blob: cc07c3be2705f1a3432117b1c687fb679813bb5a [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 {
David S. Millerf6b72b622011-07-14 07:53:20 -0700240 u16 hh_len;
David S. Miller5c25f682011-07-13 00:51:10 -0700241 u16 __pad;
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800242 seqlock_t hh_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243
244 /* cached hardware header; allow for machine alignment needs. */
245#define HH_DATA_MOD 16
246#define HH_DATA_OFF(__len) \
Jiri Benc5ba0eac2005-06-02 16:48:05 -0700247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248#define HH_DATA_ALIGN(__len) \
249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251};
252
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500253/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 * Alternative is:
255 * dev->hard_header_len ? (dev->hard_header_len +
256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 *
258 * We could use other alignment values, but we must maintain the
259 * relationship HH alignment <= LL alignment.
260 */
261#define LL_RESERVED_SPACE(dev) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700262 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700264 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700266struct header_ops {
267 int (*create) (struct sk_buff *skb, struct net_device *dev,
268 unsigned short type, const void *daddr,
Eric Dumazet95c96172012-04-15 05:58:06 +0000269 const void *saddr, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700270 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
David S. Millere69dd332011-07-12 23:28:12 -0700271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
Willem de Bruijn2793a232016-03-09 21:58:32 -0500275 bool (*validate)(const char *ll_header, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700276};
277
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278/* These flag bits are private to the generic network queueing
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500279 * layer; they may not be explicitly referenced by any other
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 * code.
281 */
282
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800283enum netdev_state_t {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 __LINK_STATE_START,
285 __LINK_STATE_PRESENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 __LINK_STATE_NOCARRIER,
Stefan Rompfb00055a2006-03-20 17:09:11 -0800287 __LINK_STATE_LINKWATCH_PENDING,
288 __LINK_STATE_DORMANT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289};
290
291
292/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500293 * This structure holds boot-time configured netdevice settings. They
Graf Yangfe2918b2009-02-05 21:26:19 -0800294 * are then used in the device probing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295 */
296struct netdev_boot_setup {
297 char name[IFNAMSIZ];
298 struct ifmap map;
299};
300#define NETDEV_BOOT_SETUP_MAX 8
301
Joe Perchesf629d202013-09-26 14:48:15 -0700302int __init netdev_boot_setup(char *str);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303
304/*
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700305 * Structure for NAPI scheduling similar to tasklet but with weighting
306 */
307struct napi_struct {
308 /* The poll_list must only be managed by the entity which
309 * changes the state of the NAPI_STATE_SCHED bit. This means
310 * whoever atomically sets that bit can add this napi_struct
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500311 * to the per-CPU poll_list, and whoever clears that bit
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700312 * can remove from the list right before clearing the bit.
313 */
314 struct list_head poll_list;
315
316 unsigned long state;
317 int weight;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000318 unsigned int gro_count;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700319 int (*poll)(struct napi_struct *, int);
320#ifdef CONFIG_NETPOLL
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700321 int poll_owner;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700322#endif
Herbert Xu5d38a072009-01-04 16:13:40 -0800323 struct net_device *dev;
Herbert Xud565b0a2008-12-15 23:38:52 -0800324 struct sk_buff *gro_list;
Herbert Xu5d38a072009-01-04 16:13:40 -0800325 struct sk_buff *skb;
Eric Dumazet3b47d302014-11-06 21:09:44 -0800326 struct hrtimer timer;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000327 struct list_head dev_list;
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300328 struct hlist_node napi_hash_node;
329 unsigned int napi_id;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700330};
331
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800332enum {
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700333 NAPI_STATE_SCHED, /* Poll is scheduled */
Eric Dumazet39e6c822017-02-28 10:34:50 -0800334 NAPI_STATE_MISSED, /* reschedule a napi */
David S. Millera0a46192008-01-07 20:35:07 -0800335 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800336 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eric Dumazetd64b5e82015-11-18 06:31:00 -0800337 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
338 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
Eric Dumazet217f6972016-11-15 10:15:11 -0800339 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
340};
341
342enum {
Eric Dumazet39e6c822017-02-28 10:34:50 -0800343 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
344 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
345 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
346 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
347 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
348 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
349 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700350};
351
Ben Hutchings5b252f02009-10-29 07:17:09 +0000352enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700353 GRO_MERGED,
354 GRO_MERGED_FREE,
355 GRO_HELD,
356 GRO_NORMAL,
357 GRO_DROP,
Steffen Klassert25393d32017-02-15 09:39:44 +0100358 GRO_CONSUMED,
Herbert Xud1c76af2009-03-16 10:50:02 -0700359};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000360typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700361
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000362/*
363 * enum rx_handler_result - Possible return values for rx_handlers.
364 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365 * further.
366 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
367 * case skb->dev was changed by rx_handler.
368 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500369 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000370 *
371 * rx_handlers are functions called from inside __netif_receive_skb(), to do
372 * special processing of the skb, prior to delivery to protocol handlers.
373 *
374 * Currently, a net_device can only have a single rx_handler registered. Trying
375 * to register a second rx_handler will return -EBUSY.
376 *
377 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
378 * To unregister a rx_handler on a net_device, use
379 * netdev_rx_handler_unregister().
380 *
381 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
382 * do with the skb.
383 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500384 * If the rx_handler consumed the skb in some way, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000385 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500386 * the skb to be delivered in some other way.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000387 *
388 * If the rx_handler changed skb->dev, to divert the skb to another
389 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
390 * new device will be called if it exists.
391 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500392 * If the rx_handler decides the skb should be ignored, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000393 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400394 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000395 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500396 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000397 * delivered, it should return RX_HANDLER_PASS.
398 *
399 * A device without a registered rx_handler will behave as if rx_handler
400 * returned RX_HANDLER_PASS.
401 */
402
403enum rx_handler_result {
404 RX_HANDLER_CONSUMED,
405 RX_HANDLER_ANOTHER,
406 RX_HANDLER_EXACT,
407 RX_HANDLER_PASS,
408};
409typedef enum rx_handler_result rx_handler_result_t;
410typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000411
Joe Perchesf629d202013-09-26 14:48:15 -0700412void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700413void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700414
David S. Miller4d295152012-03-07 21:02:35 -0500415static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800416{
417 return test_bit(NAPI_STATE_DISABLE, &n->state);
418}
419
Eric Dumazet39e6c822017-02-28 10:34:50 -0800420bool napi_schedule_prep(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700421
422/**
423 * napi_schedule - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500424 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700425 *
426 * Schedule NAPI poll routine to be called if it is not already
427 * running.
428 */
429static inline void napi_schedule(struct napi_struct *n)
430{
431 if (napi_schedule_prep(n))
432 __napi_schedule(n);
433}
434
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700435/**
436 * napi_schedule_irqoff - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500437 * @n: NAPI context
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700438 *
439 * Variant of napi_schedule(), assuming hard irqs are masked.
440 */
441static inline void napi_schedule_irqoff(struct napi_struct *n)
442{
443 if (napi_schedule_prep(n))
444 __napi_schedule_irqoff(n);
445}
446
Roland Dreierbfe13f52007-10-09 15:47:37 -0700447/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
David S. Miller4d295152012-03-07 21:02:35 -0500448static inline bool napi_reschedule(struct napi_struct *napi)
Roland Dreierbfe13f52007-10-09 15:47:37 -0700449{
450 if (napi_schedule_prep(napi)) {
451 __napi_schedule(napi);
David S. Miller4d295152012-03-07 21:02:35 -0500452 return true;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700453 }
David S. Miller4d295152012-03-07 21:02:35 -0500454 return false;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700455}
456
Eric Dumazet364b6052016-11-15 10:15:13 -0800457bool napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700458/**
459 * napi_complete - NAPI processing complete
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500460 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700461 *
462 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800463 * Consider using napi_complete_done() instead.
Eric Dumazet364b6052016-11-15 10:15:13 -0800464 * Return false if device should avoid rearming interrupts.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700465 */
Eric Dumazet364b6052016-11-15 10:15:13 -0800466static inline bool napi_complete(struct napi_struct *n)
Eric Dumazet3b47d302014-11-06 21:09:44 -0800467{
468 return napi_complete_done(n, 0);
469}
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700470
471/**
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300472 * napi_hash_del - remove a NAPI from global table
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500473 * @napi: NAPI context
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300474 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500475 * Warning: caller must observe RCU grace period
Eric Dumazet34cbe272015-11-18 06:31:02 -0800476 * before freeing memory containing @napi, if
477 * this function returns true.
Eric Dumazet93d05d42015-11-18 06:31:03 -0800478 * Note: core networking stack automatically calls it
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500479 * from netif_napi_del().
Eric Dumazet93d05d42015-11-18 06:31:03 -0800480 * Drivers might want to call this helper to combine all
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500481 * the needed RCU grace periods into a single one.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300482 */
Eric Dumazet34cbe272015-11-18 06:31:02 -0800483bool napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300484
485/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700486 * napi_disable - prevent NAPI from scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500487 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700488 *
489 * Stop NAPI from being scheduled on this context.
490 * Waits till any outstanding processing completes.
491 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800492void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700493
494/**
495 * napi_enable - enable NAPI scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500496 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700497 *
498 * Resume NAPI from being scheduled on this context.
499 * Must be paired with napi_disable.
500 */
501static inline void napi_enable(struct napi_struct *n)
502{
503 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100504 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700505 clear_bit(NAPI_STATE_SCHED, &n->state);
Neil Horman2d8bff1262015-09-23 14:57:58 -0400506 clear_bit(NAPI_STATE_NPSVC, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700507}
508
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700509/**
510 * napi_synchronize - wait until NAPI is not running
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500511 * @n: NAPI context
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700512 *
513 * Wait until NAPI is done being scheduled on this context.
514 * Waits till any outstanding processing completes but
515 * does not disable future activations.
516 */
517static inline void napi_synchronize(const struct napi_struct *n)
518{
Arnd Bergmannfacc4322016-01-22 11:43:44 +0100519 if (IS_ENABLED(CONFIG_SMP))
520 while (test_bit(NAPI_STATE_SCHED, &n->state))
521 msleep(1);
522 else
523 barrier();
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700524}
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700525
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800526enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000527 __QUEUE_STATE_DRV_XOFF,
528 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700529 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700530};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200531
532#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
533#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
534#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
535
536#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
537#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 QUEUE_STATE_FROZEN)
539#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
540 QUEUE_STATE_FROZEN)
541
Tom Herbert734664982011-11-28 16:32:44 +0000542/*
543 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
544 * netif_tx_* functions below are used to manipulate this flag. The
545 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
546 * queue independently. The netif_xmit_*stopped functions below are called
547 * to check if the queue has been stopped by the driver or stack (either
548 * of the XOFF bits are set in the state). Drivers should not need to call
549 * netif_xmit*stopped functions, they should only be using netif_tx_*.
550 */
David S. Miller79d16382008-07-08 23:14:46 -0700551
David S. Millerbb949fb2008-07-08 16:55:56 -0700552struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700553/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500554 * read-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700555 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700556 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700557 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700558 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000559#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000560 struct kobject kobj;
561#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000562#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
563 int numa_node;
564#endif
Florian Westphalc0ef0792016-05-03 03:29:09 +0200565 unsigned long tx_maxrate;
566 /*
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
569 */
570 unsigned long trans_timeout;
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700571/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500572 * write-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700573 */
574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
575 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700576 /*
Florian Westphal9b366272016-05-03 16:33:14 +0200577 * Time (in jiffies) of last Tx
Eric Dumazet9d214932009-05-17 20:55:16 -0700578 */
579 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000580
Tom Herbert114cf582011-11-28 16:33:09 +0000581 unsigned long state;
582
583#ifdef CONFIG_BQL
584 struct dql dql;
585#endif
David S. Millere8a04642008-07-17 00:34:19 -0700586} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700587
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000588static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
589{
590#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
591 return q->numa_node;
592#else
Changli Gaob236da62010-12-14 03:09:15 +0000593 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000594#endif
595}
596
597static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 q->numa_node = node;
601#endif
602}
603
Eric Dumazetdf334542010-03-24 19:13:54 +0000604#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000605/*
606 * This structure holds an RPS map which can be of variable length. The
607 * map is an array of CPUs.
608 */
609struct rps_map {
610 unsigned int len;
611 struct rcu_head rcu;
612 u16 cpus[0];
613};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000614#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000615
Tom Herbertfec5e652010-04-16 16:01:27 -0700616/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000617 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
618 * tail pointer for that CPU's input queue at the time of last enqueue, and
619 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700620 */
621struct rps_dev_flow {
622 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000623 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700624 unsigned int last_qtail;
625};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000626#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700627
628/*
629 * The rps_dev_flow_table structure contains a table of flow mappings.
630 */
631struct rps_dev_flow_table {
632 unsigned int mask;
633 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700634 struct rps_dev_flow flows[0];
635};
636#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000637 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700638
639/*
640 * The rps_sock_flow_table contains mappings of flows to the last CPU
641 * on which they were processed by the application (set in recvmsg).
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500642 * Each entry is a 32bit value. Upper part is the high-order bits
643 * of flow hash, lower part is CPU number.
Eric Dumazet567e4b72015-02-06 12:59:01 -0800644 * rps_cpu_mask is used to partition the space, depending on number of
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500645 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
646 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
Eric Dumazet567e4b72015-02-06 12:59:01 -0800647 * meaning we use 32-6=26 bits for the hash.
Tom Herbertfec5e652010-04-16 16:01:27 -0700648 */
649struct rps_sock_flow_table {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800650 u32 mask;
Eric Dumazet93c1af62015-02-08 20:39:13 -0800651
652 u32 ents[0] ____cacheline_aligned_in_smp;
Tom Herbertfec5e652010-04-16 16:01:27 -0700653};
Eric Dumazet567e4b72015-02-06 12:59:01 -0800654#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
Tom Herbertfec5e652010-04-16 16:01:27 -0700655
656#define RPS_NO_CPU 0xffff
657
Eric Dumazet567e4b72015-02-06 12:59:01 -0800658extern u32 rps_cpu_mask;
659extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
660
Tom Herbertfec5e652010-04-16 16:01:27 -0700661static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
662 u32 hash)
663{
664 if (table && hash) {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800665 unsigned int index = hash & table->mask;
666 u32 val = hash & ~rps_cpu_mask;
Tom Herbertfec5e652010-04-16 16:01:27 -0700667
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500668 /* We only give a hint, preemption can change CPU under us */
Eric Dumazet567e4b72015-02-06 12:59:01 -0800669 val |= raw_smp_processor_id();
Tom Herbertfec5e652010-04-16 16:01:27 -0700670
Eric Dumazet567e4b72015-02-06 12:59:01 -0800671 if (table->ents[index] != val)
672 table->ents[index] = val;
Tom Herbertfec5e652010-04-16 16:01:27 -0700673 }
674}
675
Ben Hutchingsc4454772011-01-19 11:03:53 +0000676#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700677bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
678 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000679#endif
Michael Daltona953be52014-01-16 22:23:28 -0800680#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000681
Tom Herbert0a9627f2010-03-16 08:03:29 +0000682/* This structure contains an instance of an RX queue. */
683struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800684#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000685 struct rps_map __rcu *rps_map;
686 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800687#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000688 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000689 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000690} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800691
692/*
693 * RX queue sysfs structures and functions.
694 */
695struct rx_queue_attribute {
696 struct attribute attr;
697 ssize_t (*show)(struct netdev_rx_queue *queue,
698 struct rx_queue_attribute *attr, char *buf);
699 ssize_t (*store)(struct netdev_rx_queue *queue,
700 struct rx_queue_attribute *attr, const char *buf, size_t len);
701};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800702
Tom Herbertbf264142010-11-26 08:36:09 +0000703#ifdef CONFIG_XPS
704/*
705 * This structure holds an XPS map which can be of variable length. The
706 * map is an array of queues.
707 */
708struct xps_map {
709 unsigned int len;
710 unsigned int alloc_len;
711 struct rcu_head rcu;
712 u16 queues[0];
713};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000714#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Helge Dellerc59f4192015-10-25 10:00:32 +0100715#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
716 - sizeof(struct xps_map)) / sizeof(u16))
Tom Herbertbf264142010-11-26 08:36:09 +0000717
718/*
719 * This structure holds all XPS maps for device. Maps are indexed by CPU.
720 */
721struct xps_dev_maps {
722 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000723 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000724};
Alexander Duyck184c4492016-10-28 11:50:13 -0400725#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
726 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
Tom Herbertbf264142010-11-26 08:36:09 +0000727#endif /* CONFIG_XPS */
728
John Fastabend4f57c082011-01-17 08:06:04 +0000729#define TC_MAX_QUEUE 16
730#define TC_BITMASK 15
731/* HW offloaded queuing disciplines txq count and offset maps */
732struct netdev_tc_txq {
733 u16 count;
734 u16 offset;
735};
736
Neerav Parikh68bad942012-01-04 20:23:39 +0000737#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
738/*
739 * This structure is to hold information about the device
740 * configured to run FCoE protocol stack.
741 */
742struct netdev_fcoe_hbainfo {
743 char manufacturer[64];
744 char serial_number[64];
745 char hardware_version[64];
746 char driver_version[64];
747 char optionrom_version[64];
748 char firmware_version[64];
749 char model[256];
750 char model_description[256];
751};
752#endif
753
Jiri Pirko02637fc2014-11-28 14:34:16 +0100754#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200755
Jiri Pirko02637fc2014-11-28 14:34:16 +0100756/* This structure holds a unique identifier to identify some
757 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200758 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100759struct netdev_phys_item_id {
760 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200761 unsigned char id_len;
762};
763
Scott Feldmand754f982015-07-18 18:24:49 -0700764static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
765 struct netdev_phys_item_id *b)
766{
767 return a->id_len == b->id_len &&
768 memcmp(a->id, b->id, a->id_len) == 0;
769}
770
Daniel Borkmann99932d42014-02-16 15:55:20 +0100771typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
772 struct sk_buff *skb);
773
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800774/* These structures hold the attributes of qdisc and classifiers
John Fastabend16e5cc62016-02-16 21:16:43 -0800775 * that are being passed to the netdevice through the setup_tc op.
776 */
777enum {
778 TC_SETUP_MQPRIO,
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800779 TC_SETUP_CLSU32,
Amir Vadai5b33f482016-03-08 12:42:29 +0200780 TC_SETUP_CLSFLOWER,
Yotam Gigib87f7932016-07-21 12:03:12 +0200781 TC_SETUP_MATCHALL,
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100782 TC_SETUP_CLSBPF,
John Fastabend16e5cc62016-02-16 21:16:43 -0800783};
784
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800785struct tc_cls_u32_offload;
786
John Fastabend16e5cc62016-02-16 21:16:43 -0800787struct tc_to_netdev {
788 unsigned int type;
789 union {
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800790 struct tc_cls_u32_offload *cls_u32;
Amir Vadai5b33f482016-03-08 12:42:29 +0200791 struct tc_cls_flower_offload *cls_flower;
Yotam Gigib87f7932016-07-21 12:03:12 +0200792 struct tc_cls_matchall_offload *cls_mall;
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100793 struct tc_cls_bpf_offload *cls_bpf;
Amritha Nambiar56f36ac2017-03-15 10:39:25 -0700794 struct tc_mqprio_qopt *mqprio;
John Fastabend16e5cc62016-02-16 21:16:43 -0800795 };
Hadar Hen Zion7091d8c2016-12-01 14:06:37 +0200796 bool egress_dev;
John Fastabend16e5cc62016-02-16 21:16:43 -0800797};
798
Brenden Blancoa7862b42016-07-19 12:16:48 -0700799/* These structures hold the attributes of xdp state that are being passed
800 * to the netdevice through the xdp op.
801 */
802enum xdp_netdev_command {
803 /* Set or clear a bpf program used in the earliest stages of packet
804 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
805 * is responsible for calling bpf_prog_put on any old progs that are
806 * stored. In case of error, the callee need not release the new prog
807 * reference, but on success it takes ownership and must bpf_prog_put
808 * when it is no longer used.
809 */
810 XDP_SETUP_PROG,
811 /* Check if a bpf program is set on the device. The callee should
812 * return true if a program is currently attached and running.
813 */
814 XDP_QUERY_PROG,
815};
816
817struct netdev_xdp {
818 enum xdp_netdev_command command;
819 union {
820 /* XDP_SETUP_PROG */
821 struct bpf_prog *prog;
822 /* XDP_QUERY_PROG */
823 bool prog_attached;
824 };
825};
John Fastabend16e5cc62016-02-16 21:16:43 -0800826
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800827/*
828 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800829 * The following hooks can be defined; unless noted otherwise, they are
830 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800831 *
832 * int (*ndo_init)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500833 * This function is called once when a network device is registered.
834 * The network device can use this for any late stage initialization
835 * or semantic validation. It can fail with an error code which will
836 * be propagated back to register_netdev.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800837 *
838 * void (*ndo_uninit)(struct net_device *dev);
839 * This function is called when device is unregistered or when registration
840 * fails. It is not called if init fails.
841 *
842 * int (*ndo_open)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500843 * This function is called when a network device transitions to the up
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800844 * state.
845 *
846 * int (*ndo_stop)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500847 * This function is called when a network device transitions to the down
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800848 * state.
849 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000850 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
851 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800852 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030853 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
854 * the queue before that can happen; it's for obsolete devices and weird
855 * corner cases, but the stack really does a non-trivial amount
856 * of useless work if you return NETDEV_TX_BUSY.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500857 * Required; cannot be NULL.
Stephen Hemminger00829822008-11-20 20:14:53 -0800858 *
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -0800859 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
860 * struct net_device *dev
861 * netdev_features_t features);
862 * Called by core transmit path to determine if device is capable of
863 * performing offload operations on a given packet. This is to give
864 * the device an opportunity to implement any restrictions that cannot
865 * be otherwise expressed by feature flags. The check is called with
866 * the set of features that the stack has calculated and it returns
867 * those the driver believes to be appropriate.
Eric Dumazetcdba7562016-01-06 06:53:50 -0800868 *
Jason Wangf663dd92014-01-10 16:18:26 +0800869 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100870 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500871 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800872 * transmit queues.
873 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800874 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
875 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500876 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800877 *
878 * void (*ndo_set_rx_mode)(struct net_device *dev);
879 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000880 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500881 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800882 *
883 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
884 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700885 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500886 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800887 *
888 * int (*ndo_validate_addr)(struct net_device *dev);
889 * Test if Media Access Control address is valid for the device.
890 *
891 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500892 * Called when a user requests an ioctl which can't be handled by
893 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800894 * not supported error code.
895 *
896 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
897 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500898 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800899 * interface (PCI) for low level management.
900 *
901 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
902 * Called when a user wants to change the Maximum Transfer Unit
903 * of a device. If not defined, any request to change MTU will
904 * will return an error.
905 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800906 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500907 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800908 * for dev->watchdog ticks.
909 *
stephen hemmingerbc1f4472017-01-06 19:12:52 -0800910 * void (*ndo_get_stats64)(struct net_device *dev,
911 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700912 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800913 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000914 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000915 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
916 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700917 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000918 * (which should normally be dev->stats) and return a pointer to
919 * it. The structure may be changed asynchronously only if each
920 * field is written atomically.
921 * 3. Update dev->stats asynchronously and atomically, and define
922 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800923 *
Or Gerlitz3df5b3c2016-11-22 23:09:54 +0200924 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
Nogah Frankel2c9d85d2016-09-16 15:05:36 +0200925 * Return true if this device supports offload stats of this attr_id.
926 *
927 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
928 * void *attr_data)
929 * Get statistics for offload operations by attr_id. Write it into the
930 * attr_data pointer.
931 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530932 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500933 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000934 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800935 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530936 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500937 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000938 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800939 *
940 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000941 *
942 * SR-IOV management functions.
943 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
Moshe Shemesh79aab092016-09-22 12:11:15 +0300944 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
945 * u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400946 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
947 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000948 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000949 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000950 * int (*ndo_get_vf_config)(struct net_device *dev,
951 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300952 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700953 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
954 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300955 *
956 * Enable or disable the VF ability to query its RSS Redirection Table and
957 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500958 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300959 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700960 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
Florian Fainelli6a4bc2b2017-01-26 14:44:17 -0800961 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
962 * __be16 protocol, struct tc_to_netdev *tc);
963 * Called to setup any 'tc' scheduler, classifier or action on @dev.
964 * This is always called from the stack with the rtnl lock held and netif
965 * tx queues stopped. This allows the netdevice to perform queue
966 * management safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000967 *
Yi Zoue9bce842011-03-09 08:48:03 +0000968 * Fiber Channel over Ethernet (FCoE) offload functions.
969 * int (*ndo_fcoe_enable)(struct net_device *dev);
970 * Called when the FCoE protocol stack wants to start using LLD for FCoE
971 * so the underlying device can perform whatever needed configuration or
972 * initialization to support acceleration of FCoE traffic.
973 *
974 * int (*ndo_fcoe_disable)(struct net_device *dev);
975 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
976 * so the underlying device can perform whatever needed clean-ups to
977 * stop supporting acceleration of FCoE traffic.
978 *
979 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
980 * struct scatterlist *sgl, unsigned int sgc);
981 * Called when the FCoE Initiator wants to initialize an I/O that
982 * is a possible candidate for Direct Data Placement (DDP). The LLD can
983 * perform necessary setup and returns 1 to indicate the device is set up
984 * successfully to perform DDP on this I/O, otherwise this returns 0.
985 *
986 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
987 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
988 * indicated by the FC exchange id 'xid', so the underlying device can
989 * clean up and reuse resources for later DDP requests.
990 *
991 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
992 * struct scatterlist *sgl, unsigned int sgc);
993 * Called when the FCoE Target wants to initialize an I/O that
994 * is a possible candidate for Direct Data Placement (DDP). The LLD can
995 * perform necessary setup and returns 1 to indicate the device is set up
996 * successfully to perform DDP on this I/O, otherwise this returns 0.
997 *
Neerav Parikh68bad942012-01-04 20:23:39 +0000998 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
999 * struct netdev_fcoe_hbainfo *hbainfo);
1000 * Called when the FCoE Protocol stack wants information on the underlying
1001 * device. This information is utilized by the FCoE protocol stack to
1002 * register attributes with Fiber Channel management service as per the
1003 * FC-GS Fabric Device Management Information(FDMI) specification.
1004 *
Yi Zoue9bce842011-03-09 08:48:03 +00001005 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1006 * Called when the underlying device wants to override default World Wide
1007 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1008 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1009 * protocol stack to use.
1010 *
Ben Hutchingsc4454772011-01-19 11:03:53 +00001011 * RFS acceleration.
1012 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1013 * u16 rxq_index, u32 flow_id);
1014 * Set hardware filter for RFS. rxq_index is the target queue index;
1015 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1016 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001017 *
Jiri Pirko8b98a702013-01-03 22:49:02 +00001018 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001019 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1020 * Called to make another netdev an underling.
1021 *
1022 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1023 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +00001024 *
1025 * Feature/offload setting functions.
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -08001026 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1027 * netdev_features_t features);
1028 * Adjusts the requested feature flags according to device-specific
1029 * constraints, and returns the resulting flags. Must not modify
1030 * the device state.
1031 *
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001032 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001033 * Called to update device configuration to new features. Passed
1034 * feature set might be less than what was returned by ndo_fix_features()).
1035 * Must return >0 or -errno if it changed dev->features itself.
1036 *
stephen hemmingeredc7d572012-10-01 12:32:33 +00001037 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1038 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001039 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +00001040 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +00001041 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1042 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001043 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001044 * Deletes the FDB entry from dev coresponding to addr.
1045 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001046 * struct net_device *dev, struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001047 * int *idx)
John Fastabend77162022012-04-15 06:43:56 +00001048 * Used to add FDB entries to dump requests. Implementers should add
1049 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001050 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001051 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1052 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001053 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001054 * struct net_device *dev, u32 filter_mask,
1055 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001056 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1057 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001058 *
1059 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1060 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1061 * which do not represent real hardware may define this to allow their
1062 * userspace components to manage their virtual carrier state. Devices
1063 * that determine carrier state from physical hardware properties (eg
1064 * network cables) or protocol-dependent mechanisms (eg
1065 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001066 *
1067 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001068 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001069 * Called to get ID of physical port of this device. If driver does
1070 * not implement this, it is assumed that the hw is not able to have
1071 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001072 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001073 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1074 * struct udp_tunnel_info *ti);
1075 * Called by UDP tunnel to notify a driver about the UDP port and socket
1076 * address family that a UDP tunnel is listnening to. It is called only
1077 * when a new port starts listening. The operation is protected by the
1078 * RTNL.
1079 *
1080 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1081 * struct udp_tunnel_info *ti);
1082 * Called by UDP tunnel to notify the driver about a UDP port and socket
1083 * address family that the UDP tunnel is not listening to anymore. The
1084 * operation is protected by the RTNL.
1085 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001086 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1087 * struct net_device *dev)
1088 * Called by upper layer devices to accelerate switching or other
1089 * station functionality into hardware. 'pdev is the lowerdev
1090 * to use for the offload and 'dev' is the net device that will
1091 * back the offload. Returns a pointer to the private structure
1092 * the upper layer will maintain.
1093 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1094 * Called by upper layer device to delete the station created
1095 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1096 * the station and priv is the structure returned by the add
1097 * operation.
1098 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1099 * struct net_device *dev,
1100 * void *priv);
1101 * Callback to use for xmit over the accelerated station. This
1102 * is used in place of ndo_start_xmit on accelerated net
1103 * devices.
John Fastabend822b3b22015-03-18 14:57:33 +02001104 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1105 * int queue_index, u32 maxrate);
1106 * Called when a user wants to set a max-rate limitation of specific
1107 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001108 * int (*ndo_get_iflink)(const struct net_device *dev);
1109 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001110 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001111 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001112 * This function is used to pass protocol port error state information
1113 * to the switch driver. The switch driver can react to the proto_down
1114 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001115 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1116 * This function is used to get egress tunnel information for given skb.
1117 * This is useful for retrieving outer tunnel header parameters while
1118 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001119 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1120 * This function is used to specify the headroom that the skb must
1121 * consider when allocation skb during packet reception. Setting
1122 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001123 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001124 * default value.
Brenden Blancoa7862b42016-07-19 12:16:48 -07001125 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1126 * This function is used to set or query state related to XDP on the
1127 * netdevice. See definition of enum xdp_netdev_command for details.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001128 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001129 */
1130struct net_device_ops {
1131 int (*ndo_init)(struct net_device *dev);
1132 void (*ndo_uninit)(struct net_device *dev);
1133 int (*ndo_open)(struct net_device *dev);
1134 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001135 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1136 struct net_device *dev);
1137 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1138 struct net_device *dev,
1139 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001140 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001141 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001142 void *accel_priv,
1143 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001144 void (*ndo_change_rx_flags)(struct net_device *dev,
1145 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001146 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001147 int (*ndo_set_mac_address)(struct net_device *dev,
1148 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001149 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001150 int (*ndo_do_ioctl)(struct net_device *dev,
1151 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001152 int (*ndo_set_config)(struct net_device *dev,
1153 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001154 int (*ndo_change_mtu)(struct net_device *dev,
1155 int new_mtu);
1156 int (*ndo_neigh_setup)(struct net_device *dev,
1157 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001158 void (*ndo_tx_timeout) (struct net_device *dev);
1159
stephen hemmingerbc1f4472017-01-06 19:12:52 -08001160 void (*ndo_get_stats64)(struct net_device *dev,
1161 struct rtnl_link_stats64 *storage);
Or Gerlitz3df5b3c2016-11-22 23:09:54 +02001162 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
Nogah Frankel2c9d85d2016-09-16 15:05:36 +02001163 int (*ndo_get_offload_stats)(int attr_id,
1164 const struct net_device *dev,
1165 void *attr_data);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001166 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1167
Jiri Pirko8e586132011-12-08 19:52:37 -05001168 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001169 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001170 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001171 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001172#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001173 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001174 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001175 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001176 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001177#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001178 int (*ndo_set_vf_mac)(struct net_device *dev,
1179 int queue, u8 *mac);
1180 int (*ndo_set_vf_vlan)(struct net_device *dev,
Moshe Shemesh79aab092016-09-22 12:11:15 +03001181 int queue, u16 vlan,
1182 u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001183 int (*ndo_set_vf_rate)(struct net_device *dev,
1184 int vf, int min_tx_rate,
1185 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001186 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1187 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001188 int (*ndo_set_vf_trust)(struct net_device *dev,
1189 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001190 int (*ndo_get_vf_config)(struct net_device *dev,
1191 int vf,
1192 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001193 int (*ndo_set_vf_link_state)(struct net_device *dev,
1194 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001195 int (*ndo_get_vf_stats)(struct net_device *dev,
1196 int vf,
1197 struct ifla_vf_stats
1198 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001199 int (*ndo_set_vf_port)(struct net_device *dev,
1200 int vf,
1201 struct nlattr *port[]);
1202 int (*ndo_get_vf_port)(struct net_device *dev,
1203 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001204 int (*ndo_set_vf_guid)(struct net_device *dev,
1205 int vf, u64 guid,
1206 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001207 int (*ndo_set_vf_rss_query_en)(
1208 struct net_device *dev,
1209 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001210 int (*ndo_setup_tc)(struct net_device *dev,
1211 u32 handle,
1212 __be16 protocol,
1213 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001214#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001215 int (*ndo_fcoe_enable)(struct net_device *dev);
1216 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001217 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1218 u16 xid,
1219 struct scatterlist *sgl,
1220 unsigned int sgc);
1221 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1222 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001223 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1224 u16 xid,
1225 struct scatterlist *sgl,
1226 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001227 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1228 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001229#endif
1230
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001231#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001232#define NETDEV_FCOE_WWNN 0
1233#define NETDEV_FCOE_WWPN 1
1234 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1235 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001236#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001237
Ben Hutchingsc4454772011-01-19 11:03:53 +00001238#ifdef CONFIG_RFS_ACCEL
1239 int (*ndo_rx_flow_steer)(struct net_device *dev,
1240 const struct sk_buff *skb,
1241 u16 rxq_index,
1242 u32 flow_id);
1243#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001244 int (*ndo_add_slave)(struct net_device *dev,
1245 struct net_device *slave_dev);
1246 int (*ndo_del_slave)(struct net_device *dev,
1247 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001248 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1249 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001250 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001251 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001252 int (*ndo_neigh_construct)(struct net_device *dev,
1253 struct neighbour *n);
1254 void (*ndo_neigh_destroy)(struct net_device *dev,
1255 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001256
1257 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001258 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001259 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001260 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001261 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001262 u16 flags);
1263 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001264 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001265 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001266 const unsigned char *addr,
1267 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001268 int (*ndo_fdb_dump)(struct sk_buff *skb,
1269 struct netlink_callback *cb,
1270 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001271 struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001272 int *idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001273
1274 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001275 struct nlmsghdr *nlh,
1276 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001277 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1278 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001279 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001280 u32 filter_mask,
1281 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001282 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001283 struct nlmsghdr *nlh,
1284 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001285 int (*ndo_change_carrier)(struct net_device *dev,
1286 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001287 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001288 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001289 int (*ndo_get_phys_port_name)(struct net_device *dev,
1290 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001291 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1292 struct udp_tunnel_info *ti);
1293 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1294 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001295 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1296 struct net_device *dev);
1297 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1298 void *priv);
1299
1300 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1301 struct net_device *dev,
1302 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001303 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001304 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1305 int queue_index,
1306 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001307 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001308 int (*ndo_change_proto_down)(struct net_device *dev,
1309 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001310 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1311 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001312 void (*ndo_set_rx_headroom)(struct net_device *dev,
1313 int needed_headroom);
Brenden Blancoa7862b42016-07-19 12:16:48 -07001314 int (*ndo_xdp)(struct net_device *dev,
1315 struct netdev_xdp *xdp);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001316};
1317
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001318/**
1319 * enum net_device_priv_flags - &struct net_device priv_flags
1320 *
1321 * These are the &struct net_device, they are only set internally
1322 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001323 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001324 * during any kernel release.
1325 *
1326 * You should have a pretty good reason to be extending these flags.
1327 *
1328 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1329 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001330 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001331 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001332 * @IFF_WAN_HDLC: WAN HDLC device
1333 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1334 * release skb->dst
1335 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1336 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1337 * @IFF_MACVLAN_PORT: device used as macvlan port
1338 * @IFF_BRIDGE_PORT: device used as bridge port
1339 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1340 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1341 * @IFF_UNICAST_FLT: Supports unicast filtering
1342 * @IFF_TEAM_PORT: device used as team port
1343 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1344 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1345 * change when it's running
1346 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001347 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1348 * underlying stacked devices
1349 * @IFF_IPVLAN_MASTER: IPvlan master device
1350 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001351 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001352 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001353 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001354 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001355 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001356 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001357 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1358 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001359 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001360 */
1361enum netdev_priv_flags {
1362 IFF_802_1Q_VLAN = 1<<0,
1363 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001364 IFF_BONDING = 1<<2,
1365 IFF_ISATAP = 1<<3,
1366 IFF_WAN_HDLC = 1<<4,
1367 IFF_XMIT_DST_RELEASE = 1<<5,
1368 IFF_DONT_BRIDGE = 1<<6,
1369 IFF_DISABLE_NETPOLL = 1<<7,
1370 IFF_MACVLAN_PORT = 1<<8,
1371 IFF_BRIDGE_PORT = 1<<9,
1372 IFF_OVS_DATAPATH = 1<<10,
1373 IFF_TX_SKB_SHARING = 1<<11,
1374 IFF_UNICAST_FLT = 1<<12,
1375 IFF_TEAM_PORT = 1<<13,
1376 IFF_SUPP_NOFCS = 1<<14,
1377 IFF_LIVE_ADDR_CHANGE = 1<<15,
1378 IFF_MACVLAN = 1<<16,
1379 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1380 IFF_IPVLAN_MASTER = 1<<18,
1381 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001382 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001383 IFF_NO_QUEUE = 1<<21,
1384 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001385 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001386 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001387 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001388 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001389 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001390};
1391
1392#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1393#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001394#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001395#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001396#define IFF_WAN_HDLC IFF_WAN_HDLC
1397#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1398#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1399#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1400#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1401#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1402#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1403#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1404#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1405#define IFF_TEAM_PORT IFF_TEAM_PORT
1406#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1407#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1408#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001409#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001410#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1411#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001412#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001413#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001414#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001415#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001416#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001417#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001418#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001419
Karoly Kemeny536721b2014-07-30 20:27:36 +02001420/**
1421 * struct net_device - The DEVICE structure.
1422 * Actually, this whole structure is a big mistake. It mixes I/O
1423 * data with strictly "high-level" data, and it has to know about
1424 * almost every data structure used in the INET module.
1425 *
1426 * @name: This is the first field of the "visible" part of this structure
1427 * (i.e. as seen by users in the "Space.c" file). It is the name
1428 * of the interface.
1429 *
1430 * @name_hlist: Device name hash chain, please keep it close to name[]
1431 * @ifalias: SNMP alias
1432 * @mem_end: Shared memory end
1433 * @mem_start: Shared memory start
1434 * @base_addr: Device I/O address
1435 * @irq: Device IRQ number
1436 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001437 * @carrier_changes: Stats to monitor carrier on<->off transitions
1438 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001439 * @state: Generic network queuing layer state, see netdev_state_t
1440 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001441 * @napi_list: List entry used for polling NAPI devices
1442 * @unreg_list: List entry when we are unregistering the
1443 * device; see the function unregister_netdev
1444 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001445 * @ptype_all: Device-specific packet handlers for all protocols
1446 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001447 *
1448 * @adj_list: Directly linked devices, like slaves for bonding
Karoly Kemeny536721b2014-07-30 20:27:36 +02001449 * @features: Currently active device features
1450 * @hw_features: User-changeable features
1451 *
1452 * @wanted_features: User-requested features
1453 * @vlan_features: Mask of features inheritable by VLAN devices
1454 *
1455 * @hw_enc_features: Mask of features inherited by encapsulating devices
1456 * This field indicates what encapsulation
1457 * offloads the hardware is capable of doing,
1458 * and drivers will need to set them appropriately.
1459 *
1460 * @mpls_features: Mask of features inheritable by MPLS
1461 *
1462 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001463 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001464 *
1465 * @stats: Statistics struct, which was left as a legacy, use
1466 * rtnl_link_stats64 instead
1467 *
1468 * @rx_dropped: Dropped packets by core network,
1469 * do not use this in drivers
1470 * @tx_dropped: Dropped packets by core network,
1471 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001472 * @rx_nohandler: nohandler dropped packets by core network on
1473 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001474 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001475 * @wireless_handlers: List of functions to handle Wireless Extensions,
1476 * instead of ioctl,
1477 * see <net/iw_handler.h> for details.
1478 * @wireless_data: Instance data managed by the core of wireless extensions
1479 *
1480 * @netdev_ops: Includes several pointers to callbacks,
1481 * if one wants to override the ndo_*() functions
1482 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001483 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1484 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001485 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001486 * of Layer 2 headers.
1487 *
1488 * @flags: Interface flags (a la BSD)
1489 * @priv_flags: Like 'flags' but invisible to userspace,
1490 * see if.h for the definitions
1491 * @gflags: Global flags ( kept as legacy )
1492 * @padded: How much padding added by alloc_netdev()
1493 * @operstate: RFC2863 operstate
1494 * @link_mode: Mapping policy to operstate
1495 * @if_port: Selectable AUI, TP, ...
1496 * @dma: DMA channel
1497 * @mtu: Interface MTU value
Jarod Wilson61e84622016-10-07 22:04:33 -04001498 * @min_mtu: Interface Minimum MTU value
1499 * @max_mtu: Interface Maximum MTU value
Karoly Kemeny536721b2014-07-30 20:27:36 +02001500 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001501 * @hard_header_len: Maximum hardware header length.
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05001502 * @min_header_len: Minimum hardware header length
Karoly Kemeny536721b2014-07-30 20:27:36 +02001503 *
1504 * @needed_headroom: Extra headroom the hardware may need, but not in all
1505 * cases can this be guaranteed
1506 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1507 * cases can this be guaranteed. Some cases also use
1508 * LL_MAX_HEADER instead to allocate the skb
1509 *
1510 * interface address info:
1511 *
1512 * @perm_addr: Permanent hw address
1513 * @addr_assign_type: Hw address assignment type
1514 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001515 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001516 * @dev_id: Used to differentiate devices that share
1517 * the same link layer address
1518 * @dev_port: Used to differentiate devices that share
1519 * the same function
1520 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001521 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001522 * has been enabled due to the need to listen to
1523 * additional unicast addresses in a device that
1524 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001525 * @uc: unicast mac addresses
1526 * @mc: multicast mac addresses
1527 * @dev_addrs: list of device hw addresses
1528 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001529 * @promiscuity: Number of times the NIC is told to work in
1530 * promiscuous mode; if it becomes 0 the NIC will
1531 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001532 * @allmulti: Counter, enables or disables allmulticast mode
1533 *
1534 * @vlan_info: VLAN info
1535 * @dsa_ptr: dsa specific data
1536 * @tipc_ptr: TIPC specific data
1537 * @atalk_ptr: AppleTalk link
1538 * @ip_ptr: IPv4 specific data
1539 * @dn_ptr: DECnet specific data
1540 * @ip6_ptr: IPv6 specific data
1541 * @ax25_ptr: AX.25 specific data
1542 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1543 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001544 * @dev_addr: Hw address (before bcast,
1545 * because most packets are unicast)
1546 *
1547 * @_rx: Array of RX queues
1548 * @num_rx_queues: Number of RX queues
1549 * allocated at register_netdev() time
1550 * @real_num_rx_queues: Number of RX queues currently active in device
1551 *
1552 * @rx_handler: handler for received packets
1553 * @rx_handler_data: XXX: need comments on this one
1554 * @ingress_queue: XXX: need comments on this one
1555 * @broadcast: hw bcast address
1556 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001557 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1558 * indexed by RX queue number. Assigned by driver.
1559 * This must only be set if the ndo_rx_flow_steer
1560 * operation is defined
1561 * @index_hlist: Device index hash chain
1562 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001563 * @_tx: Array of TX queues
1564 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1565 * @real_num_tx_queues: Number of TX queues currently active in device
1566 * @qdisc: Root qdisc from userspace point of view
1567 * @tx_queue_len: Max frames per queue allowed
1568 * @tx_global_lock: XXX: need comments on this one
1569 *
1570 * @xps_maps: XXX: need comments on this one
1571 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001572 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001573 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001574 * @watchdog_timer: List of timers
1575 *
1576 * @pcpu_refcnt: Number of references to this device
1577 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001578 * @link_watch_list: XXX: need comments on this one
1579 *
1580 * @reg_state: Register/unregister state machine
1581 * @dismantle: Device is going to be freed
1582 * @rtnl_link_state: This enum represents the phases of creating
1583 * a new link
1584 *
1585 * @destructor: Called from unregister,
1586 * can be used to call free_netdev
1587 * @npinfo: XXX: need comments on this one
1588 * @nd_net: Network namespace this network device is inside
1589 *
1590 * @ml_priv: Mid-layer private
1591 * @lstats: Loopback statistics
1592 * @tstats: Tunnel statistics
1593 * @dstats: Dummy statistics
1594 * @vstats: Virtual ethernet statistics
1595 *
1596 * @garp_port: GARP
1597 * @mrp_port: MRP
1598 *
1599 * @dev: Class/net/name entry
1600 * @sysfs_groups: Space for optional device, statistics and wireless
1601 * sysfs groups
1602 *
1603 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1604 * @rtnl_link_ops: Rtnl_link_ops
1605 *
1606 * @gso_max_size: Maximum size of generic segmentation offload
1607 * @gso_max_segs: Maximum number of segments that can be passed to the
1608 * NIC for GSO
1609 *
1610 * @dcbnl_ops: Data Center Bridging netlink ops
1611 * @num_tc: Number of traffic classes in the net device
1612 * @tc_to_txq: XXX: need comments on this one
Randy Dunlap920c1cd2016-11-21 18:28:36 -08001613 * @prio_tc_map: XXX: need comments on this one
Karoly Kemeny536721b2014-07-30 20:27:36 +02001614 *
1615 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1616 *
1617 * @priomap: XXX: need comments on this one
1618 * @phydev: Physical device may attach itself
1619 * for hardware timestamping
1620 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001621 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1622 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001623 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001624 * @proto_down: protocol port state information can be sent to the
1625 * switch driver and used to set the phys state of the
1626 * switch port.
1627 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 * FIXME: cleanup struct net_device such that network protocol info
1629 * moves out.
1630 */
1631
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001632struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001634 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001635 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 /*
1637 * I/O specific fields
1638 * FIXME: Merge these and struct ifmap into one
1639 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001640 unsigned long mem_end;
1641 unsigned long mem_start;
1642 unsigned long base_addr;
1643 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001645 atomic_t carrier_changes;
1646
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001648 * Some hardware also needs these fields (state,dev_list,
1649 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 * part of the usual set specified in Space.c.
1651 */
1652
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 unsigned long state;
1654
Pavel Emelianov7562f872007-05-03 15:13:45 -07001655 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001656 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001657 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001658 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001659 struct list_head ptype_all;
1660 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001661
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001662 struct {
1663 struct list_head upper;
1664 struct list_head lower;
1665 } adj_list;
1666
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001667 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001668 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001669 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001670 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001671 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001672 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001673 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001674
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001676 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677
Rusty Russellc45d2862007-03-28 14:29:08 -07001678 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001679
Eric Dumazet015f0682014-03-27 08:45:56 -07001680 atomic_long_t rx_dropped;
1681 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001682 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683
Johannes Bergb86e0282007-04-26 20:48:23 -07001684#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001685 const struct iw_handler_def *wireless_handlers;
1686 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001687#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001688 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001689 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001690#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001691 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001692#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001693#ifdef CONFIG_NET_L3_MASTER_DEV
1694 const struct l3mdev_ops *l3mdev_ops;
1695#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001696#if IS_ENABLED(CONFIG_IPV6)
1697 const struct ndisc_ops *ndisc_ops;
1698#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001700 const struct header_ops *header_ops;
1701
Karoly Kemeny536721b2014-07-30 20:27:36 +02001702 unsigned int flags;
1703 unsigned int priv_flags;
1704
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001706 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707
Karoly Kemeny536721b2014-07-30 20:27:36 +02001708 unsigned char operstate;
1709 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001710
Karoly Kemeny536721b2014-07-30 20:27:36 +02001711 unsigned char if_port;
1712 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001713
Karoly Kemeny536721b2014-07-30 20:27:36 +02001714 unsigned int mtu;
Jarod Wilson61e84622016-10-07 22:04:33 -04001715 unsigned int min_mtu;
1716 unsigned int max_mtu;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001717 unsigned short type;
1718 unsigned short hard_header_len;
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05001719 unsigned short min_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720
Johannes Bergf5184d22008-05-12 20:48:31 -07001721 unsigned short needed_headroom;
1722 unsigned short needed_tailroom;
1723
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001725 unsigned char perm_addr[MAX_ADDR_LEN];
1726 unsigned char addr_assign_type;
1727 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001728 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001729 unsigned short dev_id;
1730 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001731 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001732 unsigned char name_assign_type;
1733 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001734 struct netdev_hw_addr_list uc;
1735 struct netdev_hw_addr_list mc;
1736 struct netdev_hw_addr_list dev_addrs;
1737
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001738#ifdef CONFIG_SYSFS
1739 struct kset *queues_kset;
1740#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001741 unsigned int promiscuity;
1742 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001745 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001746
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001747#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001748 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001749#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001750#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001751 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001752#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001753#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001754 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001755#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001756 void *atalk_ptr;
1757 struct in_device __rcu *ip_ptr;
1758 struct dn_dev __rcu *dn_ptr;
1759 struct inet6_dev __rcu *ip6_ptr;
1760 void *ax25_ptr;
1761 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001762 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001763#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1764 struct mpls_dev __rcu *mpls_ptr;
1765#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001767/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001768 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001769 */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001770 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001771 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001772
Michael Daltona953be52014-01-16 22:23:28 -08001773#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001774 struct netdev_rx_queue *_rx;
1775
Tom Herbert0a9627f2010-03-16 08:03:29 +00001776 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001777 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001778#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001779
Eric Dumazet3b47d302014-11-06 21:09:44 -08001780 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001781 rx_handler_func_t __rcu *rx_handler;
1782 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001783
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001784#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001785 struct tcf_proto __rcu *ingress_cl_list;
1786#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001787 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001788#ifdef CONFIG_NETFILTER_INGRESS
Aaron Conolee3b37f12016-09-21 11:35:07 -04001789 struct nf_hook_entry __rcu *nf_hooks_ingress;
Pablo Neirae687ad62015-05-13 18:19:38 +02001790#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001791
Karoly Kemeny536721b2014-07-30 20:27:36 +02001792 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001793#ifdef CONFIG_RFS_ACCEL
1794 struct cpu_rmap *rx_cpu_rmap;
1795#endif
1796 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001797
1798/*
1799 * Cache lines mostly used on transmit path
1800 */
David S. Millere8a04642008-07-17 00:34:19 -07001801 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1802 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001803 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001804 struct Qdisc *qdisc;
Jiri Kosina59cc1f62016-08-10 11:05:15 +02001805#ifdef CONFIG_NET_SCHED
1806 DECLARE_HASHTABLE (qdisc_hash, 4);
1807#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001808 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001809 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001810 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001811
Tom Herbertbf264142010-11-26 08:36:09 +00001812#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001813 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001814#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001815#ifdef CONFIG_NET_CLS_ACT
1816 struct tcf_proto __rcu *egress_cl_list;
1817#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001818
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001819 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001820 struct timer_list watchdog_timer;
1821
Eric Dumazet29b44332010-10-11 10:22:12 +00001822 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824
Eric Dumazete014deb2009-11-17 05:59:21 +00001825 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001826
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001828 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829 NETREG_UNREGISTERING, /* called unregister_netdevice */
1830 NETREG_UNREGISTERED, /* completed unregister todo */
1831 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001832 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001833 } reg_state:8;
1834
Karoly Kemeny536721b2014-07-30 20:27:36 +02001835 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001836
1837 enum {
1838 RTNL_LINK_INITIALIZED,
1839 RTNL_LINK_INITIALIZING,
1840 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001841
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001842 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001845 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846#endif
David S. Millereae792b2008-07-15 03:03:33 -07001847
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001848 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001849
David S. Miller49517042008-05-12 03:29:11 -07001850 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001851 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001852 void *ml_priv;
1853 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001854 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001855 struct pcpu_dstats __percpu *dstats;
1856 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001857 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001858
Tobias Klauserfb585b42017-02-10 16:43:50 +01001859#if IS_ENABLED(CONFIG_GARP)
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001860 struct garp_port __rcu *garp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001861#endif
1862#if IS_ENABLED(CONFIG_MRP)
David Wardfebf0182013-02-08 17:17:06 +00001863 struct mrp_port __rcu *mrp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001864#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001866 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001867 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001868 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001869
Patrick McHardy38f7b872007-06-13 12:03:51 -07001870 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001871
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001872 /* for setting kernel sock attribute on TCP connection setup */
1873#define GSO_MAX_SIZE 65536
1874 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001875#define GSO_MAX_SEGS 65535
1876 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001877
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001878#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001879 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001880#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001881 u8 num_tc;
1882 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1883 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001884
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001885#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001886 unsigned int fcoe_ddp_xid;
1887#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001888#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001889 struct netprio_map __rcu *priomap;
1890#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001891 struct phy_device *phydev;
1892 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001893 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001894 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001896#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897
1898#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899
David S. Millere8a04642008-07-17 00:34:19 -07001900static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001901int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1902{
1903 return dev->prio_tc_map[prio & TC_BITMASK];
1904}
1905
1906static inline
1907int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1908{
1909 if (tc >= dev->num_tc)
1910 return -EINVAL;
1911
1912 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1913 return 0;
1914}
1915
Alexander Duyck8d059b02016-10-28 11:43:49 -04001916int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
Alexander Duyck9cf1f6a2016-10-28 11:43:20 -04001917void netdev_reset_tc(struct net_device *dev);
1918int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1919int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
John Fastabend4f57c082011-01-17 08:06:04 +00001920
1921static inline
1922int netdev_get_num_tc(struct net_device *dev)
1923{
1924 return dev->num_tc;
1925}
1926
1927static inline
David S. Millere8a04642008-07-17 00:34:19 -07001928struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1929 unsigned int index)
1930{
1931 return &dev->_tx[index];
1932}
1933
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001934static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1935 const struct sk_buff *skb)
1936{
1937 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1938}
1939
David S. Millere8a04642008-07-17 00:34:19 -07001940static inline void netdev_for_each_tx_queue(struct net_device *dev,
1941 void (*f)(struct net_device *,
1942 struct netdev_queue *,
1943 void *),
1944 void *arg)
1945{
1946 unsigned int i;
1947
1948 for (i = 0; i < dev->num_tx_queues; i++)
1949 f(dev, &dev->_tx[i], arg);
1950}
1951
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001952#define netdev_lockdep_set_classes(dev) \
1953{ \
1954 static struct lock_class_key qdisc_tx_busylock_key; \
1955 static struct lock_class_key qdisc_running_key; \
1956 static struct lock_class_key qdisc_xmit_lock_key; \
1957 static struct lock_class_key dev_addr_list_lock_key; \
1958 unsigned int i; \
1959 \
1960 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1961 (dev)->qdisc_running_key = &qdisc_running_key; \
1962 lockdep_set_class(&(dev)->addr_list_lock, \
1963 &dev_addr_list_lock_key); \
1964 for (i = 0; i < (dev)->num_tx_queues; i++) \
1965 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1966 &qdisc_xmit_lock_key); \
1967}
1968
Joe Perchesf629d202013-09-26 14:48:15 -07001969struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001970 struct sk_buff *skb,
1971 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001972
Paolo Abeni871b6422016-02-26 10:45:37 +01001973/* returns the headroom that the master device needs to take in account
1974 * when forwarding to this dev
1975 */
1976static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1977{
1978 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1979}
1980
1981static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1982{
1983 if (dev->netdev_ops->ndo_set_rx_headroom)
1984 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1985}
1986
1987/* set the device rx headroom to the dev's default */
1988static inline void netdev_reset_rx_headroom(struct net_device *dev)
1989{
1990 netdev_set_rx_headroom(dev, -1);
1991}
1992
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001993/*
1994 * Net namespace inlines
1995 */
1996static inline
1997struct net *dev_net(const struct net_device *dev)
1998{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001999 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002000}
2001
2002static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002003void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002004{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05002005 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002006}
2007
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002008/**
2009 * netdev_priv - access network device private data
2010 * @dev: network device
2011 *
2012 * Get network device private data
2013 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002014static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002016 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017}
2018
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019/* Set the sysfs physical device reference for the network logical device
2020 * if set prior to registration will cause a symlink during initialization.
2021 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002022#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023
Marcel Holtmann384912e2009-08-31 21:08:19 +00002024/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002025 * fine-grained identification of different network device types. For
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002026 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
Marcel Holtmann384912e2009-08-31 21:08:19 +00002027 */
2028#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2029
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002030/* Default NAPI poll() weight
2031 * Device drivers are strongly advised to not use bigger value
2032 */
2033#define NAPI_POLL_WEIGHT 64
2034
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002035/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002036 * netif_napi_add - initialize a NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002037 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002038 * @napi: NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002039 * @poll: polling function
2040 * @weight: default weight
2041 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002042 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2043 * *any* of the other NAPI-related functions.
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002044 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002045void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2046 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002047
Alexander Duyckd8156532008-07-08 15:13:05 -07002048/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002049 * netif_tx_napi_add - initialize a NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002050 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002051 * @napi: NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002052 * @poll: polling function
2053 * @weight: default weight
2054 *
2055 * This variant of netif_napi_add() should be used from drivers using NAPI
2056 * to exclusively poll a TX queue.
2057 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2058 */
2059static inline void netif_tx_napi_add(struct net_device *dev,
2060 struct napi_struct *napi,
2061 int (*poll)(struct napi_struct *, int),
2062 int weight)
2063{
2064 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2065 netif_napi_add(dev, napi, poll, weight);
2066}
2067
2068/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002069 * netif_napi_del - remove a NAPI context
2070 * @napi: NAPI context
Alexander Duyckd8156532008-07-08 15:13:05 -07002071 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002072 * netif_napi_del() removes a NAPI context from the network device NAPI list
Alexander Duyckd8156532008-07-08 15:13:05 -07002073 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002074void netif_napi_del(struct napi_struct *napi);
2075
2076struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002077 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002078 void *frag0;
Herbert Xu78a478d2009-05-26 18:50:21 +00002079
Herbert Xu74895942009-05-26 18:50:27 +00002080 /* Length of frag0. */
2081 unsigned int frag0_len;
2082
Herbert Xu86911732009-01-29 14:19:50 +00002083 /* This indicates where we are processing relative to skb->data. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002084 int data_offset;
Herbert Xu86911732009-01-29 14:19:50 +00002085
Herbert Xud565b0a2008-12-15 23:38:52 -08002086 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002087 u16 flush;
2088
2089 /* Save the IP ID here and check when we get to the transport layer */
2090 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002091
2092 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002093 u16 count;
2094
Tom Herbert15e23962015-02-10 16:30:31 -08002095 /* Start offset for remote checksum offload */
2096 u16 gro_remcsum_start;
2097
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002098 /* jiffies when first packet was created/queued */
2099 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002100
Tom Herbertafe93322014-09-17 12:25:57 -07002101 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002102 u16 proto;
2103
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002104 /* This is non-zero if the packet may be of the same flow. */
2105 u8 same_flow:1;
2106
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002107 /* Used in tunnel GRO receive */
2108 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002109
2110 /* GRO checksum is valid */
2111 u8 csum_valid:1;
2112
Tom Herbert662880f2014-08-27 21:26:56 -07002113 /* Number of checksums via CHECKSUM_UNNECESSARY */
2114 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002115
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002116 /* Free the skb? */
2117 u8 free:2;
2118#define NAPI_GRO_FREE 1
2119#define NAPI_GRO_FREE_STOLEN_HEAD 2
2120
Tom Herbertefc98d02014-10-03 15:48:08 -07002121 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2122 u8 is_ipv6:1;
2123
Alexander Duycka0ca1532016-04-05 09:13:39 -07002124 /* Used in GRE, set in fou/gue_gro_receive */
2125 u8 is_fou:1;
2126
Alexander Duyck15305452016-04-10 21:44:57 -04002127 /* Used to determine if flush_id can be ignored */
2128 u8 is_atomic:1;
2129
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002130 /* Number of gro_receive callbacks this packet already went through */
2131 u8 recursion_counter:4;
2132
2133 /* 1 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002134
Jerry Chubf5a7552014-01-07 10:23:19 -08002135 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2136 __wsum csum;
2137
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002138 /* used in skb_gro_receive() slow path */
2139 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002140};
2141
2142#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002143
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002144#define GRO_RECURSION_LIMIT 15
2145static inline int gro_recursion_inc_test(struct sk_buff *skb)
2146{
2147 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2148}
2149
2150typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2151static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2152 struct sk_buff **head,
2153 struct sk_buff *skb)
2154{
2155 if (unlikely(gro_recursion_inc_test(skb))) {
2156 NAPI_GRO_CB(skb)->flush |= 1;
2157 return NULL;
2158 }
2159
2160 return cb(head, skb);
2161}
2162
2163typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2164 struct sk_buff *);
2165static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2166 struct sock *sk,
2167 struct sk_buff **head,
2168 struct sk_buff *skb)
2169{
2170 if (unlikely(gro_recursion_inc_test(skb))) {
2171 NAPI_GRO_CB(skb)->flush |= 1;
2172 return NULL;
2173 }
2174
2175 return cb(sk, head, skb);
2176}
2177
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002179 __be16 type; /* This is really htons(ether_type). */
2180 struct net_device *dev; /* NULL is wildcarded here */
2181 int (*func) (struct sk_buff *,
2182 struct net_device *,
2183 struct packet_type *,
2184 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002185 bool (*id_match)(struct packet_type *ptype,
2186 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002187 void *af_packet_priv;
2188 struct list_head list;
2189};
2190
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002191struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2193 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002195 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002196 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002197};
2198
2199struct packet_offload {
2200 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002201 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002202 struct offload_callbacks callbacks;
2203 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204};
2205
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002206/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002207struct pcpu_sw_netstats {
2208 u64 rx_packets;
2209 u64 rx_bytes;
2210 u64 tx_packets;
2211 u64 tx_bytes;
2212 struct u64_stats_sync syncp;
2213};
2214
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002215#define __netdev_alloc_pcpu_stats(type, gfp) \
2216({ \
2217 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2218 if (pcpu_stats) { \
2219 int __cpu; \
2220 for_each_possible_cpu(__cpu) { \
2221 typeof(type) *stat; \
2222 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2223 u64_stats_init(&stat->syncp); \
2224 } \
2225 } \
2226 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002227})
2228
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002229#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002230 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002231
Jiri Pirko764f5e52015-12-03 12:12:12 +01002232enum netdev_lag_tx_type {
2233 NETDEV_LAG_TX_TYPE_UNKNOWN,
2234 NETDEV_LAG_TX_TYPE_RANDOM,
2235 NETDEV_LAG_TX_TYPE_BROADCAST,
2236 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2237 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2238 NETDEV_LAG_TX_TYPE_HASH,
2239};
2240
2241struct netdev_lag_upper_info {
2242 enum netdev_lag_tx_type tx_type;
2243};
2244
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002245struct netdev_lag_lower_state_info {
2246 u8 link_up : 1,
2247 tx_enabled : 1;
2248};
2249
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250#include <linux/notifier.h>
2251
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002252/* netdevice notifier chain. Please remember to update the rtnetlink
2253 * notification exclusion list in rtnetlink_event() when adding new
2254 * types.
2255 */
2256#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2257#define NETDEV_DOWN 0x0002
2258#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2259 detected a hardware crash and restarted
2260 - we can use this eg to kick tcp sessions
2261 once done */
2262#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2263#define NETDEV_REGISTER 0x0005
2264#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002265#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002266#define NETDEV_CHANGEADDR 0x0008
2267#define NETDEV_GOING_DOWN 0x0009
2268#define NETDEV_CHANGENAME 0x000A
2269#define NETDEV_FEAT_CHANGE 0x000B
2270#define NETDEV_BONDING_FAILOVER 0x000C
2271#define NETDEV_PRE_UP 0x000D
2272#define NETDEV_PRE_TYPE_CHANGE 0x000E
2273#define NETDEV_POST_TYPE_CHANGE 0x000F
2274#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002275#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002276#define NETDEV_RELEASE 0x0012
2277#define NETDEV_NOTIFY_PEERS 0x0013
2278#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002279#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002280#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002281#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002282#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002283#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002284#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002285#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002286#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Jason Wang08294a22016-06-30 14:45:35 +08002287#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002288
Joe Perchesf629d202013-09-26 14:48:15 -07002289int register_netdevice_notifier(struct notifier_block *nb);
2290int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002291
2292struct netdev_notifier_info {
2293 struct net_device *dev;
2294};
2295
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002296struct netdev_notifier_change_info {
2297 struct netdev_notifier_info info; /* must be first */
2298 unsigned int flags_changed;
2299};
2300
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002301struct netdev_notifier_changeupper_info {
2302 struct netdev_notifier_info info; /* must be first */
2303 struct net_device *upper_dev; /* new upper dev */
2304 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002305 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002306 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002307};
2308
Jiri Pirko04d48262015-12-03 12:12:15 +01002309struct netdev_notifier_changelowerstate_info {
2310 struct netdev_notifier_info info; /* must be first */
2311 void *lower_state_info; /* is lower dev state */
2312};
2313
Cong Wang75538c22013-05-29 11:30:50 +08002314static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2315 struct net_device *dev)
2316{
2317 info->dev = dev;
2318}
2319
Jiri Pirko351638e2013-05-28 01:30:21 +00002320static inline struct net_device *
2321netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2322{
2323 return info->dev;
2324}
2325
Joe Perchesf629d202013-09-26 14:48:15 -07002326int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002327
2328
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329extern rwlock_t dev_base_lock; /* Device list lock */
2330
Eric W. Biederman881d9662007-09-17 11:56:21 -07002331#define for_each_netdev(net, d) \
2332 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002333#define for_each_netdev_reverse(net, d) \
2334 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002335#define for_each_netdev_rcu(net, d) \
2336 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002337#define for_each_netdev_safe(net, d, n) \
2338 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2339#define for_each_netdev_continue(net, d) \
2340 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002341#define for_each_netdev_continue_rcu(net, d) \
2342 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002343#define for_each_netdev_in_bond_rcu(bond, slave) \
2344 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002345 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002346#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2347
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002348static inline struct net_device *next_net_device(struct net_device *dev)
2349{
2350 struct list_head *lh;
2351 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002352
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002353 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002354 lh = dev->dev_list.next;
2355 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2356}
2357
Eric Dumazetce81b762009-11-11 17:34:30 +00002358static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2359{
2360 struct list_head *lh;
2361 struct net *net;
2362
2363 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002364 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002365 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2366}
2367
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002368static inline struct net_device *first_net_device(struct net *net)
2369{
2370 return list_empty(&net->dev_base_head) ? NULL :
2371 net_device_entry(net->dev_base_head.next);
2372}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002373
Eric Dumazetccf43432011-01-26 18:08:02 +00002374static inline struct net_device *first_net_device_rcu(struct net *net)
2375{
2376 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2377
2378 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2379}
2380
Joe Perchesf629d202013-09-26 14:48:15 -07002381int netdev_boot_setup_check(struct net_device *dev);
2382unsigned long netdev_boot_base(const char *prefix, int unit);
2383struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2384 const char *hwaddr);
2385struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2386struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2387void dev_add_pack(struct packet_type *pt);
2388void dev_remove_pack(struct packet_type *pt);
2389void __dev_remove_pack(struct packet_type *pt);
2390void dev_add_offload(struct packet_offload *po);
2391void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002392
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002393int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002394int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002395struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2396 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002397struct net_device *dev_get_by_name(struct net *net, const char *name);
2398struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2399struct net_device *__dev_get_by_name(struct net *net, const char *name);
2400int dev_alloc_name(struct net_device *dev, const char *name);
2401int dev_open(struct net_device *dev);
2402int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002403int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002404void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002405int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002406int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002407int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002408int register_netdevice(struct net_device *dev);
2409void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2410void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002411static inline void unregister_netdevice(struct net_device *dev)
2412{
2413 unregister_netdevice_queue(dev, NULL);
2414}
2415
Joe Perchesf629d202013-09-26 14:48:15 -07002416int netdev_refcnt_read(const struct net_device *dev);
2417void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002418void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002419void synchronize_net(void);
2420int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002421
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002422DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002423#define XMIT_RECURSION_LIMIT 10
2424
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002425static inline int dev_recursion_level(void)
2426{
2427 return this_cpu_read(xmit_recursion);
2428}
2429
Joe Perchesf629d202013-09-26 14:48:15 -07002430struct net_device *dev_get_by_index(struct net *net, int ifindex);
2431struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2432struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2433int netdev_get_name(struct net *net, char *name, int ifindex);
2434int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002435int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002436
2437static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2438{
2439 return NAPI_GRO_CB(skb)->data_offset;
2440}
2441
2442static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2443{
2444 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2445}
2446
2447static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2448{
2449 NAPI_GRO_CB(skb)->data_offset += len;
2450}
2451
Herbert Xua5b1cf22009-05-26 18:50:28 +00002452static inline void *skb_gro_header_fast(struct sk_buff *skb,
2453 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002454{
Herbert Xu78a478d2009-05-26 18:50:21 +00002455 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002456}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002457
Herbert Xua5b1cf22009-05-26 18:50:28 +00002458static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2459{
2460 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2461}
2462
Herbert Xu57ea52a2017-01-10 12:24:15 -08002463static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2464{
2465 NAPI_GRO_CB(skb)->frag0 = NULL;
2466 NAPI_GRO_CB(skb)->frag0_len = 0;
2467}
2468
Herbert Xua5b1cf22009-05-26 18:50:28 +00002469static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2470 unsigned int offset)
2471{
Herbert Xu17dd7592011-07-27 06:16:28 -07002472 if (!pskb_may_pull(skb, hlen))
2473 return NULL;
2474
Herbert Xu57ea52a2017-01-10 12:24:15 -08002475 skb_gro_frag0_invalidate(skb);
Herbert Xu17dd7592011-07-27 06:16:28 -07002476 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002477}
2478
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002479static inline void *skb_gro_network_header(struct sk_buff *skb)
2480{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002481 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2482 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002483}
2484
Jerry Chubf5a7552014-01-07 10:23:19 -08002485static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2486 const void *start, unsigned int len)
2487{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002488 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002489 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2490 csum_partial(start, len, 0));
2491}
2492
Tom Herbert573e8fc2014-08-22 13:33:47 -07002493/* GRO checksum functions. These are logical equivalents of the normal
2494 * checksum functions (in skbuff.h) except that they operate on the GRO
2495 * offsets and fields in sk_buff.
2496 */
2497
2498__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2499
Tom Herbert15e23962015-02-10 16:30:31 -08002500static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2501{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002502 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002503}
2504
Tom Herbert573e8fc2014-08-22 13:33:47 -07002505static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2506 bool zero_okay,
2507 __sum16 check)
2508{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002509 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2510 skb_checksum_start_offset(skb) <
2511 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002512 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002513 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002514 (!zero_okay || check));
2515}
2516
2517static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2518 __wsum psum)
2519{
2520 if (NAPI_GRO_CB(skb)->csum_valid &&
2521 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2522 return 0;
2523
2524 NAPI_GRO_CB(skb)->csum = psum;
2525
2526 return __skb_gro_checksum_complete(skb);
2527}
2528
Tom Herbert573e8fc2014-08-22 13:33:47 -07002529static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2530{
Tom Herbert662880f2014-08-27 21:26:56 -07002531 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2532 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2533 NAPI_GRO_CB(skb)->csum_cnt--;
2534 } else {
2535 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2536 * verified a new top level checksum or an encapsulated one
2537 * during GRO. This saves work if we fallback to normal path.
2538 */
2539 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002540 }
2541}
2542
2543#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2544 compute_pseudo) \
2545({ \
2546 __sum16 __ret = 0; \
2547 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2548 __ret = __skb_gro_checksum_validate_complete(skb, \
2549 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002550 if (__ret) \
2551 __skb_mark_checksum_bad(skb); \
2552 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002553 skb_gro_incr_csum_unnecessary(skb); \
2554 __ret; \
2555})
2556
2557#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2558 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2559
2560#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2561 compute_pseudo) \
2562 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2563
2564#define skb_gro_checksum_simple_validate(skb) \
2565 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2566
Tom Herbertd96535a2014-08-31 15:12:42 -07002567static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2568{
2569 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2570 !NAPI_GRO_CB(skb)->csum_valid);
2571}
2572
2573static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2574 __sum16 check, __wsum pseudo)
2575{
2576 NAPI_GRO_CB(skb)->csum = ~pseudo;
2577 NAPI_GRO_CB(skb)->csum_valid = 1;
2578}
2579
2580#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2581do { \
2582 if (__skb_gro_checksum_convert_check(skb)) \
2583 __skb_gro_checksum_convert(skb, check, \
2584 compute_pseudo(skb, proto)); \
2585} while (0)
2586
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002587struct gro_remcsum {
2588 int offset;
2589 __wsum delta;
2590};
2591
2592static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2593{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002594 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002595 grc->delta = 0;
2596}
2597
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002598static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2599 unsigned int off, size_t hdrlen,
2600 int start, int offset,
2601 struct gro_remcsum *grc,
2602 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002603{
2604 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002605 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002606
2607 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2608
Tom Herbert15e23962015-02-10 16:30:31 -08002609 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002610 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2611 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002612 }
2613
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002614 ptr = skb_gro_header_fast(skb, off);
2615 if (skb_gro_header_hard(skb, off + plen)) {
2616 ptr = skb_gro_header_slow(skb, off + plen, off);
2617 if (!ptr)
2618 return NULL;
2619 }
2620
2621 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2622 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002623
2624 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002625 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002626
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002627 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002628 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002629
2630 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002631}
2632
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002633static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2634 struct gro_remcsum *grc)
2635{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002636 void *ptr;
2637 size_t plen = grc->offset + sizeof(u16);
2638
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002639 if (!grc->delta)
2640 return;
2641
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002642 ptr = skb_gro_header_fast(skb, grc->offset);
2643 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2644 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2645 if (!ptr)
2646 return;
2647 }
2648
2649 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002650}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002651
Steffen Klassert25393d32017-02-15 09:39:44 +01002652#ifdef CONFIG_XFRM_OFFLOAD
2653static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2654{
2655 if (PTR_ERR(pp) != -EINPROGRESS)
2656 NAPI_GRO_CB(skb)->flush |= flush;
2657}
2658#else
Steffen Klassert5f114162017-02-15 09:39:39 +01002659static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2660{
2661 NAPI_GRO_CB(skb)->flush |= flush;
2662}
Steffen Klassert25393d32017-02-15 09:39:44 +01002663#endif
Steffen Klassert5f114162017-02-15 09:39:39 +01002664
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002665static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2666 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002667 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002668 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002669{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002670 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002671 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002672
2673 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002674}
2675
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002676static inline int dev_parse_header(const struct sk_buff *skb,
2677 unsigned char *haddr)
2678{
2679 const struct net_device *dev = skb->dev;
2680
Patrick McHardy1b833362007-10-18 05:09:28 -07002681 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002682 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002683 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002684}
2685
Willem de Bruijn2793a232016-03-09 21:58:32 -05002686/* ll_header must have at least hard_header_len allocated */
2687static inline bool dev_validate_header(const struct net_device *dev,
2688 char *ll_header, int len)
2689{
2690 if (likely(len >= dev->hard_header_len))
2691 return true;
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05002692 if (len < dev->min_header_len)
2693 return false;
Willem de Bruijn2793a232016-03-09 21:58:32 -05002694
2695 if (capable(CAP_SYS_RAWIO)) {
2696 memset(ll_header + len, 0, dev->hard_header_len - len);
2697 return true;
2698 }
2699
2700 if (dev->header_ops && dev->header_ops->validate)
2701 return dev->header_ops->validate(ll_header, len);
2702
2703 return false;
2704}
2705
Linus Torvalds1da177e2005-04-16 15:20:36 -07002706typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002707int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002708static inline int unregister_gifconf(unsigned int family)
2709{
2710 return register_gifconf(family, NULL);
2711}
2712
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002713#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002714#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002715struct sd_flow_limit {
2716 u64 count;
2717 unsigned int num_buckets;
2718 unsigned int history_head;
2719 u16 history[FLOW_LIMIT_HISTORY];
2720 u8 buckets[];
2721};
2722
2723extern int netdev_flow_limit_table_len;
2724#endif /* CONFIG_NET_FLOW_LIMIT */
2725
Linus Torvalds1da177e2005-04-16 15:20:36 -07002726/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002727 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002728 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002729struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002730 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002731 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002732
Changli Gaodee42872010-05-02 05:42:16 +00002733 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002734 unsigned int processed;
2735 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002736 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002737#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002738 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002739#endif
2740#ifdef CONFIG_NET_FLOW_LIMIT
2741 struct sd_flow_limit __rcu *flow_limit;
2742#endif
2743 struct Qdisc *output_queue;
2744 struct Qdisc **output_queue_tailp;
2745 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002746
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002747#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002748 /* input_queue_head should be written by cpu owning this struct,
2749 * and only read by other cpus. Worth using a cache line.
2750 */
2751 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2752
2753 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002754 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002755 struct softnet_data *rps_ipi_next;
2756 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002757 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002758#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002759 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002760 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002761 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002762
Linus Torvalds1da177e2005-04-16 15:20:36 -07002763};
2764
Tom Herbert76cc8b12010-05-20 18:37:59 +00002765static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002766{
2767#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002768 sd->input_queue_head++;
2769#endif
2770}
2771
2772static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2773 unsigned int *qtail)
2774{
2775#ifdef CONFIG_RPS
2776 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002777#endif
2778}
2779
Tom Herbert0a9627f2010-03-16 08:03:29 +00002780DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002781
Joe Perchesf629d202013-09-26 14:48:15 -07002782void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002783void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002784
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002785static inline void netif_tx_schedule_all(struct net_device *dev)
2786{
2787 unsigned int i;
2788
2789 for (i = 0; i < dev->num_tx_queues; i++)
2790 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2791}
2792
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002793static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002794{
Tom Herbert734664982011-11-28 16:32:44 +00002795 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002796}
2797
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002798/**
2799 * netif_start_queue - allow transmit
2800 * @dev: network device
2801 *
2802 * Allow upper layers to call the device hard_start_xmit routine.
2803 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804static inline void netif_start_queue(struct net_device *dev)
2805{
David S. Millere8a04642008-07-17 00:34:19 -07002806 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002807}
2808
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002809static inline void netif_tx_start_all_queues(struct net_device *dev)
2810{
2811 unsigned int i;
2812
2813 for (i = 0; i < dev->num_tx_queues; i++) {
2814 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2815 netif_tx_start_queue(txq);
2816 }
2817}
2818
John Fastabend46e5da40a2014-09-12 20:04:52 -07002819void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002820
Dave Jonesd29f7492008-07-22 14:09:06 -07002821/**
2822 * netif_wake_queue - restart transmit
2823 * @dev: network device
2824 *
2825 * Allow upper layers to call the device hard_start_xmit routine.
2826 * Used for flow control when transmit resources are available.
2827 */
David S. Miller79d16382008-07-08 23:14:46 -07002828static inline void netif_wake_queue(struct net_device *dev)
2829{
David S. Millere8a04642008-07-17 00:34:19 -07002830 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002831}
2832
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002833static inline void netif_tx_wake_all_queues(struct net_device *dev)
2834{
2835 unsigned int i;
2836
2837 for (i = 0; i < dev->num_tx_queues; i++) {
2838 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2839 netif_tx_wake_queue(txq);
2840 }
2841}
2842
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002843static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002844{
Tom Herbert734664982011-11-28 16:32:44 +00002845 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002846}
2847
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002848/**
2849 * netif_stop_queue - stop transmitted packets
2850 * @dev: network device
2851 *
2852 * Stop upper layers calling the device hard_start_xmit routine.
2853 * Used for flow control when transmit resources are unavailable.
2854 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002855static inline void netif_stop_queue(struct net_device *dev)
2856{
David S. Millere8a04642008-07-17 00:34:19 -07002857 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002858}
2859
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002860void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002861
David S. Miller4d295152012-03-07 21:02:35 -05002862static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002863{
Tom Herbert734664982011-11-28 16:32:44 +00002864 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002865}
2866
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002867/**
2868 * netif_queue_stopped - test if transmit queue is flowblocked
2869 * @dev: network device
2870 *
2871 * Test if transmit queue on device is currently unable to send.
2872 */
David S. Miller4d295152012-03-07 21:02:35 -05002873static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002874{
David S. Millere8a04642008-07-17 00:34:19 -07002875 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002876}
2877
David S. Miller4d295152012-03-07 21:02:35 -05002878static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002879{
Tom Herbert734664982011-11-28 16:32:44 +00002880 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2881}
2882
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002883static inline bool
2884netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002885{
2886 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2887}
2888
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002889static inline bool
2890netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2891{
2892 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2893}
2894
Eric Dumazet53511452014-10-08 08:19:27 -07002895/**
2896 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2897 * @dev_queue: pointer to transmit queue
2898 *
2899 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002900 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002901 */
2902static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2903{
2904#ifdef CONFIG_BQL
2905 prefetchw(&dev_queue->dql.num_queued);
2906#endif
2907}
2908
2909/**
2910 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2911 * @dev_queue: pointer to transmit queue
2912 *
2913 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002914 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002915 */
2916static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2917{
2918#ifdef CONFIG_BQL
2919 prefetchw(&dev_queue->dql.limit);
2920#endif
2921}
2922
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002923static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2924 unsigned int bytes)
2925{
Tom Herbert114cf582011-11-28 16:33:09 +00002926#ifdef CONFIG_BQL
2927 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002928
2929 if (likely(dql_avail(&dev_queue->dql) >= 0))
2930 return;
2931
2932 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2933
2934 /*
2935 * The XOFF flag must be set before checking the dql_avail below,
2936 * because in netdev_tx_completed_queue we update the dql_completed
2937 * before checking the XOFF flag.
2938 */
2939 smp_mb();
2940
2941 /* check again in case another CPU has just made room avail */
2942 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2943 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002944#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002945}
2946
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002947/**
2948 * netdev_sent_queue - report the number of bytes queued to hardware
2949 * @dev: network device
2950 * @bytes: number of bytes queued to the hardware device queue
2951 *
2952 * Report the number of bytes queued for sending/completion to the network
2953 * device hardware queue. @bytes should be a good approximation and should
2954 * exactly match netdev_completed_queue() @bytes
2955 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002956static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2957{
2958 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2959}
2960
2961static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002962 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002963{
Tom Herbert114cf582011-11-28 16:33:09 +00002964#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002965 if (unlikely(!bytes))
2966 return;
2967
2968 dql_completed(&dev_queue->dql, bytes);
2969
2970 /*
2971 * Without the memory barrier there is a small possiblity that
2972 * netdev_tx_sent_queue will miss the update and cause the queue to
2973 * be stopped forever
2974 */
2975 smp_mb();
2976
2977 if (dql_avail(&dev_queue->dql) < 0)
2978 return;
2979
2980 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2981 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00002982#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002983}
2984
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002985/**
2986 * netdev_completed_queue - report bytes and packets completed by device
2987 * @dev: network device
2988 * @pkts: actual number of packets sent over the medium
2989 * @bytes: actual number of bytes sent over the medium
2990 *
2991 * Report the number of bytes and packets transmitted by the network device
2992 * hardware queue over the physical medium, @bytes must exactly match the
2993 * @bytes amount passed to netdev_sent_queue()
2994 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002995static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00002996 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002997{
2998 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2999}
3000
3001static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3002{
Tom Herbert114cf582011-11-28 16:33:09 +00003003#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003004 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003005 dql_reset(&q->dql);
3006#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003007}
3008
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003009/**
3010 * netdev_reset_queue - reset the packets and bytes count of a network device
3011 * @dev_queue: network device
3012 *
3013 * Reset the bytes and packet count of a network device and clear the
3014 * software flow control OFF bit for this network device
3015 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003016static inline void netdev_reset_queue(struct net_device *dev_queue)
3017{
3018 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003019}
3020
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003021/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003022 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3023 * @dev: network device
3024 * @queue_index: given tx queue index
3025 *
3026 * Returns 0 if given tx queue index >= number of device tx queues,
3027 * otherwise returns the originally passed tx queue index.
3028 */
3029static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3030{
3031 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3032 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3033 dev->name, queue_index,
3034 dev->real_num_tx_queues);
3035 return 0;
3036 }
3037
3038 return queue_index;
3039}
3040
3041/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003042 * netif_running - test if up
3043 * @dev: network device
3044 *
3045 * Test if the device has been brought up.
3046 */
David S. Miller4d295152012-03-07 21:02:35 -05003047static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003048{
3049 return test_bit(__LINK_STATE_START, &dev->state);
3050}
3051
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003052/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003053 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003054 * stop, and a check if it's stopped. All other device management is
3055 * done at the overall netdevice level.
3056 * Also test the device if we're multiqueue.
3057 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003058
3059/**
3060 * netif_start_subqueue - allow sending packets on subqueue
3061 * @dev: network device
3062 * @queue_index: sub queue index
3063 *
3064 * Start individual transmit queue of a device with multiple transmit queues.
3065 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003066static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3067{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003068 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003069
3070 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003071}
3072
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003073/**
3074 * netif_stop_subqueue - stop sending packets on subqueue
3075 * @dev: network device
3076 * @queue_index: sub queue index
3077 *
3078 * Stop individual transmit queue of a device with multiple transmit queues.
3079 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003080static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3081{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003082 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003083 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003084}
3085
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003086/**
3087 * netif_subqueue_stopped - test status of subqueue
3088 * @dev: network device
3089 * @queue_index: sub queue index
3090 *
3091 * Check individual transmit queue of a device with multiple transmit queues.
3092 */
David S. Miller4d295152012-03-07 21:02:35 -05003093static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3094 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003095{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003096 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003097
3098 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003099}
3100
David S. Miller4d295152012-03-07 21:02:35 -05003101static inline bool netif_subqueue_stopped(const struct net_device *dev,
3102 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003103{
3104 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3105}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003106
Florian Fainelli738b35c2017-01-11 21:13:02 -08003107/**
3108 * netif_wake_subqueue - allow sending packets on subqueue
3109 * @dev: network device
3110 * @queue_index: sub queue index
3111 *
3112 * Resume individual transmit queue of a device with multiple transmit queues.
3113 */
3114static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3115{
3116 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3117
3118 netif_tx_wake_queue(txq);
3119}
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003120
Alexander Duyck537c00d2013-01-10 08:57:02 +00003121#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003122int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003123 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003124#else
3125static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003126 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003127 u16 index)
3128{
3129 return 0;
3130}
3131#endif
3132
Jiri Pirko5605c762015-05-12 14:56:12 +02003133u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3134 unsigned int num_tx_queues);
3135
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003136/*
3137 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3138 * as a distribution range limit for the returned value.
3139 */
3140static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003141 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003142{
3143 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3144}
3145
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003146/**
3147 * netif_is_multiqueue - test if device has multiple transmit queues
3148 * @dev: network device
3149 *
3150 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003151 */
David S. Miller4d295152012-03-07 21:02:35 -05003152static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003153{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003154 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003155}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003156
Joe Perchesf629d202013-09-26 14:48:15 -07003157int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003158
Michael Daltona953be52014-01-16 22:23:28 -08003159#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003160int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003161#else
3162static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3163 unsigned int rxq)
3164{
3165 return 0;
3166}
3167#endif
3168
Michael Daltona953be52014-01-16 22:23:28 -08003169#ifdef CONFIG_SYSFS
3170static inline unsigned int get_netdev_rx_queue_index(
3171 struct netdev_rx_queue *queue)
3172{
3173 struct net_device *dev = queue->dev;
3174 int index = queue - dev->_rx;
3175
3176 BUG_ON(index >= dev->num_rx_queues);
3177 return index;
3178}
3179#endif
3180
Yuval Mintz16917b82012-07-01 03:18:50 +00003181#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003182int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003183
Eric Dumazete6247022013-12-05 04:45:08 -08003184enum skb_free_reason {
3185 SKB_REASON_CONSUMED,
3186 SKB_REASON_DROPPED,
3187};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003188
Eric Dumazete6247022013-12-05 04:45:08 -08003189void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3190void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3191
3192/*
3193 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3194 * interrupt context or with hardware interrupts being disabled.
3195 * (in_irq() || irqs_disabled())
3196 *
3197 * We provide four helpers that can be used in following contexts :
3198 *
3199 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3200 * replacing kfree_skb(skb)
3201 *
3202 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3203 * Typically used in place of consume_skb(skb) in TX completion path
3204 *
3205 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3206 * replacing kfree_skb(skb)
3207 *
3208 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3209 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003210 */
Eric Dumazete6247022013-12-05 04:45:08 -08003211static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3212{
3213 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3214}
3215
3216static inline void dev_consume_skb_irq(struct sk_buff *skb)
3217{
3218 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3219}
3220
3221static inline void dev_kfree_skb_any(struct sk_buff *skb)
3222{
3223 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3224}
3225
3226static inline void dev_consume_skb_any(struct sk_buff *skb)
3227{
3228 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3229}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003230
Joe Perchesf629d202013-09-26 14:48:15 -07003231int netif_rx(struct sk_buff *skb);
3232int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003233int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003234gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3235void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3236struct sk_buff *napi_get_frags(struct napi_struct *napi);
3237gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003238struct packet_offload *gro_find_receive_by_type(__be16 type);
3239struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003240
3241static inline void napi_free_frags(struct napi_struct *napi)
3242{
3243 kfree_skb(napi->skb);
3244 napi->skb = NULL;
3245}
3246
Mahesh Bandewar24b27fc2016-09-01 22:18:34 -07003247bool netdev_is_rx_handler_busy(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003248int netdev_rx_handler_register(struct net_device *dev,
3249 rx_handler_func_t *rx_handler,
3250 void *rx_handler_data);
3251void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003252
Joe Perchesf629d202013-09-26 14:48:15 -07003253bool dev_valid_name(const char *name);
3254int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3255int dev_ethtool(struct net *net, struct ifreq *);
3256unsigned int dev_get_flags(const struct net_device *);
3257int __dev_change_flags(struct net_device *, unsigned int flags);
3258int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003259void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3260 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003261int dev_change_name(struct net_device *, const char *);
3262int dev_set_alias(struct net_device *, const char *, size_t);
3263int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3264int dev_set_mtu(struct net_device *, int);
3265void dev_set_group(struct net_device *, int);
3266int dev_set_mac_address(struct net_device *, struct sockaddr *);
3267int dev_change_carrier(struct net_device *, bool new_carrier);
3268int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003269 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003270int dev_get_phys_port_name(struct net_device *dev,
3271 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003272int dev_change_proto_down(struct net_device *dev, bool proto_down);
Daniel Borkmann85de8572016-11-28 23:16:54 +01003273int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003274struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003275struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3276 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08003277int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003278int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003279bool is_skb_forwardable(const struct net_device *dev,
3280 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003281
Martin KaFai Lau4e3264d2016-11-09 15:36:33 -08003282static __always_inline int ____dev_forward_skb(struct net_device *dev,
3283 struct sk_buff *skb)
3284{
3285 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3286 unlikely(!is_skb_forwardable(dev, skb))) {
3287 atomic_long_inc(&dev->rx_dropped);
3288 kfree_skb(skb);
3289 return NET_RX_DROP;
3290 }
3291
3292 skb_scrub_packet(skb, true);
3293 skb->priority = 0;
3294 return 0;
3295}
3296
David Ahern74b20582016-05-10 11:19:50 -07003297void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3298
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003299extern int netdev_budget;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003300
3301/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003302void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003303
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003304/**
3305 * dev_put - release reference to device
3306 * @dev: network device
3307 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003308 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003309 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003310static inline void dev_put(struct net_device *dev)
3311{
Christoph Lameter933393f2011-12-22 11:58:51 -06003312 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003313}
3314
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003315/**
3316 * dev_hold - get reference to device
3317 * @dev: network device
3318 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003319 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003320 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003321static inline void dev_hold(struct net_device *dev)
3322{
Christoph Lameter933393f2011-12-22 11:58:51 -06003323 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003324}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003325
3326/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3327 * and _off may be called from IRQ context, but it is caller
3328 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003329 *
3330 * The name carrier is inappropriate, these functions should really be
3331 * called netif_lowerlayer_*() because they represent the state of any
3332 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003333 */
3334
Joe Perchesf629d202013-09-26 14:48:15 -07003335void linkwatch_init_dev(struct net_device *dev);
3336void linkwatch_fire_event(struct net_device *dev);
3337void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003338
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003339/**
3340 * netif_carrier_ok - test if carrier present
3341 * @dev: network device
3342 *
3343 * Check if carrier is present on device
3344 */
David S. Miller4d295152012-03-07 21:02:35 -05003345static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003346{
3347 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3348}
3349
Joe Perchesf629d202013-09-26 14:48:15 -07003350unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003351
Joe Perchesf629d202013-09-26 14:48:15 -07003352void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003353
Joe Perchesf629d202013-09-26 14:48:15 -07003354void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003355
Joe Perchesf629d202013-09-26 14:48:15 -07003356void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003357
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003358/**
3359 * netif_dormant_on - mark device as dormant.
3360 * @dev: network device
3361 *
3362 * Mark device as dormant (as per RFC2863).
3363 *
3364 * The dormant state indicates that the relevant interface is not
3365 * actually in a condition to pass packets (i.e., it is not 'up') but is
3366 * in a "pending" state, waiting for some external event. For "on-
3367 * demand" interfaces, this new state identifies the situation where the
3368 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003369 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003370static inline void netif_dormant_on(struct net_device *dev)
3371{
3372 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3373 linkwatch_fire_event(dev);
3374}
3375
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003376/**
3377 * netif_dormant_off - set device as not dormant.
3378 * @dev: network device
3379 *
3380 * Device is not in dormant state.
3381 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003382static inline void netif_dormant_off(struct net_device *dev)
3383{
3384 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3385 linkwatch_fire_event(dev);
3386}
3387
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003388/**
3389 * netif_dormant - test if carrier present
3390 * @dev: network device
3391 *
3392 * Check if carrier is present on device
3393 */
David S. Miller4d295152012-03-07 21:02:35 -05003394static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003395{
3396 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3397}
3398
3399
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003400/**
3401 * netif_oper_up - test if device is operational
3402 * @dev: network device
3403 *
3404 * Check if carrier is operational
3405 */
David S. Miller4d295152012-03-07 21:02:35 -05003406static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003407{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003408 return (dev->operstate == IF_OPER_UP ||
3409 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3410}
3411
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003412/**
3413 * netif_device_present - is device available or removed
3414 * @dev: network device
3415 *
3416 * Check if device has not been removed from system.
3417 */
David S. Miller4d295152012-03-07 21:02:35 -05003418static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003419{
3420 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3421}
3422
Joe Perchesf629d202013-09-26 14:48:15 -07003423void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003424
Joe Perchesf629d202013-09-26 14:48:15 -07003425void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003426
3427/*
3428 * Network interface message level settings
3429 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003430
3431enum {
3432 NETIF_MSG_DRV = 0x0001,
3433 NETIF_MSG_PROBE = 0x0002,
3434 NETIF_MSG_LINK = 0x0004,
3435 NETIF_MSG_TIMER = 0x0008,
3436 NETIF_MSG_IFDOWN = 0x0010,
3437 NETIF_MSG_IFUP = 0x0020,
3438 NETIF_MSG_RX_ERR = 0x0040,
3439 NETIF_MSG_TX_ERR = 0x0080,
3440 NETIF_MSG_TX_QUEUED = 0x0100,
3441 NETIF_MSG_INTR = 0x0200,
3442 NETIF_MSG_TX_DONE = 0x0400,
3443 NETIF_MSG_RX_STATUS = 0x0800,
3444 NETIF_MSG_PKTDATA = 0x1000,
3445 NETIF_MSG_HW = 0x2000,
3446 NETIF_MSG_WOL = 0x4000,
3447};
3448
3449#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3450#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3451#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3452#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3453#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3454#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3455#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3456#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3457#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3458#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3459#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3460#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3461#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3462#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3463#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3464
3465static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3466{
3467 /* use default */
3468 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3469 return default_msg_enable_bits;
3470 if (debug_value == 0) /* no output */
3471 return 0;
3472 /* set low N bits */
3473 return (1 << debug_value) - 1;
3474}
3475
David S. Millerc773e842008-07-08 23:13:53 -07003476static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003477{
David S. Millerc773e842008-07-08 23:13:53 -07003478 spin_lock(&txq->_xmit_lock);
3479 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003480}
3481
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003482static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3483{
3484 __acquire(&txq->_xmit_lock);
3485 return true;
3486}
3487
3488static inline void __netif_tx_release(struct netdev_queue *txq)
3489{
3490 __release(&txq->_xmit_lock);
3491}
3492
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003493static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3494{
3495 spin_lock_bh(&txq->_xmit_lock);
3496 txq->xmit_lock_owner = smp_processor_id();
3497}
3498
David S. Miller4d295152012-03-07 21:02:35 -05003499static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003500{
David S. Miller4d295152012-03-07 21:02:35 -05003501 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003502 if (likely(ok))
3503 txq->xmit_lock_owner = smp_processor_id();
3504 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003505}
3506
David S. Millerc773e842008-07-08 23:13:53 -07003507static inline void __netif_tx_unlock(struct netdev_queue *txq)
3508{
3509 txq->xmit_lock_owner = -1;
3510 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003511}
3512
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003513static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3514{
3515 txq->xmit_lock_owner = -1;
3516 spin_unlock_bh(&txq->_xmit_lock);
3517}
3518
Eric Dumazet08baf562009-05-25 22:58:01 -07003519static inline void txq_trans_update(struct netdev_queue *txq)
3520{
3521 if (txq->xmit_lock_owner != -1)
3522 txq->trans_start = jiffies;
3523}
3524
Florian Westphalba162f82016-05-03 16:31:00 +02003525/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3526static inline void netif_trans_update(struct net_device *dev)
3527{
Florian Westphal9b366272016-05-03 16:33:14 +02003528 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3529
3530 if (txq->trans_start != jiffies)
3531 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003532}
3533
David S. Millerc3f26a22008-07-31 16:58:50 -07003534/**
3535 * netif_tx_lock - grab network device transmit lock
3536 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003537 *
3538 * Get network device transmit lock
3539 */
3540static inline void netif_tx_lock(struct net_device *dev)
3541{
3542 unsigned int i;
3543 int cpu;
3544
3545 spin_lock(&dev->tx_global_lock);
3546 cpu = smp_processor_id();
3547 for (i = 0; i < dev->num_tx_queues; i++) {
3548 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3549
3550 /* We are the only thread of execution doing a
3551 * freeze, but we have to grab the _xmit_lock in
3552 * order to synchronize with threads which are in
3553 * the ->hard_start_xmit() handler and already
3554 * checked the frozen bit.
3555 */
3556 __netif_tx_lock(txq, cpu);
3557 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3558 __netif_tx_unlock(txq);
3559 }
3560}
3561
3562static inline void netif_tx_lock_bh(struct net_device *dev)
3563{
3564 local_bh_disable();
3565 netif_tx_lock(dev);
3566}
3567
Herbert Xu932ff272006-06-09 12:20:56 -07003568static inline void netif_tx_unlock(struct net_device *dev)
3569{
David S. Millere8a04642008-07-17 00:34:19 -07003570 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003571
David S. Millere8a04642008-07-17 00:34:19 -07003572 for (i = 0; i < dev->num_tx_queues; i++) {
3573 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003574
David S. Millerc3f26a22008-07-31 16:58:50 -07003575 /* No need to grab the _xmit_lock here. If the
3576 * queue is not stopped for another reason, we
3577 * force a schedule.
3578 */
3579 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003580 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003581 }
3582 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003583}
3584
3585static inline void netif_tx_unlock_bh(struct net_device *dev)
3586{
David S. Millere8a04642008-07-17 00:34:19 -07003587 netif_tx_unlock(dev);
3588 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003589}
3590
David S. Millerc773e842008-07-08 23:13:53 -07003591#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003592 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003593 __netif_tx_lock(txq, cpu); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003594 } else { \
3595 __netif_tx_acquire(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003596 } \
3597}
3598
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003599#define HARD_TX_TRYLOCK(dev, txq) \
3600 (((dev->features & NETIF_F_LLTX) == 0) ? \
3601 __netif_tx_trylock(txq) : \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003602 __netif_tx_acquire(txq))
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003603
David S. Millerc773e842008-07-08 23:13:53 -07003604#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003605 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003606 __netif_tx_unlock(txq); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003607 } else { \
3608 __netif_tx_release(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003609 } \
3610}
3611
Linus Torvalds1da177e2005-04-16 15:20:36 -07003612static inline void netif_tx_disable(struct net_device *dev)
3613{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003614 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003615 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003616
David S. Millerc3f26a22008-07-31 16:58:50 -07003617 local_bh_disable();
3618 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003619 for (i = 0; i < dev->num_tx_queues; i++) {
3620 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003621
3622 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003623 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003624 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003625 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003626 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003627}
3628
David S. Millere308a5d2008-07-15 00:13:44 -07003629static inline void netif_addr_lock(struct net_device *dev)
3630{
3631 spin_lock(&dev->addr_list_lock);
3632}
3633
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003634static inline void netif_addr_lock_nested(struct net_device *dev)
3635{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003636 int subclass = SINGLE_DEPTH_NESTING;
3637
3638 if (dev->netdev_ops->ndo_get_lock_subclass)
3639 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3640
3641 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003642}
3643
David S. Millere308a5d2008-07-15 00:13:44 -07003644static inline void netif_addr_lock_bh(struct net_device *dev)
3645{
3646 spin_lock_bh(&dev->addr_list_lock);
3647}
3648
3649static inline void netif_addr_unlock(struct net_device *dev)
3650{
3651 spin_unlock(&dev->addr_list_lock);
3652}
3653
3654static inline void netif_addr_unlock_bh(struct net_device *dev)
3655{
3656 spin_unlock_bh(&dev->addr_list_lock);
3657}
3658
Jiri Pirkof001fde2009-05-05 02:48:28 +00003659/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003660 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003661 * rcu_read_lock held.
3662 */
3663#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003664 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003665
Linus Torvalds1da177e2005-04-16 15:20:36 -07003666/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3667
Joe Perchesf629d202013-09-26 14:48:15 -07003668void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003669
3670/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003671struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003672 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003673 void (*setup)(struct net_device *),
3674 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003675#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3676 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003677
Tom Gundersenc835a672014-07-14 16:37:24 +02003678#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3679 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3680 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003681
Joe Perchesf629d202013-09-26 14:48:15 -07003682int register_netdev(struct net_device *dev);
3683void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003684
Jiri Pirko22bedad32010-04-01 21:22:57 +00003685/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003686int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3687 struct netdev_hw_addr_list *from_list, int addr_len);
3688void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3689 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003690int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3691 struct net_device *dev,
3692 int (*sync)(struct net_device *, const unsigned char *),
3693 int (*unsync)(struct net_device *,
3694 const unsigned char *));
3695void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3696 struct net_device *dev,
3697 int (*unsync)(struct net_device *,
3698 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003699void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003700
Jiri Pirkof001fde2009-05-05 02:48:28 +00003701/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003702int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3703 unsigned char addr_type);
3704int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3705 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003706void dev_addr_flush(struct net_device *dev);
3707int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003708
3709/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003710int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3711int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3712int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3713int dev_uc_sync(struct net_device *to, struct net_device *from);
3714int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3715void dev_uc_unsync(struct net_device *to, struct net_device *from);
3716void dev_uc_flush(struct net_device *dev);
3717void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003718
Alexander Duyck670e5b82014-05-28 18:44:46 -07003719/**
3720 * __dev_uc_sync - Synchonize device's unicast list
3721 * @dev: device to sync
3722 * @sync: function to call if address should be added
3723 * @unsync: function to call if address should be removed
3724 *
3725 * Add newly added addresses to the interface, and release
3726 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003727 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003728static inline int __dev_uc_sync(struct net_device *dev,
3729 int (*sync)(struct net_device *,
3730 const unsigned char *),
3731 int (*unsync)(struct net_device *,
3732 const unsigned char *))
3733{
3734 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3735}
3736
3737/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003738 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003739 * @dev: device to sync
3740 * @unsync: function to call if address should be removed
3741 *
3742 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003743 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003744static inline void __dev_uc_unsync(struct net_device *dev,
3745 int (*unsync)(struct net_device *,
3746 const unsigned char *))
3747{
3748 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3749}
3750
Jiri Pirko22bedad32010-04-01 21:22:57 +00003751/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003752int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3753int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3754int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3755int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3756int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3757int dev_mc_sync(struct net_device *to, struct net_device *from);
3758int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3759void dev_mc_unsync(struct net_device *to, struct net_device *from);
3760void dev_mc_flush(struct net_device *dev);
3761void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003762
Alexander Duyck670e5b82014-05-28 18:44:46 -07003763/**
3764 * __dev_mc_sync - Synchonize device's multicast list
3765 * @dev: device to sync
3766 * @sync: function to call if address should be added
3767 * @unsync: function to call if address should be removed
3768 *
3769 * Add newly added addresses to the interface, and release
3770 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003771 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003772static inline int __dev_mc_sync(struct net_device *dev,
3773 int (*sync)(struct net_device *,
3774 const unsigned char *),
3775 int (*unsync)(struct net_device *,
3776 const unsigned char *))
3777{
3778 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3779}
3780
3781/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003782 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003783 * @dev: device to sync
3784 * @unsync: function to call if address should be removed
3785 *
3786 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003787 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003788static inline void __dev_mc_unsync(struct net_device *dev,
3789 int (*unsync)(struct net_device *,
3790 const unsigned char *))
3791{
3792 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3793}
3794
Linus Torvalds1da177e2005-04-16 15:20:36 -07003795/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003796void dev_set_rx_mode(struct net_device *dev);
3797void __dev_set_rx_mode(struct net_device *dev);
3798int dev_set_promiscuity(struct net_device *dev, int inc);
3799int dev_set_allmulti(struct net_device *dev, int inc);
3800void netdev_state_change(struct net_device *dev);
3801void netdev_notify_peers(struct net_device *dev);
3802void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003803/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003804void dev_load(struct net *net, const char *name);
3805struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3806 struct rtnl_link_stats64 *storage);
3807void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3808 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003809
3810extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003811extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003812extern int weight_p;
Matthias Tafelmeier3d48b532016-12-29 21:37:21 +01003813extern int dev_weight_rx_bias;
3814extern int dev_weight_tx_bias;
3815extern int dev_rx_weight;
3816extern int dev_tx_weight;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003817
Joe Perchesf629d202013-09-26 14:48:15 -07003818bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003819struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3820 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003821struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3822 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003823
3824/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003825#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3826 for (iter = &(dev)->adj_list.upper, \
3827 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3828 updev; \
3829 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3830
David Ahern1a3f0602016-10-17 19:15:44 -07003831int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3832 int (*fn)(struct net_device *upper_dev,
3833 void *data),
3834 void *data);
3835
3836bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3837 struct net_device *upper_dev);
3838
Joe Perchesf629d202013-09-26 14:48:15 -07003839void *netdev_lower_get_next_private(struct net_device *dev,
3840 struct list_head **iter);
3841void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3842 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003843
3844#define netdev_for_each_lower_private(dev, priv, iter) \
3845 for (iter = (dev)->adj_list.lower.next, \
3846 priv = netdev_lower_get_next_private(dev, &(iter)); \
3847 priv; \
3848 priv = netdev_lower_get_next_private(dev, &(iter)))
3849
3850#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3851 for (iter = &(dev)->adj_list.lower, \
3852 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3853 priv; \
3854 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3855
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003856void *netdev_lower_get_next(struct net_device *dev,
3857 struct list_head **iter);
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003858
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003859#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003860 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003861 ldev = netdev_lower_get_next(dev, &(iter)); \
3862 ldev; \
3863 ldev = netdev_lower_get_next(dev, &(iter)))
3864
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003865struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3866 struct list_head **iter);
3867struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3868 struct list_head **iter);
3869
David Ahern1a3f0602016-10-17 19:15:44 -07003870int netdev_walk_all_lower_dev(struct net_device *dev,
3871 int (*fn)(struct net_device *lower_dev,
3872 void *data),
3873 void *data);
3874int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3875 int (*fn)(struct net_device *lower_dev,
3876 void *data),
3877 void *data);
3878
Joe Perchesf629d202013-09-26 14:48:15 -07003879void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003880void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003881struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3882struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3883int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3884int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003885 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003886 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003887void netdev_upper_dev_unlink(struct net_device *dev,
3888 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003889void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003890void *netdev_lower_dev_get_private(struct net_device *dev,
3891 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003892void netdev_lower_state_changed(struct net_device *lower_dev,
3893 void *lower_state_info);
Eric Dumazet960fb622014-11-16 06:23:05 -08003894
3895/* RSS keys are 40 or 52 bytes long */
3896#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003897extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003898void netdev_rss_key_fill(void *buffer, size_t len);
3899
Sabrina Dubroca952fcfd2016-08-12 16:10:33 +02003900int dev_get_nest_level(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003901int skb_checksum_help(struct sk_buff *skb);
3902struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3903 netdev_features_t features, bool tx_path);
3904struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3905 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003906
Moni Shoua61bd3852015-02-03 16:48:29 +02003907struct netdev_bonding_info {
3908 ifslave slave;
3909 ifbond master;
3910};
3911
3912struct netdev_notifier_bonding_info {
3913 struct netdev_notifier_info info; /* must be first */
3914 struct netdev_bonding_info bonding_info;
3915};
3916
3917void netdev_bonding_info_change(struct net_device *dev,
3918 struct netdev_bonding_info *bonding_info);
3919
Cong Wang12b00042013-02-05 16:36:38 +00003920static inline
3921struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3922{
3923 return __skb_gso_segment(skb, features, true);
3924}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003925__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003926
3927static inline bool can_checksum_protocol(netdev_features_t features,
3928 __be16 protocol)
3929{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003930 if (protocol == htons(ETH_P_FCOE))
3931 return !!(features & NETIF_F_FCOE_CRC);
3932
3933 /* Assume this is an IP checksum (not SCTP CRC) */
3934
3935 if (features & NETIF_F_HW_CSUM) {
3936 /* Can checksum everything */
3937 return true;
3938 }
3939
3940 switch (protocol) {
3941 case htons(ETH_P_IP):
3942 return !!(features & NETIF_F_IP_CSUM);
3943 case htons(ETH_P_IPV6):
3944 return !!(features & NETIF_F_IPV6_CSUM);
3945 default:
3946 return false;
3947 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003948}
Cong Wang12b00042013-02-05 16:36:38 +00003949
Herbert Xufb286bb2005-11-10 13:01:24 -08003950#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003951void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003952#else
3953static inline void netdev_rx_csum_fault(struct net_device *dev)
3954{
3955}
3956#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003957/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003958void net_enable_timestamp(void);
3959void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003960
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003961#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003962int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003963#else
3964#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003965#endif
3966
David S. Miller47982482014-08-22 16:21:53 -07003967static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003968 struct sk_buff *skb, struct net_device *dev,
3969 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003970{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003971 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003972 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003973}
3974
David S. Miller10b3ad82014-08-29 21:07:24 -07003975static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003976 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07003977{
3978 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07003979 int rc;
David S. Miller47982482014-08-22 16:21:53 -07003980
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003981 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07003982 if (rc == NETDEV_TX_OK)
3983 txq_trans_update(txq);
3984
3985 return rc;
David S. Miller47982482014-08-22 16:21:53 -07003986}
3987
Linus Torvalds42a2d922013-11-13 17:40:34 +09003988int netdev_class_create_file_ns(struct class_attribute *class_attr,
3989 const void *ns);
3990void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3991 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04003992
3993static inline int netdev_class_create_file(struct class_attribute *class_attr)
3994{
3995 return netdev_class_create_file_ns(class_attr, NULL);
3996}
3997
3998static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3999{
4000 netdev_class_remove_file_ns(class_attr, NULL);
4001}
Jay Vosburghb8a97872008-06-13 18:12:04 -07004002
Johannes Berg04600792010-08-05 17:45:15 +02004003extern struct kobj_ns_type_operations net_ns_type_operations;
4004
Joe Perchesf629d202013-09-26 14:48:15 -07004005const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07004006
Joe Perchesf629d202013-09-26 14:48:15 -07004007void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004008
Michal Kubečekda081432014-05-20 08:29:25 +02004009static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4010 netdev_features_t f2)
4011{
Tom Herbertc8cd0982015-12-14 11:19:44 -08004012 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4013 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08004014 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004015 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08004016 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004017 }
Michal Kubečekda081432014-05-20 08:29:25 +02004018
Tom Herbertc8cd0982015-12-14 11:19:44 -08004019 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02004020}
4021
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004022static inline netdev_features_t netdev_get_wanted_features(
4023 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00004024{
4025 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4026}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004027netdev_features_t netdev_increment_features(netdev_features_t all,
4028 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00004029
4030/* Allow TSO being used on stacked device :
4031 * Performing the GSO segmentation before last device
4032 * is a performance improvement.
4033 */
4034static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4035 netdev_features_t mask)
4036{
4037 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4038}
4039
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004040int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004041void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004042void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004043
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004044void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4045 struct net_device *dev);
4046
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004047netdev_features_t passthru_features_check(struct sk_buff *skb,
4048 struct net_device *dev,
4049 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004050netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004051
David S. Miller4d295152012-03-07 21:02:35 -05004052static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004053{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004054 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004055
4056 /* check flags correspondence */
4057 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4058 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4059 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4060 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004061 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004062 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4063 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004064 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4065 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004066 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4067 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004068 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4069 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 -04004070 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004071 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004072 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004073
Herbert Xubcd76112006-06-30 13:36:35 -07004074 return (features & feature) == feature;
4075}
4076
David S. Miller4d295152012-03-07 21:02:35 -05004077static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004078{
Herbert Xu278b2512009-06-03 21:20:51 -07004079 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004080 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004081}
4082
Johannes Berg8b86a612015-04-17 15:45:04 +02004083static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004084 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004085{
Jesse Grossfc741212011-01-09 06:23:32 +00004086 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004087 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4088 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004089}
4090
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004091static inline void netif_set_gso_max_size(struct net_device *dev,
4092 unsigned int size)
4093{
4094 dev->gso_max_size = size;
4095}
4096
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004097static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4098 int pulled_hlen, u16 mac_offset,
4099 int mac_len)
4100{
4101 skb->protocol = protocol;
4102 skb->encapsulation = 1;
4103 skb_push(skb, pulled_hlen);
4104 skb_reset_transport_header(skb);
4105 skb->mac_header = mac_offset;
4106 skb->network_header = skb->mac_header + mac_len;
4107 skb->mac_len = mac_len;
4108}
4109
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004110static inline bool netif_is_macsec(const struct net_device *dev)
4111{
4112 return dev->priv_flags & IFF_MACSEC;
4113}
4114
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004115static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004116{
4117 return dev->priv_flags & IFF_MACVLAN;
4118}
4119
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004120static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004121{
4122 return dev->priv_flags & IFF_MACVLAN_PORT;
4123}
4124
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004125static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004126{
4127 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4128}
4129
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004130static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004131{
4132 return dev->priv_flags & IFF_IPVLAN_MASTER;
4133}
4134
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004135static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004136{
4137 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4138}
4139
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004140static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004141{
4142 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4143}
4144
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004145static inline bool netif_supports_nofcs(struct net_device *dev)
4146{
4147 return dev->priv_flags & IFF_SUPP_NOFCS;
4148}
4149
David Ahern007979e2015-09-29 20:07:10 -07004150static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004151{
David Ahern007979e2015-09-29 20:07:10 -07004152 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004153}
4154
David Ahernfee6d4c2015-10-05 08:51:24 -07004155static inline bool netif_is_l3_slave(const struct net_device *dev)
4156{
4157 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4158}
4159
Jiri Pirko0894ae32015-08-27 09:31:19 +02004160static inline bool netif_is_bridge_master(const struct net_device *dev)
4161{
4162 return dev->priv_flags & IFF_EBRIDGE;
4163}
4164
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004165static inline bool netif_is_bridge_port(const struct net_device *dev)
4166{
4167 return dev->priv_flags & IFF_BRIDGE_PORT;
4168}
4169
Jiri Pirko35d4e172015-08-27 09:31:20 +02004170static inline bool netif_is_ovs_master(const struct net_device *dev)
4171{
4172 return dev->priv_flags & IFF_OPENVSWITCH;
4173}
4174
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004175static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004176{
4177 return dev->priv_flags & IFF_TEAM;
4178}
4179
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004180static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004181{
4182 return dev->priv_flags & IFF_TEAM_PORT;
4183}
4184
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004185static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004186{
4187 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4188}
4189
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004190static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004191{
4192 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4193}
4194
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004195static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4196{
4197 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4198}
4199
Eric Dumazet02875872014-10-05 18:38:35 -07004200/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4201static inline void netif_keep_dst(struct net_device *dev)
4202{
4203 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4204}
4205
Paolo Abeni18d3df32016-07-14 18:00:10 +02004206/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4207static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4208{
4209 /* TODO: reserve and use an additional IFF bit, if we get more users */
4210 return dev->priv_flags & IFF_MACSEC;
4211}
4212
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004213extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004214
Joe Perches571ba422010-02-09 11:49:47 +00004215/* Logging, debugging and troubleshooting/diagnostic helpers. */
4216
4217/* netdev_printk helpers, similar to dev_printk */
4218
4219static inline const char *netdev_name(const struct net_device *dev)
4220{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004221 if (!dev->name[0] || strchr(dev->name, '%'))
4222 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004223 return dev->name;
4224}
4225
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004226static inline const char *netdev_reg_state(const struct net_device *dev)
4227{
4228 switch (dev->reg_state) {
4229 case NETREG_UNINITIALIZED: return " (uninitialized)";
4230 case NETREG_REGISTERED: return "";
4231 case NETREG_UNREGISTERING: return " (unregistering)";
4232 case NETREG_UNREGISTERED: return " (unregistered)";
4233 case NETREG_RELEASED: return " (released)";
4234 case NETREG_DUMMY: return " (dummy)";
4235 }
4236
4237 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4238 return " (unknown)";
4239}
4240
Joe Perchesf629d202013-09-26 14:48:15 -07004241__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004242void netdev_printk(const char *level, const struct net_device *dev,
4243 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004244__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004245void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004246__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004247void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004248__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004249void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004250__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004251void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004252__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004253void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004254__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004255void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004256__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004257void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004258
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004259#define MODULE_ALIAS_NETDEV(device) \
4260 MODULE_ALIAS("netdev-" device)
4261
Jim Cromieb558c962011-12-19 17:11:18 -05004262#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004263#define netdev_dbg(__dev, format, args...) \
4264do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004265 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004266} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004267#elif defined(DEBUG)
4268#define netdev_dbg(__dev, format, args...) \
4269 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004270#else
4271#define netdev_dbg(__dev, format, args...) \
4272({ \
4273 if (0) \
4274 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004275})
4276#endif
4277
4278#if defined(VERBOSE_DEBUG)
4279#define netdev_vdbg netdev_dbg
4280#else
4281
4282#define netdev_vdbg(dev, format, args...) \
4283({ \
4284 if (0) \
4285 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4286 0; \
4287})
4288#endif
4289
4290/*
4291 * netdev_WARN() acts like dev_printk(), but with the key difference
4292 * of using a WARN/WARN_ON to get the message out, including the
4293 * file/line information and a backtrace.
4294 */
4295#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004296 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4297 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004298
Joe Perchesb3d95c52010-02-09 11:49:49 +00004299/* netif printk helpers, similar to netdev_printk */
4300
4301#define netif_printk(priv, type, level, dev, fmt, args...) \
4302do { \
4303 if (netif_msg_##type(priv)) \
4304 netdev_printk(level, (dev), fmt, ##args); \
4305} while (0)
4306
Joe Perchesf45f4322010-06-27 01:02:36 +00004307#define netif_level(level, priv, type, dev, fmt, args...) \
4308do { \
4309 if (netif_msg_##type(priv)) \
4310 netdev_##level(dev, fmt, ##args); \
4311} while (0)
4312
Joe Perchesb3d95c52010-02-09 11:49:49 +00004313#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004314 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004315#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004316 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004317#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004318 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004319#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004320 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004321#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004322 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004323#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004324 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004325#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004326 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004327
Joe Perches0053ea92012-05-30 07:43:34 +00004328#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004329#define netif_dbg(priv, type, netdev, format, args...) \
4330do { \
4331 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004332 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004333} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004334#elif defined(DEBUG)
4335#define netif_dbg(priv, type, dev, format, args...) \
4336 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004337#else
4338#define netif_dbg(priv, type, dev, format, args...) \
4339({ \
4340 if (0) \
4341 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4342 0; \
4343})
4344#endif
4345
Edward Creef617f272017-01-27 15:02:26 +00004346/* if @cond then downgrade to debug, else print at @level */
4347#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4348 do { \
4349 if (cond) \
4350 netif_dbg(priv, type, netdev, fmt, ##args); \
4351 else \
4352 netif_ ## level(priv, type, netdev, fmt, ##args); \
4353 } while (0)
4354
Joe Perchesb3d95c52010-02-09 11:49:49 +00004355#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004356#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004357#else
4358#define netif_vdbg(priv, type, dev, format, args...) \
4359({ \
4360 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004361 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004362 0; \
4363})
4364#endif
Joe Perches571ba422010-02-09 11:49:47 +00004365
Cong Wang900ff8c2013-02-18 19:20:33 +00004366/*
4367 * The list of packet types we will receive (as opposed to discard)
4368 * and the routines to invoke.
4369 *
4370 * Why 16. Because with 16 the only overlap we get on a hash of the
4371 * low nibble of the protocol value is RARP/SNAP/X.25.
4372 *
4373 * NOTE: That is no longer true with the addition of VLAN tags. Not
4374 * sure which should go first, but I bet it won't make much
4375 * difference if we are running VLANs. The good news is that
4376 * this protocol won't be in the list unless compiled in, so
4377 * the average user (w/out VLANs) will not be adversely affected.
4378 * --BLG
4379 *
4380 * 0800 IP
4381 * 8100 802.1Q VLAN
4382 * 0001 802.3
4383 * 0002 AX.25
4384 * 0004 802.2
4385 * 8035 RARP
4386 * 0005 SNAP
4387 * 0805 X.25
4388 * 0806 ARP
4389 * 8137 IPX
4390 * 0009 Localtalk
4391 * 86DD IPv6
4392 */
4393#define PTYPE_HASH_SIZE (16)
4394#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4395
Jiri Pirko385a1542009-05-27 15:48:07 -07004396#endif /* _LINUX_NETDEVICE_H */