blob: 97ae0ac513eeec64da23b1df65f17e13a1e1ab1d [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 Pirko6683ece32010-02-04 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 *
Eric Dumazetcdba7562016-01-06 06:53:50 -0800869 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
870 * netdev_features_t features);
871 * Adjusts the requested feature flags according to device-specific
872 * constraints, and returns the resulting flags. Must not modify
873 * the device state.
874 *
Jason Wangf663dd92014-01-10 16:18:26 +0800875 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100876 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500877 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800878 * transmit queues.
879 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800880 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
881 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500882 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800883 *
884 * void (*ndo_set_rx_mode)(struct net_device *dev);
885 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000886 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500887 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800888 *
889 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
890 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700891 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500892 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800893 *
894 * int (*ndo_validate_addr)(struct net_device *dev);
895 * Test if Media Access Control address is valid for the device.
896 *
897 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500898 * Called when a user requests an ioctl which can't be handled by
899 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800900 * not supported error code.
901 *
902 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
903 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500904 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800905 * interface (PCI) for low level management.
906 *
907 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
908 * Called when a user wants to change the Maximum Transfer Unit
909 * of a device. If not defined, any request to change MTU will
910 * will return an error.
911 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800912 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500913 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800914 * for dev->watchdog ticks.
915 *
stephen hemmingerbc1f4472017-01-06 19:12:52 -0800916 * void (*ndo_get_stats64)(struct net_device *dev,
917 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700918 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800919 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000920 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000921 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
922 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700923 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000924 * (which should normally be dev->stats) and return a pointer to
925 * it. The structure may be changed asynchronously only if each
926 * field is written atomically.
927 * 3. Update dev->stats asynchronously and atomically, and define
928 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800929 *
Or Gerlitz3df5b3c2016-11-22 23:09:54 +0200930 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
Nogah Frankel2c9d85d2016-09-16 15:05:36 +0200931 * Return true if this device supports offload stats of this attr_id.
932 *
933 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
934 * void *attr_data)
935 * Get statistics for offload operations by attr_id. Write it into the
936 * attr_data pointer.
937 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530938 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500939 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000940 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800941 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530942 * int (*ndo_vlan_rx_kill_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 unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800945 *
946 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000947 *
948 * SR-IOV management functions.
949 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
Moshe Shemesh79aab092016-09-22 12:11:15 +0300950 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
951 * u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400952 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
953 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000954 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000955 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000956 * int (*ndo_get_vf_config)(struct net_device *dev,
957 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300958 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700959 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
960 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300961 *
962 * Enable or disable the VF ability to query its RSS Redirection Table and
963 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500964 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300965 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700966 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +0000967 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
968 * Called to setup 'tc' number of traffic classes in the net device. This
969 * is always called from the stack with the rtnl lock held and netif tx
970 * queues stopped. This allows the netdevice to perform queue management
971 * safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000972 *
Yi Zoue9bce842011-03-09 08:48:03 +0000973 * Fiber Channel over Ethernet (FCoE) offload functions.
974 * int (*ndo_fcoe_enable)(struct net_device *dev);
975 * Called when the FCoE protocol stack wants to start using LLD for FCoE
976 * so the underlying device can perform whatever needed configuration or
977 * initialization to support acceleration of FCoE traffic.
978 *
979 * int (*ndo_fcoe_disable)(struct net_device *dev);
980 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
981 * so the underlying device can perform whatever needed clean-ups to
982 * stop supporting acceleration of FCoE traffic.
983 *
984 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
985 * struct scatterlist *sgl, unsigned int sgc);
986 * Called when the FCoE Initiator wants to initialize an I/O that
987 * is a possible candidate for Direct Data Placement (DDP). The LLD can
988 * perform necessary setup and returns 1 to indicate the device is set up
989 * successfully to perform DDP on this I/O, otherwise this returns 0.
990 *
991 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
992 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
993 * indicated by the FC exchange id 'xid', so the underlying device can
994 * clean up and reuse resources for later DDP requests.
995 *
996 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
997 * struct scatterlist *sgl, unsigned int sgc);
998 * Called when the FCoE Target wants to initialize an I/O that
999 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1000 * perform necessary setup and returns 1 to indicate the device is set up
1001 * successfully to perform DDP on this I/O, otherwise this returns 0.
1002 *
Neerav Parikh68bad942012-01-04 20:23:39 +00001003 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1004 * struct netdev_fcoe_hbainfo *hbainfo);
1005 * Called when the FCoE Protocol stack wants information on the underlying
1006 * device. This information is utilized by the FCoE protocol stack to
1007 * register attributes with Fiber Channel management service as per the
1008 * FC-GS Fabric Device Management Information(FDMI) specification.
1009 *
Yi Zoue9bce842011-03-09 08:48:03 +00001010 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1011 * Called when the underlying device wants to override default World Wide
1012 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1013 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1014 * protocol stack to use.
1015 *
Ben Hutchingsc4454772011-01-19 11:03:53 +00001016 * RFS acceleration.
1017 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1018 * u16 rxq_index, u32 flow_id);
1019 * Set hardware filter for RFS. rxq_index is the target queue index;
1020 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1021 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001022 *
Jiri Pirko8b98a702013-01-03 22:49:02 +00001023 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001024 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1025 * Called to make another netdev an underling.
1026 *
1027 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1028 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +00001029 *
1030 * Feature/offload setting functions.
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001031 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001032 * Called to update device configuration to new features. Passed
1033 * feature set might be less than what was returned by ndo_fix_features()).
1034 * Must return >0 or -errno if it changed dev->features itself.
1035 *
stephen hemmingeredc7d572012-10-01 12:32:33 +00001036 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1037 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001038 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +00001039 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +00001040 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1041 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001042 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001043 * Deletes the FDB entry from dev coresponding to addr.
1044 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001045 * struct net_device *dev, struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001046 * int *idx)
John Fastabend77162022012-04-15 06:43:56 +00001047 * Used to add FDB entries to dump requests. Implementers should add
1048 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001049 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001050 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1051 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001052 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001053 * struct net_device *dev, u32 filter_mask,
1054 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001055 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1056 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001057 *
1058 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1059 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1060 * which do not represent real hardware may define this to allow their
1061 * userspace components to manage their virtual carrier state. Devices
1062 * that determine carrier state from physical hardware properties (eg
1063 * network cables) or protocol-dependent mechanisms (eg
1064 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001065 *
1066 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001067 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001068 * Called to get ID of physical port of this device. If driver does
1069 * not implement this, it is assumed that the hw is not able to have
1070 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001071 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001072 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1073 * struct udp_tunnel_info *ti);
1074 * Called by UDP tunnel to notify a driver about the UDP port and socket
1075 * address family that a UDP tunnel is listnening to. It is called only
1076 * when a new port starts listening. The operation is protected by the
1077 * RTNL.
1078 *
1079 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1080 * struct udp_tunnel_info *ti);
1081 * Called by UDP tunnel to notify the driver about a UDP port and socket
1082 * address family that the UDP tunnel is not listening to anymore. The
1083 * operation is protected by the RTNL.
1084 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001085 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1086 * struct net_device *dev)
1087 * Called by upper layer devices to accelerate switching or other
1088 * station functionality into hardware. 'pdev is the lowerdev
1089 * to use for the offload and 'dev' is the net device that will
1090 * back the offload. Returns a pointer to the private structure
1091 * the upper layer will maintain.
1092 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1093 * Called by upper layer device to delete the station created
1094 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1095 * the station and priv is the structure returned by the add
1096 * operation.
1097 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1098 * struct net_device *dev,
1099 * void *priv);
1100 * Callback to use for xmit over the accelerated station. This
1101 * is used in place of ndo_start_xmit on accelerated net
1102 * devices.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001103 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1104 * struct net_device *dev
1105 * netdev_features_t features);
Tom Herbert04ffcb22014-10-14 15:19:06 -07001106 * Called by core transmit path to determine if device is capable of
Jesse Gross5f352272014-12-23 22:37:26 -08001107 * performing offload operations on a given packet. This is to give
1108 * the device an opportunity to implement any restrictions that cannot
1109 * be otherwise expressed by feature flags. The check is called with
1110 * the set of features that the stack has calculated and it returns
1111 * those the driver believes to be appropriate.
John Fastabend822b3b22015-03-18 14:57:33 +02001112 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1113 * int queue_index, u32 maxrate);
1114 * Called when a user wants to set a max-rate limitation of specific
1115 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001116 * int (*ndo_get_iflink)(const struct net_device *dev);
1117 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001118 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001119 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001120 * This function is used to pass protocol port error state information
1121 * to the switch driver. The switch driver can react to the proto_down
1122 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001123 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1124 * This function is used to get egress tunnel information for given skb.
1125 * This is useful for retrieving outer tunnel header parameters while
1126 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001127 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1128 * This function is used to specify the headroom that the skb must
1129 * consider when allocation skb during packet reception. Setting
1130 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001131 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001132 * default value.
Brenden Blancoa7862b42016-07-19 12:16:48 -07001133 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1134 * This function is used to set or query state related to XDP on the
1135 * netdevice. See definition of enum xdp_netdev_command for details.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001136 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001137 */
1138struct net_device_ops {
1139 int (*ndo_init)(struct net_device *dev);
1140 void (*ndo_uninit)(struct net_device *dev);
1141 int (*ndo_open)(struct net_device *dev);
1142 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001143 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1144 struct net_device *dev);
1145 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1146 struct net_device *dev,
1147 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001148 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001149 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001150 void *accel_priv,
1151 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001152 void (*ndo_change_rx_flags)(struct net_device *dev,
1153 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001154 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001155 int (*ndo_set_mac_address)(struct net_device *dev,
1156 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001157 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001158 int (*ndo_do_ioctl)(struct net_device *dev,
1159 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001160 int (*ndo_set_config)(struct net_device *dev,
1161 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001162 int (*ndo_change_mtu)(struct net_device *dev,
1163 int new_mtu);
1164 int (*ndo_neigh_setup)(struct net_device *dev,
1165 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001166 void (*ndo_tx_timeout) (struct net_device *dev);
1167
stephen hemmingerbc1f4472017-01-06 19:12:52 -08001168 void (*ndo_get_stats64)(struct net_device *dev,
1169 struct rtnl_link_stats64 *storage);
Or Gerlitz3df5b3c2016-11-22 23:09:54 +02001170 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
Nogah Frankel2c9d85d2016-09-16 15:05:36 +02001171 int (*ndo_get_offload_stats)(int attr_id,
1172 const struct net_device *dev,
1173 void *attr_data);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001174 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1175
Jiri Pirko8e586132011-12-08 19:52:37 -05001176 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001177 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001178 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001179 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001180#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001181 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001182 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001183 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001184 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001185#endif
Cong Wange0d10952013-08-01 11:10:25 +08001186#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir8b80cda2013-07-10 17:13:26 +03001187 int (*ndo_busy_poll)(struct napi_struct *dev);
Eliezer Tamir06021292013-06-10 11:39:50 +03001188#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001189 int (*ndo_set_vf_mac)(struct net_device *dev,
1190 int queue, u8 *mac);
1191 int (*ndo_set_vf_vlan)(struct net_device *dev,
Moshe Shemesh79aab092016-09-22 12:11:15 +03001192 int queue, u16 vlan,
1193 u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001194 int (*ndo_set_vf_rate)(struct net_device *dev,
1195 int vf, int min_tx_rate,
1196 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001197 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1198 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001199 int (*ndo_set_vf_trust)(struct net_device *dev,
1200 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001201 int (*ndo_get_vf_config)(struct net_device *dev,
1202 int vf,
1203 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001204 int (*ndo_set_vf_link_state)(struct net_device *dev,
1205 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001206 int (*ndo_get_vf_stats)(struct net_device *dev,
1207 int vf,
1208 struct ifla_vf_stats
1209 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001210 int (*ndo_set_vf_port)(struct net_device *dev,
1211 int vf,
1212 struct nlattr *port[]);
1213 int (*ndo_get_vf_port)(struct net_device *dev,
1214 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001215 int (*ndo_set_vf_guid)(struct net_device *dev,
1216 int vf, u64 guid,
1217 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001218 int (*ndo_set_vf_rss_query_en)(
1219 struct net_device *dev,
1220 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001221 int (*ndo_setup_tc)(struct net_device *dev,
1222 u32 handle,
1223 __be16 protocol,
1224 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001225#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001226 int (*ndo_fcoe_enable)(struct net_device *dev);
1227 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001228 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1229 u16 xid,
1230 struct scatterlist *sgl,
1231 unsigned int sgc);
1232 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1233 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001234 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1235 u16 xid,
1236 struct scatterlist *sgl,
1237 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001238 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1239 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001240#endif
1241
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001242#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001243#define NETDEV_FCOE_WWNN 0
1244#define NETDEV_FCOE_WWPN 1
1245 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1246 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001247#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001248
Ben Hutchingsc4454772011-01-19 11:03:53 +00001249#ifdef CONFIG_RFS_ACCEL
1250 int (*ndo_rx_flow_steer)(struct net_device *dev,
1251 const struct sk_buff *skb,
1252 u16 rxq_index,
1253 u32 flow_id);
1254#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001255 int (*ndo_add_slave)(struct net_device *dev,
1256 struct net_device *slave_dev);
1257 int (*ndo_del_slave)(struct net_device *dev,
1258 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001259 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1260 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001261 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001262 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001263 int (*ndo_neigh_construct)(struct net_device *dev,
1264 struct neighbour *n);
1265 void (*ndo_neigh_destroy)(struct net_device *dev,
1266 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001267
1268 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001269 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001270 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001271 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001272 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001273 u16 flags);
1274 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001275 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001276 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001277 const unsigned char *addr,
1278 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001279 int (*ndo_fdb_dump)(struct sk_buff *skb,
1280 struct netlink_callback *cb,
1281 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001282 struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001283 int *idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001284
1285 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001286 struct nlmsghdr *nlh,
1287 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001288 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1289 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001290 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001291 u32 filter_mask,
1292 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001293 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001294 struct nlmsghdr *nlh,
1295 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001296 int (*ndo_change_carrier)(struct net_device *dev,
1297 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001298 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001299 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001300 int (*ndo_get_phys_port_name)(struct net_device *dev,
1301 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001302 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1303 struct udp_tunnel_info *ti);
1304 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1305 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001306 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1307 struct net_device *dev);
1308 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1309 void *priv);
1310
1311 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1312 struct net_device *dev,
1313 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001314 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001315 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1316 int queue_index,
1317 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001318 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001319 int (*ndo_change_proto_down)(struct net_device *dev,
1320 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001321 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1322 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001323 void (*ndo_set_rx_headroom)(struct net_device *dev,
1324 int needed_headroom);
Brenden Blancoa7862b42016-07-19 12:16:48 -07001325 int (*ndo_xdp)(struct net_device *dev,
1326 struct netdev_xdp *xdp);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001327};
1328
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001329/**
1330 * enum net_device_priv_flags - &struct net_device priv_flags
1331 *
1332 * These are the &struct net_device, they are only set internally
1333 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001334 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001335 * during any kernel release.
1336 *
1337 * You should have a pretty good reason to be extending these flags.
1338 *
1339 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1340 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001341 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001342 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001343 * @IFF_WAN_HDLC: WAN HDLC device
1344 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1345 * release skb->dst
1346 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1347 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1348 * @IFF_MACVLAN_PORT: device used as macvlan port
1349 * @IFF_BRIDGE_PORT: device used as bridge port
1350 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1351 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1352 * @IFF_UNICAST_FLT: Supports unicast filtering
1353 * @IFF_TEAM_PORT: device used as team port
1354 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1355 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1356 * change when it's running
1357 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001358 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1359 * underlying stacked devices
1360 * @IFF_IPVLAN_MASTER: IPvlan master device
1361 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001362 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001363 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001364 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001365 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001366 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001367 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001368 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1369 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001370 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001371 */
1372enum netdev_priv_flags {
1373 IFF_802_1Q_VLAN = 1<<0,
1374 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001375 IFF_BONDING = 1<<2,
1376 IFF_ISATAP = 1<<3,
1377 IFF_WAN_HDLC = 1<<4,
1378 IFF_XMIT_DST_RELEASE = 1<<5,
1379 IFF_DONT_BRIDGE = 1<<6,
1380 IFF_DISABLE_NETPOLL = 1<<7,
1381 IFF_MACVLAN_PORT = 1<<8,
1382 IFF_BRIDGE_PORT = 1<<9,
1383 IFF_OVS_DATAPATH = 1<<10,
1384 IFF_TX_SKB_SHARING = 1<<11,
1385 IFF_UNICAST_FLT = 1<<12,
1386 IFF_TEAM_PORT = 1<<13,
1387 IFF_SUPP_NOFCS = 1<<14,
1388 IFF_LIVE_ADDR_CHANGE = 1<<15,
1389 IFF_MACVLAN = 1<<16,
1390 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1391 IFF_IPVLAN_MASTER = 1<<18,
1392 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001393 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001394 IFF_NO_QUEUE = 1<<21,
1395 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001396 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001397 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001398 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001399 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001400 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001401};
1402
1403#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1404#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001405#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001406#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001407#define IFF_WAN_HDLC IFF_WAN_HDLC
1408#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1409#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1410#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1411#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1412#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1413#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1414#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1415#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1416#define IFF_TEAM_PORT IFF_TEAM_PORT
1417#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1418#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1419#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001420#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001421#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1422#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001423#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001424#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001425#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001426#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001427#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001428#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001429#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001430
Karoly Kemeny536721b2014-07-30 20:27:36 +02001431/**
1432 * struct net_device - The DEVICE structure.
1433 * Actually, this whole structure is a big mistake. It mixes I/O
1434 * data with strictly "high-level" data, and it has to know about
1435 * almost every data structure used in the INET module.
1436 *
1437 * @name: This is the first field of the "visible" part of this structure
1438 * (i.e. as seen by users in the "Space.c" file). It is the name
1439 * of the interface.
1440 *
1441 * @name_hlist: Device name hash chain, please keep it close to name[]
1442 * @ifalias: SNMP alias
1443 * @mem_end: Shared memory end
1444 * @mem_start: Shared memory start
1445 * @base_addr: Device I/O address
1446 * @irq: Device IRQ number
1447 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001448 * @carrier_changes: Stats to monitor carrier on<->off transitions
1449 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001450 * @state: Generic network queuing layer state, see netdev_state_t
1451 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001452 * @napi_list: List entry used for polling NAPI devices
1453 * @unreg_list: List entry when we are unregistering the
1454 * device; see the function unregister_netdev
1455 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001456 * @ptype_all: Device-specific packet handlers for all protocols
1457 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001458 *
1459 * @adj_list: Directly linked devices, like slaves for bonding
Karoly Kemeny536721b2014-07-30 20:27:36 +02001460 * @features: Currently active device features
1461 * @hw_features: User-changeable features
1462 *
1463 * @wanted_features: User-requested features
1464 * @vlan_features: Mask of features inheritable by VLAN devices
1465 *
1466 * @hw_enc_features: Mask of features inherited by encapsulating devices
1467 * This field indicates what encapsulation
1468 * offloads the hardware is capable of doing,
1469 * and drivers will need to set them appropriately.
1470 *
1471 * @mpls_features: Mask of features inheritable by MPLS
1472 *
1473 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001474 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001475 *
1476 * @stats: Statistics struct, which was left as a legacy, use
1477 * rtnl_link_stats64 instead
1478 *
1479 * @rx_dropped: Dropped packets by core network,
1480 * do not use this in drivers
1481 * @tx_dropped: Dropped packets by core network,
1482 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001483 * @rx_nohandler: nohandler dropped packets by core network on
1484 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001485 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001486 * @wireless_handlers: List of functions to handle Wireless Extensions,
1487 * instead of ioctl,
1488 * see <net/iw_handler.h> for details.
1489 * @wireless_data: Instance data managed by the core of wireless extensions
1490 *
1491 * @netdev_ops: Includes several pointers to callbacks,
1492 * if one wants to override the ndo_*() functions
1493 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001494 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1495 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001496 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001497 * of Layer 2 headers.
1498 *
1499 * @flags: Interface flags (a la BSD)
1500 * @priv_flags: Like 'flags' but invisible to userspace,
1501 * see if.h for the definitions
1502 * @gflags: Global flags ( kept as legacy )
1503 * @padded: How much padding added by alloc_netdev()
1504 * @operstate: RFC2863 operstate
1505 * @link_mode: Mapping policy to operstate
1506 * @if_port: Selectable AUI, TP, ...
1507 * @dma: DMA channel
1508 * @mtu: Interface MTU value
Jarod Wilson61e84622016-10-07 22:04:33 -04001509 * @min_mtu: Interface Minimum MTU value
1510 * @max_mtu: Interface Maximum MTU value
Karoly Kemeny536721b2014-07-30 20:27:36 +02001511 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001512 * @hard_header_len: Maximum hardware header length.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001513 *
1514 * @needed_headroom: Extra headroom the hardware may need, but not in all
1515 * cases can this be guaranteed
1516 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1517 * cases can this be guaranteed. Some cases also use
1518 * LL_MAX_HEADER instead to allocate the skb
1519 *
1520 * interface address info:
1521 *
1522 * @perm_addr: Permanent hw address
1523 * @addr_assign_type: Hw address assignment type
1524 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001525 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001526 * @dev_id: Used to differentiate devices that share
1527 * the same link layer address
1528 * @dev_port: Used to differentiate devices that share
1529 * the same function
1530 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001531 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001532 * has been enabled due to the need to listen to
1533 * additional unicast addresses in a device that
1534 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001535 * @uc: unicast mac addresses
1536 * @mc: multicast mac addresses
1537 * @dev_addrs: list of device hw addresses
1538 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001539 * @promiscuity: Number of times the NIC is told to work in
1540 * promiscuous mode; if it becomes 0 the NIC will
1541 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001542 * @allmulti: Counter, enables or disables allmulticast mode
1543 *
1544 * @vlan_info: VLAN info
1545 * @dsa_ptr: dsa specific data
1546 * @tipc_ptr: TIPC specific data
1547 * @atalk_ptr: AppleTalk link
1548 * @ip_ptr: IPv4 specific data
1549 * @dn_ptr: DECnet specific data
1550 * @ip6_ptr: IPv6 specific data
1551 * @ax25_ptr: AX.25 specific data
1552 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1553 *
1554 * @last_rx: Time of last Rx
1555 * @dev_addr: Hw address (before bcast,
1556 * because most packets are unicast)
1557 *
1558 * @_rx: Array of RX queues
1559 * @num_rx_queues: Number of RX queues
1560 * allocated at register_netdev() time
1561 * @real_num_rx_queues: Number of RX queues currently active in device
1562 *
1563 * @rx_handler: handler for received packets
1564 * @rx_handler_data: XXX: need comments on this one
1565 * @ingress_queue: XXX: need comments on this one
1566 * @broadcast: hw bcast address
1567 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001568 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1569 * indexed by RX queue number. Assigned by driver.
1570 * This must only be set if the ndo_rx_flow_steer
1571 * operation is defined
1572 * @index_hlist: Device index hash chain
1573 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001574 * @_tx: Array of TX queues
1575 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1576 * @real_num_tx_queues: Number of TX queues currently active in device
1577 * @qdisc: Root qdisc from userspace point of view
1578 * @tx_queue_len: Max frames per queue allowed
1579 * @tx_global_lock: XXX: need comments on this one
1580 *
1581 * @xps_maps: XXX: need comments on this one
1582 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001583 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001584 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001585 * @watchdog_timer: List of timers
1586 *
1587 * @pcpu_refcnt: Number of references to this device
1588 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001589 * @link_watch_list: XXX: need comments on this one
1590 *
1591 * @reg_state: Register/unregister state machine
1592 * @dismantle: Device is going to be freed
1593 * @rtnl_link_state: This enum represents the phases of creating
1594 * a new link
1595 *
1596 * @destructor: Called from unregister,
1597 * can be used to call free_netdev
1598 * @npinfo: XXX: need comments on this one
1599 * @nd_net: Network namespace this network device is inside
1600 *
1601 * @ml_priv: Mid-layer private
1602 * @lstats: Loopback statistics
1603 * @tstats: Tunnel statistics
1604 * @dstats: Dummy statistics
1605 * @vstats: Virtual ethernet statistics
1606 *
1607 * @garp_port: GARP
1608 * @mrp_port: MRP
1609 *
1610 * @dev: Class/net/name entry
1611 * @sysfs_groups: Space for optional device, statistics and wireless
1612 * sysfs groups
1613 *
1614 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1615 * @rtnl_link_ops: Rtnl_link_ops
1616 *
1617 * @gso_max_size: Maximum size of generic segmentation offload
1618 * @gso_max_segs: Maximum number of segments that can be passed to the
1619 * NIC for GSO
1620 *
1621 * @dcbnl_ops: Data Center Bridging netlink ops
1622 * @num_tc: Number of traffic classes in the net device
1623 * @tc_to_txq: XXX: need comments on this one
Randy Dunlap920c1cd2016-11-21 18:28:36 -08001624 * @prio_tc_map: XXX: need comments on this one
Karoly Kemeny536721b2014-07-30 20:27:36 +02001625 *
1626 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1627 *
1628 * @priomap: XXX: need comments on this one
1629 * @phydev: Physical device may attach itself
1630 * for hardware timestamping
1631 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001632 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1633 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001634 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001635 * @proto_down: protocol port state information can be sent to the
1636 * switch driver and used to set the phys state of the
1637 * switch port.
1638 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639 * FIXME: cleanup struct net_device such that network protocol info
1640 * moves out.
1641 */
1642
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001643struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001645 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001646 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 /*
1648 * I/O specific fields
1649 * FIXME: Merge these and struct ifmap into one
1650 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001651 unsigned long mem_end;
1652 unsigned long mem_start;
1653 unsigned long base_addr;
1654 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001656 atomic_t carrier_changes;
1657
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001659 * Some hardware also needs these fields (state,dev_list,
1660 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 * part of the usual set specified in Space.c.
1662 */
1663
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 unsigned long state;
1665
Pavel Emelianov7562f872007-05-03 15:13:45 -07001666 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001667 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001668 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001669 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001670 struct list_head ptype_all;
1671 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001672
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001673 struct {
1674 struct list_head upper;
1675 struct list_head lower;
1676 } adj_list;
1677
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001678 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001679 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001680 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001681 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001682 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001683 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001684 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001685
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001687 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688
Rusty Russellc45d2862007-03-28 14:29:08 -07001689 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001690
Eric Dumazet015f0682014-03-27 08:45:56 -07001691 atomic_long_t rx_dropped;
1692 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001693 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694
Johannes Bergb86e0282007-04-26 20:48:23 -07001695#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001696 const struct iw_handler_def *wireless_handlers;
1697 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001698#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001699 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001700 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001701#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001702 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001703#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001704#ifdef CONFIG_NET_L3_MASTER_DEV
1705 const struct l3mdev_ops *l3mdev_ops;
1706#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001707#if IS_ENABLED(CONFIG_IPV6)
1708 const struct ndisc_ops *ndisc_ops;
1709#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001711 const struct header_ops *header_ops;
1712
Karoly Kemeny536721b2014-07-30 20:27:36 +02001713 unsigned int flags;
1714 unsigned int priv_flags;
1715
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001717 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718
Karoly Kemeny536721b2014-07-30 20:27:36 +02001719 unsigned char operstate;
1720 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001721
Karoly Kemeny536721b2014-07-30 20:27:36 +02001722 unsigned char if_port;
1723 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001724
Karoly Kemeny536721b2014-07-30 20:27:36 +02001725 unsigned int mtu;
Jarod Wilson61e84622016-10-07 22:04:33 -04001726 unsigned int min_mtu;
1727 unsigned int max_mtu;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001728 unsigned short type;
1729 unsigned short hard_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730
Johannes Bergf5184d22008-05-12 20:48:31 -07001731 unsigned short needed_headroom;
1732 unsigned short needed_tailroom;
1733
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001735 unsigned char perm_addr[MAX_ADDR_LEN];
1736 unsigned char addr_assign_type;
1737 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001738 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001739 unsigned short dev_id;
1740 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001741 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001742 unsigned char name_assign_type;
1743 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001744 struct netdev_hw_addr_list uc;
1745 struct netdev_hw_addr_list mc;
1746 struct netdev_hw_addr_list dev_addrs;
1747
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001748#ifdef CONFIG_SYSFS
1749 struct kset *queues_kset;
1750#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001751 unsigned int promiscuity;
1752 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001755 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001756
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001757#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001758 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001759#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001760#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001761 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001762#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001763#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001764 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001765#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001766 void *atalk_ptr;
1767 struct in_device __rcu *ip_ptr;
1768 struct dn_dev __rcu *dn_ptr;
1769 struct inet6_dev __rcu *ip6_ptr;
1770 void *ax25_ptr;
1771 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001772 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001773#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1774 struct mpls_dev __rcu *mpls_ptr;
1775#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001777/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001778 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001779 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001780 unsigned long last_rx;
Eric Dumazet4dc89132010-08-31 07:40:16 +00001781
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001782 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001783 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001784
Michael Daltona953be52014-01-16 22:23:28 -08001785#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001786 struct netdev_rx_queue *_rx;
1787
Tom Herbert0a9627f2010-03-16 08:03:29 +00001788 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001789 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001790#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001791
Eric Dumazet3b47d302014-11-06 21:09:44 -08001792 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001793 rx_handler_func_t __rcu *rx_handler;
1794 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001795
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001796#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001797 struct tcf_proto __rcu *ingress_cl_list;
1798#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001799 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001800#ifdef CONFIG_NETFILTER_INGRESS
Aaron Conolee3b37f12016-09-21 11:35:07 -04001801 struct nf_hook_entry __rcu *nf_hooks_ingress;
Pablo Neirae687ad62015-05-13 18:19:38 +02001802#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001803
Karoly Kemeny536721b2014-07-30 20:27:36 +02001804 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001805#ifdef CONFIG_RFS_ACCEL
1806 struct cpu_rmap *rx_cpu_rmap;
1807#endif
1808 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001809
1810/*
1811 * Cache lines mostly used on transmit path
1812 */
David S. Millere8a04642008-07-17 00:34:19 -07001813 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1814 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001815 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001816 struct Qdisc *qdisc;
Jiri Kosina59cc1f62016-08-10 11:05:15 +02001817#ifdef CONFIG_NET_SCHED
1818 DECLARE_HASHTABLE (qdisc_hash, 4);
1819#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001820 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001821 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001822 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001823
Tom Herbertbf264142010-11-26 08:36:09 +00001824#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001825 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001826#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001827#ifdef CONFIG_NET_CLS_ACT
1828 struct tcf_proto __rcu *egress_cl_list;
1829#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001830
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001831 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001832 struct timer_list watchdog_timer;
1833
Eric Dumazet29b44332010-10-11 10:22:12 +00001834 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836
Eric Dumazete014deb2009-11-17 05:59:21 +00001837 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001838
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001840 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001841 NETREG_UNREGISTERING, /* called unregister_netdevice */
1842 NETREG_UNREGISTERED, /* completed unregister todo */
1843 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001844 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001845 } reg_state:8;
1846
Karoly Kemeny536721b2014-07-30 20:27:36 +02001847 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001848
1849 enum {
1850 RTNL_LINK_INITIALIZED,
1851 RTNL_LINK_INITIALIZING,
1852 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001854 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001857 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858#endif
David S. Millereae792b2008-07-15 03:03:33 -07001859
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001860 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001861
David S. Miller49517042008-05-12 03:29:11 -07001862 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001863 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001864 void *ml_priv;
1865 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001866 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001867 struct pcpu_dstats __percpu *dstats;
1868 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001869 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001870
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001871 struct garp_port __rcu *garp_port;
David Wardfebf0182013-02-08 17:17:06 +00001872 struct mrp_port __rcu *mrp_port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001874 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001875 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001876 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001877
Patrick McHardy38f7b872007-06-13 12:03:51 -07001878 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001879
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001880 /* for setting kernel sock attribute on TCP connection setup */
1881#define GSO_MAX_SIZE 65536
1882 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001883#define GSO_MAX_SEGS 65535
1884 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001885
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001886#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001887 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001888#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001889 u8 num_tc;
1890 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1891 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001892
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001893#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001894 unsigned int fcoe_ddp_xid;
1895#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001896#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001897 struct netprio_map __rcu *priomap;
1898#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001899 struct phy_device *phydev;
1900 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001901 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001902 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001904#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905
1906#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907
David S. Millere8a04642008-07-17 00:34:19 -07001908static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001909int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1910{
1911 return dev->prio_tc_map[prio & TC_BITMASK];
1912}
1913
1914static inline
1915int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1916{
1917 if (tc >= dev->num_tc)
1918 return -EINVAL;
1919
1920 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1921 return 0;
1922}
1923
Alexander Duyck8d059b02016-10-28 11:43:49 -04001924int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
Alexander Duyck9cf1f6a2016-10-28 11:43:20 -04001925void netdev_reset_tc(struct net_device *dev);
1926int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1927int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
John Fastabend4f57c082011-01-17 08:06:04 +00001928
1929static inline
1930int netdev_get_num_tc(struct net_device *dev)
1931{
1932 return dev->num_tc;
1933}
1934
1935static inline
David S. Millere8a04642008-07-17 00:34:19 -07001936struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1937 unsigned int index)
1938{
1939 return &dev->_tx[index];
1940}
1941
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001942static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1943 const struct sk_buff *skb)
1944{
1945 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1946}
1947
David S. Millere8a04642008-07-17 00:34:19 -07001948static inline void netdev_for_each_tx_queue(struct net_device *dev,
1949 void (*f)(struct net_device *,
1950 struct netdev_queue *,
1951 void *),
1952 void *arg)
1953{
1954 unsigned int i;
1955
1956 for (i = 0; i < dev->num_tx_queues; i++)
1957 f(dev, &dev->_tx[i], arg);
1958}
1959
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001960#define netdev_lockdep_set_classes(dev) \
1961{ \
1962 static struct lock_class_key qdisc_tx_busylock_key; \
1963 static struct lock_class_key qdisc_running_key; \
1964 static struct lock_class_key qdisc_xmit_lock_key; \
1965 static struct lock_class_key dev_addr_list_lock_key; \
1966 unsigned int i; \
1967 \
1968 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1969 (dev)->qdisc_running_key = &qdisc_running_key; \
1970 lockdep_set_class(&(dev)->addr_list_lock, \
1971 &dev_addr_list_lock_key); \
1972 for (i = 0; i < (dev)->num_tx_queues; i++) \
1973 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1974 &qdisc_xmit_lock_key); \
1975}
1976
Joe Perchesf629d202013-09-26 14:48:15 -07001977struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001978 struct sk_buff *skb,
1979 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001980
Paolo Abeni871b6422016-02-26 10:45:37 +01001981/* returns the headroom that the master device needs to take in account
1982 * when forwarding to this dev
1983 */
1984static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1985{
1986 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1987}
1988
1989static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1990{
1991 if (dev->netdev_ops->ndo_set_rx_headroom)
1992 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1993}
1994
1995/* set the device rx headroom to the dev's default */
1996static inline void netdev_reset_rx_headroom(struct net_device *dev)
1997{
1998 netdev_set_rx_headroom(dev, -1);
1999}
2000
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002001/*
2002 * Net namespace inlines
2003 */
2004static inline
2005struct net *dev_net(const struct net_device *dev)
2006{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002007 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002008}
2009
2010static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002011void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002012{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05002013 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002014}
2015
Florian Fainelli3e8a72d2014-08-27 17:04:46 -07002016static inline bool netdev_uses_dsa(struct net_device *dev)
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00002017{
Alexander Duyck3fc88672014-09-12 23:12:46 -07002018#if IS_ENABLED(CONFIG_NET_DSA)
Florian Fainelli5aed85c2014-08-27 17:04:52 -07002019 if (dev->dsa_ptr != NULL)
2020 return dsa_uses_tagged_protocol(dev->dsa_ptr);
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00002021#endif
Florian Fainelli5aed85c2014-08-27 17:04:52 -07002022 return false;
Lennert Buytenhek396138f2008-10-07 13:46:07 +00002023}
2024
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002025/**
2026 * netdev_priv - access network device private data
2027 * @dev: network device
2028 *
2029 * Get network device private data
2030 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002031static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002033 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034}
2035
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036/* Set the sysfs physical device reference for the network logical device
2037 * if set prior to registration will cause a symlink during initialization.
2038 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002039#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040
Marcel Holtmann384912e2009-08-31 21:08:19 +00002041/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002042 * fine-grained identification of different network device types. For
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002043 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
Marcel Holtmann384912e2009-08-31 21:08:19 +00002044 */
2045#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2046
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002047/* Default NAPI poll() weight
2048 * Device drivers are strongly advised to not use bigger value
2049 */
2050#define NAPI_POLL_WEIGHT 64
2051
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002052/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002053 * netif_napi_add - initialize a NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002054 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002055 * @napi: NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002056 * @poll: polling function
2057 * @weight: default weight
2058 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002059 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2060 * *any* of the other NAPI-related functions.
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002061 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002062void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2063 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002064
Alexander Duyckd8156532008-07-08 15:13:05 -07002065/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002066 * netif_tx_napi_add - initialize a NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002067 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002068 * @napi: NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002069 * @poll: polling function
2070 * @weight: default weight
2071 *
2072 * This variant of netif_napi_add() should be used from drivers using NAPI
2073 * to exclusively poll a TX queue.
2074 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2075 */
2076static inline void netif_tx_napi_add(struct net_device *dev,
2077 struct napi_struct *napi,
2078 int (*poll)(struct napi_struct *, int),
2079 int weight)
2080{
2081 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2082 netif_napi_add(dev, napi, poll, weight);
2083}
2084
2085/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002086 * netif_napi_del - remove a NAPI context
2087 * @napi: NAPI context
Alexander Duyckd8156532008-07-08 15:13:05 -07002088 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002089 * netif_napi_del() removes a NAPI context from the network device NAPI list
Alexander Duyckd8156532008-07-08 15:13:05 -07002090 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002091void netif_napi_del(struct napi_struct *napi);
2092
2093struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002094 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002095 void *frag0;
Herbert Xu78a478d2009-05-26 18:50:21 +00002096
Herbert Xu74895942009-05-26 18:50:27 +00002097 /* Length of frag0. */
2098 unsigned int frag0_len;
2099
Herbert Xu86911732009-01-29 14:19:50 +00002100 /* This indicates where we are processing relative to skb->data. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002101 int data_offset;
Herbert Xu86911732009-01-29 14:19:50 +00002102
Herbert Xud565b0a2008-12-15 23:38:52 -08002103 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002104 u16 flush;
2105
2106 /* Save the IP ID here and check when we get to the transport layer */
2107 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002108
2109 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002110 u16 count;
2111
Tom Herbert15e23962015-02-10 16:30:31 -08002112 /* Start offset for remote checksum offload */
2113 u16 gro_remcsum_start;
2114
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002115 /* jiffies when first packet was created/queued */
2116 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002117
Tom Herbertafe93322014-09-17 12:25:57 -07002118 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002119 u16 proto;
2120
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002121 /* This is non-zero if the packet may be of the same flow. */
2122 u8 same_flow:1;
2123
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002124 /* Used in tunnel GRO receive */
2125 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002126
2127 /* GRO checksum is valid */
2128 u8 csum_valid:1;
2129
Tom Herbert662880f2014-08-27 21:26:56 -07002130 /* Number of checksums via CHECKSUM_UNNECESSARY */
2131 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002132
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002133 /* Free the skb? */
2134 u8 free:2;
2135#define NAPI_GRO_FREE 1
2136#define NAPI_GRO_FREE_STOLEN_HEAD 2
2137
Tom Herbertefc98d02014-10-03 15:48:08 -07002138 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2139 u8 is_ipv6:1;
2140
Alexander Duycka0ca1532016-04-05 09:13:39 -07002141 /* Used in GRE, set in fou/gue_gro_receive */
2142 u8 is_fou:1;
2143
Alexander Duyck15305452016-04-10 21:44:57 -04002144 /* Used to determine if flush_id can be ignored */
2145 u8 is_atomic:1;
2146
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002147 /* Number of gro_receive callbacks this packet already went through */
2148 u8 recursion_counter:4;
2149
2150 /* 1 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002151
Jerry Chubf5a7552014-01-07 10:23:19 -08002152 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2153 __wsum csum;
2154
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002155 /* used in skb_gro_receive() slow path */
2156 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002157};
2158
2159#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002160
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002161#define GRO_RECURSION_LIMIT 15
2162static inline int gro_recursion_inc_test(struct sk_buff *skb)
2163{
2164 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2165}
2166
2167typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2168static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2169 struct sk_buff **head,
2170 struct sk_buff *skb)
2171{
2172 if (unlikely(gro_recursion_inc_test(skb))) {
2173 NAPI_GRO_CB(skb)->flush |= 1;
2174 return NULL;
2175 }
2176
2177 return cb(head, skb);
2178}
2179
2180typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2181 struct sk_buff *);
2182static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2183 struct sock *sk,
2184 struct sk_buff **head,
2185 struct sk_buff *skb)
2186{
2187 if (unlikely(gro_recursion_inc_test(skb))) {
2188 NAPI_GRO_CB(skb)->flush |= 1;
2189 return NULL;
2190 }
2191
2192 return cb(sk, head, skb);
2193}
2194
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002196 __be16 type; /* This is really htons(ether_type). */
2197 struct net_device *dev; /* NULL is wildcarded here */
2198 int (*func) (struct sk_buff *,
2199 struct net_device *,
2200 struct packet_type *,
2201 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002202 bool (*id_match)(struct packet_type *ptype,
2203 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204 void *af_packet_priv;
2205 struct list_head list;
2206};
2207
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002208struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2210 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002212 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002213 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002214};
2215
2216struct packet_offload {
2217 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002218 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002219 struct offload_callbacks callbacks;
2220 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221};
2222
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002223/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002224struct pcpu_sw_netstats {
2225 u64 rx_packets;
2226 u64 rx_bytes;
2227 u64 tx_packets;
2228 u64 tx_bytes;
2229 struct u64_stats_sync syncp;
2230};
2231
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002232#define __netdev_alloc_pcpu_stats(type, gfp) \
2233({ \
2234 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2235 if (pcpu_stats) { \
2236 int __cpu; \
2237 for_each_possible_cpu(__cpu) { \
2238 typeof(type) *stat; \
2239 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2240 u64_stats_init(&stat->syncp); \
2241 } \
2242 } \
2243 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002244})
2245
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002246#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002247 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002248
Jiri Pirko764f5e52015-12-03 12:12:12 +01002249enum netdev_lag_tx_type {
2250 NETDEV_LAG_TX_TYPE_UNKNOWN,
2251 NETDEV_LAG_TX_TYPE_RANDOM,
2252 NETDEV_LAG_TX_TYPE_BROADCAST,
2253 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2254 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2255 NETDEV_LAG_TX_TYPE_HASH,
2256};
2257
2258struct netdev_lag_upper_info {
2259 enum netdev_lag_tx_type tx_type;
2260};
2261
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002262struct netdev_lag_lower_state_info {
2263 u8 link_up : 1,
2264 tx_enabled : 1;
2265};
2266
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267#include <linux/notifier.h>
2268
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002269/* netdevice notifier chain. Please remember to update the rtnetlink
2270 * notification exclusion list in rtnetlink_event() when adding new
2271 * types.
2272 */
2273#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2274#define NETDEV_DOWN 0x0002
2275#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2276 detected a hardware crash and restarted
2277 - we can use this eg to kick tcp sessions
2278 once done */
2279#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2280#define NETDEV_REGISTER 0x0005
2281#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002282#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002283#define NETDEV_CHANGEADDR 0x0008
2284#define NETDEV_GOING_DOWN 0x0009
2285#define NETDEV_CHANGENAME 0x000A
2286#define NETDEV_FEAT_CHANGE 0x000B
2287#define NETDEV_BONDING_FAILOVER 0x000C
2288#define NETDEV_PRE_UP 0x000D
2289#define NETDEV_PRE_TYPE_CHANGE 0x000E
2290#define NETDEV_POST_TYPE_CHANGE 0x000F
2291#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002292#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002293#define NETDEV_RELEASE 0x0012
2294#define NETDEV_NOTIFY_PEERS 0x0013
2295#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002296#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002297#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002298#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002299#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002300#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002301#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002302#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002303#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Jason Wang08294a22016-06-30 14:45:35 +08002304#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002305
Joe Perchesf629d202013-09-26 14:48:15 -07002306int register_netdevice_notifier(struct notifier_block *nb);
2307int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002308
2309struct netdev_notifier_info {
2310 struct net_device *dev;
2311};
2312
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002313struct netdev_notifier_change_info {
2314 struct netdev_notifier_info info; /* must be first */
2315 unsigned int flags_changed;
2316};
2317
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002318struct netdev_notifier_changeupper_info {
2319 struct netdev_notifier_info info; /* must be first */
2320 struct net_device *upper_dev; /* new upper dev */
2321 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002322 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002323 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002324};
2325
Jiri Pirko04d48262015-12-03 12:12:15 +01002326struct netdev_notifier_changelowerstate_info {
2327 struct netdev_notifier_info info; /* must be first */
2328 void *lower_state_info; /* is lower dev state */
2329};
2330
Cong Wang75538c22013-05-29 11:30:50 +08002331static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2332 struct net_device *dev)
2333{
2334 info->dev = dev;
2335}
2336
Jiri Pirko351638e2013-05-28 01:30:21 +00002337static inline struct net_device *
2338netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2339{
2340 return info->dev;
2341}
2342
Joe Perchesf629d202013-09-26 14:48:15 -07002343int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002344
2345
Linus Torvalds1da177e2005-04-16 15:20:36 -07002346extern rwlock_t dev_base_lock; /* Device list lock */
2347
Eric W. Biederman881d9662007-09-17 11:56:21 -07002348#define for_each_netdev(net, d) \
2349 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002350#define for_each_netdev_reverse(net, d) \
2351 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002352#define for_each_netdev_rcu(net, d) \
2353 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002354#define for_each_netdev_safe(net, d, n) \
2355 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2356#define for_each_netdev_continue(net, d) \
2357 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002358#define for_each_netdev_continue_rcu(net, d) \
2359 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002360#define for_each_netdev_in_bond_rcu(bond, slave) \
2361 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002362 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002363#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2364
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002365static inline struct net_device *next_net_device(struct net_device *dev)
2366{
2367 struct list_head *lh;
2368 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002369
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002370 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002371 lh = dev->dev_list.next;
2372 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2373}
2374
Eric Dumazetce81b762009-11-11 17:34:30 +00002375static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2376{
2377 struct list_head *lh;
2378 struct net *net;
2379
2380 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002381 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002382 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2383}
2384
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002385static inline struct net_device *first_net_device(struct net *net)
2386{
2387 return list_empty(&net->dev_base_head) ? NULL :
2388 net_device_entry(net->dev_base_head.next);
2389}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002390
Eric Dumazetccf43432011-01-26 18:08:02 +00002391static inline struct net_device *first_net_device_rcu(struct net *net)
2392{
2393 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2394
2395 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2396}
2397
Joe Perchesf629d202013-09-26 14:48:15 -07002398int netdev_boot_setup_check(struct net_device *dev);
2399unsigned long netdev_boot_base(const char *prefix, int unit);
2400struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2401 const char *hwaddr);
2402struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2403struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2404void dev_add_pack(struct packet_type *pt);
2405void dev_remove_pack(struct packet_type *pt);
2406void __dev_remove_pack(struct packet_type *pt);
2407void dev_add_offload(struct packet_offload *po);
2408void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002409
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002410int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002411int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002412struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2413 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002414struct net_device *dev_get_by_name(struct net *net, const char *name);
2415struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2416struct net_device *__dev_get_by_name(struct net *net, const char *name);
2417int dev_alloc_name(struct net_device *dev, const char *name);
2418int dev_open(struct net_device *dev);
2419int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002420int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002421void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002422int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002423int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002424int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002425int register_netdevice(struct net_device *dev);
2426void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2427void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002428static inline void unregister_netdevice(struct net_device *dev)
2429{
2430 unregister_netdevice_queue(dev, NULL);
2431}
2432
Joe Perchesf629d202013-09-26 14:48:15 -07002433int netdev_refcnt_read(const struct net_device *dev);
2434void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002435void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002436void synchronize_net(void);
2437int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002438
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002439DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002440#define XMIT_RECURSION_LIMIT 10
2441
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002442static inline int dev_recursion_level(void)
2443{
2444 return this_cpu_read(xmit_recursion);
2445}
2446
Joe Perchesf629d202013-09-26 14:48:15 -07002447struct net_device *dev_get_by_index(struct net *net, int ifindex);
2448struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2449struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2450int netdev_get_name(struct net *net, char *name, int ifindex);
2451int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002452int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002453
2454static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2455{
2456 return NAPI_GRO_CB(skb)->data_offset;
2457}
2458
2459static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2460{
2461 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2462}
2463
2464static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2465{
2466 NAPI_GRO_CB(skb)->data_offset += len;
2467}
2468
Herbert Xua5b1cf22009-05-26 18:50:28 +00002469static inline void *skb_gro_header_fast(struct sk_buff *skb,
2470 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002471{
Herbert Xu78a478d2009-05-26 18:50:21 +00002472 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002473}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002474
Herbert Xua5b1cf22009-05-26 18:50:28 +00002475static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2476{
2477 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2478}
2479
Herbert Xu57ea52a2017-01-10 12:24:15 -08002480static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2481{
2482 NAPI_GRO_CB(skb)->frag0 = NULL;
2483 NAPI_GRO_CB(skb)->frag0_len = 0;
2484}
2485
Herbert Xua5b1cf22009-05-26 18:50:28 +00002486static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2487 unsigned int offset)
2488{
Herbert Xu17dd7592011-07-27 06:16:28 -07002489 if (!pskb_may_pull(skb, hlen))
2490 return NULL;
2491
Herbert Xu57ea52a2017-01-10 12:24:15 -08002492 skb_gro_frag0_invalidate(skb);
Herbert Xu17dd7592011-07-27 06:16:28 -07002493 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002494}
2495
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002496static inline void *skb_gro_network_header(struct sk_buff *skb)
2497{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002498 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2499 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002500}
2501
Jerry Chubf5a7552014-01-07 10:23:19 -08002502static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2503 const void *start, unsigned int len)
2504{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002505 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002506 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2507 csum_partial(start, len, 0));
2508}
2509
Tom Herbert573e8fc2014-08-22 13:33:47 -07002510/* GRO checksum functions. These are logical equivalents of the normal
2511 * checksum functions (in skbuff.h) except that they operate on the GRO
2512 * offsets and fields in sk_buff.
2513 */
2514
2515__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2516
Tom Herbert15e23962015-02-10 16:30:31 -08002517static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2518{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002519 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002520}
2521
Tom Herbert573e8fc2014-08-22 13:33:47 -07002522static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2523 bool zero_okay,
2524 __sum16 check)
2525{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002526 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2527 skb_checksum_start_offset(skb) <
2528 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002529 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002530 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002531 (!zero_okay || check));
2532}
2533
2534static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2535 __wsum psum)
2536{
2537 if (NAPI_GRO_CB(skb)->csum_valid &&
2538 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2539 return 0;
2540
2541 NAPI_GRO_CB(skb)->csum = psum;
2542
2543 return __skb_gro_checksum_complete(skb);
2544}
2545
Tom Herbert573e8fc2014-08-22 13:33:47 -07002546static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2547{
Tom Herbert662880f2014-08-27 21:26:56 -07002548 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2549 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2550 NAPI_GRO_CB(skb)->csum_cnt--;
2551 } else {
2552 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2553 * verified a new top level checksum or an encapsulated one
2554 * during GRO. This saves work if we fallback to normal path.
2555 */
2556 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002557 }
2558}
2559
2560#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2561 compute_pseudo) \
2562({ \
2563 __sum16 __ret = 0; \
2564 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2565 __ret = __skb_gro_checksum_validate_complete(skb, \
2566 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002567 if (__ret) \
2568 __skb_mark_checksum_bad(skb); \
2569 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002570 skb_gro_incr_csum_unnecessary(skb); \
2571 __ret; \
2572})
2573
2574#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2575 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2576
2577#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2578 compute_pseudo) \
2579 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2580
2581#define skb_gro_checksum_simple_validate(skb) \
2582 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2583
Tom Herbertd96535a2014-08-31 15:12:42 -07002584static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2585{
2586 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2587 !NAPI_GRO_CB(skb)->csum_valid);
2588}
2589
2590static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2591 __sum16 check, __wsum pseudo)
2592{
2593 NAPI_GRO_CB(skb)->csum = ~pseudo;
2594 NAPI_GRO_CB(skb)->csum_valid = 1;
2595}
2596
2597#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2598do { \
2599 if (__skb_gro_checksum_convert_check(skb)) \
2600 __skb_gro_checksum_convert(skb, check, \
2601 compute_pseudo(skb, proto)); \
2602} while (0)
2603
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002604struct gro_remcsum {
2605 int offset;
2606 __wsum delta;
2607};
2608
2609static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2610{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002611 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002612 grc->delta = 0;
2613}
2614
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002615static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2616 unsigned int off, size_t hdrlen,
2617 int start, int offset,
2618 struct gro_remcsum *grc,
2619 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002620{
2621 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002622 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002623
2624 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2625
Tom Herbert15e23962015-02-10 16:30:31 -08002626 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002627 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2628 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002629 }
2630
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002631 ptr = skb_gro_header_fast(skb, off);
2632 if (skb_gro_header_hard(skb, off + plen)) {
2633 ptr = skb_gro_header_slow(skb, off + plen, off);
2634 if (!ptr)
2635 return NULL;
2636 }
2637
2638 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2639 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002640
2641 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002642 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002643
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002644 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002645 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002646
2647 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002648}
2649
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002650static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2651 struct gro_remcsum *grc)
2652{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002653 void *ptr;
2654 size_t plen = grc->offset + sizeof(u16);
2655
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002656 if (!grc->delta)
2657 return;
2658
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002659 ptr = skb_gro_header_fast(skb, grc->offset);
2660 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2661 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2662 if (!ptr)
2663 return;
2664 }
2665
2666 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002667}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002668
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002669static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2670 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002671 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002672 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002673{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002674 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002675 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002676
2677 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002678}
2679
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002680static inline int dev_parse_header(const struct sk_buff *skb,
2681 unsigned char *haddr)
2682{
2683 const struct net_device *dev = skb->dev;
2684
Patrick McHardy1b833362007-10-18 05:09:28 -07002685 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002686 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002687 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002688}
2689
Willem de Bruijn2793a232016-03-09 21:58:32 -05002690/* ll_header must have at least hard_header_len allocated */
2691static inline bool dev_validate_header(const struct net_device *dev,
2692 char *ll_header, int len)
2693{
2694 if (likely(len >= dev->hard_header_len))
2695 return true;
2696
2697 if (capable(CAP_SYS_RAWIO)) {
2698 memset(ll_header + len, 0, dev->hard_header_len - len);
2699 return true;
2700 }
2701
2702 if (dev->header_ops && dev->header_ops->validate)
2703 return dev->header_ops->validate(ll_header, len);
2704
2705 return false;
2706}
2707
Linus Torvalds1da177e2005-04-16 15:20:36 -07002708typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002709int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002710static inline int unregister_gifconf(unsigned int family)
2711{
2712 return register_gifconf(family, NULL);
2713}
2714
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002715#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002716#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002717struct sd_flow_limit {
2718 u64 count;
2719 unsigned int num_buckets;
2720 unsigned int history_head;
2721 u16 history[FLOW_LIMIT_HISTORY];
2722 u8 buckets[];
2723};
2724
2725extern int netdev_flow_limit_table_len;
2726#endif /* CONFIG_NET_FLOW_LIMIT */
2727
Linus Torvalds1da177e2005-04-16 15:20:36 -07002728/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002729 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002730 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002731struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002732 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002733 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002734
Changli Gaodee42872010-05-02 05:42:16 +00002735 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002736 unsigned int processed;
2737 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002738 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002739#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002740 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002741#endif
2742#ifdef CONFIG_NET_FLOW_LIMIT
2743 struct sd_flow_limit __rcu *flow_limit;
2744#endif
2745 struct Qdisc *output_queue;
2746 struct Qdisc **output_queue_tailp;
2747 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002748
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002749#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002750 /* input_queue_head should be written by cpu owning this struct,
2751 * and only read by other cpus. Worth using a cache line.
2752 */
2753 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2754
2755 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002756 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002757 struct softnet_data *rps_ipi_next;
2758 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002759 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002760#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002761 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002762 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002763 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002764
Linus Torvalds1da177e2005-04-16 15:20:36 -07002765};
2766
Tom Herbert76cc8b12010-05-20 18:37:59 +00002767static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002768{
2769#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002770 sd->input_queue_head++;
2771#endif
2772}
2773
2774static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2775 unsigned int *qtail)
2776{
2777#ifdef CONFIG_RPS
2778 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002779#endif
2780}
2781
Tom Herbert0a9627f2010-03-16 08:03:29 +00002782DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002783
Joe Perchesf629d202013-09-26 14:48:15 -07002784void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002785void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002786
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002787static inline void netif_tx_schedule_all(struct net_device *dev)
2788{
2789 unsigned int i;
2790
2791 for (i = 0; i < dev->num_tx_queues; i++)
2792 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2793}
2794
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002795static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002796{
Tom Herbert734664982011-11-28 16:32:44 +00002797 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002798}
2799
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002800/**
2801 * netif_start_queue - allow transmit
2802 * @dev: network device
2803 *
2804 * Allow upper layers to call the device hard_start_xmit routine.
2805 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002806static inline void netif_start_queue(struct net_device *dev)
2807{
David S. Millere8a04642008-07-17 00:34:19 -07002808 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002809}
2810
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002811static inline void netif_tx_start_all_queues(struct net_device *dev)
2812{
2813 unsigned int i;
2814
2815 for (i = 0; i < dev->num_tx_queues; i++) {
2816 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2817 netif_tx_start_queue(txq);
2818 }
2819}
2820
John Fastabend46e5da40a2014-09-12 20:04:52 -07002821void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002822
Dave Jonesd29f7492008-07-22 14:09:06 -07002823/**
2824 * netif_wake_queue - restart transmit
2825 * @dev: network device
2826 *
2827 * Allow upper layers to call the device hard_start_xmit routine.
2828 * Used for flow control when transmit resources are available.
2829 */
David S. Miller79d16382008-07-08 23:14:46 -07002830static inline void netif_wake_queue(struct net_device *dev)
2831{
David S. Millere8a04642008-07-17 00:34:19 -07002832 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002833}
2834
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002835static inline void netif_tx_wake_all_queues(struct net_device *dev)
2836{
2837 unsigned int i;
2838
2839 for (i = 0; i < dev->num_tx_queues; i++) {
2840 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2841 netif_tx_wake_queue(txq);
2842 }
2843}
2844
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002845static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002846{
Tom Herbert734664982011-11-28 16:32:44 +00002847 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002848}
2849
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002850/**
2851 * netif_stop_queue - stop transmitted packets
2852 * @dev: network device
2853 *
2854 * Stop upper layers calling the device hard_start_xmit routine.
2855 * Used for flow control when transmit resources are unavailable.
2856 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002857static inline void netif_stop_queue(struct net_device *dev)
2858{
David S. Millere8a04642008-07-17 00:34:19 -07002859 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002860}
2861
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002862void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002863
David S. Miller4d295152012-03-07 21:02:35 -05002864static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002865{
Tom Herbert734664982011-11-28 16:32:44 +00002866 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002867}
2868
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002869/**
2870 * netif_queue_stopped - test if transmit queue is flowblocked
2871 * @dev: network device
2872 *
2873 * Test if transmit queue on device is currently unable to send.
2874 */
David S. Miller4d295152012-03-07 21:02:35 -05002875static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002876{
David S. Millere8a04642008-07-17 00:34:19 -07002877 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002878}
2879
David S. Miller4d295152012-03-07 21:02:35 -05002880static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002881{
Tom Herbert734664982011-11-28 16:32:44 +00002882 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2883}
2884
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002885static inline bool
2886netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002887{
2888 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2889}
2890
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002891static inline bool
2892netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2893{
2894 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2895}
2896
Eric Dumazet53511452014-10-08 08:19:27 -07002897/**
2898 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2899 * @dev_queue: pointer to transmit queue
2900 *
2901 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002902 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002903 */
2904static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2905{
2906#ifdef CONFIG_BQL
2907 prefetchw(&dev_queue->dql.num_queued);
2908#endif
2909}
2910
2911/**
2912 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2913 * @dev_queue: pointer to transmit queue
2914 *
2915 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002916 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002917 */
2918static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2919{
2920#ifdef CONFIG_BQL
2921 prefetchw(&dev_queue->dql.limit);
2922#endif
2923}
2924
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002925static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2926 unsigned int bytes)
2927{
Tom Herbert114cf582011-11-28 16:33:09 +00002928#ifdef CONFIG_BQL
2929 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002930
2931 if (likely(dql_avail(&dev_queue->dql) >= 0))
2932 return;
2933
2934 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2935
2936 /*
2937 * The XOFF flag must be set before checking the dql_avail below,
2938 * because in netdev_tx_completed_queue we update the dql_completed
2939 * before checking the XOFF flag.
2940 */
2941 smp_mb();
2942
2943 /* check again in case another CPU has just made room avail */
2944 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2945 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002946#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002947}
2948
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002949/**
2950 * netdev_sent_queue - report the number of bytes queued to hardware
2951 * @dev: network device
2952 * @bytes: number of bytes queued to the hardware device queue
2953 *
2954 * Report the number of bytes queued for sending/completion to the network
2955 * device hardware queue. @bytes should be a good approximation and should
2956 * exactly match netdev_completed_queue() @bytes
2957 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002958static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2959{
2960 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2961}
2962
2963static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002964 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002965{
Tom Herbert114cf582011-11-28 16:33:09 +00002966#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002967 if (unlikely(!bytes))
2968 return;
2969
2970 dql_completed(&dev_queue->dql, bytes);
2971
2972 /*
2973 * Without the memory barrier there is a small possiblity that
2974 * netdev_tx_sent_queue will miss the update and cause the queue to
2975 * be stopped forever
2976 */
2977 smp_mb();
2978
2979 if (dql_avail(&dev_queue->dql) < 0)
2980 return;
2981
2982 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2983 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00002984#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002985}
2986
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002987/**
2988 * netdev_completed_queue - report bytes and packets completed by device
2989 * @dev: network device
2990 * @pkts: actual number of packets sent over the medium
2991 * @bytes: actual number of bytes sent over the medium
2992 *
2993 * Report the number of bytes and packets transmitted by the network device
2994 * hardware queue over the physical medium, @bytes must exactly match the
2995 * @bytes amount passed to netdev_sent_queue()
2996 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002997static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00002998 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002999{
3000 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3001}
3002
3003static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3004{
Tom Herbert114cf582011-11-28 16:33:09 +00003005#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003006 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003007 dql_reset(&q->dql);
3008#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003009}
3010
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003011/**
3012 * netdev_reset_queue - reset the packets and bytes count of a network device
3013 * @dev_queue: network device
3014 *
3015 * Reset the bytes and packet count of a network device and clear the
3016 * software flow control OFF bit for this network device
3017 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003018static inline void netdev_reset_queue(struct net_device *dev_queue)
3019{
3020 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003021}
3022
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003023/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003024 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3025 * @dev: network device
3026 * @queue_index: given tx queue index
3027 *
3028 * Returns 0 if given tx queue index >= number of device tx queues,
3029 * otherwise returns the originally passed tx queue index.
3030 */
3031static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3032{
3033 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3034 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3035 dev->name, queue_index,
3036 dev->real_num_tx_queues);
3037 return 0;
3038 }
3039
3040 return queue_index;
3041}
3042
3043/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003044 * netif_running - test if up
3045 * @dev: network device
3046 *
3047 * Test if the device has been brought up.
3048 */
David S. Miller4d295152012-03-07 21:02:35 -05003049static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003050{
3051 return test_bit(__LINK_STATE_START, &dev->state);
3052}
3053
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003054/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003055 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003056 * stop, and a check if it's stopped. All other device management is
3057 * done at the overall netdevice level.
3058 * Also test the device if we're multiqueue.
3059 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003060
3061/**
3062 * netif_start_subqueue - allow sending packets on subqueue
3063 * @dev: network device
3064 * @queue_index: sub queue index
3065 *
3066 * Start individual transmit queue of a device with multiple transmit queues.
3067 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003068static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3069{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003070 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003071
3072 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003073}
3074
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003075/**
3076 * netif_stop_subqueue - stop sending packets on subqueue
3077 * @dev: network device
3078 * @queue_index: sub queue index
3079 *
3080 * Stop individual transmit queue of a device with multiple transmit queues.
3081 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003082static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3083{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003084 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003085 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003086}
3087
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003088/**
3089 * netif_subqueue_stopped - test status of subqueue
3090 * @dev: network device
3091 * @queue_index: sub queue index
3092 *
3093 * Check individual transmit queue of a device with multiple transmit queues.
3094 */
David S. Miller4d295152012-03-07 21:02:35 -05003095static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3096 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003097{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003098 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003099
3100 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003101}
3102
David S. Miller4d295152012-03-07 21:02:35 -05003103static inline bool netif_subqueue_stopped(const struct net_device *dev,
3104 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003105{
3106 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3107}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003108
Florian Fainelli738b35c2017-01-11 21:13:02 -08003109/**
3110 * netif_wake_subqueue - allow sending packets on subqueue
3111 * @dev: network device
3112 * @queue_index: sub queue index
3113 *
3114 * Resume individual transmit queue of a device with multiple transmit queues.
3115 */
3116static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3117{
3118 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3119
3120 netif_tx_wake_queue(txq);
3121}
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003122
Alexander Duyck537c00d2013-01-10 08:57:02 +00003123#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003124int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003125 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003126#else
3127static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003128 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003129 u16 index)
3130{
3131 return 0;
3132}
3133#endif
3134
Jiri Pirko5605c762015-05-12 14:56:12 +02003135u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3136 unsigned int num_tx_queues);
3137
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003138/*
3139 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3140 * as a distribution range limit for the returned value.
3141 */
3142static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003143 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003144{
3145 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3146}
3147
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003148/**
3149 * netif_is_multiqueue - test if device has multiple transmit queues
3150 * @dev: network device
3151 *
3152 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003153 */
David S. Miller4d295152012-03-07 21:02:35 -05003154static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003155{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003156 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003157}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003158
Joe Perchesf629d202013-09-26 14:48:15 -07003159int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003160
Michael Daltona953be52014-01-16 22:23:28 -08003161#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003162int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003163#else
3164static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3165 unsigned int rxq)
3166{
3167 return 0;
3168}
3169#endif
3170
Michael Daltona953be52014-01-16 22:23:28 -08003171#ifdef CONFIG_SYSFS
3172static inline unsigned int get_netdev_rx_queue_index(
3173 struct netdev_rx_queue *queue)
3174{
3175 struct net_device *dev = queue->dev;
3176 int index = queue - dev->_rx;
3177
3178 BUG_ON(index >= dev->num_rx_queues);
3179 return index;
3180}
3181#endif
3182
Yuval Mintz16917b82012-07-01 03:18:50 +00003183#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003184int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003185
Eric Dumazete6247022013-12-05 04:45:08 -08003186enum skb_free_reason {
3187 SKB_REASON_CONSUMED,
3188 SKB_REASON_DROPPED,
3189};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003190
Eric Dumazete6247022013-12-05 04:45:08 -08003191void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3192void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3193
3194/*
3195 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3196 * interrupt context or with hardware interrupts being disabled.
3197 * (in_irq() || irqs_disabled())
3198 *
3199 * We provide four helpers that can be used in following contexts :
3200 *
3201 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3202 * replacing kfree_skb(skb)
3203 *
3204 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3205 * Typically used in place of consume_skb(skb) in TX completion path
3206 *
3207 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3208 * replacing kfree_skb(skb)
3209 *
3210 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3211 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003212 */
Eric Dumazete6247022013-12-05 04:45:08 -08003213static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3214{
3215 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3216}
3217
3218static inline void dev_consume_skb_irq(struct sk_buff *skb)
3219{
3220 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3221}
3222
3223static inline void dev_kfree_skb_any(struct sk_buff *skb)
3224{
3225 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3226}
3227
3228static inline void dev_consume_skb_any(struct sk_buff *skb)
3229{
3230 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3231}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003232
Joe Perchesf629d202013-09-26 14:48:15 -07003233int netif_rx(struct sk_buff *skb);
3234int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003235int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003236gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3237void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3238struct sk_buff *napi_get_frags(struct napi_struct *napi);
3239gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003240struct packet_offload *gro_find_receive_by_type(__be16 type);
3241struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003242
3243static inline void napi_free_frags(struct napi_struct *napi)
3244{
3245 kfree_skb(napi->skb);
3246 napi->skb = NULL;
3247}
3248
Mahesh Bandewar24b27fc2016-09-01 22:18:34 -07003249bool netdev_is_rx_handler_busy(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003250int netdev_rx_handler_register(struct net_device *dev,
3251 rx_handler_func_t *rx_handler,
3252 void *rx_handler_data);
3253void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003254
Joe Perchesf629d202013-09-26 14:48:15 -07003255bool dev_valid_name(const char *name);
3256int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3257int dev_ethtool(struct net *net, struct ifreq *);
3258unsigned int dev_get_flags(const struct net_device *);
3259int __dev_change_flags(struct net_device *, unsigned int flags);
3260int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003261void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3262 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003263int dev_change_name(struct net_device *, const char *);
3264int dev_set_alias(struct net_device *, const char *, size_t);
3265int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3266int dev_set_mtu(struct net_device *, int);
3267void dev_set_group(struct net_device *, int);
3268int dev_set_mac_address(struct net_device *, struct sockaddr *);
3269int dev_change_carrier(struct net_device *, bool new_carrier);
3270int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003271 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003272int dev_get_phys_port_name(struct net_device *dev,
3273 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003274int dev_change_proto_down(struct net_device *dev, bool proto_down);
Daniel Borkmann85de8572016-11-28 23:16:54 +01003275int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003276struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003277struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3278 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08003279int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003280int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003281bool is_skb_forwardable(const struct net_device *dev,
3282 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003283
Martin KaFai Lau4e3264d2016-11-09 15:36:33 -08003284static __always_inline int ____dev_forward_skb(struct net_device *dev,
3285 struct sk_buff *skb)
3286{
3287 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3288 unlikely(!is_skb_forwardable(dev, skb))) {
3289 atomic_long_inc(&dev->rx_dropped);
3290 kfree_skb(skb);
3291 return NET_RX_DROP;
3292 }
3293
3294 skb_scrub_packet(skb, true);
3295 skb->priority = 0;
3296 return 0;
3297}
3298
David Ahern74b20582016-05-10 11:19:50 -07003299void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3300
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003301extern int netdev_budget;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003302
3303/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003304void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003305
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003306/**
3307 * dev_put - release reference to device
3308 * @dev: network device
3309 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003310 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003311 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003312static inline void dev_put(struct net_device *dev)
3313{
Christoph Lameter933393f2011-12-22 11:58:51 -06003314 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003315}
3316
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003317/**
3318 * dev_hold - get reference to device
3319 * @dev: network device
3320 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003321 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003322 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003323static inline void dev_hold(struct net_device *dev)
3324{
Christoph Lameter933393f2011-12-22 11:58:51 -06003325 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003326}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003327
3328/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3329 * and _off may be called from IRQ context, but it is caller
3330 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003331 *
3332 * The name carrier is inappropriate, these functions should really be
3333 * called netif_lowerlayer_*() because they represent the state of any
3334 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003335 */
3336
Joe Perchesf629d202013-09-26 14:48:15 -07003337void linkwatch_init_dev(struct net_device *dev);
3338void linkwatch_fire_event(struct net_device *dev);
3339void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003340
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003341/**
3342 * netif_carrier_ok - test if carrier present
3343 * @dev: network device
3344 *
3345 * Check if carrier is present on device
3346 */
David S. Miller4d295152012-03-07 21:02:35 -05003347static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003348{
3349 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3350}
3351
Joe Perchesf629d202013-09-26 14:48:15 -07003352unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003353
Joe Perchesf629d202013-09-26 14:48:15 -07003354void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003355
Joe Perchesf629d202013-09-26 14:48:15 -07003356void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003357
Joe Perchesf629d202013-09-26 14:48:15 -07003358void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003359
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003360/**
3361 * netif_dormant_on - mark device as dormant.
3362 * @dev: network device
3363 *
3364 * Mark device as dormant (as per RFC2863).
3365 *
3366 * The dormant state indicates that the relevant interface is not
3367 * actually in a condition to pass packets (i.e., it is not 'up') but is
3368 * in a "pending" state, waiting for some external event. For "on-
3369 * demand" interfaces, this new state identifies the situation where the
3370 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003371 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003372static inline void netif_dormant_on(struct net_device *dev)
3373{
3374 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3375 linkwatch_fire_event(dev);
3376}
3377
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003378/**
3379 * netif_dormant_off - set device as not dormant.
3380 * @dev: network device
3381 *
3382 * Device is not in dormant state.
3383 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003384static inline void netif_dormant_off(struct net_device *dev)
3385{
3386 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3387 linkwatch_fire_event(dev);
3388}
3389
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003390/**
3391 * netif_dormant - test if carrier present
3392 * @dev: network device
3393 *
3394 * Check if carrier is present on device
3395 */
David S. Miller4d295152012-03-07 21:02:35 -05003396static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003397{
3398 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3399}
3400
3401
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003402/**
3403 * netif_oper_up - test if device is operational
3404 * @dev: network device
3405 *
3406 * Check if carrier is operational
3407 */
David S. Miller4d295152012-03-07 21:02:35 -05003408static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003409{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003410 return (dev->operstate == IF_OPER_UP ||
3411 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3412}
3413
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003414/**
3415 * netif_device_present - is device available or removed
3416 * @dev: network device
3417 *
3418 * Check if device has not been removed from system.
3419 */
David S. Miller4d295152012-03-07 21:02:35 -05003420static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003421{
3422 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3423}
3424
Joe Perchesf629d202013-09-26 14:48:15 -07003425void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003426
Joe Perchesf629d202013-09-26 14:48:15 -07003427void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003428
3429/*
3430 * Network interface message level settings
3431 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003432
3433enum {
3434 NETIF_MSG_DRV = 0x0001,
3435 NETIF_MSG_PROBE = 0x0002,
3436 NETIF_MSG_LINK = 0x0004,
3437 NETIF_MSG_TIMER = 0x0008,
3438 NETIF_MSG_IFDOWN = 0x0010,
3439 NETIF_MSG_IFUP = 0x0020,
3440 NETIF_MSG_RX_ERR = 0x0040,
3441 NETIF_MSG_TX_ERR = 0x0080,
3442 NETIF_MSG_TX_QUEUED = 0x0100,
3443 NETIF_MSG_INTR = 0x0200,
3444 NETIF_MSG_TX_DONE = 0x0400,
3445 NETIF_MSG_RX_STATUS = 0x0800,
3446 NETIF_MSG_PKTDATA = 0x1000,
3447 NETIF_MSG_HW = 0x2000,
3448 NETIF_MSG_WOL = 0x4000,
3449};
3450
3451#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3452#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3453#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3454#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3455#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3456#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3457#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3458#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3459#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3460#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3461#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3462#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3463#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3464#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3465#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3466
3467static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3468{
3469 /* use default */
3470 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3471 return default_msg_enable_bits;
3472 if (debug_value == 0) /* no output */
3473 return 0;
3474 /* set low N bits */
3475 return (1 << debug_value) - 1;
3476}
3477
David S. Millerc773e842008-07-08 23:13:53 -07003478static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003479{
David S. Millerc773e842008-07-08 23:13:53 -07003480 spin_lock(&txq->_xmit_lock);
3481 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003482}
3483
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003484static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3485{
3486 __acquire(&txq->_xmit_lock);
3487 return true;
3488}
3489
3490static inline void __netif_tx_release(struct netdev_queue *txq)
3491{
3492 __release(&txq->_xmit_lock);
3493}
3494
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003495static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3496{
3497 spin_lock_bh(&txq->_xmit_lock);
3498 txq->xmit_lock_owner = smp_processor_id();
3499}
3500
David S. Miller4d295152012-03-07 21:02:35 -05003501static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003502{
David S. Miller4d295152012-03-07 21:02:35 -05003503 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003504 if (likely(ok))
3505 txq->xmit_lock_owner = smp_processor_id();
3506 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003507}
3508
David S. Millerc773e842008-07-08 23:13:53 -07003509static inline void __netif_tx_unlock(struct netdev_queue *txq)
3510{
3511 txq->xmit_lock_owner = -1;
3512 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003513}
3514
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003515static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3516{
3517 txq->xmit_lock_owner = -1;
3518 spin_unlock_bh(&txq->_xmit_lock);
3519}
3520
Eric Dumazet08baf562009-05-25 22:58:01 -07003521static inline void txq_trans_update(struct netdev_queue *txq)
3522{
3523 if (txq->xmit_lock_owner != -1)
3524 txq->trans_start = jiffies;
3525}
3526
Florian Westphalba162f82016-05-03 16:31:00 +02003527/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3528static inline void netif_trans_update(struct net_device *dev)
3529{
Florian Westphal9b366272016-05-03 16:33:14 +02003530 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3531
3532 if (txq->trans_start != jiffies)
3533 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003534}
3535
David S. Millerc3f26a22008-07-31 16:58:50 -07003536/**
3537 * netif_tx_lock - grab network device transmit lock
3538 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003539 *
3540 * Get network device transmit lock
3541 */
3542static inline void netif_tx_lock(struct net_device *dev)
3543{
3544 unsigned int i;
3545 int cpu;
3546
3547 spin_lock(&dev->tx_global_lock);
3548 cpu = smp_processor_id();
3549 for (i = 0; i < dev->num_tx_queues; i++) {
3550 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3551
3552 /* We are the only thread of execution doing a
3553 * freeze, but we have to grab the _xmit_lock in
3554 * order to synchronize with threads which are in
3555 * the ->hard_start_xmit() handler and already
3556 * checked the frozen bit.
3557 */
3558 __netif_tx_lock(txq, cpu);
3559 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3560 __netif_tx_unlock(txq);
3561 }
3562}
3563
3564static inline void netif_tx_lock_bh(struct net_device *dev)
3565{
3566 local_bh_disable();
3567 netif_tx_lock(dev);
3568}
3569
Herbert Xu932ff272006-06-09 12:20:56 -07003570static inline void netif_tx_unlock(struct net_device *dev)
3571{
David S. Millere8a04642008-07-17 00:34:19 -07003572 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003573
David S. Millere8a04642008-07-17 00:34:19 -07003574 for (i = 0; i < dev->num_tx_queues; i++) {
3575 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003576
David S. Millerc3f26a22008-07-31 16:58:50 -07003577 /* No need to grab the _xmit_lock here. If the
3578 * queue is not stopped for another reason, we
3579 * force a schedule.
3580 */
3581 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003582 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003583 }
3584 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003585}
3586
3587static inline void netif_tx_unlock_bh(struct net_device *dev)
3588{
David S. Millere8a04642008-07-17 00:34:19 -07003589 netif_tx_unlock(dev);
3590 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003591}
3592
David S. Millerc773e842008-07-08 23:13:53 -07003593#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003594 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003595 __netif_tx_lock(txq, cpu); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003596 } else { \
3597 __netif_tx_acquire(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003598 } \
3599}
3600
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003601#define HARD_TX_TRYLOCK(dev, txq) \
3602 (((dev->features & NETIF_F_LLTX) == 0) ? \
3603 __netif_tx_trylock(txq) : \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003604 __netif_tx_acquire(txq))
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003605
David S. Millerc773e842008-07-08 23:13:53 -07003606#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003607 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003608 __netif_tx_unlock(txq); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003609 } else { \
3610 __netif_tx_release(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003611 } \
3612}
3613
Linus Torvalds1da177e2005-04-16 15:20:36 -07003614static inline void netif_tx_disable(struct net_device *dev)
3615{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003616 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003617 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003618
David S. Millerc3f26a22008-07-31 16:58:50 -07003619 local_bh_disable();
3620 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003621 for (i = 0; i < dev->num_tx_queues; i++) {
3622 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003623
3624 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003625 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003626 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003627 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003628 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003629}
3630
David S. Millere308a5d2008-07-15 00:13:44 -07003631static inline void netif_addr_lock(struct net_device *dev)
3632{
3633 spin_lock(&dev->addr_list_lock);
3634}
3635
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003636static inline void netif_addr_lock_nested(struct net_device *dev)
3637{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003638 int subclass = SINGLE_DEPTH_NESTING;
3639
3640 if (dev->netdev_ops->ndo_get_lock_subclass)
3641 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3642
3643 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003644}
3645
David S. Millere308a5d2008-07-15 00:13:44 -07003646static inline void netif_addr_lock_bh(struct net_device *dev)
3647{
3648 spin_lock_bh(&dev->addr_list_lock);
3649}
3650
3651static inline void netif_addr_unlock(struct net_device *dev)
3652{
3653 spin_unlock(&dev->addr_list_lock);
3654}
3655
3656static inline void netif_addr_unlock_bh(struct net_device *dev)
3657{
3658 spin_unlock_bh(&dev->addr_list_lock);
3659}
3660
Jiri Pirkof001fde2009-05-05 02:48:28 +00003661/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003662 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003663 * rcu_read_lock held.
3664 */
3665#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003666 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003667
Linus Torvalds1da177e2005-04-16 15:20:36 -07003668/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3669
Joe Perchesf629d202013-09-26 14:48:15 -07003670void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003671
3672/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003673struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003674 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003675 void (*setup)(struct net_device *),
3676 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003677#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3678 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003679
Tom Gundersenc835a672014-07-14 16:37:24 +02003680#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3681 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3682 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003683
Joe Perchesf629d202013-09-26 14:48:15 -07003684int register_netdev(struct net_device *dev);
3685void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003686
Jiri Pirko22bedad32010-04-01 21:22:57 +00003687/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003688int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3689 struct netdev_hw_addr_list *from_list, int addr_len);
3690void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3691 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003692int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3693 struct net_device *dev,
3694 int (*sync)(struct net_device *, const unsigned char *),
3695 int (*unsync)(struct net_device *,
3696 const unsigned char *));
3697void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3698 struct net_device *dev,
3699 int (*unsync)(struct net_device *,
3700 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003701void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003702
Jiri Pirkof001fde2009-05-05 02:48:28 +00003703/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003704int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3705 unsigned char addr_type);
3706int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3707 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003708void dev_addr_flush(struct net_device *dev);
3709int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003710
3711/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003712int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3713int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3714int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3715int dev_uc_sync(struct net_device *to, struct net_device *from);
3716int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3717void dev_uc_unsync(struct net_device *to, struct net_device *from);
3718void dev_uc_flush(struct net_device *dev);
3719void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003720
Alexander Duyck670e5b82014-05-28 18:44:46 -07003721/**
3722 * __dev_uc_sync - Synchonize device's unicast list
3723 * @dev: device to sync
3724 * @sync: function to call if address should be added
3725 * @unsync: function to call if address should be removed
3726 *
3727 * Add newly added addresses to the interface, and release
3728 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003729 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003730static inline int __dev_uc_sync(struct net_device *dev,
3731 int (*sync)(struct net_device *,
3732 const unsigned char *),
3733 int (*unsync)(struct net_device *,
3734 const unsigned char *))
3735{
3736 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3737}
3738
3739/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003740 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003741 * @dev: device to sync
3742 * @unsync: function to call if address should be removed
3743 *
3744 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003745 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003746static inline void __dev_uc_unsync(struct net_device *dev,
3747 int (*unsync)(struct net_device *,
3748 const unsigned char *))
3749{
3750 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3751}
3752
Jiri Pirko22bedad32010-04-01 21:22:57 +00003753/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003754int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3755int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3756int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3757int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3758int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3759int dev_mc_sync(struct net_device *to, struct net_device *from);
3760int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3761void dev_mc_unsync(struct net_device *to, struct net_device *from);
3762void dev_mc_flush(struct net_device *dev);
3763void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003764
Alexander Duyck670e5b82014-05-28 18:44:46 -07003765/**
3766 * __dev_mc_sync - Synchonize device's multicast list
3767 * @dev: device to sync
3768 * @sync: function to call if address should be added
3769 * @unsync: function to call if address should be removed
3770 *
3771 * Add newly added addresses to the interface, and release
3772 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003773 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003774static inline int __dev_mc_sync(struct net_device *dev,
3775 int (*sync)(struct net_device *,
3776 const unsigned char *),
3777 int (*unsync)(struct net_device *,
3778 const unsigned char *))
3779{
3780 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3781}
3782
3783/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003784 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003785 * @dev: device to sync
3786 * @unsync: function to call if address should be removed
3787 *
3788 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003789 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003790static inline void __dev_mc_unsync(struct net_device *dev,
3791 int (*unsync)(struct net_device *,
3792 const unsigned char *))
3793{
3794 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3795}
3796
Linus Torvalds1da177e2005-04-16 15:20:36 -07003797/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003798void dev_set_rx_mode(struct net_device *dev);
3799void __dev_set_rx_mode(struct net_device *dev);
3800int dev_set_promiscuity(struct net_device *dev, int inc);
3801int dev_set_allmulti(struct net_device *dev, int inc);
3802void netdev_state_change(struct net_device *dev);
3803void netdev_notify_peers(struct net_device *dev);
3804void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003805/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003806void dev_load(struct net *net, const char *name);
3807struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3808 struct rtnl_link_stats64 *storage);
3809void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3810 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003811
3812extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003813extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003814extern int weight_p;
Matthias Tafelmeier3d48b532016-12-29 21:37:21 +01003815extern int dev_weight_rx_bias;
3816extern int dev_weight_tx_bias;
3817extern int dev_rx_weight;
3818extern int dev_tx_weight;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003819
Joe Perchesf629d202013-09-26 14:48:15 -07003820bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003821struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3822 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003823struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3824 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003825
3826/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003827#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3828 for (iter = &(dev)->adj_list.upper, \
3829 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3830 updev; \
3831 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3832
David Ahern1a3f0602016-10-17 19:15:44 -07003833int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3834 int (*fn)(struct net_device *upper_dev,
3835 void *data),
3836 void *data);
3837
3838bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3839 struct net_device *upper_dev);
3840
Joe Perchesf629d202013-09-26 14:48:15 -07003841void *netdev_lower_get_next_private(struct net_device *dev,
3842 struct list_head **iter);
3843void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3844 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003845
3846#define netdev_for_each_lower_private(dev, priv, iter) \
3847 for (iter = (dev)->adj_list.lower.next, \
3848 priv = netdev_lower_get_next_private(dev, &(iter)); \
3849 priv; \
3850 priv = netdev_lower_get_next_private(dev, &(iter)))
3851
3852#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3853 for (iter = &(dev)->adj_list.lower, \
3854 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3855 priv; \
3856 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3857
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003858void *netdev_lower_get_next(struct net_device *dev,
3859 struct list_head **iter);
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003860
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003861#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003862 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003863 ldev = netdev_lower_get_next(dev, &(iter)); \
3864 ldev; \
3865 ldev = netdev_lower_get_next(dev, &(iter)))
3866
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003867struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3868 struct list_head **iter);
3869struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3870 struct list_head **iter);
3871
David Ahern1a3f0602016-10-17 19:15:44 -07003872int netdev_walk_all_lower_dev(struct net_device *dev,
3873 int (*fn)(struct net_device *lower_dev,
3874 void *data),
3875 void *data);
3876int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3877 int (*fn)(struct net_device *lower_dev,
3878 void *data),
3879 void *data);
3880
Joe Perchesf629d202013-09-26 14:48:15 -07003881void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003882void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003883struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3884struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3885int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3886int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003887 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003888 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003889void netdev_upper_dev_unlink(struct net_device *dev,
3890 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003891void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003892void *netdev_lower_dev_get_private(struct net_device *dev,
3893 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003894void netdev_lower_state_changed(struct net_device *lower_dev,
3895 void *lower_state_info);
Jiri Pirko18bfb922016-07-05 11:27:38 +02003896int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3897 struct neighbour *n);
3898void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3899 struct neighbour *n);
Eric Dumazet960fb622014-11-16 06:23:05 -08003900
3901/* RSS keys are 40 or 52 bytes long */
3902#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003903extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003904void netdev_rss_key_fill(void *buffer, size_t len);
3905
Sabrina Dubroca952fcfd2016-08-12 16:10:33 +02003906int dev_get_nest_level(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003907int skb_checksum_help(struct sk_buff *skb);
3908struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3909 netdev_features_t features, bool tx_path);
3910struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3911 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003912
Moni Shoua61bd3852015-02-03 16:48:29 +02003913struct netdev_bonding_info {
3914 ifslave slave;
3915 ifbond master;
3916};
3917
3918struct netdev_notifier_bonding_info {
3919 struct netdev_notifier_info info; /* must be first */
3920 struct netdev_bonding_info bonding_info;
3921};
3922
3923void netdev_bonding_info_change(struct net_device *dev,
3924 struct netdev_bonding_info *bonding_info);
3925
Cong Wang12b00042013-02-05 16:36:38 +00003926static inline
3927struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3928{
3929 return __skb_gso_segment(skb, features, true);
3930}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003931__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003932
3933static inline bool can_checksum_protocol(netdev_features_t features,
3934 __be16 protocol)
3935{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003936 if (protocol == htons(ETH_P_FCOE))
3937 return !!(features & NETIF_F_FCOE_CRC);
3938
3939 /* Assume this is an IP checksum (not SCTP CRC) */
3940
3941 if (features & NETIF_F_HW_CSUM) {
3942 /* Can checksum everything */
3943 return true;
3944 }
3945
3946 switch (protocol) {
3947 case htons(ETH_P_IP):
3948 return !!(features & NETIF_F_IP_CSUM);
3949 case htons(ETH_P_IPV6):
3950 return !!(features & NETIF_F_IPV6_CSUM);
3951 default:
3952 return false;
3953 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003954}
Cong Wang12b00042013-02-05 16:36:38 +00003955
Herbert Xufb286bb2005-11-10 13:01:24 -08003956#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003957void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003958#else
3959static inline void netdev_rx_csum_fault(struct net_device *dev)
3960{
3961}
3962#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003963/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003964void net_enable_timestamp(void);
3965void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003966
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003967#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003968int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003969#else
3970#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003971#endif
3972
David S. Miller47982482014-08-22 16:21:53 -07003973static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003974 struct sk_buff *skb, struct net_device *dev,
3975 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003976{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003977 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003978 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003979}
3980
David S. Miller10b3ad82014-08-29 21:07:24 -07003981static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003982 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07003983{
3984 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07003985 int rc;
David S. Miller47982482014-08-22 16:21:53 -07003986
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003987 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07003988 if (rc == NETDEV_TX_OK)
3989 txq_trans_update(txq);
3990
3991 return rc;
David S. Miller47982482014-08-22 16:21:53 -07003992}
3993
Linus Torvalds42a2d922013-11-13 17:40:34 +09003994int netdev_class_create_file_ns(struct class_attribute *class_attr,
3995 const void *ns);
3996void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3997 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04003998
3999static inline int netdev_class_create_file(struct class_attribute *class_attr)
4000{
4001 return netdev_class_create_file_ns(class_attr, NULL);
4002}
4003
4004static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4005{
4006 netdev_class_remove_file_ns(class_attr, NULL);
4007}
Jay Vosburghb8a97872008-06-13 18:12:04 -07004008
Johannes Berg04600792010-08-05 17:45:15 +02004009extern struct kobj_ns_type_operations net_ns_type_operations;
4010
Joe Perchesf629d202013-09-26 14:48:15 -07004011const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07004012
Joe Perchesf629d202013-09-26 14:48:15 -07004013void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004014
Michal Kubečekda081432014-05-20 08:29:25 +02004015static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4016 netdev_features_t f2)
4017{
Tom Herbertc8cd0982015-12-14 11:19:44 -08004018 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4019 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08004020 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004021 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08004022 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004023 }
Michal Kubečekda081432014-05-20 08:29:25 +02004024
Tom Herbertc8cd0982015-12-14 11:19:44 -08004025 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02004026}
4027
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004028static inline netdev_features_t netdev_get_wanted_features(
4029 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00004030{
4031 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4032}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004033netdev_features_t netdev_increment_features(netdev_features_t all,
4034 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00004035
4036/* Allow TSO being used on stacked device :
4037 * Performing the GSO segmentation before last device
4038 * is a performance improvement.
4039 */
4040static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4041 netdev_features_t mask)
4042{
4043 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4044}
4045
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004046int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004047void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004048void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004049
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004050void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4051 struct net_device *dev);
4052
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004053netdev_features_t passthru_features_check(struct sk_buff *skb,
4054 struct net_device *dev,
4055 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004056netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004057
David S. Miller4d295152012-03-07 21:02:35 -05004058static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004059{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004060 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004061
4062 /* check flags correspondence */
4063 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4064 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4065 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4066 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004067 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004068 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4069 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004070 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4071 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004072 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4073 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004074 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4075 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 -04004076 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004077 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004078 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004079
Herbert Xubcd76112006-06-30 13:36:35 -07004080 return (features & feature) == feature;
4081}
4082
David S. Miller4d295152012-03-07 21:02:35 -05004083static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004084{
Herbert Xu278b2512009-06-03 21:20:51 -07004085 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004086 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004087}
4088
Johannes Berg8b86a612015-04-17 15:45:04 +02004089static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004090 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004091{
Jesse Grossfc741212011-01-09 06:23:32 +00004092 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004093 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4094 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004095}
4096
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004097static inline void netif_set_gso_max_size(struct net_device *dev,
4098 unsigned int size)
4099{
4100 dev->gso_max_size = size;
4101}
4102
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004103static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4104 int pulled_hlen, u16 mac_offset,
4105 int mac_len)
4106{
4107 skb->protocol = protocol;
4108 skb->encapsulation = 1;
4109 skb_push(skb, pulled_hlen);
4110 skb_reset_transport_header(skb);
4111 skb->mac_header = mac_offset;
4112 skb->network_header = skb->mac_header + mac_len;
4113 skb->mac_len = mac_len;
4114}
4115
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004116static inline bool netif_is_macsec(const struct net_device *dev)
4117{
4118 return dev->priv_flags & IFF_MACSEC;
4119}
4120
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004121static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004122{
4123 return dev->priv_flags & IFF_MACVLAN;
4124}
4125
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004126static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004127{
4128 return dev->priv_flags & IFF_MACVLAN_PORT;
4129}
4130
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004131static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004132{
4133 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4134}
4135
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004136static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004137{
4138 return dev->priv_flags & IFF_IPVLAN_MASTER;
4139}
4140
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004141static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004142{
4143 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4144}
4145
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004146static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004147{
4148 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4149}
4150
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004151static inline bool netif_supports_nofcs(struct net_device *dev)
4152{
4153 return dev->priv_flags & IFF_SUPP_NOFCS;
4154}
4155
David Ahern007979e2015-09-29 20:07:10 -07004156static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004157{
David Ahern007979e2015-09-29 20:07:10 -07004158 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004159}
4160
David Ahernfee6d4c2015-10-05 08:51:24 -07004161static inline bool netif_is_l3_slave(const struct net_device *dev)
4162{
4163 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4164}
4165
Jiri Pirko0894ae32015-08-27 09:31:19 +02004166static inline bool netif_is_bridge_master(const struct net_device *dev)
4167{
4168 return dev->priv_flags & IFF_EBRIDGE;
4169}
4170
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004171static inline bool netif_is_bridge_port(const struct net_device *dev)
4172{
4173 return dev->priv_flags & IFF_BRIDGE_PORT;
4174}
4175
Jiri Pirko35d4e172015-08-27 09:31:20 +02004176static inline bool netif_is_ovs_master(const struct net_device *dev)
4177{
4178 return dev->priv_flags & IFF_OPENVSWITCH;
4179}
4180
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004181static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004182{
4183 return dev->priv_flags & IFF_TEAM;
4184}
4185
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004186static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004187{
4188 return dev->priv_flags & IFF_TEAM_PORT;
4189}
4190
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004191static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004192{
4193 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4194}
4195
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004196static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004197{
4198 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4199}
4200
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004201static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4202{
4203 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4204}
4205
Eric Dumazet02875872014-10-05 18:38:35 -07004206/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4207static inline void netif_keep_dst(struct net_device *dev)
4208{
4209 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4210}
4211
Paolo Abeni18d3df32016-07-14 18:00:10 +02004212/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4213static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4214{
4215 /* TODO: reserve and use an additional IFF bit, if we get more users */
4216 return dev->priv_flags & IFF_MACSEC;
4217}
4218
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004219extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004220
Joe Perches571ba422010-02-09 11:49:47 +00004221/* Logging, debugging and troubleshooting/diagnostic helpers. */
4222
4223/* netdev_printk helpers, similar to dev_printk */
4224
4225static inline const char *netdev_name(const struct net_device *dev)
4226{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004227 if (!dev->name[0] || strchr(dev->name, '%'))
4228 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004229 return dev->name;
4230}
4231
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004232static inline const char *netdev_reg_state(const struct net_device *dev)
4233{
4234 switch (dev->reg_state) {
4235 case NETREG_UNINITIALIZED: return " (uninitialized)";
4236 case NETREG_REGISTERED: return "";
4237 case NETREG_UNREGISTERING: return " (unregistering)";
4238 case NETREG_UNREGISTERED: return " (unregistered)";
4239 case NETREG_RELEASED: return " (released)";
4240 case NETREG_DUMMY: return " (dummy)";
4241 }
4242
4243 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4244 return " (unknown)";
4245}
4246
Joe Perchesf629d202013-09-26 14:48:15 -07004247__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004248void netdev_printk(const char *level, const struct net_device *dev,
4249 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004250__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004251void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004252__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004253void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004254__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004255void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004256__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004257void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004258__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004259void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004260__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004261void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004262__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004263void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004264
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004265#define MODULE_ALIAS_NETDEV(device) \
4266 MODULE_ALIAS("netdev-" device)
4267
Jim Cromieb558c962011-12-19 17:11:18 -05004268#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004269#define netdev_dbg(__dev, format, args...) \
4270do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004271 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004272} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004273#elif defined(DEBUG)
4274#define netdev_dbg(__dev, format, args...) \
4275 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004276#else
4277#define netdev_dbg(__dev, format, args...) \
4278({ \
4279 if (0) \
4280 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004281})
4282#endif
4283
4284#if defined(VERBOSE_DEBUG)
4285#define netdev_vdbg netdev_dbg
4286#else
4287
4288#define netdev_vdbg(dev, format, args...) \
4289({ \
4290 if (0) \
4291 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4292 0; \
4293})
4294#endif
4295
4296/*
4297 * netdev_WARN() acts like dev_printk(), but with the key difference
4298 * of using a WARN/WARN_ON to get the message out, including the
4299 * file/line information and a backtrace.
4300 */
4301#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004302 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4303 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004304
Joe Perchesb3d95c52010-02-09 11:49:49 +00004305/* netif printk helpers, similar to netdev_printk */
4306
4307#define netif_printk(priv, type, level, dev, fmt, args...) \
4308do { \
4309 if (netif_msg_##type(priv)) \
4310 netdev_printk(level, (dev), fmt, ##args); \
4311} while (0)
4312
Joe Perchesf45f4322010-06-27 01:02:36 +00004313#define netif_level(level, priv, type, dev, fmt, args...) \
4314do { \
4315 if (netif_msg_##type(priv)) \
4316 netdev_##level(dev, fmt, ##args); \
4317} while (0)
4318
Joe Perchesb3d95c52010-02-09 11:49:49 +00004319#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004320 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004321#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004322 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004323#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004324 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004325#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004326 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004327#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004328 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004329#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004330 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004331#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004332 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004333
Joe Perches0053ea92012-05-30 07:43:34 +00004334#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004335#define netif_dbg(priv, type, netdev, format, args...) \
4336do { \
4337 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004338 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004339} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004340#elif defined(DEBUG)
4341#define netif_dbg(priv, type, dev, format, args...) \
4342 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004343#else
4344#define netif_dbg(priv, type, dev, format, args...) \
4345({ \
4346 if (0) \
4347 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4348 0; \
4349})
4350#endif
4351
4352#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004353#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004354#else
4355#define netif_vdbg(priv, type, dev, format, args...) \
4356({ \
4357 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004358 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004359 0; \
4360})
4361#endif
Joe Perches571ba422010-02-09 11:49:47 +00004362
Cong Wang900ff8c2013-02-18 19:20:33 +00004363/*
4364 * The list of packet types we will receive (as opposed to discard)
4365 * and the routines to invoke.
4366 *
4367 * Why 16. Because with 16 the only overlap we get on a hash of the
4368 * low nibble of the protocol value is RARP/SNAP/X.25.
4369 *
4370 * NOTE: That is no longer true with the addition of VLAN tags. Not
4371 * sure which should go first, but I bet it won't make much
4372 * difference if we are running VLANs. The good news is that
4373 * this protocol won't be in the list unless compiled in, so
4374 * the average user (w/out VLANs) will not be adversely affected.
4375 * --BLG
4376 *
4377 * 0800 IP
4378 * 8100 802.1Q VLAN
4379 * 0001 802.3
4380 * 0002 AX.25
4381 * 0004 802.2
4382 * 8035 RARP
4383 * 0005 SNAP
4384 * 0805 X.25
4385 * 0806 ARP
4386 * 8137 IPX
4387 * 0009 Localtalk
4388 * 86DD IPv6
4389 */
4390#define PTYPE_HASH_SIZE (16)
4391#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4392
Jiri Pirko385a1542009-05-27 15:48:07 -07004393#endif /* _LINUX_NETDEVICE_H */