blob: 8c5c8cdc7b977250dab9cab2a508249eb3814e32 [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>
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -080044#ifdef CONFIG_DCB
Alexander Duyck2f90b862008-11-20 20:52:10 -080045#include <net/dcbnl.h>
46#endif
Neil Horman5bc14212011-11-22 05:10:51 +000047#include <net/netprio_cgroup.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020048
Michał Mirosława59e2ec2011-11-15 15:29:55 +000049#include <linux/netdev_features.h>
John Fastabend77162022012-04-15 06:43:56 +000050#include <linux/neighbour.h>
David Howells607ca462012-10-13 10:46:48 +010051#include <uapi/linux/netdevice.h>
Moni Shoua61bd3852015-02-03 16:48:29 +020052#include <uapi/linux/if_bonding.h>
John Fastabende4c67342016-02-16 21:16:15 -080053#include <uapi/linux/pkt_cls.h>
Jiri Kosina59cc1f62016-08-10 11:05:15 +020054#include <linux/hashtable.h>
Michał Mirosława59e2ec2011-11-15 15:29:55 +000055
Jeff Moyer115c1d62005-06-22 22:05:31 -070056struct netpoll_info;
Paul Gortmaker313162d2012-01-30 11:46:54 -050057struct device;
Richard Cochranc1f19b52010-07-17 08:49:36 +000058struct phy_device;
Andrew Lunnc6e970a2017-03-28 23:45:06 +020059struct dsa_switch_tree;
60
Johannes Berg704232c2007-04-23 12:20:05 -070061/* 802.11 specific */
62struct wireless_dev;
Alexander Aring98a18b62014-11-02 06:44:54 +010063/* 802.15.4 specific */
64struct wpan_dev;
Robert Shearman03c57742015-04-22 11:14:37 +010065struct mpls_dev;
Alexander Duyck7c46a642016-06-16 12:21:00 -070066/* UDP Tunnel offloads */
67struct udp_tunnel_info;
Brenden Blancoa7862b42016-07-19 12:16:48 -070068struct bpf_prog;
Linus Torvalds1da177e2005-04-16 15:20:36 -070069
Joe Perchesf629d202013-09-26 14:48:15 -070070void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000072
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000073/* Backlog congestion levels */
74#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75#define NET_RX_DROP 1 /* packet dropped */
76
Patrick McHardy572a9d72009-11-10 06:14:14 +000077/*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -050089 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
Patrick McHardy572a9d72009-11-10 06:14:14 +000091 * others are propagated to higher layers.
92 */
93
94/* qdisc ->enqueue() return codes. */
95#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000096#define NET_XMIT_DROP 0x01 /* skb dropped */
97#define NET_XMIT_CN 0x02 /* congestion notification */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000098#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -0200100/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +0000103#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000106/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000107#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000108
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000109enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000113};
114typedef enum netdev_tx netdev_tx_t;
115
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000116/*
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 */
120static inline bool dev_xmit_complete(int rc)
121{
122 /*
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 */
128 if (likely(rc < NET_XMIT_MASK))
129 return true;
130
131 return false;
132}
133
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500135 * Compute the worst-case header length according to the protocols
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 * used.
137 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800138
KY Srinivasanc0eb4542015-12-01 16:43:10 -0800139#if defined(CONFIG_HYPERV_NET)
140# define LL_MAX_HEADER 128
141#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700142# if defined(CONFIG_MAC80211_MESH)
143# define LL_MAX_HEADER 128
144# else
145# define LL_MAX_HEADER 96
146# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700148# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149#endif
150
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000151#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153#define MAX_HEADER LL_MAX_HEADER
154#else
155#define MAX_HEADER (LL_MAX_HEADER + 48)
156#endif
157
158/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800162
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800163struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187};
188
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189
190#include <linux/cache.h>
191#include <linux/skbuff.h>
192
Eric Dumazetadc93002011-11-17 03:13:26 +0000193#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100194#include <linux/static_key.h>
195extern struct static_key rps_needed;
Eric Dumazet13bfff22016-12-07 08:29:10 -0800196extern struct static_key rfs_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000197#endif
198
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199struct neighbour;
200struct neigh_parms;
201struct sk_buff;
202
Jiri Pirkof001fde2009-05-05 02:48:28 +0000203struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
206 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000207#define NETDEV_HW_ADDR_T_LAN 1
208#define NETDEV_HW_ADDR_T_SAN 2
209#define NETDEV_HW_ADDR_T_SLAVE 3
210#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000211#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000212 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000213 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700214 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400215 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000216 struct rcu_head rcu_head;
217};
218
Jiri Pirko31278e72009-06-17 01:12:19 +0000219struct netdev_hw_addr_list {
220 struct list_head list;
221 int count;
222};
223
Jiri Pirko22bedad32010-04-01 21:22:57 +0000224#define netdev_hw_addr_list_count(l) ((l)->count)
225#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226#define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
228
229#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800231#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800233
Jiri Pirko22bedad32010-04-01 21:22:57 +0000234#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700236#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece2010-02-04 10:22:25 -0800238
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800239struct hh_cache {
Alexey Dobriyan5b3dc2f2017-04-10 11:11:17 +0300240 unsigned int hh_len;
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 */
Eric Dumazet39e6c822017-02-28 10:34:50 -0800333 NAPI_STATE_MISSED, /* reschedule a napi */
David S. Millera0a46192008-01-07 20:35:07 -0800334 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800335 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eric Dumazetd64b5e82015-11-18 06:31:00 -0800336 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
337 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
Eric Dumazet217f6972016-11-15 10:15:11 -0800338 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
339};
340
341enum {
Eric Dumazet39e6c822017-02-28 10:34:50 -0800342 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
343 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
344 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
345 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
346 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
347 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
348 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700349};
350
Ben Hutchings5b252f02009-10-29 07:17:09 +0000351enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700352 GRO_MERGED,
353 GRO_MERGED_FREE,
354 GRO_HELD,
355 GRO_NORMAL,
356 GRO_DROP,
Steffen Klassert25393d32017-02-15 09:39:44 +0100357 GRO_CONSUMED,
Herbert Xud1c76af2009-03-16 10:50:02 -0700358};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000359typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700360
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000361/*
362 * enum rx_handler_result - Possible return values for rx_handlers.
363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
364 * further.
365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366 * case skb->dev was changed by rx_handler.
367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000369 *
370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
371 * special processing of the skb, prior to delivery to protocol handlers.
372 *
373 * Currently, a net_device can only have a single rx_handler registered. Trying
374 * to register a second rx_handler will return -EBUSY.
375 *
376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377 * To unregister a rx_handler on a net_device, use
378 * netdev_rx_handler_unregister().
379 *
380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
381 * do with the skb.
382 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500383 * If the rx_handler consumed the skb in some way, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500385 * the skb to be delivered in some other way.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000386 *
387 * If the rx_handler changed skb->dev, to divert the skb to another
388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389 * new device will be called if it exists.
390 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500391 * If the rx_handler decides the skb should be ignored, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400393 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000394 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000396 * delivered, it should return RX_HANDLER_PASS.
397 *
398 * A device without a registered rx_handler will behave as if rx_handler
399 * returned RX_HANDLER_PASS.
400 */
401
402enum rx_handler_result {
403 RX_HANDLER_CONSUMED,
404 RX_HANDLER_ANOTHER,
405 RX_HANDLER_EXACT,
406 RX_HANDLER_PASS,
407};
408typedef enum rx_handler_result rx_handler_result_t;
409typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000410
Joe Perchesf629d202013-09-26 14:48:15 -0700411void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700412void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700413
David S. Miller4d295152012-03-07 21:02:35 -0500414static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800415{
416 return test_bit(NAPI_STATE_DISABLE, &n->state);
417}
418
Eric Dumazet39e6c822017-02-28 10:34:50 -0800419bool napi_schedule_prep(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700420
421/**
422 * napi_schedule - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500423 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700424 *
425 * Schedule NAPI poll routine to be called if it is not already
426 * running.
427 */
428static inline void napi_schedule(struct napi_struct *n)
429{
430 if (napi_schedule_prep(n))
431 __napi_schedule(n);
432}
433
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700434/**
435 * napi_schedule_irqoff - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500436 * @n: NAPI context
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700437 *
438 * Variant of napi_schedule(), assuming hard irqs are masked.
439 */
440static inline void napi_schedule_irqoff(struct napi_struct *n)
441{
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
444}
445
Roland Dreierbfe13f52007-10-09 15:47:37 -0700446/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
David S. Miller4d295152012-03-07 21:02:35 -0500447static inline bool napi_reschedule(struct napi_struct *napi)
Roland Dreierbfe13f52007-10-09 15:47:37 -0700448{
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
David S. Miller4d295152012-03-07 21:02:35 -0500451 return true;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700452 }
David S. Miller4d295152012-03-07 21:02:35 -0500453 return false;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700454}
455
Eric Dumazet364b6052016-11-15 10:15:13 -0800456bool napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700457/**
458 * napi_complete - NAPI processing complete
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500459 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700460 *
461 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800462 * Consider using napi_complete_done() instead.
Eric Dumazet364b6052016-11-15 10:15:13 -0800463 * Return false if device should avoid rearming interrupts.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700464 */
Eric Dumazet364b6052016-11-15 10:15:13 -0800465static inline bool napi_complete(struct napi_struct *n)
Eric Dumazet3b47d302014-11-06 21:09:44 -0800466{
467 return napi_complete_done(n, 0);
468}
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700469
470/**
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300471 * napi_hash_del - remove a NAPI from global table
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500472 * @napi: NAPI context
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300473 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500474 * Warning: caller must observe RCU grace period
Eric Dumazet34cbe272015-11-18 06:31:02 -0800475 * before freeing memory containing @napi, if
476 * this function returns true.
Eric Dumazet93d05d42015-11-18 06:31:03 -0800477 * Note: core networking stack automatically calls it
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500478 * from netif_napi_del().
Eric Dumazet93d05d42015-11-18 06:31:03 -0800479 * Drivers might want to call this helper to combine all
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500480 * the needed RCU grace periods into a single one.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300481 */
Eric Dumazet34cbe272015-11-18 06:31:02 -0800482bool napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300483
484/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700485 * napi_disable - prevent NAPI from scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500486 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700487 *
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
490 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800491void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700492
493/**
494 * napi_enable - enable NAPI scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500495 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700496 *
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
499 */
500static inline void napi_enable(struct napi_struct *n)
501{
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100503 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700504 clear_bit(NAPI_STATE_SCHED, &n->state);
Neil Horman2d8bff1262015-09-23 14:57:58 -0400505 clear_bit(NAPI_STATE_NPSVC, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700506}
507
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700508/**
509 * napi_synchronize - wait until NAPI is not running
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500510 * @n: NAPI context
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700511 *
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
515 */
516static inline void napi_synchronize(const struct napi_struct *n)
517{
Arnd Bergmannfacc4322016-01-22 11:43:44 +0100518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
520 msleep(1);
521 else
522 barrier();
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700523}
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700524
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800525enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700528 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700529};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200530
531#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
534
535#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
537 QUEUE_STATE_FROZEN)
538#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
539 QUEUE_STATE_FROZEN)
540
Tom Herbert734664982011-11-28 16:32:44 +0000541/*
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
549 */
David S. Miller79d16382008-07-08 23:14:46 -0700550
David S. Millerbb949fb2008-07-08 16:55:56 -0700551struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700552/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500553 * read-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700554 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700555 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700556 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700557 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000558#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000559 struct kobject kobj;
560#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000561#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 int numa_node;
563#endif
Florian Westphalc0ef0792016-05-03 03:29:09 +0200564 unsigned long tx_maxrate;
565 /*
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
568 */
569 unsigned long trans_timeout;
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700570/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500571 * write-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700575 /*
Florian Westphal9b366272016-05-03 16:33:14 +0200576 * Time (in jiffies) of last Tx
Eric Dumazet9d214932009-05-17 20:55:16 -0700577 */
578 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000579
Tom Herbert114cf582011-11-28 16:33:09 +0000580 unsigned long state;
581
582#ifdef CONFIG_BQL
583 struct dql dql;
584#endif
David S. Millere8a04642008-07-17 00:34:19 -0700585} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700586
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000587static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
588{
589#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 return q->numa_node;
591#else
Changli Gaob236da62010-12-14 03:09:15 +0000592 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000593#endif
594}
595
596static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
597{
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 q->numa_node = node;
600#endif
601}
602
Eric Dumazetdf334542010-03-24 19:13:54 +0000603#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000604/*
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
607 */
608struct rps_map {
609 unsigned int len;
610 struct rcu_head rcu;
611 u16 cpus[0];
612};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000613#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000614
Tom Herbertfec5e652010-04-16 16:01:27 -0700615/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700619 */
620struct rps_dev_flow {
621 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000622 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700623 unsigned int last_qtail;
624};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000625#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700626
627/*
628 * The rps_dev_flow_table structure contains a table of flow mappings.
629 */
630struct rps_dev_flow_table {
631 unsigned int mask;
632 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700633 struct rps_dev_flow flows[0];
634};
635#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000636 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700637
638/*
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
Eric Dumazet567e4b72015-02-06 12:59:01 -0800643 * rps_cpu_mask is used to partition the space, depending on number of
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
Eric Dumazet567e4b72015-02-06 12:59:01 -0800646 * meaning we use 32-6=26 bits for the hash.
Tom Herbertfec5e652010-04-16 16:01:27 -0700647 */
648struct rps_sock_flow_table {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800649 u32 mask;
Eric Dumazet93c1af62015-02-08 20:39:13 -0800650
651 u32 ents[0] ____cacheline_aligned_in_smp;
Tom Herbertfec5e652010-04-16 16:01:27 -0700652};
Eric Dumazet567e4b72015-02-06 12:59:01 -0800653#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
Tom Herbertfec5e652010-04-16 16:01:27 -0700654
655#define RPS_NO_CPU 0xffff
656
Eric Dumazet567e4b72015-02-06 12:59:01 -0800657extern u32 rps_cpu_mask;
658extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
659
Tom Herbertfec5e652010-04-16 16:01:27 -0700660static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
661 u32 hash)
662{
663 if (table && hash) {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800664 unsigned int index = hash & table->mask;
665 u32 val = hash & ~rps_cpu_mask;
Tom Herbertfec5e652010-04-16 16:01:27 -0700666
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500667 /* We only give a hint, preemption can change CPU under us */
Eric Dumazet567e4b72015-02-06 12:59:01 -0800668 val |= raw_smp_processor_id();
Tom Herbertfec5e652010-04-16 16:01:27 -0700669
Eric Dumazet567e4b72015-02-06 12:59:01 -0800670 if (table->ents[index] != val)
671 table->ents[index] = val;
Tom Herbertfec5e652010-04-16 16:01:27 -0700672 }
673}
674
Ben Hutchingsc4454772011-01-19 11:03:53 +0000675#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700676bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
677 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000678#endif
Michael Daltona953be52014-01-16 22:23:28 -0800679#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000680
Tom Herbert0a9627f2010-03-16 08:03:29 +0000681/* This structure contains an instance of an RX queue. */
682struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800683#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000684 struct rps_map __rcu *rps_map;
685 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800686#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000687 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000688 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000689} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800690
691/*
692 * RX queue sysfs structures and functions.
693 */
694struct rx_queue_attribute {
695 struct attribute attr;
696 ssize_t (*show)(struct netdev_rx_queue *queue,
697 struct rx_queue_attribute *attr, char *buf);
698 ssize_t (*store)(struct netdev_rx_queue *queue,
699 struct rx_queue_attribute *attr, const char *buf, size_t len);
700};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800701
Tom Herbertbf264142010-11-26 08:36:09 +0000702#ifdef CONFIG_XPS
703/*
704 * This structure holds an XPS map which can be of variable length. The
705 * map is an array of queues.
706 */
707struct xps_map {
708 unsigned int len;
709 unsigned int alloc_len;
710 struct rcu_head rcu;
711 u16 queues[0];
712};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000713#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Helge Dellerc59f4192015-10-25 10:00:32 +0100714#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
715 - sizeof(struct xps_map)) / sizeof(u16))
Tom Herbertbf264142010-11-26 08:36:09 +0000716
717/*
718 * This structure holds all XPS maps for device. Maps are indexed by CPU.
719 */
720struct xps_dev_maps {
721 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000722 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000723};
Alexander Duyck184c4492016-10-28 11:50:13 -0400724#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
725 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
Tom Herbertbf264142010-11-26 08:36:09 +0000726#endif /* CONFIG_XPS */
727
John Fastabend4f57c082011-01-17 08:06:04 +0000728#define TC_MAX_QUEUE 16
729#define TC_BITMASK 15
730/* HW offloaded queuing disciplines txq count and offset maps */
731struct netdev_tc_txq {
732 u16 count;
733 u16 offset;
734};
735
Neerav Parikh68bad942012-01-04 20:23:39 +0000736#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
737/*
738 * This structure is to hold information about the device
739 * configured to run FCoE protocol stack.
740 */
741struct netdev_fcoe_hbainfo {
742 char manufacturer[64];
743 char serial_number[64];
744 char hardware_version[64];
745 char driver_version[64];
746 char optionrom_version[64];
747 char firmware_version[64];
748 char model[256];
749 char model_description[256];
750};
751#endif
752
Jiri Pirko02637fc2014-11-28 14:34:16 +0100753#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200754
Jiri Pirko02637fc2014-11-28 14:34:16 +0100755/* This structure holds a unique identifier to identify some
756 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200757 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100758struct netdev_phys_item_id {
759 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200760 unsigned char id_len;
761};
762
Scott Feldmand754f982015-07-18 18:24:49 -0700763static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
764 struct netdev_phys_item_id *b)
765{
766 return a->id_len == b->id_len &&
767 memcmp(a->id, b->id, a->id_len) == 0;
768}
769
Daniel Borkmann99932d42014-02-16 15:55:20 +0100770typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
771 struct sk_buff *skb);
772
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800773/* These structures hold the attributes of qdisc and classifiers
John Fastabend16e5cc62016-02-16 21:16:43 -0800774 * that are being passed to the netdevice through the setup_tc op.
775 */
776enum {
777 TC_SETUP_MQPRIO,
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800778 TC_SETUP_CLSU32,
Amir Vadai5b33f482016-03-08 12:42:29 +0200779 TC_SETUP_CLSFLOWER,
Yotam Gigib87f7932016-07-21 12:03:12 +0200780 TC_SETUP_MATCHALL,
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100781 TC_SETUP_CLSBPF,
John Fastabend16e5cc62016-02-16 21:16:43 -0800782};
783
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800784struct tc_cls_u32_offload;
785
John Fastabend16e5cc62016-02-16 21:16:43 -0800786struct tc_to_netdev {
787 unsigned int type;
788 union {
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800789 struct tc_cls_u32_offload *cls_u32;
Amir Vadai5b33f482016-03-08 12:42:29 +0200790 struct tc_cls_flower_offload *cls_flower;
Yotam Gigib87f7932016-07-21 12:03:12 +0200791 struct tc_cls_matchall_offload *cls_mall;
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100792 struct tc_cls_bpf_offload *cls_bpf;
Amritha Nambiar56f36ac2017-03-15 10:39:25 -0700793 struct tc_mqprio_qopt *mqprio;
John Fastabend16e5cc62016-02-16 21:16:43 -0800794 };
Hadar Hen Zion7091d8c2016-12-01 14:06:37 +0200795 bool egress_dev;
John Fastabend16e5cc62016-02-16 21:16:43 -0800796};
797
Brenden Blancoa7862b42016-07-19 12:16:48 -0700798/* These structures hold the attributes of xdp state that are being passed
799 * to the netdevice through the xdp op.
800 */
801enum xdp_netdev_command {
802 /* Set or clear a bpf program used in the earliest stages of packet
803 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
804 * is responsible for calling bpf_prog_put on any old progs that are
805 * stored. In case of error, the callee need not release the new prog
806 * reference, but on success it takes ownership and must bpf_prog_put
807 * when it is no longer used.
808 */
809 XDP_SETUP_PROG,
810 /* Check if a bpf program is set on the device. The callee should
811 * return true if a program is currently attached and running.
812 */
813 XDP_QUERY_PROG,
814};
815
816struct netdev_xdp {
817 enum xdp_netdev_command command;
818 union {
819 /* XDP_SETUP_PROG */
820 struct bpf_prog *prog;
821 /* XDP_QUERY_PROG */
822 bool prog_attached;
823 };
824};
John Fastabend16e5cc62016-02-16 21:16:43 -0800825
Steffen Klassertd77e38e2017-04-14 10:06:10 +0200826#ifdef CONFIG_XFRM_OFFLOAD
827struct xfrmdev_ops {
828 int (*xdo_dev_state_add) (struct xfrm_state *x);
829 void (*xdo_dev_state_delete) (struct xfrm_state *x);
830 void (*xdo_dev_state_free) (struct xfrm_state *x);
831 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
832 struct xfrm_state *x);
833};
834#endif
835
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800836/*
837 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800840 *
841 * int (*ndo_init)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800846 *
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
850 *
851 * int (*ndo_open)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500852 * This function is called when a network device transitions to the up
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800853 * state.
854 *
855 * int (*ndo_stop)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500856 * This function is called when a network device transitions to the down
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800857 * state.
858 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800861 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500866 * Required; cannot be NULL.
Stephen Hemminger00829822008-11-20 20:14:53 -0800867 *
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -0800868 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
869 * struct net_device *dev
870 * netdev_features_t features);
871 * Called by core transmit path to determine if device is capable of
872 * performing offload operations on a given packet. This is to give
873 * the device an opportunity to implement any restrictions that cannot
874 * be otherwise expressed by feature flags. The check is called with
875 * the set of features that the stack has calculated and it returns
876 * those the driver believes to be appropriate.
Eric Dumazetcdba7562016-01-06 06:53:50 -0800877 *
Jason Wangf663dd92014-01-10 16:18:26 +0800878 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100879 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500880 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800881 * transmit queues.
882 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800883 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
884 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500885 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800886 *
887 * void (*ndo_set_rx_mode)(struct net_device *dev);
888 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000889 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500890 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800891 *
892 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
893 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700894 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500895 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800896 *
897 * int (*ndo_validate_addr)(struct net_device *dev);
898 * Test if Media Access Control address is valid for the device.
899 *
900 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500901 * Called when a user requests an ioctl which can't be handled by
902 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800903 * not supported error code.
904 *
905 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
906 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500907 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800908 * interface (PCI) for low level management.
909 *
910 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
911 * Called when a user wants to change the Maximum Transfer Unit
912 * of a device. If not defined, any request to change MTU will
913 * will return an error.
914 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800915 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500916 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800917 * for dev->watchdog ticks.
918 *
stephen hemmingerbc1f4472017-01-06 19:12:52 -0800919 * void (*ndo_get_stats64)(struct net_device *dev,
920 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700921 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800922 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000923 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000924 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
925 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700926 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000927 * (which should normally be dev->stats) and return a pointer to
928 * it. The structure may be changed asynchronously only if each
929 * field is written atomically.
930 * 3. Update dev->stats asynchronously and atomically, and define
931 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800932 *
Or Gerlitz3df5b3c2016-11-22 23:09:54 +0200933 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
Nogah Frankel2c9d85d2016-09-16 15:05:36 +0200934 * Return true if this device supports offload stats of this attr_id.
935 *
936 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
937 * void *attr_data)
938 * Get statistics for offload operations by attr_id. Write it into the
939 * attr_data pointer.
940 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530941 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500942 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000943 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800944 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530945 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500946 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000947 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800948 *
949 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000950 *
951 * SR-IOV management functions.
952 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
Moshe Shemesh79aab092016-09-22 12:11:15 +0300953 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
954 * u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400955 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
956 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000957 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000958 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000959 * int (*ndo_get_vf_config)(struct net_device *dev,
960 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300961 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700962 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
963 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300964 *
965 * Enable or disable the VF ability to query its RSS Redirection Table and
966 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500967 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300968 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700969 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
Florian Fainelli6a4bc2b2017-01-26 14:44:17 -0800970 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
971 * __be16 protocol, struct tc_to_netdev *tc);
972 * Called to setup any 'tc' scheduler, classifier or action on @dev.
973 * This is always called from the stack with the rtnl lock held and netif
974 * tx queues stopped. This allows the netdevice to perform queue
975 * management safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000976 *
Yi Zoue9bce842011-03-09 08:48:03 +0000977 * Fiber Channel over Ethernet (FCoE) offload functions.
978 * int (*ndo_fcoe_enable)(struct net_device *dev);
979 * Called when the FCoE protocol stack wants to start using LLD for FCoE
980 * so the underlying device can perform whatever needed configuration or
981 * initialization to support acceleration of FCoE traffic.
982 *
983 * int (*ndo_fcoe_disable)(struct net_device *dev);
984 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
985 * so the underlying device can perform whatever needed clean-ups to
986 * stop supporting acceleration of FCoE traffic.
987 *
988 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
989 * struct scatterlist *sgl, unsigned int sgc);
990 * Called when the FCoE Initiator wants to initialize an I/O that
991 * is a possible candidate for Direct Data Placement (DDP). The LLD can
992 * perform necessary setup and returns 1 to indicate the device is set up
993 * successfully to perform DDP on this I/O, otherwise this returns 0.
994 *
995 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
996 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
997 * indicated by the FC exchange id 'xid', so the underlying device can
998 * clean up and reuse resources for later DDP requests.
999 *
1000 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1001 * struct scatterlist *sgl, unsigned int sgc);
1002 * Called when the FCoE Target wants to initialize an I/O that
1003 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1004 * perform necessary setup and returns 1 to indicate the device is set up
1005 * successfully to perform DDP on this I/O, otherwise this returns 0.
1006 *
Neerav Parikh68bad942012-01-04 20:23:39 +00001007 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1008 * struct netdev_fcoe_hbainfo *hbainfo);
1009 * Called when the FCoE Protocol stack wants information on the underlying
1010 * device. This information is utilized by the FCoE protocol stack to
1011 * register attributes with Fiber Channel management service as per the
1012 * FC-GS Fabric Device Management Information(FDMI) specification.
1013 *
Yi Zoue9bce842011-03-09 08:48:03 +00001014 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1015 * Called when the underlying device wants to override default World Wide
1016 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1017 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1018 * protocol stack to use.
1019 *
Ben Hutchingsc4454772011-01-19 11:03:53 +00001020 * RFS acceleration.
1021 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1022 * u16 rxq_index, u32 flow_id);
1023 * Set hardware filter for RFS. rxq_index is the target queue index;
1024 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1025 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001026 *
Jiri Pirko8b98a702013-01-03 22:49:02 +00001027 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001028 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1029 * Called to make another netdev an underling.
1030 *
1031 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1032 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +00001033 *
1034 * Feature/offload setting functions.
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -08001035 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1036 * netdev_features_t features);
1037 * Adjusts the requested feature flags according to device-specific
1038 * constraints, and returns the resulting flags. Must not modify
1039 * the device state.
1040 *
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001041 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001042 * Called to update device configuration to new features. Passed
1043 * feature set might be less than what was returned by ndo_fix_features()).
1044 * Must return >0 or -errno if it changed dev->features itself.
1045 *
stephen hemmingeredc7d572012-10-01 12:32:33 +00001046 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1047 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001048 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +00001049 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +00001050 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1051 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001052 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001053 * Deletes the FDB entry from dev coresponding to addr.
1054 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001055 * struct net_device *dev, struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001056 * int *idx)
John Fastabend77162022012-04-15 06:43:56 +00001057 * Used to add FDB entries to dump requests. Implementers should add
1058 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001059 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001060 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1061 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001062 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001063 * struct net_device *dev, u32 filter_mask,
1064 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001065 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1066 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001067 *
1068 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1069 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1070 * which do not represent real hardware may define this to allow their
1071 * userspace components to manage their virtual carrier state. Devices
1072 * that determine carrier state from physical hardware properties (eg
1073 * network cables) or protocol-dependent mechanisms (eg
1074 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001075 *
1076 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001077 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001078 * Called to get ID of physical port of this device. If driver does
1079 * not implement this, it is assumed that the hw is not able to have
1080 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001081 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001082 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1083 * struct udp_tunnel_info *ti);
1084 * Called by UDP tunnel to notify a driver about the UDP port and socket
1085 * address family that a UDP tunnel is listnening to. It is called only
1086 * when a new port starts listening. The operation is protected by the
1087 * RTNL.
1088 *
1089 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1090 * struct udp_tunnel_info *ti);
1091 * Called by UDP tunnel to notify the driver about a UDP port and socket
1092 * address family that the UDP tunnel is not listening to anymore. The
1093 * operation is protected by the RTNL.
1094 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001095 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1096 * struct net_device *dev)
1097 * Called by upper layer devices to accelerate switching or other
1098 * station functionality into hardware. 'pdev is the lowerdev
1099 * to use for the offload and 'dev' is the net device that will
1100 * back the offload. Returns a pointer to the private structure
1101 * the upper layer will maintain.
1102 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1103 * Called by upper layer device to delete the station created
1104 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1105 * the station and priv is the structure returned by the add
1106 * operation.
1107 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1108 * struct net_device *dev,
1109 * void *priv);
1110 * Callback to use for xmit over the accelerated station. This
1111 * is used in place of ndo_start_xmit on accelerated net
1112 * devices.
John Fastabend822b3b22015-03-18 14:57:33 +02001113 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1114 * int queue_index, u32 maxrate);
1115 * Called when a user wants to set a max-rate limitation of specific
1116 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001117 * int (*ndo_get_iflink)(const struct net_device *dev);
1118 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001119 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001120 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001121 * This function is used to pass protocol port error state information
1122 * to the switch driver. The switch driver can react to the proto_down
1123 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001124 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1125 * This function is used to get egress tunnel information for given skb.
1126 * This is useful for retrieving outer tunnel header parameters while
1127 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001128 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1129 * This function is used to specify the headroom that the skb must
1130 * consider when allocation skb during packet reception. Setting
1131 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001132 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001133 * default value.
Brenden Blancoa7862b42016-07-19 12:16:48 -07001134 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1135 * This function is used to set or query state related to XDP on the
1136 * netdevice. See definition of enum xdp_netdev_command for details.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001137 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001138 */
1139struct net_device_ops {
1140 int (*ndo_init)(struct net_device *dev);
1141 void (*ndo_uninit)(struct net_device *dev);
1142 int (*ndo_open)(struct net_device *dev);
1143 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001144 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1145 struct net_device *dev);
1146 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1147 struct net_device *dev,
1148 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001149 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001150 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001151 void *accel_priv,
1152 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001153 void (*ndo_change_rx_flags)(struct net_device *dev,
1154 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001155 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001156 int (*ndo_set_mac_address)(struct net_device *dev,
1157 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001158 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001159 int (*ndo_do_ioctl)(struct net_device *dev,
1160 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001161 int (*ndo_set_config)(struct net_device *dev,
1162 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001163 int (*ndo_change_mtu)(struct net_device *dev,
1164 int new_mtu);
1165 int (*ndo_neigh_setup)(struct net_device *dev,
1166 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001167 void (*ndo_tx_timeout) (struct net_device *dev);
1168
stephen hemmingerbc1f4472017-01-06 19:12:52 -08001169 void (*ndo_get_stats64)(struct net_device *dev,
1170 struct rtnl_link_stats64 *storage);
Or Gerlitz3df5b3c2016-11-22 23:09:54 +02001171 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
Nogah Frankel2c9d85d2016-09-16 15:05:36 +02001172 int (*ndo_get_offload_stats)(int attr_id,
1173 const struct net_device *dev,
1174 void *attr_data);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001175 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1176
Jiri Pirko8e586132011-12-08 19:52:37 -05001177 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001178 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001179 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001180 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001181#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001182 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001183 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001184 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001185 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001186#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001187 int (*ndo_set_vf_mac)(struct net_device *dev,
1188 int queue, u8 *mac);
1189 int (*ndo_set_vf_vlan)(struct net_device *dev,
Moshe Shemesh79aab092016-09-22 12:11:15 +03001190 int queue, u16 vlan,
1191 u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001192 int (*ndo_set_vf_rate)(struct net_device *dev,
1193 int vf, int min_tx_rate,
1194 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001195 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1196 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001197 int (*ndo_set_vf_trust)(struct net_device *dev,
1198 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001199 int (*ndo_get_vf_config)(struct net_device *dev,
1200 int vf,
1201 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001202 int (*ndo_set_vf_link_state)(struct net_device *dev,
1203 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001204 int (*ndo_get_vf_stats)(struct net_device *dev,
1205 int vf,
1206 struct ifla_vf_stats
1207 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001208 int (*ndo_set_vf_port)(struct net_device *dev,
1209 int vf,
1210 struct nlattr *port[]);
1211 int (*ndo_get_vf_port)(struct net_device *dev,
1212 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001213 int (*ndo_set_vf_guid)(struct net_device *dev,
1214 int vf, u64 guid,
1215 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001216 int (*ndo_set_vf_rss_query_en)(
1217 struct net_device *dev,
1218 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001219 int (*ndo_setup_tc)(struct net_device *dev,
1220 u32 handle,
1221 __be16 protocol,
1222 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001223#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001224 int (*ndo_fcoe_enable)(struct net_device *dev);
1225 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001226 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1227 u16 xid,
1228 struct scatterlist *sgl,
1229 unsigned int sgc);
1230 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1231 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001232 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1233 u16 xid,
1234 struct scatterlist *sgl,
1235 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001236 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1237 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001238#endif
1239
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001240#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001241#define NETDEV_FCOE_WWNN 0
1242#define NETDEV_FCOE_WWPN 1
1243 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1244 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001245#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001246
Ben Hutchingsc4454772011-01-19 11:03:53 +00001247#ifdef CONFIG_RFS_ACCEL
1248 int (*ndo_rx_flow_steer)(struct net_device *dev,
1249 const struct sk_buff *skb,
1250 u16 rxq_index,
1251 u32 flow_id);
1252#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001253 int (*ndo_add_slave)(struct net_device *dev,
1254 struct net_device *slave_dev);
1255 int (*ndo_del_slave)(struct net_device *dev,
1256 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001257 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1258 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001259 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001260 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001261 int (*ndo_neigh_construct)(struct net_device *dev,
1262 struct neighbour *n);
1263 void (*ndo_neigh_destroy)(struct net_device *dev,
1264 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001265
1266 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001267 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001268 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001269 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001270 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001271 u16 flags);
1272 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001273 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001274 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001275 const unsigned char *addr,
1276 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001277 int (*ndo_fdb_dump)(struct sk_buff *skb,
1278 struct netlink_callback *cb,
1279 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001280 struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001281 int *idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001282
1283 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001284 struct nlmsghdr *nlh,
1285 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001286 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1287 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001288 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001289 u32 filter_mask,
1290 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001291 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001292 struct nlmsghdr *nlh,
1293 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001294 int (*ndo_change_carrier)(struct net_device *dev,
1295 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001296 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001297 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001298 int (*ndo_get_phys_port_name)(struct net_device *dev,
1299 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001300 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1301 struct udp_tunnel_info *ti);
1302 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1303 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001304 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1305 struct net_device *dev);
1306 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1307 void *priv);
1308
1309 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1310 struct net_device *dev,
1311 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001312 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001313 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1314 int queue_index,
1315 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001316 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001317 int (*ndo_change_proto_down)(struct net_device *dev,
1318 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001319 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1320 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001321 void (*ndo_set_rx_headroom)(struct net_device *dev,
1322 int needed_headroom);
Brenden Blancoa7862b42016-07-19 12:16:48 -07001323 int (*ndo_xdp)(struct net_device *dev,
1324 struct netdev_xdp *xdp);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001325};
1326
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001327/**
1328 * enum net_device_priv_flags - &struct net_device priv_flags
1329 *
1330 * These are the &struct net_device, they are only set internally
1331 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001332 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001333 * during any kernel release.
1334 *
1335 * You should have a pretty good reason to be extending these flags.
1336 *
1337 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1338 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001339 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001340 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001341 * @IFF_WAN_HDLC: WAN HDLC device
1342 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1343 * release skb->dst
1344 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1345 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1346 * @IFF_MACVLAN_PORT: device used as macvlan port
1347 * @IFF_BRIDGE_PORT: device used as bridge port
1348 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1349 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1350 * @IFF_UNICAST_FLT: Supports unicast filtering
1351 * @IFF_TEAM_PORT: device used as team port
1352 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1353 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1354 * change when it's running
1355 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001356 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1357 * underlying stacked devices
1358 * @IFF_IPVLAN_MASTER: IPvlan master device
1359 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001360 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001361 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001362 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001363 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001364 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001365 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001366 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1367 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001368 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001369 */
1370enum netdev_priv_flags {
1371 IFF_802_1Q_VLAN = 1<<0,
1372 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001373 IFF_BONDING = 1<<2,
1374 IFF_ISATAP = 1<<3,
1375 IFF_WAN_HDLC = 1<<4,
1376 IFF_XMIT_DST_RELEASE = 1<<5,
1377 IFF_DONT_BRIDGE = 1<<6,
1378 IFF_DISABLE_NETPOLL = 1<<7,
1379 IFF_MACVLAN_PORT = 1<<8,
1380 IFF_BRIDGE_PORT = 1<<9,
1381 IFF_OVS_DATAPATH = 1<<10,
1382 IFF_TX_SKB_SHARING = 1<<11,
1383 IFF_UNICAST_FLT = 1<<12,
1384 IFF_TEAM_PORT = 1<<13,
1385 IFF_SUPP_NOFCS = 1<<14,
1386 IFF_LIVE_ADDR_CHANGE = 1<<15,
1387 IFF_MACVLAN = 1<<16,
1388 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1389 IFF_IPVLAN_MASTER = 1<<18,
1390 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001391 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001392 IFF_NO_QUEUE = 1<<21,
1393 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001394 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001395 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001396 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001397 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001398 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001399};
1400
1401#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1402#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001403#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001404#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001405#define IFF_WAN_HDLC IFF_WAN_HDLC
1406#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1407#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1408#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1409#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1410#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1411#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1412#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1413#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1414#define IFF_TEAM_PORT IFF_TEAM_PORT
1415#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1416#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1417#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001418#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001419#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1420#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001421#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001422#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001423#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001424#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001425#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001426#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001427#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001428
Karoly Kemeny536721b2014-07-30 20:27:36 +02001429/**
1430 * struct net_device - The DEVICE structure.
1431 * Actually, this whole structure is a big mistake. It mixes I/O
1432 * data with strictly "high-level" data, and it has to know about
1433 * almost every data structure used in the INET module.
1434 *
1435 * @name: This is the first field of the "visible" part of this structure
1436 * (i.e. as seen by users in the "Space.c" file). It is the name
1437 * of the interface.
1438 *
1439 * @name_hlist: Device name hash chain, please keep it close to name[]
1440 * @ifalias: SNMP alias
1441 * @mem_end: Shared memory end
1442 * @mem_start: Shared memory start
1443 * @base_addr: Device I/O address
1444 * @irq: Device IRQ number
1445 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001446 * @carrier_changes: Stats to monitor carrier on<->off transitions
1447 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001448 * @state: Generic network queuing layer state, see netdev_state_t
1449 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001450 * @napi_list: List entry used for polling NAPI devices
1451 * @unreg_list: List entry when we are unregistering the
1452 * device; see the function unregister_netdev
1453 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001454 * @ptype_all: Device-specific packet handlers for all protocols
1455 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001456 *
1457 * @adj_list: Directly linked devices, like slaves for bonding
Karoly Kemeny536721b2014-07-30 20:27:36 +02001458 * @features: Currently active device features
1459 * @hw_features: User-changeable features
1460 *
1461 * @wanted_features: User-requested features
1462 * @vlan_features: Mask of features inheritable by VLAN devices
1463 *
1464 * @hw_enc_features: Mask of features inherited by encapsulating devices
1465 * This field indicates what encapsulation
1466 * offloads the hardware is capable of doing,
1467 * and drivers will need to set them appropriately.
1468 *
1469 * @mpls_features: Mask of features inheritable by MPLS
1470 *
1471 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001472 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001473 *
1474 * @stats: Statistics struct, which was left as a legacy, use
1475 * rtnl_link_stats64 instead
1476 *
1477 * @rx_dropped: Dropped packets by core network,
1478 * do not use this in drivers
1479 * @tx_dropped: Dropped packets by core network,
1480 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001481 * @rx_nohandler: nohandler dropped packets by core network on
1482 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001483 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001484 * @wireless_handlers: List of functions to handle Wireless Extensions,
1485 * instead of ioctl,
1486 * see <net/iw_handler.h> for details.
1487 * @wireless_data: Instance data managed by the core of wireless extensions
1488 *
1489 * @netdev_ops: Includes several pointers to callbacks,
1490 * if one wants to override the ndo_*() functions
1491 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001492 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1493 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001494 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001495 * of Layer 2 headers.
1496 *
1497 * @flags: Interface flags (a la BSD)
1498 * @priv_flags: Like 'flags' but invisible to userspace,
1499 * see if.h for the definitions
1500 * @gflags: Global flags ( kept as legacy )
1501 * @padded: How much padding added by alloc_netdev()
1502 * @operstate: RFC2863 operstate
1503 * @link_mode: Mapping policy to operstate
1504 * @if_port: Selectable AUI, TP, ...
1505 * @dma: DMA channel
1506 * @mtu: Interface MTU value
Jarod Wilson61e84622016-10-07 22:04:33 -04001507 * @min_mtu: Interface Minimum MTU value
1508 * @max_mtu: Interface Maximum MTU value
Karoly Kemeny536721b2014-07-30 20:27:36 +02001509 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001510 * @hard_header_len: Maximum hardware header length.
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05001511 * @min_header_len: Minimum hardware header length
Karoly Kemeny536721b2014-07-30 20:27:36 +02001512 *
1513 * @needed_headroom: Extra headroom the hardware may need, but not in all
1514 * cases can this be guaranteed
1515 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1516 * cases can this be guaranteed. Some cases also use
1517 * LL_MAX_HEADER instead to allocate the skb
1518 *
1519 * interface address info:
1520 *
1521 * @perm_addr: Permanent hw address
1522 * @addr_assign_type: Hw address assignment type
1523 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001524 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001525 * @dev_id: Used to differentiate devices that share
1526 * the same link layer address
1527 * @dev_port: Used to differentiate devices that share
1528 * the same function
1529 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001530 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001531 * has been enabled due to the need to listen to
1532 * additional unicast addresses in a device that
1533 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001534 * @uc: unicast mac addresses
1535 * @mc: multicast mac addresses
1536 * @dev_addrs: list of device hw addresses
1537 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001538 * @promiscuity: Number of times the NIC is told to work in
1539 * promiscuous mode; if it becomes 0 the NIC will
1540 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001541 * @allmulti: Counter, enables or disables allmulticast mode
1542 *
1543 * @vlan_info: VLAN info
1544 * @dsa_ptr: dsa specific data
1545 * @tipc_ptr: TIPC specific data
1546 * @atalk_ptr: AppleTalk link
1547 * @ip_ptr: IPv4 specific data
1548 * @dn_ptr: DECnet specific data
1549 * @ip6_ptr: IPv6 specific data
1550 * @ax25_ptr: AX.25 specific data
1551 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1552 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001553 * @dev_addr: Hw address (before bcast,
1554 * because most packets are unicast)
1555 *
1556 * @_rx: Array of RX queues
1557 * @num_rx_queues: Number of RX queues
1558 * allocated at register_netdev() time
1559 * @real_num_rx_queues: Number of RX queues currently active in device
1560 *
1561 * @rx_handler: handler for received packets
1562 * @rx_handler_data: XXX: need comments on this one
1563 * @ingress_queue: XXX: need comments on this one
1564 * @broadcast: hw bcast address
1565 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001566 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1567 * indexed by RX queue number. Assigned by driver.
1568 * This must only be set if the ndo_rx_flow_steer
1569 * operation is defined
1570 * @index_hlist: Device index hash chain
1571 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001572 * @_tx: Array of TX queues
1573 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1574 * @real_num_tx_queues: Number of TX queues currently active in device
1575 * @qdisc: Root qdisc from userspace point of view
1576 * @tx_queue_len: Max frames per queue allowed
1577 * @tx_global_lock: XXX: need comments on this one
1578 *
1579 * @xps_maps: XXX: need comments on this one
1580 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001581 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001582 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001583 * @watchdog_timer: List of timers
1584 *
1585 * @pcpu_refcnt: Number of references to this device
1586 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001587 * @link_watch_list: XXX: need comments on this one
1588 *
1589 * @reg_state: Register/unregister state machine
1590 * @dismantle: Device is going to be freed
1591 * @rtnl_link_state: This enum represents the phases of creating
1592 * a new link
1593 *
1594 * @destructor: Called from unregister,
1595 * can be used to call free_netdev
1596 * @npinfo: XXX: need comments on this one
1597 * @nd_net: Network namespace this network device is inside
1598 *
1599 * @ml_priv: Mid-layer private
1600 * @lstats: Loopback statistics
1601 * @tstats: Tunnel statistics
1602 * @dstats: Dummy statistics
1603 * @vstats: Virtual ethernet statistics
1604 *
1605 * @garp_port: GARP
1606 * @mrp_port: MRP
1607 *
1608 * @dev: Class/net/name entry
1609 * @sysfs_groups: Space for optional device, statistics and wireless
1610 * sysfs groups
1611 *
1612 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1613 * @rtnl_link_ops: Rtnl_link_ops
1614 *
1615 * @gso_max_size: Maximum size of generic segmentation offload
1616 * @gso_max_segs: Maximum number of segments that can be passed to the
1617 * NIC for GSO
1618 *
1619 * @dcbnl_ops: Data Center Bridging netlink ops
1620 * @num_tc: Number of traffic classes in the net device
1621 * @tc_to_txq: XXX: need comments on this one
Randy Dunlap920c1cd2016-11-21 18:28:36 -08001622 * @prio_tc_map: XXX: need comments on this one
Karoly Kemeny536721b2014-07-30 20:27:36 +02001623 *
1624 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1625 *
1626 * @priomap: XXX: need comments on this one
1627 * @phydev: Physical device may attach itself
1628 * for hardware timestamping
1629 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001630 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1631 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001632 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001633 * @proto_down: protocol port state information can be sent to the
1634 * switch driver and used to set the phys state of the
1635 * switch port.
1636 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 * FIXME: cleanup struct net_device such that network protocol info
1638 * moves out.
1639 */
1640
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001641struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001643 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001644 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 /*
1646 * I/O specific fields
1647 * FIXME: Merge these and struct ifmap into one
1648 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001649 unsigned long mem_end;
1650 unsigned long mem_start;
1651 unsigned long base_addr;
1652 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001654 atomic_t carrier_changes;
1655
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001657 * Some hardware also needs these fields (state,dev_list,
1658 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 * part of the usual set specified in Space.c.
1660 */
1661
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 unsigned long state;
1663
Pavel Emelianov7562f872007-05-03 15:13:45 -07001664 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001665 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001666 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001667 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001668 struct list_head ptype_all;
1669 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001670
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001671 struct {
1672 struct list_head upper;
1673 struct list_head lower;
1674 } adj_list;
1675
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001676 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001677 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001678 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001679 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001680 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001681 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001682 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001683
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001685 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686
Rusty Russellc45d2862007-03-28 14:29:08 -07001687 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001688
Eric Dumazet015f0682014-03-27 08:45:56 -07001689 atomic_long_t rx_dropped;
1690 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001691 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692
Johannes Bergb86e0282007-04-26 20:48:23 -07001693#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001694 const struct iw_handler_def *wireless_handlers;
1695 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001696#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001697 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001698 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001699#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001700 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001701#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001702#ifdef CONFIG_NET_L3_MASTER_DEV
1703 const struct l3mdev_ops *l3mdev_ops;
1704#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001705#if IS_ENABLED(CONFIG_IPV6)
1706 const struct ndisc_ops *ndisc_ops;
1707#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708
Steffen Klassertd77e38e2017-04-14 10:06:10 +02001709#ifdef CONFIG_XFRM
1710 const struct xfrmdev_ops *xfrmdev_ops;
1711#endif
1712
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001713 const struct header_ops *header_ops;
1714
Karoly Kemeny536721b2014-07-30 20:27:36 +02001715 unsigned int flags;
1716 unsigned int priv_flags;
1717
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001719 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720
Karoly Kemeny536721b2014-07-30 20:27:36 +02001721 unsigned char operstate;
1722 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001723
Karoly Kemeny536721b2014-07-30 20:27:36 +02001724 unsigned char if_port;
1725 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001726
Karoly Kemeny536721b2014-07-30 20:27:36 +02001727 unsigned int mtu;
Jarod Wilson61e84622016-10-07 22:04:33 -04001728 unsigned int min_mtu;
1729 unsigned int max_mtu;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001730 unsigned short type;
1731 unsigned short hard_header_len;
Alexey Dobriyand92be7a2017-04-10 11:25:26 +03001732 unsigned char min_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733
Johannes Bergf5184d22008-05-12 20:48:31 -07001734 unsigned short needed_headroom;
1735 unsigned short needed_tailroom;
1736
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001738 unsigned char perm_addr[MAX_ADDR_LEN];
1739 unsigned char addr_assign_type;
1740 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001741 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001742 unsigned short dev_id;
1743 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001744 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001745 unsigned char name_assign_type;
1746 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001747 struct netdev_hw_addr_list uc;
1748 struct netdev_hw_addr_list mc;
1749 struct netdev_hw_addr_list dev_addrs;
1750
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001751#ifdef CONFIG_SYSFS
1752 struct kset *queues_kset;
1753#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001754 unsigned int promiscuity;
1755 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001758 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001759
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001760#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001761 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001762#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001763#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001764 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001765#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001766#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001767 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001768#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001769 void *atalk_ptr;
1770 struct in_device __rcu *ip_ptr;
1771 struct dn_dev __rcu *dn_ptr;
1772 struct inet6_dev __rcu *ip6_ptr;
1773 void *ax25_ptr;
1774 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001775 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001776#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1777 struct mpls_dev __rcu *mpls_ptr;
1778#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001780/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001781 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001782 */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001783 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001784 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001785
Michael Daltona953be52014-01-16 22:23:28 -08001786#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001787 struct netdev_rx_queue *_rx;
1788
Tom Herbert0a9627f2010-03-16 08:03:29 +00001789 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001790 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001791#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001792
Eric Dumazet7acedaf2017-04-25 11:36:52 -07001793 struct bpf_prog __rcu *xdp_prog;
Eric Dumazet3b47d302014-11-06 21:09:44 -08001794 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001795 rx_handler_func_t __rcu *rx_handler;
1796 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001797
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001798#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001799 struct tcf_proto __rcu *ingress_cl_list;
1800#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001801 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001802#ifdef CONFIG_NETFILTER_INGRESS
Aaron Conolee3b37f12016-09-21 11:35:07 -04001803 struct nf_hook_entry __rcu *nf_hooks_ingress;
Pablo Neirae687ad62015-05-13 18:19:38 +02001804#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001805
Karoly Kemeny536721b2014-07-30 20:27:36 +02001806 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001807#ifdef CONFIG_RFS_ACCEL
1808 struct cpu_rmap *rx_cpu_rmap;
1809#endif
1810 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001811
1812/*
1813 * Cache lines mostly used on transmit path
1814 */
David S. Millere8a04642008-07-17 00:34:19 -07001815 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1816 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001817 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001818 struct Qdisc *qdisc;
Jiri Kosina59cc1f62016-08-10 11:05:15 +02001819#ifdef CONFIG_NET_SCHED
1820 DECLARE_HASHTABLE (qdisc_hash, 4);
1821#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001822 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001823 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001824 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001825
Tom Herbertbf264142010-11-26 08:36:09 +00001826#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001827 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001828#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001829#ifdef CONFIG_NET_CLS_ACT
1830 struct tcf_proto __rcu *egress_cl_list;
1831#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001832
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001833 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001834 struct timer_list watchdog_timer;
1835
Eric Dumazet29b44332010-10-11 10:22:12 +00001836 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838
Eric Dumazete014deb2009-11-17 05:59:21 +00001839 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001840
Linus Torvalds1da177e2005-04-16 15:20:36 -07001841 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001842 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843 NETREG_UNREGISTERING, /* called unregister_netdevice */
1844 NETREG_UNREGISTERED, /* completed unregister todo */
1845 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001846 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001847 } reg_state:8;
1848
Karoly Kemeny536721b2014-07-30 20:27:36 +02001849 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001850
1851 enum {
1852 RTNL_LINK_INITIALIZED,
1853 RTNL_LINK_INITIALIZING,
1854 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001856 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001859 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860#endif
David S. Millereae792b2008-07-15 03:03:33 -07001861
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001862 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001863
David S. Miller49517042008-05-12 03:29:11 -07001864 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001865 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001866 void *ml_priv;
1867 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001868 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001869 struct pcpu_dstats __percpu *dstats;
1870 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001871 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001872
Tobias Klauserfb585b42017-02-10 16:43:50 +01001873#if IS_ENABLED(CONFIG_GARP)
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001874 struct garp_port __rcu *garp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001875#endif
1876#if IS_ENABLED(CONFIG_MRP)
David Wardfebf0182013-02-08 17:17:06 +00001877 struct mrp_port __rcu *mrp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001878#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001880 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001881 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001882 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001883
Patrick McHardy38f7b872007-06-13 12:03:51 -07001884 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001885
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001886 /* for setting kernel sock attribute on TCP connection setup */
1887#define GSO_MAX_SIZE 65536
1888 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001889#define GSO_MAX_SEGS 65535
1890 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001891
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001892#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001893 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001894#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001895 u8 num_tc;
1896 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1897 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001898
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001899#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001900 unsigned int fcoe_ddp_xid;
1901#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001902#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001903 struct netprio_map __rcu *priomap;
1904#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001905 struct phy_device *phydev;
1906 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001907 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001908 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001910#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911
David S. Millerb5cdae32017-04-18 15:36:58 -04001912static inline bool netif_elide_gro(const struct net_device *dev)
1913{
1914 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1915 return true;
1916 return false;
1917}
1918
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920
David S. Millere8a04642008-07-17 00:34:19 -07001921static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001922int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1923{
1924 return dev->prio_tc_map[prio & TC_BITMASK];
1925}
1926
1927static inline
1928int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1929{
1930 if (tc >= dev->num_tc)
1931 return -EINVAL;
1932
1933 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1934 return 0;
1935}
1936
Alexander Duyck8d059b02016-10-28 11:43:49 -04001937int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
Alexander Duyck9cf1f6a2016-10-28 11:43:20 -04001938void netdev_reset_tc(struct net_device *dev);
1939int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1940int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
John Fastabend4f57c082011-01-17 08:06:04 +00001941
1942static inline
1943int netdev_get_num_tc(struct net_device *dev)
1944{
1945 return dev->num_tc;
1946}
1947
1948static inline
David S. Millere8a04642008-07-17 00:34:19 -07001949struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1950 unsigned int index)
1951{
1952 return &dev->_tx[index];
1953}
1954
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001955static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1956 const struct sk_buff *skb)
1957{
1958 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1959}
1960
David S. Millere8a04642008-07-17 00:34:19 -07001961static inline void netdev_for_each_tx_queue(struct net_device *dev,
1962 void (*f)(struct net_device *,
1963 struct netdev_queue *,
1964 void *),
1965 void *arg)
1966{
1967 unsigned int i;
1968
1969 for (i = 0; i < dev->num_tx_queues; i++)
1970 f(dev, &dev->_tx[i], arg);
1971}
1972
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001973#define netdev_lockdep_set_classes(dev) \
1974{ \
1975 static struct lock_class_key qdisc_tx_busylock_key; \
1976 static struct lock_class_key qdisc_running_key; \
1977 static struct lock_class_key qdisc_xmit_lock_key; \
1978 static struct lock_class_key dev_addr_list_lock_key; \
1979 unsigned int i; \
1980 \
1981 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1982 (dev)->qdisc_running_key = &qdisc_running_key; \
1983 lockdep_set_class(&(dev)->addr_list_lock, \
1984 &dev_addr_list_lock_key); \
1985 for (i = 0; i < (dev)->num_tx_queues; i++) \
1986 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1987 &qdisc_xmit_lock_key); \
1988}
1989
Joe Perchesf629d202013-09-26 14:48:15 -07001990struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001991 struct sk_buff *skb,
1992 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001993
Paolo Abeni871b6422016-02-26 10:45:37 +01001994/* returns the headroom that the master device needs to take in account
1995 * when forwarding to this dev
1996 */
1997static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1998{
1999 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2000}
2001
2002static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2003{
2004 if (dev->netdev_ops->ndo_set_rx_headroom)
2005 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2006}
2007
2008/* set the device rx headroom to the dev's default */
2009static inline void netdev_reset_rx_headroom(struct net_device *dev)
2010{
2011 netdev_set_rx_headroom(dev, -1);
2012}
2013
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002014/*
2015 * Net namespace inlines
2016 */
2017static inline
2018struct net *dev_net(const struct net_device *dev)
2019{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002020 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002021}
2022
2023static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002024void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002025{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05002026 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002027}
2028
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002029/**
2030 * netdev_priv - access network device private data
2031 * @dev: network device
2032 *
2033 * Get network device private data
2034 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002035static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002037 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038}
2039
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040/* Set the sysfs physical device reference for the network logical device
2041 * if set prior to registration will cause a symlink during initialization.
2042 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002043#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044
Marcel Holtmann384912e2009-08-31 21:08:19 +00002045/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002046 * fine-grained identification of different network device types. For
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002047 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
Marcel Holtmann384912e2009-08-31 21:08:19 +00002048 */
2049#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2050
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002051/* Default NAPI poll() weight
2052 * Device drivers are strongly advised to not use bigger value
2053 */
2054#define NAPI_POLL_WEIGHT 64
2055
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002056/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002057 * netif_napi_add - initialize a NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002058 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002059 * @napi: NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002060 * @poll: polling function
2061 * @weight: default weight
2062 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002063 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2064 * *any* of the other NAPI-related functions.
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002065 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002066void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2067 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002068
Alexander Duyckd8156532008-07-08 15:13:05 -07002069/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002070 * netif_tx_napi_add - initialize a NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002071 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002072 * @napi: NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002073 * @poll: polling function
2074 * @weight: default weight
2075 *
2076 * This variant of netif_napi_add() should be used from drivers using NAPI
2077 * to exclusively poll a TX queue.
2078 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2079 */
2080static inline void netif_tx_napi_add(struct net_device *dev,
2081 struct napi_struct *napi,
2082 int (*poll)(struct napi_struct *, int),
2083 int weight)
2084{
2085 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2086 netif_napi_add(dev, napi, poll, weight);
2087}
2088
2089/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002090 * netif_napi_del - remove a NAPI context
2091 * @napi: NAPI context
Alexander Duyckd8156532008-07-08 15:13:05 -07002092 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002093 * netif_napi_del() removes a NAPI context from the network device NAPI list
Alexander Duyckd8156532008-07-08 15:13:05 -07002094 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002095void netif_napi_del(struct napi_struct *napi);
2096
2097struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002098 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002099 void *frag0;
Herbert Xu78a478d2009-05-26 18:50:21 +00002100
Herbert Xu74895942009-05-26 18:50:27 +00002101 /* Length of frag0. */
2102 unsigned int frag0_len;
2103
Herbert Xu86911732009-01-29 14:19:50 +00002104 /* This indicates where we are processing relative to skb->data. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002105 int data_offset;
Herbert Xu86911732009-01-29 14:19:50 +00002106
Herbert Xud565b0a2008-12-15 23:38:52 -08002107 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002108 u16 flush;
2109
2110 /* Save the IP ID here and check when we get to the transport layer */
2111 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002112
2113 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002114 u16 count;
2115
Tom Herbert15e23962015-02-10 16:30:31 -08002116 /* Start offset for remote checksum offload */
2117 u16 gro_remcsum_start;
2118
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002119 /* jiffies when first packet was created/queued */
2120 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002121
Tom Herbertafe93322014-09-17 12:25:57 -07002122 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002123 u16 proto;
2124
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002125 /* This is non-zero if the packet may be of the same flow. */
2126 u8 same_flow:1;
2127
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002128 /* Used in tunnel GRO receive */
2129 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002130
2131 /* GRO checksum is valid */
2132 u8 csum_valid:1;
2133
Tom Herbert662880f2014-08-27 21:26:56 -07002134 /* Number of checksums via CHECKSUM_UNNECESSARY */
2135 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002136
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002137 /* Free the skb? */
2138 u8 free:2;
2139#define NAPI_GRO_FREE 1
2140#define NAPI_GRO_FREE_STOLEN_HEAD 2
2141
Tom Herbertefc98d02014-10-03 15:48:08 -07002142 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2143 u8 is_ipv6:1;
2144
Alexander Duycka0ca1532016-04-05 09:13:39 -07002145 /* Used in GRE, set in fou/gue_gro_receive */
2146 u8 is_fou:1;
2147
Alexander Duyck15305452016-04-10 21:44:57 -04002148 /* Used to determine if flush_id can be ignored */
2149 u8 is_atomic:1;
2150
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002151 /* Number of gro_receive callbacks this packet already went through */
2152 u8 recursion_counter:4;
2153
2154 /* 1 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002155
Jerry Chubf5a7552014-01-07 10:23:19 -08002156 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2157 __wsum csum;
2158
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002159 /* used in skb_gro_receive() slow path */
2160 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002161};
2162
2163#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002164
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002165#define GRO_RECURSION_LIMIT 15
2166static inline int gro_recursion_inc_test(struct sk_buff *skb)
2167{
2168 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2169}
2170
2171typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2172static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2173 struct sk_buff **head,
2174 struct sk_buff *skb)
2175{
2176 if (unlikely(gro_recursion_inc_test(skb))) {
2177 NAPI_GRO_CB(skb)->flush |= 1;
2178 return NULL;
2179 }
2180
2181 return cb(head, skb);
2182}
2183
2184typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2185 struct sk_buff *);
2186static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2187 struct sock *sk,
2188 struct sk_buff **head,
2189 struct sk_buff *skb)
2190{
2191 if (unlikely(gro_recursion_inc_test(skb))) {
2192 NAPI_GRO_CB(skb)->flush |= 1;
2193 return NULL;
2194 }
2195
2196 return cb(sk, head, skb);
2197}
2198
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002200 __be16 type; /* This is really htons(ether_type). */
2201 struct net_device *dev; /* NULL is wildcarded here */
2202 int (*func) (struct sk_buff *,
2203 struct net_device *,
2204 struct packet_type *,
2205 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002206 bool (*id_match)(struct packet_type *ptype,
2207 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002208 void *af_packet_priv;
2209 struct list_head list;
2210};
2211
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002212struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2214 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002216 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002217 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002218};
2219
2220struct packet_offload {
2221 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002222 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002223 struct offload_callbacks callbacks;
2224 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225};
2226
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002227/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002228struct pcpu_sw_netstats {
2229 u64 rx_packets;
2230 u64 rx_bytes;
2231 u64 tx_packets;
2232 u64 tx_bytes;
2233 struct u64_stats_sync syncp;
2234};
2235
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002236#define __netdev_alloc_pcpu_stats(type, gfp) \
2237({ \
2238 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2239 if (pcpu_stats) { \
2240 int __cpu; \
2241 for_each_possible_cpu(__cpu) { \
2242 typeof(type) *stat; \
2243 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2244 u64_stats_init(&stat->syncp); \
2245 } \
2246 } \
2247 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002248})
2249
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002250#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002251 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002252
Jiri Pirko764f5e52015-12-03 12:12:12 +01002253enum netdev_lag_tx_type {
2254 NETDEV_LAG_TX_TYPE_UNKNOWN,
2255 NETDEV_LAG_TX_TYPE_RANDOM,
2256 NETDEV_LAG_TX_TYPE_BROADCAST,
2257 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2258 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2259 NETDEV_LAG_TX_TYPE_HASH,
2260};
2261
2262struct netdev_lag_upper_info {
2263 enum netdev_lag_tx_type tx_type;
2264};
2265
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002266struct netdev_lag_lower_state_info {
2267 u8 link_up : 1,
2268 tx_enabled : 1;
2269};
2270
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271#include <linux/notifier.h>
2272
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002273/* netdevice notifier chain. Please remember to update the rtnetlink
2274 * notification exclusion list in rtnetlink_event() when adding new
2275 * types.
2276 */
2277#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2278#define NETDEV_DOWN 0x0002
2279#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2280 detected a hardware crash and restarted
2281 - we can use this eg to kick tcp sessions
2282 once done */
2283#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2284#define NETDEV_REGISTER 0x0005
2285#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002286#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002287#define NETDEV_CHANGEADDR 0x0008
2288#define NETDEV_GOING_DOWN 0x0009
2289#define NETDEV_CHANGENAME 0x000A
2290#define NETDEV_FEAT_CHANGE 0x000B
2291#define NETDEV_BONDING_FAILOVER 0x000C
2292#define NETDEV_PRE_UP 0x000D
2293#define NETDEV_PRE_TYPE_CHANGE 0x000E
2294#define NETDEV_POST_TYPE_CHANGE 0x000F
2295#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002296#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002297#define NETDEV_RELEASE 0x0012
2298#define NETDEV_NOTIFY_PEERS 0x0013
2299#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002300#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002301#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002302#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002303#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002304#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002305#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002306#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002307#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Jason Wang08294a22016-06-30 14:45:35 +08002308#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002309
Joe Perchesf629d202013-09-26 14:48:15 -07002310int register_netdevice_notifier(struct notifier_block *nb);
2311int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002312
2313struct netdev_notifier_info {
2314 struct net_device *dev;
2315};
2316
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002317struct netdev_notifier_change_info {
2318 struct netdev_notifier_info info; /* must be first */
2319 unsigned int flags_changed;
2320};
2321
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002322struct netdev_notifier_changeupper_info {
2323 struct netdev_notifier_info info; /* must be first */
2324 struct net_device *upper_dev; /* new upper dev */
2325 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002326 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002327 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002328};
2329
Jiri Pirko04d48262015-12-03 12:12:15 +01002330struct netdev_notifier_changelowerstate_info {
2331 struct netdev_notifier_info info; /* must be first */
2332 void *lower_state_info; /* is lower dev state */
2333};
2334
Cong Wang75538c22013-05-29 11:30:50 +08002335static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2336 struct net_device *dev)
2337{
2338 info->dev = dev;
2339}
2340
Jiri Pirko351638e2013-05-28 01:30:21 +00002341static inline struct net_device *
2342netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2343{
2344 return info->dev;
2345}
2346
Joe Perchesf629d202013-09-26 14:48:15 -07002347int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002348
2349
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350extern rwlock_t dev_base_lock; /* Device list lock */
2351
Eric W. Biederman881d9662007-09-17 11:56:21 -07002352#define for_each_netdev(net, d) \
2353 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002354#define for_each_netdev_reverse(net, d) \
2355 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002356#define for_each_netdev_rcu(net, d) \
2357 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002358#define for_each_netdev_safe(net, d, n) \
2359 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2360#define for_each_netdev_continue(net, d) \
2361 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002362#define for_each_netdev_continue_rcu(net, d) \
2363 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002364#define for_each_netdev_in_bond_rcu(bond, slave) \
2365 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002366 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002367#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2368
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002369static inline struct net_device *next_net_device(struct net_device *dev)
2370{
2371 struct list_head *lh;
2372 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002373
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002374 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002375 lh = dev->dev_list.next;
2376 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2377}
2378
Eric Dumazetce81b762009-11-11 17:34:30 +00002379static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2380{
2381 struct list_head *lh;
2382 struct net *net;
2383
2384 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002385 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002386 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2387}
2388
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002389static inline struct net_device *first_net_device(struct net *net)
2390{
2391 return list_empty(&net->dev_base_head) ? NULL :
2392 net_device_entry(net->dev_base_head.next);
2393}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002394
Eric Dumazetccf43432011-01-26 18:08:02 +00002395static inline struct net_device *first_net_device_rcu(struct net *net)
2396{
2397 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2398
2399 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2400}
2401
Joe Perchesf629d202013-09-26 14:48:15 -07002402int netdev_boot_setup_check(struct net_device *dev);
2403unsigned long netdev_boot_base(const char *prefix, int unit);
2404struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2405 const char *hwaddr);
2406struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2407struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2408void dev_add_pack(struct packet_type *pt);
2409void dev_remove_pack(struct packet_type *pt);
2410void __dev_remove_pack(struct packet_type *pt);
2411void dev_add_offload(struct packet_offload *po);
2412void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002413
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002414int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002415int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002416struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2417 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002418struct net_device *dev_get_by_name(struct net *net, const char *name);
2419struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2420struct net_device *__dev_get_by_name(struct net *net, const char *name);
2421int dev_alloc_name(struct net_device *dev, const char *name);
2422int dev_open(struct net_device *dev);
2423int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002424int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002425void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002426int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002427int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002428int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002429int register_netdevice(struct net_device *dev);
2430void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2431void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002432static inline void unregister_netdevice(struct net_device *dev)
2433{
2434 unregister_netdevice_queue(dev, NULL);
2435}
2436
Joe Perchesf629d202013-09-26 14:48:15 -07002437int netdev_refcnt_read(const struct net_device *dev);
2438void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002439void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002440void synchronize_net(void);
2441int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002442
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002443DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002444#define XMIT_RECURSION_LIMIT 10
2445
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002446static inline int dev_recursion_level(void)
2447{
2448 return this_cpu_read(xmit_recursion);
2449}
2450
Joe Perchesf629d202013-09-26 14:48:15 -07002451struct net_device *dev_get_by_index(struct net *net, int ifindex);
2452struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2453struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2454int netdev_get_name(struct net *net, char *name, int ifindex);
2455int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002456int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002457
2458static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2459{
2460 return NAPI_GRO_CB(skb)->data_offset;
2461}
2462
2463static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2464{
2465 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2466}
2467
2468static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2469{
2470 NAPI_GRO_CB(skb)->data_offset += len;
2471}
2472
Herbert Xua5b1cf22009-05-26 18:50:28 +00002473static inline void *skb_gro_header_fast(struct sk_buff *skb,
2474 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002475{
Herbert Xu78a478d2009-05-26 18:50:21 +00002476 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002477}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002478
Herbert Xua5b1cf22009-05-26 18:50:28 +00002479static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2480{
2481 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2482}
2483
Herbert Xu57ea52a2017-01-10 12:24:15 -08002484static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2485{
2486 NAPI_GRO_CB(skb)->frag0 = NULL;
2487 NAPI_GRO_CB(skb)->frag0_len = 0;
2488}
2489
Herbert Xua5b1cf22009-05-26 18:50:28 +00002490static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2491 unsigned int offset)
2492{
Herbert Xu17dd7592011-07-27 06:16:28 -07002493 if (!pskb_may_pull(skb, hlen))
2494 return NULL;
2495
Herbert Xu57ea52a2017-01-10 12:24:15 -08002496 skb_gro_frag0_invalidate(skb);
Herbert Xu17dd7592011-07-27 06:16:28 -07002497 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002498}
2499
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002500static inline void *skb_gro_network_header(struct sk_buff *skb)
2501{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002502 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2503 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002504}
2505
Jerry Chubf5a7552014-01-07 10:23:19 -08002506static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2507 const void *start, unsigned int len)
2508{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002509 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002510 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2511 csum_partial(start, len, 0));
2512}
2513
Tom Herbert573e8fc2014-08-22 13:33:47 -07002514/* GRO checksum functions. These are logical equivalents of the normal
2515 * checksum functions (in skbuff.h) except that they operate on the GRO
2516 * offsets and fields in sk_buff.
2517 */
2518
2519__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2520
Tom Herbert15e23962015-02-10 16:30:31 -08002521static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2522{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002523 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002524}
2525
Tom Herbert573e8fc2014-08-22 13:33:47 -07002526static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2527 bool zero_okay,
2528 __sum16 check)
2529{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002530 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2531 skb_checksum_start_offset(skb) <
2532 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002533 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002534 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002535 (!zero_okay || check));
2536}
2537
2538static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2539 __wsum psum)
2540{
2541 if (NAPI_GRO_CB(skb)->csum_valid &&
2542 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2543 return 0;
2544
2545 NAPI_GRO_CB(skb)->csum = psum;
2546
2547 return __skb_gro_checksum_complete(skb);
2548}
2549
Tom Herbert573e8fc2014-08-22 13:33:47 -07002550static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2551{
Tom Herbert662880f2014-08-27 21:26:56 -07002552 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2553 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2554 NAPI_GRO_CB(skb)->csum_cnt--;
2555 } else {
2556 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2557 * verified a new top level checksum or an encapsulated one
2558 * during GRO. This saves work if we fallback to normal path.
2559 */
2560 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002561 }
2562}
2563
2564#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2565 compute_pseudo) \
2566({ \
2567 __sum16 __ret = 0; \
2568 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2569 __ret = __skb_gro_checksum_validate_complete(skb, \
2570 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002571 if (__ret) \
2572 __skb_mark_checksum_bad(skb); \
2573 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002574 skb_gro_incr_csum_unnecessary(skb); \
2575 __ret; \
2576})
2577
2578#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2579 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2580
2581#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2582 compute_pseudo) \
2583 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2584
2585#define skb_gro_checksum_simple_validate(skb) \
2586 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2587
Tom Herbertd96535a2014-08-31 15:12:42 -07002588static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2589{
2590 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2591 !NAPI_GRO_CB(skb)->csum_valid);
2592}
2593
2594static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2595 __sum16 check, __wsum pseudo)
2596{
2597 NAPI_GRO_CB(skb)->csum = ~pseudo;
2598 NAPI_GRO_CB(skb)->csum_valid = 1;
2599}
2600
2601#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2602do { \
2603 if (__skb_gro_checksum_convert_check(skb)) \
2604 __skb_gro_checksum_convert(skb, check, \
2605 compute_pseudo(skb, proto)); \
2606} while (0)
2607
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002608struct gro_remcsum {
2609 int offset;
2610 __wsum delta;
2611};
2612
2613static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2614{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002615 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002616 grc->delta = 0;
2617}
2618
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002619static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2620 unsigned int off, size_t hdrlen,
2621 int start, int offset,
2622 struct gro_remcsum *grc,
2623 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002624{
2625 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002626 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002627
2628 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2629
Tom Herbert15e23962015-02-10 16:30:31 -08002630 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002631 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2632 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002633 }
2634
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002635 ptr = skb_gro_header_fast(skb, off);
2636 if (skb_gro_header_hard(skb, off + plen)) {
2637 ptr = skb_gro_header_slow(skb, off + plen, off);
2638 if (!ptr)
2639 return NULL;
2640 }
2641
2642 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2643 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002644
2645 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002646 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002647
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002648 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002649 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002650
2651 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002652}
2653
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002654static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2655 struct gro_remcsum *grc)
2656{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002657 void *ptr;
2658 size_t plen = grc->offset + sizeof(u16);
2659
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002660 if (!grc->delta)
2661 return;
2662
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002663 ptr = skb_gro_header_fast(skb, grc->offset);
2664 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2665 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2666 if (!ptr)
2667 return;
2668 }
2669
2670 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002671}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002672
Steffen Klassert25393d32017-02-15 09:39:44 +01002673#ifdef CONFIG_XFRM_OFFLOAD
2674static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2675{
2676 if (PTR_ERR(pp) != -EINPROGRESS)
2677 NAPI_GRO_CB(skb)->flush |= flush;
2678}
2679#else
Steffen Klassert5f114162017-02-15 09:39:39 +01002680static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2681{
2682 NAPI_GRO_CB(skb)->flush |= flush;
2683}
Steffen Klassert25393d32017-02-15 09:39:44 +01002684#endif
Steffen Klassert5f114162017-02-15 09:39:39 +01002685
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002686static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2687 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002688 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002689 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002690{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002691 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002692 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002693
2694 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002695}
2696
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002697static inline int dev_parse_header(const struct sk_buff *skb,
2698 unsigned char *haddr)
2699{
2700 const struct net_device *dev = skb->dev;
2701
Patrick McHardy1b833362007-10-18 05:09:28 -07002702 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002703 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002704 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002705}
2706
Willem de Bruijn2793a232016-03-09 21:58:32 -05002707/* ll_header must have at least hard_header_len allocated */
2708static inline bool dev_validate_header(const struct net_device *dev,
2709 char *ll_header, int len)
2710{
2711 if (likely(len >= dev->hard_header_len))
2712 return true;
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05002713 if (len < dev->min_header_len)
2714 return false;
Willem de Bruijn2793a232016-03-09 21:58:32 -05002715
2716 if (capable(CAP_SYS_RAWIO)) {
2717 memset(ll_header + len, 0, dev->hard_header_len - len);
2718 return true;
2719 }
2720
2721 if (dev->header_ops && dev->header_ops->validate)
2722 return dev->header_ops->validate(ll_header, len);
2723
2724 return false;
2725}
2726
Linus Torvalds1da177e2005-04-16 15:20:36 -07002727typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002728int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002729static inline int unregister_gifconf(unsigned int family)
2730{
2731 return register_gifconf(family, NULL);
2732}
2733
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002734#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002735#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002736struct sd_flow_limit {
2737 u64 count;
2738 unsigned int num_buckets;
2739 unsigned int history_head;
2740 u16 history[FLOW_LIMIT_HISTORY];
2741 u8 buckets[];
2742};
2743
2744extern int netdev_flow_limit_table_len;
2745#endif /* CONFIG_NET_FLOW_LIMIT */
2746
Linus Torvalds1da177e2005-04-16 15:20:36 -07002747/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002748 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002749 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002750struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002751 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002752 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002753
Changli Gaodee42872010-05-02 05:42:16 +00002754 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002755 unsigned int processed;
2756 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002757 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002758#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002759 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002760#endif
2761#ifdef CONFIG_NET_FLOW_LIMIT
2762 struct sd_flow_limit __rcu *flow_limit;
2763#endif
2764 struct Qdisc *output_queue;
2765 struct Qdisc **output_queue_tailp;
2766 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002767
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002768#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002769 /* input_queue_head should be written by cpu owning this struct,
2770 * and only read by other cpus. Worth using a cache line.
2771 */
2772 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2773
2774 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002775 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002776 struct softnet_data *rps_ipi_next;
2777 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002778 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002779#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002780 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002781 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002782 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002783
Linus Torvalds1da177e2005-04-16 15:20:36 -07002784};
2785
Tom Herbert76cc8b12010-05-20 18:37:59 +00002786static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002787{
2788#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002789 sd->input_queue_head++;
2790#endif
2791}
2792
2793static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2794 unsigned int *qtail)
2795{
2796#ifdef CONFIG_RPS
2797 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002798#endif
2799}
2800
Tom Herbert0a9627f2010-03-16 08:03:29 +00002801DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002802
Joe Perchesf629d202013-09-26 14:48:15 -07002803void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002804void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002805
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002806static inline void netif_tx_schedule_all(struct net_device *dev)
2807{
2808 unsigned int i;
2809
2810 for (i = 0; i < dev->num_tx_queues; i++)
2811 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2812}
2813
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002814static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002815{
Tom Herbert734664982011-11-28 16:32:44 +00002816 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002817}
2818
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002819/**
2820 * netif_start_queue - allow transmit
2821 * @dev: network device
2822 *
2823 * Allow upper layers to call the device hard_start_xmit routine.
2824 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002825static inline void netif_start_queue(struct net_device *dev)
2826{
David S. Millere8a04642008-07-17 00:34:19 -07002827 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002828}
2829
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002830static inline void netif_tx_start_all_queues(struct net_device *dev)
2831{
2832 unsigned int i;
2833
2834 for (i = 0; i < dev->num_tx_queues; i++) {
2835 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2836 netif_tx_start_queue(txq);
2837 }
2838}
2839
John Fastabend46e5da40a2014-09-12 20:04:52 -07002840void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002841
Dave Jonesd29f7492008-07-22 14:09:06 -07002842/**
2843 * netif_wake_queue - restart transmit
2844 * @dev: network device
2845 *
2846 * Allow upper layers to call the device hard_start_xmit routine.
2847 * Used for flow control when transmit resources are available.
2848 */
David S. Miller79d16382008-07-08 23:14:46 -07002849static inline void netif_wake_queue(struct net_device *dev)
2850{
David S. Millere8a04642008-07-17 00:34:19 -07002851 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002852}
2853
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002854static inline void netif_tx_wake_all_queues(struct net_device *dev)
2855{
2856 unsigned int i;
2857
2858 for (i = 0; i < dev->num_tx_queues; i++) {
2859 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2860 netif_tx_wake_queue(txq);
2861 }
2862}
2863
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002864static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002865{
Tom Herbert734664982011-11-28 16:32:44 +00002866 set_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_stop_queue - stop transmitted packets
2871 * @dev: network device
2872 *
2873 * Stop upper layers calling the device hard_start_xmit routine.
2874 * Used for flow control when transmit resources are unavailable.
2875 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002876static inline void netif_stop_queue(struct net_device *dev)
2877{
David S. Millere8a04642008-07-17 00:34:19 -07002878 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002879}
2880
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002881void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002882
David S. Miller4d295152012-03-07 21:02:35 -05002883static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002884{
Tom Herbert734664982011-11-28 16:32:44 +00002885 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002886}
2887
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002888/**
2889 * netif_queue_stopped - test if transmit queue is flowblocked
2890 * @dev: network device
2891 *
2892 * Test if transmit queue on device is currently unable to send.
2893 */
David S. Miller4d295152012-03-07 21:02:35 -05002894static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002895{
David S. Millere8a04642008-07-17 00:34:19 -07002896 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002897}
2898
David S. Miller4d295152012-03-07 21:02:35 -05002899static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002900{
Tom Herbert734664982011-11-28 16:32:44 +00002901 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2902}
2903
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002904static inline bool
2905netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002906{
2907 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2908}
2909
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002910static inline bool
2911netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2912{
2913 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2914}
2915
Eric Dumazet53511452014-10-08 08:19:27 -07002916/**
2917 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2918 * @dev_queue: pointer to transmit queue
2919 *
2920 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002921 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002922 */
2923static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2924{
2925#ifdef CONFIG_BQL
2926 prefetchw(&dev_queue->dql.num_queued);
2927#endif
2928}
2929
2930/**
2931 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2932 * @dev_queue: pointer to transmit queue
2933 *
2934 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002935 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002936 */
2937static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2938{
2939#ifdef CONFIG_BQL
2940 prefetchw(&dev_queue->dql.limit);
2941#endif
2942}
2943
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002944static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2945 unsigned int bytes)
2946{
Tom Herbert114cf582011-11-28 16:33:09 +00002947#ifdef CONFIG_BQL
2948 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002949
2950 if (likely(dql_avail(&dev_queue->dql) >= 0))
2951 return;
2952
2953 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2954
2955 /*
2956 * The XOFF flag must be set before checking the dql_avail below,
2957 * because in netdev_tx_completed_queue we update the dql_completed
2958 * before checking the XOFF flag.
2959 */
2960 smp_mb();
2961
2962 /* check again in case another CPU has just made room avail */
2963 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2964 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002965#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002966}
2967
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002968/**
2969 * netdev_sent_queue - report the number of bytes queued to hardware
2970 * @dev: network device
2971 * @bytes: number of bytes queued to the hardware device queue
2972 *
2973 * Report the number of bytes queued for sending/completion to the network
2974 * device hardware queue. @bytes should be a good approximation and should
2975 * exactly match netdev_completed_queue() @bytes
2976 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002977static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2978{
2979 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2980}
2981
2982static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002983 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002984{
Tom Herbert114cf582011-11-28 16:33:09 +00002985#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002986 if (unlikely(!bytes))
2987 return;
2988
2989 dql_completed(&dev_queue->dql, bytes);
2990
2991 /*
2992 * Without the memory barrier there is a small possiblity that
2993 * netdev_tx_sent_queue will miss the update and cause the queue to
2994 * be stopped forever
2995 */
2996 smp_mb();
2997
2998 if (dql_avail(&dev_queue->dql) < 0)
2999 return;
3000
3001 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3002 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00003003#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003004}
3005
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003006/**
3007 * netdev_completed_queue - report bytes and packets completed by device
3008 * @dev: network device
3009 * @pkts: actual number of packets sent over the medium
3010 * @bytes: actual number of bytes sent over the medium
3011 *
3012 * Report the number of bytes and packets transmitted by the network device
3013 * hardware queue over the physical medium, @bytes must exactly match the
3014 * @bytes amount passed to netdev_sent_queue()
3015 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003016static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00003017 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003018{
3019 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3020}
3021
3022static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3023{
Tom Herbert114cf582011-11-28 16:33:09 +00003024#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003025 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003026 dql_reset(&q->dql);
3027#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003028}
3029
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003030/**
3031 * netdev_reset_queue - reset the packets and bytes count of a network device
3032 * @dev_queue: network device
3033 *
3034 * Reset the bytes and packet count of a network device and clear the
3035 * software flow control OFF bit for this network device
3036 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003037static inline void netdev_reset_queue(struct net_device *dev_queue)
3038{
3039 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003040}
3041
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003042/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003043 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3044 * @dev: network device
3045 * @queue_index: given tx queue index
3046 *
3047 * Returns 0 if given tx queue index >= number of device tx queues,
3048 * otherwise returns the originally passed tx queue index.
3049 */
3050static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3051{
3052 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3053 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3054 dev->name, queue_index,
3055 dev->real_num_tx_queues);
3056 return 0;
3057 }
3058
3059 return queue_index;
3060}
3061
3062/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003063 * netif_running - test if up
3064 * @dev: network device
3065 *
3066 * Test if the device has been brought up.
3067 */
David S. Miller4d295152012-03-07 21:02:35 -05003068static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003069{
3070 return test_bit(__LINK_STATE_START, &dev->state);
3071}
3072
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003073/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003074 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003075 * stop, and a check if it's stopped. All other device management is
3076 * done at the overall netdevice level.
3077 * Also test the device if we're multiqueue.
3078 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003079
3080/**
3081 * netif_start_subqueue - allow sending packets on subqueue
3082 * @dev: network device
3083 * @queue_index: sub queue index
3084 *
3085 * Start individual transmit queue of a device with multiple transmit queues.
3086 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003087static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3088{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003089 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003090
3091 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003092}
3093
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003094/**
3095 * netif_stop_subqueue - stop sending packets on subqueue
3096 * @dev: network device
3097 * @queue_index: sub queue index
3098 *
3099 * Stop individual transmit queue of a device with multiple transmit queues.
3100 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003101static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3102{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003103 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003104 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003105}
3106
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003107/**
3108 * netif_subqueue_stopped - test status of subqueue
3109 * @dev: network device
3110 * @queue_index: sub queue index
3111 *
3112 * Check individual transmit queue of a device with multiple transmit queues.
3113 */
David S. Miller4d295152012-03-07 21:02:35 -05003114static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3115 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003116{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003117 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003118
3119 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003120}
3121
David S. Miller4d295152012-03-07 21:02:35 -05003122static inline bool netif_subqueue_stopped(const struct net_device *dev,
3123 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003124{
3125 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3126}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003127
Florian Fainelli738b35c2017-01-11 21:13:02 -08003128/**
3129 * netif_wake_subqueue - allow sending packets on subqueue
3130 * @dev: network device
3131 * @queue_index: sub queue index
3132 *
3133 * Resume individual transmit queue of a device with multiple transmit queues.
3134 */
3135static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3136{
3137 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3138
3139 netif_tx_wake_queue(txq);
3140}
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003141
Alexander Duyck537c00d2013-01-10 08:57:02 +00003142#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003143int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003144 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003145#else
3146static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003147 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003148 u16 index)
3149{
3150 return 0;
3151}
3152#endif
3153
Jiri Pirko5605c762015-05-12 14:56:12 +02003154u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3155 unsigned int num_tx_queues);
3156
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003157/*
3158 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3159 * as a distribution range limit for the returned value.
3160 */
3161static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003162 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003163{
3164 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3165}
3166
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003167/**
3168 * netif_is_multiqueue - test if device has multiple transmit queues
3169 * @dev: network device
3170 *
3171 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003172 */
David S. Miller4d295152012-03-07 21:02:35 -05003173static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003174{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003175 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003176}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003177
Joe Perchesf629d202013-09-26 14:48:15 -07003178int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003179
Michael Daltona953be52014-01-16 22:23:28 -08003180#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003181int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003182#else
3183static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3184 unsigned int rxq)
3185{
3186 return 0;
3187}
3188#endif
3189
Michael Daltona953be52014-01-16 22:23:28 -08003190#ifdef CONFIG_SYSFS
3191static inline unsigned int get_netdev_rx_queue_index(
3192 struct netdev_rx_queue *queue)
3193{
3194 struct net_device *dev = queue->dev;
3195 int index = queue - dev->_rx;
3196
3197 BUG_ON(index >= dev->num_rx_queues);
3198 return index;
3199}
3200#endif
3201
Yuval Mintz16917b82012-07-01 03:18:50 +00003202#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003203int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003204
Eric Dumazete6247022013-12-05 04:45:08 -08003205enum skb_free_reason {
3206 SKB_REASON_CONSUMED,
3207 SKB_REASON_DROPPED,
3208};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003209
Eric Dumazete6247022013-12-05 04:45:08 -08003210void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3211void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3212
3213/*
3214 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3215 * interrupt context or with hardware interrupts being disabled.
3216 * (in_irq() || irqs_disabled())
3217 *
3218 * We provide four helpers that can be used in following contexts :
3219 *
3220 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3221 * replacing kfree_skb(skb)
3222 *
3223 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3224 * Typically used in place of consume_skb(skb) in TX completion path
3225 *
3226 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3227 * replacing kfree_skb(skb)
3228 *
3229 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3230 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003231 */
Eric Dumazete6247022013-12-05 04:45:08 -08003232static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3233{
3234 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3235}
3236
3237static inline void dev_consume_skb_irq(struct sk_buff *skb)
3238{
3239 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3240}
3241
3242static inline void dev_kfree_skb_any(struct sk_buff *skb)
3243{
3244 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3245}
3246
3247static inline void dev_consume_skb_any(struct sk_buff *skb)
3248{
3249 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3250}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003251
Joe Perchesf629d202013-09-26 14:48:15 -07003252int netif_rx(struct sk_buff *skb);
3253int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003254int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003255gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3256void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3257struct sk_buff *napi_get_frags(struct napi_struct *napi);
3258gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003259struct packet_offload *gro_find_receive_by_type(__be16 type);
3260struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003261
3262static inline void napi_free_frags(struct napi_struct *napi)
3263{
3264 kfree_skb(napi->skb);
3265 napi->skb = NULL;
3266}
3267
Mahesh Bandewar24b27fc2016-09-01 22:18:34 -07003268bool netdev_is_rx_handler_busy(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003269int netdev_rx_handler_register(struct net_device *dev,
3270 rx_handler_func_t *rx_handler,
3271 void *rx_handler_data);
3272void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003273
Joe Perchesf629d202013-09-26 14:48:15 -07003274bool dev_valid_name(const char *name);
3275int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3276int dev_ethtool(struct net *net, struct ifreq *);
3277unsigned int dev_get_flags(const struct net_device *);
3278int __dev_change_flags(struct net_device *, unsigned int flags);
3279int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003280void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3281 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003282int dev_change_name(struct net_device *, const char *);
3283int dev_set_alias(struct net_device *, const char *, size_t);
3284int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3285int dev_set_mtu(struct net_device *, int);
3286void dev_set_group(struct net_device *, int);
3287int dev_set_mac_address(struct net_device *, struct sockaddr *);
3288int dev_change_carrier(struct net_device *, bool new_carrier);
3289int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003290 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003291int dev_get_phys_port_name(struct net_device *dev,
3292 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003293int dev_change_proto_down(struct net_device *dev, bool proto_down);
Daniel Borkmann85de8572016-11-28 23:16:54 +01003294int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003295struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003296struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3297 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08003298int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003299int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003300bool is_skb_forwardable(const struct net_device *dev,
3301 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003302
Martin KaFai Lau4e3264d2016-11-09 15:36:33 -08003303static __always_inline int ____dev_forward_skb(struct net_device *dev,
3304 struct sk_buff *skb)
3305{
3306 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3307 unlikely(!is_skb_forwardable(dev, skb))) {
3308 atomic_long_inc(&dev->rx_dropped);
3309 kfree_skb(skb);
3310 return NET_RX_DROP;
3311 }
3312
3313 skb_scrub_packet(skb, true);
3314 skb->priority = 0;
3315 return 0;
3316}
3317
David Ahern74b20582016-05-10 11:19:50 -07003318void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3319
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003320extern int netdev_budget;
Matthew Whitehead7acf8a12017-04-19 12:37:10 -04003321extern unsigned int netdev_budget_usecs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003322
3323/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003324void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003325
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003326/**
3327 * dev_put - release reference to device
3328 * @dev: network device
3329 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003330 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003331 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003332static inline void dev_put(struct net_device *dev)
3333{
Christoph Lameter933393f2011-12-22 11:58:51 -06003334 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003335}
3336
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003337/**
3338 * dev_hold - get reference to device
3339 * @dev: network device
3340 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003341 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003342 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003343static inline void dev_hold(struct net_device *dev)
3344{
Christoph Lameter933393f2011-12-22 11:58:51 -06003345 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003346}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003347
3348/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3349 * and _off may be called from IRQ context, but it is caller
3350 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003351 *
3352 * The name carrier is inappropriate, these functions should really be
3353 * called netif_lowerlayer_*() because they represent the state of any
3354 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003355 */
3356
Joe Perchesf629d202013-09-26 14:48:15 -07003357void linkwatch_init_dev(struct net_device *dev);
3358void linkwatch_fire_event(struct net_device *dev);
3359void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003360
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003361/**
3362 * netif_carrier_ok - test if carrier present
3363 * @dev: network device
3364 *
3365 * Check if carrier is present on device
3366 */
David S. Miller4d295152012-03-07 21:02:35 -05003367static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003368{
3369 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3370}
3371
Joe Perchesf629d202013-09-26 14:48:15 -07003372unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003373
Joe Perchesf629d202013-09-26 14:48:15 -07003374void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003375
Joe Perchesf629d202013-09-26 14:48:15 -07003376void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003377
Joe Perchesf629d202013-09-26 14:48:15 -07003378void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003379
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003380/**
3381 * netif_dormant_on - mark device as dormant.
3382 * @dev: network device
3383 *
3384 * Mark device as dormant (as per RFC2863).
3385 *
3386 * The dormant state indicates that the relevant interface is not
3387 * actually in a condition to pass packets (i.e., it is not 'up') but is
3388 * in a "pending" state, waiting for some external event. For "on-
3389 * demand" interfaces, this new state identifies the situation where the
3390 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003391 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003392static inline void netif_dormant_on(struct net_device *dev)
3393{
3394 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3395 linkwatch_fire_event(dev);
3396}
3397
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003398/**
3399 * netif_dormant_off - set device as not dormant.
3400 * @dev: network device
3401 *
3402 * Device is not in dormant state.
3403 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003404static inline void netif_dormant_off(struct net_device *dev)
3405{
3406 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3407 linkwatch_fire_event(dev);
3408}
3409
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003410/**
3411 * netif_dormant - test if carrier present
3412 * @dev: network device
3413 *
3414 * Check if carrier is present on device
3415 */
David S. Miller4d295152012-03-07 21:02:35 -05003416static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003417{
3418 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3419}
3420
3421
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003422/**
3423 * netif_oper_up - test if device is operational
3424 * @dev: network device
3425 *
3426 * Check if carrier is operational
3427 */
David S. Miller4d295152012-03-07 21:02:35 -05003428static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003429{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003430 return (dev->operstate == IF_OPER_UP ||
3431 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3432}
3433
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003434/**
3435 * netif_device_present - is device available or removed
3436 * @dev: network device
3437 *
3438 * Check if device has not been removed from system.
3439 */
David S. Miller4d295152012-03-07 21:02:35 -05003440static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003441{
3442 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3443}
3444
Joe Perchesf629d202013-09-26 14:48:15 -07003445void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003446
Joe Perchesf629d202013-09-26 14:48:15 -07003447void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003448
3449/*
3450 * Network interface message level settings
3451 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003452
3453enum {
3454 NETIF_MSG_DRV = 0x0001,
3455 NETIF_MSG_PROBE = 0x0002,
3456 NETIF_MSG_LINK = 0x0004,
3457 NETIF_MSG_TIMER = 0x0008,
3458 NETIF_MSG_IFDOWN = 0x0010,
3459 NETIF_MSG_IFUP = 0x0020,
3460 NETIF_MSG_RX_ERR = 0x0040,
3461 NETIF_MSG_TX_ERR = 0x0080,
3462 NETIF_MSG_TX_QUEUED = 0x0100,
3463 NETIF_MSG_INTR = 0x0200,
3464 NETIF_MSG_TX_DONE = 0x0400,
3465 NETIF_MSG_RX_STATUS = 0x0800,
3466 NETIF_MSG_PKTDATA = 0x1000,
3467 NETIF_MSG_HW = 0x2000,
3468 NETIF_MSG_WOL = 0x4000,
3469};
3470
3471#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3472#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3473#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3474#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3475#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3476#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3477#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3478#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3479#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3480#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3481#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3482#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3483#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3484#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3485#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3486
3487static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3488{
3489 /* use default */
3490 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3491 return default_msg_enable_bits;
3492 if (debug_value == 0) /* no output */
3493 return 0;
3494 /* set low N bits */
3495 return (1 << debug_value) - 1;
3496}
3497
David S. Millerc773e842008-07-08 23:13:53 -07003498static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003499{
David S. Millerc773e842008-07-08 23:13:53 -07003500 spin_lock(&txq->_xmit_lock);
3501 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003502}
3503
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003504static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3505{
3506 __acquire(&txq->_xmit_lock);
3507 return true;
3508}
3509
3510static inline void __netif_tx_release(struct netdev_queue *txq)
3511{
3512 __release(&txq->_xmit_lock);
3513}
3514
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003515static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3516{
3517 spin_lock_bh(&txq->_xmit_lock);
3518 txq->xmit_lock_owner = smp_processor_id();
3519}
3520
David S. Miller4d295152012-03-07 21:02:35 -05003521static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003522{
David S. Miller4d295152012-03-07 21:02:35 -05003523 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003524 if (likely(ok))
3525 txq->xmit_lock_owner = smp_processor_id();
3526 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003527}
3528
David S. Millerc773e842008-07-08 23:13:53 -07003529static inline void __netif_tx_unlock(struct netdev_queue *txq)
3530{
3531 txq->xmit_lock_owner = -1;
3532 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003533}
3534
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003535static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3536{
3537 txq->xmit_lock_owner = -1;
3538 spin_unlock_bh(&txq->_xmit_lock);
3539}
3540
Eric Dumazet08baf562009-05-25 22:58:01 -07003541static inline void txq_trans_update(struct netdev_queue *txq)
3542{
3543 if (txq->xmit_lock_owner != -1)
3544 txq->trans_start = jiffies;
3545}
3546
Florian Westphalba162f82016-05-03 16:31:00 +02003547/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3548static inline void netif_trans_update(struct net_device *dev)
3549{
Florian Westphal9b366272016-05-03 16:33:14 +02003550 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3551
3552 if (txq->trans_start != jiffies)
3553 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003554}
3555
David S. Millerc3f26a22008-07-31 16:58:50 -07003556/**
3557 * netif_tx_lock - grab network device transmit lock
3558 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003559 *
3560 * Get network device transmit lock
3561 */
3562static inline void netif_tx_lock(struct net_device *dev)
3563{
3564 unsigned int i;
3565 int cpu;
3566
3567 spin_lock(&dev->tx_global_lock);
3568 cpu = smp_processor_id();
3569 for (i = 0; i < dev->num_tx_queues; i++) {
3570 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3571
3572 /* We are the only thread of execution doing a
3573 * freeze, but we have to grab the _xmit_lock in
3574 * order to synchronize with threads which are in
3575 * the ->hard_start_xmit() handler and already
3576 * checked the frozen bit.
3577 */
3578 __netif_tx_lock(txq, cpu);
3579 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3580 __netif_tx_unlock(txq);
3581 }
3582}
3583
3584static inline void netif_tx_lock_bh(struct net_device *dev)
3585{
3586 local_bh_disable();
3587 netif_tx_lock(dev);
3588}
3589
Herbert Xu932ff272006-06-09 12:20:56 -07003590static inline void netif_tx_unlock(struct net_device *dev)
3591{
David S. Millere8a04642008-07-17 00:34:19 -07003592 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003593
David S. Millere8a04642008-07-17 00:34:19 -07003594 for (i = 0; i < dev->num_tx_queues; i++) {
3595 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003596
David S. Millerc3f26a22008-07-31 16:58:50 -07003597 /* No need to grab the _xmit_lock here. If the
3598 * queue is not stopped for another reason, we
3599 * force a schedule.
3600 */
3601 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003602 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003603 }
3604 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003605}
3606
3607static inline void netif_tx_unlock_bh(struct net_device *dev)
3608{
David S. Millere8a04642008-07-17 00:34:19 -07003609 netif_tx_unlock(dev);
3610 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003611}
3612
David S. Millerc773e842008-07-08 23:13:53 -07003613#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003614 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003615 __netif_tx_lock(txq, cpu); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003616 } else { \
3617 __netif_tx_acquire(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003618 } \
3619}
3620
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003621#define HARD_TX_TRYLOCK(dev, txq) \
3622 (((dev->features & NETIF_F_LLTX) == 0) ? \
3623 __netif_tx_trylock(txq) : \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003624 __netif_tx_acquire(txq))
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003625
David S. Millerc773e842008-07-08 23:13:53 -07003626#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003627 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003628 __netif_tx_unlock(txq); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003629 } else { \
3630 __netif_tx_release(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003631 } \
3632}
3633
Linus Torvalds1da177e2005-04-16 15:20:36 -07003634static inline void netif_tx_disable(struct net_device *dev)
3635{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003636 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003637 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003638
David S. Millerc3f26a22008-07-31 16:58:50 -07003639 local_bh_disable();
3640 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003641 for (i = 0; i < dev->num_tx_queues; i++) {
3642 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003643
3644 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003645 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003646 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003647 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003648 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003649}
3650
David S. Millere308a5d2008-07-15 00:13:44 -07003651static inline void netif_addr_lock(struct net_device *dev)
3652{
3653 spin_lock(&dev->addr_list_lock);
3654}
3655
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003656static inline void netif_addr_lock_nested(struct net_device *dev)
3657{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003658 int subclass = SINGLE_DEPTH_NESTING;
3659
3660 if (dev->netdev_ops->ndo_get_lock_subclass)
3661 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3662
3663 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003664}
3665
David S. Millere308a5d2008-07-15 00:13:44 -07003666static inline void netif_addr_lock_bh(struct net_device *dev)
3667{
3668 spin_lock_bh(&dev->addr_list_lock);
3669}
3670
3671static inline void netif_addr_unlock(struct net_device *dev)
3672{
3673 spin_unlock(&dev->addr_list_lock);
3674}
3675
3676static inline void netif_addr_unlock_bh(struct net_device *dev)
3677{
3678 spin_unlock_bh(&dev->addr_list_lock);
3679}
3680
Jiri Pirkof001fde2009-05-05 02:48:28 +00003681/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003682 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003683 * rcu_read_lock held.
3684 */
3685#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003686 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003687
Linus Torvalds1da177e2005-04-16 15:20:36 -07003688/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3689
Joe Perchesf629d202013-09-26 14:48:15 -07003690void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003691
3692/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003693struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003694 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003695 void (*setup)(struct net_device *),
3696 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003697#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3698 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003699
Tom Gundersenc835a672014-07-14 16:37:24 +02003700#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3701 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3702 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003703
Joe Perchesf629d202013-09-26 14:48:15 -07003704int register_netdev(struct net_device *dev);
3705void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003706
Jiri Pirko22bedad32010-04-01 21:22:57 +00003707/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003708int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3709 struct netdev_hw_addr_list *from_list, int addr_len);
3710void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3711 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003712int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3713 struct net_device *dev,
3714 int (*sync)(struct net_device *, const unsigned char *),
3715 int (*unsync)(struct net_device *,
3716 const unsigned char *));
3717void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3718 struct net_device *dev,
3719 int (*unsync)(struct net_device *,
3720 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003721void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003722
Jiri Pirkof001fde2009-05-05 02:48:28 +00003723/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003724int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3725 unsigned char addr_type);
3726int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3727 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003728void dev_addr_flush(struct net_device *dev);
3729int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003730
3731/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003732int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3733int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3734int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3735int dev_uc_sync(struct net_device *to, struct net_device *from);
3736int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3737void dev_uc_unsync(struct net_device *to, struct net_device *from);
3738void dev_uc_flush(struct net_device *dev);
3739void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003740
Alexander Duyck670e5b82014-05-28 18:44:46 -07003741/**
3742 * __dev_uc_sync - Synchonize device's unicast list
3743 * @dev: device to sync
3744 * @sync: function to call if address should be added
3745 * @unsync: function to call if address should be removed
3746 *
3747 * Add newly added addresses to the interface, and release
3748 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003749 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003750static inline int __dev_uc_sync(struct net_device *dev,
3751 int (*sync)(struct net_device *,
3752 const unsigned char *),
3753 int (*unsync)(struct net_device *,
3754 const unsigned char *))
3755{
3756 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3757}
3758
3759/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003760 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003761 * @dev: device to sync
3762 * @unsync: function to call if address should be removed
3763 *
3764 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003765 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003766static inline void __dev_uc_unsync(struct net_device *dev,
3767 int (*unsync)(struct net_device *,
3768 const unsigned char *))
3769{
3770 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3771}
3772
Jiri Pirko22bedad32010-04-01 21:22:57 +00003773/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003774int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3775int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3776int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3777int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3778int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3779int dev_mc_sync(struct net_device *to, struct net_device *from);
3780int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3781void dev_mc_unsync(struct net_device *to, struct net_device *from);
3782void dev_mc_flush(struct net_device *dev);
3783void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003784
Alexander Duyck670e5b82014-05-28 18:44:46 -07003785/**
3786 * __dev_mc_sync - Synchonize device's multicast list
3787 * @dev: device to sync
3788 * @sync: function to call if address should be added
3789 * @unsync: function to call if address should be removed
3790 *
3791 * Add newly added addresses to the interface, and release
3792 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003793 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003794static inline int __dev_mc_sync(struct net_device *dev,
3795 int (*sync)(struct net_device *,
3796 const unsigned char *),
3797 int (*unsync)(struct net_device *,
3798 const unsigned char *))
3799{
3800 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3801}
3802
3803/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003804 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003805 * @dev: device to sync
3806 * @unsync: function to call if address should be removed
3807 *
3808 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003809 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003810static inline void __dev_mc_unsync(struct net_device *dev,
3811 int (*unsync)(struct net_device *,
3812 const unsigned char *))
3813{
3814 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3815}
3816
Linus Torvalds1da177e2005-04-16 15:20:36 -07003817/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003818void dev_set_rx_mode(struct net_device *dev);
3819void __dev_set_rx_mode(struct net_device *dev);
3820int dev_set_promiscuity(struct net_device *dev, int inc);
3821int dev_set_allmulti(struct net_device *dev, int inc);
3822void netdev_state_change(struct net_device *dev);
3823void netdev_notify_peers(struct net_device *dev);
3824void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003825/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003826void dev_load(struct net *net, const char *name);
3827struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3828 struct rtnl_link_stats64 *storage);
3829void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3830 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003831
3832extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003833extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003834extern int weight_p;
Matthias Tafelmeier3d48b532016-12-29 21:37:21 +01003835extern int dev_weight_rx_bias;
3836extern int dev_weight_tx_bias;
3837extern int dev_rx_weight;
3838extern int dev_tx_weight;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003839
Joe Perchesf629d202013-09-26 14:48:15 -07003840bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003841struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3842 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003843struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3844 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003845
3846/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003847#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3848 for (iter = &(dev)->adj_list.upper, \
3849 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3850 updev; \
3851 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3852
David Ahern1a3f0602016-10-17 19:15:44 -07003853int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3854 int (*fn)(struct net_device *upper_dev,
3855 void *data),
3856 void *data);
3857
3858bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3859 struct net_device *upper_dev);
3860
Joe Perchesf629d202013-09-26 14:48:15 -07003861void *netdev_lower_get_next_private(struct net_device *dev,
3862 struct list_head **iter);
3863void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3864 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003865
3866#define netdev_for_each_lower_private(dev, priv, iter) \
3867 for (iter = (dev)->adj_list.lower.next, \
3868 priv = netdev_lower_get_next_private(dev, &(iter)); \
3869 priv; \
3870 priv = netdev_lower_get_next_private(dev, &(iter)))
3871
3872#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3873 for (iter = &(dev)->adj_list.lower, \
3874 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3875 priv; \
3876 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3877
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003878void *netdev_lower_get_next(struct net_device *dev,
3879 struct list_head **iter);
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003880
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003881#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003882 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003883 ldev = netdev_lower_get_next(dev, &(iter)); \
3884 ldev; \
3885 ldev = netdev_lower_get_next(dev, &(iter)))
3886
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003887struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3888 struct list_head **iter);
3889struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3890 struct list_head **iter);
3891
David Ahern1a3f0602016-10-17 19:15:44 -07003892int netdev_walk_all_lower_dev(struct net_device *dev,
3893 int (*fn)(struct net_device *lower_dev,
3894 void *data),
3895 void *data);
3896int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3897 int (*fn)(struct net_device *lower_dev,
3898 void *data),
3899 void *data);
3900
Joe Perchesf629d202013-09-26 14:48:15 -07003901void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003902void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003903struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3904struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3905int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3906int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003907 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003908 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003909void netdev_upper_dev_unlink(struct net_device *dev,
3910 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003911void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003912void *netdev_lower_dev_get_private(struct net_device *dev,
3913 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003914void netdev_lower_state_changed(struct net_device *lower_dev,
3915 void *lower_state_info);
Eric Dumazet960fb622014-11-16 06:23:05 -08003916
3917/* RSS keys are 40 or 52 bytes long */
3918#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003919extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003920void netdev_rss_key_fill(void *buffer, size_t len);
3921
Sabrina Dubroca952fcfd2016-08-12 16:10:33 +02003922int dev_get_nest_level(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003923int skb_checksum_help(struct sk_buff *skb);
3924struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3925 netdev_features_t features, bool tx_path);
3926struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3927 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003928
Moni Shoua61bd3852015-02-03 16:48:29 +02003929struct netdev_bonding_info {
3930 ifslave slave;
3931 ifbond master;
3932};
3933
3934struct netdev_notifier_bonding_info {
3935 struct netdev_notifier_info info; /* must be first */
3936 struct netdev_bonding_info bonding_info;
3937};
3938
3939void netdev_bonding_info_change(struct net_device *dev,
3940 struct netdev_bonding_info *bonding_info);
3941
Cong Wang12b00042013-02-05 16:36:38 +00003942static inline
3943struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3944{
3945 return __skb_gso_segment(skb, features, true);
3946}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003947__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003948
3949static inline bool can_checksum_protocol(netdev_features_t features,
3950 __be16 protocol)
3951{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003952 if (protocol == htons(ETH_P_FCOE))
3953 return !!(features & NETIF_F_FCOE_CRC);
3954
3955 /* Assume this is an IP checksum (not SCTP CRC) */
3956
3957 if (features & NETIF_F_HW_CSUM) {
3958 /* Can checksum everything */
3959 return true;
3960 }
3961
3962 switch (protocol) {
3963 case htons(ETH_P_IP):
3964 return !!(features & NETIF_F_IP_CSUM);
3965 case htons(ETH_P_IPV6):
3966 return !!(features & NETIF_F_IPV6_CSUM);
3967 default:
3968 return false;
3969 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003970}
Cong Wang12b00042013-02-05 16:36:38 +00003971
Herbert Xufb286bb2005-11-10 13:01:24 -08003972#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003973void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003974#else
3975static inline void netdev_rx_csum_fault(struct net_device *dev)
3976{
3977}
3978#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003979/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003980void net_enable_timestamp(void);
3981void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003982
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003983#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003984int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003985#else
3986#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003987#endif
3988
David S. Miller47982482014-08-22 16:21:53 -07003989static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003990 struct sk_buff *skb, struct net_device *dev,
3991 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003992{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003993 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003994 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003995}
3996
David S. Miller10b3ad82014-08-29 21:07:24 -07003997static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003998 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07003999{
4000 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07004001 int rc;
David S. Miller47982482014-08-22 16:21:53 -07004002
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004003 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07004004 if (rc == NETDEV_TX_OK)
4005 txq_trans_update(txq);
4006
4007 return rc;
David S. Miller47982482014-08-22 16:21:53 -07004008}
4009
Linus Torvalds42a2d922013-11-13 17:40:34 +09004010int netdev_class_create_file_ns(struct class_attribute *class_attr,
4011 const void *ns);
4012void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4013 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04004014
4015static inline int netdev_class_create_file(struct class_attribute *class_attr)
4016{
4017 return netdev_class_create_file_ns(class_attr, NULL);
4018}
4019
4020static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4021{
4022 netdev_class_remove_file_ns(class_attr, NULL);
4023}
Jay Vosburghb8a97872008-06-13 18:12:04 -07004024
Johannes Berg04600792010-08-05 17:45:15 +02004025extern struct kobj_ns_type_operations net_ns_type_operations;
4026
Joe Perchesf629d202013-09-26 14:48:15 -07004027const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07004028
Joe Perchesf629d202013-09-26 14:48:15 -07004029void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004030
Michal Kubečekda081432014-05-20 08:29:25 +02004031static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4032 netdev_features_t f2)
4033{
Tom Herbertc8cd0982015-12-14 11:19:44 -08004034 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4035 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08004036 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004037 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08004038 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004039 }
Michal Kubečekda081432014-05-20 08:29:25 +02004040
Tom Herbertc8cd0982015-12-14 11:19:44 -08004041 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02004042}
4043
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004044static inline netdev_features_t netdev_get_wanted_features(
4045 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00004046{
4047 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4048}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004049netdev_features_t netdev_increment_features(netdev_features_t all,
4050 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00004051
4052/* Allow TSO being used on stacked device :
4053 * Performing the GSO segmentation before last device
4054 * is a performance improvement.
4055 */
4056static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4057 netdev_features_t mask)
4058{
4059 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4060}
4061
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004062int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004063void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004064void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004065
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004066void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4067 struct net_device *dev);
4068
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004069netdev_features_t passthru_features_check(struct sk_buff *skb,
4070 struct net_device *dev,
4071 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004072netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004073
David S. Miller4d295152012-03-07 21:02:35 -05004074static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004075{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004076 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004077
4078 /* check flags correspondence */
4079 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4080 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4081 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4082 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004083 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004084 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4085 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004086 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4087 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004088 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4089 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004090 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4091 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 -04004092 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004093 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004094 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Steffen Klassertc7ef8f02017-04-14 10:05:36 +02004095 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004096
Herbert Xubcd76112006-06-30 13:36:35 -07004097 return (features & feature) == feature;
4098}
4099
David S. Miller4d295152012-03-07 21:02:35 -05004100static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004101{
Herbert Xu278b2512009-06-03 21:20:51 -07004102 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004103 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004104}
4105
Johannes Berg8b86a612015-04-17 15:45:04 +02004106static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004107 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004108{
Jesse Grossfc741212011-01-09 06:23:32 +00004109 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004110 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4111 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004112}
4113
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004114static inline void netif_set_gso_max_size(struct net_device *dev,
4115 unsigned int size)
4116{
4117 dev->gso_max_size = size;
4118}
4119
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004120static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4121 int pulled_hlen, u16 mac_offset,
4122 int mac_len)
4123{
4124 skb->protocol = protocol;
4125 skb->encapsulation = 1;
4126 skb_push(skb, pulled_hlen);
4127 skb_reset_transport_header(skb);
4128 skb->mac_header = mac_offset;
4129 skb->network_header = skb->mac_header + mac_len;
4130 skb->mac_len = mac_len;
4131}
4132
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004133static inline bool netif_is_macsec(const struct net_device *dev)
4134{
4135 return dev->priv_flags & IFF_MACSEC;
4136}
4137
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004138static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004139{
4140 return dev->priv_flags & IFF_MACVLAN;
4141}
4142
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004143static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004144{
4145 return dev->priv_flags & IFF_MACVLAN_PORT;
4146}
4147
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004148static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004149{
4150 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4151}
4152
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004153static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004154{
4155 return dev->priv_flags & IFF_IPVLAN_MASTER;
4156}
4157
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004158static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004159{
4160 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4161}
4162
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004163static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004164{
4165 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4166}
4167
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004168static inline bool netif_supports_nofcs(struct net_device *dev)
4169{
4170 return dev->priv_flags & IFF_SUPP_NOFCS;
4171}
4172
David Ahern007979e2015-09-29 20:07:10 -07004173static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004174{
David Ahern007979e2015-09-29 20:07:10 -07004175 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004176}
4177
David Ahernfee6d4c2015-10-05 08:51:24 -07004178static inline bool netif_is_l3_slave(const struct net_device *dev)
4179{
4180 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4181}
4182
Jiri Pirko0894ae32015-08-27 09:31:19 +02004183static inline bool netif_is_bridge_master(const struct net_device *dev)
4184{
4185 return dev->priv_flags & IFF_EBRIDGE;
4186}
4187
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004188static inline bool netif_is_bridge_port(const struct net_device *dev)
4189{
4190 return dev->priv_flags & IFF_BRIDGE_PORT;
4191}
4192
Jiri Pirko35d4e172015-08-27 09:31:20 +02004193static inline bool netif_is_ovs_master(const struct net_device *dev)
4194{
4195 return dev->priv_flags & IFF_OPENVSWITCH;
4196}
4197
Jiri Pirko5be66142017-04-18 16:55:36 +02004198static inline bool netif_is_ovs_port(const struct net_device *dev)
4199{
4200 return dev->priv_flags & IFF_OVS_DATAPATH;
4201}
4202
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004203static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004204{
4205 return dev->priv_flags & IFF_TEAM;
4206}
4207
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004208static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004209{
4210 return dev->priv_flags & IFF_TEAM_PORT;
4211}
4212
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004213static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004214{
4215 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4216}
4217
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004218static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004219{
4220 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4221}
4222
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004223static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4224{
4225 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4226}
4227
Eric Dumazet02875872014-10-05 18:38:35 -07004228/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4229static inline void netif_keep_dst(struct net_device *dev)
4230{
4231 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4232}
4233
Paolo Abeni18d3df32016-07-14 18:00:10 +02004234/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4235static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4236{
4237 /* TODO: reserve and use an additional IFF bit, if we get more users */
4238 return dev->priv_flags & IFF_MACSEC;
4239}
4240
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004241extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004242
Joe Perches571ba422010-02-09 11:49:47 +00004243/* Logging, debugging and troubleshooting/diagnostic helpers. */
4244
4245/* netdev_printk helpers, similar to dev_printk */
4246
4247static inline const char *netdev_name(const struct net_device *dev)
4248{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004249 if (!dev->name[0] || strchr(dev->name, '%'))
4250 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004251 return dev->name;
4252}
4253
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004254static inline const char *netdev_reg_state(const struct net_device *dev)
4255{
4256 switch (dev->reg_state) {
4257 case NETREG_UNINITIALIZED: return " (uninitialized)";
4258 case NETREG_REGISTERED: return "";
4259 case NETREG_UNREGISTERING: return " (unregistering)";
4260 case NETREG_UNREGISTERED: return " (unregistered)";
4261 case NETREG_RELEASED: return " (released)";
4262 case NETREG_DUMMY: return " (dummy)";
4263 }
4264
4265 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4266 return " (unknown)";
4267}
4268
Joe Perchesf629d202013-09-26 14:48:15 -07004269__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004270void netdev_printk(const char *level, const struct net_device *dev,
4271 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004272__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004273void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004274__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004275void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004276__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004277void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004278__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004279void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004280__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004281void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004282__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004283void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004284__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004285void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004286
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004287#define MODULE_ALIAS_NETDEV(device) \
4288 MODULE_ALIAS("netdev-" device)
4289
Jim Cromieb558c962011-12-19 17:11:18 -05004290#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004291#define netdev_dbg(__dev, format, args...) \
4292do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004293 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004294} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004295#elif defined(DEBUG)
4296#define netdev_dbg(__dev, format, args...) \
4297 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004298#else
4299#define netdev_dbg(__dev, format, args...) \
4300({ \
4301 if (0) \
4302 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004303})
4304#endif
4305
4306#if defined(VERBOSE_DEBUG)
4307#define netdev_vdbg netdev_dbg
4308#else
4309
4310#define netdev_vdbg(dev, format, args...) \
4311({ \
4312 if (0) \
4313 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4314 0; \
4315})
4316#endif
4317
4318/*
4319 * netdev_WARN() acts like dev_printk(), but with the key difference
4320 * of using a WARN/WARN_ON to get the message out, including the
4321 * file/line information and a backtrace.
4322 */
4323#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004324 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4325 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004326
Joe Perchesb3d95c52010-02-09 11:49:49 +00004327/* netif printk helpers, similar to netdev_printk */
4328
4329#define netif_printk(priv, type, level, dev, fmt, args...) \
4330do { \
4331 if (netif_msg_##type(priv)) \
4332 netdev_printk(level, (dev), fmt, ##args); \
4333} while (0)
4334
Joe Perchesf45f4322010-06-27 01:02:36 +00004335#define netif_level(level, priv, type, dev, fmt, args...) \
4336do { \
4337 if (netif_msg_##type(priv)) \
4338 netdev_##level(dev, fmt, ##args); \
4339} while (0)
4340
Joe Perchesb3d95c52010-02-09 11:49:49 +00004341#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004342 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004343#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004344 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004345#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004346 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004347#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004348 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004349#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004350 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004351#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004352 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004353#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004354 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004355
Joe Perches0053ea92012-05-30 07:43:34 +00004356#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004357#define netif_dbg(priv, type, netdev, format, args...) \
4358do { \
4359 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004360 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004361} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004362#elif defined(DEBUG)
4363#define netif_dbg(priv, type, dev, format, args...) \
4364 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004365#else
4366#define netif_dbg(priv, type, dev, format, args...) \
4367({ \
4368 if (0) \
4369 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4370 0; \
4371})
4372#endif
4373
Edward Creef617f272017-01-27 15:02:26 +00004374/* if @cond then downgrade to debug, else print at @level */
4375#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4376 do { \
4377 if (cond) \
4378 netif_dbg(priv, type, netdev, fmt, ##args); \
4379 else \
4380 netif_ ## level(priv, type, netdev, fmt, ##args); \
4381 } while (0)
4382
Joe Perchesb3d95c52010-02-09 11:49:49 +00004383#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004384#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004385#else
4386#define netif_vdbg(priv, type, dev, format, args...) \
4387({ \
4388 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004389 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004390 0; \
4391})
4392#endif
Joe Perches571ba422010-02-09 11:49:47 +00004393
Cong Wang900ff8c2013-02-18 19:20:33 +00004394/*
4395 * The list of packet types we will receive (as opposed to discard)
4396 * and the routines to invoke.
4397 *
4398 * Why 16. Because with 16 the only overlap we get on a hash of the
4399 * low nibble of the protocol value is RARP/SNAP/X.25.
4400 *
4401 * NOTE: That is no longer true with the addition of VLAN tags. Not
4402 * sure which should go first, but I bet it won't make much
4403 * difference if we are running VLANs. The good news is that
4404 * this protocol won't be in the list unless compiled in, so
4405 * the average user (w/out VLANs) will not be adversely affected.
4406 * --BLG
4407 *
4408 * 0800 IP
4409 * 8100 802.1Q VLAN
4410 * 0001 802.3
4411 * 0002 AX.25
4412 * 0004 802.2
4413 * 8035 RARP
4414 * 0005 SNAP
4415 * 0805 X.25
4416 * 0806 ARP
4417 * 8137 IPX
4418 * 0009 Localtalk
4419 * 86DD IPv6
4420 */
4421#define PTYPE_HASH_SIZE (16)
4422#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4423
Jiri Pirko385a1542009-05-27 15:48:07 -07004424#endif /* _LINUX_NETDEVICE_H */