blob: 70ad0291d517b41cd1d4a3c2c7fc046f13e27980 [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>
Jiri Kosina59cc1f62016-08-10 11:05:15 +020055#include <linux/hashtable.h>
Michał Mirosława59e2ec2011-11-15 15:29:55 +000056
Jeff Moyer115c1d62005-06-22 22:05:31 -070057struct netpoll_info;
Paul Gortmaker313162d2012-01-30 11:46:54 -050058struct device;
Richard Cochranc1f19b52010-07-17 08:49:36 +000059struct phy_device;
Johannes Berg704232c2007-04-23 12:20:05 -070060/* 802.11 specific */
61struct wireless_dev;
Alexander Aring98a18b62014-11-02 06:44:54 +010062/* 802.15.4 specific */
63struct wpan_dev;
Robert Shearman03c57742015-04-22 11:14:37 +010064struct mpls_dev;
Alexander Duyck7c46a642016-06-16 12:21:00 -070065/* UDP Tunnel offloads */
66struct udp_tunnel_info;
Brenden Blancoa7862b42016-07-19 12:16:48 -070067struct bpf_prog;
Linus Torvalds1da177e2005-04-16 15:20:36 -070068
Joe Perchesf629d202013-09-26 14:48:15 -070069void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000071
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000072/* Backlog congestion levels */
73#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74#define NET_RX_DROP 1 /* packet dropped */
75
Patrick McHardy572a9d72009-11-10 06:14:14 +000076/*
77 * Transmit return codes: transmit return codes originate from three different
78 * namespaces:
79 *
80 * - qdisc return codes
81 * - driver transmit return codes
82 * - errno values
83 *
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -050088 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
Patrick McHardy572a9d72009-11-10 06:14:14 +000090 * others are propagated to higher layers.
91 */
92
93/* qdisc ->enqueue() return codes. */
94#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000095#define NET_XMIT_DROP 0x01 /* skb dropped */
96#define NET_XMIT_CN 0x02 /* congestion notification */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000097#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -070098
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -020099/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +0000102#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
104
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000105/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000106#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000107
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000108enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000112};
113typedef enum netdev_tx netdev_tx_t;
114
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000115/*
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
118 */
119static inline bool dev_xmit_complete(int rc)
120{
121 /*
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
126 */
127 if (likely(rc < NET_XMIT_MASK))
128 return true;
129
130 return false;
131}
132
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500134 * Compute the worst-case header length according to the protocols
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 * used.
136 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800137
KY Srinivasanc0eb4542015-12-01 16:43:10 -0800138#if defined(CONFIG_HYPERV_NET)
139# define LL_MAX_HEADER 128
140#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700141# if defined(CONFIG_MAC80211_MESH)
142# define LL_MAX_HEADER 128
143# else
144# define LL_MAX_HEADER 96
145# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700147# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148#endif
149
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000150#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152#define MAX_HEADER LL_MAX_HEADER
153#else
154#define MAX_HEADER (LL_MAX_HEADER + 48)
155#endif
156
157/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800161
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800162struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
186};
187
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188
189#include <linux/cache.h>
190#include <linux/skbuff.h>
191
Eric Dumazetadc93002011-11-17 03:13:26 +0000192#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100193#include <linux/static_key.h>
194extern struct static_key rps_needed;
Eric Dumazet13bfff22016-12-07 08:29:10 -0800195extern struct static_key rfs_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000196#endif
197
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198struct neighbour;
199struct neigh_parms;
200struct sk_buff;
201
Jiri Pirkof001fde2009-05-05 02:48:28 +0000202struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
205 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000206#define NETDEV_HW_ADDR_T_LAN 1
207#define NETDEV_HW_ADDR_T_SAN 2
208#define NETDEV_HW_ADDR_T_SLAVE 3
209#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000210#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000211 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000212 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700213 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400214 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000215 struct rcu_head rcu_head;
216};
217
Jiri Pirko31278e72009-06-17 01:12:19 +0000218struct netdev_hw_addr_list {
219 struct list_head list;
220 int count;
221};
222
Jiri Pirko22bedad32010-04-01 21:22:57 +0000223#define netdev_hw_addr_list_count(l) ((l)->count)
224#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225#define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
227
228#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800230#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800232
Jiri Pirko22bedad32010-04-01 21:22:57 +0000233#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700235#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000236 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece2010-02-04 10:22:25 -0800237
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800238struct hh_cache {
David S. Millerf6b72b622011-07-14 07:53:20 -0700239 u16 hh_len;
David S. Miller5c25f682011-07-13 00:51:10 -0700240 u16 __pad;
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800241 seqlock_t hh_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242
243 /* cached hardware header; allow for machine alignment needs. */
244#define HH_DATA_MOD 16
245#define HH_DATA_OFF(__len) \
Jiri Benc5ba0eac2005-06-02 16:48:05 -0700246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247#define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250};
251
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500252/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260#define LL_RESERVED_SPACE(dev) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700265struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
Eric Dumazet95c96172012-04-15 05:58:06 +0000268 const void *saddr, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
David S. Millere69dd332011-07-12 23:28:12 -0700270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
Willem de Bruijn2793a232016-03-09 21:58:32 -0500274 bool (*validate)(const char *ll_header, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700275};
276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277/* These flag bits are private to the generic network queueing
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500278 * layer; they may not be explicitly referenced by any other
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 * code.
280 */
281
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800282enum netdev_state_t {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 __LINK_STATE_NOCARRIER,
Stefan Rompfb00055a2006-03-20 17:09:11 -0800286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288};
289
290
291/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500292 * This structure holds boot-time configured netdevice settings. They
Graf Yangfe2918b2009-02-05 21:26:19 -0800293 * are then used in the device probing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 */
295struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298};
299#define NETDEV_BOOT_SETUP_MAX 8
300
Joe Perchesf629d202013-09-26 14:48:15 -0700301int __init netdev_boot_setup(char *str);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302
303/*
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500310 * to the per-CPU poll_list, and whoever clears that bit
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000317 unsigned int gro_count;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700318 int (*poll)(struct napi_struct *, int);
319#ifdef CONFIG_NETPOLL
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700320 int poll_owner;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700321#endif
Herbert Xu5d38a072009-01-04 16:13:40 -0800322 struct net_device *dev;
Herbert Xud565b0a2008-12-15 23:38:52 -0800323 struct sk_buff *gro_list;
Herbert Xu5d38a072009-01-04 16:13:40 -0800324 struct sk_buff *skb;
Eric Dumazet3b47d302014-11-06 21:09:44 -0800325 struct hrtimer timer;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000326 struct list_head dev_list;
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700329};
330
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800331enum {
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700332 NAPI_STATE_SCHED, /* Poll is scheduled */
David S. Millera0a46192008-01-07 20:35:07 -0800333 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eric Dumazetd64b5e82015-11-18 06:31:00 -0800335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
Eric Dumazet217f6972016-11-15 10:15:11 -0800337 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
338};
339
340enum {
341 NAPIF_STATE_SCHED = (1UL << NAPI_STATE_SCHED),
342 NAPIF_STATE_DISABLE = (1UL << NAPI_STATE_DISABLE),
343 NAPIF_STATE_NPSVC = (1UL << NAPI_STATE_NPSVC),
344 NAPIF_STATE_HASHED = (1UL << NAPI_STATE_HASHED),
345 NAPIF_STATE_NO_BUSY_POLL = (1UL << NAPI_STATE_NO_BUSY_POLL),
346 NAPIF_STATE_IN_BUSY_POLL = (1UL << NAPI_STATE_IN_BUSY_POLL),
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700347};
348
Ben Hutchings5b252f02009-10-29 07:17:09 +0000349enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700350 GRO_MERGED,
351 GRO_MERGED_FREE,
352 GRO_HELD,
353 GRO_NORMAL,
354 GRO_DROP,
355};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000356typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700357
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000358/*
359 * enum rx_handler_result - Possible return values for rx_handlers.
360 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
361 * further.
362 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
363 * case skb->dev was changed by rx_handler.
364 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500365 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000366 *
367 * rx_handlers are functions called from inside __netif_receive_skb(), to do
368 * special processing of the skb, prior to delivery to protocol handlers.
369 *
370 * Currently, a net_device can only have a single rx_handler registered. Trying
371 * to register a second rx_handler will return -EBUSY.
372 *
373 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
374 * To unregister a rx_handler on a net_device, use
375 * netdev_rx_handler_unregister().
376 *
377 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
378 * do with the skb.
379 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500380 * If the rx_handler consumed the skb in some way, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000381 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500382 * the skb to be delivered in some other way.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000383 *
384 * If the rx_handler changed skb->dev, to divert the skb to another
385 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
386 * new device will be called if it exists.
387 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500388 * If the rx_handler decides the skb should be ignored, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000389 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400390 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000391 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500392 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000393 * delivered, it should return RX_HANDLER_PASS.
394 *
395 * A device without a registered rx_handler will behave as if rx_handler
396 * returned RX_HANDLER_PASS.
397 */
398
399enum rx_handler_result {
400 RX_HANDLER_CONSUMED,
401 RX_HANDLER_ANOTHER,
402 RX_HANDLER_EXACT,
403 RX_HANDLER_PASS,
404};
405typedef enum rx_handler_result rx_handler_result_t;
406typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000407
Joe Perchesf629d202013-09-26 14:48:15 -0700408void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700409void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700410
David S. Miller4d295152012-03-07 21:02:35 -0500411static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800412{
413 return test_bit(NAPI_STATE_DISABLE, &n->state);
414}
415
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700416/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500417 * napi_schedule_prep - check if NAPI can be scheduled
418 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700419 *
420 * Test if NAPI routine is already running, and if not mark
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500421 * it as running. This is used as a condition variable to
David S. Millera0a46192008-01-07 20:35:07 -0800422 * insure only one NAPI poll instance runs. We also make
423 * sure there is no pending NAPI disable.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700424 */
David S. Miller4d295152012-03-07 21:02:35 -0500425static inline bool napi_schedule_prep(struct napi_struct *n)
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700426{
David S. Millera0a46192008-01-07 20:35:07 -0800427 return !napi_disable_pending(n) &&
428 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700429}
430
431/**
432 * napi_schedule - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500433 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700434 *
435 * Schedule NAPI poll routine to be called if it is not already
436 * running.
437 */
438static inline void napi_schedule(struct napi_struct *n)
439{
440 if (napi_schedule_prep(n))
441 __napi_schedule(n);
442}
443
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700444/**
445 * napi_schedule_irqoff - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500446 * @n: NAPI context
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700447 *
448 * Variant of napi_schedule(), assuming hard irqs are masked.
449 */
450static inline void napi_schedule_irqoff(struct napi_struct *n)
451{
452 if (napi_schedule_prep(n))
453 __napi_schedule_irqoff(n);
454}
455
Roland Dreierbfe13f52007-10-09 15:47:37 -0700456/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
David S. Miller4d295152012-03-07 21:02:35 -0500457static inline bool napi_reschedule(struct napi_struct *napi)
Roland Dreierbfe13f52007-10-09 15:47:37 -0700458{
459 if (napi_schedule_prep(napi)) {
460 __napi_schedule(napi);
David S. Miller4d295152012-03-07 21:02:35 -0500461 return true;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700462 }
David S. Miller4d295152012-03-07 21:02:35 -0500463 return false;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700464}
465
Eric Dumazet364b6052016-11-15 10:15:13 -0800466bool __napi_complete(struct napi_struct *n);
467bool napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700468/**
469 * napi_complete - NAPI processing complete
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500470 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700471 *
472 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800473 * Consider using napi_complete_done() instead.
Eric Dumazet364b6052016-11-15 10:15:13 -0800474 * Return false if device should avoid rearming interrupts.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700475 */
Eric Dumazet364b6052016-11-15 10:15:13 -0800476static inline bool napi_complete(struct napi_struct *n)
Eric Dumazet3b47d302014-11-06 21:09:44 -0800477{
478 return napi_complete_done(n, 0);
479}
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700480
481/**
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300482 * napi_hash_del - remove a NAPI from global table
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500483 * @napi: NAPI context
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300484 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500485 * Warning: caller must observe RCU grace period
Eric Dumazet34cbe272015-11-18 06:31:02 -0800486 * before freeing memory containing @napi, if
487 * this function returns true.
Eric Dumazet93d05d42015-11-18 06:31:03 -0800488 * Note: core networking stack automatically calls it
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500489 * from netif_napi_del().
Eric Dumazet93d05d42015-11-18 06:31:03 -0800490 * Drivers might want to call this helper to combine all
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500491 * the needed RCU grace periods into a single one.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300492 */
Eric Dumazet34cbe272015-11-18 06:31:02 -0800493bool napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300494
495/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700496 * napi_disable - prevent NAPI from scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500497 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700498 *
499 * Stop NAPI from being scheduled on this context.
500 * Waits till any outstanding processing completes.
501 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800502void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700503
504/**
505 * napi_enable - enable NAPI scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500506 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700507 *
508 * Resume NAPI from being scheduled on this context.
509 * Must be paired with napi_disable.
510 */
511static inline void napi_enable(struct napi_struct *n)
512{
513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100514 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700515 clear_bit(NAPI_STATE_SCHED, &n->state);
Neil Horman2d8bff1262015-09-23 14:57:58 -0400516 clear_bit(NAPI_STATE_NPSVC, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700517}
518
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700519/**
520 * napi_synchronize - wait until NAPI is not running
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500521 * @n: NAPI context
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700522 *
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
526 */
527static inline void napi_synchronize(const struct napi_struct *n)
528{
Arnd Bergmannfacc4322016-01-22 11:43:44 +0100529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
531 msleep(1);
532 else
533 barrier();
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700534}
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700535
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800536enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000537 __QUEUE_STATE_DRV_XOFF,
538 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700539 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700540};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200541
542#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
543#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
544#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
545
546#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 QUEUE_STATE_FROZEN)
549#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
550 QUEUE_STATE_FROZEN)
551
Tom Herbert734664982011-11-28 16:32:44 +0000552/*
553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
554 * netif_tx_* functions below are used to manipulate this flag. The
555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556 * queue independently. The netif_xmit_*stopped functions below are called
557 * to check if the queue has been stopped by the driver or stack (either
558 * of the XOFF bits are set in the state). Drivers should not need to call
559 * netif_xmit*stopped functions, they should only be using netif_tx_*.
560 */
David S. Miller79d16382008-07-08 23:14:46 -0700561
David S. Millerbb949fb2008-07-08 16:55:56 -0700562struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700563/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500564 * read-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700565 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700566 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700567 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700568 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000569#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000570 struct kobject kobj;
571#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000572#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 int numa_node;
574#endif
Florian Westphalc0ef0792016-05-03 03:29:09 +0200575 unsigned long tx_maxrate;
576 /*
577 * Number of TX timeouts for this queue
578 * (/sys/class/net/DEV/Q/trans_timeout)
579 */
580 unsigned long trans_timeout;
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700581/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500582 * write-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700583 */
584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
585 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700586 /*
Florian Westphal9b366272016-05-03 16:33:14 +0200587 * Time (in jiffies) of last Tx
Eric Dumazet9d214932009-05-17 20:55:16 -0700588 */
589 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000590
Tom Herbert114cf582011-11-28 16:33:09 +0000591 unsigned long state;
592
593#ifdef CONFIG_BQL
594 struct dql dql;
595#endif
David S. Millere8a04642008-07-17 00:34:19 -0700596} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700597
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000598static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599{
600#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 return q->numa_node;
602#else
Changli Gaob236da62010-12-14 03:09:15 +0000603 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000604#endif
605}
606
607static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608{
609#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 q->numa_node = node;
611#endif
612}
613
Eric Dumazetdf334542010-03-24 19:13:54 +0000614#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000615/*
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
618 */
619struct rps_map {
620 unsigned int len;
621 struct rcu_head rcu;
622 u16 cpus[0];
623};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000624#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000625
Tom Herbertfec5e652010-04-16 16:01:27 -0700626/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628 * tail pointer for that CPU's input queue at the time of last enqueue, and
629 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700630 */
631struct rps_dev_flow {
632 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000633 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700634 unsigned int last_qtail;
635};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000636#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700637
638/*
639 * The rps_dev_flow_table structure contains a table of flow mappings.
640 */
641struct rps_dev_flow_table {
642 unsigned int mask;
643 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700644 struct rps_dev_flow flows[0];
645};
646#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000647 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700648
649/*
650 * The rps_sock_flow_table contains mappings of flows to the last CPU
651 * on which they were processed by the application (set in recvmsg).
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500652 * Each entry is a 32bit value. Upper part is the high-order bits
653 * of flow hash, lower part is CPU number.
Eric Dumazet567e4b72015-02-06 12:59:01 -0800654 * rps_cpu_mask is used to partition the space, depending on number of
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
Eric Dumazet567e4b72015-02-06 12:59:01 -0800657 * meaning we use 32-6=26 bits for the hash.
Tom Herbertfec5e652010-04-16 16:01:27 -0700658 */
659struct rps_sock_flow_table {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800660 u32 mask;
Eric Dumazet93c1af62015-02-08 20:39:13 -0800661
662 u32 ents[0] ____cacheline_aligned_in_smp;
Tom Herbertfec5e652010-04-16 16:01:27 -0700663};
Eric Dumazet567e4b72015-02-06 12:59:01 -0800664#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
Tom Herbertfec5e652010-04-16 16:01:27 -0700665
666#define RPS_NO_CPU 0xffff
667
Eric Dumazet567e4b72015-02-06 12:59:01 -0800668extern u32 rps_cpu_mask;
669extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670
Tom Herbertfec5e652010-04-16 16:01:27 -0700671static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
672 u32 hash)
673{
674 if (table && hash) {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
Tom Herbertfec5e652010-04-16 16:01:27 -0700677
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500678 /* We only give a hint, preemption can change CPU under us */
Eric Dumazet567e4b72015-02-06 12:59:01 -0800679 val |= raw_smp_processor_id();
Tom Herbertfec5e652010-04-16 16:01:27 -0700680
Eric Dumazet567e4b72015-02-06 12:59:01 -0800681 if (table->ents[index] != val)
682 table->ents[index] = val;
Tom Herbertfec5e652010-04-16 16:01:27 -0700683 }
684}
685
Ben Hutchingsc4454772011-01-19 11:03:53 +0000686#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700687bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
688 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000689#endif
Michael Daltona953be52014-01-16 22:23:28 -0800690#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000691
Tom Herbert0a9627f2010-03-16 08:03:29 +0000692/* This structure contains an instance of an RX queue. */
693struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800694#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000695 struct rps_map __rcu *rps_map;
696 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800697#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000698 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000699 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000700} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800701
702/*
703 * RX queue sysfs structures and functions.
704 */
705struct rx_queue_attribute {
706 struct attribute attr;
707 ssize_t (*show)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, char *buf);
709 ssize_t (*store)(struct netdev_rx_queue *queue,
710 struct rx_queue_attribute *attr, const char *buf, size_t len);
711};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800712
Tom Herbertbf264142010-11-26 08:36:09 +0000713#ifdef CONFIG_XPS
714/*
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
717 */
718struct xps_map {
719 unsigned int len;
720 unsigned int alloc_len;
721 struct rcu_head rcu;
722 u16 queues[0];
723};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000724#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Helge Dellerc59f4192015-10-25 10:00:32 +0100725#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726 - sizeof(struct xps_map)) / sizeof(u16))
Tom Herbertbf264142010-11-26 08:36:09 +0000727
728/*
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 */
731struct xps_dev_maps {
732 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000733 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000734};
Alexander Duyck184c4492016-10-28 11:50:13 -0400735#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
736 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
Tom Herbertbf264142010-11-26 08:36:09 +0000737#endif /* CONFIG_XPS */
738
John Fastabend4f57c082011-01-17 08:06:04 +0000739#define TC_MAX_QUEUE 16
740#define TC_BITMASK 15
741/* HW offloaded queuing disciplines txq count and offset maps */
742struct netdev_tc_txq {
743 u16 count;
744 u16 offset;
745};
746
Neerav Parikh68bad942012-01-04 20:23:39 +0000747#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748/*
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
751 */
752struct netdev_fcoe_hbainfo {
753 char manufacturer[64];
754 char serial_number[64];
755 char hardware_version[64];
756 char driver_version[64];
757 char optionrom_version[64];
758 char firmware_version[64];
759 char model[256];
760 char model_description[256];
761};
762#endif
763
Jiri Pirko02637fc2014-11-28 14:34:16 +0100764#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200765
Jiri Pirko02637fc2014-11-28 14:34:16 +0100766/* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200768 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100769struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200771 unsigned char id_len;
772};
773
Scott Feldmand754f982015-07-18 18:24:49 -0700774static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
776{
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
779}
780
Daniel Borkmann99932d42014-02-16 15:55:20 +0100781typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
783
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800784/* These structures hold the attributes of qdisc and classifiers
John Fastabend16e5cc62016-02-16 21:16:43 -0800785 * that are being passed to the netdevice through the setup_tc op.
786 */
787enum {
788 TC_SETUP_MQPRIO,
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800789 TC_SETUP_CLSU32,
Amir Vadai5b33f482016-03-08 12:42:29 +0200790 TC_SETUP_CLSFLOWER,
Yotam Gigib87f7932016-07-21 12:03:12 +0200791 TC_SETUP_MATCHALL,
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100792 TC_SETUP_CLSBPF,
John Fastabend16e5cc62016-02-16 21:16:43 -0800793};
794
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800795struct tc_cls_u32_offload;
796
John Fastabend16e5cc62016-02-16 21:16:43 -0800797struct tc_to_netdev {
798 unsigned int type;
799 union {
800 u8 tc;
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800801 struct tc_cls_u32_offload *cls_u32;
Amir Vadai5b33f482016-03-08 12:42:29 +0200802 struct tc_cls_flower_offload *cls_flower;
Yotam Gigib87f7932016-07-21 12:03:12 +0200803 struct tc_cls_matchall_offload *cls_mall;
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100804 struct tc_cls_bpf_offload *cls_bpf;
John Fastabend16e5cc62016-02-16 21:16:43 -0800805 };
Hadar Hen Zion7091d8c2016-12-01 14:06:37 +0200806 bool egress_dev;
John Fastabend16e5cc62016-02-16 21:16:43 -0800807};
808
Brenden Blancoa7862b42016-07-19 12:16:48 -0700809/* These structures hold the attributes of xdp state that are being passed
810 * to the netdevice through the xdp op.
811 */
812enum xdp_netdev_command {
813 /* Set or clear a bpf program used in the earliest stages of packet
814 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
815 * is responsible for calling bpf_prog_put on any old progs that are
816 * stored. In case of error, the callee need not release the new prog
817 * reference, but on success it takes ownership and must bpf_prog_put
818 * when it is no longer used.
819 */
820 XDP_SETUP_PROG,
821 /* Check if a bpf program is set on the device. The callee should
822 * return true if a program is currently attached and running.
823 */
824 XDP_QUERY_PROG,
825};
826
827struct netdev_xdp {
828 enum xdp_netdev_command command;
829 union {
830 /* XDP_SETUP_PROG */
831 struct bpf_prog *prog;
832 /* XDP_QUERY_PROG */
833 bool prog_attached;
834 };
835};
John Fastabend16e5cc62016-02-16 21:16:43 -0800836
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800837/*
838 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800839 * The following hooks can be defined; unless noted otherwise, they are
840 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800841 *
842 * int (*ndo_init)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500843 * This function is called once when a network device is registered.
844 * The network device can use this for any late stage initialization
845 * or semantic validation. It can fail with an error code which will
846 * be propagated back to register_netdev.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800847 *
848 * void (*ndo_uninit)(struct net_device *dev);
849 * This function is called when device is unregistered or when registration
850 * fails. It is not called if init fails.
851 *
852 * int (*ndo_open)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500853 * This function is called when a network device transitions to the up
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800854 * state.
855 *
856 * int (*ndo_stop)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500857 * This function is called when a network device transitions to the down
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800858 * state.
859 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000860 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
861 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800862 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030863 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
864 * the queue before that can happen; it's for obsolete devices and weird
865 * corner cases, but the stack really does a non-trivial amount
866 * of useless work if you return NETDEV_TX_BUSY.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500867 * Required; cannot be NULL.
Stephen Hemminger00829822008-11-20 20:14:53 -0800868 *
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -0800869 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
870 * struct net_device *dev
871 * netdev_features_t features);
872 * Called by core transmit path to determine if device is capable of
873 * performing offload operations on a given packet. This is to give
874 * the device an opportunity to implement any restrictions that cannot
875 * be otherwise expressed by feature flags. The check is called with
876 * the set of features that the stack has calculated and it returns
877 * those the driver believes to be appropriate.
Eric Dumazetcdba7562016-01-06 06:53:50 -0800878 *
Jason Wangf663dd92014-01-10 16:18:26 +0800879 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100880 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500881 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800882 * transmit queues.
883 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800884 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
885 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500886 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800887 *
888 * void (*ndo_set_rx_mode)(struct net_device *dev);
889 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000890 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500891 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800892 *
893 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
894 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700895 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500896 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800897 *
898 * int (*ndo_validate_addr)(struct net_device *dev);
899 * Test if Media Access Control address is valid for the device.
900 *
901 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500902 * Called when a user requests an ioctl which can't be handled by
903 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800904 * not supported error code.
905 *
906 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
907 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500908 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800909 * interface (PCI) for low level management.
910 *
911 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
912 * Called when a user wants to change the Maximum Transfer Unit
913 * of a device. If not defined, any request to change MTU will
914 * will return an error.
915 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800916 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500917 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800918 * for dev->watchdog ticks.
919 *
Ben Hutchings3cfde792010-07-09 09:11:52 +0000920 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
Eric Dumazet28172732010-07-07 14:58:56 -0700921 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700922 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800923 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000924 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000925 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
926 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700927 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000928 * (which should normally be dev->stats) and return a pointer to
929 * it. The structure may be changed asynchronously only if each
930 * field is written atomically.
931 * 3. Update dev->stats asynchronously and atomically, and define
932 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800933 *
Or Gerlitz3df5b3c2016-11-22 23:09:54 +0200934 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
Nogah Frankel2c9d85d2016-09-16 15:05:36 +0200935 * Return true if this device supports offload stats of this attr_id.
936 *
937 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
938 * void *attr_data)
939 * Get statistics for offload operations by attr_id. Write it into the
940 * attr_data pointer.
941 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530942 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500943 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000944 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800945 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530946 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500947 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000948 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800949 *
950 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000951 *
952 * SR-IOV management functions.
953 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
Moshe Shemesh79aab092016-09-22 12:11:15 +0300954 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
955 * u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400956 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
957 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000958 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000959 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000960 * int (*ndo_get_vf_config)(struct net_device *dev,
961 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300962 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700963 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
964 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300965 *
966 * Enable or disable the VF ability to query its RSS Redirection Table and
967 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500968 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300969 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700970 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +0000971 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
972 * Called to setup 'tc' number of traffic classes in the net device. This
973 * is always called from the stack with the rtnl lock held and netif tx
974 * queues stopped. This allows the netdevice to perform queue management
975 * safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000976 *
Yi Zoue9bce842011-03-09 08:48:03 +0000977 * Fiber Channel over Ethernet (FCoE) offload functions.
978 * int (*ndo_fcoe_enable)(struct net_device *dev);
979 * Called when the FCoE protocol stack wants to start using LLD for FCoE
980 * so the underlying device can perform whatever needed configuration or
981 * initialization to support acceleration of FCoE traffic.
982 *
983 * int (*ndo_fcoe_disable)(struct net_device *dev);
984 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
985 * so the underlying device can perform whatever needed clean-ups to
986 * stop supporting acceleration of FCoE traffic.
987 *
988 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
989 * struct scatterlist *sgl, unsigned int sgc);
990 * Called when the FCoE Initiator wants to initialize an I/O that
991 * is a possible candidate for Direct Data Placement (DDP). The LLD can
992 * perform necessary setup and returns 1 to indicate the device is set up
993 * successfully to perform DDP on this I/O, otherwise this returns 0.
994 *
995 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
996 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
997 * indicated by the FC exchange id 'xid', so the underlying device can
998 * clean up and reuse resources for later DDP requests.
999 *
1000 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1001 * struct scatterlist *sgl, unsigned int sgc);
1002 * Called when the FCoE Target wants to initialize an I/O that
1003 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1004 * perform necessary setup and returns 1 to indicate the device is set up
1005 * successfully to perform DDP on this I/O, otherwise this returns 0.
1006 *
Neerav Parikh68bad942012-01-04 20:23:39 +00001007 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1008 * struct netdev_fcoe_hbainfo *hbainfo);
1009 * Called when the FCoE Protocol stack wants information on the underlying
1010 * device. This information is utilized by the FCoE protocol stack to
1011 * register attributes with Fiber Channel management service as per the
1012 * FC-GS Fabric Device Management Information(FDMI) specification.
1013 *
Yi Zoue9bce842011-03-09 08:48:03 +00001014 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1015 * Called when the underlying device wants to override default World Wide
1016 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1017 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1018 * protocol stack to use.
1019 *
Ben Hutchingsc4454772011-01-19 11:03:53 +00001020 * RFS acceleration.
1021 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1022 * u16 rxq_index, u32 flow_id);
1023 * Set hardware filter for RFS. rxq_index is the target queue index;
1024 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1025 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001026 *
Jiri Pirko8b98a702013-01-03 22:49:02 +00001027 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001028 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1029 * Called to make another netdev an underling.
1030 *
1031 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1032 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +00001033 *
1034 * Feature/offload setting functions.
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -08001035 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1036 * netdev_features_t features);
1037 * Adjusts the requested feature flags according to device-specific
1038 * constraints, and returns the resulting flags. Must not modify
1039 * the device state.
1040 *
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001041 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001042 * Called to update device configuration to new features. Passed
1043 * feature set might be less than what was returned by ndo_fix_features()).
1044 * Must return >0 or -errno if it changed dev->features itself.
1045 *
stephen hemmingeredc7d572012-10-01 12:32:33 +00001046 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1047 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001048 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +00001049 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +00001050 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1051 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001052 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001053 * Deletes the FDB entry from dev coresponding to addr.
1054 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001055 * struct net_device *dev, struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001056 * int *idx)
John Fastabend77162022012-04-15 06:43:56 +00001057 * Used to add FDB entries to dump requests. Implementers should add
1058 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001059 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001060 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1061 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001062 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001063 * struct net_device *dev, u32 filter_mask,
1064 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001065 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1066 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001067 *
1068 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1069 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1070 * which do not represent real hardware may define this to allow their
1071 * userspace components to manage their virtual carrier state. Devices
1072 * that determine carrier state from physical hardware properties (eg
1073 * network cables) or protocol-dependent mechanisms (eg
1074 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001075 *
1076 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001077 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001078 * Called to get ID of physical port of this device. If driver does
1079 * not implement this, it is assumed that the hw is not able to have
1080 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001081 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001082 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1083 * struct udp_tunnel_info *ti);
1084 * Called by UDP tunnel to notify a driver about the UDP port and socket
1085 * address family that a UDP tunnel is listnening to. It is called only
1086 * when a new port starts listening. The operation is protected by the
1087 * RTNL.
1088 *
1089 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1090 * struct udp_tunnel_info *ti);
1091 * Called by UDP tunnel to notify the driver about a UDP port and socket
1092 * address family that the UDP tunnel is not listening to anymore. The
1093 * operation is protected by the RTNL.
1094 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001095 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1096 * struct net_device *dev)
1097 * Called by upper layer devices to accelerate switching or other
1098 * station functionality into hardware. 'pdev is the lowerdev
1099 * to use for the offload and 'dev' is the net device that will
1100 * back the offload. Returns a pointer to the private structure
1101 * the upper layer will maintain.
1102 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1103 * Called by upper layer device to delete the station created
1104 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1105 * the station and priv is the structure returned by the add
1106 * operation.
1107 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1108 * struct net_device *dev,
1109 * void *priv);
1110 * Callback to use for xmit over the accelerated station. This
1111 * is used in place of ndo_start_xmit on accelerated net
1112 * devices.
John Fastabend822b3b22015-03-18 14:57:33 +02001113 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1114 * int queue_index, u32 maxrate);
1115 * Called when a user wants to set a max-rate limitation of specific
1116 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001117 * int (*ndo_get_iflink)(const struct net_device *dev);
1118 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001119 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001120 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001121 * This function is used to pass protocol port error state information
1122 * to the switch driver. The switch driver can react to the proto_down
1123 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001124 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1125 * This function is used to get egress tunnel information for given skb.
1126 * This is useful for retrieving outer tunnel header parameters while
1127 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001128 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1129 * This function is used to specify the headroom that the skb must
1130 * consider when allocation skb during packet reception. Setting
1131 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001132 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001133 * default value.
Brenden Blancoa7862b42016-07-19 12:16:48 -07001134 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1135 * This function is used to set or query state related to XDP on the
1136 * netdevice. See definition of enum xdp_netdev_command for details.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001137 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001138 */
1139struct net_device_ops {
1140 int (*ndo_init)(struct net_device *dev);
1141 void (*ndo_uninit)(struct net_device *dev);
1142 int (*ndo_open)(struct net_device *dev);
1143 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001144 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1145 struct net_device *dev);
1146 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1147 struct net_device *dev,
1148 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001149 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001150 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001151 void *accel_priv,
1152 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001153 void (*ndo_change_rx_flags)(struct net_device *dev,
1154 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001155 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001156 int (*ndo_set_mac_address)(struct net_device *dev,
1157 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001158 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001159 int (*ndo_do_ioctl)(struct net_device *dev,
1160 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001161 int (*ndo_set_config)(struct net_device *dev,
1162 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001163 int (*ndo_change_mtu)(struct net_device *dev,
1164 int new_mtu);
1165 int (*ndo_neigh_setup)(struct net_device *dev,
1166 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001167 void (*ndo_tx_timeout) (struct net_device *dev);
1168
Eric Dumazet28172732010-07-07 14:58:56 -07001169 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1170 struct rtnl_link_stats64 *storage);
Or Gerlitz3df5b3c2016-11-22 23:09:54 +02001171 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
Nogah Frankel2c9d85d2016-09-16 15:05:36 +02001172 int (*ndo_get_offload_stats)(int attr_id,
1173 const struct net_device *dev,
1174 void *attr_data);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001175 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1176
Jiri Pirko8e586132011-12-08 19:52:37 -05001177 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001178 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001179 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001180 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001181#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001182 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001183 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001184 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001185 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001186#endif
Cong Wange0d10952013-08-01 11:10:25 +08001187#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir8b80cda2013-07-10 17:13:26 +03001188 int (*ndo_busy_poll)(struct napi_struct *dev);
Eliezer Tamir06021292013-06-10 11:39:50 +03001189#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001190 int (*ndo_set_vf_mac)(struct net_device *dev,
1191 int queue, u8 *mac);
1192 int (*ndo_set_vf_vlan)(struct net_device *dev,
Moshe Shemesh79aab092016-09-22 12:11:15 +03001193 int queue, u16 vlan,
1194 u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001195 int (*ndo_set_vf_rate)(struct net_device *dev,
1196 int vf, int min_tx_rate,
1197 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001198 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1199 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001200 int (*ndo_set_vf_trust)(struct net_device *dev,
1201 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001202 int (*ndo_get_vf_config)(struct net_device *dev,
1203 int vf,
1204 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001205 int (*ndo_set_vf_link_state)(struct net_device *dev,
1206 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001207 int (*ndo_get_vf_stats)(struct net_device *dev,
1208 int vf,
1209 struct ifla_vf_stats
1210 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001211 int (*ndo_set_vf_port)(struct net_device *dev,
1212 int vf,
1213 struct nlattr *port[]);
1214 int (*ndo_get_vf_port)(struct net_device *dev,
1215 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001216 int (*ndo_set_vf_guid)(struct net_device *dev,
1217 int vf, u64 guid,
1218 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001219 int (*ndo_set_vf_rss_query_en)(
1220 struct net_device *dev,
1221 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001222 int (*ndo_setup_tc)(struct net_device *dev,
1223 u32 handle,
1224 __be16 protocol,
1225 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001226#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001227 int (*ndo_fcoe_enable)(struct net_device *dev);
1228 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001229 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1230 u16 xid,
1231 struct scatterlist *sgl,
1232 unsigned int sgc);
1233 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1234 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001235 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1236 u16 xid,
1237 struct scatterlist *sgl,
1238 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001239 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1240 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001241#endif
1242
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001243#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001244#define NETDEV_FCOE_WWNN 0
1245#define NETDEV_FCOE_WWPN 1
1246 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1247 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001248#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001249
Ben Hutchingsc4454772011-01-19 11:03:53 +00001250#ifdef CONFIG_RFS_ACCEL
1251 int (*ndo_rx_flow_steer)(struct net_device *dev,
1252 const struct sk_buff *skb,
1253 u16 rxq_index,
1254 u32 flow_id);
1255#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001256 int (*ndo_add_slave)(struct net_device *dev,
1257 struct net_device *slave_dev);
1258 int (*ndo_del_slave)(struct net_device *dev,
1259 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001260 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1261 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001262 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001263 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001264 int (*ndo_neigh_construct)(struct net_device *dev,
1265 struct neighbour *n);
1266 void (*ndo_neigh_destroy)(struct net_device *dev,
1267 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001268
1269 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001270 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001271 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001272 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001273 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001274 u16 flags);
1275 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001276 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001277 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001278 const unsigned char *addr,
1279 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001280 int (*ndo_fdb_dump)(struct sk_buff *skb,
1281 struct netlink_callback *cb,
1282 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001283 struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001284 int *idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001285
1286 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001287 struct nlmsghdr *nlh,
1288 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001289 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1290 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001291 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001292 u32 filter_mask,
1293 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001294 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001295 struct nlmsghdr *nlh,
1296 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001297 int (*ndo_change_carrier)(struct net_device *dev,
1298 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001299 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001300 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001301 int (*ndo_get_phys_port_name)(struct net_device *dev,
1302 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001303 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1304 struct udp_tunnel_info *ti);
1305 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1306 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001307 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1308 struct net_device *dev);
1309 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1310 void *priv);
1311
1312 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1313 struct net_device *dev,
1314 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001315 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001316 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1317 int queue_index,
1318 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001319 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001320 int (*ndo_change_proto_down)(struct net_device *dev,
1321 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001322 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1323 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001324 void (*ndo_set_rx_headroom)(struct net_device *dev,
1325 int needed_headroom);
Brenden Blancoa7862b42016-07-19 12:16:48 -07001326 int (*ndo_xdp)(struct net_device *dev,
1327 struct netdev_xdp *xdp);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001328};
1329
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001330/**
1331 * enum net_device_priv_flags - &struct net_device priv_flags
1332 *
1333 * These are the &struct net_device, they are only set internally
1334 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001335 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001336 * during any kernel release.
1337 *
1338 * You should have a pretty good reason to be extending these flags.
1339 *
1340 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1341 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001342 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001343 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001344 * @IFF_WAN_HDLC: WAN HDLC device
1345 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1346 * release skb->dst
1347 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1348 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1349 * @IFF_MACVLAN_PORT: device used as macvlan port
1350 * @IFF_BRIDGE_PORT: device used as bridge port
1351 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1352 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1353 * @IFF_UNICAST_FLT: Supports unicast filtering
1354 * @IFF_TEAM_PORT: device used as team port
1355 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1356 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1357 * change when it's running
1358 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001359 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1360 * underlying stacked devices
1361 * @IFF_IPVLAN_MASTER: IPvlan master device
1362 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001363 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001364 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001365 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001366 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001367 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001368 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001369 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1370 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001371 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001372 */
1373enum netdev_priv_flags {
1374 IFF_802_1Q_VLAN = 1<<0,
1375 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001376 IFF_BONDING = 1<<2,
1377 IFF_ISATAP = 1<<3,
1378 IFF_WAN_HDLC = 1<<4,
1379 IFF_XMIT_DST_RELEASE = 1<<5,
1380 IFF_DONT_BRIDGE = 1<<6,
1381 IFF_DISABLE_NETPOLL = 1<<7,
1382 IFF_MACVLAN_PORT = 1<<8,
1383 IFF_BRIDGE_PORT = 1<<9,
1384 IFF_OVS_DATAPATH = 1<<10,
1385 IFF_TX_SKB_SHARING = 1<<11,
1386 IFF_UNICAST_FLT = 1<<12,
1387 IFF_TEAM_PORT = 1<<13,
1388 IFF_SUPP_NOFCS = 1<<14,
1389 IFF_LIVE_ADDR_CHANGE = 1<<15,
1390 IFF_MACVLAN = 1<<16,
1391 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1392 IFF_IPVLAN_MASTER = 1<<18,
1393 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001394 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001395 IFF_NO_QUEUE = 1<<21,
1396 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001397 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001398 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001399 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001400 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001401 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001402};
1403
1404#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1405#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001406#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001407#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001408#define IFF_WAN_HDLC IFF_WAN_HDLC
1409#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1410#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1411#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1412#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1413#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1414#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1415#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1416#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1417#define IFF_TEAM_PORT IFF_TEAM_PORT
1418#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1419#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1420#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001421#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001422#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1423#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001424#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001425#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001426#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001427#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001428#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001429#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001430#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001431
Karoly Kemeny536721b2014-07-30 20:27:36 +02001432/**
1433 * struct net_device - The DEVICE structure.
1434 * Actually, this whole structure is a big mistake. It mixes I/O
1435 * data with strictly "high-level" data, and it has to know about
1436 * almost every data structure used in the INET module.
1437 *
1438 * @name: This is the first field of the "visible" part of this structure
1439 * (i.e. as seen by users in the "Space.c" file). It is the name
1440 * of the interface.
1441 *
1442 * @name_hlist: Device name hash chain, please keep it close to name[]
1443 * @ifalias: SNMP alias
1444 * @mem_end: Shared memory end
1445 * @mem_start: Shared memory start
1446 * @base_addr: Device I/O address
1447 * @irq: Device IRQ number
1448 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001449 * @carrier_changes: Stats to monitor carrier on<->off transitions
1450 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001451 * @state: Generic network queuing layer state, see netdev_state_t
1452 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001453 * @napi_list: List entry used for polling NAPI devices
1454 * @unreg_list: List entry when we are unregistering the
1455 * device; see the function unregister_netdev
1456 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001457 * @ptype_all: Device-specific packet handlers for all protocols
1458 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001459 *
1460 * @adj_list: Directly linked devices, like slaves for bonding
Karoly Kemeny536721b2014-07-30 20:27:36 +02001461 * @features: Currently active device features
1462 * @hw_features: User-changeable features
1463 *
1464 * @wanted_features: User-requested features
1465 * @vlan_features: Mask of features inheritable by VLAN devices
1466 *
1467 * @hw_enc_features: Mask of features inherited by encapsulating devices
1468 * This field indicates what encapsulation
1469 * offloads the hardware is capable of doing,
1470 * and drivers will need to set them appropriately.
1471 *
1472 * @mpls_features: Mask of features inheritable by MPLS
1473 *
1474 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001475 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001476 *
1477 * @stats: Statistics struct, which was left as a legacy, use
1478 * rtnl_link_stats64 instead
1479 *
1480 * @rx_dropped: Dropped packets by core network,
1481 * do not use this in drivers
1482 * @tx_dropped: Dropped packets by core network,
1483 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001484 * @rx_nohandler: nohandler dropped packets by core network on
1485 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001486 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001487 * @wireless_handlers: List of functions to handle Wireless Extensions,
1488 * instead of ioctl,
1489 * see <net/iw_handler.h> for details.
1490 * @wireless_data: Instance data managed by the core of wireless extensions
1491 *
1492 * @netdev_ops: Includes several pointers to callbacks,
1493 * if one wants to override the ndo_*() functions
1494 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001495 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1496 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001497 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001498 * of Layer 2 headers.
1499 *
1500 * @flags: Interface flags (a la BSD)
1501 * @priv_flags: Like 'flags' but invisible to userspace,
1502 * see if.h for the definitions
1503 * @gflags: Global flags ( kept as legacy )
1504 * @padded: How much padding added by alloc_netdev()
1505 * @operstate: RFC2863 operstate
1506 * @link_mode: Mapping policy to operstate
1507 * @if_port: Selectable AUI, TP, ...
1508 * @dma: DMA channel
1509 * @mtu: Interface MTU value
Jarod Wilson61e84622016-10-07 22:04:33 -04001510 * @min_mtu: Interface Minimum MTU value
1511 * @max_mtu: Interface Maximum MTU value
Karoly Kemeny536721b2014-07-30 20:27:36 +02001512 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001513 * @hard_header_len: Maximum hardware header length.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001514 *
1515 * @needed_headroom: Extra headroom the hardware may need, but not in all
1516 * cases can this be guaranteed
1517 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1518 * cases can this be guaranteed. Some cases also use
1519 * LL_MAX_HEADER instead to allocate the skb
1520 *
1521 * interface address info:
1522 *
1523 * @perm_addr: Permanent hw address
1524 * @addr_assign_type: Hw address assignment type
1525 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001526 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001527 * @dev_id: Used to differentiate devices that share
1528 * the same link layer address
1529 * @dev_port: Used to differentiate devices that share
1530 * the same function
1531 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001532 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001533 * has been enabled due to the need to listen to
1534 * additional unicast addresses in a device that
1535 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001536 * @uc: unicast mac addresses
1537 * @mc: multicast mac addresses
1538 * @dev_addrs: list of device hw addresses
1539 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001540 * @promiscuity: Number of times the NIC is told to work in
1541 * promiscuous mode; if it becomes 0 the NIC will
1542 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001543 * @allmulti: Counter, enables or disables allmulticast mode
1544 *
1545 * @vlan_info: VLAN info
1546 * @dsa_ptr: dsa specific data
1547 * @tipc_ptr: TIPC specific data
1548 * @atalk_ptr: AppleTalk link
1549 * @ip_ptr: IPv4 specific data
1550 * @dn_ptr: DECnet specific data
1551 * @ip6_ptr: IPv6 specific data
1552 * @ax25_ptr: AX.25 specific data
1553 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1554 *
1555 * @last_rx: Time of last Rx
1556 * @dev_addr: Hw address (before bcast,
1557 * because most packets are unicast)
1558 *
1559 * @_rx: Array of RX queues
1560 * @num_rx_queues: Number of RX queues
1561 * allocated at register_netdev() time
1562 * @real_num_rx_queues: Number of RX queues currently active in device
1563 *
1564 * @rx_handler: handler for received packets
1565 * @rx_handler_data: XXX: need comments on this one
1566 * @ingress_queue: XXX: need comments on this one
1567 * @broadcast: hw bcast address
1568 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001569 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1570 * indexed by RX queue number. Assigned by driver.
1571 * This must only be set if the ndo_rx_flow_steer
1572 * operation is defined
1573 * @index_hlist: Device index hash chain
1574 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001575 * @_tx: Array of TX queues
1576 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1577 * @real_num_tx_queues: Number of TX queues currently active in device
1578 * @qdisc: Root qdisc from userspace point of view
1579 * @tx_queue_len: Max frames per queue allowed
1580 * @tx_global_lock: XXX: need comments on this one
1581 *
1582 * @xps_maps: XXX: need comments on this one
1583 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001584 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001585 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001586 * @watchdog_timer: List of timers
1587 *
1588 * @pcpu_refcnt: Number of references to this device
1589 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001590 * @link_watch_list: XXX: need comments on this one
1591 *
1592 * @reg_state: Register/unregister state machine
1593 * @dismantle: Device is going to be freed
1594 * @rtnl_link_state: This enum represents the phases of creating
1595 * a new link
1596 *
1597 * @destructor: Called from unregister,
1598 * can be used to call free_netdev
1599 * @npinfo: XXX: need comments on this one
1600 * @nd_net: Network namespace this network device is inside
1601 *
1602 * @ml_priv: Mid-layer private
1603 * @lstats: Loopback statistics
1604 * @tstats: Tunnel statistics
1605 * @dstats: Dummy statistics
1606 * @vstats: Virtual ethernet statistics
1607 *
1608 * @garp_port: GARP
1609 * @mrp_port: MRP
1610 *
1611 * @dev: Class/net/name entry
1612 * @sysfs_groups: Space for optional device, statistics and wireless
1613 * sysfs groups
1614 *
1615 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1616 * @rtnl_link_ops: Rtnl_link_ops
1617 *
1618 * @gso_max_size: Maximum size of generic segmentation offload
1619 * @gso_max_segs: Maximum number of segments that can be passed to the
1620 * NIC for GSO
1621 *
1622 * @dcbnl_ops: Data Center Bridging netlink ops
1623 * @num_tc: Number of traffic classes in the net device
1624 * @tc_to_txq: XXX: need comments on this one
Randy Dunlap920c1cd2016-11-21 18:28:36 -08001625 * @prio_tc_map: XXX: need comments on this one
Karoly Kemeny536721b2014-07-30 20:27:36 +02001626 *
1627 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1628 *
1629 * @priomap: XXX: need comments on this one
1630 * @phydev: Physical device may attach itself
1631 * for hardware timestamping
1632 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001633 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1634 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001635 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001636 * @proto_down: protocol port state information can be sent to the
1637 * switch driver and used to set the phys state of the
1638 * switch port.
1639 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 * FIXME: cleanup struct net_device such that network protocol info
1641 * moves out.
1642 */
1643
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001644struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001646 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001647 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 /*
1649 * I/O specific fields
1650 * FIXME: Merge these and struct ifmap into one
1651 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001652 unsigned long mem_end;
1653 unsigned long mem_start;
1654 unsigned long base_addr;
1655 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001657 atomic_t carrier_changes;
1658
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001660 * Some hardware also needs these fields (state,dev_list,
1661 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 * part of the usual set specified in Space.c.
1663 */
1664
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 unsigned long state;
1666
Pavel Emelianov7562f872007-05-03 15:13:45 -07001667 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001668 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001669 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001670 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001671 struct list_head ptype_all;
1672 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001673
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001674 struct {
1675 struct list_head upper;
1676 struct list_head lower;
1677 } adj_list;
1678
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001679 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001680 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001681 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001682 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001683 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001684 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001685 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001686
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001688 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689
Rusty Russellc45d2862007-03-28 14:29:08 -07001690 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001691
Eric Dumazet015f0682014-03-27 08:45:56 -07001692 atomic_long_t rx_dropped;
1693 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001694 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695
Johannes Bergb86e0282007-04-26 20:48:23 -07001696#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001697 const struct iw_handler_def *wireless_handlers;
1698 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001699#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001700 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001701 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001702#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001703 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001704#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001705#ifdef CONFIG_NET_L3_MASTER_DEV
1706 const struct l3mdev_ops *l3mdev_ops;
1707#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001708#if IS_ENABLED(CONFIG_IPV6)
1709 const struct ndisc_ops *ndisc_ops;
1710#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001712 const struct header_ops *header_ops;
1713
Karoly Kemeny536721b2014-07-30 20:27:36 +02001714 unsigned int flags;
1715 unsigned int priv_flags;
1716
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001718 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719
Karoly Kemeny536721b2014-07-30 20:27:36 +02001720 unsigned char operstate;
1721 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001722
Karoly Kemeny536721b2014-07-30 20:27:36 +02001723 unsigned char if_port;
1724 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001725
Karoly Kemeny536721b2014-07-30 20:27:36 +02001726 unsigned int mtu;
Jarod Wilson61e84622016-10-07 22:04:33 -04001727 unsigned int min_mtu;
1728 unsigned int max_mtu;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001729 unsigned short type;
1730 unsigned short hard_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
Johannes Bergf5184d22008-05-12 20:48:31 -07001732 unsigned short needed_headroom;
1733 unsigned short needed_tailroom;
1734
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001736 unsigned char perm_addr[MAX_ADDR_LEN];
1737 unsigned char addr_assign_type;
1738 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001739 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001740 unsigned short dev_id;
1741 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001742 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001743 unsigned char name_assign_type;
1744 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001745 struct netdev_hw_addr_list uc;
1746 struct netdev_hw_addr_list mc;
1747 struct netdev_hw_addr_list dev_addrs;
1748
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001749#ifdef CONFIG_SYSFS
1750 struct kset *queues_kset;
1751#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001752 unsigned int promiscuity;
1753 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001756 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001757
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001758#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001759 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001760#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001761#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001762 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001763#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001764#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001765 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001766#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001767 void *atalk_ptr;
1768 struct in_device __rcu *ip_ptr;
1769 struct dn_dev __rcu *dn_ptr;
1770 struct inet6_dev __rcu *ip6_ptr;
1771 void *ax25_ptr;
1772 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001773 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001774#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1775 struct mpls_dev __rcu *mpls_ptr;
1776#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001778/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001779 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001780 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001781 unsigned long last_rx;
Eric Dumazet4dc89132010-08-31 07:40:16 +00001782
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001783 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001784 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001785
Michael Daltona953be52014-01-16 22:23:28 -08001786#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001787 struct netdev_rx_queue *_rx;
1788
Tom Herbert0a9627f2010-03-16 08:03:29 +00001789 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001790 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001791#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001792
Eric Dumazet3b47d302014-11-06 21:09:44 -08001793 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001794 rx_handler_func_t __rcu *rx_handler;
1795 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001796
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001797#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001798 struct tcf_proto __rcu *ingress_cl_list;
1799#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001800 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001801#ifdef CONFIG_NETFILTER_INGRESS
Aaron Conolee3b37f12016-09-21 11:35:07 -04001802 struct nf_hook_entry __rcu *nf_hooks_ingress;
Pablo Neirae687ad62015-05-13 18:19:38 +02001803#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001804
Karoly Kemeny536721b2014-07-30 20:27:36 +02001805 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001806#ifdef CONFIG_RFS_ACCEL
1807 struct cpu_rmap *rx_cpu_rmap;
1808#endif
1809 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001810
1811/*
1812 * Cache lines mostly used on transmit path
1813 */
David S. Millere8a04642008-07-17 00:34:19 -07001814 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1815 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001816 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001817 struct Qdisc *qdisc;
Jiri Kosina59cc1f62016-08-10 11:05:15 +02001818#ifdef CONFIG_NET_SCHED
1819 DECLARE_HASHTABLE (qdisc_hash, 4);
1820#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001821 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001822 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001823 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001824
Tom Herbertbf264142010-11-26 08:36:09 +00001825#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001826 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001827#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001828#ifdef CONFIG_NET_CLS_ACT
1829 struct tcf_proto __rcu *egress_cl_list;
1830#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001831
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001832 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001833 struct timer_list watchdog_timer;
1834
Eric Dumazet29b44332010-10-11 10:22:12 +00001835 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837
Eric Dumazete014deb2009-11-17 05:59:21 +00001838 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001839
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001841 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842 NETREG_UNREGISTERING, /* called unregister_netdevice */
1843 NETREG_UNREGISTERED, /* completed unregister todo */
1844 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001845 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001846 } reg_state:8;
1847
Karoly Kemeny536721b2014-07-30 20:27:36 +02001848 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001849
1850 enum {
1851 RTNL_LINK_INITIALIZED,
1852 RTNL_LINK_INITIALIZING,
1853 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001855 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001858 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859#endif
David S. Millereae792b2008-07-15 03:03:33 -07001860
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001861 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001862
David S. Miller49517042008-05-12 03:29:11 -07001863 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001864 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001865 void *ml_priv;
1866 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001867 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001868 struct pcpu_dstats __percpu *dstats;
1869 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001870 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001871
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001872 struct garp_port __rcu *garp_port;
David Wardfebf0182013-02-08 17:17:06 +00001873 struct mrp_port __rcu *mrp_port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001875 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001876 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001877 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001878
Patrick McHardy38f7b872007-06-13 12:03:51 -07001879 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001880
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001881 /* for setting kernel sock attribute on TCP connection setup */
1882#define GSO_MAX_SIZE 65536
1883 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001884#define GSO_MAX_SEGS 65535
1885 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001886
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001887#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001888 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001889#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001890 u8 num_tc;
1891 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1892 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001893
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001894#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001895 unsigned int fcoe_ddp_xid;
1896#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001897#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001898 struct netprio_map __rcu *priomap;
1899#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001900 struct phy_device *phydev;
1901 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001902 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001903 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001905#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906
1907#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908
David S. Millere8a04642008-07-17 00:34:19 -07001909static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001910int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1911{
1912 return dev->prio_tc_map[prio & TC_BITMASK];
1913}
1914
1915static inline
1916int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1917{
1918 if (tc >= dev->num_tc)
1919 return -EINVAL;
1920
1921 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1922 return 0;
1923}
1924
Alexander Duyck8d059b02016-10-28 11:43:49 -04001925int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
Alexander Duyck9cf1f6a2016-10-28 11:43:20 -04001926void netdev_reset_tc(struct net_device *dev);
1927int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1928int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
John Fastabend4f57c082011-01-17 08:06:04 +00001929
1930static inline
1931int netdev_get_num_tc(struct net_device *dev)
1932{
1933 return dev->num_tc;
1934}
1935
1936static inline
David S. Millere8a04642008-07-17 00:34:19 -07001937struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1938 unsigned int index)
1939{
1940 return &dev->_tx[index];
1941}
1942
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001943static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1944 const struct sk_buff *skb)
1945{
1946 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1947}
1948
David S. Millere8a04642008-07-17 00:34:19 -07001949static inline void netdev_for_each_tx_queue(struct net_device *dev,
1950 void (*f)(struct net_device *,
1951 struct netdev_queue *,
1952 void *),
1953 void *arg)
1954{
1955 unsigned int i;
1956
1957 for (i = 0; i < dev->num_tx_queues; i++)
1958 f(dev, &dev->_tx[i], arg);
1959}
1960
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001961#define netdev_lockdep_set_classes(dev) \
1962{ \
1963 static struct lock_class_key qdisc_tx_busylock_key; \
1964 static struct lock_class_key qdisc_running_key; \
1965 static struct lock_class_key qdisc_xmit_lock_key; \
1966 static struct lock_class_key dev_addr_list_lock_key; \
1967 unsigned int i; \
1968 \
1969 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1970 (dev)->qdisc_running_key = &qdisc_running_key; \
1971 lockdep_set_class(&(dev)->addr_list_lock, \
1972 &dev_addr_list_lock_key); \
1973 for (i = 0; i < (dev)->num_tx_queues; i++) \
1974 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1975 &qdisc_xmit_lock_key); \
1976}
1977
Joe Perchesf629d202013-09-26 14:48:15 -07001978struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001979 struct sk_buff *skb,
1980 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001981
Paolo Abeni871b6422016-02-26 10:45:37 +01001982/* returns the headroom that the master device needs to take in account
1983 * when forwarding to this dev
1984 */
1985static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1986{
1987 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1988}
1989
1990static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1991{
1992 if (dev->netdev_ops->ndo_set_rx_headroom)
1993 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1994}
1995
1996/* set the device rx headroom to the dev's default */
1997static inline void netdev_reset_rx_headroom(struct net_device *dev)
1998{
1999 netdev_set_rx_headroom(dev, -1);
2000}
2001
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002002/*
2003 * Net namespace inlines
2004 */
2005static inline
2006struct net *dev_net(const struct net_device *dev)
2007{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002008 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002009}
2010
2011static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002012void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002013{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05002014 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002015}
2016
Florian Fainelli3e8a72d2014-08-27 17:04:46 -07002017static inline bool netdev_uses_dsa(struct net_device *dev)
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00002018{
Alexander Duyck3fc88672014-09-12 23:12:46 -07002019#if IS_ENABLED(CONFIG_NET_DSA)
Florian Fainelli5aed85c2014-08-27 17:04:52 -07002020 if (dev->dsa_ptr != NULL)
2021 return dsa_uses_tagged_protocol(dev->dsa_ptr);
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00002022#endif
Florian Fainelli5aed85c2014-08-27 17:04:52 -07002023 return false;
Lennert Buytenhek396138f2008-10-07 13:46:07 +00002024}
2025
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002026/**
2027 * netdev_priv - access network device private data
2028 * @dev: network device
2029 *
2030 * Get network device private data
2031 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002032static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002034 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035}
2036
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037/* Set the sysfs physical device reference for the network logical device
2038 * if set prior to registration will cause a symlink during initialization.
2039 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002040#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041
Marcel Holtmann384912e2009-08-31 21:08:19 +00002042/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002043 * fine-grained identification of different network device types. For
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002044 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
Marcel Holtmann384912e2009-08-31 21:08:19 +00002045 */
2046#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2047
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002048/* Default NAPI poll() weight
2049 * Device drivers are strongly advised to not use bigger value
2050 */
2051#define NAPI_POLL_WEIGHT 64
2052
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002053/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002054 * netif_napi_add - initialize a NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002055 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002056 * @napi: NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002057 * @poll: polling function
2058 * @weight: default weight
2059 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002060 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2061 * *any* of the other NAPI-related functions.
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002062 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002063void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2064 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002065
Alexander Duyckd8156532008-07-08 15:13:05 -07002066/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002067 * netif_tx_napi_add - initialize a NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002068 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002069 * @napi: NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002070 * @poll: polling function
2071 * @weight: default weight
2072 *
2073 * This variant of netif_napi_add() should be used from drivers using NAPI
2074 * to exclusively poll a TX queue.
2075 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2076 */
2077static inline void netif_tx_napi_add(struct net_device *dev,
2078 struct napi_struct *napi,
2079 int (*poll)(struct napi_struct *, int),
2080 int weight)
2081{
2082 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2083 netif_napi_add(dev, napi, poll, weight);
2084}
2085
2086/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002087 * netif_napi_del - remove a NAPI context
2088 * @napi: NAPI context
Alexander Duyckd8156532008-07-08 15:13:05 -07002089 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002090 * netif_napi_del() removes a NAPI context from the network device NAPI list
Alexander Duyckd8156532008-07-08 15:13:05 -07002091 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002092void netif_napi_del(struct napi_struct *napi);
2093
2094struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002095 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002096 void *frag0;
Herbert Xu78a478d2009-05-26 18:50:21 +00002097
Herbert Xu74895942009-05-26 18:50:27 +00002098 /* Length of frag0. */
2099 unsigned int frag0_len;
2100
Herbert Xu86911732009-01-29 14:19:50 +00002101 /* This indicates where we are processing relative to skb->data. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002102 int data_offset;
Herbert Xu86911732009-01-29 14:19:50 +00002103
Herbert Xud565b0a2008-12-15 23:38:52 -08002104 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002105 u16 flush;
2106
2107 /* Save the IP ID here and check when we get to the transport layer */
2108 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002109
2110 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002111 u16 count;
2112
Tom Herbert15e23962015-02-10 16:30:31 -08002113 /* Start offset for remote checksum offload */
2114 u16 gro_remcsum_start;
2115
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002116 /* jiffies when first packet was created/queued */
2117 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002118
Tom Herbertafe93322014-09-17 12:25:57 -07002119 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002120 u16 proto;
2121
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002122 /* This is non-zero if the packet may be of the same flow. */
2123 u8 same_flow:1;
2124
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002125 /* Used in tunnel GRO receive */
2126 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002127
2128 /* GRO checksum is valid */
2129 u8 csum_valid:1;
2130
Tom Herbert662880f2014-08-27 21:26:56 -07002131 /* Number of checksums via CHECKSUM_UNNECESSARY */
2132 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002133
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002134 /* Free the skb? */
2135 u8 free:2;
2136#define NAPI_GRO_FREE 1
2137#define NAPI_GRO_FREE_STOLEN_HEAD 2
2138
Tom Herbertefc98d02014-10-03 15:48:08 -07002139 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2140 u8 is_ipv6:1;
2141
Alexander Duycka0ca1532016-04-05 09:13:39 -07002142 /* Used in GRE, set in fou/gue_gro_receive */
2143 u8 is_fou:1;
2144
Alexander Duyck15305452016-04-10 21:44:57 -04002145 /* Used to determine if flush_id can be ignored */
2146 u8 is_atomic:1;
2147
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002148 /* Number of gro_receive callbacks this packet already went through */
2149 u8 recursion_counter:4;
2150
2151 /* 1 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002152
Jerry Chubf5a7552014-01-07 10:23:19 -08002153 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2154 __wsum csum;
2155
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002156 /* used in skb_gro_receive() slow path */
2157 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002158};
2159
2160#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002161
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002162#define GRO_RECURSION_LIMIT 15
2163static inline int gro_recursion_inc_test(struct sk_buff *skb)
2164{
2165 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2166}
2167
2168typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2169static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2170 struct sk_buff **head,
2171 struct sk_buff *skb)
2172{
2173 if (unlikely(gro_recursion_inc_test(skb))) {
2174 NAPI_GRO_CB(skb)->flush |= 1;
2175 return NULL;
2176 }
2177
2178 return cb(head, skb);
2179}
2180
2181typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2182 struct sk_buff *);
2183static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2184 struct sock *sk,
2185 struct sk_buff **head,
2186 struct sk_buff *skb)
2187{
2188 if (unlikely(gro_recursion_inc_test(skb))) {
2189 NAPI_GRO_CB(skb)->flush |= 1;
2190 return NULL;
2191 }
2192
2193 return cb(sk, head, skb);
2194}
2195
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002197 __be16 type; /* This is really htons(ether_type). */
2198 struct net_device *dev; /* NULL is wildcarded here */
2199 int (*func) (struct sk_buff *,
2200 struct net_device *,
2201 struct packet_type *,
2202 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002203 bool (*id_match)(struct packet_type *ptype,
2204 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205 void *af_packet_priv;
2206 struct list_head list;
2207};
2208
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002209struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2211 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002213 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002214 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002215};
2216
2217struct packet_offload {
2218 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002219 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002220 struct offload_callbacks callbacks;
2221 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222};
2223
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002224/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002225struct pcpu_sw_netstats {
2226 u64 rx_packets;
2227 u64 rx_bytes;
2228 u64 tx_packets;
2229 u64 tx_bytes;
2230 struct u64_stats_sync syncp;
2231};
2232
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002233#define __netdev_alloc_pcpu_stats(type, gfp) \
2234({ \
2235 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2236 if (pcpu_stats) { \
2237 int __cpu; \
2238 for_each_possible_cpu(__cpu) { \
2239 typeof(type) *stat; \
2240 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2241 u64_stats_init(&stat->syncp); \
2242 } \
2243 } \
2244 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002245})
2246
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002247#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002248 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002249
Jiri Pirko764f5e52015-12-03 12:12:12 +01002250enum netdev_lag_tx_type {
2251 NETDEV_LAG_TX_TYPE_UNKNOWN,
2252 NETDEV_LAG_TX_TYPE_RANDOM,
2253 NETDEV_LAG_TX_TYPE_BROADCAST,
2254 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2255 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2256 NETDEV_LAG_TX_TYPE_HASH,
2257};
2258
2259struct netdev_lag_upper_info {
2260 enum netdev_lag_tx_type tx_type;
2261};
2262
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002263struct netdev_lag_lower_state_info {
2264 u8 link_up : 1,
2265 tx_enabled : 1;
2266};
2267
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268#include <linux/notifier.h>
2269
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002270/* netdevice notifier chain. Please remember to update the rtnetlink
2271 * notification exclusion list in rtnetlink_event() when adding new
2272 * types.
2273 */
2274#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2275#define NETDEV_DOWN 0x0002
2276#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2277 detected a hardware crash and restarted
2278 - we can use this eg to kick tcp sessions
2279 once done */
2280#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2281#define NETDEV_REGISTER 0x0005
2282#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002283#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002284#define NETDEV_CHANGEADDR 0x0008
2285#define NETDEV_GOING_DOWN 0x0009
2286#define NETDEV_CHANGENAME 0x000A
2287#define NETDEV_FEAT_CHANGE 0x000B
2288#define NETDEV_BONDING_FAILOVER 0x000C
2289#define NETDEV_PRE_UP 0x000D
2290#define NETDEV_PRE_TYPE_CHANGE 0x000E
2291#define NETDEV_POST_TYPE_CHANGE 0x000F
2292#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002293#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002294#define NETDEV_RELEASE 0x0012
2295#define NETDEV_NOTIFY_PEERS 0x0013
2296#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002297#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002298#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002299#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002300#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002301#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002302#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002303#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002304#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Jason Wang08294a22016-06-30 14:45:35 +08002305#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002306
Joe Perchesf629d202013-09-26 14:48:15 -07002307int register_netdevice_notifier(struct notifier_block *nb);
2308int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002309
2310struct netdev_notifier_info {
2311 struct net_device *dev;
2312};
2313
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002314struct netdev_notifier_change_info {
2315 struct netdev_notifier_info info; /* must be first */
2316 unsigned int flags_changed;
2317};
2318
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002319struct netdev_notifier_changeupper_info {
2320 struct netdev_notifier_info info; /* must be first */
2321 struct net_device *upper_dev; /* new upper dev */
2322 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002323 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002324 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002325};
2326
Jiri Pirko04d48262015-12-03 12:12:15 +01002327struct netdev_notifier_changelowerstate_info {
2328 struct netdev_notifier_info info; /* must be first */
2329 void *lower_state_info; /* is lower dev state */
2330};
2331
Cong Wang75538c22013-05-29 11:30:50 +08002332static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2333 struct net_device *dev)
2334{
2335 info->dev = dev;
2336}
2337
Jiri Pirko351638e2013-05-28 01:30:21 +00002338static inline struct net_device *
2339netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2340{
2341 return info->dev;
2342}
2343
Joe Perchesf629d202013-09-26 14:48:15 -07002344int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002345
2346
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347extern rwlock_t dev_base_lock; /* Device list lock */
2348
Eric W. Biederman881d9662007-09-17 11:56:21 -07002349#define for_each_netdev(net, d) \
2350 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002351#define for_each_netdev_reverse(net, d) \
2352 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002353#define for_each_netdev_rcu(net, d) \
2354 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002355#define for_each_netdev_safe(net, d, n) \
2356 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2357#define for_each_netdev_continue(net, d) \
2358 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002359#define for_each_netdev_continue_rcu(net, d) \
2360 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002361#define for_each_netdev_in_bond_rcu(bond, slave) \
2362 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002363 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002364#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2365
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002366static inline struct net_device *next_net_device(struct net_device *dev)
2367{
2368 struct list_head *lh;
2369 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002370
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002371 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002372 lh = dev->dev_list.next;
2373 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2374}
2375
Eric Dumazetce81b762009-11-11 17:34:30 +00002376static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2377{
2378 struct list_head *lh;
2379 struct net *net;
2380
2381 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002382 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002383 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2384}
2385
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002386static inline struct net_device *first_net_device(struct net *net)
2387{
2388 return list_empty(&net->dev_base_head) ? NULL :
2389 net_device_entry(net->dev_base_head.next);
2390}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002391
Eric Dumazetccf43432011-01-26 18:08:02 +00002392static inline struct net_device *first_net_device_rcu(struct net *net)
2393{
2394 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2395
2396 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2397}
2398
Joe Perchesf629d202013-09-26 14:48:15 -07002399int netdev_boot_setup_check(struct net_device *dev);
2400unsigned long netdev_boot_base(const char *prefix, int unit);
2401struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2402 const char *hwaddr);
2403struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2404struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2405void dev_add_pack(struct packet_type *pt);
2406void dev_remove_pack(struct packet_type *pt);
2407void __dev_remove_pack(struct packet_type *pt);
2408void dev_add_offload(struct packet_offload *po);
2409void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002410
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002411int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002412int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002413struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2414 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002415struct net_device *dev_get_by_name(struct net *net, const char *name);
2416struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2417struct net_device *__dev_get_by_name(struct net *net, const char *name);
2418int dev_alloc_name(struct net_device *dev, const char *name);
2419int dev_open(struct net_device *dev);
2420int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002421int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002422void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002423int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002424int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002425int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002426int register_netdevice(struct net_device *dev);
2427void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2428void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002429static inline void unregister_netdevice(struct net_device *dev)
2430{
2431 unregister_netdevice_queue(dev, NULL);
2432}
2433
Joe Perchesf629d202013-09-26 14:48:15 -07002434int netdev_refcnt_read(const struct net_device *dev);
2435void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002436void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002437void synchronize_net(void);
2438int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002439
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002440DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002441#define XMIT_RECURSION_LIMIT 10
2442
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002443static inline int dev_recursion_level(void)
2444{
2445 return this_cpu_read(xmit_recursion);
2446}
2447
Joe Perchesf629d202013-09-26 14:48:15 -07002448struct net_device *dev_get_by_index(struct net *net, int ifindex);
2449struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2450struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2451int netdev_get_name(struct net *net, char *name, int ifindex);
2452int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002453int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002454
2455static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2456{
2457 return NAPI_GRO_CB(skb)->data_offset;
2458}
2459
2460static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2461{
2462 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2463}
2464
2465static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2466{
2467 NAPI_GRO_CB(skb)->data_offset += len;
2468}
2469
Herbert Xua5b1cf22009-05-26 18:50:28 +00002470static inline void *skb_gro_header_fast(struct sk_buff *skb,
2471 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002472{
Herbert Xu78a478d2009-05-26 18:50:21 +00002473 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002474}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002475
Herbert Xua5b1cf22009-05-26 18:50:28 +00002476static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2477{
2478 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2479}
2480
Herbert Xu57ea52a2017-01-10 12:24:15 -08002481static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2482{
2483 NAPI_GRO_CB(skb)->frag0 = NULL;
2484 NAPI_GRO_CB(skb)->frag0_len = 0;
2485}
2486
Herbert Xua5b1cf22009-05-26 18:50:28 +00002487static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2488 unsigned int offset)
2489{
Herbert Xu17dd7592011-07-27 06:16:28 -07002490 if (!pskb_may_pull(skb, hlen))
2491 return NULL;
2492
Herbert Xu57ea52a2017-01-10 12:24:15 -08002493 skb_gro_frag0_invalidate(skb);
Herbert Xu17dd7592011-07-27 06:16:28 -07002494 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002495}
2496
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002497static inline void *skb_gro_network_header(struct sk_buff *skb)
2498{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002499 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2500 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002501}
2502
Jerry Chubf5a7552014-01-07 10:23:19 -08002503static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2504 const void *start, unsigned int len)
2505{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002506 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002507 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2508 csum_partial(start, len, 0));
2509}
2510
Tom Herbert573e8fc2014-08-22 13:33:47 -07002511/* GRO checksum functions. These are logical equivalents of the normal
2512 * checksum functions (in skbuff.h) except that they operate on the GRO
2513 * offsets and fields in sk_buff.
2514 */
2515
2516__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2517
Tom Herbert15e23962015-02-10 16:30:31 -08002518static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2519{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002520 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002521}
2522
Tom Herbert573e8fc2014-08-22 13:33:47 -07002523static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2524 bool zero_okay,
2525 __sum16 check)
2526{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002527 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2528 skb_checksum_start_offset(skb) <
2529 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002530 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002531 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002532 (!zero_okay || check));
2533}
2534
2535static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2536 __wsum psum)
2537{
2538 if (NAPI_GRO_CB(skb)->csum_valid &&
2539 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2540 return 0;
2541
2542 NAPI_GRO_CB(skb)->csum = psum;
2543
2544 return __skb_gro_checksum_complete(skb);
2545}
2546
Tom Herbert573e8fc2014-08-22 13:33:47 -07002547static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2548{
Tom Herbert662880f2014-08-27 21:26:56 -07002549 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2550 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2551 NAPI_GRO_CB(skb)->csum_cnt--;
2552 } else {
2553 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2554 * verified a new top level checksum or an encapsulated one
2555 * during GRO. This saves work if we fallback to normal path.
2556 */
2557 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002558 }
2559}
2560
2561#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2562 compute_pseudo) \
2563({ \
2564 __sum16 __ret = 0; \
2565 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2566 __ret = __skb_gro_checksum_validate_complete(skb, \
2567 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002568 if (__ret) \
2569 __skb_mark_checksum_bad(skb); \
2570 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002571 skb_gro_incr_csum_unnecessary(skb); \
2572 __ret; \
2573})
2574
2575#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2576 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2577
2578#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2579 compute_pseudo) \
2580 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2581
2582#define skb_gro_checksum_simple_validate(skb) \
2583 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2584
Tom Herbertd96535a2014-08-31 15:12:42 -07002585static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2586{
2587 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2588 !NAPI_GRO_CB(skb)->csum_valid);
2589}
2590
2591static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2592 __sum16 check, __wsum pseudo)
2593{
2594 NAPI_GRO_CB(skb)->csum = ~pseudo;
2595 NAPI_GRO_CB(skb)->csum_valid = 1;
2596}
2597
2598#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2599do { \
2600 if (__skb_gro_checksum_convert_check(skb)) \
2601 __skb_gro_checksum_convert(skb, check, \
2602 compute_pseudo(skb, proto)); \
2603} while (0)
2604
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002605struct gro_remcsum {
2606 int offset;
2607 __wsum delta;
2608};
2609
2610static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2611{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002612 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002613 grc->delta = 0;
2614}
2615
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002616static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2617 unsigned int off, size_t hdrlen,
2618 int start, int offset,
2619 struct gro_remcsum *grc,
2620 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002621{
2622 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002623 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002624
2625 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2626
Tom Herbert15e23962015-02-10 16:30:31 -08002627 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002628 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2629 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002630 }
2631
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002632 ptr = skb_gro_header_fast(skb, off);
2633 if (skb_gro_header_hard(skb, off + plen)) {
2634 ptr = skb_gro_header_slow(skb, off + plen, off);
2635 if (!ptr)
2636 return NULL;
2637 }
2638
2639 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2640 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002641
2642 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002643 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002644
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002645 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002646 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002647
2648 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002649}
2650
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002651static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2652 struct gro_remcsum *grc)
2653{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002654 void *ptr;
2655 size_t plen = grc->offset + sizeof(u16);
2656
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002657 if (!grc->delta)
2658 return;
2659
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002660 ptr = skb_gro_header_fast(skb, grc->offset);
2661 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2662 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2663 if (!ptr)
2664 return;
2665 }
2666
2667 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002668}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002669
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002670static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2671 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002672 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002673 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002674{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002675 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002676 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002677
2678 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002679}
2680
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002681static inline int dev_parse_header(const struct sk_buff *skb,
2682 unsigned char *haddr)
2683{
2684 const struct net_device *dev = skb->dev;
2685
Patrick McHardy1b833362007-10-18 05:09:28 -07002686 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002687 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002688 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002689}
2690
Willem de Bruijn2793a232016-03-09 21:58:32 -05002691/* ll_header must have at least hard_header_len allocated */
2692static inline bool dev_validate_header(const struct net_device *dev,
2693 char *ll_header, int len)
2694{
2695 if (likely(len >= dev->hard_header_len))
2696 return true;
2697
2698 if (capable(CAP_SYS_RAWIO)) {
2699 memset(ll_header + len, 0, dev->hard_header_len - len);
2700 return true;
2701 }
2702
2703 if (dev->header_ops && dev->header_ops->validate)
2704 return dev->header_ops->validate(ll_header, len);
2705
2706 return false;
2707}
2708
Linus Torvalds1da177e2005-04-16 15:20:36 -07002709typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002710int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002711static inline int unregister_gifconf(unsigned int family)
2712{
2713 return register_gifconf(family, NULL);
2714}
2715
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002716#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002717#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002718struct sd_flow_limit {
2719 u64 count;
2720 unsigned int num_buckets;
2721 unsigned int history_head;
2722 u16 history[FLOW_LIMIT_HISTORY];
2723 u8 buckets[];
2724};
2725
2726extern int netdev_flow_limit_table_len;
2727#endif /* CONFIG_NET_FLOW_LIMIT */
2728
Linus Torvalds1da177e2005-04-16 15:20:36 -07002729/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002730 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002731 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002732struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002733 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002734 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002735
Changli Gaodee42872010-05-02 05:42:16 +00002736 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002737 unsigned int processed;
2738 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002739 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002740#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002741 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002742#endif
2743#ifdef CONFIG_NET_FLOW_LIMIT
2744 struct sd_flow_limit __rcu *flow_limit;
2745#endif
2746 struct Qdisc *output_queue;
2747 struct Qdisc **output_queue_tailp;
2748 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002749
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002750#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002751 /* input_queue_head should be written by cpu owning this struct,
2752 * and only read by other cpus. Worth using a cache line.
2753 */
2754 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2755
2756 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002757 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002758 struct softnet_data *rps_ipi_next;
2759 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002760 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002761#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002762 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002763 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002764 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002765
Linus Torvalds1da177e2005-04-16 15:20:36 -07002766};
2767
Tom Herbert76cc8b12010-05-20 18:37:59 +00002768static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002769{
2770#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002771 sd->input_queue_head++;
2772#endif
2773}
2774
2775static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2776 unsigned int *qtail)
2777{
2778#ifdef CONFIG_RPS
2779 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002780#endif
2781}
2782
Tom Herbert0a9627f2010-03-16 08:03:29 +00002783DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002784
Joe Perchesf629d202013-09-26 14:48:15 -07002785void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002786void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002787
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002788static inline void netif_tx_schedule_all(struct net_device *dev)
2789{
2790 unsigned int i;
2791
2792 for (i = 0; i < dev->num_tx_queues; i++)
2793 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2794}
2795
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002796static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002797{
Tom Herbert734664982011-11-28 16:32:44 +00002798 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002799}
2800
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002801/**
2802 * netif_start_queue - allow transmit
2803 * @dev: network device
2804 *
2805 * Allow upper layers to call the device hard_start_xmit routine.
2806 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002807static inline void netif_start_queue(struct net_device *dev)
2808{
David S. Millere8a04642008-07-17 00:34:19 -07002809 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002810}
2811
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002812static inline void netif_tx_start_all_queues(struct net_device *dev)
2813{
2814 unsigned int i;
2815
2816 for (i = 0; i < dev->num_tx_queues; i++) {
2817 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2818 netif_tx_start_queue(txq);
2819 }
2820}
2821
John Fastabend46e5da40a2014-09-12 20:04:52 -07002822void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002823
Dave Jonesd29f7492008-07-22 14:09:06 -07002824/**
2825 * netif_wake_queue - restart transmit
2826 * @dev: network device
2827 *
2828 * Allow upper layers to call the device hard_start_xmit routine.
2829 * Used for flow control when transmit resources are available.
2830 */
David S. Miller79d16382008-07-08 23:14:46 -07002831static inline void netif_wake_queue(struct net_device *dev)
2832{
David S. Millere8a04642008-07-17 00:34:19 -07002833 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002834}
2835
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002836static inline void netif_tx_wake_all_queues(struct net_device *dev)
2837{
2838 unsigned int i;
2839
2840 for (i = 0; i < dev->num_tx_queues; i++) {
2841 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2842 netif_tx_wake_queue(txq);
2843 }
2844}
2845
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002846static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002847{
Tom Herbert734664982011-11-28 16:32:44 +00002848 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002849}
2850
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002851/**
2852 * netif_stop_queue - stop transmitted packets
2853 * @dev: network device
2854 *
2855 * Stop upper layers calling the device hard_start_xmit routine.
2856 * Used for flow control when transmit resources are unavailable.
2857 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002858static inline void netif_stop_queue(struct net_device *dev)
2859{
David S. Millere8a04642008-07-17 00:34:19 -07002860 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002861}
2862
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002863void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002864
David S. Miller4d295152012-03-07 21:02:35 -05002865static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002866{
Tom Herbert734664982011-11-28 16:32:44 +00002867 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002868}
2869
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002870/**
2871 * netif_queue_stopped - test if transmit queue is flowblocked
2872 * @dev: network device
2873 *
2874 * Test if transmit queue on device is currently unable to send.
2875 */
David S. Miller4d295152012-03-07 21:02:35 -05002876static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002877{
David S. Millere8a04642008-07-17 00:34:19 -07002878 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002879}
2880
David S. Miller4d295152012-03-07 21:02:35 -05002881static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002882{
Tom Herbert734664982011-11-28 16:32:44 +00002883 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2884}
2885
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002886static inline bool
2887netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002888{
2889 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2890}
2891
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002892static inline bool
2893netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2894{
2895 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2896}
2897
Eric Dumazet53511452014-10-08 08:19:27 -07002898/**
2899 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2900 * @dev_queue: pointer to transmit queue
2901 *
2902 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002903 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002904 */
2905static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2906{
2907#ifdef CONFIG_BQL
2908 prefetchw(&dev_queue->dql.num_queued);
2909#endif
2910}
2911
2912/**
2913 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2914 * @dev_queue: pointer to transmit queue
2915 *
2916 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002917 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002918 */
2919static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2920{
2921#ifdef CONFIG_BQL
2922 prefetchw(&dev_queue->dql.limit);
2923#endif
2924}
2925
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002926static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2927 unsigned int bytes)
2928{
Tom Herbert114cf582011-11-28 16:33:09 +00002929#ifdef CONFIG_BQL
2930 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002931
2932 if (likely(dql_avail(&dev_queue->dql) >= 0))
2933 return;
2934
2935 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2936
2937 /*
2938 * The XOFF flag must be set before checking the dql_avail below,
2939 * because in netdev_tx_completed_queue we update the dql_completed
2940 * before checking the XOFF flag.
2941 */
2942 smp_mb();
2943
2944 /* check again in case another CPU has just made room avail */
2945 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2946 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002947#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002948}
2949
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002950/**
2951 * netdev_sent_queue - report the number of bytes queued to hardware
2952 * @dev: network device
2953 * @bytes: number of bytes queued to the hardware device queue
2954 *
2955 * Report the number of bytes queued for sending/completion to the network
2956 * device hardware queue. @bytes should be a good approximation and should
2957 * exactly match netdev_completed_queue() @bytes
2958 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002959static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2960{
2961 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2962}
2963
2964static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002965 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002966{
Tom Herbert114cf582011-11-28 16:33:09 +00002967#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002968 if (unlikely(!bytes))
2969 return;
2970
2971 dql_completed(&dev_queue->dql, bytes);
2972
2973 /*
2974 * Without the memory barrier there is a small possiblity that
2975 * netdev_tx_sent_queue will miss the update and cause the queue to
2976 * be stopped forever
2977 */
2978 smp_mb();
2979
2980 if (dql_avail(&dev_queue->dql) < 0)
2981 return;
2982
2983 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2984 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00002985#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002986}
2987
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002988/**
2989 * netdev_completed_queue - report bytes and packets completed by device
2990 * @dev: network device
2991 * @pkts: actual number of packets sent over the medium
2992 * @bytes: actual number of bytes sent over the medium
2993 *
2994 * Report the number of bytes and packets transmitted by the network device
2995 * hardware queue over the physical medium, @bytes must exactly match the
2996 * @bytes amount passed to netdev_sent_queue()
2997 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002998static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00002999 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003000{
3001 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3002}
3003
3004static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3005{
Tom Herbert114cf582011-11-28 16:33:09 +00003006#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003007 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003008 dql_reset(&q->dql);
3009#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003010}
3011
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003012/**
3013 * netdev_reset_queue - reset the packets and bytes count of a network device
3014 * @dev_queue: network device
3015 *
3016 * Reset the bytes and packet count of a network device and clear the
3017 * software flow control OFF bit for this network device
3018 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003019static inline void netdev_reset_queue(struct net_device *dev_queue)
3020{
3021 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003022}
3023
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003024/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003025 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3026 * @dev: network device
3027 * @queue_index: given tx queue index
3028 *
3029 * Returns 0 if given tx queue index >= number of device tx queues,
3030 * otherwise returns the originally passed tx queue index.
3031 */
3032static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3033{
3034 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3035 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3036 dev->name, queue_index,
3037 dev->real_num_tx_queues);
3038 return 0;
3039 }
3040
3041 return queue_index;
3042}
3043
3044/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003045 * netif_running - test if up
3046 * @dev: network device
3047 *
3048 * Test if the device has been brought up.
3049 */
David S. Miller4d295152012-03-07 21:02:35 -05003050static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003051{
3052 return test_bit(__LINK_STATE_START, &dev->state);
3053}
3054
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003055/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003056 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003057 * stop, and a check if it's stopped. All other device management is
3058 * done at the overall netdevice level.
3059 * Also test the device if we're multiqueue.
3060 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003061
3062/**
3063 * netif_start_subqueue - allow sending packets on subqueue
3064 * @dev: network device
3065 * @queue_index: sub queue index
3066 *
3067 * Start individual transmit queue of a device with multiple transmit queues.
3068 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003069static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3070{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003071 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003072
3073 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003074}
3075
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003076/**
3077 * netif_stop_subqueue - stop sending packets on subqueue
3078 * @dev: network device
3079 * @queue_index: sub queue index
3080 *
3081 * Stop individual transmit queue of a device with multiple transmit queues.
3082 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003083static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3084{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003085 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003086 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003087}
3088
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003089/**
3090 * netif_subqueue_stopped - test status of subqueue
3091 * @dev: network device
3092 * @queue_index: sub queue index
3093 *
3094 * Check individual transmit queue of a device with multiple transmit queues.
3095 */
David S. Miller4d295152012-03-07 21:02:35 -05003096static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3097 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003098{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003099 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003100
3101 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003102}
3103
David S. Miller4d295152012-03-07 21:02:35 -05003104static inline bool netif_subqueue_stopped(const struct net_device *dev,
3105 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003106{
3107 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3108}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003109
John Fastabend46e5da40a2014-09-12 20:04:52 -07003110void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003111
Alexander Duyck537c00d2013-01-10 08:57:02 +00003112#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003113int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003114 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003115#else
3116static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003117 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003118 u16 index)
3119{
3120 return 0;
3121}
3122#endif
3123
Jiri Pirko5605c762015-05-12 14:56:12 +02003124u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3125 unsigned int num_tx_queues);
3126
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003127/*
3128 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3129 * as a distribution range limit for the returned value.
3130 */
3131static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003132 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003133{
3134 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3135}
3136
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003137/**
3138 * netif_is_multiqueue - test if device has multiple transmit queues
3139 * @dev: network device
3140 *
3141 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003142 */
David S. Miller4d295152012-03-07 21:02:35 -05003143static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003144{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003145 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003146}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003147
Joe Perchesf629d202013-09-26 14:48:15 -07003148int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003149
Michael Daltona953be52014-01-16 22:23:28 -08003150#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003151int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003152#else
3153static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3154 unsigned int rxq)
3155{
3156 return 0;
3157}
3158#endif
3159
Michael Daltona953be52014-01-16 22:23:28 -08003160#ifdef CONFIG_SYSFS
3161static inline unsigned int get_netdev_rx_queue_index(
3162 struct netdev_rx_queue *queue)
3163{
3164 struct net_device *dev = queue->dev;
3165 int index = queue - dev->_rx;
3166
3167 BUG_ON(index >= dev->num_rx_queues);
3168 return index;
3169}
3170#endif
3171
Yuval Mintz16917b82012-07-01 03:18:50 +00003172#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003173int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003174
Eric Dumazete6247022013-12-05 04:45:08 -08003175enum skb_free_reason {
3176 SKB_REASON_CONSUMED,
3177 SKB_REASON_DROPPED,
3178};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003179
Eric Dumazete6247022013-12-05 04:45:08 -08003180void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3181void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3182
3183/*
3184 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3185 * interrupt context or with hardware interrupts being disabled.
3186 * (in_irq() || irqs_disabled())
3187 *
3188 * We provide four helpers that can be used in following contexts :
3189 *
3190 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3191 * replacing kfree_skb(skb)
3192 *
3193 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3194 * Typically used in place of consume_skb(skb) in TX completion path
3195 *
3196 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3197 * replacing kfree_skb(skb)
3198 *
3199 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3200 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003201 */
Eric Dumazete6247022013-12-05 04:45:08 -08003202static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3203{
3204 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3205}
3206
3207static inline void dev_consume_skb_irq(struct sk_buff *skb)
3208{
3209 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3210}
3211
3212static inline void dev_kfree_skb_any(struct sk_buff *skb)
3213{
3214 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3215}
3216
3217static inline void dev_consume_skb_any(struct sk_buff *skb)
3218{
3219 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3220}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003221
Joe Perchesf629d202013-09-26 14:48:15 -07003222int netif_rx(struct sk_buff *skb);
3223int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003224int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003225gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3226void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3227struct sk_buff *napi_get_frags(struct napi_struct *napi);
3228gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003229struct packet_offload *gro_find_receive_by_type(__be16 type);
3230struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003231
3232static inline void napi_free_frags(struct napi_struct *napi)
3233{
3234 kfree_skb(napi->skb);
3235 napi->skb = NULL;
3236}
3237
Mahesh Bandewar24b27fc2016-09-01 22:18:34 -07003238bool netdev_is_rx_handler_busy(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003239int netdev_rx_handler_register(struct net_device *dev,
3240 rx_handler_func_t *rx_handler,
3241 void *rx_handler_data);
3242void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003243
Joe Perchesf629d202013-09-26 14:48:15 -07003244bool dev_valid_name(const char *name);
3245int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3246int dev_ethtool(struct net *net, struct ifreq *);
3247unsigned int dev_get_flags(const struct net_device *);
3248int __dev_change_flags(struct net_device *, unsigned int flags);
3249int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003250void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3251 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003252int dev_change_name(struct net_device *, const char *);
3253int dev_set_alias(struct net_device *, const char *, size_t);
3254int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3255int dev_set_mtu(struct net_device *, int);
3256void dev_set_group(struct net_device *, int);
3257int dev_set_mac_address(struct net_device *, struct sockaddr *);
3258int dev_change_carrier(struct net_device *, bool new_carrier);
3259int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003260 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003261int dev_get_phys_port_name(struct net_device *dev,
3262 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003263int dev_change_proto_down(struct net_device *dev, bool proto_down);
Daniel Borkmann85de8572016-11-28 23:16:54 +01003264int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003265struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003266struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3267 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08003268int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003269int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003270bool is_skb_forwardable(const struct net_device *dev,
3271 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003272
Martin KaFai Lau4e3264d2016-11-09 15:36:33 -08003273static __always_inline int ____dev_forward_skb(struct net_device *dev,
3274 struct sk_buff *skb)
3275{
3276 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3277 unlikely(!is_skb_forwardable(dev, skb))) {
3278 atomic_long_inc(&dev->rx_dropped);
3279 kfree_skb(skb);
3280 return NET_RX_DROP;
3281 }
3282
3283 skb_scrub_packet(skb, true);
3284 skb->priority = 0;
3285 return 0;
3286}
3287
David Ahern74b20582016-05-10 11:19:50 -07003288void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3289
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003290extern int netdev_budget;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003291
3292/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003293void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003294
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003295/**
3296 * dev_put - release reference to device
3297 * @dev: network device
3298 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003299 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003300 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003301static inline void dev_put(struct net_device *dev)
3302{
Christoph Lameter933393f2011-12-22 11:58:51 -06003303 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003304}
3305
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003306/**
3307 * dev_hold - get reference to device
3308 * @dev: network device
3309 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003310 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003311 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003312static inline void dev_hold(struct net_device *dev)
3313{
Christoph Lameter933393f2011-12-22 11:58:51 -06003314 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003315}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003316
3317/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3318 * and _off may be called from IRQ context, but it is caller
3319 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003320 *
3321 * The name carrier is inappropriate, these functions should really be
3322 * called netif_lowerlayer_*() because they represent the state of any
3323 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003324 */
3325
Joe Perchesf629d202013-09-26 14:48:15 -07003326void linkwatch_init_dev(struct net_device *dev);
3327void linkwatch_fire_event(struct net_device *dev);
3328void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003329
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003330/**
3331 * netif_carrier_ok - test if carrier present
3332 * @dev: network device
3333 *
3334 * Check if carrier is present on device
3335 */
David S. Miller4d295152012-03-07 21:02:35 -05003336static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003337{
3338 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3339}
3340
Joe Perchesf629d202013-09-26 14:48:15 -07003341unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003342
Joe Perchesf629d202013-09-26 14:48:15 -07003343void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003344
Joe Perchesf629d202013-09-26 14:48:15 -07003345void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003346
Joe Perchesf629d202013-09-26 14:48:15 -07003347void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003348
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003349/**
3350 * netif_dormant_on - mark device as dormant.
3351 * @dev: network device
3352 *
3353 * Mark device as dormant (as per RFC2863).
3354 *
3355 * The dormant state indicates that the relevant interface is not
3356 * actually in a condition to pass packets (i.e., it is not 'up') but is
3357 * in a "pending" state, waiting for some external event. For "on-
3358 * demand" interfaces, this new state identifies the situation where the
3359 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003360 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003361static inline void netif_dormant_on(struct net_device *dev)
3362{
3363 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3364 linkwatch_fire_event(dev);
3365}
3366
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003367/**
3368 * netif_dormant_off - set device as not dormant.
3369 * @dev: network device
3370 *
3371 * Device is not in dormant state.
3372 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003373static inline void netif_dormant_off(struct net_device *dev)
3374{
3375 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3376 linkwatch_fire_event(dev);
3377}
3378
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003379/**
3380 * netif_dormant - test if carrier present
3381 * @dev: network device
3382 *
3383 * Check if carrier is present on device
3384 */
David S. Miller4d295152012-03-07 21:02:35 -05003385static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003386{
3387 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3388}
3389
3390
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003391/**
3392 * netif_oper_up - test if device is operational
3393 * @dev: network device
3394 *
3395 * Check if carrier is operational
3396 */
David S. Miller4d295152012-03-07 21:02:35 -05003397static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003398{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003399 return (dev->operstate == IF_OPER_UP ||
3400 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3401}
3402
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003403/**
3404 * netif_device_present - is device available or removed
3405 * @dev: network device
3406 *
3407 * Check if device has not been removed from system.
3408 */
David S. Miller4d295152012-03-07 21:02:35 -05003409static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003410{
3411 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3412}
3413
Joe Perchesf629d202013-09-26 14:48:15 -07003414void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003415
Joe Perchesf629d202013-09-26 14:48:15 -07003416void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003417
3418/*
3419 * Network interface message level settings
3420 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003421
3422enum {
3423 NETIF_MSG_DRV = 0x0001,
3424 NETIF_MSG_PROBE = 0x0002,
3425 NETIF_MSG_LINK = 0x0004,
3426 NETIF_MSG_TIMER = 0x0008,
3427 NETIF_MSG_IFDOWN = 0x0010,
3428 NETIF_MSG_IFUP = 0x0020,
3429 NETIF_MSG_RX_ERR = 0x0040,
3430 NETIF_MSG_TX_ERR = 0x0080,
3431 NETIF_MSG_TX_QUEUED = 0x0100,
3432 NETIF_MSG_INTR = 0x0200,
3433 NETIF_MSG_TX_DONE = 0x0400,
3434 NETIF_MSG_RX_STATUS = 0x0800,
3435 NETIF_MSG_PKTDATA = 0x1000,
3436 NETIF_MSG_HW = 0x2000,
3437 NETIF_MSG_WOL = 0x4000,
3438};
3439
3440#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3441#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3442#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3443#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3444#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3445#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3446#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3447#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3448#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3449#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3450#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3451#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3452#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3453#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3454#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3455
3456static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3457{
3458 /* use default */
3459 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3460 return default_msg_enable_bits;
3461 if (debug_value == 0) /* no output */
3462 return 0;
3463 /* set low N bits */
3464 return (1 << debug_value) - 1;
3465}
3466
David S. Millerc773e842008-07-08 23:13:53 -07003467static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003468{
David S. Millerc773e842008-07-08 23:13:53 -07003469 spin_lock(&txq->_xmit_lock);
3470 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003471}
3472
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003473static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3474{
3475 __acquire(&txq->_xmit_lock);
3476 return true;
3477}
3478
3479static inline void __netif_tx_release(struct netdev_queue *txq)
3480{
3481 __release(&txq->_xmit_lock);
3482}
3483
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003484static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3485{
3486 spin_lock_bh(&txq->_xmit_lock);
3487 txq->xmit_lock_owner = smp_processor_id();
3488}
3489
David S. Miller4d295152012-03-07 21:02:35 -05003490static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003491{
David S. Miller4d295152012-03-07 21:02:35 -05003492 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003493 if (likely(ok))
3494 txq->xmit_lock_owner = smp_processor_id();
3495 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003496}
3497
David S. Millerc773e842008-07-08 23:13:53 -07003498static inline void __netif_tx_unlock(struct netdev_queue *txq)
3499{
3500 txq->xmit_lock_owner = -1;
3501 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003502}
3503
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003504static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3505{
3506 txq->xmit_lock_owner = -1;
3507 spin_unlock_bh(&txq->_xmit_lock);
3508}
3509
Eric Dumazet08baf562009-05-25 22:58:01 -07003510static inline void txq_trans_update(struct netdev_queue *txq)
3511{
3512 if (txq->xmit_lock_owner != -1)
3513 txq->trans_start = jiffies;
3514}
3515
Florian Westphalba162f82016-05-03 16:31:00 +02003516/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3517static inline void netif_trans_update(struct net_device *dev)
3518{
Florian Westphal9b366272016-05-03 16:33:14 +02003519 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3520
3521 if (txq->trans_start != jiffies)
3522 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003523}
3524
David S. Millerc3f26a22008-07-31 16:58:50 -07003525/**
3526 * netif_tx_lock - grab network device transmit lock
3527 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003528 *
3529 * Get network device transmit lock
3530 */
3531static inline void netif_tx_lock(struct net_device *dev)
3532{
3533 unsigned int i;
3534 int cpu;
3535
3536 spin_lock(&dev->tx_global_lock);
3537 cpu = smp_processor_id();
3538 for (i = 0; i < dev->num_tx_queues; i++) {
3539 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3540
3541 /* We are the only thread of execution doing a
3542 * freeze, but we have to grab the _xmit_lock in
3543 * order to synchronize with threads which are in
3544 * the ->hard_start_xmit() handler and already
3545 * checked the frozen bit.
3546 */
3547 __netif_tx_lock(txq, cpu);
3548 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3549 __netif_tx_unlock(txq);
3550 }
3551}
3552
3553static inline void netif_tx_lock_bh(struct net_device *dev)
3554{
3555 local_bh_disable();
3556 netif_tx_lock(dev);
3557}
3558
Herbert Xu932ff272006-06-09 12:20:56 -07003559static inline void netif_tx_unlock(struct net_device *dev)
3560{
David S. Millere8a04642008-07-17 00:34:19 -07003561 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003562
David S. Millere8a04642008-07-17 00:34:19 -07003563 for (i = 0; i < dev->num_tx_queues; i++) {
3564 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003565
David S. Millerc3f26a22008-07-31 16:58:50 -07003566 /* No need to grab the _xmit_lock here. If the
3567 * queue is not stopped for another reason, we
3568 * force a schedule.
3569 */
3570 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003571 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003572 }
3573 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003574}
3575
3576static inline void netif_tx_unlock_bh(struct net_device *dev)
3577{
David S. Millere8a04642008-07-17 00:34:19 -07003578 netif_tx_unlock(dev);
3579 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003580}
3581
David S. Millerc773e842008-07-08 23:13:53 -07003582#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003583 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003584 __netif_tx_lock(txq, cpu); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003585 } else { \
3586 __netif_tx_acquire(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003587 } \
3588}
3589
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003590#define HARD_TX_TRYLOCK(dev, txq) \
3591 (((dev->features & NETIF_F_LLTX) == 0) ? \
3592 __netif_tx_trylock(txq) : \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003593 __netif_tx_acquire(txq))
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003594
David S. Millerc773e842008-07-08 23:13:53 -07003595#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003596 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003597 __netif_tx_unlock(txq); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003598 } else { \
3599 __netif_tx_release(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003600 } \
3601}
3602
Linus Torvalds1da177e2005-04-16 15:20:36 -07003603static inline void netif_tx_disable(struct net_device *dev)
3604{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003605 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003606 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003607
David S. Millerc3f26a22008-07-31 16:58:50 -07003608 local_bh_disable();
3609 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003610 for (i = 0; i < dev->num_tx_queues; i++) {
3611 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003612
3613 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003614 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003615 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003616 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003617 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003618}
3619
David S. Millere308a5d2008-07-15 00:13:44 -07003620static inline void netif_addr_lock(struct net_device *dev)
3621{
3622 spin_lock(&dev->addr_list_lock);
3623}
3624
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003625static inline void netif_addr_lock_nested(struct net_device *dev)
3626{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003627 int subclass = SINGLE_DEPTH_NESTING;
3628
3629 if (dev->netdev_ops->ndo_get_lock_subclass)
3630 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3631
3632 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003633}
3634
David S. Millere308a5d2008-07-15 00:13:44 -07003635static inline void netif_addr_lock_bh(struct net_device *dev)
3636{
3637 spin_lock_bh(&dev->addr_list_lock);
3638}
3639
3640static inline void netif_addr_unlock(struct net_device *dev)
3641{
3642 spin_unlock(&dev->addr_list_lock);
3643}
3644
3645static inline void netif_addr_unlock_bh(struct net_device *dev)
3646{
3647 spin_unlock_bh(&dev->addr_list_lock);
3648}
3649
Jiri Pirkof001fde2009-05-05 02:48:28 +00003650/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003651 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003652 * rcu_read_lock held.
3653 */
3654#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003655 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003656
Linus Torvalds1da177e2005-04-16 15:20:36 -07003657/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3658
Joe Perchesf629d202013-09-26 14:48:15 -07003659void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003660
3661/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003662struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003663 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003664 void (*setup)(struct net_device *),
3665 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003666#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3667 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003668
Tom Gundersenc835a672014-07-14 16:37:24 +02003669#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3670 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3671 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003672
Joe Perchesf629d202013-09-26 14:48:15 -07003673int register_netdev(struct net_device *dev);
3674void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003675
Jiri Pirko22bedad32010-04-01 21:22:57 +00003676/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003677int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3678 struct netdev_hw_addr_list *from_list, int addr_len);
3679void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3680 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003681int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3682 struct net_device *dev,
3683 int (*sync)(struct net_device *, const unsigned char *),
3684 int (*unsync)(struct net_device *,
3685 const unsigned char *));
3686void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3687 struct net_device *dev,
3688 int (*unsync)(struct net_device *,
3689 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003690void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003691
Jiri Pirkof001fde2009-05-05 02:48:28 +00003692/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003693int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3694 unsigned char addr_type);
3695int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3696 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003697void dev_addr_flush(struct net_device *dev);
3698int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003699
3700/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003701int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3702int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3703int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3704int dev_uc_sync(struct net_device *to, struct net_device *from);
3705int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3706void dev_uc_unsync(struct net_device *to, struct net_device *from);
3707void dev_uc_flush(struct net_device *dev);
3708void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003709
Alexander Duyck670e5b82014-05-28 18:44:46 -07003710/**
3711 * __dev_uc_sync - Synchonize device's unicast list
3712 * @dev: device to sync
3713 * @sync: function to call if address should be added
3714 * @unsync: function to call if address should be removed
3715 *
3716 * Add newly added addresses to the interface, and release
3717 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003718 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003719static inline int __dev_uc_sync(struct net_device *dev,
3720 int (*sync)(struct net_device *,
3721 const unsigned char *),
3722 int (*unsync)(struct net_device *,
3723 const unsigned char *))
3724{
3725 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3726}
3727
3728/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003729 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003730 * @dev: device to sync
3731 * @unsync: function to call if address should be removed
3732 *
3733 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003734 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003735static inline void __dev_uc_unsync(struct net_device *dev,
3736 int (*unsync)(struct net_device *,
3737 const unsigned char *))
3738{
3739 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3740}
3741
Jiri Pirko22bedad32010-04-01 21:22:57 +00003742/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003743int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3744int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3745int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3746int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3747int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3748int dev_mc_sync(struct net_device *to, struct net_device *from);
3749int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3750void dev_mc_unsync(struct net_device *to, struct net_device *from);
3751void dev_mc_flush(struct net_device *dev);
3752void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003753
Alexander Duyck670e5b82014-05-28 18:44:46 -07003754/**
3755 * __dev_mc_sync - Synchonize device's multicast list
3756 * @dev: device to sync
3757 * @sync: function to call if address should be added
3758 * @unsync: function to call if address should be removed
3759 *
3760 * Add newly added addresses to the interface, and release
3761 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003762 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003763static inline int __dev_mc_sync(struct net_device *dev,
3764 int (*sync)(struct net_device *,
3765 const unsigned char *),
3766 int (*unsync)(struct net_device *,
3767 const unsigned char *))
3768{
3769 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3770}
3771
3772/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003773 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003774 * @dev: device to sync
3775 * @unsync: function to call if address should be removed
3776 *
3777 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003778 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003779static inline void __dev_mc_unsync(struct net_device *dev,
3780 int (*unsync)(struct net_device *,
3781 const unsigned char *))
3782{
3783 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3784}
3785
Linus Torvalds1da177e2005-04-16 15:20:36 -07003786/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003787void dev_set_rx_mode(struct net_device *dev);
3788void __dev_set_rx_mode(struct net_device *dev);
3789int dev_set_promiscuity(struct net_device *dev, int inc);
3790int dev_set_allmulti(struct net_device *dev, int inc);
3791void netdev_state_change(struct net_device *dev);
3792void netdev_notify_peers(struct net_device *dev);
3793void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003794/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003795void dev_load(struct net *net, const char *name);
3796struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3797 struct rtnl_link_stats64 *storage);
3798void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3799 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003800
3801extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003802extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003803extern int weight_p;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003804
Joe Perchesf629d202013-09-26 14:48:15 -07003805bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003806struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3807 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003808struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3809 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003810
3811/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003812#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3813 for (iter = &(dev)->adj_list.upper, \
3814 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3815 updev; \
3816 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3817
David Ahern1a3f0602016-10-17 19:15:44 -07003818int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3819 int (*fn)(struct net_device *upper_dev,
3820 void *data),
3821 void *data);
3822
3823bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3824 struct net_device *upper_dev);
3825
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
David Ahern1a3f0602016-10-17 19:15:44 -07003857int netdev_walk_all_lower_dev(struct net_device *dev,
3858 int (*fn)(struct net_device *lower_dev,
3859 void *data),
3860 void *data);
3861int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3862 int (*fn)(struct net_device *lower_dev,
3863 void *data),
3864 void *data);
3865
Joe Perchesf629d202013-09-26 14:48:15 -07003866void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003867void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003868struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3869struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3870int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3871int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003872 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003873 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003874void netdev_upper_dev_unlink(struct net_device *dev,
3875 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003876void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003877void *netdev_lower_dev_get_private(struct net_device *dev,
3878 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003879void netdev_lower_state_changed(struct net_device *lower_dev,
3880 void *lower_state_info);
Jiri Pirko18bfb9242016-07-05 11:27:38 +02003881int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3882 struct neighbour *n);
3883void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3884 struct neighbour *n);
Eric Dumazet960fb622014-11-16 06:23:05 -08003885
3886/* RSS keys are 40 or 52 bytes long */
3887#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003888extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003889void netdev_rss_key_fill(void *buffer, size_t len);
3890
Sabrina Dubroca952fcfd2016-08-12 16:10:33 +02003891int dev_get_nest_level(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003892int skb_checksum_help(struct sk_buff *skb);
3893struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3894 netdev_features_t features, bool tx_path);
3895struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3896 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003897
Moni Shoua61bd3852015-02-03 16:48:29 +02003898struct netdev_bonding_info {
3899 ifslave slave;
3900 ifbond master;
3901};
3902
3903struct netdev_notifier_bonding_info {
3904 struct netdev_notifier_info info; /* must be first */
3905 struct netdev_bonding_info bonding_info;
3906};
3907
3908void netdev_bonding_info_change(struct net_device *dev,
3909 struct netdev_bonding_info *bonding_info);
3910
Cong Wang12b00042013-02-05 16:36:38 +00003911static inline
3912struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3913{
3914 return __skb_gso_segment(skb, features, true);
3915}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003916__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003917
3918static inline bool can_checksum_protocol(netdev_features_t features,
3919 __be16 protocol)
3920{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003921 if (protocol == htons(ETH_P_FCOE))
3922 return !!(features & NETIF_F_FCOE_CRC);
3923
3924 /* Assume this is an IP checksum (not SCTP CRC) */
3925
3926 if (features & NETIF_F_HW_CSUM) {
3927 /* Can checksum everything */
3928 return true;
3929 }
3930
3931 switch (protocol) {
3932 case htons(ETH_P_IP):
3933 return !!(features & NETIF_F_IP_CSUM);
3934 case htons(ETH_P_IPV6):
3935 return !!(features & NETIF_F_IPV6_CSUM);
3936 default:
3937 return false;
3938 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003939}
Cong Wang12b00042013-02-05 16:36:38 +00003940
Herbert Xufb286bb2005-11-10 13:01:24 -08003941#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003942void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003943#else
3944static inline void netdev_rx_csum_fault(struct net_device *dev)
3945{
3946}
3947#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003948/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003949void net_enable_timestamp(void);
3950void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003951
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003952#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003953int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003954#else
3955#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003956#endif
3957
David S. Miller47982482014-08-22 16:21:53 -07003958static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003959 struct sk_buff *skb, struct net_device *dev,
3960 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003961{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003962 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003963 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003964}
3965
David S. Miller10b3ad82014-08-29 21:07:24 -07003966static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003967 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07003968{
3969 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07003970 int rc;
David S. Miller47982482014-08-22 16:21:53 -07003971
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003972 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07003973 if (rc == NETDEV_TX_OK)
3974 txq_trans_update(txq);
3975
3976 return rc;
David S. Miller47982482014-08-22 16:21:53 -07003977}
3978
Linus Torvalds42a2d922013-11-13 17:40:34 +09003979int netdev_class_create_file_ns(struct class_attribute *class_attr,
3980 const void *ns);
3981void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3982 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04003983
3984static inline int netdev_class_create_file(struct class_attribute *class_attr)
3985{
3986 return netdev_class_create_file_ns(class_attr, NULL);
3987}
3988
3989static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3990{
3991 netdev_class_remove_file_ns(class_attr, NULL);
3992}
Jay Vosburghb8a97872008-06-13 18:12:04 -07003993
Johannes Berg04600792010-08-05 17:45:15 +02003994extern struct kobj_ns_type_operations net_ns_type_operations;
3995
Joe Perchesf629d202013-09-26 14:48:15 -07003996const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07003997
Joe Perchesf629d202013-09-26 14:48:15 -07003998void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003999
Michal Kubečekda081432014-05-20 08:29:25 +02004000static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4001 netdev_features_t f2)
4002{
Tom Herbertc8cd0982015-12-14 11:19:44 -08004003 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4004 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08004005 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004006 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08004007 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004008 }
Michal Kubečekda081432014-05-20 08:29:25 +02004009
Tom Herbertc8cd0982015-12-14 11:19:44 -08004010 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02004011}
4012
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004013static inline netdev_features_t netdev_get_wanted_features(
4014 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00004015{
4016 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4017}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004018netdev_features_t netdev_increment_features(netdev_features_t all,
4019 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00004020
4021/* Allow TSO being used on stacked device :
4022 * Performing the GSO segmentation before last device
4023 * is a performance improvement.
4024 */
4025static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4026 netdev_features_t mask)
4027{
4028 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4029}
4030
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004031int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004032void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004033void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004034
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004035void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4036 struct net_device *dev);
4037
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004038netdev_features_t passthru_features_check(struct sk_buff *skb,
4039 struct net_device *dev,
4040 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004041netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004042
David S. Miller4d295152012-03-07 21:02:35 -05004043static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004044{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004045 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004046
4047 /* check flags correspondence */
4048 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4049 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4050 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4051 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004052 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004053 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4054 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004055 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4056 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004057 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4058 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004059 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4060 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 -04004061 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004062 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004063 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004064
Herbert Xubcd76112006-06-30 13:36:35 -07004065 return (features & feature) == feature;
4066}
4067
David S. Miller4d295152012-03-07 21:02:35 -05004068static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004069{
Herbert Xu278b2512009-06-03 21:20:51 -07004070 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004071 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004072}
4073
Johannes Berg8b86a612015-04-17 15:45:04 +02004074static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004075 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004076{
Jesse Grossfc741212011-01-09 06:23:32 +00004077 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004078 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4079 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004080}
4081
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004082static inline void netif_set_gso_max_size(struct net_device *dev,
4083 unsigned int size)
4084{
4085 dev->gso_max_size = size;
4086}
4087
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004088static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4089 int pulled_hlen, u16 mac_offset,
4090 int mac_len)
4091{
4092 skb->protocol = protocol;
4093 skb->encapsulation = 1;
4094 skb_push(skb, pulled_hlen);
4095 skb_reset_transport_header(skb);
4096 skb->mac_header = mac_offset;
4097 skb->network_header = skb->mac_header + mac_len;
4098 skb->mac_len = mac_len;
4099}
4100
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004101static inline bool netif_is_macsec(const struct net_device *dev)
4102{
4103 return dev->priv_flags & IFF_MACSEC;
4104}
4105
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004106static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004107{
4108 return dev->priv_flags & IFF_MACVLAN;
4109}
4110
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004111static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004112{
4113 return dev->priv_flags & IFF_MACVLAN_PORT;
4114}
4115
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004116static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004117{
4118 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4119}
4120
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004121static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004122{
4123 return dev->priv_flags & IFF_IPVLAN_MASTER;
4124}
4125
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004126static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004127{
4128 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4129}
4130
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004131static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004132{
4133 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4134}
4135
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004136static inline bool netif_supports_nofcs(struct net_device *dev)
4137{
4138 return dev->priv_flags & IFF_SUPP_NOFCS;
4139}
4140
David Ahern007979e2015-09-29 20:07:10 -07004141static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004142{
David Ahern007979e2015-09-29 20:07:10 -07004143 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004144}
4145
David Ahernfee6d4c2015-10-05 08:51:24 -07004146static inline bool netif_is_l3_slave(const struct net_device *dev)
4147{
4148 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4149}
4150
Jiri Pirko0894ae32015-08-27 09:31:19 +02004151static inline bool netif_is_bridge_master(const struct net_device *dev)
4152{
4153 return dev->priv_flags & IFF_EBRIDGE;
4154}
4155
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004156static inline bool netif_is_bridge_port(const struct net_device *dev)
4157{
4158 return dev->priv_flags & IFF_BRIDGE_PORT;
4159}
4160
Jiri Pirko35d4e172015-08-27 09:31:20 +02004161static inline bool netif_is_ovs_master(const struct net_device *dev)
4162{
4163 return dev->priv_flags & IFF_OPENVSWITCH;
4164}
4165
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004166static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004167{
4168 return dev->priv_flags & IFF_TEAM;
4169}
4170
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004171static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004172{
4173 return dev->priv_flags & IFF_TEAM_PORT;
4174}
4175
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004176static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004177{
4178 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4179}
4180
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004181static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004182{
4183 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4184}
4185
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004186static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4187{
4188 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4189}
4190
Eric Dumazet02875872014-10-05 18:38:35 -07004191/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4192static inline void netif_keep_dst(struct net_device *dev)
4193{
4194 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4195}
4196
Paolo Abeni18d3df32016-07-14 18:00:10 +02004197/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4198static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4199{
4200 /* TODO: reserve and use an additional IFF bit, if we get more users */
4201 return dev->priv_flags & IFF_MACSEC;
4202}
4203
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004204extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004205
Joe Perches571ba422010-02-09 11:49:47 +00004206/* Logging, debugging and troubleshooting/diagnostic helpers. */
4207
4208/* netdev_printk helpers, similar to dev_printk */
4209
4210static inline const char *netdev_name(const struct net_device *dev)
4211{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004212 if (!dev->name[0] || strchr(dev->name, '%'))
4213 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004214 return dev->name;
4215}
4216
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004217static inline const char *netdev_reg_state(const struct net_device *dev)
4218{
4219 switch (dev->reg_state) {
4220 case NETREG_UNINITIALIZED: return " (uninitialized)";
4221 case NETREG_REGISTERED: return "";
4222 case NETREG_UNREGISTERING: return " (unregistering)";
4223 case NETREG_UNREGISTERED: return " (unregistered)";
4224 case NETREG_RELEASED: return " (released)";
4225 case NETREG_DUMMY: return " (dummy)";
4226 }
4227
4228 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4229 return " (unknown)";
4230}
4231
Joe Perchesf629d202013-09-26 14:48:15 -07004232__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004233void netdev_printk(const char *level, const struct net_device *dev,
4234 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004235__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004236void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004237__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004238void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004239__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004240void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004241__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004242void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004243__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004244void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004245__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004246void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004247__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004248void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004249
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004250#define MODULE_ALIAS_NETDEV(device) \
4251 MODULE_ALIAS("netdev-" device)
4252
Jim Cromieb558c962011-12-19 17:11:18 -05004253#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004254#define netdev_dbg(__dev, format, args...) \
4255do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004256 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004257} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004258#elif defined(DEBUG)
4259#define netdev_dbg(__dev, format, args...) \
4260 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004261#else
4262#define netdev_dbg(__dev, format, args...) \
4263({ \
4264 if (0) \
4265 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004266})
4267#endif
4268
4269#if defined(VERBOSE_DEBUG)
4270#define netdev_vdbg netdev_dbg
4271#else
4272
4273#define netdev_vdbg(dev, format, args...) \
4274({ \
4275 if (0) \
4276 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4277 0; \
4278})
4279#endif
4280
4281/*
4282 * netdev_WARN() acts like dev_printk(), but with the key difference
4283 * of using a WARN/WARN_ON to get the message out, including the
4284 * file/line information and a backtrace.
4285 */
4286#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004287 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4288 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004289
Joe Perchesb3d95c52010-02-09 11:49:49 +00004290/* netif printk helpers, similar to netdev_printk */
4291
4292#define netif_printk(priv, type, level, dev, fmt, args...) \
4293do { \
4294 if (netif_msg_##type(priv)) \
4295 netdev_printk(level, (dev), fmt, ##args); \
4296} while (0)
4297
Joe Perchesf45f4322010-06-27 01:02:36 +00004298#define netif_level(level, priv, type, dev, fmt, args...) \
4299do { \
4300 if (netif_msg_##type(priv)) \
4301 netdev_##level(dev, fmt, ##args); \
4302} while (0)
4303
Joe Perchesb3d95c52010-02-09 11:49:49 +00004304#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004305 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004306#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004307 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004308#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004309 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004310#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004311 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004312#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004313 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004314#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004315 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004316#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004317 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004318
Joe Perches0053ea92012-05-30 07:43:34 +00004319#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004320#define netif_dbg(priv, type, netdev, format, args...) \
4321do { \
4322 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004323 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004324} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004325#elif defined(DEBUG)
4326#define netif_dbg(priv, type, dev, format, args...) \
4327 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004328#else
4329#define netif_dbg(priv, type, dev, format, args...) \
4330({ \
4331 if (0) \
4332 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4333 0; \
4334})
4335#endif
4336
4337#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004338#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004339#else
4340#define netif_vdbg(priv, type, dev, format, args...) \
4341({ \
4342 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004343 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004344 0; \
4345})
4346#endif
Joe Perches571ba422010-02-09 11:49:47 +00004347
Cong Wang900ff8c2013-02-18 19:20:33 +00004348/*
4349 * The list of packet types we will receive (as opposed to discard)
4350 * and the routines to invoke.
4351 *
4352 * Why 16. Because with 16 the only overlap we get on a hash of the
4353 * low nibble of the protocol value is RARP/SNAP/X.25.
4354 *
4355 * NOTE: That is no longer true with the addition of VLAN tags. Not
4356 * sure which should go first, but I bet it won't make much
4357 * difference if we are running VLANs. The good news is that
4358 * this protocol won't be in the list unless compiled in, so
4359 * the average user (w/out VLANs) will not be adversely affected.
4360 * --BLG
4361 *
4362 * 0800 IP
4363 * 8100 802.1Q VLAN
4364 * 0001 802.3
4365 * 0002 AX.25
4366 * 0004 802.2
4367 * 8035 RARP
4368 * 0005 SNAP
4369 * 0805 X.25
4370 * 0806 ARP
4371 * 8137 IPX
4372 * 0009 Localtalk
4373 * 86DD IPv6
4374 */
4375#define PTYPE_HASH_SIZE (16)
4376#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4377
Jiri Pirko385a1542009-05-27 15:48:07 -07004378#endif /* _LINUX_NETDEVICE_H */