blob: 3a788bf0affdcd80a1282c5e4f4a5dc59e56c8ca [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
Al Virod7fe0f22006-12-03 23:15:30 -050028#include <linux/timer.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050029#include <linux/bug.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070030#include <linux/delay.h>
Arun Sharma600634972011-07-26 16:09:06 -070031#include <linux/atomic.h>
Eric Dumazet53511452014-10-08 08:19:27 -070032#include <linux/prefetch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <asm/cache.h>
34#include <asm/byteorder.h>
35
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include <linux/percpu.h>
David S. Miller4d5b78c2009-05-06 16:52:51 -070037#include <linux/rculist.h>
Chris Leechdb217332006-06-17 21:24:58 -070038#include <linux/dmaengine.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070039#include <linux/workqueue.h>
Tom Herbert114cf582011-11-28 16:33:09 +000040#include <linux/dynamic_queue_limits.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
Patrick McHardyb1b67dd2009-04-20 04:49:28 +000042#include <linux/ethtool.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020043#include <net/net_namespace.h>
Lennert Buytenhekcf85d082008-10-07 13:45:02 +000044#include <net/dsa.h>
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -080045#ifdef CONFIG_DCB
Alexander Duyck2f90b862008-11-20 20:52:10 -080046#include <net/dcbnl.h>
47#endif
Neil Horman5bc14212011-11-22 05:10:51 +000048#include <net/netprio_cgroup.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020049
Michał Mirosława59e2ec2011-11-15 15:29:55 +000050#include <linux/netdev_features.h>
John Fastabend77162022012-04-15 06:43:56 +000051#include <linux/neighbour.h>
David Howells607ca462012-10-13 10:46:48 +010052#include <uapi/linux/netdevice.h>
Moni Shoua61bd3852015-02-03 16:48:29 +020053#include <uapi/linux/if_bonding.h>
John Fastabende4c67342016-02-16 21:16:15 -080054#include <uapi/linux/pkt_cls.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;
Robert Shearman03c57742015-04-22 11:14:37 +010063struct mpls_dev;
Alexander Duyck7c46a642016-06-16 12:21:00 -070064/* UDP Tunnel offloads */
65struct udp_tunnel_info;
Brenden Blancoa7862b42016-07-19 12:16:48 -070066struct bpf_prog;
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
Joe Perchesf629d202013-09-26 14:48:15 -070068void netdev_set_default_ethtool_ops(struct net_device *dev,
69 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000070
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000071/* Backlog congestion levels */
72#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
73#define NET_RX_DROP 1 /* packet dropped */
74
Patrick McHardy572a9d72009-11-10 06:14:14 +000075/*
76 * Transmit return codes: transmit return codes originate from three different
77 * namespaces:
78 *
79 * - qdisc return codes
80 * - driver transmit return codes
81 * - errno values
82 *
83 * Drivers are allowed to return any one of those in their hard_start_xmit()
84 * function. Real network devices commonly used with qdiscs should only return
85 * the driver transmit return codes though - when qdiscs are used, the actual
86 * transmission happens asynchronously, so the value is not propagated to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -050087 * higher layers. Virtual network devices transmit synchronously; in this case
88 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
Patrick McHardy572a9d72009-11-10 06:14:14 +000089 * others are propagated to higher layers.
90 */
91
92/* qdisc ->enqueue() return codes. */
93#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000094#define NET_XMIT_DROP 0x01 /* skb dropped */
95#define NET_XMIT_CN 0x02 /* congestion notification */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000096#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -020098/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
99 * indicates that the device will soon be dropping packets, or already drops
100 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +0000101#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
103
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000104/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000105#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000106
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000107enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000108 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000109 NETDEV_TX_OK = 0x00, /* driver took care of packet */
110 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000111};
112typedef enum netdev_tx netdev_tx_t;
113
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000114/*
115 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
116 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
117 */
118static inline bool dev_xmit_complete(int rc)
119{
120 /*
121 * Positive cases with an skb consumed by a driver:
122 * - successful transmission (rc == NETDEV_TX_OK)
123 * - error while transmitting (rc < 0)
124 * - error while queueing to a different device (rc & NET_XMIT_MASK)
125 */
126 if (likely(rc < NET_XMIT_MASK))
127 return true;
128
129 return false;
130}
131
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500133 * Compute the worst-case header length according to the protocols
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 * used.
135 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800136
KY Srinivasanc0eb4542015-12-01 16:43:10 -0800137#if defined(CONFIG_HYPERV_NET)
138# define LL_MAX_HEADER 128
139#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700140# if defined(CONFIG_MAC80211_MESH)
141# define LL_MAX_HEADER 128
142# else
143# define LL_MAX_HEADER 96
144# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700146# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147#endif
148
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000149#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
150 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151#define MAX_HEADER LL_MAX_HEADER
152#else
153#define MAX_HEADER (LL_MAX_HEADER + 48)
154#endif
155
156/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000157 * Old network device statistics. Fields are native words
158 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800160
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800161struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000162 unsigned long rx_packets;
163 unsigned long tx_packets;
164 unsigned long rx_bytes;
165 unsigned long tx_bytes;
166 unsigned long rx_errors;
167 unsigned long tx_errors;
168 unsigned long rx_dropped;
169 unsigned long tx_dropped;
170 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000173 unsigned long rx_over_errors;
174 unsigned long rx_crc_errors;
175 unsigned long rx_frame_errors;
176 unsigned long rx_fifo_errors;
177 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 unsigned long tx_aborted_errors;
179 unsigned long tx_carrier_errors;
180 unsigned long tx_fifo_errors;
181 unsigned long tx_heartbeat_errors;
182 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 unsigned long rx_compressed;
184 unsigned long tx_compressed;
185};
186
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187
188#include <linux/cache.h>
189#include <linux/skbuff.h>
190
Eric Dumazetadc93002011-11-17 03:13:26 +0000191#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100192#include <linux/static_key.h>
193extern struct static_key rps_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000194#endif
195
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196struct neighbour;
197struct neigh_parms;
198struct sk_buff;
199
Jiri Pirkof001fde2009-05-05 02:48:28 +0000200struct netdev_hw_addr {
201 struct list_head list;
202 unsigned char addr[MAX_ADDR_LEN];
203 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000204#define NETDEV_HW_ADDR_T_LAN 1
205#define NETDEV_HW_ADDR_T_SAN 2
206#define NETDEV_HW_ADDR_T_SLAVE 3
207#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000208#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000209 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000210 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700211 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400212 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000213 struct rcu_head rcu_head;
214};
215
Jiri Pirko31278e72009-06-17 01:12:19 +0000216struct netdev_hw_addr_list {
217 struct list_head list;
218 int count;
219};
220
Jiri Pirko22bedad32010-04-01 21:22:57 +0000221#define netdev_hw_addr_list_count(l) ((l)->count)
222#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
223#define netdev_hw_addr_list_for_each(ha, l) \
224 list_for_each_entry(ha, &(l)->list, list)
225
226#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
227#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800228#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000229 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800230
Jiri Pirko22bedad32010-04-01 21:22:57 +0000231#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
232#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700233#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000234 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece32010-02-04 10:22:25 -0800235
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800236struct hh_cache {
David S. Millerf6b72b622011-07-14 07:53:20 -0700237 u16 hh_len;
David S. Miller5c25f682011-07-13 00:51:10 -0700238 u16 __pad;
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800239 seqlock_t hh_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
241 /* cached hardware header; allow for machine alignment needs. */
242#define HH_DATA_MOD 16
243#define HH_DATA_OFF(__len) \
Jiri Benc5ba0eac2005-06-02 16:48:05 -0700244 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245#define HH_DATA_ALIGN(__len) \
246 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
247 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
248};
249
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500250/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 * Alternative is:
252 * dev->hard_header_len ? (dev->hard_header_len +
253 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
254 *
255 * We could use other alignment values, but we must maintain the
256 * relationship HH alignment <= LL alignment.
257 */
258#define LL_RESERVED_SPACE(dev) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700259 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700261 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700263struct header_ops {
264 int (*create) (struct sk_buff *skb, struct net_device *dev,
265 unsigned short type, const void *daddr,
Eric Dumazet95c96172012-04-15 05:58:06 +0000266 const void *saddr, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700267 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
David S. Millere69dd332011-07-12 23:28:12 -0700268 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700269 void (*cache_update)(struct hh_cache *hh,
270 const struct net_device *dev,
271 const unsigned char *haddr);
Willem de Bruijn2793a232016-03-09 21:58:32 -0500272 bool (*validate)(const char *ll_header, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700273};
274
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275/* These flag bits are private to the generic network queueing
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500276 * layer; they may not be explicitly referenced by any other
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 * code.
278 */
279
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800280enum netdev_state_t {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 __LINK_STATE_START,
282 __LINK_STATE_PRESENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 __LINK_STATE_NOCARRIER,
Stefan Rompfb00055a2006-03-20 17:09:11 -0800284 __LINK_STATE_LINKWATCH_PENDING,
285 __LINK_STATE_DORMANT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286};
287
288
289/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500290 * This structure holds boot-time configured netdevice settings. They
Graf Yangfe2918b2009-02-05 21:26:19 -0800291 * are then used in the device probing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 */
293struct netdev_boot_setup {
294 char name[IFNAMSIZ];
295 struct ifmap map;
296};
297#define NETDEV_BOOT_SETUP_MAX 8
298
Joe Perchesf629d202013-09-26 14:48:15 -0700299int __init netdev_boot_setup(char *str);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300
301/*
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700302 * Structure for NAPI scheduling similar to tasklet but with weighting
303 */
304struct napi_struct {
305 /* The poll_list must only be managed by the entity which
306 * changes the state of the NAPI_STATE_SCHED bit. This means
307 * whoever atomically sets that bit can add this napi_struct
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500308 * to the per-CPU poll_list, and whoever clears that bit
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700309 * can remove from the list right before clearing the bit.
310 */
311 struct list_head poll_list;
312
313 unsigned long state;
314 int weight;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000315 unsigned int gro_count;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700316 int (*poll)(struct napi_struct *, int);
317#ifdef CONFIG_NETPOLL
318 spinlock_t poll_lock;
319 int poll_owner;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700320#endif
Herbert Xu5d38a072009-01-04 16:13:40 -0800321 struct net_device *dev;
Herbert Xud565b0a2008-12-15 23:38:52 -0800322 struct sk_buff *gro_list;
Herbert Xu5d38a072009-01-04 16:13:40 -0800323 struct sk_buff *skb;
Eric Dumazet3b47d302014-11-06 21:09:44 -0800324 struct hrtimer timer;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000325 struct list_head dev_list;
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300326 struct hlist_node napi_hash_node;
327 unsigned int napi_id;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700328};
329
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800330enum {
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700331 NAPI_STATE_SCHED, /* Poll is scheduled */
David S. Millera0a46192008-01-07 20:35:07 -0800332 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800333 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eric Dumazetd64b5e82015-11-18 06:31:00 -0800334 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
335 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700336};
337
Ben Hutchings5b252f02009-10-29 07:17:09 +0000338enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700339 GRO_MERGED,
340 GRO_MERGED_FREE,
341 GRO_HELD,
342 GRO_NORMAL,
343 GRO_DROP,
344};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000345typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700346
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000347/*
348 * enum rx_handler_result - Possible return values for rx_handlers.
349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
350 * further.
351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
352 * case skb->dev was changed by rx_handler.
353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500354 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000355 *
356 * rx_handlers are functions called from inside __netif_receive_skb(), to do
357 * special processing of the skb, prior to delivery to protocol handlers.
358 *
359 * Currently, a net_device can only have a single rx_handler registered. Trying
360 * to register a second rx_handler will return -EBUSY.
361 *
362 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
363 * To unregister a rx_handler on a net_device, use
364 * netdev_rx_handler_unregister().
365 *
366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
367 * do with the skb.
368 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500369 * If the rx_handler consumed the skb in some way, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500371 * the skb to be delivered in some other way.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000372 *
373 * If the rx_handler changed skb->dev, to divert the skb to another
374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
375 * new device will be called if it exists.
376 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500377 * If the rx_handler decides the skb should be ignored, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400379 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000380 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500381 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000382 * delivered, it should return RX_HANDLER_PASS.
383 *
384 * A device without a registered rx_handler will behave as if rx_handler
385 * returned RX_HANDLER_PASS.
386 */
387
388enum rx_handler_result {
389 RX_HANDLER_CONSUMED,
390 RX_HANDLER_ANOTHER,
391 RX_HANDLER_EXACT,
392 RX_HANDLER_PASS,
393};
394typedef enum rx_handler_result rx_handler_result_t;
395typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000396
Joe Perchesf629d202013-09-26 14:48:15 -0700397void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700398void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700399
David S. Miller4d295152012-03-07 21:02:35 -0500400static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800401{
402 return test_bit(NAPI_STATE_DISABLE, &n->state);
403}
404
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700405/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500406 * napi_schedule_prep - check if NAPI can be scheduled
407 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700408 *
409 * Test if NAPI routine is already running, and if not mark
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500410 * it as running. This is used as a condition variable to
David S. Millera0a46192008-01-07 20:35:07 -0800411 * insure only one NAPI poll instance runs. We also make
412 * sure there is no pending NAPI disable.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700413 */
David S. Miller4d295152012-03-07 21:02:35 -0500414static inline bool napi_schedule_prep(struct napi_struct *n)
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700415{
David S. Millera0a46192008-01-07 20:35:07 -0800416 return !napi_disable_pending(n) &&
417 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700418}
419
420/**
421 * napi_schedule - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500422 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700423 *
424 * Schedule NAPI poll routine to be called if it is not already
425 * running.
426 */
427static inline void napi_schedule(struct napi_struct *n)
428{
429 if (napi_schedule_prep(n))
430 __napi_schedule(n);
431}
432
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700433/**
434 * napi_schedule_irqoff - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500435 * @n: NAPI context
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700436 *
437 * Variant of napi_schedule(), assuming hard irqs are masked.
438 */
439static inline void napi_schedule_irqoff(struct napi_struct *n)
440{
441 if (napi_schedule_prep(n))
442 __napi_schedule_irqoff(n);
443}
444
Roland Dreierbfe13f52007-10-09 15:47:37 -0700445/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
David S. Miller4d295152012-03-07 21:02:35 -0500446static inline bool napi_reschedule(struct napi_struct *napi)
Roland Dreierbfe13f52007-10-09 15:47:37 -0700447{
448 if (napi_schedule_prep(napi)) {
449 __napi_schedule(napi);
David S. Miller4d295152012-03-07 21:02:35 -0500450 return true;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700451 }
David S. Miller4d295152012-03-07 21:02:35 -0500452 return false;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700453}
454
Eric Dumazet3b47d302014-11-06 21:09:44 -0800455void __napi_complete(struct napi_struct *n);
456void napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700457/**
458 * napi_complete - NAPI processing complete
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500459 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700460 *
461 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800462 * Consider using napi_complete_done() instead.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700463 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800464static inline void napi_complete(struct napi_struct *n)
465{
466 return napi_complete_done(n, 0);
467}
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700468
469/**
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300470 * napi_hash_add - add a NAPI to global hashtable
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500471 * @napi: NAPI context
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300472 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500473 * Generate a new napi_id and store a @napi under it in napi_hash.
474 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
Eric Dumazet93d05d42015-11-18 06:31:03 -0800475 * Note: This is normally automatically done from netif_napi_add(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500476 * so might disappear in a future Linux version.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300477 */
Joe Perchesf629d202013-09-26 14:48:15 -0700478void napi_hash_add(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300479
480/**
481 * napi_hash_del - remove a NAPI from global table
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500482 * @napi: NAPI context
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300483 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500484 * Warning: caller must observe RCU grace period
Eric Dumazet34cbe272015-11-18 06:31:02 -0800485 * before freeing memory containing @napi, if
486 * this function returns true.
Eric Dumazet93d05d42015-11-18 06:31:03 -0800487 * Note: core networking stack automatically calls it
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500488 * from netif_napi_del().
Eric Dumazet93d05d42015-11-18 06:31:03 -0800489 * Drivers might want to call this helper to combine all
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500490 * the needed RCU grace periods into a single one.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300491 */
Eric Dumazet34cbe272015-11-18 06:31:02 -0800492bool napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300493
494/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700495 * napi_disable - prevent NAPI from scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500496 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700497 *
498 * Stop NAPI from being scheduled on this context.
499 * Waits till any outstanding processing completes.
500 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800501void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700502
503/**
504 * napi_enable - enable NAPI scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500505 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700506 *
507 * Resume NAPI from being scheduled on this context.
508 * Must be paired with napi_disable.
509 */
510static inline void napi_enable(struct napi_struct *n)
511{
512 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100513 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700514 clear_bit(NAPI_STATE_SCHED, &n->state);
Neil Horman2d8bff1262015-09-23 14:57:58 -0400515 clear_bit(NAPI_STATE_NPSVC, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700516}
517
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700518/**
519 * napi_synchronize - wait until NAPI is not running
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500520 * @n: NAPI context
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700521 *
522 * Wait until NAPI is done being scheduled on this context.
523 * Waits till any outstanding processing completes but
524 * does not disable future activations.
525 */
526static inline void napi_synchronize(const struct napi_struct *n)
527{
Arnd Bergmannfacc4322016-01-22 11:43:44 +0100528 if (IS_ENABLED(CONFIG_SMP))
529 while (test_bit(NAPI_STATE_SCHED, &n->state))
530 msleep(1);
531 else
532 barrier();
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700533}
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700534
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800535enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000536 __QUEUE_STATE_DRV_XOFF,
537 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700538 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700539};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200540
541#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
542#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
543#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
544
545#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
546#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
547 QUEUE_STATE_FROZEN)
548#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
549 QUEUE_STATE_FROZEN)
550
Tom Herbert734664982011-11-28 16:32:44 +0000551/*
552 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
553 * netif_tx_* functions below are used to manipulate this flag. The
554 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
555 * queue independently. The netif_xmit_*stopped functions below are called
556 * to check if the queue has been stopped by the driver or stack (either
557 * of the XOFF bits are set in the state). Drivers should not need to call
558 * netif_xmit*stopped functions, they should only be using netif_tx_*.
559 */
David S. Miller79d16382008-07-08 23:14:46 -0700560
David S. Millerbb949fb2008-07-08 16:55:56 -0700561struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700562/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500563 * read-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700564 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700565 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700566 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700567 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000568#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000569 struct kobject kobj;
570#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000571#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
572 int numa_node;
573#endif
Florian Westphalc0ef0792016-05-03 03:29:09 +0200574 unsigned long tx_maxrate;
575 /*
576 * Number of TX timeouts for this queue
577 * (/sys/class/net/DEV/Q/trans_timeout)
578 */
579 unsigned long trans_timeout;
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700580/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500581 * write-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700582 */
583 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
584 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700585 /*
Florian Westphal9b366272016-05-03 16:33:14 +0200586 * Time (in jiffies) of last Tx
Eric Dumazet9d214932009-05-17 20:55:16 -0700587 */
588 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000589
Tom Herbert114cf582011-11-28 16:33:09 +0000590 unsigned long state;
591
592#ifdef CONFIG_BQL
593 struct dql dql;
594#endif
David S. Millere8a04642008-07-17 00:34:19 -0700595} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700596
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000597static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 return q->numa_node;
601#else
Changli Gaob236da62010-12-14 03:09:15 +0000602 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000603#endif
604}
605
606static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
607{
608#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
609 q->numa_node = node;
610#endif
611}
612
Eric Dumazetdf334542010-03-24 19:13:54 +0000613#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000614/*
615 * This structure holds an RPS map which can be of variable length. The
616 * map is an array of CPUs.
617 */
618struct rps_map {
619 unsigned int len;
620 struct rcu_head rcu;
621 u16 cpus[0];
622};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000623#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000624
Tom Herbertfec5e652010-04-16 16:01:27 -0700625/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627 * tail pointer for that CPU's input queue at the time of last enqueue, and
628 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700629 */
630struct rps_dev_flow {
631 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000632 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700633 unsigned int last_qtail;
634};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000635#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700636
637/*
638 * The rps_dev_flow_table structure contains a table of flow mappings.
639 */
640struct rps_dev_flow_table {
641 unsigned int mask;
642 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700643 struct rps_dev_flow flows[0];
644};
645#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000646 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700647
648/*
649 * The rps_sock_flow_table contains mappings of flows to the last CPU
650 * on which they were processed by the application (set in recvmsg).
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500651 * Each entry is a 32bit value. Upper part is the high-order bits
652 * of flow hash, lower part is CPU number.
Eric Dumazet567e4b72015-02-06 12:59:01 -0800653 * rps_cpu_mask is used to partition the space, depending on number of
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500654 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
Eric Dumazet567e4b72015-02-06 12:59:01 -0800656 * meaning we use 32-6=26 bits for the hash.
Tom Herbertfec5e652010-04-16 16:01:27 -0700657 */
658struct rps_sock_flow_table {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800659 u32 mask;
Eric Dumazet93c1af62015-02-08 20:39:13 -0800660
661 u32 ents[0] ____cacheline_aligned_in_smp;
Tom Herbertfec5e652010-04-16 16:01:27 -0700662};
Eric Dumazet567e4b72015-02-06 12:59:01 -0800663#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
Tom Herbertfec5e652010-04-16 16:01:27 -0700664
665#define RPS_NO_CPU 0xffff
666
Eric Dumazet567e4b72015-02-06 12:59:01 -0800667extern u32 rps_cpu_mask;
668extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
669
Tom Herbertfec5e652010-04-16 16:01:27 -0700670static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
671 u32 hash)
672{
673 if (table && hash) {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800674 unsigned int index = hash & table->mask;
675 u32 val = hash & ~rps_cpu_mask;
Tom Herbertfec5e652010-04-16 16:01:27 -0700676
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500677 /* We only give a hint, preemption can change CPU under us */
Eric Dumazet567e4b72015-02-06 12:59:01 -0800678 val |= raw_smp_processor_id();
Tom Herbertfec5e652010-04-16 16:01:27 -0700679
Eric Dumazet567e4b72015-02-06 12:59:01 -0800680 if (table->ents[index] != val)
681 table->ents[index] = val;
Tom Herbertfec5e652010-04-16 16:01:27 -0700682 }
683}
684
Ben Hutchingsc4454772011-01-19 11:03:53 +0000685#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700686bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
687 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000688#endif
Michael Daltona953be52014-01-16 22:23:28 -0800689#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000690
Tom Herbert0a9627f2010-03-16 08:03:29 +0000691/* This structure contains an instance of an RX queue. */
692struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800693#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000694 struct rps_map __rcu *rps_map;
695 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800696#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000697 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000698 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000699} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800700
701/*
702 * RX queue sysfs structures and functions.
703 */
704struct rx_queue_attribute {
705 struct attribute attr;
706 ssize_t (*show)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, char *buf);
708 ssize_t (*store)(struct netdev_rx_queue *queue,
709 struct rx_queue_attribute *attr, const char *buf, size_t len);
710};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800711
Tom Herbertbf264142010-11-26 08:36:09 +0000712#ifdef CONFIG_XPS
713/*
714 * This structure holds an XPS map which can be of variable length. The
715 * map is an array of queues.
716 */
717struct xps_map {
718 unsigned int len;
719 unsigned int alloc_len;
720 struct rcu_head rcu;
721 u16 queues[0];
722};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000723#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Helge Dellerc59f4192015-10-25 10:00:32 +0100724#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725 - sizeof(struct xps_map)) / sizeof(u16))
Tom Herbertbf264142010-11-26 08:36:09 +0000726
727/*
728 * This structure holds all XPS maps for device. Maps are indexed by CPU.
729 */
730struct xps_dev_maps {
731 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000732 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000733};
734#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
735 (nr_cpu_ids * sizeof(struct xps_map *)))
736#endif /* CONFIG_XPS */
737
John Fastabend4f57c082011-01-17 08:06:04 +0000738#define TC_MAX_QUEUE 16
739#define TC_BITMASK 15
740/* HW offloaded queuing disciplines txq count and offset maps */
741struct netdev_tc_txq {
742 u16 count;
743 u16 offset;
744};
745
Neerav Parikh68bad942012-01-04 20:23:39 +0000746#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
747/*
748 * This structure is to hold information about the device
749 * configured to run FCoE protocol stack.
750 */
751struct netdev_fcoe_hbainfo {
752 char manufacturer[64];
753 char serial_number[64];
754 char hardware_version[64];
755 char driver_version[64];
756 char optionrom_version[64];
757 char firmware_version[64];
758 char model[256];
759 char model_description[256];
760};
761#endif
762
Jiri Pirko02637fc2014-11-28 14:34:16 +0100763#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200764
Jiri Pirko02637fc2014-11-28 14:34:16 +0100765/* This structure holds a unique identifier to identify some
766 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200767 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100768struct netdev_phys_item_id {
769 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200770 unsigned char id_len;
771};
772
Scott Feldmand754f982015-07-18 18:24:49 -0700773static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
774 struct netdev_phys_item_id *b)
775{
776 return a->id_len == b->id_len &&
777 memcmp(a->id, b->id, a->id_len) == 0;
778}
779
Daniel Borkmann99932d42014-02-16 15:55:20 +0100780typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
781 struct sk_buff *skb);
782
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800783/* These structures hold the attributes of qdisc and classifiers
John Fastabend16e5cc62016-02-16 21:16:43 -0800784 * that are being passed to the netdevice through the setup_tc op.
785 */
786enum {
787 TC_SETUP_MQPRIO,
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800788 TC_SETUP_CLSU32,
Amir Vadai5b33f482016-03-08 12:42:29 +0200789 TC_SETUP_CLSFLOWER,
Yotam Gigib87f7932016-07-21 12:03:12 +0200790 TC_SETUP_MATCHALL,
John Fastabend16e5cc62016-02-16 21:16:43 -0800791};
792
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800793struct tc_cls_u32_offload;
794
John Fastabend16e5cc62016-02-16 21:16:43 -0800795struct tc_to_netdev {
796 unsigned int type;
797 union {
798 u8 tc;
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800799 struct tc_cls_u32_offload *cls_u32;
Amir Vadai5b33f482016-03-08 12:42:29 +0200800 struct tc_cls_flower_offload *cls_flower;
Yotam Gigib87f7932016-07-21 12:03:12 +0200801 struct tc_cls_matchall_offload *cls_mall;
John Fastabend16e5cc62016-02-16 21:16:43 -0800802 };
803};
804
Brenden Blancoa7862b42016-07-19 12:16:48 -0700805/* These structures hold the attributes of xdp state that are being passed
806 * to the netdevice through the xdp op.
807 */
808enum xdp_netdev_command {
809 /* Set or clear a bpf program used in the earliest stages of packet
810 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
811 * is responsible for calling bpf_prog_put on any old progs that are
812 * stored. In case of error, the callee need not release the new prog
813 * reference, but on success it takes ownership and must bpf_prog_put
814 * when it is no longer used.
815 */
816 XDP_SETUP_PROG,
817 /* Check if a bpf program is set on the device. The callee should
818 * return true if a program is currently attached and running.
819 */
820 XDP_QUERY_PROG,
821};
822
823struct netdev_xdp {
824 enum xdp_netdev_command command;
825 union {
826 /* XDP_SETUP_PROG */
827 struct bpf_prog *prog;
828 /* XDP_QUERY_PROG */
829 bool prog_attached;
830 };
831};
John Fastabend16e5cc62016-02-16 21:16:43 -0800832
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800833/*
834 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800835 * The following hooks can be defined; unless noted otherwise, they are
836 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800837 *
838 * int (*ndo_init)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500839 * This function is called once when a network device is registered.
840 * The network device can use this for any late stage initialization
841 * or semantic validation. It can fail with an error code which will
842 * be propagated back to register_netdev.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800843 *
844 * void (*ndo_uninit)(struct net_device *dev);
845 * This function is called when device is unregistered or when registration
846 * fails. It is not called if init fails.
847 *
848 * int (*ndo_open)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500849 * This function is called when a network device transitions to the up
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800850 * state.
851 *
852 * int (*ndo_stop)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500853 * This function is called when a network device transitions to the down
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800854 * state.
855 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000856 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
857 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800858 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030859 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
860 * the queue before that can happen; it's for obsolete devices and weird
861 * corner cases, but the stack really does a non-trivial amount
862 * of useless work if you return NETDEV_TX_BUSY.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500863 * Required; cannot be NULL.
Stephen Hemminger00829822008-11-20 20:14:53 -0800864 *
Eric Dumazetcdba7562016-01-06 06:53:50 -0800865 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
866 * netdev_features_t features);
867 * Adjusts the requested feature flags according to device-specific
868 * constraints, and returns the resulting flags. Must not modify
869 * the device state.
870 *
Jason Wangf663dd92014-01-10 16:18:26 +0800871 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100872 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500873 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800874 * transmit queues.
875 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800876 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
877 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500878 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800879 *
880 * void (*ndo_set_rx_mode)(struct net_device *dev);
881 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000882 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500883 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800884 *
885 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
886 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700887 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500888 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800889 *
890 * int (*ndo_validate_addr)(struct net_device *dev);
891 * Test if Media Access Control address is valid for the device.
892 *
893 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500894 * Called when a user requests an ioctl which can't be handled by
895 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800896 * not supported error code.
897 *
898 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
899 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500900 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800901 * interface (PCI) for low level management.
902 *
903 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
904 * Called when a user wants to change the Maximum Transfer Unit
905 * of a device. If not defined, any request to change MTU will
906 * will return an error.
907 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800908 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500909 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800910 * for dev->watchdog ticks.
911 *
Ben Hutchings3cfde792010-07-09 09:11:52 +0000912 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
Eric Dumazet28172732010-07-07 14:58:56 -0700913 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700914 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800915 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000916 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000917 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
918 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700919 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000920 * (which should normally be dev->stats) and return a pointer to
921 * it. The structure may be changed asynchronously only if each
922 * field is written atomically.
923 * 3. Update dev->stats asynchronously and atomically, and define
924 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800925 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530926 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500927 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000928 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800929 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530930 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500931 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000932 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800933 *
934 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000935 *
936 * SR-IOV management functions.
937 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
938 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400939 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
940 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000941 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000942 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000943 * int (*ndo_get_vf_config)(struct net_device *dev,
944 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300945 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700946 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
947 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300948 *
949 * Enable or disable the VF ability to query its RSS Redirection Table and
950 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500951 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300952 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700953 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +0000954 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
955 * Called to setup 'tc' number of traffic classes in the net device. This
956 * is always called from the stack with the rtnl lock held and netif tx
957 * queues stopped. This allows the netdevice to perform queue management
958 * safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000959 *
Yi Zoue9bce842011-03-09 08:48:03 +0000960 * Fiber Channel over Ethernet (FCoE) offload functions.
961 * int (*ndo_fcoe_enable)(struct net_device *dev);
962 * Called when the FCoE protocol stack wants to start using LLD for FCoE
963 * so the underlying device can perform whatever needed configuration or
964 * initialization to support acceleration of FCoE traffic.
965 *
966 * int (*ndo_fcoe_disable)(struct net_device *dev);
967 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
968 * so the underlying device can perform whatever needed clean-ups to
969 * stop supporting acceleration of FCoE traffic.
970 *
971 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
972 * struct scatterlist *sgl, unsigned int sgc);
973 * Called when the FCoE Initiator wants to initialize an I/O that
974 * is a possible candidate for Direct Data Placement (DDP). The LLD can
975 * perform necessary setup and returns 1 to indicate the device is set up
976 * successfully to perform DDP on this I/O, otherwise this returns 0.
977 *
978 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
979 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
980 * indicated by the FC exchange id 'xid', so the underlying device can
981 * clean up and reuse resources for later DDP requests.
982 *
983 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
984 * struct scatterlist *sgl, unsigned int sgc);
985 * Called when the FCoE Target wants to initialize an I/O that
986 * is a possible candidate for Direct Data Placement (DDP). The LLD can
987 * perform necessary setup and returns 1 to indicate the device is set up
988 * successfully to perform DDP on this I/O, otherwise this returns 0.
989 *
Neerav Parikh68bad942012-01-04 20:23:39 +0000990 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
991 * struct netdev_fcoe_hbainfo *hbainfo);
992 * Called when the FCoE Protocol stack wants information on the underlying
993 * device. This information is utilized by the FCoE protocol stack to
994 * register attributes with Fiber Channel management service as per the
995 * FC-GS Fabric Device Management Information(FDMI) specification.
996 *
Yi Zoue9bce842011-03-09 08:48:03 +0000997 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
998 * Called when the underlying device wants to override default World Wide
999 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1000 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1001 * protocol stack to use.
1002 *
Ben Hutchingsc4454772011-01-19 11:03:53 +00001003 * RFS acceleration.
1004 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1005 * u16 rxq_index, u32 flow_id);
1006 * Set hardware filter for RFS. rxq_index is the target queue index;
1007 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1008 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001009 *
Jiri Pirko8b98a702013-01-03 22:49:02 +00001010 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001011 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1012 * Called to make another netdev an underling.
1013 *
1014 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1015 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +00001016 *
1017 * Feature/offload setting functions.
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001018 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001019 * Called to update device configuration to new features. Passed
1020 * feature set might be less than what was returned by ndo_fix_features()).
1021 * Must return >0 or -errno if it changed dev->features itself.
1022 *
stephen hemmingeredc7d572012-10-01 12:32:33 +00001023 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1024 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001025 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +00001026 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +00001027 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1028 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001029 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001030 * Deletes the FDB entry from dev coresponding to addr.
1031 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001032 * struct net_device *dev, struct net_device *filter_dev,
1033 * int idx)
John Fastabend77162022012-04-15 06:43:56 +00001034 * Used to add FDB entries to dump requests. Implementers should add
1035 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001036 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001037 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1038 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001039 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001040 * struct net_device *dev, u32 filter_mask,
1041 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001042 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1043 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001044 *
1045 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1046 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1047 * which do not represent real hardware may define this to allow their
1048 * userspace components to manage their virtual carrier state. Devices
1049 * that determine carrier state from physical hardware properties (eg
1050 * network cables) or protocol-dependent mechanisms (eg
1051 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001052 *
1053 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001054 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001055 * Called to get ID of physical port of this device. If driver does
1056 * not implement this, it is assumed that the hw is not able to have
1057 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001058 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001059 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1060 * struct udp_tunnel_info *ti);
1061 * Called by UDP tunnel to notify a driver about the UDP port and socket
1062 * address family that a UDP tunnel is listnening to. It is called only
1063 * when a new port starts listening. The operation is protected by the
1064 * RTNL.
1065 *
1066 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1067 * struct udp_tunnel_info *ti);
1068 * Called by UDP tunnel to notify the driver about a UDP port and socket
1069 * address family that the UDP tunnel is not listening to anymore. The
1070 * operation is protected by the RTNL.
1071 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001072 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1073 * struct net_device *dev)
1074 * Called by upper layer devices to accelerate switching or other
1075 * station functionality into hardware. 'pdev is the lowerdev
1076 * to use for the offload and 'dev' is the net device that will
1077 * back the offload. Returns a pointer to the private structure
1078 * the upper layer will maintain.
1079 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1080 * Called by upper layer device to delete the station created
1081 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1082 * the station and priv is the structure returned by the add
1083 * operation.
1084 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1085 * struct net_device *dev,
1086 * void *priv);
1087 * Callback to use for xmit over the accelerated station. This
1088 * is used in place of ndo_start_xmit on accelerated net
1089 * devices.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001090 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1091 * struct net_device *dev
1092 * netdev_features_t features);
Tom Herbert04ffcb22014-10-14 15:19:06 -07001093 * Called by core transmit path to determine if device is capable of
Jesse Gross5f352272014-12-23 22:37:26 -08001094 * performing offload operations on a given packet. This is to give
1095 * the device an opportunity to implement any restrictions that cannot
1096 * be otherwise expressed by feature flags. The check is called with
1097 * the set of features that the stack has calculated and it returns
1098 * those the driver believes to be appropriate.
John Fastabend822b3b22015-03-18 14:57:33 +02001099 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1100 * int queue_index, u32 maxrate);
1101 * Called when a user wants to set a max-rate limitation of specific
1102 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001103 * int (*ndo_get_iflink)(const struct net_device *dev);
1104 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001105 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001106 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001107 * This function is used to pass protocol port error state information
1108 * to the switch driver. The switch driver can react to the proto_down
1109 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001110 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1111 * This function is used to get egress tunnel information for given skb.
1112 * This is useful for retrieving outer tunnel header parameters while
1113 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001114 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1115 * This function is used to specify the headroom that the skb must
1116 * consider when allocation skb during packet reception. Setting
1117 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001118 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001119 * default value.
Brenden Blancoa7862b42016-07-19 12:16:48 -07001120 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1121 * This function is used to set or query state related to XDP on the
1122 * netdevice. See definition of enum xdp_netdev_command for details.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001123 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001124 */
1125struct net_device_ops {
1126 int (*ndo_init)(struct net_device *dev);
1127 void (*ndo_uninit)(struct net_device *dev);
1128 int (*ndo_open)(struct net_device *dev);
1129 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001130 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1131 struct net_device *dev);
1132 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1133 struct net_device *dev,
1134 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001135 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001136 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001137 void *accel_priv,
1138 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001139 void (*ndo_change_rx_flags)(struct net_device *dev,
1140 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001141 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001142 int (*ndo_set_mac_address)(struct net_device *dev,
1143 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001144 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001145 int (*ndo_do_ioctl)(struct net_device *dev,
1146 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001147 int (*ndo_set_config)(struct net_device *dev,
1148 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001149 int (*ndo_change_mtu)(struct net_device *dev,
1150 int new_mtu);
1151 int (*ndo_neigh_setup)(struct net_device *dev,
1152 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001153 void (*ndo_tx_timeout) (struct net_device *dev);
1154
Eric Dumazet28172732010-07-07 14:58:56 -07001155 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1156 struct rtnl_link_stats64 *storage);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001157 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1158
Jiri Pirko8e586132011-12-08 19:52:37 -05001159 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001160 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001161 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001162 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001163#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001164 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001165 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001166 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001167 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001168#endif
Cong Wange0d10952013-08-01 11:10:25 +08001169#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir8b80cda2013-07-10 17:13:26 +03001170 int (*ndo_busy_poll)(struct napi_struct *dev);
Eliezer Tamir06021292013-06-10 11:39:50 +03001171#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001172 int (*ndo_set_vf_mac)(struct net_device *dev,
1173 int queue, u8 *mac);
1174 int (*ndo_set_vf_vlan)(struct net_device *dev,
1175 int queue, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001176 int (*ndo_set_vf_rate)(struct net_device *dev,
1177 int vf, int min_tx_rate,
1178 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001179 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1180 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001181 int (*ndo_set_vf_trust)(struct net_device *dev,
1182 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001183 int (*ndo_get_vf_config)(struct net_device *dev,
1184 int vf,
1185 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001186 int (*ndo_set_vf_link_state)(struct net_device *dev,
1187 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001188 int (*ndo_get_vf_stats)(struct net_device *dev,
1189 int vf,
1190 struct ifla_vf_stats
1191 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001192 int (*ndo_set_vf_port)(struct net_device *dev,
1193 int vf,
1194 struct nlattr *port[]);
1195 int (*ndo_get_vf_port)(struct net_device *dev,
1196 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001197 int (*ndo_set_vf_guid)(struct net_device *dev,
1198 int vf, u64 guid,
1199 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001200 int (*ndo_set_vf_rss_query_en)(
1201 struct net_device *dev,
1202 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001203 int (*ndo_setup_tc)(struct net_device *dev,
1204 u32 handle,
1205 __be16 protocol,
1206 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001207#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001208 int (*ndo_fcoe_enable)(struct net_device *dev);
1209 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001210 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1211 u16 xid,
1212 struct scatterlist *sgl,
1213 unsigned int sgc);
1214 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1215 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001216 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1217 u16 xid,
1218 struct scatterlist *sgl,
1219 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001220 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1221 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001222#endif
1223
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001224#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001225#define NETDEV_FCOE_WWNN 0
1226#define NETDEV_FCOE_WWPN 1
1227 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1228 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001229#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001230
Ben Hutchingsc4454772011-01-19 11:03:53 +00001231#ifdef CONFIG_RFS_ACCEL
1232 int (*ndo_rx_flow_steer)(struct net_device *dev,
1233 const struct sk_buff *skb,
1234 u16 rxq_index,
1235 u32 flow_id);
1236#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001237 int (*ndo_add_slave)(struct net_device *dev,
1238 struct net_device *slave_dev);
1239 int (*ndo_del_slave)(struct net_device *dev,
1240 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001241 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1242 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001243 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001244 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001245 int (*ndo_neigh_construct)(struct net_device *dev,
1246 struct neighbour *n);
1247 void (*ndo_neigh_destroy)(struct net_device *dev,
1248 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001249
1250 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001251 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001252 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001253 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001254 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001255 u16 flags);
1256 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001257 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001258 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001259 const unsigned char *addr,
1260 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001261 int (*ndo_fdb_dump)(struct sk_buff *skb,
1262 struct netlink_callback *cb,
1263 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001264 struct net_device *filter_dev,
John Fastabend77162022012-04-15 06:43:56 +00001265 int idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001266
1267 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001268 struct nlmsghdr *nlh,
1269 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001270 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1271 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001272 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001273 u32 filter_mask,
1274 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001275 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001276 struct nlmsghdr *nlh,
1277 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001278 int (*ndo_change_carrier)(struct net_device *dev,
1279 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001280 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001281 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001282 int (*ndo_get_phys_port_name)(struct net_device *dev,
1283 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001284 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1285 struct udp_tunnel_info *ti);
1286 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1287 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001288 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1289 struct net_device *dev);
1290 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1291 void *priv);
1292
1293 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1294 struct net_device *dev,
1295 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001296 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001297 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1298 int queue_index,
1299 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001300 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001301 int (*ndo_change_proto_down)(struct net_device *dev,
1302 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001303 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1304 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001305 void (*ndo_set_rx_headroom)(struct net_device *dev,
1306 int needed_headroom);
Brenden Blancoa7862b42016-07-19 12:16:48 -07001307 int (*ndo_xdp)(struct net_device *dev,
1308 struct netdev_xdp *xdp);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001309};
1310
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001311/**
1312 * enum net_device_priv_flags - &struct net_device priv_flags
1313 *
1314 * These are the &struct net_device, they are only set internally
1315 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001316 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001317 * during any kernel release.
1318 *
1319 * You should have a pretty good reason to be extending these flags.
1320 *
1321 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1322 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001323 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001324 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001325 * @IFF_WAN_HDLC: WAN HDLC device
1326 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1327 * release skb->dst
1328 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1329 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1330 * @IFF_MACVLAN_PORT: device used as macvlan port
1331 * @IFF_BRIDGE_PORT: device used as bridge port
1332 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1333 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1334 * @IFF_UNICAST_FLT: Supports unicast filtering
1335 * @IFF_TEAM_PORT: device used as team port
1336 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1337 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1338 * change when it's running
1339 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001340 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1341 * underlying stacked devices
1342 * @IFF_IPVLAN_MASTER: IPvlan master device
1343 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001344 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001345 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001346 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001347 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001348 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001349 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001350 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1351 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001352 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001353 */
1354enum netdev_priv_flags {
1355 IFF_802_1Q_VLAN = 1<<0,
1356 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001357 IFF_BONDING = 1<<2,
1358 IFF_ISATAP = 1<<3,
1359 IFF_WAN_HDLC = 1<<4,
1360 IFF_XMIT_DST_RELEASE = 1<<5,
1361 IFF_DONT_BRIDGE = 1<<6,
1362 IFF_DISABLE_NETPOLL = 1<<7,
1363 IFF_MACVLAN_PORT = 1<<8,
1364 IFF_BRIDGE_PORT = 1<<9,
1365 IFF_OVS_DATAPATH = 1<<10,
1366 IFF_TX_SKB_SHARING = 1<<11,
1367 IFF_UNICAST_FLT = 1<<12,
1368 IFF_TEAM_PORT = 1<<13,
1369 IFF_SUPP_NOFCS = 1<<14,
1370 IFF_LIVE_ADDR_CHANGE = 1<<15,
1371 IFF_MACVLAN = 1<<16,
1372 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1373 IFF_IPVLAN_MASTER = 1<<18,
1374 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001375 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001376 IFF_NO_QUEUE = 1<<21,
1377 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001378 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001379 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001380 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001381 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001382 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001383};
1384
1385#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1386#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001387#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001388#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001389#define IFF_WAN_HDLC IFF_WAN_HDLC
1390#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1391#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1392#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1393#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1394#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1395#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1396#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1397#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1398#define IFF_TEAM_PORT IFF_TEAM_PORT
1399#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1400#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1401#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001402#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001403#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1404#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001405#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001406#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001407#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001408#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001409#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001410#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001411#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001412
Karoly Kemeny536721b2014-07-30 20:27:36 +02001413/**
1414 * struct net_device - The DEVICE structure.
1415 * Actually, this whole structure is a big mistake. It mixes I/O
1416 * data with strictly "high-level" data, and it has to know about
1417 * almost every data structure used in the INET module.
1418 *
1419 * @name: This is the first field of the "visible" part of this structure
1420 * (i.e. as seen by users in the "Space.c" file). It is the name
1421 * of the interface.
1422 *
1423 * @name_hlist: Device name hash chain, please keep it close to name[]
1424 * @ifalias: SNMP alias
1425 * @mem_end: Shared memory end
1426 * @mem_start: Shared memory start
1427 * @base_addr: Device I/O address
1428 * @irq: Device IRQ number
1429 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001430 * @carrier_changes: Stats to monitor carrier on<->off transitions
1431 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001432 * @state: Generic network queuing layer state, see netdev_state_t
1433 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001434 * @napi_list: List entry used for polling NAPI devices
1435 * @unreg_list: List entry when we are unregistering the
1436 * device; see the function unregister_netdev
1437 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001438 * @ptype_all: Device-specific packet handlers for all protocols
1439 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001440 *
1441 * @adj_list: Directly linked devices, like slaves for bonding
1442 * @all_adj_list: All linked devices, *including* neighbours
1443 * @features: Currently active device features
1444 * @hw_features: User-changeable features
1445 *
1446 * @wanted_features: User-requested features
1447 * @vlan_features: Mask of features inheritable by VLAN devices
1448 *
1449 * @hw_enc_features: Mask of features inherited by encapsulating devices
1450 * This field indicates what encapsulation
1451 * offloads the hardware is capable of doing,
1452 * and drivers will need to set them appropriately.
1453 *
1454 * @mpls_features: Mask of features inheritable by MPLS
1455 *
1456 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001457 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001458 *
1459 * @stats: Statistics struct, which was left as a legacy, use
1460 * rtnl_link_stats64 instead
1461 *
1462 * @rx_dropped: Dropped packets by core network,
1463 * do not use this in drivers
1464 * @tx_dropped: Dropped packets by core network,
1465 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001466 * @rx_nohandler: nohandler dropped packets by core network on
1467 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001468 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001469 * @wireless_handlers: List of functions to handle Wireless Extensions,
1470 * instead of ioctl,
1471 * see <net/iw_handler.h> for details.
1472 * @wireless_data: Instance data managed by the core of wireless extensions
1473 *
1474 * @netdev_ops: Includes several pointers to callbacks,
1475 * if one wants to override the ndo_*() functions
1476 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001477 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1478 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001479 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001480 * of Layer 2 headers.
1481 *
1482 * @flags: Interface flags (a la BSD)
1483 * @priv_flags: Like 'flags' but invisible to userspace,
1484 * see if.h for the definitions
1485 * @gflags: Global flags ( kept as legacy )
1486 * @padded: How much padding added by alloc_netdev()
1487 * @operstate: RFC2863 operstate
1488 * @link_mode: Mapping policy to operstate
1489 * @if_port: Selectable AUI, TP, ...
1490 * @dma: DMA channel
1491 * @mtu: Interface MTU value
1492 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001493 * @hard_header_len: Maximum hardware header length.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001494 *
1495 * @needed_headroom: Extra headroom the hardware may need, but not in all
1496 * cases can this be guaranteed
1497 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1498 * cases can this be guaranteed. Some cases also use
1499 * LL_MAX_HEADER instead to allocate the skb
1500 *
1501 * interface address info:
1502 *
1503 * @perm_addr: Permanent hw address
1504 * @addr_assign_type: Hw address assignment type
1505 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001506 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001507 * @dev_id: Used to differentiate devices that share
1508 * the same link layer address
1509 * @dev_port: Used to differentiate devices that share
1510 * the same function
1511 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001512 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001513 * has been enabled due to the need to listen to
1514 * additional unicast addresses in a device that
1515 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001516 * @uc: unicast mac addresses
1517 * @mc: multicast mac addresses
1518 * @dev_addrs: list of device hw addresses
1519 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001520 * @promiscuity: Number of times the NIC is told to work in
1521 * promiscuous mode; if it becomes 0 the NIC will
1522 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001523 * @allmulti: Counter, enables or disables allmulticast mode
1524 *
1525 * @vlan_info: VLAN info
1526 * @dsa_ptr: dsa specific data
1527 * @tipc_ptr: TIPC specific data
1528 * @atalk_ptr: AppleTalk link
1529 * @ip_ptr: IPv4 specific data
1530 * @dn_ptr: DECnet specific data
1531 * @ip6_ptr: IPv6 specific data
1532 * @ax25_ptr: AX.25 specific data
1533 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1534 *
1535 * @last_rx: Time of last Rx
1536 * @dev_addr: Hw address (before bcast,
1537 * because most packets are unicast)
1538 *
1539 * @_rx: Array of RX queues
1540 * @num_rx_queues: Number of RX queues
1541 * allocated at register_netdev() time
1542 * @real_num_rx_queues: Number of RX queues currently active in device
1543 *
1544 * @rx_handler: handler for received packets
1545 * @rx_handler_data: XXX: need comments on this one
1546 * @ingress_queue: XXX: need comments on this one
1547 * @broadcast: hw bcast address
1548 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001549 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1550 * indexed by RX queue number. Assigned by driver.
1551 * This must only be set if the ndo_rx_flow_steer
1552 * operation is defined
1553 * @index_hlist: Device index hash chain
1554 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001555 * @_tx: Array of TX queues
1556 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1557 * @real_num_tx_queues: Number of TX queues currently active in device
1558 * @qdisc: Root qdisc from userspace point of view
1559 * @tx_queue_len: Max frames per queue allowed
1560 * @tx_global_lock: XXX: need comments on this one
1561 *
1562 * @xps_maps: XXX: need comments on this one
1563 *
Scott Feldman0c4f6912015-07-18 18:24:48 -07001564 * @offload_fwd_mark: Offload device fwding mark
1565 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001566 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001567 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001568 * @watchdog_timer: List of timers
1569 *
1570 * @pcpu_refcnt: Number of references to this device
1571 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001572 * @link_watch_list: XXX: need comments on this one
1573 *
1574 * @reg_state: Register/unregister state machine
1575 * @dismantle: Device is going to be freed
1576 * @rtnl_link_state: This enum represents the phases of creating
1577 * a new link
1578 *
1579 * @destructor: Called from unregister,
1580 * can be used to call free_netdev
1581 * @npinfo: XXX: need comments on this one
1582 * @nd_net: Network namespace this network device is inside
1583 *
1584 * @ml_priv: Mid-layer private
1585 * @lstats: Loopback statistics
1586 * @tstats: Tunnel statistics
1587 * @dstats: Dummy statistics
1588 * @vstats: Virtual ethernet statistics
1589 *
1590 * @garp_port: GARP
1591 * @mrp_port: MRP
1592 *
1593 * @dev: Class/net/name entry
1594 * @sysfs_groups: Space for optional device, statistics and wireless
1595 * sysfs groups
1596 *
1597 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1598 * @rtnl_link_ops: Rtnl_link_ops
1599 *
1600 * @gso_max_size: Maximum size of generic segmentation offload
1601 * @gso_max_segs: Maximum number of segments that can be passed to the
1602 * NIC for GSO
1603 *
1604 * @dcbnl_ops: Data Center Bridging netlink ops
1605 * @num_tc: Number of traffic classes in the net device
1606 * @tc_to_txq: XXX: need comments on this one
1607 * @prio_tc_map XXX: need comments on this one
1608 *
1609 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1610 *
1611 * @priomap: XXX: need comments on this one
1612 * @phydev: Physical device may attach itself
1613 * for hardware timestamping
1614 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001615 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1616 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001617 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001618 * @proto_down: protocol port state information can be sent to the
1619 * switch driver and used to set the phys state of the
1620 * switch port.
1621 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 * FIXME: cleanup struct net_device such that network protocol info
1623 * moves out.
1624 */
1625
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001626struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001628 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001629 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 /*
1631 * I/O specific fields
1632 * FIXME: Merge these and struct ifmap into one
1633 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001634 unsigned long mem_end;
1635 unsigned long mem_start;
1636 unsigned long base_addr;
1637 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001639 atomic_t carrier_changes;
1640
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001642 * Some hardware also needs these fields (state,dev_list,
1643 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644 * part of the usual set specified in Space.c.
1645 */
1646
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 unsigned long state;
1648
Pavel Emelianov7562f872007-05-03 15:13:45 -07001649 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001650 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001651 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001652 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001653 struct list_head ptype_all;
1654 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001655
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001656 struct {
1657 struct list_head upper;
1658 struct list_head lower;
1659 } adj_list;
1660
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001661 struct {
1662 struct list_head upper;
1663 struct list_head lower;
1664 } all_adj_list;
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001665
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001666 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001667 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001668 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001669 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001670 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001671 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001672 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001673
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001675 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676
Rusty Russellc45d2862007-03-28 14:29:08 -07001677 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001678
Eric Dumazet015f0682014-03-27 08:45:56 -07001679 atomic_long_t rx_dropped;
1680 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001681 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682
Johannes Bergb86e0282007-04-26 20:48:23 -07001683#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001684 const struct iw_handler_def *wireless_handlers;
1685 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001686#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001687 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001688 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001689#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001690 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001691#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001692#ifdef CONFIG_NET_L3_MASTER_DEV
1693 const struct l3mdev_ops *l3mdev_ops;
1694#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001695#if IS_ENABLED(CONFIG_IPV6)
1696 const struct ndisc_ops *ndisc_ops;
1697#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001699 const struct header_ops *header_ops;
1700
Karoly Kemeny536721b2014-07-30 20:27:36 +02001701 unsigned int flags;
1702 unsigned int priv_flags;
1703
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001705 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706
Karoly Kemeny536721b2014-07-30 20:27:36 +02001707 unsigned char operstate;
1708 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001709
Karoly Kemeny536721b2014-07-30 20:27:36 +02001710 unsigned char if_port;
1711 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001712
Karoly Kemeny536721b2014-07-30 20:27:36 +02001713 unsigned int mtu;
1714 unsigned short type;
1715 unsigned short hard_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716
Johannes Bergf5184d22008-05-12 20:48:31 -07001717 unsigned short needed_headroom;
1718 unsigned short needed_tailroom;
1719
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001721 unsigned char perm_addr[MAX_ADDR_LEN];
1722 unsigned char addr_assign_type;
1723 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001724 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001725 unsigned short dev_id;
1726 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001727 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001728 unsigned char name_assign_type;
1729 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001730 struct netdev_hw_addr_list uc;
1731 struct netdev_hw_addr_list mc;
1732 struct netdev_hw_addr_list dev_addrs;
1733
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001734#ifdef CONFIG_SYSFS
1735 struct kset *queues_kset;
1736#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001737 unsigned int promiscuity;
1738 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001741 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001742
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001743#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001744 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001745#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001746#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001747 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001748#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001749#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001750 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001751#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001752 void *atalk_ptr;
1753 struct in_device __rcu *ip_ptr;
1754 struct dn_dev __rcu *dn_ptr;
1755 struct inet6_dev __rcu *ip6_ptr;
1756 void *ax25_ptr;
1757 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001758 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001759#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1760 struct mpls_dev __rcu *mpls_ptr;
1761#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001763/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001764 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001765 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001766 unsigned long last_rx;
Eric Dumazet4dc89132010-08-31 07:40:16 +00001767
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001768 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001769 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001770
Michael Daltona953be52014-01-16 22:23:28 -08001771#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001772 struct netdev_rx_queue *_rx;
1773
Tom Herbert0a9627f2010-03-16 08:03:29 +00001774 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001775 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001776#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001777
Eric Dumazet3b47d302014-11-06 21:09:44 -08001778 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001779 rx_handler_func_t __rcu *rx_handler;
1780 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001781
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001782#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001783 struct tcf_proto __rcu *ingress_cl_list;
1784#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001785 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001786#ifdef CONFIG_NETFILTER_INGRESS
1787 struct list_head nf_hooks_ingress;
1788#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001789
Karoly Kemeny536721b2014-07-30 20:27:36 +02001790 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001791#ifdef CONFIG_RFS_ACCEL
1792 struct cpu_rmap *rx_cpu_rmap;
1793#endif
1794 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001795
1796/*
1797 * Cache lines mostly used on transmit path
1798 */
David S. Millere8a04642008-07-17 00:34:19 -07001799 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1800 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001801 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001802 struct Qdisc *qdisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001803 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001804 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001805 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001806
Tom Herbertbf264142010-11-26 08:36:09 +00001807#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001808 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001809#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001810#ifdef CONFIG_NET_CLS_ACT
1811 struct tcf_proto __rcu *egress_cl_list;
1812#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001813#ifdef CONFIG_NET_SWITCHDEV
1814 u32 offload_fwd_mark;
1815#endif
1816
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001817 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001818 struct timer_list watchdog_timer;
1819
Eric Dumazet29b44332010-10-11 10:22:12 +00001820 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822
Eric Dumazete014deb2009-11-17 05:59:21 +00001823 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001824
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001826 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827 NETREG_UNREGISTERING, /* called unregister_netdevice */
1828 NETREG_UNREGISTERED, /* completed unregister todo */
1829 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001830 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001831 } reg_state:8;
1832
Karoly Kemeny536721b2014-07-30 20:27:36 +02001833 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001834
1835 enum {
1836 RTNL_LINK_INITIALIZED,
1837 RTNL_LINK_INITIALIZING,
1838 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001840 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001841
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001843 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844#endif
David S. Millereae792b2008-07-15 03:03:33 -07001845
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001846 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001847
David S. Miller49517042008-05-12 03:29:11 -07001848 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001849 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001850 void *ml_priv;
1851 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001852 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001853 struct pcpu_dstats __percpu *dstats;
1854 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001855 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001856
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001857 struct garp_port __rcu *garp_port;
David Wardfebf0182013-02-08 17:17:06 +00001858 struct mrp_port __rcu *mrp_port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001860 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001861 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001862 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001863
Patrick McHardy38f7b872007-06-13 12:03:51 -07001864 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001865
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001866 /* for setting kernel sock attribute on TCP connection setup */
1867#define GSO_MAX_SIZE 65536
1868 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001869#define GSO_MAX_SEGS 65535
1870 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001871
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001872#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001873 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001874#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001875 u8 num_tc;
1876 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1877 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001878
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001879#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001880 unsigned int fcoe_ddp_xid;
1881#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001882#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001883 struct netprio_map __rcu *priomap;
1884#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001885 struct phy_device *phydev;
1886 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001887 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001888 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001890#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891
1892#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893
David S. Millere8a04642008-07-17 00:34:19 -07001894static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001895int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1896{
1897 return dev->prio_tc_map[prio & TC_BITMASK];
1898}
1899
1900static inline
1901int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1902{
1903 if (tc >= dev->num_tc)
1904 return -EINVAL;
1905
1906 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1907 return 0;
1908}
1909
1910static inline
1911void netdev_reset_tc(struct net_device *dev)
1912{
1913 dev->num_tc = 0;
1914 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1915 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1916}
1917
1918static inline
1919int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1920{
1921 if (tc >= dev->num_tc)
1922 return -EINVAL;
1923
1924 dev->tc_to_txq[tc].count = count;
1925 dev->tc_to_txq[tc].offset = offset;
1926 return 0;
1927}
1928
1929static inline
1930int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1931{
1932 if (num_tc > TC_MAX_QUEUE)
1933 return -EINVAL;
1934
1935 dev->num_tc = num_tc;
1936 return 0;
1937}
1938
1939static inline
1940int netdev_get_num_tc(struct net_device *dev)
1941{
1942 return dev->num_tc;
1943}
1944
1945static inline
David S. Millere8a04642008-07-17 00:34:19 -07001946struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1947 unsigned int index)
1948{
1949 return &dev->_tx[index];
1950}
1951
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001952static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1953 const struct sk_buff *skb)
1954{
1955 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1956}
1957
David S. Millere8a04642008-07-17 00:34:19 -07001958static inline void netdev_for_each_tx_queue(struct net_device *dev,
1959 void (*f)(struct net_device *,
1960 struct netdev_queue *,
1961 void *),
1962 void *arg)
1963{
1964 unsigned int i;
1965
1966 for (i = 0; i < dev->num_tx_queues; i++)
1967 f(dev, &dev->_tx[i], arg);
1968}
1969
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001970#define netdev_lockdep_set_classes(dev) \
1971{ \
1972 static struct lock_class_key qdisc_tx_busylock_key; \
1973 static struct lock_class_key qdisc_running_key; \
1974 static struct lock_class_key qdisc_xmit_lock_key; \
1975 static struct lock_class_key dev_addr_list_lock_key; \
1976 unsigned int i; \
1977 \
1978 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1979 (dev)->qdisc_running_key = &qdisc_running_key; \
1980 lockdep_set_class(&(dev)->addr_list_lock, \
1981 &dev_addr_list_lock_key); \
1982 for (i = 0; i < (dev)->num_tx_queues; i++) \
1983 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1984 &qdisc_xmit_lock_key); \
1985}
1986
Joe Perchesf629d202013-09-26 14:48:15 -07001987struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001988 struct sk_buff *skb,
1989 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001990
Paolo Abeni871b6422016-02-26 10:45:37 +01001991/* returns the headroom that the master device needs to take in account
1992 * when forwarding to this dev
1993 */
1994static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1995{
1996 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1997}
1998
1999static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2000{
2001 if (dev->netdev_ops->ndo_set_rx_headroom)
2002 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2003}
2004
2005/* set the device rx headroom to the dev's default */
2006static inline void netdev_reset_rx_headroom(struct net_device *dev)
2007{
2008 netdev_set_rx_headroom(dev, -1);
2009}
2010
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002011/*
2012 * Net namespace inlines
2013 */
2014static inline
2015struct net *dev_net(const struct net_device *dev)
2016{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002017 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002018}
2019
2020static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002021void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002022{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05002023 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002024}
2025
Florian Fainelli3e8a72d2014-08-27 17:04:46 -07002026static inline bool netdev_uses_dsa(struct net_device *dev)
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00002027{
Alexander Duyck3fc88672014-09-12 23:12:46 -07002028#if IS_ENABLED(CONFIG_NET_DSA)
Florian Fainelli5aed85c2014-08-27 17:04:52 -07002029 if (dev->dsa_ptr != NULL)
2030 return dsa_uses_tagged_protocol(dev->dsa_ptr);
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00002031#endif
Florian Fainelli5aed85c2014-08-27 17:04:52 -07002032 return false;
Lennert Buytenhek396138f2008-10-07 13:46:07 +00002033}
2034
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002035/**
2036 * netdev_priv - access network device private data
2037 * @dev: network device
2038 *
2039 * Get network device private data
2040 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002041static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002043 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044}
2045
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046/* Set the sysfs physical device reference for the network logical device
2047 * if set prior to registration will cause a symlink during initialization.
2048 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002049#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050
Marcel Holtmann384912e2009-08-31 21:08:19 +00002051/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002052 * fine-grained identification of different network device types. For
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002053 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
Marcel Holtmann384912e2009-08-31 21:08:19 +00002054 */
2055#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2056
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002057/* Default NAPI poll() weight
2058 * Device drivers are strongly advised to not use bigger value
2059 */
2060#define NAPI_POLL_WEIGHT 64
2061
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002062/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002063 * netif_napi_add - initialize a NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002064 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002065 * @napi: NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002066 * @poll: polling function
2067 * @weight: default weight
2068 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002069 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2070 * *any* of the other NAPI-related functions.
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002071 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002072void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2073 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002074
Alexander Duyckd8156532008-07-08 15:13:05 -07002075/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002076 * netif_tx_napi_add - initialize a NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002077 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002078 * @napi: NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002079 * @poll: polling function
2080 * @weight: default weight
2081 *
2082 * This variant of netif_napi_add() should be used from drivers using NAPI
2083 * to exclusively poll a TX queue.
2084 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2085 */
2086static inline void netif_tx_napi_add(struct net_device *dev,
2087 struct napi_struct *napi,
2088 int (*poll)(struct napi_struct *, int),
2089 int weight)
2090{
2091 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2092 netif_napi_add(dev, napi, poll, weight);
2093}
2094
2095/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002096 * netif_napi_del - remove a NAPI context
2097 * @napi: NAPI context
Alexander Duyckd8156532008-07-08 15:13:05 -07002098 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002099 * netif_napi_del() removes a NAPI context from the network device NAPI list
Alexander Duyckd8156532008-07-08 15:13:05 -07002100 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002101void netif_napi_del(struct napi_struct *napi);
2102
2103struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002104 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002105 void *frag0;
Herbert Xu78a478d2009-05-26 18:50:21 +00002106
Herbert Xu74895942009-05-26 18:50:27 +00002107 /* Length of frag0. */
2108 unsigned int frag0_len;
2109
Herbert Xu86911732009-01-29 14:19:50 +00002110 /* This indicates where we are processing relative to skb->data. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002111 int data_offset;
Herbert Xu86911732009-01-29 14:19:50 +00002112
Herbert Xud565b0a2008-12-15 23:38:52 -08002113 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002114 u16 flush;
2115
2116 /* Save the IP ID here and check when we get to the transport layer */
2117 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002118
2119 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002120 u16 count;
2121
Tom Herbert15e23962015-02-10 16:30:31 -08002122 /* Start offset for remote checksum offload */
2123 u16 gro_remcsum_start;
2124
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002125 /* jiffies when first packet was created/queued */
2126 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002127
Tom Herbertafe93322014-09-17 12:25:57 -07002128 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002129 u16 proto;
2130
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002131 /* This is non-zero if the packet may be of the same flow. */
2132 u8 same_flow:1;
2133
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002134 /* Used in tunnel GRO receive */
2135 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002136
2137 /* GRO checksum is valid */
2138 u8 csum_valid:1;
2139
Tom Herbert662880f2014-08-27 21:26:56 -07002140 /* Number of checksums via CHECKSUM_UNNECESSARY */
2141 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002142
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002143 /* Free the skb? */
2144 u8 free:2;
2145#define NAPI_GRO_FREE 1
2146#define NAPI_GRO_FREE_STOLEN_HEAD 2
2147
Tom Herbertefc98d02014-10-03 15:48:08 -07002148 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2149 u8 is_ipv6:1;
2150
Alexander Duycka0ca1532016-04-05 09:13:39 -07002151 /* Used in GRE, set in fou/gue_gro_receive */
2152 u8 is_fou:1;
2153
Alexander Duyck15305452016-04-10 21:44:57 -04002154 /* Used to determine if flush_id can be ignored */
2155 u8 is_atomic:1;
2156
2157 /* 5 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002158
Jerry Chubf5a7552014-01-07 10:23:19 -08002159 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2160 __wsum csum;
2161
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002162 /* used in skb_gro_receive() slow path */
2163 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002164};
2165
2166#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002167
Linus Torvalds1da177e2005-04-16 15:20:36 -07002168struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002169 __be16 type; /* This is really htons(ether_type). */
2170 struct net_device *dev; /* NULL is wildcarded here */
2171 int (*func) (struct sk_buff *,
2172 struct net_device *,
2173 struct packet_type *,
2174 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002175 bool (*id_match)(struct packet_type *ptype,
2176 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177 void *af_packet_priv;
2178 struct list_head list;
2179};
2180
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002181struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002182 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2183 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002185 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002186 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002187};
2188
2189struct packet_offload {
2190 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002191 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002192 struct offload_callbacks callbacks;
2193 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194};
2195
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002196/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002197struct pcpu_sw_netstats {
2198 u64 rx_packets;
2199 u64 rx_bytes;
2200 u64 tx_packets;
2201 u64 tx_bytes;
2202 struct u64_stats_sync syncp;
2203};
2204
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002205#define __netdev_alloc_pcpu_stats(type, gfp) \
2206({ \
2207 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2208 if (pcpu_stats) { \
2209 int __cpu; \
2210 for_each_possible_cpu(__cpu) { \
2211 typeof(type) *stat; \
2212 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2213 u64_stats_init(&stat->syncp); \
2214 } \
2215 } \
2216 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002217})
2218
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002219#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002220 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002221
Jiri Pirko764f5e52015-12-03 12:12:12 +01002222enum netdev_lag_tx_type {
2223 NETDEV_LAG_TX_TYPE_UNKNOWN,
2224 NETDEV_LAG_TX_TYPE_RANDOM,
2225 NETDEV_LAG_TX_TYPE_BROADCAST,
2226 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2227 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2228 NETDEV_LAG_TX_TYPE_HASH,
2229};
2230
2231struct netdev_lag_upper_info {
2232 enum netdev_lag_tx_type tx_type;
2233};
2234
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002235struct netdev_lag_lower_state_info {
2236 u8 link_up : 1,
2237 tx_enabled : 1;
2238};
2239
Linus Torvalds1da177e2005-04-16 15:20:36 -07002240#include <linux/notifier.h>
2241
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002242/* netdevice notifier chain. Please remember to update the rtnetlink
2243 * notification exclusion list in rtnetlink_event() when adding new
2244 * types.
2245 */
2246#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2247#define NETDEV_DOWN 0x0002
2248#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2249 detected a hardware crash and restarted
2250 - we can use this eg to kick tcp sessions
2251 once done */
2252#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2253#define NETDEV_REGISTER 0x0005
2254#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002255#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002256#define NETDEV_CHANGEADDR 0x0008
2257#define NETDEV_GOING_DOWN 0x0009
2258#define NETDEV_CHANGENAME 0x000A
2259#define NETDEV_FEAT_CHANGE 0x000B
2260#define NETDEV_BONDING_FAILOVER 0x000C
2261#define NETDEV_PRE_UP 0x000D
2262#define NETDEV_PRE_TYPE_CHANGE 0x000E
2263#define NETDEV_POST_TYPE_CHANGE 0x000F
2264#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002265#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002266#define NETDEV_RELEASE 0x0012
2267#define NETDEV_NOTIFY_PEERS 0x0013
2268#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002269#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002270#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002271#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002272#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002273#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002274#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002275#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002276#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Jason Wang08294a22016-06-30 14:45:35 +08002277#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002278
Joe Perchesf629d202013-09-26 14:48:15 -07002279int register_netdevice_notifier(struct notifier_block *nb);
2280int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002281
2282struct netdev_notifier_info {
2283 struct net_device *dev;
2284};
2285
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002286struct netdev_notifier_change_info {
2287 struct netdev_notifier_info info; /* must be first */
2288 unsigned int flags_changed;
2289};
2290
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002291struct netdev_notifier_changeupper_info {
2292 struct netdev_notifier_info info; /* must be first */
2293 struct net_device *upper_dev; /* new upper dev */
2294 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002295 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002296 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002297};
2298
Jiri Pirko04d48262015-12-03 12:12:15 +01002299struct netdev_notifier_changelowerstate_info {
2300 struct netdev_notifier_info info; /* must be first */
2301 void *lower_state_info; /* is lower dev state */
2302};
2303
Cong Wang75538c22013-05-29 11:30:50 +08002304static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2305 struct net_device *dev)
2306{
2307 info->dev = dev;
2308}
2309
Jiri Pirko351638e2013-05-28 01:30:21 +00002310static inline struct net_device *
2311netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2312{
2313 return info->dev;
2314}
2315
Joe Perchesf629d202013-09-26 14:48:15 -07002316int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002317
2318
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319extern rwlock_t dev_base_lock; /* Device list lock */
2320
Eric W. Biederman881d9662007-09-17 11:56:21 -07002321#define for_each_netdev(net, d) \
2322 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002323#define for_each_netdev_reverse(net, d) \
2324 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002325#define for_each_netdev_rcu(net, d) \
2326 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002327#define for_each_netdev_safe(net, d, n) \
2328 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2329#define for_each_netdev_continue(net, d) \
2330 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002331#define for_each_netdev_continue_rcu(net, d) \
2332 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002333#define for_each_netdev_in_bond_rcu(bond, slave) \
2334 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002335 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002336#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2337
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002338static inline struct net_device *next_net_device(struct net_device *dev)
2339{
2340 struct list_head *lh;
2341 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002342
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002343 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002344 lh = dev->dev_list.next;
2345 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2346}
2347
Eric Dumazetce81b762009-11-11 17:34:30 +00002348static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2349{
2350 struct list_head *lh;
2351 struct net *net;
2352
2353 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002354 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002355 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2356}
2357
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002358static inline struct net_device *first_net_device(struct net *net)
2359{
2360 return list_empty(&net->dev_base_head) ? NULL :
2361 net_device_entry(net->dev_base_head.next);
2362}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002363
Eric Dumazetccf43432011-01-26 18:08:02 +00002364static inline struct net_device *first_net_device_rcu(struct net *net)
2365{
2366 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2367
2368 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2369}
2370
Joe Perchesf629d202013-09-26 14:48:15 -07002371int netdev_boot_setup_check(struct net_device *dev);
2372unsigned long netdev_boot_base(const char *prefix, int unit);
2373struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2374 const char *hwaddr);
2375struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2376struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2377void dev_add_pack(struct packet_type *pt);
2378void dev_remove_pack(struct packet_type *pt);
2379void __dev_remove_pack(struct packet_type *pt);
2380void dev_add_offload(struct packet_offload *po);
2381void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002383int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002384int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002385struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2386 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002387struct net_device *dev_get_by_name(struct net *net, const char *name);
2388struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2389struct net_device *__dev_get_by_name(struct net *net, const char *name);
2390int dev_alloc_name(struct net_device *dev, const char *name);
2391int dev_open(struct net_device *dev);
2392int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002393int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002394void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002395int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002396int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002397int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002398int register_netdevice(struct net_device *dev);
2399void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2400void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002401static inline void unregister_netdevice(struct net_device *dev)
2402{
2403 unregister_netdevice_queue(dev, NULL);
2404}
2405
Joe Perchesf629d202013-09-26 14:48:15 -07002406int netdev_refcnt_read(const struct net_device *dev);
2407void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002408void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002409void synchronize_net(void);
2410int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002411
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002412DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002413#define XMIT_RECURSION_LIMIT 10
2414
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002415static inline int dev_recursion_level(void)
2416{
2417 return this_cpu_read(xmit_recursion);
2418}
2419
Joe Perchesf629d202013-09-26 14:48:15 -07002420struct net_device *dev_get_by_index(struct net *net, int ifindex);
2421struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2422struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2423int netdev_get_name(struct net *net, char *name, int ifindex);
2424int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002425int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002426
2427static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2428{
2429 return NAPI_GRO_CB(skb)->data_offset;
2430}
2431
2432static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2433{
2434 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2435}
2436
2437static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2438{
2439 NAPI_GRO_CB(skb)->data_offset += len;
2440}
2441
Herbert Xua5b1cf22009-05-26 18:50:28 +00002442static inline void *skb_gro_header_fast(struct sk_buff *skb,
2443 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002444{
Herbert Xu78a478d2009-05-26 18:50:21 +00002445 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002446}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447
Herbert Xua5b1cf22009-05-26 18:50:28 +00002448static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2449{
2450 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2451}
2452
2453static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2454 unsigned int offset)
2455{
Herbert Xu17dd7592011-07-27 06:16:28 -07002456 if (!pskb_may_pull(skb, hlen))
2457 return NULL;
2458
Herbert Xua5b1cf22009-05-26 18:50:28 +00002459 NAPI_GRO_CB(skb)->frag0 = NULL;
2460 NAPI_GRO_CB(skb)->frag0_len = 0;
Herbert Xu17dd7592011-07-27 06:16:28 -07002461 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002462}
2463
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002464static inline void *skb_gro_network_header(struct sk_buff *skb)
2465{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002466 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2467 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002468}
2469
Jerry Chubf5a7552014-01-07 10:23:19 -08002470static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2471 const void *start, unsigned int len)
2472{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002473 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002474 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2475 csum_partial(start, len, 0));
2476}
2477
Tom Herbert573e8fc2014-08-22 13:33:47 -07002478/* GRO checksum functions. These are logical equivalents of the normal
2479 * checksum functions (in skbuff.h) except that they operate on the GRO
2480 * offsets and fields in sk_buff.
2481 */
2482
2483__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2484
Tom Herbert15e23962015-02-10 16:30:31 -08002485static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2486{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002487 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002488}
2489
Tom Herbert573e8fc2014-08-22 13:33:47 -07002490static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2491 bool zero_okay,
2492 __sum16 check)
2493{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002494 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2495 skb_checksum_start_offset(skb) <
2496 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002497 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002498 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002499 (!zero_okay || check));
2500}
2501
2502static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2503 __wsum psum)
2504{
2505 if (NAPI_GRO_CB(skb)->csum_valid &&
2506 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2507 return 0;
2508
2509 NAPI_GRO_CB(skb)->csum = psum;
2510
2511 return __skb_gro_checksum_complete(skb);
2512}
2513
Tom Herbert573e8fc2014-08-22 13:33:47 -07002514static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2515{
Tom Herbert662880f2014-08-27 21:26:56 -07002516 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2517 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2518 NAPI_GRO_CB(skb)->csum_cnt--;
2519 } else {
2520 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2521 * verified a new top level checksum or an encapsulated one
2522 * during GRO. This saves work if we fallback to normal path.
2523 */
2524 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002525 }
2526}
2527
2528#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2529 compute_pseudo) \
2530({ \
2531 __sum16 __ret = 0; \
2532 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2533 __ret = __skb_gro_checksum_validate_complete(skb, \
2534 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002535 if (__ret) \
2536 __skb_mark_checksum_bad(skb); \
2537 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002538 skb_gro_incr_csum_unnecessary(skb); \
2539 __ret; \
2540})
2541
2542#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2543 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2544
2545#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2546 compute_pseudo) \
2547 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2548
2549#define skb_gro_checksum_simple_validate(skb) \
2550 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2551
Tom Herbertd96535a2014-08-31 15:12:42 -07002552static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2553{
2554 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2555 !NAPI_GRO_CB(skb)->csum_valid);
2556}
2557
2558static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2559 __sum16 check, __wsum pseudo)
2560{
2561 NAPI_GRO_CB(skb)->csum = ~pseudo;
2562 NAPI_GRO_CB(skb)->csum_valid = 1;
2563}
2564
2565#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2566do { \
2567 if (__skb_gro_checksum_convert_check(skb)) \
2568 __skb_gro_checksum_convert(skb, check, \
2569 compute_pseudo(skb, proto)); \
2570} while (0)
2571
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002572struct gro_remcsum {
2573 int offset;
2574 __wsum delta;
2575};
2576
2577static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2578{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002579 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002580 grc->delta = 0;
2581}
2582
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002583static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2584 unsigned int off, size_t hdrlen,
2585 int start, int offset,
2586 struct gro_remcsum *grc,
2587 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002588{
2589 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002590 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002591
2592 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2593
Tom Herbert15e23962015-02-10 16:30:31 -08002594 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002595 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2596 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002597 }
2598
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002599 ptr = skb_gro_header_fast(skb, off);
2600 if (skb_gro_header_hard(skb, off + plen)) {
2601 ptr = skb_gro_header_slow(skb, off + plen, off);
2602 if (!ptr)
2603 return NULL;
2604 }
2605
2606 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2607 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002608
2609 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002610 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002611
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002612 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002613 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002614
2615 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002616}
2617
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002618static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2619 struct gro_remcsum *grc)
2620{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002621 void *ptr;
2622 size_t plen = grc->offset + sizeof(u16);
2623
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002624 if (!grc->delta)
2625 return;
2626
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002627 ptr = skb_gro_header_fast(skb, grc->offset);
2628 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2629 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2630 if (!ptr)
2631 return;
2632 }
2633
2634 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002635}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002636
Tom Herbert6ae23ad2015-12-14 11:19:46 -08002637struct skb_csum_offl_spec {
2638 __u16 ipv4_okay:1,
2639 ipv6_okay:1,
2640 encap_okay:1,
2641 ip_options_okay:1,
2642 ext_hdrs_okay:1,
2643 tcp_okay:1,
2644 udp_okay:1,
2645 sctp_okay:1,
2646 vlan_okay:1,
2647 no_encapped_ipv6:1,
2648 no_not_encapped:1;
2649};
2650
2651bool __skb_csum_offload_chk(struct sk_buff *skb,
2652 const struct skb_csum_offl_spec *spec,
2653 bool *csum_encapped,
2654 bool csum_help);
2655
2656static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2657 const struct skb_csum_offl_spec *spec,
2658 bool *csum_encapped,
2659 bool csum_help)
2660{
2661 if (skb->ip_summed != CHECKSUM_PARTIAL)
2662 return false;
2663
2664 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2665}
2666
2667static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2668 const struct skb_csum_offl_spec *spec)
2669{
2670 bool csum_encapped;
2671
2672 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2673}
2674
2675static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2676{
2677 static const struct skb_csum_offl_spec csum_offl_spec = {
2678 .ipv4_okay = 1,
2679 .ip_options_okay = 1,
2680 .ipv6_okay = 1,
2681 .vlan_okay = 1,
2682 .tcp_okay = 1,
2683 .udp_okay = 1,
2684 };
2685
2686 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2687}
2688
2689static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2690{
2691 static const struct skb_csum_offl_spec csum_offl_spec = {
2692 .ipv4_okay = 1,
2693 .ip_options_okay = 1,
2694 .tcp_okay = 1,
2695 .udp_okay = 1,
2696 .vlan_okay = 1,
2697 };
2698
2699 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2700}
2701
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002702static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2703 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002704 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002705 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002706{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002707 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002708 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002709
2710 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002711}
2712
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002713static inline int dev_parse_header(const struct sk_buff *skb,
2714 unsigned char *haddr)
2715{
2716 const struct net_device *dev = skb->dev;
2717
Patrick McHardy1b833362007-10-18 05:09:28 -07002718 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002719 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002720 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002721}
2722
Willem de Bruijn2793a232016-03-09 21:58:32 -05002723/* ll_header must have at least hard_header_len allocated */
2724static inline bool dev_validate_header(const struct net_device *dev,
2725 char *ll_header, int len)
2726{
2727 if (likely(len >= dev->hard_header_len))
2728 return true;
2729
2730 if (capable(CAP_SYS_RAWIO)) {
2731 memset(ll_header + len, 0, dev->hard_header_len - len);
2732 return true;
2733 }
2734
2735 if (dev->header_ops && dev->header_ops->validate)
2736 return dev->header_ops->validate(ll_header, len);
2737
2738 return false;
2739}
2740
Linus Torvalds1da177e2005-04-16 15:20:36 -07002741typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002742int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002743static inline int unregister_gifconf(unsigned int family)
2744{
2745 return register_gifconf(family, NULL);
2746}
2747
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002748#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002749#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002750struct sd_flow_limit {
2751 u64 count;
2752 unsigned int num_buckets;
2753 unsigned int history_head;
2754 u16 history[FLOW_LIMIT_HISTORY];
2755 u8 buckets[];
2756};
2757
2758extern int netdev_flow_limit_table_len;
2759#endif /* CONFIG_NET_FLOW_LIMIT */
2760
Linus Torvalds1da177e2005-04-16 15:20:36 -07002761/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002762 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002763 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002764struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002765 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002766 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002767
Changli Gaodee42872010-05-02 05:42:16 +00002768 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002769 unsigned int processed;
2770 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002771 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002772#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002773 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002774#endif
2775#ifdef CONFIG_NET_FLOW_LIMIT
2776 struct sd_flow_limit __rcu *flow_limit;
2777#endif
2778 struct Qdisc *output_queue;
2779 struct Qdisc **output_queue_tailp;
2780 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002781
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002782#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002783 /* input_queue_head should be written by cpu owning this struct,
2784 * and only read by other cpus. Worth using a cache line.
2785 */
2786 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2787
2788 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002789 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002790 struct softnet_data *rps_ipi_next;
2791 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002792 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002793#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002794 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002795 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002796 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002797
Linus Torvalds1da177e2005-04-16 15:20:36 -07002798};
2799
Tom Herbert76cc8b12010-05-20 18:37:59 +00002800static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002801{
2802#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002803 sd->input_queue_head++;
2804#endif
2805}
2806
2807static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2808 unsigned int *qtail)
2809{
2810#ifdef CONFIG_RPS
2811 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002812#endif
2813}
2814
Tom Herbert0a9627f2010-03-16 08:03:29 +00002815DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002816
Joe Perchesf629d202013-09-26 14:48:15 -07002817void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002818void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002819
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002820static inline void netif_tx_schedule_all(struct net_device *dev)
2821{
2822 unsigned int i;
2823
2824 for (i = 0; i < dev->num_tx_queues; i++)
2825 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2826}
2827
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002828static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002829{
Tom Herbert734664982011-11-28 16:32:44 +00002830 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002831}
2832
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002833/**
2834 * netif_start_queue - allow transmit
2835 * @dev: network device
2836 *
2837 * Allow upper layers to call the device hard_start_xmit routine.
2838 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002839static inline void netif_start_queue(struct net_device *dev)
2840{
David S. Millere8a04642008-07-17 00:34:19 -07002841 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002842}
2843
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002844static inline void netif_tx_start_all_queues(struct net_device *dev)
2845{
2846 unsigned int i;
2847
2848 for (i = 0; i < dev->num_tx_queues; i++) {
2849 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2850 netif_tx_start_queue(txq);
2851 }
2852}
2853
John Fastabend46e5da40a2014-09-12 20:04:52 -07002854void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002855
Dave Jonesd29f7492008-07-22 14:09:06 -07002856/**
2857 * netif_wake_queue - restart transmit
2858 * @dev: network device
2859 *
2860 * Allow upper layers to call the device hard_start_xmit routine.
2861 * Used for flow control when transmit resources are available.
2862 */
David S. Miller79d16382008-07-08 23:14:46 -07002863static inline void netif_wake_queue(struct net_device *dev)
2864{
David S. Millere8a04642008-07-17 00:34:19 -07002865 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002866}
2867
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002868static inline void netif_tx_wake_all_queues(struct net_device *dev)
2869{
2870 unsigned int i;
2871
2872 for (i = 0; i < dev->num_tx_queues; i++) {
2873 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2874 netif_tx_wake_queue(txq);
2875 }
2876}
2877
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002878static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002879{
Tom Herbert734664982011-11-28 16:32:44 +00002880 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002881}
2882
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002883/**
2884 * netif_stop_queue - stop transmitted packets
2885 * @dev: network device
2886 *
2887 * Stop upper layers calling the device hard_start_xmit routine.
2888 * Used for flow control when transmit resources are unavailable.
2889 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002890static inline void netif_stop_queue(struct net_device *dev)
2891{
David S. Millere8a04642008-07-17 00:34:19 -07002892 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002893}
2894
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002895void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002896
David S. Miller4d295152012-03-07 21:02:35 -05002897static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002898{
Tom Herbert734664982011-11-28 16:32:44 +00002899 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002900}
2901
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002902/**
2903 * netif_queue_stopped - test if transmit queue is flowblocked
2904 * @dev: network device
2905 *
2906 * Test if transmit queue on device is currently unable to send.
2907 */
David S. Miller4d295152012-03-07 21:02:35 -05002908static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002909{
David S. Millere8a04642008-07-17 00:34:19 -07002910 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002911}
2912
David S. Miller4d295152012-03-07 21:02:35 -05002913static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002914{
Tom Herbert734664982011-11-28 16:32:44 +00002915 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2916}
2917
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002918static inline bool
2919netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002920{
2921 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2922}
2923
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002924static inline bool
2925netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2926{
2927 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2928}
2929
Eric Dumazet53511452014-10-08 08:19:27 -07002930/**
2931 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2932 * @dev_queue: pointer to transmit queue
2933 *
2934 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002935 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002936 */
2937static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2938{
2939#ifdef CONFIG_BQL
2940 prefetchw(&dev_queue->dql.num_queued);
2941#endif
2942}
2943
2944/**
2945 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2946 * @dev_queue: pointer to transmit queue
2947 *
2948 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002949 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002950 */
2951static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2952{
2953#ifdef CONFIG_BQL
2954 prefetchw(&dev_queue->dql.limit);
2955#endif
2956}
2957
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002958static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2959 unsigned int bytes)
2960{
Tom Herbert114cf582011-11-28 16:33:09 +00002961#ifdef CONFIG_BQL
2962 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002963
2964 if (likely(dql_avail(&dev_queue->dql) >= 0))
2965 return;
2966
2967 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2968
2969 /*
2970 * The XOFF flag must be set before checking the dql_avail below,
2971 * because in netdev_tx_completed_queue we update the dql_completed
2972 * before checking the XOFF flag.
2973 */
2974 smp_mb();
2975
2976 /* check again in case another CPU has just made room avail */
2977 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2978 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002979#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002980}
2981
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002982/**
2983 * netdev_sent_queue - report the number of bytes queued to hardware
2984 * @dev: network device
2985 * @bytes: number of bytes queued to the hardware device queue
2986 *
2987 * Report the number of bytes queued for sending/completion to the network
2988 * device hardware queue. @bytes should be a good approximation and should
2989 * exactly match netdev_completed_queue() @bytes
2990 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002991static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2992{
2993 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2994}
2995
2996static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002997 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002998{
Tom Herbert114cf582011-11-28 16:33:09 +00002999#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00003000 if (unlikely(!bytes))
3001 return;
3002
3003 dql_completed(&dev_queue->dql, bytes);
3004
3005 /*
3006 * Without the memory barrier there is a small possiblity that
3007 * netdev_tx_sent_queue will miss the update and cause the queue to
3008 * be stopped forever
3009 */
3010 smp_mb();
3011
3012 if (dql_avail(&dev_queue->dql) < 0)
3013 return;
3014
3015 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3016 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00003017#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003018}
3019
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003020/**
3021 * netdev_completed_queue - report bytes and packets completed by device
3022 * @dev: network device
3023 * @pkts: actual number of packets sent over the medium
3024 * @bytes: actual number of bytes sent over the medium
3025 *
3026 * Report the number of bytes and packets transmitted by the network device
3027 * hardware queue over the physical medium, @bytes must exactly match the
3028 * @bytes amount passed to netdev_sent_queue()
3029 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003030static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00003031 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003032{
3033 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3034}
3035
3036static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3037{
Tom Herbert114cf582011-11-28 16:33:09 +00003038#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003039 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003040 dql_reset(&q->dql);
3041#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003042}
3043
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003044/**
3045 * netdev_reset_queue - reset the packets and bytes count of a network device
3046 * @dev_queue: network device
3047 *
3048 * Reset the bytes and packet count of a network device and clear the
3049 * software flow control OFF bit for this network device
3050 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003051static inline void netdev_reset_queue(struct net_device *dev_queue)
3052{
3053 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003054}
3055
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003056/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003057 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3058 * @dev: network device
3059 * @queue_index: given tx queue index
3060 *
3061 * Returns 0 if given tx queue index >= number of device tx queues,
3062 * otherwise returns the originally passed tx queue index.
3063 */
3064static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3065{
3066 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3067 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3068 dev->name, queue_index,
3069 dev->real_num_tx_queues);
3070 return 0;
3071 }
3072
3073 return queue_index;
3074}
3075
3076/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003077 * netif_running - test if up
3078 * @dev: network device
3079 *
3080 * Test if the device has been brought up.
3081 */
David S. Miller4d295152012-03-07 21:02:35 -05003082static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003083{
3084 return test_bit(__LINK_STATE_START, &dev->state);
3085}
3086
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003087/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003088 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003089 * stop, and a check if it's stopped. All other device management is
3090 * done at the overall netdevice level.
3091 * Also test the device if we're multiqueue.
3092 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003093
3094/**
3095 * netif_start_subqueue - allow sending packets on subqueue
3096 * @dev: network device
3097 * @queue_index: sub queue index
3098 *
3099 * Start individual transmit queue of a device with multiple transmit queues.
3100 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003101static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3102{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003103 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003104
3105 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003106}
3107
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003108/**
3109 * netif_stop_subqueue - stop sending packets on subqueue
3110 * @dev: network device
3111 * @queue_index: sub queue index
3112 *
3113 * Stop individual transmit queue of a device with multiple transmit queues.
3114 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003115static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3116{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003117 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003118 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003119}
3120
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003121/**
3122 * netif_subqueue_stopped - test status of subqueue
3123 * @dev: network device
3124 * @queue_index: sub queue index
3125 *
3126 * Check individual transmit queue of a device with multiple transmit queues.
3127 */
David S. Miller4d295152012-03-07 21:02:35 -05003128static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3129 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003130{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003131 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003132
3133 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003134}
3135
David S. Miller4d295152012-03-07 21:02:35 -05003136static inline bool netif_subqueue_stopped(const struct net_device *dev,
3137 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003138{
3139 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3140}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003141
John Fastabend46e5da40a2014-09-12 20:04:52 -07003142void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003143
Alexander Duyck537c00d2013-01-10 08:57:02 +00003144#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003145int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003146 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003147#else
3148static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003149 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003150 u16 index)
3151{
3152 return 0;
3153}
3154#endif
3155
Jiri Pirko5605c762015-05-12 14:56:12 +02003156u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3157 unsigned int num_tx_queues);
3158
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003159/*
3160 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3161 * as a distribution range limit for the returned value.
3162 */
3163static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003164 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003165{
3166 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3167}
3168
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003169/**
3170 * netif_is_multiqueue - test if device has multiple transmit queues
3171 * @dev: network device
3172 *
3173 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003174 */
David S. Miller4d295152012-03-07 21:02:35 -05003175static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003176{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003177 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003178}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003179
Joe Perchesf629d202013-09-26 14:48:15 -07003180int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003181
Michael Daltona953be52014-01-16 22:23:28 -08003182#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003183int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003184#else
3185static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3186 unsigned int rxq)
3187{
3188 return 0;
3189}
3190#endif
3191
Michael Daltona953be52014-01-16 22:23:28 -08003192#ifdef CONFIG_SYSFS
3193static inline unsigned int get_netdev_rx_queue_index(
3194 struct netdev_rx_queue *queue)
3195{
3196 struct net_device *dev = queue->dev;
3197 int index = queue - dev->_rx;
3198
3199 BUG_ON(index >= dev->num_rx_queues);
3200 return index;
3201}
3202#endif
3203
Yuval Mintz16917b82012-07-01 03:18:50 +00003204#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003205int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003206
Eric Dumazete6247022013-12-05 04:45:08 -08003207enum skb_free_reason {
3208 SKB_REASON_CONSUMED,
3209 SKB_REASON_DROPPED,
3210};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003211
Eric Dumazete6247022013-12-05 04:45:08 -08003212void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3213void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3214
3215/*
3216 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3217 * interrupt context or with hardware interrupts being disabled.
3218 * (in_irq() || irqs_disabled())
3219 *
3220 * We provide four helpers that can be used in following contexts :
3221 *
3222 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3223 * replacing kfree_skb(skb)
3224 *
3225 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3226 * Typically used in place of consume_skb(skb) in TX completion path
3227 *
3228 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3229 * replacing kfree_skb(skb)
3230 *
3231 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3232 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003233 */
Eric Dumazete6247022013-12-05 04:45:08 -08003234static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3235{
3236 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3237}
3238
3239static inline void dev_consume_skb_irq(struct sk_buff *skb)
3240{
3241 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3242}
3243
3244static inline void dev_kfree_skb_any(struct sk_buff *skb)
3245{
3246 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3247}
3248
3249static inline void dev_consume_skb_any(struct sk_buff *skb)
3250{
3251 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3252}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003253
Joe Perchesf629d202013-09-26 14:48:15 -07003254int netif_rx(struct sk_buff *skb);
3255int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003256int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003257gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3258void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3259struct sk_buff *napi_get_frags(struct napi_struct *napi);
3260gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003261struct packet_offload *gro_find_receive_by_type(__be16 type);
3262struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003263
3264static inline void napi_free_frags(struct napi_struct *napi)
3265{
3266 kfree_skb(napi->skb);
3267 napi->skb = NULL;
3268}
3269
Joe Perchesf629d202013-09-26 14:48:15 -07003270int netdev_rx_handler_register(struct net_device *dev,
3271 rx_handler_func_t *rx_handler,
3272 void *rx_handler_data);
3273void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003274
Joe Perchesf629d202013-09-26 14:48:15 -07003275bool dev_valid_name(const char *name);
3276int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3277int dev_ethtool(struct net *net, struct ifreq *);
3278unsigned int dev_get_flags(const struct net_device *);
3279int __dev_change_flags(struct net_device *, unsigned int flags);
3280int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003281void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3282 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003283int dev_change_name(struct net_device *, const char *);
3284int dev_set_alias(struct net_device *, const char *, size_t);
3285int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3286int dev_set_mtu(struct net_device *, int);
3287void dev_set_group(struct net_device *, int);
3288int dev_set_mac_address(struct net_device *, struct sockaddr *);
3289int dev_change_carrier(struct net_device *, bool new_carrier);
3290int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003291 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003292int dev_get_phys_port_name(struct net_device *dev,
3293 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003294int dev_change_proto_down(struct net_device *dev, bool proto_down);
Brenden Blancoa7862b42016-07-19 12:16:48 -07003295int dev_change_xdp_fd(struct net_device *dev, int fd);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003296struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003297struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3298 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08003299int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003300int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003301bool is_skb_forwardable(const struct net_device *dev,
3302 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003303
David Ahern74b20582016-05-10 11:19:50 -07003304void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3305
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003306extern int netdev_budget;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003307
3308/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003309void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003310
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003311/**
3312 * dev_put - release reference to device
3313 * @dev: network device
3314 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003315 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003316 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003317static inline void dev_put(struct net_device *dev)
3318{
Christoph Lameter933393f2011-12-22 11:58:51 -06003319 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003320}
3321
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003322/**
3323 * dev_hold - get reference to device
3324 * @dev: network device
3325 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003326 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003327 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003328static inline void dev_hold(struct net_device *dev)
3329{
Christoph Lameter933393f2011-12-22 11:58:51 -06003330 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003331}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003332
3333/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3334 * and _off may be called from IRQ context, but it is caller
3335 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003336 *
3337 * The name carrier is inappropriate, these functions should really be
3338 * called netif_lowerlayer_*() because they represent the state of any
3339 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003340 */
3341
Joe Perchesf629d202013-09-26 14:48:15 -07003342void linkwatch_init_dev(struct net_device *dev);
3343void linkwatch_fire_event(struct net_device *dev);
3344void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003345
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003346/**
3347 * netif_carrier_ok - test if carrier present
3348 * @dev: network device
3349 *
3350 * Check if carrier is present on device
3351 */
David S. Miller4d295152012-03-07 21:02:35 -05003352static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003353{
3354 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3355}
3356
Joe Perchesf629d202013-09-26 14:48:15 -07003357unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003358
Joe Perchesf629d202013-09-26 14:48:15 -07003359void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003360
Joe Perchesf629d202013-09-26 14:48:15 -07003361void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003362
Joe Perchesf629d202013-09-26 14:48:15 -07003363void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003364
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003365/**
3366 * netif_dormant_on - mark device as dormant.
3367 * @dev: network device
3368 *
3369 * Mark device as dormant (as per RFC2863).
3370 *
3371 * The dormant state indicates that the relevant interface is not
3372 * actually in a condition to pass packets (i.e., it is not 'up') but is
3373 * in a "pending" state, waiting for some external event. For "on-
3374 * demand" interfaces, this new state identifies the situation where the
3375 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003376 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003377static inline void netif_dormant_on(struct net_device *dev)
3378{
3379 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3380 linkwatch_fire_event(dev);
3381}
3382
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003383/**
3384 * netif_dormant_off - set device as not dormant.
3385 * @dev: network device
3386 *
3387 * Device is not in dormant state.
3388 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003389static inline void netif_dormant_off(struct net_device *dev)
3390{
3391 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3392 linkwatch_fire_event(dev);
3393}
3394
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003395/**
3396 * netif_dormant - test if carrier present
3397 * @dev: network device
3398 *
3399 * Check if carrier is present on device
3400 */
David S. Miller4d295152012-03-07 21:02:35 -05003401static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003402{
3403 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3404}
3405
3406
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003407/**
3408 * netif_oper_up - test if device is operational
3409 * @dev: network device
3410 *
3411 * Check if carrier is operational
3412 */
David S. Miller4d295152012-03-07 21:02:35 -05003413static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003414{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003415 return (dev->operstate == IF_OPER_UP ||
3416 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3417}
3418
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003419/**
3420 * netif_device_present - is device available or removed
3421 * @dev: network device
3422 *
3423 * Check if device has not been removed from system.
3424 */
David S. Miller4d295152012-03-07 21:02:35 -05003425static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003426{
3427 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3428}
3429
Joe Perchesf629d202013-09-26 14:48:15 -07003430void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003431
Joe Perchesf629d202013-09-26 14:48:15 -07003432void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003433
3434/*
3435 * Network interface message level settings
3436 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003437
3438enum {
3439 NETIF_MSG_DRV = 0x0001,
3440 NETIF_MSG_PROBE = 0x0002,
3441 NETIF_MSG_LINK = 0x0004,
3442 NETIF_MSG_TIMER = 0x0008,
3443 NETIF_MSG_IFDOWN = 0x0010,
3444 NETIF_MSG_IFUP = 0x0020,
3445 NETIF_MSG_RX_ERR = 0x0040,
3446 NETIF_MSG_TX_ERR = 0x0080,
3447 NETIF_MSG_TX_QUEUED = 0x0100,
3448 NETIF_MSG_INTR = 0x0200,
3449 NETIF_MSG_TX_DONE = 0x0400,
3450 NETIF_MSG_RX_STATUS = 0x0800,
3451 NETIF_MSG_PKTDATA = 0x1000,
3452 NETIF_MSG_HW = 0x2000,
3453 NETIF_MSG_WOL = 0x4000,
3454};
3455
3456#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3457#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3458#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3459#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3460#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3461#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3462#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3463#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3464#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3465#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3466#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3467#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3468#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3469#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3470#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3471
3472static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3473{
3474 /* use default */
3475 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3476 return default_msg_enable_bits;
3477 if (debug_value == 0) /* no output */
3478 return 0;
3479 /* set low N bits */
3480 return (1 << debug_value) - 1;
3481}
3482
David S. Millerc773e842008-07-08 23:13:53 -07003483static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003484{
David S. Millerc773e842008-07-08 23:13:53 -07003485 spin_lock(&txq->_xmit_lock);
3486 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003487}
3488
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003489static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3490{
3491 spin_lock_bh(&txq->_xmit_lock);
3492 txq->xmit_lock_owner = smp_processor_id();
3493}
3494
David S. Miller4d295152012-03-07 21:02:35 -05003495static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003496{
David S. Miller4d295152012-03-07 21:02:35 -05003497 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003498 if (likely(ok))
3499 txq->xmit_lock_owner = smp_processor_id();
3500 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003501}
3502
David S. Millerc773e842008-07-08 23:13:53 -07003503static inline void __netif_tx_unlock(struct netdev_queue *txq)
3504{
3505 txq->xmit_lock_owner = -1;
3506 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003507}
3508
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003509static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3510{
3511 txq->xmit_lock_owner = -1;
3512 spin_unlock_bh(&txq->_xmit_lock);
3513}
3514
Eric Dumazet08baf562009-05-25 22:58:01 -07003515static inline void txq_trans_update(struct netdev_queue *txq)
3516{
3517 if (txq->xmit_lock_owner != -1)
3518 txq->trans_start = jiffies;
3519}
3520
Florian Westphalba162f82016-05-03 16:31:00 +02003521/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3522static inline void netif_trans_update(struct net_device *dev)
3523{
Florian Westphal9b366272016-05-03 16:33:14 +02003524 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3525
3526 if (txq->trans_start != jiffies)
3527 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003528}
3529
David S. Millerc3f26a22008-07-31 16:58:50 -07003530/**
3531 * netif_tx_lock - grab network device transmit lock
3532 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003533 *
3534 * Get network device transmit lock
3535 */
3536static inline void netif_tx_lock(struct net_device *dev)
3537{
3538 unsigned int i;
3539 int cpu;
3540
3541 spin_lock(&dev->tx_global_lock);
3542 cpu = smp_processor_id();
3543 for (i = 0; i < dev->num_tx_queues; i++) {
3544 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3545
3546 /* We are the only thread of execution doing a
3547 * freeze, but we have to grab the _xmit_lock in
3548 * order to synchronize with threads which are in
3549 * the ->hard_start_xmit() handler and already
3550 * checked the frozen bit.
3551 */
3552 __netif_tx_lock(txq, cpu);
3553 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3554 __netif_tx_unlock(txq);
3555 }
3556}
3557
3558static inline void netif_tx_lock_bh(struct net_device *dev)
3559{
3560 local_bh_disable();
3561 netif_tx_lock(dev);
3562}
3563
Herbert Xu932ff272006-06-09 12:20:56 -07003564static inline void netif_tx_unlock(struct net_device *dev)
3565{
David S. Millere8a04642008-07-17 00:34:19 -07003566 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003567
David S. Millere8a04642008-07-17 00:34:19 -07003568 for (i = 0; i < dev->num_tx_queues; i++) {
3569 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003570
David S. Millerc3f26a22008-07-31 16:58:50 -07003571 /* No need to grab the _xmit_lock here. If the
3572 * queue is not stopped for another reason, we
3573 * force a schedule.
3574 */
3575 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003576 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003577 }
3578 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003579}
3580
3581static inline void netif_tx_unlock_bh(struct net_device *dev)
3582{
David S. Millere8a04642008-07-17 00:34:19 -07003583 netif_tx_unlock(dev);
3584 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003585}
3586
David S. Millerc773e842008-07-08 23:13:53 -07003587#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003588 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003589 __netif_tx_lock(txq, cpu); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003590 } \
3591}
3592
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003593#define HARD_TX_TRYLOCK(dev, txq) \
3594 (((dev->features & NETIF_F_LLTX) == 0) ? \
3595 __netif_tx_trylock(txq) : \
3596 true )
3597
David S. Millerc773e842008-07-08 23:13:53 -07003598#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003599 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003600 __netif_tx_unlock(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003601 } \
3602}
3603
Linus Torvalds1da177e2005-04-16 15:20:36 -07003604static inline void netif_tx_disable(struct net_device *dev)
3605{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003606 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003607 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003608
David S. Millerc3f26a22008-07-31 16:58:50 -07003609 local_bh_disable();
3610 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003611 for (i = 0; i < dev->num_tx_queues; i++) {
3612 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003613
3614 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003615 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003616 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003617 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003618 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003619}
3620
David S. Millere308a5d2008-07-15 00:13:44 -07003621static inline void netif_addr_lock(struct net_device *dev)
3622{
3623 spin_lock(&dev->addr_list_lock);
3624}
3625
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003626static inline void netif_addr_lock_nested(struct net_device *dev)
3627{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003628 int subclass = SINGLE_DEPTH_NESTING;
3629
3630 if (dev->netdev_ops->ndo_get_lock_subclass)
3631 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3632
3633 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003634}
3635
David S. Millere308a5d2008-07-15 00:13:44 -07003636static inline void netif_addr_lock_bh(struct net_device *dev)
3637{
3638 spin_lock_bh(&dev->addr_list_lock);
3639}
3640
3641static inline void netif_addr_unlock(struct net_device *dev)
3642{
3643 spin_unlock(&dev->addr_list_lock);
3644}
3645
3646static inline void netif_addr_unlock_bh(struct net_device *dev)
3647{
3648 spin_unlock_bh(&dev->addr_list_lock);
3649}
3650
Jiri Pirkof001fde2009-05-05 02:48:28 +00003651/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003652 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003653 * rcu_read_lock held.
3654 */
3655#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003656 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003657
Linus Torvalds1da177e2005-04-16 15:20:36 -07003658/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3659
Joe Perchesf629d202013-09-26 14:48:15 -07003660void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003661
3662/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003663struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003664 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003665 void (*setup)(struct net_device *),
3666 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003667#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3668 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003669
Tom Gundersenc835a672014-07-14 16:37:24 +02003670#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3671 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3672 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003673
Joe Perchesf629d202013-09-26 14:48:15 -07003674int register_netdev(struct net_device *dev);
3675void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003676
Jiri Pirko22bedad32010-04-01 21:22:57 +00003677/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003678int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3679 struct netdev_hw_addr_list *from_list, int addr_len);
3680void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3681 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003682int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3683 struct net_device *dev,
3684 int (*sync)(struct net_device *, const unsigned char *),
3685 int (*unsync)(struct net_device *,
3686 const unsigned char *));
3687void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3688 struct net_device *dev,
3689 int (*unsync)(struct net_device *,
3690 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003691void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003692
Jiri Pirkof001fde2009-05-05 02:48:28 +00003693/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003694int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3695 unsigned char addr_type);
3696int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3697 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003698void dev_addr_flush(struct net_device *dev);
3699int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003700
3701/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003702int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3703int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3704int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3705int dev_uc_sync(struct net_device *to, struct net_device *from);
3706int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3707void dev_uc_unsync(struct net_device *to, struct net_device *from);
3708void dev_uc_flush(struct net_device *dev);
3709void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003710
Alexander Duyck670e5b82014-05-28 18:44:46 -07003711/**
3712 * __dev_uc_sync - Synchonize device's unicast list
3713 * @dev: device to sync
3714 * @sync: function to call if address should be added
3715 * @unsync: function to call if address should be removed
3716 *
3717 * Add newly added addresses to the interface, and release
3718 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003719 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003720static inline int __dev_uc_sync(struct net_device *dev,
3721 int (*sync)(struct net_device *,
3722 const unsigned char *),
3723 int (*unsync)(struct net_device *,
3724 const unsigned char *))
3725{
3726 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3727}
3728
3729/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003730 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003731 * @dev: device to sync
3732 * @unsync: function to call if address should be removed
3733 *
3734 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003735 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003736static inline void __dev_uc_unsync(struct net_device *dev,
3737 int (*unsync)(struct net_device *,
3738 const unsigned char *))
3739{
3740 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3741}
3742
Jiri Pirko22bedad32010-04-01 21:22:57 +00003743/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003744int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3745int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3746int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3747int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3748int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3749int dev_mc_sync(struct net_device *to, struct net_device *from);
3750int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3751void dev_mc_unsync(struct net_device *to, struct net_device *from);
3752void dev_mc_flush(struct net_device *dev);
3753void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003754
Alexander Duyck670e5b82014-05-28 18:44:46 -07003755/**
3756 * __dev_mc_sync - Synchonize device's multicast list
3757 * @dev: device to sync
3758 * @sync: function to call if address should be added
3759 * @unsync: function to call if address should be removed
3760 *
3761 * Add newly added addresses to the interface, and release
3762 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003763 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003764static inline int __dev_mc_sync(struct net_device *dev,
3765 int (*sync)(struct net_device *,
3766 const unsigned char *),
3767 int (*unsync)(struct net_device *,
3768 const unsigned char *))
3769{
3770 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3771}
3772
3773/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003774 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003775 * @dev: device to sync
3776 * @unsync: function to call if address should be removed
3777 *
3778 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003779 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003780static inline void __dev_mc_unsync(struct net_device *dev,
3781 int (*unsync)(struct net_device *,
3782 const unsigned char *))
3783{
3784 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3785}
3786
Linus Torvalds1da177e2005-04-16 15:20:36 -07003787/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003788void dev_set_rx_mode(struct net_device *dev);
3789void __dev_set_rx_mode(struct net_device *dev);
3790int dev_set_promiscuity(struct net_device *dev, int inc);
3791int dev_set_allmulti(struct net_device *dev, int inc);
3792void netdev_state_change(struct net_device *dev);
3793void netdev_notify_peers(struct net_device *dev);
3794void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003795/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003796void dev_load(struct net *net, const char *name);
3797struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3798 struct rtnl_link_stats64 *storage);
3799void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3800 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003801
3802extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003803extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003804extern int weight_p;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003805
Joe Perchesf629d202013-09-26 14:48:15 -07003806bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003807struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3808 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003809struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3810 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003811
3812/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003813#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3814 for (iter = &(dev)->adj_list.upper, \
3815 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3816 updev; \
3817 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3818
3819/* iterate through upper list, must be called under RCU read lock */
Veaceslav Falico2f268f12013-09-25 09:20:07 +02003820#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3821 for (iter = &(dev)->all_adj_list.upper, \
3822 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3823 updev; \
3824 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003825
Joe Perchesf629d202013-09-26 14:48:15 -07003826void *netdev_lower_get_next_private(struct net_device *dev,
3827 struct list_head **iter);
3828void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3829 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003830
3831#define netdev_for_each_lower_private(dev, priv, iter) \
3832 for (iter = (dev)->adj_list.lower.next, \
3833 priv = netdev_lower_get_next_private(dev, &(iter)); \
3834 priv; \
3835 priv = netdev_lower_get_next_private(dev, &(iter)))
3836
3837#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3838 for (iter = &(dev)->adj_list.lower, \
3839 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3840 priv; \
3841 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3842
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003843void *netdev_lower_get_next(struct net_device *dev,
3844 struct list_head **iter);
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003845
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003846#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003847 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003848 ldev = netdev_lower_get_next(dev, &(iter)); \
3849 ldev; \
3850 ldev = netdev_lower_get_next(dev, &(iter)))
3851
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003852struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3853 struct list_head **iter);
3854struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3855 struct list_head **iter);
3856
3857#define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3858 for (iter = (dev)->all_adj_list.lower.next, \
3859 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3860 ldev; \
3861 ldev = netdev_all_lower_get_next(dev, &(iter)))
3862
3863#define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
3864 for (iter = (dev)->all_adj_list.lower.next, \
3865 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3866 ldev; \
3867 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3868
Joe Perchesf629d202013-09-26 14:48:15 -07003869void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003870void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003871struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3872struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3873int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3874int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003875 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003876 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003877void netdev_upper_dev_unlink(struct net_device *dev,
3878 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003879void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003880void *netdev_lower_dev_get_private(struct net_device *dev,
3881 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003882void netdev_lower_state_changed(struct net_device *lower_dev,
3883 void *lower_state_info);
Jiri Pirko18bfb922016-07-05 11:27:38 +02003884int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3885 struct neighbour *n);
3886void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3887 struct neighbour *n);
Eric Dumazet960fb622014-11-16 06:23:05 -08003888
3889/* RSS keys are 40 or 52 bytes long */
3890#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003891extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003892void netdev_rss_key_fill(void *buffer, size_t len);
3893
Sabrina Dubroca952fcfd2016-08-12 16:10:33 +02003894int dev_get_nest_level(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003895int skb_checksum_help(struct sk_buff *skb);
3896struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3897 netdev_features_t features, bool tx_path);
3898struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3899 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003900
Moni Shoua61bd3852015-02-03 16:48:29 +02003901struct netdev_bonding_info {
3902 ifslave slave;
3903 ifbond master;
3904};
3905
3906struct netdev_notifier_bonding_info {
3907 struct netdev_notifier_info info; /* must be first */
3908 struct netdev_bonding_info bonding_info;
3909};
3910
3911void netdev_bonding_info_change(struct net_device *dev,
3912 struct netdev_bonding_info *bonding_info);
3913
Cong Wang12b00042013-02-05 16:36:38 +00003914static inline
3915struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3916{
3917 return __skb_gso_segment(skb, features, true);
3918}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003919__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003920
3921static inline bool can_checksum_protocol(netdev_features_t features,
3922 __be16 protocol)
3923{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003924 if (protocol == htons(ETH_P_FCOE))
3925 return !!(features & NETIF_F_FCOE_CRC);
3926
3927 /* Assume this is an IP checksum (not SCTP CRC) */
3928
3929 if (features & NETIF_F_HW_CSUM) {
3930 /* Can checksum everything */
3931 return true;
3932 }
3933
3934 switch (protocol) {
3935 case htons(ETH_P_IP):
3936 return !!(features & NETIF_F_IP_CSUM);
3937 case htons(ETH_P_IPV6):
3938 return !!(features & NETIF_F_IPV6_CSUM);
3939 default:
3940 return false;
3941 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003942}
Cong Wang12b00042013-02-05 16:36:38 +00003943
Tom Herbert6ae23ad2015-12-14 11:19:46 -08003944/* Map an ethertype into IP protocol if possible */
3945static inline int eproto_to_ipproto(int eproto)
3946{
3947 switch (eproto) {
3948 case htons(ETH_P_IP):
3949 return IPPROTO_IP;
3950 case htons(ETH_P_IPV6):
3951 return IPPROTO_IPV6;
3952 default:
3953 return -1;
3954 }
3955}
3956
Herbert Xufb286bb2005-11-10 13:01:24 -08003957#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003958void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003959#else
3960static inline void netdev_rx_csum_fault(struct net_device *dev)
3961{
3962}
3963#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003964/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003965void net_enable_timestamp(void);
3966void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003967
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003968#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003969int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003970#else
3971#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003972#endif
3973
David S. Miller47982482014-08-22 16:21:53 -07003974static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003975 struct sk_buff *skb, struct net_device *dev,
3976 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003977{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003978 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003979 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003980}
3981
David S. Miller10b3ad82014-08-29 21:07:24 -07003982static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003983 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07003984{
3985 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07003986 int rc;
David S. Miller47982482014-08-22 16:21:53 -07003987
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003988 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07003989 if (rc == NETDEV_TX_OK)
3990 txq_trans_update(txq);
3991
3992 return rc;
David S. Miller47982482014-08-22 16:21:53 -07003993}
3994
Linus Torvalds42a2d922013-11-13 17:40:34 +09003995int netdev_class_create_file_ns(struct class_attribute *class_attr,
3996 const void *ns);
3997void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3998 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04003999
4000static inline int netdev_class_create_file(struct class_attribute *class_attr)
4001{
4002 return netdev_class_create_file_ns(class_attr, NULL);
4003}
4004
4005static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4006{
4007 netdev_class_remove_file_ns(class_attr, NULL);
4008}
Jay Vosburghb8a97872008-06-13 18:12:04 -07004009
Johannes Berg04600792010-08-05 17:45:15 +02004010extern struct kobj_ns_type_operations net_ns_type_operations;
4011
Joe Perchesf629d202013-09-26 14:48:15 -07004012const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07004013
Joe Perchesf629d202013-09-26 14:48:15 -07004014void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004015
Michal Kubečekda081432014-05-20 08:29:25 +02004016static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4017 netdev_features_t f2)
4018{
Tom Herbertc8cd0982015-12-14 11:19:44 -08004019 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4020 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08004021 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004022 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08004023 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004024 }
Michal Kubečekda081432014-05-20 08:29:25 +02004025
Tom Herbertc8cd0982015-12-14 11:19:44 -08004026 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02004027}
4028
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004029static inline netdev_features_t netdev_get_wanted_features(
4030 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00004031{
4032 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4033}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004034netdev_features_t netdev_increment_features(netdev_features_t all,
4035 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00004036
4037/* Allow TSO being used on stacked device :
4038 * Performing the GSO segmentation before last device
4039 * is a performance improvement.
4040 */
4041static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4042 netdev_features_t mask)
4043{
4044 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4045}
4046
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004047int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004048void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004049void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004050
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004051void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4052 struct net_device *dev);
4053
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004054netdev_features_t passthru_features_check(struct sk_buff *skb,
4055 struct net_device *dev,
4056 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004057netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004058
David S. Miller4d295152012-03-07 21:02:35 -05004059static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004060{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004061 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004062
4063 /* check flags correspondence */
4064 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4065 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4066 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4067 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004068 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004069 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4070 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004071 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4072 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004073 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4074 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004075 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4076 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
Alexander Duyck802ab552016-04-10 21:45:03 -04004077 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004078 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004079 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004080
Herbert Xubcd76112006-06-30 13:36:35 -07004081 return (features & feature) == feature;
4082}
4083
David S. Miller4d295152012-03-07 21:02:35 -05004084static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004085{
Herbert Xu278b2512009-06-03 21:20:51 -07004086 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004087 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004088}
4089
Johannes Berg8b86a612015-04-17 15:45:04 +02004090static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004091 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004092{
Jesse Grossfc741212011-01-09 06:23:32 +00004093 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004094 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4095 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004096}
4097
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004098static inline void netif_set_gso_max_size(struct net_device *dev,
4099 unsigned int size)
4100{
4101 dev->gso_max_size = size;
4102}
4103
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004104static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4105 int pulled_hlen, u16 mac_offset,
4106 int mac_len)
4107{
4108 skb->protocol = protocol;
4109 skb->encapsulation = 1;
4110 skb_push(skb, pulled_hlen);
4111 skb_reset_transport_header(skb);
4112 skb->mac_header = mac_offset;
4113 skb->network_header = skb->mac_header + mac_len;
4114 skb->mac_len = mac_len;
4115}
4116
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004117static inline bool netif_is_macsec(const struct net_device *dev)
4118{
4119 return dev->priv_flags & IFF_MACSEC;
4120}
4121
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004122static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004123{
4124 return dev->priv_flags & IFF_MACVLAN;
4125}
4126
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004127static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004128{
4129 return dev->priv_flags & IFF_MACVLAN_PORT;
4130}
4131
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004132static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004133{
4134 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4135}
4136
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004137static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004138{
4139 return dev->priv_flags & IFF_IPVLAN_MASTER;
4140}
4141
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004142static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004143{
4144 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4145}
4146
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004147static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004148{
4149 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4150}
4151
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004152static inline bool netif_supports_nofcs(struct net_device *dev)
4153{
4154 return dev->priv_flags & IFF_SUPP_NOFCS;
4155}
4156
David Ahern007979e2015-09-29 20:07:10 -07004157static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004158{
David Ahern007979e2015-09-29 20:07:10 -07004159 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004160}
4161
David Ahernfee6d4c2015-10-05 08:51:24 -07004162static inline bool netif_is_l3_slave(const struct net_device *dev)
4163{
4164 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4165}
4166
Jiri Pirko0894ae32015-08-27 09:31:19 +02004167static inline bool netif_is_bridge_master(const struct net_device *dev)
4168{
4169 return dev->priv_flags & IFF_EBRIDGE;
4170}
4171
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004172static inline bool netif_is_bridge_port(const struct net_device *dev)
4173{
4174 return dev->priv_flags & IFF_BRIDGE_PORT;
4175}
4176
Jiri Pirko35d4e172015-08-27 09:31:20 +02004177static inline bool netif_is_ovs_master(const struct net_device *dev)
4178{
4179 return dev->priv_flags & IFF_OPENVSWITCH;
4180}
4181
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004182static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004183{
4184 return dev->priv_flags & IFF_TEAM;
4185}
4186
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004187static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004188{
4189 return dev->priv_flags & IFF_TEAM_PORT;
4190}
4191
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004192static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004193{
4194 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4195}
4196
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004197static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004198{
4199 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4200}
4201
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004202static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4203{
4204 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4205}
4206
Eric Dumazet02875872014-10-05 18:38:35 -07004207/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4208static inline void netif_keep_dst(struct net_device *dev)
4209{
4210 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4211}
4212
Paolo Abeni18d3df32016-07-14 18:00:10 +02004213/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4214static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4215{
4216 /* TODO: reserve and use an additional IFF bit, if we get more users */
4217 return dev->priv_flags & IFF_MACSEC;
4218}
4219
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004220extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004221
Joe Perches571ba422010-02-09 11:49:47 +00004222/* Logging, debugging and troubleshooting/diagnostic helpers. */
4223
4224/* netdev_printk helpers, similar to dev_printk */
4225
4226static inline const char *netdev_name(const struct net_device *dev)
4227{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004228 if (!dev->name[0] || strchr(dev->name, '%'))
4229 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004230 return dev->name;
4231}
4232
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004233static inline const char *netdev_reg_state(const struct net_device *dev)
4234{
4235 switch (dev->reg_state) {
4236 case NETREG_UNINITIALIZED: return " (uninitialized)";
4237 case NETREG_REGISTERED: return "";
4238 case NETREG_UNREGISTERING: return " (unregistering)";
4239 case NETREG_UNREGISTERED: return " (unregistered)";
4240 case NETREG_RELEASED: return " (released)";
4241 case NETREG_DUMMY: return " (dummy)";
4242 }
4243
4244 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4245 return " (unknown)";
4246}
4247
Joe Perchesf629d202013-09-26 14:48:15 -07004248__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004249void netdev_printk(const char *level, const struct net_device *dev,
4250 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004251__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004252void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004253__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004254void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004255__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004256void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004257__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004258void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004259__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004260void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004261__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004262void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004263__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004264void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004265
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004266#define MODULE_ALIAS_NETDEV(device) \
4267 MODULE_ALIAS("netdev-" device)
4268
Jim Cromieb558c962011-12-19 17:11:18 -05004269#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004270#define netdev_dbg(__dev, format, args...) \
4271do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004272 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004273} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004274#elif defined(DEBUG)
4275#define netdev_dbg(__dev, format, args...) \
4276 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004277#else
4278#define netdev_dbg(__dev, format, args...) \
4279({ \
4280 if (0) \
4281 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004282})
4283#endif
4284
4285#if defined(VERBOSE_DEBUG)
4286#define netdev_vdbg netdev_dbg
4287#else
4288
4289#define netdev_vdbg(dev, format, args...) \
4290({ \
4291 if (0) \
4292 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4293 0; \
4294})
4295#endif
4296
4297/*
4298 * netdev_WARN() acts like dev_printk(), but with the key difference
4299 * of using a WARN/WARN_ON to get the message out, including the
4300 * file/line information and a backtrace.
4301 */
4302#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004303 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4304 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004305
Joe Perchesb3d95c52010-02-09 11:49:49 +00004306/* netif printk helpers, similar to netdev_printk */
4307
4308#define netif_printk(priv, type, level, dev, fmt, args...) \
4309do { \
4310 if (netif_msg_##type(priv)) \
4311 netdev_printk(level, (dev), fmt, ##args); \
4312} while (0)
4313
Joe Perchesf45f4322010-06-27 01:02:36 +00004314#define netif_level(level, priv, type, dev, fmt, args...) \
4315do { \
4316 if (netif_msg_##type(priv)) \
4317 netdev_##level(dev, fmt, ##args); \
4318} while (0)
4319
Joe Perchesb3d95c52010-02-09 11:49:49 +00004320#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004321 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004322#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004323 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004324#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004325 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004326#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004327 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004328#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004329 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004330#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004331 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004332#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004333 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004334
Joe Perches0053ea92012-05-30 07:43:34 +00004335#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004336#define netif_dbg(priv, type, netdev, format, args...) \
4337do { \
4338 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004339 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004340} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004341#elif defined(DEBUG)
4342#define netif_dbg(priv, type, dev, format, args...) \
4343 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004344#else
4345#define netif_dbg(priv, type, dev, format, args...) \
4346({ \
4347 if (0) \
4348 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4349 0; \
4350})
4351#endif
4352
4353#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004354#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004355#else
4356#define netif_vdbg(priv, type, dev, format, args...) \
4357({ \
4358 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004359 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004360 0; \
4361})
4362#endif
Joe Perches571ba422010-02-09 11:49:47 +00004363
Cong Wang900ff8c2013-02-18 19:20:33 +00004364/*
4365 * The list of packet types we will receive (as opposed to discard)
4366 * and the routines to invoke.
4367 *
4368 * Why 16. Because with 16 the only overlap we get on a hash of the
4369 * low nibble of the protocol value is RARP/SNAP/X.25.
4370 *
4371 * NOTE: That is no longer true with the addition of VLAN tags. Not
4372 * sure which should go first, but I bet it won't make much
4373 * difference if we are running VLANs. The good news is that
4374 * this protocol won't be in the list unless compiled in, so
4375 * the average user (w/out VLANs) will not be adversely affected.
4376 * --BLG
4377 *
4378 * 0800 IP
4379 * 8100 802.1Q VLAN
4380 * 0001 802.3
4381 * 0002 AX.25
4382 * 0004 802.2
4383 * 8035 RARP
4384 * 0005 SNAP
4385 * 0805 X.25
4386 * 0806 ARP
4387 * 8137 IPX
4388 * 0009 Localtalk
4389 * 86DD IPv6
4390 */
4391#define PTYPE_HASH_SIZE (16)
4392#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4393
Jiri Pirko385a1542009-05-27 15:48:07 -07004394#endif /* _LINUX_NETDEVICE_H */