
3689}
3690
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08003691static inline bool netif_is_macvlan_port(struct net_device *dev)
3692{
3693 return dev->priv_flags & IFF_MACVLAN_PORT;
3694}
3695
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08003696static inline bool netif_is_ipvlan(struct net_device *dev)
3697{
3698 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3699}
3700
3701static inline bool netif_is_ipvlan_port(struct net_device *dev)
3702{
3703 return dev->priv_flags & IFF_IPVLAN_MASTER;
3704}
3705
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00003706static inline bool netif_is_bond_master(struct net_device *dev)
3707{
3708 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3709}
3710
David S. Miller4d295152012-03-07 21:02:35 -05003711static inline bool netif_is_bond_slave(struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00003712{
3713 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3714}
3715
Ben Greear3bdc0eb2012-02-11 15:39:30 +00003716static inline bool netif_supports_nofcs(struct net_device *dev)
3717{
3718 return dev->priv_flags & IFF_SUPP_NOFCS;
3719}
3720
Eric Dumazet02875872014-10-05 18:38:35 -07003721/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3722static inline void netif_keep_dst(struct net_device *dev)
3723{
3724 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3725}
3726
Eric W. Biederman505d4f72008-11-07 22:54:20 -08003727extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00003728
Joe Perches571ba422010-02-09 11:49:47 +00003729/* Logging, debugging and troubleshooting/diagnostic helpers. */
3730
3731/* netdev_printk helpers, similar to dev_printk */
3732
3733static inline const char *netdev_name(const struct net_device *dev)
3734{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02003735 if (!dev->name[0] || strchr(dev->name, '%'))
3736 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00003737 return dev->name;
3738}
3739
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02003740static inline const char *netdev_reg_state(const struct net_device *dev)
3741{
3742 switch (dev->reg_state) {
3743 case NETREG_UNINITIALIZED: return " (uninitialized)";
3744 case NETREG_REGISTERED: return "";
3745 case NETREG_UNREGISTERING: return " (unregistering)";
3746 case NETREG_UNREGISTERED: return " (unregistered)";
3747 case NETREG_RELEASED: return " (released)";
3748 case NETREG_DUMMY: return " (dummy)";
3749 }
3750
3751 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3752 return " (unknown)";
3753}
3754
Joe Perchesf629d202013-09-26 14:48:15 -07003755__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07003756void netdev_printk(const char *level, const struct net_device *dev,
3757 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003758__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003759void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003760__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003761void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003762__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003763void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003764__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003765void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003766__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003767void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003768__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003769void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003770__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003771void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00003772
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03003773#define MODULE_ALIAS_NETDEV(device) \
3774 MODULE_ALIAS("netdev-" device)
3775
Jim Cromieb558c962011-12-19 17:11:18 -05003776#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00003777#define netdev_dbg(__dev, format, args...) \
3778do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04003779 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00003780} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05003781#elif defined(DEBUG)
3782#define netdev_dbg(__dev, format, args...) \
3783 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00003784#else
3785#define netdev_dbg(__dev, format, args...) \
3786({ \
3787 if (0) \
3788 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00003789})
3790#endif
3791
3792#if defined(VERBOSE_DEBUG)
3793#define netdev_vdbg netdev_dbg
3794#else
3795
3796#define netdev_vdbg(dev, format, args...) \
3797({ \
3798 if (0) \
3799 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3800 0; \
3801})
3802#endif
3803
3804/*
3805 * netdev_WARN() acts like dev_printk(), but with the key difference
3806 * of using a WARN/WARN_ON to get the message out, including the
3807 * file/line information and a backtrace.
3808 */
3809#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02003810 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3811 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00003812
Joe Perchesb3d95c52010-02-09 11:49:49 +00003813/* netif printk helpers, similar to netdev_printk */
3814
3815#define netif_printk(priv, type, level, dev, fmt, args...) \
3816do { \
3817 if (netif_msg_##type(priv)) \
3818 netdev_printk(level, (dev), fmt, ##args); \
3819} while (0)
3820
Joe Perchesf45f4322010-06-27 01:02:36 +00003821#define netif_level(level, priv, type, dev, fmt, args...) \
3822do { \
3823 if (netif_msg_##type(priv)) \
3824 netdev_##level(dev, fmt, ##args); \
3825} while (0)
3826
Joe Perchesb3d95c52010-02-09 11:49:49 +00003827#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003828 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003829#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003830 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003831#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003832 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003833#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003834 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003835#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003836 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003837#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003838 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003839#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003840 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003841
Joe Perches0053ea92012-05-30 07:43:34 +00003842#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003843#define netif_dbg(priv, type, netdev, format, args...) \
3844do { \
3845 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04003846 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00003847} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00003848#elif defined(DEBUG)
3849#define netif_dbg(priv, type, dev, format, args...) \
3850 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003851#else
3852#define netif_dbg(priv, type, dev, format, args...) \
3853({ \
3854 if (0) \
3855 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3856 0; \
3857})
3858#endif
3859
3860#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00003861#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00003862#else
3863#define netif_vdbg(priv, type, dev, format, args...) \
3864({ \
3865 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00003866 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00003867 0; \
3868})
3869#endif
Joe Perches571ba422010-02-09 11:49:47 +00003870
Cong Wang900ff8c2013-02-18 19:20:33 +00003871/*
3872 * The list of packet types we will receive (as opposed to discard)
3873 * and the routines to invoke.
3874 *
3875 * Why 16. Because with 16 the only overlap we get on a hash of the
3876 * low nibble of the protocol value is RARP/SNAP/X.25.
3877 *
3878 * NOTE: That is no longer true with the addition of VLAN tags. Not
3879 * sure which should go first, but I bet it won't make much
3880 * difference if we are running VLANs. The good news is that
3881 * this protocol won't be in the list unless compiled in, so
3882 * the average user (w/out VLANs) will not be adversely affected.
3883 * --BLG
3884 *
3885 * 0800 IP
3886 * 8100 802.1Q VLAN
3887 * 0001 802.3
3888 * 0002 AX.25
3889 * 0004 802.2
3890 * 8035 RARP
3891 * 0005 SNAP
3892 * 0805 X.25
3893 * 0806 ARP
3894 * 8137 IPX
3895 * 0009 Localtalk
3896 * 86DD IPv6
3897 */
3898#define PTYPE_HASH_SIZE (16)
3899#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3900
Jiri Pirko385a1542009-05-27 15:48:07 -07003901#endif /* _LINUX_NETDEVICE_H */